Fast and slow cyclic nucleotide-dissociation sites in cAMP-dependent protein kinase are transposed in type Ibeta cGMP-dependent protein kinase.

Both cyclic GMP-dependent protein kinase (cGK) and cyclic AMP-dependent protein kinase (cAK) contain two distinct cyclic nucleotide-binding sites referred to as fast and slow sites based on cyclic nucleotide dissociation behavior. In cAK, the fast site lies amino-terminal to the slow site, and sequence homologies between cAK and cGK have suggested similar positioning for the sites in cGK. Recombinant human type Iβ cGK (wild type (WT) cGK) was overexpressed, and the properties of purified WT cGK and native type Iβ cGK were similar. cGK was mutated singly at Thr-193 (T193A, T193V, and T193S) and Thr-317 (T317A, T317V, and T317S), which have been predicted to provide cGMP specificity in the cGMP-binding sites of cGK; a double mutant (T193A/T317A) was produced also. Compared with WT cGK, half-maximal activation (Ka) of mutant cGKs by cGMP was increased 2- (T317A), 27- (T193A), or 63-fold (T193A/T317A), but the Ka for cAMP of these mutants was essentially unchanged. The T193A and T193V mutants had a large increase in the rate of the slow component of [3H]cGMP dissociation, but in the T317A and T317V mutants, there was no change in the slow component. The T193S and T317S mutants had only minor effects on [3H]cGMP dissociation, thus establishing the importance of the hydroxyl group of Thr-193 and −317 for cGMP binding to cGK. Thus, in type Iβ cGK, the slow cGMP-binding site is identified as the amino-terminal site in contrast to the order assigned to the fast and slow cAMP-binding sites of cAK.

Both cyclic GMP-dependent protein kinase (cGK) and cyclic AMP-dependent protein kinase (cAK) contain two distinct cyclic nucleotide-binding sites referred to as fast and slow sites based on cyclic nucleotide dissociation behavior. In cAK, the fast site lies amino-terminal to the slow site, and sequence homologies between cAK and cGK have suggested similar positioning for the sites in cGK. Recombinant human type I␤ cGK (wild type (WT) cGK) was overexpressed, and the properties of purified WT cGK and native type I␤ cGK were similar. cGK was mutated singly at Thr-193 (T193A, T193V, and T193S) and Thr-317 (T317A, T317V, and T317S), which have been predicted to provide cGMP specificity in the cGMP-binding sites of cGK; a double mutant (T193A/ T317A) was produced also. Compared with WT cGK, half-maximal activation (K a ) of mutant cGKs by cGMP was increased 2-(T317A), 27-(T193A), or 63-fold (T193A/ T317A), but the K a for cAMP of these mutants was essentially unchanged. The T193A and T193V mutants had a large increase in the rate of the slow component of [ 3 H]cGMP dissociation, but in the T317A and T317V mutants, there was no change in the slow component. The T193S and T317S mutants had only minor effects on [ 3 H]cGMP dissociation, thus establishing the importance of the hydroxyl group of Thr-193 and -317 for cGMP binding to cGK. Thus, in type I␤ cGK, the slow cGMP-binding site is identified as the amino-terminal site in contrast to the order assigned to the fast and slow cAMP-binding sites of cAK.
The cGMP-dependent protein kinase (cGK) 1 and cAMP-dependent protein kinase (cAK) are homologous enzymes that are preferentially activated by cGMP and cAMP (1), respectively, although cross-activation of either cGK by cAMP (2) or cAK by cGMP (3) has been demonstrated under physiological and pathological conditions. The two cyclic nucleotide-binding sites in cAK preferentially bind cAMP over cGMP with a Ͼ50-fold selectivity, have different cAMP analog specificities, and have different dissociation rates for cAMP (4 -7). Because of the latter feature, these cAMP-binding sites generally have been referred to as the fast and slow cAMP-dissociation sites (see Footnote 2 of Ref. 8). Biochemical studies of type II cAK first suggested that the more amino-terminal cyclic nucleotide-binding site is the fast site whereas the adjacent site is the slow site (9), and similar approaches were used to demonstrate the same structural order for the binding sites of type I cAK (10). Molecular modeling (11) and results of site-directed mutagenesis of the binding sites (12,13) supported these findings.
Like cAK, each subunit of cGK ( Fig. 1, top) contains two cyclic nucleotide-binding sites that are homologous in sequence but have distinct kinetic characteristics. The cyclic nucleotidebinding sites of cGK preferentially bind cGMP over cAMP with a Ͼ100-fold selectivity, have different cGMP analog specificities, and have extremely different dissociation rates for cGMP (14). Additionally, these sites show significant amino acid sequence homology to the two kinetically distinct cAMP-binding sites of cAK. The more amino-terminal site of cGK shows higher homology to the more amino-terminal site of cAK, and conversely, the more carboxyl-terminal site of cGK shows higher homology to the more carboxyl-terminal site of cAK (15). Based solely on the sequence homology between the cyclic nucleotide-binding sites in cAK and cGK, it has been thought that the more amino-terminal cGMP-binding site of cGK is the fast site and its adjacent site is the slow site. However, no experimental evidence has been presented to support this prediction.
Studies of cGK have been facilitated by the cloning of cDNAs for type I␣ cGK from bovine tracheal smooth muscle (16) and type I␤ cGK from human placenta (17) and show that these two isoenzymes have identical amino acid sequences in the cyclic nucleotide-binding domains. Type II cGK recently has been cloned from mouse brain (18), rat intestine (19), and rat brain (20) and shows only 45% amino acid sequence identity of its cyclic nucleotide-binding domains with those of the type I cGKs. The presence of fast and slow cGMP-dissociation sites in cGK probably has physiological significance. The very large difference in cGMP dissociation rates allows for cGMP binding to the slow site at low cGMP concentrations in vitro, and occupation of the slow site has been associated with partial activation of cGK (21). If this occurs in vivo, it would be expected that under conditions of low to intermediate levels of cGMP, slow-site occupation is the main determinant of cGK activity, and substantial elevations in cGMP would cause further increases in cGK activity due to occupation of the fast site.
Determination of the crystal structures of the cAMP-binding domain of catabolite gene activator protein (22)(23)(24) and a truncated form of the regulatory subunit (R subunit) of cAK (25) has provided a framework on which to model conserved residues of the evolutionarily related cyclic nucleotide-binding sites of cGK (26) and ion channels (27). A conserved threonine or serine is present in the sites that are selective for binding cGMP, whereas an alanine occupies the same position in sites that are selective for cAMP (28). Substitution of a threonine for the alanine in the cAMP-binding sites of type I␣ R subunit of cAK significantly improves the affinity of R subunit for cGMP with little effect on the affinity for cAMP (29 -30). Thus, these results experimentally verify the prediction by Weber et al. (26) that a threonine in this position provides for cGMP/cAMP selectivity. Based on sequence homology, it has been predicted that the conserved threonine in cGMP-gated ion channels would also impart cGMP selectivity (30). In fact, mutation of this threonine in cGMP-gated ion channels results in the loss of cGMP/cAMP selectivity by decreasing the affinity of the channel for cGMP while slightly increasing its affinity for cAMP (31). These results are consistent with the hypothesis that the threonine hydroxyl forms a hydrogen bond with the 2-NH 2 group of cGMP. The importance of the conserved threonines in the cyclic nucleotide-binding sites of cGK and the location of the slow and fast cGMP-binding sites in the primary sequence were examined by mutating the conserved threonine residue in each cGMP-binding site of the human type I␤ cGK.

EXPERIMENTAL PROCEDURES
Subcloning and Mutagenesis of Human Type I␤ cGK cDNA-Three partial cDNA clones for the wild type human type I␤ cGMP-dependent protein kinase (cGK) (17) were subcloned to produce the entire cGK open reading frame. The 580-base pair EcoRI/NcoI fragment of clone 3 (containing 65 base pairs of 5Ј-nontranslated region) and the 894-base pair SacI/HincII fragment of clone 1 (containing 219 base pairs of 3Ј-nontranslated region) were excised using the respective restriction endonucleases (New England Biolabs). Fragments were ligated into analogous restriction sites in clone 2 to produce the full-length clone of 2058 base pairs in the pUC18 vector which was propagated in Escherichia coli (TG1, Amersham Corp.). The 2335-base pair EcoRI/HincII fragment containing the full-length cGK clone was ligated into the EcoRI and SmaI unique sites of the baculovirus transfer vector pVL1392, (generous gift from T. R. Soderling, Oregon Health Science Center, Portland, OR), to produce nonfusion proteins. The 1441-base pair EcoRI/SacI fragment containing the coding region for both cGMPbinding sites was ligated into pBluescript IIKSϩ (Stratagene) for oligonucleotide-directed mutagenesis based on the method of Kunkel et al. (32) using a commercially available kit (Invitrogen). Oligonucleotides (Vanderbilt University DNA Core Facility) with the following sequences were constructed to be complementary to the cDNA-coding strand except for the underlined nucleotides that serve to mutate the cGMPbinding site threonines to the residues indicated. T193A, 5Ј-CTTGACG-GTCGCTGCCCGGGTACAGTTGTA-3Ј; T193S, 5Ј-CTTGACGGTCGC- The double mutant (T193A/T317A) also was produced prior to subcloning back into the parent pVL1392-hcGKI␤ vector. pVL1392 vectors and pBluescript IIKSϩ vectors were propagated in E. coli (DH5␣ and XL1 Blue, respectively). All plasmids was sequenced manually (U. S. Biochemical Corp.) or on an Automated Biosystems, Inc. DNA Sequencer 373A by the dideoxy chain-termination method (33) prior to co-transfection. The human and bovine type I␤ isoforms are identical on the predicted amino acid sequence level except lysine 280 in bovine is a threonine in human and asparagine at position 290 in bovine is a serine in human. Both of these changes lie between the two cGMP-binding sites (17).
Expression of Wild Type and Mutant cGK-All tissue culture procedures were performed in Sf9 insect cells (Spodoptera frugiperda, Invitrogen) maintained at 27°C in Excell 401 serum-free media (JRH Biosciences). pVL1392-hcGKI␤ transfer vectors (4 g) were co-transfected with linear wild type Autographa californica nuclear polyhedrosis virus DNA (1 g) using cationic liposomes (Invitrogen) as a carrier. Recombinant viruses were harvested 3 days after co-transfection and purified by two rounds of agarose overlay plaque assays (34). Purified recombinant viruses were amplified by infection of Sf9 cells for 7 days in 25-cm 2 flasks (Falcon). Infected cells were harvested by gentle rapping and pelleted by centrifugation (1000 ϫ g for 10 min at 20°C). High titer viruses in the supernatant were harvested as extracellular viral particles and stored at 4°C as stock for further experiments or screened for the presence of hcGKI␤ recombined genes by polymerase chain reaction techniques (34) using commercially available polymerase chain reaction techniques primers (Invitrogen). The cell pellet containing cGK was resuspended in cold, sterile KPEM (10 mM potassium phosphate, pH 6.8, 1 mM EDTA, 25 mM 2-mercaptoethanol) plus the protease inhibitors pepstatin A (1 g/ml) and leupeptin (0.5 g/ml) (Sigma) and homogenized on ice by a 10-s burst in an Ultra-Turrax microhomogenizer (Janke and Kunkel). Crude extracts were tested as described below for cGK activity, [ 3 H]cGMP-binding activity, and immunoreactivity to rabbit anti-bovine lung type I␣ cGK antibodies to confirm expression of wild type and mutant cGK.
Purification of Wild Type and Mutant cGK-10 8 cells were infected with recombinant baculoviruses at a multiplicity of infection of 1 in a 100-ml spinner flask (Bellco) at 100 rpm. At 96 h postinfection, cells were harvested by centrifugation (1000 ϫ g for 10 min at 20°C), and the cell pellet was frozen in a dry ice/ethanol bath for 20 min prior to storage at Ϫ70°C until analyzed. All subsequent steps were performed at 4°C and were modified from purification procedures for bovine type I␤ cGK (35). The cell pellet was thawed on ice in 10 ml of KPEM plus protease inhibitors (see above), resuspended, and homogenized in 100 ml of KPEM plus protease inhibitors by 4 ϫ 10-s bursts in the microhomogenizer. The crude extract was centrifuged (20,000 ϫ g for 30 min at 4°C), and the supernatant, which contained 95% of the cGK activity, was applied to an aminoethylamino (C-8 position) cAMP-agarose affinity column (0.9 ϫ 8 cm) (Sigma) that had been pre-equilibrated in KPEM. The column was washed sequentially with 100 ml of KPEM containing 2 M NaCl and then 10 ml of KPEM. The cGK was immediately eluted with 10 mM cAMP in KPEM, and the fractions were tested for cGK activity. Fractions containing activity were pooled and applied to a DEAE-Sephacel column (0.9 ϫ 5 cm, Pharmacia Biotech Inc.) that had been pre-equilibrated in KPEM at 4°C. The column was washed with 150 ml of KPEM to allow for dissociation and removal of cAMP from the cGMP-binding sites of cGK bound to the DEAE-Sephacel. Proteolyzed monomeric forms of the enzyme were eluted by washing the column with 50 ml of KPEM containing 0.12 M NaCl (36). The fulllength cGK was eluted with 10 ml of KPEM containing 0.3 M NaCl. cGK was flash-frozen in liquid nitrogen after the addition of sucrose to 10% final concentration. cGK could be stored for several months at Ϫ70°C with only minor loss of activity or at 4°C for several weeks. Native type I␤ cGK was purified from bovine aorta smooth muscle by the method of Francis et al. (35).
Protein Kinase Activation Assays-The kinase activities of native, wild type and mutant cGK were determined by the method of Sodium Dodecyl Sulfate-Polyacrylamide Gel Electrophoresis of cGK-500 ng of purified protein was boiled for 5 min in the presence of 10% SDS, 2 M 2-mercaptoethanol, and bromphenol blue (1 mg/ml) and subjected to 8% SDS-polyacrylamide gel electrophoresis (41). Proteins were visualized by Coomassie Brilliant Blue staining.
Protein Quantification-Protein quantities were determined by the method of Bradford (42) using staining reagent from Bio-Rad and bovine serum albumin fraction V (Sigma) as standard. This method routinely overestimated the amount of protein by 37% (37). RESULTS A baculovirus expression system was utilized to overexpress the wild type (WT) human type I␤ cGK, cGMP-binding site A mutants (T193A, T193S, and T193V), and cGMP-binding site B mutants (T317A, T317S, and T317V) as well as the double mutant, T193A/T317A, for structure/function studies (Fig. 1,  top). Expression of recombinant cGKs was achieved by infec-tion of Sf9 cells with plaque-purified recombinant baculoviruses. The time course of expression was similar for all constructs tested (not shown). The appearance of active, fulllength cGKs was determined by measuring cGMP-binding activity and protein kinase activity in the absence and presence of cGMP and by performing immunoblots with anti-type I␣ cGK antibodies. All measures indicated an increase in expression at 48 h post-infection with maximum expression of active kinase at 96 h, which was taken as the optimal time of expression for all recombinant cGKs. Infections were harvested from 100-ml spinner flasks for subsequent purification of cGK as described under "Experimental Procedures." The yield of purified cGKs ranged from 0.6 mg to 1.3 mg per 10 8 cells infected. On SDS-polyacrylamide gel electrophoresis gels (Fig. 1, bottom) the wild type and mutant cGKs contained a major band of protein of approximately 78 kDa that was essentially homogeneous and co-migrated with native type I␤ cGK purified from bovine aorta. Each of the cGKs also crossreacted with cGK antibodies in immunoblot experiments (not shown). Only trace amounts of proteolytic products were present. All enzymes (native, wild type, and mutants) had specific kinase catalytic activities ranging from 1.5 to 2.7 mol/min/mg (Table I). These values were not considered to be different from each other since the values for the wild type or native cGK ranged from 1 to 4 mol/min/mg. The specific protein kinase activity, the K a of cGMP for kinase activation, the kinetics of cGMP dissociation, and the immunoreactivity of the recombinant wild type cGK were nearly identical to those of native bovine type I␤ cGK (Table I). This was to be expected since the two enzymes (native bovine and wild type recombinant human) differ by only two amino acids. Similar results were observed by Pohler et al. (43).
The importance of the conserved threonine in each of the cGMP-binding sites was examined by mutating these residues to alanines. The kinase catalytic activities of the wild type and mutant cGKs were then determined at increasing concentrations of cAMP (Fig. 2) or cGMP (Fig. 3), and Hill plots were constructed to determine kinase activation constants (K a ) for both cAMP (Fig. 2, bottom panel) and cGMP (Fig. 3, bottom  panel). K a is the concentration of cyclic nucleotide required for half-maximal activation of the enzyme. The K a for cAMP of all enzymes showed no significant changes (Fig. 2) indicating that the overall structure and function of the binding sites were preserved. The mutation of threonine 193 to alanine caused a 27-fold increase in the K a for cGMP, whereas the threonine to alanine mutation at position 317 resulted in only a 2-fold increase in K a for cGMP. The double threonine to alanine mutation produced a 63-fold increase in the K a . Thus, the FIG. 1. Top, linear arrangement of functional domains of cGK. One subunit of the cGMP-dependent protein kinase homodimer is represented. Locations of the cGMP-binding site mutations are indicated with asterisks. DIM is the dimerization domain, and INHIB is the inhibitory/autophosphorylation domain. Cyclic nucleotide-binding sites (A and B) have previously been referred to as sites 2 and 1, respectively. Type I␣ and type I␤ cGK differ in amino acid sequence in the dimerization and inhibitory domains only. Bottom, comparison of native, wild type, and mutant purified type I␤ cGKs. Purified cGK enzymes (500 ng) were electrophoresed on 8% SDS-polyacrylamide minigels and stained with Coomassie Brilliant Blue dye as described under "Experimental Procedures." T193A mutation produced a far greater effect on the K a for cGMP than was predicted for modification of the fast site (presumed low affinity). Conversely, the T317A mutation had a much smaller effect on the activation constant for cGMP than was predicted for a slow site (presumed high affinity) mutant. These results suggested that the high affinity cGMP-binding site in type I␤ cGK is the more amino-terminal binding site in the amino acid sequence, i.e. site A, and that the fast and slow cGMP-binding sites in type I␤ cGK are reversed from their original assignments. Autophosphorylation of cGK causes an increase in cyclic nucleotide binding affinity, resulting in a decrease in the K a of the enzyme for cyclic nucleotide (44). The kinase activation studies described above were performed under conditions (30min incubations of enzyme in the presence of cyclic nucleotide, Mg 2ϩ , and ATP) that would promote significant autophosphorylation of cGK (0.5 mol of P i incorporated per mol of subunit at 30 min). 2 We therefore repeated these determinations of K a values, but the incubation time was reduced to 5 min (0.2 mol of P i incorporated per mol of subunit at 5 min). 2 The relative order of affinities of the wild type and mutant cGKs was approximately the same in the 5-min assays as in the 30-min assays (Table I), thus confirming the importance of the threonine hydroxyl for cGMP-specific binding and also the structural order of the fast and slow cGMP-binding sites.
To confirm the importance of the conserved threonine in each cGMP-binding site and to provide further evidence that the structural order of the two sites in cGK is reversed from its previous assignment, the dissociation of [ 3 H]cGMP from native, wild type, and each threonine to alanine mutant enzyme was determined in the presence of a 100-fold excess of unlabeled cGMP (Fig. 4). Native and wild type cGK exhibited similar biphasic dissociation curves consistent with cGMP dissociation from two sites with distinctly different affinities, i.e. fast and slow dissociation sites. The initial steep slope represents dissociation of [ 3 H]cGMP from the fast site whereas the more shallow slope measures [ 3 H]cGMP dissociation from the slow site. In the T317A mutant, the slow component of the [ 3 H]cGMP dissociation curve (t 1 ⁄2 ϭ 5.6 min) was essentially unchanged from that of wild type or native type I␤ cGK (t 1 ⁄2 ϭ 5 min), but the fast component nearly disappeared. Conversely, T193A displayed a dramatic alteration of the slow component 2 Jeffrey A. Smith, personal communication.  (Table I) for native, wild type, and the T317A mutant; however, the values for the T193A mutant and double mutant were considerably lower, perhaps due to partial saturation of these lower affinity mutants even at the highest concentrations of [ 3 H]cGMP, or to partial retention of the bound [ 3 H]cGMP in the Millipore filtration binding assay. These results indicated that substituting an alanine for the conserved threonine in either cGMP-binding site of cGK markedly reduces the affinity of that site for cGMP and provided further support to the interpretation that in type I␤ cGK the slow site is amino-terminal to the fast site.
To study the selectivity of cGMP binding imparted by the conserved threonine in the cGMP-binding sites, these same threonines were converted to serine and valine, and these mutant cGKs were also analyzed for changes in kinetic properties of cGMP binding. Substitution of a serine for the threonine in the B domain (T317S) had no effect on the K a for cGMP and only a 4-fold increase in K a for cGMP when the threonine in the A domain was mutated to serine (T193S) ( Table I). Threonine to valine mutations in either site led to K a constants that were even higher than those for the threonine to alanine mutations; when compared to wild type cGK, the K a for cGMP of T193V was 29-fold higher whereas T317V was 3-fold higher. Once again, the K a of these mutants when assayed with cAMP did not change appreciably. [ 3 H]cGMP dissociation from the serine or valine mutants directly reflected the changes observed with the K a for cGMP. When the specific threonines were changed to serines, there were only slight decreases in affinity for cGMP at either site, whereas threonine to valine mutations produced much larger decreases in cGMP binding affinities. In both sites, the decreases in affinities were greater for the valine substitution than for the alanine substitution. Thus, replacement of a critical binding site threonine by serine, alanine, or valine leads to progressively decreasing affinity for cGMP, consistent with the hypothesis that a hydroxyl group is necessary for high affinity cGMP binding and kinase activation, whereas the more hydrophobic valine at this position serves as a negative determinant for either cGMP or cAMP. DISCUSSION The structural order for the two cGMP-binding sites of type I␣ cGK was originally assigned based on sequence homologies to the cyclic nucleotide-binding sites in the R subunit of cAK (15). The carboxyl-terminal cyclic nucleotide-binding site of cGK has greater homology with the carboxyl-terminal site in R subunit than with the more amino-terminal cyclic nucleotidebinding site, and the converse is true for the more aminoterminal sites in the two enzymes. Since Corbin et al. (4) and Weber et al. (11) assigned the more amino-terminal site of R subunits as the fast cAMP site, these strong sequence homologies supported the same assignment to cGK, i.e. the more amino-terminal site of cGK was predicted to be the fast site. In fact, the results presented in the current study indicated that the amino-terminal site of type I␤ cGK is the slow cGMPdissociation site. Caution should be used in extrapolating the results of the current study to type II cGK or even to type I␣ cGK even though types I␣ and I␤ cGK are identical in the sequences of their cGMP-binding sites and catalytic domains. The different amino termini of types I␣ and I␤ cGK have significant effects on the properties of the cGMP-binding sites as evidenced from different cGMP-dissociation behavior of type I␣ and I␤ (37) as well as from quite different cGMP analog selectivities (45). Low amino acid sequence identity between type I and type II cGK also make the assignment of cyclic nucleotide-binding site structural order for type II cGK particularly ambiguous.
The present results strongly support the interpretation that in type I␤ cGK the more amino-terminal cyclic nucleotidebinding site is the slow site, and the more carboxyl-terminal site is the fast site. The assignment of the structural order for the cyclic nucleotide-binding sites in cAK is based on less definitive proof than that provided here for cGK, but, if the assignment of the order of the sites in cAK is correct, then the fast and slow kinetic characteristics associated with the different sites may have developed independently after the point in evolution when cAMP/cGMP selectivity appeared. The reversal of the slow and fast sites in type I␤ cGK compared with the sites in cAK could suggest differences in the mechanism of activation of these enzymes, but it is also possible that tandem fast and slow sites that display positive cooperativity in activating catalysis, as seen for both cGK (21) and cAK (46), may work in a similar manner after cyclic nucleotide occupation to activate the catalytic domain, regardless of the structural order of cyclic nucleotide-binding sites.
The results of the present study confirm the importance of the conserved cGMP-binding site threonine for creating a high affinity attraction of cGMP compared with cAMP. In an effort to examine the role of the conserved threonines in cGMPbinding sites of cGK, the specificity and selectivity of serine, alanine, or valine replacement mutations were studied for their ability to activate cGK upon binding cGMP or cAMP. Under all conditions examined, the threonine to serine mutants were very comparable with the wild type enzyme with only slight decreases observed for kinase activation constants. While a hydroxyl group imparts cGMP-specific binding, the threonine hydroxyl group appears to be more optimally positioned than does the serine hydroxyl group to hydrogen bond to the 2-NH 2 group of cGMP. Increasing the hydrophobicity of the side chains of the amino acid at this position, as in the valine substitution, provides a negative determinant for cGMP binding in cGK as evidenced by the resulting decrease in cGMP affinity compared with the alanine substitution. While the threonine is responsible for high affinity interactions of the cGMP-binding site with cGMP, none of the current substitutions into that location served to increase the affinity of the site for cAMP. This suggests that the conserved threonine only provides high affinity cGMP-binding contacts and that some other residue(s), perhaps outside of the primary binding pocket, serve to dampen the affinity of the sites for cAMP, thereby increasing the selectivity of the site for cGMP over cAMP. Having experimentally established the location of the fast and slow cyclic nucleotide-dissociation sites of type I␤ cGK now allows for more detailed analysis of the residues responsible for imparting cGMP/cAMP selectivity as well as for addressing the cGMP-specific binding characteristics observed in other cGKs.