Progressive Cyclic Nucleotide-induced Conformational Changes in the cGMP-dependent Protein Kinase Studied by Small Angle X-ray Scattering in Solution*

Small angle scattering data from bovine lung type Iα cGMP-dependent protein kinase (PKG) in the absence of cGMP show the protein to have a highly asymmetric structure with a radius of gyration (R g ) of 45 Å and a maximum linear dimension (d max) of 165 Å. The addition of cGMP induces a marked conformational change in PKG. TheR g and d max increase 25–30%, and the protein’s mass moves further away from the center of mass; this results in an even more asymmetric structure. Fourier transform infrared spectroscopy data suggest that the conformational change induced by cGMP binding is primarily due to a topographical movement of the structural domains of PKG rather than to secondary structural changes within one or more of the individual domains. Each monomer of the dimeric PKG contains one high and one low affinity cGMP-binding site. A prominent increase in the asymmetry of PKG occurs with binding to high affinity cGMP-binding sites alone, but the full domain movements require the binding to both sets of sites. These conformational changes occurring in PKG with the progressive binding of cGMP to both sets of cGMP-binding sites correlate with past data, which have indicated that cGMP binding to both sets of sites is required for the full activation of the enzyme. These results provide the first quantitative measurement of the overall PKG structure, as well as an assessment of the structural events that accompany the activation of a protein kinase upon binding a small molecular weight ligand.

Small angle scattering data from bovine lung type I␣ cGMP-dependent protein kinase (PKG) in the absence of cGMP show the protein to have a highly asymmetric structure with a radius of gyration (R g ) of 45 Å and a maximum linear dimension (d max ) of 165 Å. The addition of cGMP induces a marked conformational change in PKG. The R g and d max increase 25-30%, and the protein's mass moves further away from the center of mass; this results in an even more asymmetric structure. Fourier transform infrared spectroscopy data suggest that the conformational change induced by cGMP binding is primarily due to a topographical movement of the structural domains of PKG rather than to secondary structural changes within one or more of the individual domains. Each monomer of the dimeric PKG contains one high and one low affinity cGMP-binding site. A prominent increase in the asymmetry of PKG occurs with binding to high affinity cGMP-binding sites alone, but the full domain movements require the binding to both sets of sites. These conformational changes occurring in PKG with the progressive binding of cGMP to both sets of cGMP-binding sites correlate with past data, which have indicated that cGMP binding to both sets of sites is required for the full activation of the enzyme. These results provide the first quantitative measurement of the overall PKG structure, as well as an assessment of the structural events that accompany the activation of a protein kinase upon binding a small molecular weight ligand.
The cGMP-dependent protein kinase (PKG) 1 is a serine/threonine protein kinase that was first discovered in 1970 (1) and subsequently shown to have a widespread distribution in mam-malian tissues. Its cyclic nucleotide-dependent activation is one major mechanism whereby the broadly diverse physiological actions of cGMP are mediated (2)(3)(4)(5)(6)(7). PKG is a homodimer, each monomer of which contains the following in sequence: (i) the amino-terminal dimerization domain, (ii) the pseudosubstrate autoinhibitory/autophosphorylation domain, (iii) the two contiguous cGMP-binding domains, and (iv) the carboxyl-terminal catalytic domain (8). PKG shares significant homology with the cAMP-dependent protein kinase (PKA) in amino acid sequence, structure, and catalytic competence, with the major difference being that the regulatory (R) and catalytic (C) components of PKA are located on separate polypeptides encoded by separate genes (8,9). This difference impacts the mechanisms of cyclic nucleotide-dependent activation of PKA and PKG. PKA activation occurs through cAMP-dependent dissociation of the protein into separate regulatory and catalytic entities, although total dissociation may not be required (10,11). In contrast, PKG activation is constrained to involve only conformational rearrangements, since dissociation into separate regulatory and catalytic entities is not possible. In this regard, the structure of PKG is more similar to many members of the protein kinase family in which the regulatory and catalytic domains exist on a single polypeptide.
The structure of PKA catalytic subunit has been solved by x-ray diffraction (12), and the structure of the catalytic domain of PKG has been modeled upon what is now recognized as the conserved core of essentially all mammalian protein kinases (13). Like the other kinases whose structures have been solved (14 -18), the catalytic domain of PKG is predicted to be bilobal with a catalytic cleft located between the two lobes (5). Protein substrate binding to PKG shares many features in common with that of PKA (19), and significant homology is also evident between the autoinhibitory pseudosubstrate domains of the two kinases (20). Both kinases also contain shared recognition determinants within their catalytic center for substrate and pseudosubstrate binding (21). Release of the autoinhibition occurs upon cyclic nucleotide binding, i.e. the cyclic nucleotidedependent activation of protein kinase activity.
The crystal structure of the two in tandem cAMP-binding sites of PKA has been solved using the ⌬ 91 truncated RI subunit of PKA (22), and the two sites are structurally similar to that of the cAMP-binding pocket of the Escherichia coli catabolite gene activation protein, CAP (23). The two cGMP-binding sites of PKG are homologous in amino acid sequence to both the cAMP-binding sites of PKA and of CAP (8), and thus there is a very high likelihood that the overall structure of the two cGMP in tandem binding sites of PKG will closely mimic that determined for the regulatory subunit of PKA (24). One noted difference exists between the cyclic nucleotide-binding sites of PKA and PKG in that the two in tandem sites have different cyclic nucleotide binding characteristics. These two sites are termed A and B in both kinases with reference to their location along the polypeptide, with site A more proximal and site B more distal to the amino terminus. In PKA, site B is the higher affinity cAMP binding site and the site into which cyclic nucleotide would first equilibrate. Site A is of lower affinity and releases bound cAMP more readily (25)(26)(27). Because of the kinetic differences in the rates of cAMP dissociation from the two sites, A and B are often termed the "fast" and "slow" sites, respectively. In type I PKG, the kinetic characteristics of these two sites are reversed. Site A is the high affinity cGMP binding site, and site B is the low affinity site (28 -30). How these differences in the kinetic properties of the cyclic nucleotide binding sites may modulate the mechanism of activation of PKG and PKA remains to be elucidated.
Both PKA and PKG holoenzymes are dimers, with interactions of the amino-terminal domains of the regulatory region/ subunit important for dimerization. Removing this region by limited proteolysis produces monomeric units of catalytic and regulatory domains/subunits that still display cyclic nucleotidedependent protein kinase activity. The protein interactions that account for dimerization of the two kinases, however, appear to be different. In PKG-I␣ there is a prominent leucine zipper motif that is predicted to provide for dimerization, and NMR spectroscopy of a peptide composed of the first 39 aminoterminal residues shows that it dimerizes by such a mechanism (31). Leucine zipper motifs are also apparent in PKG-I␤ and PKG-II, (32,33). These motifs are not present in any of the PKA isoforms (I␣, I␤, II␣, II␤) (9,34).
In this study, we have examined the conformational changes that accompany the progressive cGMP binding to PKG by small angle x-ray scattering and Fourier transform infrared (FTIR) spectroscopy. We demonstrate that cGMP causes a major conformational shift in PKG and that the full shift requires binding at both the high and low affinity cGMP binding sites. This work follows upon the previous conclusions of Wolfe et al. (35), based upon anion exchange chromatographic profiles, and of Landgraf et al. (36), from CD spectroscopy, that cGMP prompts conformational changes in PKG. In a companion paper (37), the types of conformational changes that we have observed here were identified employing quite different experimental approaches and conditions. Both sets of data provide fully complementary conclusions.

MATERIALS AND METHODS
Protein Preparation and Assay-Type I␣ PKG was purified to apparent homogeneity from bovine lung using the methods previously described (38). Purity was analyzed using 10% SDS-polyacrylamide gel electrophoresis followed by Coomassie Brilliant Blue staining. The purified protein was stored in 10 mM potassium phosphate buffer, pH 6.8, 2 mM EDTA, 25 mM ␤-mercaptoethanol, and 150 mM NaCl at 4°C. X-ray scattering experiments were performed within 1 week of purification to minimize possible time-dependent aggregation effects as well as potential loss of enzyme catalytic activity. Protein kinase catalytic activity was assayed as described previously in the absence and presence of 10 M cGMP using the heptapeptide (RKRSRAE) as substrate (39). The PKG activity was stimulated Ͼ10-fold by the addition of cGMP. The cAMP and cGMP content of the purified PKG was determined using the PKA or PKG activation assays previously described (40). The cAMP content of purified PKG was Ͻ2%, and the cGMP content was Ͻ0.1% of the concentration of cyclic nucleotide binding sites. Protein concentrations were determined by the Bio-Rad Bradford protein assay as standardized by repeated amino acid analyses of purified PKG and the determined amino acid composition (8,39).
The PKG samples used for the scattering experiments were ϳ1.5 mg/ml, as noted for each experiment. Cyclic GMP was added to the protein at the indicated stoichiometric levels from a concentrated stock solution prepared in the same buffer as that of the protein. The cGMP concentration in the stock solution was determined by measuring opti-cal density (molar extinction coefficient (⑀ 252 ) ϭ 13,700). The PKG samples used for FTIR measurements were further concentrated to 7.8 mg/ml using a Centricon-30 concentrator, followed by dialysis against 10 mM KH 2 PO 4 in D 2 O, pH 6.8 (uncorrected meter reading), containing 25 mM ␤-mercaptoethanol, with no observed change of enzymatic activity. cGMP was exchanged into D 2 O by repeated lyophilization in the presence of excess D 2 O. FTIR samples were prepared by the addition of either 10 mM cGMP in D 2 O or D 2 O alone, to give a final concentration of 50 M PKG-dimer with or without 1 mM cGMP.
FTIR Spectroscopy-FTIR spectra were recorded at 4-cm Ϫ1 resolution using a Bomem Interferometer equipped with a Perkin-Elmer solution cell with CaF 2 windows separated by a 100-m path length. FTIR data were analyzed using Bomem Grams/386 software. Typically, 1024 scans were coadded and triangularly apodized to increase the signal:noise ratio. Data were analyzed using the techniques of spectral subtraction and second derivative analysis as described by Trewhella et al. (41). Background buffer spectra were obtained using the buffer samples collected from the flow-through from the Centricon concentration procedure and treated identically as the PKG samples.
Small Angle X-ray Scattering Data Acquisition-X-ray scattering experiments were carried out using the small angle station at Beam Line 4 -2 at the Stanford Synchrotron Radiation Laboratory (42) as well as on the instrument at Los Alamos described by Heidorn and Trewhella (43). Scattering measurements were made on one sample using both instruments, and the data were carefully compared to evaluate possible radiation-induced aggregation effects arising from the very bright synchrotron source at SSRL. In addition, the synchrotron data were collected in intervals of 2 min to monitor possible time-dependent aggregation effects or other changes. Protein aggregation would be evidenced by an increase in the forward or zero angle scattering (I 0 ), which is directly proportional to the molecular weight of the scattering particle times its concentration in units of mg/ml (44). PKG samples showed no change in scattering over the 20-min data collection times used for the synchrotron experiments. In addition, there were no differences between the scattering data collected at Los Alamos and SSRL, apart from the expected improved statistics obtained with the synchrotron intensities. Thus, all of the data presented here are from the SSRL measurements for which the sample-to-detector distance was 2.3 m and the wavelength () was 1.3 Å.
Scattering data were measured as a function of scattering angle 2, and reduced to I(Q) versus Q, where Q (equal to 4sin/) is the amplitude of the scattering vector or momentum transfer of the scattered x-rays whose wavelength is . All scattering data were normalized by the incident beam intensity recorded by an ion chamber monitor before the sample. The net scattering from the protein was calculated by subtracting a buffer scattering profile measured in the same sample cell immediately before or after the protein measurement. The sample cells were made of Teflon to give a ϳ1-mm path length with circular mica windows (diameter, ϳ5 mm) that provided an area for irradiation larger than the x-ray beam (ϳ3 ϫ 1 mm 2 ). The mica window material was ϳ10 m thick. For the experiments presented here, the temperature was maintained at 15°C. X-ray scattering data were collected on PKG samples in which the molar ratio of cGMP to PKG-dimer was varied from 0 to 20 (i.e. 0 -5-fold molar excess over total number of cGMP binding sites). In the first set of experiments, we observed no time-dependent aggregation of PKG in the absence of cGMP; however, the addition of cGMP resulted in some protein aggregation after 2-3 h. Subsequent experiments were therefore conducted using samples to which cGMP was added less than 10 min before they were placed in the x-ray beam for measurement, and under such conditions no aggregation was apparent. Scattering data were collected for 10 -20 min for each sample. To demonstrate that the effects observed on the scattering are specific to cGMP, measurements were also performed using 5Ј-GMP.
Small Angle Scattering Data Analysis-The radius of gyration (R g ) forward scattering (I 0 ), and vector distribution function (P(r)) were calculated from the scattering data using both Guinier (45) and P(r) (46,47) analyses. For a homogeneous solution of monodispersed particles, the Guinier approximation, which holds for sufficiently small Q values, is as follows.
A plot of log[I(Q)] versus Q 2 thus yields R g and I 0 from the slope and the intercept with the log[I] axis, respectively. If one or two dimensions of the scattering particle are much larger than the other(s), one can multiply by Q to remove the longer dimension(s) from the scattering. This scaling yields corresponding Guinier approximations for rodlike and disklike forms, respectively, from which the radius of gyrations of cross-section and thickness, R c and R t , can be calculated. P(r) is obtained from the inverse Fourier transformation of the scattering data (46,47), using Tikhonov's regularization method (48). I 0 and R g can also be calculated as the zeroth and second moments of P(r), respectively. P(r) is the distribution of all possible vector lengths between scattering centers (atoms) within the scattering particle, and it therefore goes to zero at the maximum linear dimension, d max . R g is the root mean square distance of all scattering centers from their center of mass, which is equivalent to the root mean square of all vector lengths between scattering centers.
Determination of cGMP Site Occupancy-Purified PKG (ϳ0.6 mg/ml, 14 M total cGMP-binding sites) in 10 mM potassium phosphate buffer, pH 6.8, 2 mM EDTA, and 0.15 M sodium chloride was incubated with the indicated concentrations of [ 3 H]cGMP (15-30 Ci/mmol) at 15°C for 8 min. Four 5-l aliquots were then removed for determination of total bound [ 3 H]cGMP, and then a 50-fold excess of unlabeled cGMP was added to the remainder of the assay mixture. Aliquots of 5 l were then removed at regular intervals over a period of 2 h to determine the rate of release of bound [ 3 H]cGMP. All 5-l aliquots were added into 2 ml of saturated ammonium sulfate at 4°C, and then the total sample was filtered through 0.45-m nitrocellulose papers, prewetted with 1 ml of ice-cold saturated ammonium sulfate, using a Millipore vacuum filtration system as described previously (28). The nitrocellulose papers were washed with 3 ϫ 1 ml of saturated ammonium sulfate, dried, and then placed into vials, thoroughly agitated with 1 ml of 2% SDS and counted with 10 ml of aqueous scintillant. Separate blanks were used that contained all of the same components of the binding mixture except PKG, before and after the addition of unlabeled cGMP. Release from the low affinity "fast" site occurred within a few seconds. The t1 ⁄2 for release from the high affinity "slow" site was ϳ38 min. Site occupancy was determined in comparison with PKG saturated with 100 M [ 3 H]cGMP (7-fold excess of total cGMP to total cGMP sites).

RESULTS
Structural Parameters of PKG in the Absence of cGMP from X-ray Scattering-X-ray scattering data were collected for PKG in the absence of cGMP and over a range of cGMP:PKG stoichiometries. Fig. 1 shows the scattering profiles for PKG with no cGMP (E) and with five incremental concentrations of cGMP, and Fig. 2 shows example Guinier plots for two of these data sets at 0 and 4.4 mol of cGMP/PKG-dimer. The P(r) functions calculated from the data in Fig. 1 are shown in Fig. 3, along with the P(r) function for PKG in the presence of saturating 5Ј-GMP (dashed line). A summation of the structural parameters derived from the scattering data is presented in Table I for an extended range of cGMP:PKG stoichiometries. As demonstrated by these data, the Guinier plots obtained for PKG in both the absence and presence of cGMP are linear to low Q values (Fig. 2), and there is good agreement between the R g values calculated from the Guinier and P(r) analyses (Table  I) and the normalized I 0 values in the absence of cGMP and at each of the cGMP: PKG stoichiometries are identical within error (Table I) except for the very highest concentration of cGMP evaluated, which shows only a small increase. These three observations provide good evidence that the PKG samples are aggregation-free and not adversely affected by the x-ray irradiation. In addition, comparison of I 0 values for the PKG samples and a standard calmodulin sample (49) gave the expected ratio based on the molecular weights of the two proteins, confirming that the PKG samples were aggregation-free.
In the absence of cGMP, the PKG scattering data indicate a relatively asymmetric structure. This is evident from a very simple analysis of the R g and d max values. A spherically symmetric particle with an R g of 45.5 Å would have a diameter of indicating a second axis of rotation, and the linear region in the R t plot is smaller and shifts toward higher Q values. These data indicate that PKG becomes less disk-shaped when saturated with cGMP. As cGMP is titrated to the sample, the R t linear Q region shifts gradually outwards. The Q region shown in the figure was that used to calculate the R t value.
117 Å, which is 30% smaller than the actual measured d max for PKG of 165 Å (Table I), and its volume would be 4 times larger than that calculated for PKG. PKG has a molecular mass of 152 kDa, which equates to a calculated "dry" volume of ϳ 1.9 ϫ 10 5 Å 3 , assuming a partial specific volume for a typical protein of ϳ0.72 g/cm 3 . Since the structure of the zero cGMP:PKG appears to be quite asymmetric, R c and R t values were calculated using the respective Guinier approximations (Equations 2 and 3). Both of these analyses gave good straight line fits over a wide Q range (ϳ0.05-0.1 Å Ϫ1 for R t and ϳ0.03-0.06 Å Ϫ1 for R c , Fig. 2), indicating that the approximations are reasonable and that in the absence of cGMP, PKG appears to have one very long dimension as well as one very short dimension. These conclusions are in agreement with past physiochemical determinations of the Stokes radius and frictional coefficient, which likewise indicated that PKG is notably asymmetric (50,51).
Based upon the conclusion of asymmetry, the structural parameters of PKG were modeled as a simple ellipsoid for an initial approximation. The largest semiaxis dimension, c, can be calculated from the measured R g and R c values using the relationship R g 2 Ϫ R c 2 ϭ c 2 /5. For PKG with no cGMP, this calculation gives a value of c ϭ 89 Å that corresponds to a d max of ϳ 178 Å, which is within 8% of the value determined from P(r) ( Table I). R t for a particle with one very thin dimension is related to the thickness of that dimension, T, by the following.
From this relationship, the thinnest semiaxis dimension, b, of PKG with no cGMP (Table I) is 19 Å. For an ellipsoid, the three semiaxes are related by the equation R g 2 ϭ (a 2 ϩ b 2 ϩ c 2 )/5, where a, b, and c are the three semiaxes. Thus, an ellipsoid with c ϭ 89 Å and b ϭ 19 Å will have its third semiaxis a ϭ 46 Å, and its volume will be 3.3 ϫ 10 5 Å 3 . The hydrated volume of protein is typically as much as ϳ30% larger (52,53) than the dry volume, giving an estimated hydrated volume of PKG of ϳ2.4 ϫ 10 5 Å 3 . Given that PKG is likely to have an irregular surface and would not form a perfect ellipsoid, the estimated hydrated volume is in reasonable agreement with that calculated for the ellipsoid. The ellipsoid approximation thus appears to provide a reasonable initial description of both the overall asymmetry of the protein and of its envelope and dimensions.   Table I). The cGMP was added in exact amounts to achieve the desired stoichiometries. As is further documented below, under these conditions the concentrations of cGMP and PKG are in such marked excess of the cGMP affinity constants for the cGMPbinding sites that Ͼ99.8% of the added cGMP would be PKGbound until very close to full saturation at 4 mol of cGMP/mol of PKG-dimer is achieved. The conformational change in PKG appears complete with the saturation of the four cGMP binding sites per PKG-dimer, and no further change is observed with additional cGMP. This is particularly evident in Fig. 4, which shows plots of the R g , d max , R c , R t , and I 0 values as a function of the molar ratio of cGMP to PKG-dimer. Both R g and I 0 are calculated from the P(r) analyses, while R c and R t are obtained from Equations 2 and 3. The R g value increases by 30% from zero to saturating cGMP. Accompanying the change in R g is an equally dramatic change in the distribution of vector lengths in P(r), which shifts toward higher r values with the largest changes seen in the distances greater than 80 Å (Fig. 3). The d max also increases from 165 to 205 Å ( Fig. 4 and Table I), reaching a maximum value at a 4:1 molar stoichiometry. As R g and d max increase with increasing cGMP:PKG stoichiometries, there are also changes in the R c and R t plots. As discussed above, in the absence of cGMP the linear Q regions in the R c and R t plots are both fairly wide. With increasing cGMP, a second linear region with a different slope emerges (at Q values Ͼ 0.05 Å Ϫ1 ) in the R c plot, indicating a second axis of rotation, while the first linear region (Q ϭ 0.03 Ϫ 0.06 Å Ϫ1 ) does not change throughout the titration (Table I). The second linear region yields decreasing R c values with increasing cGMP (Table I and Fig. 4). For the R t plot, the linear region gradually becomes somewhat smaller and shifts toward higher Q values, showing that PKG becomes increasingly less disk-shaped. Beyond 4 mol of cGMP/mol of PKG-dimer, no further changes are evident in R g or R t . The changes observed in PKG conformation are specific for the cyclic nucleotide, and no conformational change was evident either in the P(r) plot (Fig. 3, dashed line) or for any of the measured scattering parameters (Table I) at 40-and 110-fold molar excess of 5Ј-GMP (i.e. at 10-and 27.5fold molar excess over total number of cGMP binding sites; data for 110-fold 5Ј-GMP not shown). If the same ellipsoid simplification is carried out as for the zero cGMP structure, the three semiaxes of the ellipsoid for the PKG in its cGMP-bound state can be calculated as 51, 9, and 121 Å, respectively, compared with the 46, 19, and 89 Å in the cGMP-free state. Thus, in the absence of cyclic nucleotide, the PKG is quite asymmetric, but cGMP binding to PKG induces an even more asymmetric form.
Progression of cGMP Binding under the Conditions Used for X-ray Scattering-The cGMP binding characteristics of PKG have been well documented by both direct cGMP binding studies and measurements of cyclic nucleotide-dependent protein kinase activation (28,29). Because small angle scattering measures the time and ensemble average, it cannot tell how the cGMP binding sites of PKG contribute to the observed conformational change as cGMP is titrated into each type of cGMPbinding site. The PKG-dimer binds 4 mol of cGMP/mol of protein, with each polypeptide chain containing two binding sites, designated A and B. The K D values for both the A and B sites are submicromolar, but they differ in cGMP affinity by a factor of ϳ10; the specific K D values are dependent upon such conditions as temperature, the presence or absence of ATP, the degree of autophosphorylation, etc. (28,29,54,55). This difference in affinity between the two sites is primarily a consequence of a marked difference in their cGMP dissociation rates (k d ), with the association rates (k a ) being quite similar (28,29). The conditions used for the previous analyses of cGMP binding parameters were different from those used in these current studies of x-ray scattering in one key parameter. Most of the binding studies have been undertaken with a concentration of PKG in the 10 -50 nM range (28,29), whereas for the scattering experiments the PKG concentration was ϳ10 M (Figs. 1-4). Under the conditions used for the scattering experiments, the concentrations of cGMP and PKG greatly exceed the K D values of either site A or B. Thus, until the level of 4 mol/mol of PKG-dimer had been achieved, it would be predicted that greater than 99.8% of added cGMP would be bound to the PKG. In the data reported (Figs. 1-4; Table I), stoichiometric binding levels were attained by adding varying amounts of cGMP to achieve partial or complete saturation of the four sites.
Under the conditions of the scattering experiments, because of the difference in the K D values for the high and low affinity cGMP sites, it would be predicted that at a stoichiometry of less than 2 mol of cGMP/mol of PKG, 80 -90% of the added cGMP would selectively equilibrate into the high affinity site. Only as the high affinity sites became saturated would cyclic nucleotide occupy the lower affinity site. To verify this prediction, cGMP binding was determined under conditions that closely mimicked those of the x-ray scattering experiments. As is evidenced by the data of Table II, when the stoichiometry of cGMP to PKG-dimer is less than 2 mol/mol, at least 70 -75% of cGMP is bound to the high affinity "slow release" site. As the stoichiometries of cGMP to PKG-dimer increase further, the nucleotide equilibrates into both sites until a stoichiometry of 4 mol/mol is FIG. 4. Top to bottom, R g , d max , R c , R t , and I 0 versus molar eq of cGMP for PKG samples. d max is obtained from the maximal r value at which P(r) goes to zero. R g and I 0 are calculated as the second and zero moments of P(r). R c and R t are obtained from the data in Fig. 1 using Equations 2 and 3. The R c shown here is R c (a) in Table I. The errors are propagated from the statistical errors of the scattering data only. The estimate of uncertainty in d max is ϳ10 Å.
achieved. The t1 ⁄2 values of cGMP dissociation from these sites (site A, ϳ38 min; site B, Ͻ5 s) under these conditions agree well with those determined previously (28,29). Equilibrium binding into these sites occurred promptly, as anticipated from the association rates. An examination of the x-ray scattering data collected at 5 and 10 min during data acquisition indicated that, within statistical limitations, they were identical to the data acquired over the full 20-min duration. This indicates that throughout the duration of the x-ray scattering experiments, PKG was at equilibrium with the different cGMP levels used.
Overall, these data indicate that under these conditions, cGMP first occupies the high affinity cGMP-binding sites, and then the low affinity cGMP binding sites are subsequently occupied with increasing cGMP:PKG stoichiometries. Of note, there is also a change in the structural parameters of PKG coincident with the increasing levels of cGMP that is readily detected upon the addition of cGMP even at less than a mol:mol stoichiometry and proceeds up to the binding of 4 mol of cGMP. Thus, the occupancy of the high affinity cGMP-binding sites and then the subsequent occupancy of the low affinity cGMPbinding sites each appear to promote conformational changes in PKG.

Examination of PKG Conformational Changes by FTIR-
The PKG holoenzyme is an assembly of two polypeptides, each with one leucine zipper dimerization domain, one autoinhibitory/autophosphorylation domain, two cGMP binding domains, and one catalytic domain with a small and a large lobe. The conformational changes induced by cGMP, as evidenced by the x-ray scattering experiments, are probably a consequence of the topographical rearrangement of these domains, a major secondary and tertiary structural rearrangement of one or more of the individual domains, or some combination of both such changes. These alternatives were evaluated by FTIR spectroscopy. The region of the FTIR spectrum known as the amide IЈ region (1600 -1700 cm Ϫ1 ) contains the conformationally sensitive stretching frequencies of the peptide backbone carbonyl groups. In this region, as might be expected for a large complex protein, the FTIR spectrum of PKG displays as a broad envelope (Fig. 5A). Difference spectra for PKG with and without cGMP showed that the protein in the absence or presence of cGMP is very similar in secondary structure content (Fig. 5C). Resolution enhancement of an FTIR spectrum can be achieved by taking the second derivative of the protein spectrum, which has the effect of accentuating subtle variations in the line slope caused by contributions from multiple bands beneath the spectral envelope as illustrated for PKG alone (Fig. 5B). From the inverted second derivative spectra, percentages of secondary structural components can be estimated from the relative areas under the peaks (56) using the secondary structure assignments to frequency of Byler and Susi (57). Table III gives the results of such an analysis for PKG and also for PKG plus 20 molar equivalents of cGMP. The major scaffolding architecture of protein domains is provided by a combination of ␣-helix and ␤-sheet. The percentages of these two components of PKG were minimally changed by cGMP addition (Table III), suggesting that the secondary structure of the individual domains of PKG are minimally affected by the binding of cGMP. This absence of significant change in the percentages of secondary structures in individual domains is also indicated by the difference spectra (Fig. 5C). The primary differences detected by FTIR were in the percentage composition of unordered regions and ␤-turn/bends (Table III). Such changes are as might be anticipated if the large conformational changes seen by the scattering data are a consequence of the topographical reorientation of the domains, with consequential changes in the linking regions. The FTIR results for PKG without cGMP are consistent with the secondary structure content determined by circular dichroism (36). Landgraf and colleagues interpreted small changes in their CD a Each mol of PKG-dimer contains 4 mol of cGMP-binding sites, i.e., 2 mol of the high affinity site and 2 mol of the low affinity site.  spectra upon cGMP binding as indicating an increase in the proportion of ␤-pleated sheet at the expense of random coil. It is, however, difficult to obtain accurate estimates of ␤-structures from CD data, particularly in the presence of a strong ␣-helix component as there is in PKG. In contrast, the amide IЈ region of the FTIR is quite sensitive to both ␣and ␤-structures, and the FTIR data showed that the percentages of these two structural motifs in PKG were very similar in the absence and presence of cGMP. The cGMP-induced structural conformational changes observed in the x-ray scattering studies thus do not appear to be due to changes in the secondary structure of the individual domains but rather are a consequence of domain movements with respect to each other.

DISCUSSION
The x-ray scattering studies presented here provide for the first time detail on the overall structural parameters of PKG and a description of the structural changes that accompany activation of a protein kinase by a small ligand, in this case cGMP. As might have been anticipated from consideration of its complex domain structure, and in agreement with previous reports (50,51), the PKG molecule is markedly asymmetric, modeling as a first estimate of its overall envelope to an ellipsoid with semiaxes dimensions 89 ϫ 46 ϫ 19 Å. The putative leucine zipper dimerization domain in PKG-I␣ includes six heptad repeats of leucine or isoleucine extending from Leu-11 to Leu-46. Based upon the known structure of leucine zippers in other proteins and on the NMR structure determination of the structure of the synthetic peptide of this region (31), it is very probable that this domain is highly asymmetric, adopting a cylindrical shape of some 40 -50 Å in length. At the carboxylterminal end of this cylinder is separately tethered the continuation of each polypeptide chain, the first part of which contains the two pseudosubstrate/autoinhibitory domains. In the cGMP-free protein, the latter are bound into the kinase catalytic cleft. Gly-62 within this autoinhibitory domain is specifically located at the position that would be occupied by the acceptor serine of the kinase substrates (20). Residues 109 -335 of each monomer of PKG-I␣ form the in tandem cGMP binding sites (8), whose structure is probably highly similar to that solved for PKA (58). Preceding this structure, residues 80 -108 are packed onto the cyclic nucleotide binding domain, as is evident in the crystal structure solved for the ⌬ 91 truncated RI subunit of PKA (22), or in the holoenzyme this region might form part of the linkage to the autoinhibitory/pseudosubstrate region. Residues 363-699 of PKG constitute the primary protein kinase bilobal catalytic core shared by all protein kinases (8,13), with the catalytic cleft, into which Gly-62 of the pseudosubstrate is bound, located between the two lobes. PKG residues 340 -362 precede the catalytic core and share homology with the A ␣-helix of the PKA (59). With the restrictions that are placed upon the overall structure of PKG by the lengths and location of the linkages between the different domains, the cyclic nucleotide binding domain needs to be primarily localized along one face of the catalytic domain. Judging from the molecular dimensions that are apparent from the crystal structures of the homologous PKA regulatory and catalytic subunits (12,22), the combined structure of PKG regulatory and catalytic domains is also predicted to be quite asymmetric. Thus, cGMP-free PKG is likely composed of a combination of three quite asymmetric structures (i.e. two regulatory-catalytic interacting domains and one leucine zipper dimerization domain). Hence, it might be anticipated that the whole molecule is asymmetric, as the scattering data now directly demonstrate.
cGMP binding clearly provokes a major structural change in PKG, as is well evidenced by the x-ray scattering profiles, with a marked increase in the maximum linear dimension of the molecule and a shift of mass away from the center of the molecule. These changes are specific for cGMP. The 5Ј-GMP analog did not mimic the changes observed for cGMP ( Fig. 3 and Table I); the binding characteristics of cGMP under the conditions of the x-ray scattering experiments matched those detailed previously for cGMP (Table II); and the maximum effect required a stoichiometry of 4 mol of cGMP/mol of PKG (Fig. 4), which equals the established number of cGMP-binding sites on PKG. This stoichiometry of binding does not match the number of binding sites for ATP, the only other nucleotide with established binding sites on PKG.
Since cGMP activates PKG, it must result in some degree of movement of the autoinhibitory site out of the catalytic cleft. In doing so, it would decrease or eliminate the degree of interaction between the first ϳ80 amino acids of the amino terminus of the protein and the more carboxyl-terminal portions of the molecule containing the catalytic domain. This might allow the molecule to unfold to some degree, thereby accounting for the movement of mass away from the center of the molecule and an increase in its asymmetry. It is known that the bilobal catalytic subunit of PKA can exist in open and closed conformations, depending on its interaction with substrate or inhibitor, and presumably also with the autoinhibitory site of the regulatory subunit (60,61). The same kind of domain movement may also exist within the catalytic domain of PKG; however, such a movement is much too small to account for the conformational changes that we have observed here. It seems most likely that the change in PKG conformation is due to either a movement that causes some separation between the regulatory and catalytic domains or that there is movement of the two PKGmonomeric units away from each other. A combination of both movements is also possible.
To gain further insight into the nature of the PKG conformational changes, we calculated the radius of gyrations of cross-section and thickness for the catalytic subunit and ⌬ 91 truncated RI subunit of the homologous PKA, using model scattering data calculated from their crystal structures (12,22). The model data give an R c of ϳ12 Å and R t of ϳ6 Å for the truncated regulatory subunit and an R c of ϳ14 Å and R t of ϳ9 Å for the catalytic subunit. In parallel studies of PKA, we have measured scattering data from a regulatory subunit-catalytic subunit monomer (⌬ 90 RIIC) obtained by chymotryptic digestion of PKA, which removes the PKA R subunit dimerization domain, leaving a monomer containing one catalytic subunit and one truncated regulatory domain. These data give an R c of ϳ16 Å and R t of ϳ11 Å for this regulatory subunit-catalytic subunit monomer. 2 It is of note that the R c value for this regulatory subunit-catalytic subunit monomer is 2-4 Å larger than the values calculated for the individual subunits, indicating that the long axes of the subunits are aligned at an angle. Assuming a similar arrangement for the regulatory and catalytic domains in an equivalent PKG-monomer, the even larger R c value for the PKG-dimer indicates that the long axes of the two PKG-monomers are also at an angle. This arrangement would give rise to the two axes of rotation observed in the R c analysis of the PKG data. Since the second R c observed for PKG with increasing cGMP approaches the values calculated for the individual PKA catalytic and truncated regulatory subunits, cGMP binding appears to cause these subdomains to align such that the angle between their long axes approaches 180°. The second axis of rotation observed with increasing cGMP is thus attributed to the long axis of each PKG-monomer. These pro-posed movements are supported by the observed changes in R t . The R t value of PKG in the absence of cGMP is the same as that of the PKA monomer, but the R t value of PKG decreases with increasing cGMP, approaching the mean value of that calculated for the PKA catalytic and regulatory subunits. These results also suggest that there is a movement of the regulatory and catalytic domains in PKG upon cGMP binding such that they extend to align more closely in a single plane.
The conformational changes in PKG were apparent even at a 1:1 stoichiometry of cGMP to PKG-dimer, were well evident with the binding of 2 mol of cGMP/mol of PKG-dimer, and continued until all four sites were saturated (Figs. 3 and 4). We have demonstrated here that under the conditions of this study, cGMP binding at the lower cGMP stoichiometries occurs principally to the high affinity sites, whereas at higher stoichiometries cGMP also saturates the low affinity sites (Table  II). Thus, the energy of binding of cGMP to the high affinity binding site alone is sufficient to induce a marked conformational change in PKG, but the total conformational changes in PKG require the binding into both the high and low affinity sites. While some activation of PKG protein kinase activity is readily evident with the binding of cGMP into just the high affinity sites, the full activation only occurs upon the saturation of both sets of sites (28,35). Thus, both the initial conformational changes evoked by the binding to the high affinity sites and the additional conformational changes that occur as a result of binding to the low affinity sites appear to be involved in, and presumably required for, maximum enzyme activation. It is of note that within the limitations of the experimental methods a nearly linear change in the structural parameters was observed throughout the binding of all 4 mol of cGMP (Fig.  4). The structural changes induced by cGMP binding into the two sets of sites thus appear to be quite similar, although the two sets of sites have distinctive nucleotide binding characteristics.