Activation by Cyclic GMP Binding Causes an Apparent Conformational Change in cGMP-dependent Protein Kinase*

Cyclic nucleotide binding activates cyclic nucleotide-dependent protein kinases, but the molecular mechanism is unknown. In the present studies, cGMP binding to type Iα or type Iβ cGMP-dependent protein kinase (PKG) caused (i) a large electronegative charge shift of each enzyme on ion exchange chromatography, (ii) an increase in the Stokes radius (>3 Å) of each enzyme, and (iii) a decreased mobility of type Iβ PKG on native gel electrophoresis. These physical changes were not detected in the monomeric form of type Iβ PKG upon activation by cGMP. However, the results of partial proteolysis of type Iα PKG revealed some degree of cGMP-induced conformational change within the PKG-monomer, since cGMP binding protects the PKG-monomer against chymotryptic cleavage. The altered sensitivity to proteolysis occurs at Met-200, which is located between the B and C α-helices in the high affinity site (site A), and implies that the cGMP-induced structural perturbations in this region may participate in activation of dimeric PKG. The cGMP-induced conformational effects observed using the physical separation methods are likely to reflect altered interactions within the dimeric PKG that are caused by structural alterations within the subunits.

Cyclic nucleotide binding activates cyclic nucleotidedependent protein kinases, but the molecular mechanism is unknown. In the present studies, cGMP binding to type I␣ or type I␤ cGMP-dependent protein kinase (PKG) caused (i) a large electronegative charge shift of each enzyme on ion exchange chromatography, (ii) an increase in the Stokes radius (>3 Å) of each enzyme, and (iii) a decreased mobility of type I␤ PKG on native gel electrophoresis. These physical changes were not detected in the monomeric form of type I␤ PKG upon activation by cGMP. However, the results of partial proteolysis of type I␣ PKG revealed some degree of cGMPinduced conformational change within the PKGmonomer, since cGMP binding protects the PKGmonomer against chymotryptic cleavage. The altered sensitivity to proteolysis occurs at Met-200, which is located between the B and C ␣-helices in the high affinity site (site A), and implies that the cGMP-induced structural perturbations in this region may participate in activation of dimeric PKG. The cGMP-induced conformational effects observed using the physical separation methods are likely to reflect altered interactions within the dimeric PKG that are caused by structural alterations within the subunits.
Protein kinases catalyze the transfer of the ␥-phosphate of ATP to serine, threonine, or tyrosine residues on many cellular proteins and play key roles in regulating such diverse physiological processes as metabolism, gene expression, cell growth, cell differentiation, cardiac and skeletal muscle contraction, smooth muscle function, and synaptic transmission. Most serine/threonine-or tyrosine-specific protein kinases are activated by allosteric ligand binding (e.g. cAMP, cGMP, calcium, insulin, epidermal growth factor), but the molecular mechanisms involved in the activation of the protein kinases by ligands are largely unknown. It has been suggested that maintenance of a low basal activity results from intramolecular interactions between the autoinhibitory domains of the kinases and their catalytic sites (1)(2)(3)(4). The binding of activating allosteric ligands could induce conformational changes that relieve autoinhibition, thus activating the protein kinases, but there is little evidence to support such conformational changes. Since many regulatory and catalytic features of protein kinases have been conserved throughout evolution, it is possible that the mo-lecular events that lead to the relief of autoinhibition of protein kinases upon binding of an activating ligand may be similar.
Two homologous serine/threonine-specific protein kinases, cGMP-dependent protein kinase (PKG) 1 and cAMP-dependent protein kinase (PKA), are major intracellular mediators for cGMP and cAMP actions. However, the activation mechanism of these enzymes by cyclic nucleotide binding is poorly understood. A number of studies have suggested that binding of cyclic nucleotides to these kinases induces a conformational change that relieves autoinhibition, thereby activating the catalytic site of the respective kinase (1,2,(5)(6)(7)(8). For instance, the proteolytic sensitivities of the autoinhibitory domains in type I PKGs are increased markedly in the presence of cGMP (7), which suggests that a conformational change has occurred that increases the solvent exposure of this region. Studies of the type I␣ PKG using far-ultraviolet circular dichroism suggest that cGMP binding to the enzyme causes changes in the circular dichroism spectra that are consistent with the induction of ␤-sheet elements from random coil (8). The circular dichroism studies were performed at very high concentrations of PKG and cGMP, which greatly exceed their physiological levels (16).
The physical changes that accompany activation of a ligandregulated protein kinase using concentrations of enzyme that approach physiological levels have not been assessed. Such structural changes can be studied using type I PKGs, since they are available as highly purified soluble enzymes. In addition, the regulatory and catalytic domains of PKG reside on a single polypeptide chain, which is typical of other members of the protein kinase family. Consequently, studies of conformational changes in PKG upon activation by cGMP may relate to the structural events that accompany the activation mechanisms utilized by other protein kinases. Insights into the activation of PKG may be particularly pertinent to PKA activation, since the overall domain structures of PKA and PKG are quite similar, and cyclic nucleotide binding may cause similar molecular perturbations in each protein kinase.
In mammalian tissues, two major isozymes of PKG, termed type I and type II, have been identified, and these derive from separate genes. The type I isozyme has at least two isoforms, designated I␣ and I␤, and these isoforms are the products of alternative mRNA splicing (9 -14). Type I␣ PKG and type I␤ PKG are dimeric enzymes that contain two identical subunits of approximately 78 kDa each. Each of these subunits contains several domains of well defined functions. The regulatory component is located in the amino-terminal segment of PKG, which includes the dimerization domain, the autoinhibitory domain containing multiple autophosphorylation sites, and two homologous cGMP-binding sites that are arranged in tandem. The cGMP-binding sites in the primary sequence are followed by the catalytic component, which contains the MgATP and protein substrate binding sites (15,16). The amino acid sequences of the two type I isoforms differ only in their amino-terminal ϳ100 residues, in which they share only 36% identity (17). Although the cyclic nucleotide-binding sites in type I␣ and type I␤ PKG have identical amino acid sequences, the two isoforms show very different cGMP binding properties, suggesting that the amino termini affect the binding of cGMP to the enzymes (10, 18 -19).
To examine possible conformational changes that are associated with the process of activation of the PKG by cGMP, four approaches have been used: (i) ion exchange chromatography, which resolves proteins based on their surface charges; (ii) gel filtration chromatography, which resolves proteins based on their masses and shapes; (iii) native polyacrylamide gel electrophoresis, which provides resolution based on both the surface charges and shapes of proteins; and (iv) partial proteolysis, which reveals alterations in structure based on the susceptibility of PKG to selective digestion by proteases. Purified type I␣ and type I␤ PKG have been used for these studies, since features that are central to the regulatory and catalytic mechanisms are likely to be conserved in these homologous kinases. The present studies showed that cGMP binding to PKG produces distinct conformational changes in the enzyme as evidenced by increased surface electronegativity, a larger Stokes radius, decreased mobility on native gel electrophoresis, and altered susceptibility to proteases. These results agree well with those reported in the accompanying manuscript (40) and complement that study.

EXPERIMENTAL PROCEDURES
Purification of PKG and Protein Kinase Assay-Bovine lung type I␣ and bovine aorta type I␤ PKG were purified to homogeneity as described by Francis et al. (20). The kinase activity of PKG was determined by the phosphocellulose paper assay using heptapeptide substrate (RKRSRAE) as described previously (10).

Preparation and [ 3 H]cAMP
Binding Assay of Cyclic Nucleotide-free Regulatory Subunit of Type II␣ PKA-The dimeric regulatory subunit of bovine heart type II␣ PKA (RII␣) was purified according to the method of Corbin et al. (1). The binding activity of RII␣ was measured by a [ 3 H]cAMP binding assay as described previously (21). Urea denaturation of RII␣ to remove cyclic nucleotides was performed as described by Poteet-Smith et al. (22). Like native RII␣, this urea-treated cAMP-free RII␣ exhibited two active cAMP-binding sites, inhibited the catalytic subunit of PKA stoichiometrically, and had a dimeric structure as verified by determination of its sedimentation coefficient.
Measurement of cAMP and cGMP-The cyclic nucleotide contents of the purified proteins were determined using the modified version of the cyclic nucleotide assay described by Corbin et al. (23). Briefly, purified proteins were placed in a boiling water bath for 2 h and then centrifuged for 7 min in a Beckman Microfuge B. The resulting supernatant fractions were removed and chilled on ice, and the concentrations of cAMP and cGMP of these supernatants were measured by PKA and PKG activation assays. For RII␣, 0.5 M HCl was added to the protein for a final concentration of 13.8 mM before boiling, and the supernatant was neutralized by adding the same volume of 1 M K 2 HPO 4 as the volume of HCl added. The sample was then diluted ϳ10-fold with H 2 O to lower the ionic strength for the assay.
Preparation of cGMP-bound PKG-Purified PKG was incubated with unlabeled cGMP (ϳ500-fold excess of cGMP-binding sites) at 4°C overnight to assure full saturation of the cGMP-binding sites of the enzyme. For native gel electrophoresis experiments, purified PKG was incubated with unlabeled cGMP or cGMP analog, ␤-phenyl-1,N 2 -etheno-8bromo-cGMP (8-Br-PET-cGMP), as described above to obtain cyclic nucleotide-bound PKG.
Ion Exchange Chromatography-Because type I␣ and type I␤ PKG have quite different affinities for DEAE-Sephacel resin, different salt gradients were used to elute the enzymes (10). Purified PKG that had been preincubated in the absence or presence of cGMP was applied to a DEAE-Sephacel column (0.9 ϫ 10 cm) equilibrated in 10 mM potassium phosphate, pH 6.8, 1 mM EDTA, and 25 mM ␤-mercaptoethanol (KPEM) at 4°C. The column was washed with 30 ml of KPEM containing 0.12 M (for type I␣ PKG) or 0.17 M (for type I␤ PKG) NaCl. The enzyme was eluted with a linear gradient (200 ml) of NaCl (0.12-0.24 M for type I␣ PKG, 0.17-0.24 M for type I␤ PKG) in KPEM, and fractions (4 ml) were collected in silanized tubes containing 100 l of bovine serum albumin (10 mg/ml) to minimize protein sticking to the glass. Fractions were assayed for PKG activity as described above to determine the elution position of the enzyme. For experiments utilizing cGMP-bound PKG, the column was preequilibrated with KPEM containing 100 M cGMP, and the KPEM elution buffer also contained 100 M cGMP to assure that the cGMP-binding sites of the enzyme remained saturated with cGMP during chromatography. The cGMP concentration in the KPEM buffer was measured by absorbance at 252 nm.
Gel Filtration Chromatography-Purified PKG that had been preincubated in the absence or presence of cGMP was combined with crystalline catalase (4 mg) in a volume of 500 l and loaded onto a Sephacryl S-300 column (0.9 ϫ 168 cm) equilibrated with KPEM buffer containing 0.1 M NaCl at 4°C. The catalase served as an internal standard for all of the gel chromatographies. The column was eluted with the same buffer, and fractions (1.5 ml) were collected and assayed for PKG activity to determine the elution position of the enzyme. Catalase was located by the absorbance at 280 nm and/or 400 nm. In experiments in which PKG was presaturated with cGMP, the column was saturated with 100 M cGMP, and the column buffer also contained 100 M cGMP. Because cGMP in the column buffer interfered with the absorbance at 280 nm, absorbance at 400 nm was used to determine the peak position of catalase for those experiments involving cGMP-bound PKG. To compare the effect of different wavelength absorbances on the elution position of catalase, absorbances at 280 and 400 nm were done for several purified PKG experiments, and there was no difference in the location of the catalase peak.
Determination of Stokes Radius, Sedimentation Coefficient, and Molecular Weight-Apparent Stokes radii of purified PKGs in the presence and absence of cGMP were determined by the method described previously (10). Mixtures of protein standards with known Stokes radii (trypsin inhibitor, 24.5 Å; ovalbumin, 30.5 Å; apoferritin, 59.4 Å) were chromatographed on Sephacryl S-300 under identical conditions as described above, and the A 280 or A 400 was measured to generate the standard curve. A standard curve was constructed by plotting (Ϫlog K av ) 1 ⁄2 versus Stokes radius. The equation used to calculate the K av was as follows.
Elution volumeϪVoid volume Inclusion volumeϪVoid volume (Eq. 1) The Stokes radii of the PKGs were determined from the standard curve based on elution volume. The sedimentation coefficients of the enzymes were determined using linear sucrose gradients (5-20%) in KPEM containing 0.15 M NaCl in the presence and absence of 100 M cGMP and proteins of known sedimentation coefficients (catalase, 11.3 S; hemoglobin, 4.6 S) as internal standards as described previously (10). The apparent molecular weights, frictional ratios, and axial ratios were calculated according to the method of Siegel and Monty (24) together with the procedures of Cohn and Edsall (25). Native Gel Electrophoresis-Samples were electrophoresed on a 9.5% polyacrylamide gel and 4% stacking gel without sodium dodecyl sulfate at 4°C using constant current (ϳ10 mA) for 5 h. Proteins were detected by Coomassie Brilliant Blue staining.
Chymotrypsin Treatment of Type I␣ PKG-To produce partial proteolysis of purified type I␣ PKG, the enzyme (0.17 mg/ml) in 10 mM KPEM containing 0.15 M NaCl was treated with chymotrypsin at 30°C in the absence or presence of cGMP (10 M) or cAMP (110 M). Aliquots were removed at various times and subjected to 10% SDS-PAGE to determine patterns of digestion. Protein bands were visualized by staining with Coomassie Brilliant Blue. For the time courses of partial digestion, the ratio of PKG to chymotrypsin ranged from 350:1 to 50:1 (w/w). For production of the R 81 -PKG-I␣ and the M 201 -PKG-I␣ fragments for physical characterization and amino acid sequence analysis, 150 g of purified bovine type I␣ PKG (in the presence or absence of 110 M cAMP, respectively) was treated with 3 g of chymotrypsin for ϳ1 h at 30°C. Chymotrypsin treatment of PKG saturated with either cGMP or cAMP produced the the same pattern of PKG fragments. To purify the fragments for characterization and amino acid sequence, cAMP was used due to its much lower affinity for PKG, thus making it easier to remove from the resulting PKG fragment. The digest was cooled to 4°C, and an aliquot of [ 3 H]H 2 O was added prior to chromatography of the sample on a standardized Sephacryl S-200 column (0.9 ϫ 55 cm) in KPEM containing 0.05 M NaCl. Fractions (0.5 ml) were collected, and the resulting fragments were located using 10% SDS-PAGE followed by Coomassie Brilliant Blue staining, [ 3 H]cGMP binding, and protein kinase activity.
Amino Acid Sequencing-Amino acid sequences for the R 81 PKG-I␣ ( 81 RKFTKSERSKDL-) and the M 201 PKG-I␣ ( 201 MRTGLIKHTEY-) fragments were determined by the protein microsequencing facility in the Department of Biochemistry, University of Washington (Seattle, WA).
[ 3 H]Cyclic GMP Binding-Cyclic GMP-binding saturation and the rate of cGMP dissociation was determined as described previously (38) using aliquots of the PKG fragments from the Sephacryl S-200 chromatography described above. For studies of [ 3 H]cGMP binding saturation, the enzyme (ϳ18 g/ml) in KPEM containing 0.05 M NaCl was incubated in the presence of histone II-AS (1 mg/ml) and [ 3 H]cGMP (0.05-7 M) at 30°C for 30 min to promote full labeling of the cGMP binding sites. The mixture was then cooled to 4°C and incubated for another 20 min prior to removal of aliquots for the 0 t determination using Millipore filtration in the presence of saturated ammonium sulfate as described previously (38). For studies of cGMP dissociation, the PKG fragments (ϳ18 g/ml) in KPEM containing 0.05 M NaCl were incubated with [ 3 H]cGMP (final concentration, 10 M; specific activity, 10 mCi/mmol) and histone II-AS (final concentration, 1 mg/ml) as above. Aliquots were then removed for determination of O t binding. A 100-fold excess of unlabeled cGMP was then added, and aliquots were removed at various times and subjected to vacuum filtration using saturated ammonium sulfate.
Materials-Heptapeptide substrate was purchased from Peninsula Laboratories. Phosphocellulose paper was from Whatman. [␥-32 P]ATP was purchased from NEN Life Science Products. cGMP was obtained from Sigma, and [ 3 H]cGMP was from Amersham Corp. 8-Br-PET-cGMP was from Biolog Life Science Institute. DEAE-Sephacel and Sephacryl S-300 resins were purchased from Pharmacia Biotech Inc. Bovine serum albumin, bovine liver catalase, soybean trypsin inhibitor, hen egg ovalbumin, chymotrypsin, and bovine hemoglobin were from Sigma. Horse spleen apoferritin was obtained from Calbiochem. Endoproteinase Lys-C was purchased from Boehringer Mannheim.

RESULTS AND DISCUSSION
The physical effects of cGMP binding and activation of type I PKGs were examined using ion exchange chromatography, gel filtration, native gel electrophoresis, and partial proteolysis.
Cyclic Nucleotide-free and cGMP-bound PKGs-Purified PKGs commonly contain some cyclic nucleotides, since the enzymes exist as both cyclic nucleotide-free and cyclic nucleotidebound forms in tissues (6), and cAMP is added to elute these enzymes from a cAMP-affinity resin during purification (20). Although extensive efforts were made in the last step of the purifications to remove cyclic nucleotides from the enzymes, it was critical to verify the absence of cyclic nucleotides in PKGs for the present studies. The cAMP and cGMP contents of the purified enzymes were measured as described under "Experimental Procedures." The cyclic nucleotide occupancy of cGMPbinding sites of purified PKGs used for the following experiments was less than 3%. Therefore, these purified type I␣ and type I␤ PKGs were considered to be cyclic nucleotide-free forms. To obtain cGMP-bound PKGs, purified enzymes were incubated with a large excess of cGMP, and equal occupation of the two intrasubunit cGMP-binding sites was verified by [ 3 H]cGMP dissociation assays (data not shown).
DEAE-Sephacel Anion Exchange Chromatography of cGMPfree and cGMP-bound PKGs-When purified cyclic nucleotidefree type I␣ PKG was chromatographed on a DEAE-Sephacel column as described under "Experimental Procedures," a single peak of PKG activity was obtained (Fig. 1, top panel). When cGMP-bound type I␣ PKG was chromatographed in the presence of buffer containing 100 M cGMP, a single PKG activity peak was also obtained, but this peak eluted at higher ionic strength (Fig. 1, bottom panel). This same pattern was also found for the cyclic nucleotide-free and -bound type I␤ PKG isoform (Fig. 2). It should be noted that different NaCl gradients were used for these chromatographies, since type I␤ PKG binds more tightly to DEAE than does the type I␣ PKG. The increased affinity of the cGMP-bound forms of types I␣ and I␤ PKGs for the positively charged DEAE resin indicated that when cyclic nucleotide was bound, the net negative surface charge of these enzymes was increased. This increased electronegativity could be due to a conformational change of these PKGs, which exposes more negatively charged amino acids; alternatively, since the cyclic phosphate of cGMP is negatively charged, the overall negative charge of these enzymes could increase when cGMP binds. The crystal structure of a truncated form of regulatory subunit of the homologous PKA was recently reported (26). In this regulatory subunit-cAMP complex, the cyclic phosphate of bound cAMP is not near the surface of the protein. The predicted structure of the cGMPbinding sites of type I PKGs based on the crystal structure of the cAMP-binding domain of the E. coli catabolite gene activator protein (27) is similar to the cAMP-binding site of PKA (26,28,29). Therefore, it seems unlikely that the charge shift of the PKGs on the DEAE matrix is due to a direct interaction of the negative charge of the cGMP molecule with the resin. A conformational change produced by cGMP binding, which causes increased net surface electronegativity of the enzymes, is a more plausible explanation.
Sephacryl S-300 Gel Filtration Chromatography of cGMPfree and cGMP-bound PKGs-Cyclic nucleotide-free and cGMP-bound type I␣ and type I␤ PKGs were also subjected to gel filtration on a Sephacryl S-300 column in the presence of the internal standard catalase as described under "Experimental Procedures." It can be seen that the cGMP-bound form of type I␣ PKG eluted significantly earlier from the column than did the cGMP-free form (Fig. 3), and a similar pattern was obtained for the type I␤ isoform (Fig. 4). These shifts in the elution position of the PKGs could be easily discerned when the PKG kinase activity profile was compared with the elution position of the internal standard catalase. The resolution of proteins using gel filtration chromatography depends on several factors including the mass and overall shape of the molecule, which together confer the Stokes radius of the protein (24,30). A collection of proteins of known Stokes radii were chromatographed on the same column, and their elution positions were used to generate a standard curve; based on this standard curve, the Stokes radii of the respective cGMP-bound PKGs were calculated to be approximately 3.5 Å larger than that of the cyclic nucleotide-free form for either type I␣ or type I␤ PKG (Table I).
Molecular Parameters of the Type I PKGs-As can be seen in Table I, the Stokes radii of type I␣ and type I␤ PKGs in the absence of cGMP were calculated to be 47.3 and 47.8 Å, respectively; in the presence of 100 M cGMP, the values were 50.7 and 51.3 Å for type I␣ and type I␤ PKGs, respectively. The sedimentation coefficients for these enzymes were determined in the presence and absence of 100 M cGMP using sucrose density gradients, but there was no significant difference in these values. The apparent molecular weights were calculated from the Stokes radii and sedimentation coefficients, and in each instance the value for the cGMP-bound enzyme was ϳ6,000 Da greater than that of the cyclic nucleotide-free enzyme. Thus, since the subunit molecular weights of the two isoforms are 76,331 Da for type I␣ PKG and 77,803 Da for type I␤ PKG (17,31), it is unlikely that cGMP caused oligomerization of the enzymes. The added mass due to four cGMP molecules bound per dimer would be only 1,468 Da, which would be insufficient to account for the shifts in elution positions on the Sephacryl S-300 column caused by cGMP binding. The calcu-lated axial ratios (Table I) indicated that the cGMP-bound enzymes are more elongated compared with the cGMP-free enzymes, and it is likely that this elongation of the enzyme causes an increase in the Stokes radius. Combining these results with those using anion exchange chromatography, activation of either isoform of PKG by cGMP binding causes an apparent conformational change with elongation of the enzyme and increased net negative surface charge.
Native Gel Electrophoresis-A third technique, native gel electrophoresis, was used to further examine differences in the physical parameters of the cyclic nucleotide-free and cyclic nucleotide-bound type I␤ PKG. Nondenatured bovine serum albumin was used as a standard marker, since it exhibited multiple bands that are believed to represent molecular weight isomers (32,33). In preliminary studies, cGMP was used for the experiment of Fig. 5A, and it did not cause a mobility shift (data not shown). The absence of an effect could have been due to rapid dissociation of cGMP from the enzyme during electrophoresis. Thus, a cGMP analog, 8-Br-PET-cGMP, was chosen for these experiments since this analog binds to PKG with significantly higher affinity than does cGMP (18), and loss of cyclic nucleotide from the PKG during electrophoresis should be minimized. Type I␤ PKG (15 g) was preincubated with a large excess of 8-Br-PET-cGMP (300 M) prior to electrophoresis on native gel. The relative mobility of the cGMP analogbound PKG was less than that of the cGMP-free PKG (Fig. 5A). When these two forms of the enzyme were electrophoresed on FIG. 2. Effect of cGMP binding on DEAE-Sephacel chromatography of type I␤ PKG. Either cyclic nucleotide-free enzyme (15 g, top) or cGMP-bound enzyme (8 g, bottom) in a volume of ϳ50 l was chromatographed and analyzed as described in Fig. 1. The results are representative of four separate experiments.

FIG. 3. Effect of cGMP binding on Sephacryl S-300 gel filtration chromatography of type I␣ PKG.
A mixture of cyclic nucleotide-free enzyme (20 g, top) and catalase (4 mg), or a mixture of cGMPbound enzyme (15 g, bottom) and catalase (4 mg) in a volume of 500 l was applied to a Sephacryl S-300 gel filtration column as described under "Experimental Procedures." PKG activity and absorbance at 280 or 400 nm of each fraction (1.5 ml) were measured to determine the elution volume of the enzyme and internal standard catalase, respectively. The data shown are representative of five experiments. the same gel system with both the gel matrix and the running buffer containing 100 M cGMP, they exhibited the same mobility (Fig. 5B), and this mobility was the same as that of the cGMP analog-bound PKG in Fig. 5A. The results suggested that the mobility shift of the type I␤ PKG that was induced in the presence of cyclic nucleotide may be due to a conformational change that diminishes migration of the PKG through the gel matrix toward the anode. The decreased mobility in the gel occurs despite the increased surface electronegativity of the cGMP-saturated PKG as measured on DEAE-Sephacel chromatography, which would be predicted to increase the mobility of the enzyme toward the anode. Therefore, the magnitude of the cGMP-induced upward shift in the mobility of the PKG on the gel may be compromised by the influence of the increased electronegativity of the enzyme. The results from these studies using native gel electrophoresis support the interpretation of the previous results from gel filtration analysis.
Effect of cGMP on a Monomeric PKG-A dimeric structure for all of the mammalian isoforms of PKG has been conserved, and a monomeric form of the enzyme from Paramecium has been reported (34). Although the type I␣ and type I␤ PKGs appear to be dimerized through a conserved leucine zipper motif near their amino termini (35), the role of dimerization in enzyme function is unclear. Several monomeric species of type I␤ PKG have been studied (7,36) and have been shown to contain the functional elements that provide for regulation and catalysis; these include the autoinhibitory domain, two cyclic nucleotide-binding domains, and the catalytic domain. Studies of the requirement of quaternary structure for the apparent conformational change in PKG were performed using a monomeric form of type I␤ PKG. Proteolytic cleavage of type I␤ PKG by endoproteinase Lys-C was performed as described by Francis et al. (7), and the enzyme fragment was purified by Sephacryl S-200 chromatography. The protease cleaves the dimeric type I␤ PKG (ϳ78-kDa subunit) carboxyl-terminal to Lys-74, thus converting the enzyme into a monomer (ϳ67-70 kDa), and the purity of the PKG fragment was confirmed using SDS-PAGE. Like the native type I␤ PKG, the monomer is a protein with an asymmetrical shape (frictional ratio, ϳ1.35). This monomeric PKG retains two active cyclic nucleotide-binding sites and has a cGMP dependence similar to that of the native type I␤ holoenzyme (the ratio of kinase activity measured in the absence of cGMP to that measured in the presence of cGMP was Ͻ0.1). The k cat is also the same as that of PKG holoenzyme.
Gel filtration chromatography was performed using both Sephacryl S-200 and Sephacryl S-100 in an effort to optimize resolution, and ovalbumin (Stokes radius, ϳ30.5 Å) was used as an internal standard. The Stokes radius of the PKG-monomer was determined to be ϳ36 Å based on the standard curve described above. However, there was no detectable shift in the elution position of this monomer produced by cGMP binding (data not shown). Assuming that this chromatographic analysis would detect a shift for the monomer if a conformational FIG. 4. Effect of cGMP binding on Sephacryl S-300 gel filtration chromatography of type I␤ PKG. A mixture of cyclic nucleotide-free enzyme (15 g, top) and catalase (4 mg) or a mixture of cGMP-bound enzyme (8 g, bottom) and catalase (4 mg) was chromatographed and analyzed as described in Fig. 3. Each panel is representative of five experiments.  5. Effect of cyclic nucleotide on mobility of type I␤ PKG on native gel electrophoresis. A, cyclic nucleotide-free (15 g) and 8-Br-PET-cGMP-bound enzymes (15 g) were subjected to electrophoresis on a 9.5% native polyacrylamide gel in the absence of SDS and stained with Coomassie Brilliant Blue as described under "Experimental Procedures." B, the same enzymes were analyzed on 9.5% native polyacrylamide gel containing 100 M cGMP. The results shown are representative of five separate experiments. change occurred, the results indicate that the holoenzyme dimeric structure is an important component of the apparent conformational change measured in these studies. The monomeric type I␤ PKG fragment, produced by digestion with endoproteinase Lys-C, was also used for native gel experiments similar to those shown in Fig. 5 to determine the importance of dimerization for the apparent shape change. Monomeric PKG that was cGMP-free or saturated with 8-Br-PET-cGMP showed the same mobility despite using different gel running times to achieve varied degrees of migration through the gel (data not shown). The combined results from gel filtration analysis and native gel electrophoresis using both the dimeric and monomeric PKG suggested that the integrity of the PKG-dimer contributes significantly to the conformational change that is detected in these studies. It is possible that activation of PKG by cGMP causes a displacement of the relative position of the two monomers within the dimer.
Despite the absence of detectable changes noted above, another approach was used to examine whether or not the PKGmonomer undergoes a conformational change upon cGMP binding. Partial chymotryptic digestion of purified type I␣ PKG (Fig. 6) converted the 78-kDa PKG-monomer to fragments of ϳ65-70 kDa and ϳ50 kDa based on their mobilities on SDS-PAGE. The rate of production of the 65-70-kDa fragment was enhanced in the presence of cyclic nucleotide (data not shown). Using sequential Edman degradation through 12 residues of this fragment, we determined that chymotrypsin cleaved type I␣ PKG in the autoinhibitory domain carboxyl-terminal to Phe-80 to produce the amino-terminally truncated R 81 -PKG-I␣. This result was consistent with the results of a previous study from this laboratory showing the autoinhibitory domain of type I PKGs to have increased sensitivity to proteases when cyclic nucleotide is bound (7). The R 81 -PKG-I␣ had a k cat that was comparable with that of intact PKG; it bound cGMP stoichiometrically (2 mol/mol protein); and it exhibited a "fast" cGMPbinding site (low affinity) and a "slow" cGMP-binding site (high affinity) (data not shown).
In the absence of cyclic nucleotide, the R 81 -PKG-I␣ cyclic nucleotide-independent PKG was further hydrolyzed to a 50-kDa fragment (Fig. 6), but cyclic nucleotide binding to the PKG profoundly retarded the rate of this cleavage. To determine the location of the site that was protected by cyclic nucleotide binding, the 50-kDa PKG fragment was further characterized. Purified type I␣ PKG was digested in the absence of cyclic nucleotide with chymotrypsin as described under "Experimental Procedures," and the resulting fragment was purified to homogeneity using Sephacryl S-200 chromatography. Aliquots from the column fractions were examined using 10% SDS-PAGE followed by Coomassie Brilliant Blue staining, and the pooled fractions containing the pure 50-kDa fragment were further characterized using [ 3 H]cGMP binding and protein kinase activity (data not shown), as well as by sequential Edman degradation through 11 cycles. The amino-terminal amino acid sequence of the 50-kDa PKG began at Met-201, consistent with a chymotrypsin cleavage at Met-200 in the extreme carboxyl end of the high affinity cGMP-binding site (residues 101-219) (31). The M 201 -PKG-I␣ bound cGMP (1.1 mol of cGMP/mol of protein) with half-maximal saturation at 130 nM cGMP at 4°C. Cyclic GMP dissociated rapidly from M 201 -PKG (Ͻ1 min at 4°C), a rate that was consistent with the presence of only the low affinity cGMP-binding site (residues 220 -340) (31). The protein kinase activity was cyclic nucleotide-independent with a k cat comparable with that of intact PKG.
These results using partial proteolysis are consistent with a cGMP-induced conformational change in at least two regions of the PKG-monomer, i.e. in the autoinhibitory domain and in the cGMP-binding domain. Cleavage within the autoinhibitory domain is stimulated by cGMP binding, whereas cleavage within the cGMP-binding domain is inhibited by cGMP binding. It seems possible that cGMP binding to the high affinity cyclic nucleotide-binding site could directly obscure the highly vulnerable chymotryptic site at Met-200. However, in the crystal structure of RI␣ (26), the site that is homologous to Met-200 in PKG is located at the junction between the B and C ␣-helices in site A. Based on the crystal structures of CAP and RI␣, this junction is quite distant from the binding pocket occupied by the cyclic nucleotide (26,39). Thus, it seems likely that cGMP binding to the binding domain of the monomeric type I␣ PKG induces a conformational change in the vicinity of the junction of the B and C ␣-helices in site A. This conformational change in the PKG-monomer undoubtedly contributes to the cGMPinduced changes observed in cGMP activation of the dimeric enzyme, but it is not detectable using the physical methods. The insights derived from the current studies will be important in interpreting the structural events that accompany cyclic nucleotide binding to and activation of the kinase.
Effect of cAMP on the Regulatory Subunit of PKA-The cAMP-free RII␣ of PKA was used as a control for three of the approaches described above. Like the intact PKG, the RII␣ is a dimer that binds four cyclic nucleotide molecules, and it has a similar Stokes radius (ϳ53.4 Å). The cAMP-free RII␣ (less than 5% cAMP occupancy) was prepared as described under "Experimental Procedures," preincubated in the absence and presence of a large excess of cAMP (ϳ500-fold excess of cAMP-binding sites) at 4°C overnight, and then subjected to DEAE ion exchange chromatography, gel filtration chromatography, and native gel electrophoresis. The RII␣ eluted from the DEAE column at ϳ0.35 M NaCl, but cAMP binding to the RII␣ did not produce a detectable electronegative charge shift on this column (data not shown). The cAMP-binding sites of the RII␣ remained saturated with cAMP throughout the DEAE chromatography, since the column buffer contained a high concentration of cAMP. The results indicated that the introduction of four additional negative charges of the cAMP molecules is not FIG. 6. Effect of cyclic nucleotide on the pattern of chymotrypsin cleavage of type I␣ PKG. Purified type I␣ PKG was partially digested with chymotrypsin as described under "Experimental Procedures" and then subjected to 10% SDS-PAGE followed by Coomassie Brilliant Blue staining. Chymotrypsin treatment was performed in the absence and presence of 10 M cGMP using a PKG:chymotrypsin ratio of 50:1 (w/w). sufficient to change the surface charge of RII␣, and thus it is not likely that the negative charges of bound cGMP can account for the increased surface electronegativity observed with the PKGs. There was also no cAMP-induced shift in the elution position of the RII␣ on the gel filtration column, and both the cAMP-free and -bound RII␣ had the same mobility on native gel electrophoresis (data not shown). Thus, the increased mass conferred by cyclic nucleotide molecules bound to RII␣ was not sufficient to shift the position of RII␣. Since this dimeric RII␣ is smaller than PKG (90 versus 156 kDa), the percentage increase in charge or mass due to four cyclic nucleotide molecules would be larger than that for PKG. Thus, if these effects in PKG were attributable solely to the added charge or mass of the bound cGMP, an even larger shift in RII␣ might have been expected for each analytical procedure used.
The absence of a discernable effect of cyclic nucleotide with the RII␣ control suggested that the shift in the migration of the PKG caused by cGMP using gel filtration, native gel electrophoresis, or DEAE chromatography was due to conformational change rather than to an increase in charge or mass conferred directly by the bound cGMP molecules. If the homologous PKA and PKG holoenzymes undergo similar molecular perturbations during cyclic nucleotide activation, the results could suggest that both the regulatory and catalytic domains of PKG are needed to detect the conformational change that occurs with the PKG holoenzyme. However, caution should be exercised when extrapolating using the concept of homology between PKA and PKG. The monomers in the RII␣ dimer are thought to be aligned in an antiparallel configuration (37), while the monomers in the PKG-dimer are suggested to be in a parallel arrangement (35). It is therefore possible that similar initial conformational perturbations occur in each protein when cyclic nucleotide binds but that these perturbations do not cause the same quaternary displacement in RII␣ as in PKG. Thus, such a displacement in PKG might be readily detectable as a chromatographic shift by the procedures used here, but this may not be the case for RII␣.
Conclusion-In the present studies, both type I␣ and type I␤ PKGs were used to assess the conformational changes produced by binding of cGMP, the activating allosteric ligand. The combined results from ion exchange chromatography, gel filtration chromatography, and native gel electrophoresis indicate that cGMP binding to PKG causes a conformational change that results in increased net negative surface charge and an apparent elongation of the enzyme. The results also suggest that the conformational change that causes the shifts requires the dimeric form of PKG, since the monomeric form did not exhibit similar shifts. In the present study, an apparent cGMP-induced conformational change in monomeric PKG could be detected only by the approach of partial proteolysis. Perhaps this more subtle conformational change within the individual subunits of the dimer perturbs the dimeric structure sufficiently to be readily detectable by the physical methods used. The results also imply that cGMP binding to type I␣ PKG induces a conformational change in the cGMP-binding domain that protects this region against proteolytic cleavage. This latter event involves altered susceptibility of the type I␣ PKGmonomer to chymotrypsin cleavage at Met-200 between the B and C ␣-helices of the high affinity cGMP-binding site (site A).
The results for PKG elongation in the current study agree well with the results of Zhao et al. (40) using small angle x-ray scattering analysis of the type I␣ PKG. In that study, cGMP binding causes a significant elongation in the overall structure of type I␣ PKG (27% increase in maximum linear dimension), which is similar to the elongation (23% increase in axial ratio) determined in the present study.
To our knowledge, this is the first evidence, using gel filtration or native gel electrophoresis, for a conformational change of a protein kinase produced by binding of an activating allosteric ligand, and these techniques could prove to be useful in examining ligand effects in other proteins. This conformational change in PKG may either cause or be associated with activation of the protein kinase catalytic activity, but the details of this process have yet to be defined. The PKG, like other protein kinases, is thought to be activated by an allosteric ligandinduced conformational change that disrupts the interaction of the autoinhibitory domain with the catalytic site. It is expected that the information provided in the present studies may give insight into processes involved in the activation of other serine/ threonine-or tyrosine-specific protein kinases, almost all of which are homologous to PKG.