The activity of cGMP-dependent protein kinase Iα is not directly regulated by oxidation-induced disulfide formation at cysteine 43

The type I cGMP-dependent protein kinases (PKGs) are key regulators of smooth muscle tone, cardiac hypertrophy, and other physiological processes. The two isoforms PKGIα and PKGIβ are thought to have unique functions because of their tissue-specific expression, different cGMP affinities, and isoform-specific protein-protein interactions. Recently, a non-canonical pathway of PKGIα activation has been proposed, in which PKGIα is activated in a cGMP-independent fashion via oxidation of Cys43, resulting in disulfide formation within the PKGIα N-terminal dimerization domain. A “redox-dead” knock-in mouse containing a C43S mutation exhibits phenotypes consistent with decreased PKGIα signaling, but the detailed mechanism of oxidation-induced PKGIα activation is unknown. Therefore, we examined oxidation-induced activation of PKGIα, and in contrast to previous findings, we observed that disulfide formation at Cys43 does not directly activate PKGIα in vitro or in intact cells. In transfected cells, phosphorylation of Ras homolog gene family member A (RhoA) and vasodilator-stimulated phosphoprotein was increased in response to 8-CPT-cGMP treatment, but not when disulfide formation in PKGIα was induced by H2O2. Using purified enzymes, we found that the Cys43 oxidation had no effect on basal kinase activity or Km and Vmax values; however, PKGIα containing the C43S mutation was less responsive to cGMP-induced activation. This reduction in cGMP affinity may in part explain the PKGIα loss-of-function phenotype of the C43S knock-in mouse. In conclusion, disulfide formation at Cys43 does not directly activate PKGIα, and the C43S-mutant PKGIα has a higher Ka for cGMP. Our results highlight that mutant enzymes should be carefully biochemically characterized before making in vivo inferences.

The type I cGMP-dependent protein kinases play key roles in regulating vascular tone, intestinal motility, memory formation, and nociception in the spinal cord (1). The kinases are activated by cGMP, and cellular cGMP levels are increased by the activity of soluble and particulate guanylate cyclases, which are activated by nitric oxide or small peptides, respectively. Conversely, cellular cGMP levels are decreased by phosphodiesterases. Pharmacological cGMP-elevating agents include nitric-oxide donors (nitroglycerin, nitroprusside), direct guanylate cyclase activators (riociguat), and phosphodiesterase inhibitors (sildenafil, tadalafil). These agents are used clinically to treat cardiac ischemia, systemic and pulmonary hypertension, and erectile dysfunction (2).
The type I PKG gene produces splice variants (PKGI␣ and PKGI␤) 2 that differ in their first ϳ100 amino acids (3). The kinases have a similar domain structure, which can be roughly divided into N-terminal regulatory and C-terminal catalytic domains. The regulatory domain contains a number of functional subdomains. At the extreme N termini, leucine/isoleucine zippers (LZs) mediate homodimerization and facilitate binding to specific interacting proteins (4 -9). After the LZs, each isoform has a unique autoinhibitory (AI) loop containing an inhibitory sequence that binds within the catalytic cleft and blocks substrate access in the absence of cGMP. The LZ and AI domains differ between PKGI␣ and PKGI␤. Next come two tandem cyclic nucleotide-binding domains (cNBD-A and cNBD-B) that have preferences for binding cGMP over cAMP (ϳ2.4-fold higher for cNBD-A and ϳ240-fold for cNBD-B (10)). The catalytic domain has an N-terminal ATP-binding lobe and a C-terminal substrate recognition lobe, and the catalytic cleft is located in the crevice between these two lobes. In the canonical PKG activation pathway, cGMP binding to the cNBDs induces a conformational change in the regulatory domain that pulls the AI loop from the catalytic cleft, thereby activating the kinase by allowing substrate access to the catalytic center.
In addition to the canonical activation pathway, PKGI␣ has been reported to be directly activated by oxidation. In an earlier paper, Landgraf et al. (11) showed that metal-induced disulfide bond formation between Cys 118 -Cys 196 and/or Cys 313 -Cys 519 increased basal kinase activity to ϳ70% of the maximum activity that could be achieved with saturating amounts of cGMP. 3  end of the PKGI␣ LZ, activated the kinase in a cGMP-independent manner. Although PKG dimer formation is stably mediated by the LZ, Cys 43 oxidation and disulfide formation cause the two peptide chains in the dimer to become covalently linked. In addition to direct kinase activation, Cys 43 oxidation has been proposed to alter the K m and V max for substrates and to increase PKGI␣ binding to specific interacting proteins (12). A knock-in mouse containing a "redox-dead" C43S-mutant PKGI␣ shows phenotypes consistent with a loss of PKGI␣ signaling (13)(14)(15)(16)(17)(18)(19), and these results have been used to argue for a predominant role for Cys 43 oxidation in PKGI␣ regulation. However, the enzymatic properties of the C43S mutant PKGI␣ were not thoroughly investigated. Exactly how Cys 43 oxidation activates PKGI␣ is unknown. To explore the underlying mechanism, we began by comparing the activity of wild-type and C43S-mutant PKGI␣ under reducing and oxidizing conditions. Surprisingly, and in contrast to previous findings, we found that wild-type PKGI␣ activity was not directly increased by disulfide formation at Cys 43 and that Cys 43 oxidation had no effect on substrate phosphorylation. In addition, we found that the C43S mutation caused PKGI␣ to have a lower sensitivity to cGMP-induced activation.

Oxidation-induced PKGI␣ dimerization does not lead to increased kinase activity in intact cells
In the original report describing activation of PKGI␣ by Cys 43 oxidation, A10 cells were transfected with wild-type or C43S-mutant PKGI␣ and treated with H 2 O 2 , and myosin light chain phosphorylation was examined (12). H 2 O 2 treatment caused increased PKGI␣ disulfide bond formation in wild-type but not C43S-mutant PKGI␣, and disulfide bond formation was correlated with decreased myosin light chain phosphorylation, presumably through PKGI␣-induced activation of myosin phosphatase (5). Using a similar rationale, we examinedwhetherPKGI␣Cys 43 oxidationledtoincreasedphosphorylation of known PKGI␣ substrates in cells. We transfected 293T cells with expression constructs for RhoA together with wild-type or C43S-mutant PKGI␣ and treated cells with 8-CPT-cGMP or H 2 O 2 . 8-CPT-cGMP-treated cells showed robust RhoA phosphorylation with both wild-type and C43Smutant PKGI␣ (Fig. 1A, top panel, compare lanes 1 and 2 and compare lane 4 and 5). As expected, treatment with H 2 O 2 caused a pronounced increase in disulfide bond formation in wild-type but not C43S-mutant PKGI␣ (shown by gel shift under non-reducing conditions; Fig. 1A, bottom panel, compare lanes 1 and 3 and compare lanes 4 and 6). Unexpectedly, H 2 O 2 treatment had no effect on RhoA phosphorylation (Fig.  1A, top panel, compare lanes 1 and 3). Similar results were seen when vasodilator-stimulated phosphoprotein (VASP) phosphorylation was examined. PKG phosphorylates VASP at Ser 157 and Ser 239 , and phosphorylation at Ser 157 causes VASP to migrate with a higher apparent molecular weight. Upon Western blotting of transfected cells probed with a VASP Ser(P) 239 antibody, VASP runs as a doublet, and 8-CPT-cGMP treatment causes VASP to completely shift to the upper band, indicating double phosphorylation of Ser 157 and Ser 239 (Fig. 1B, compare lanes 1 and 2 and compare lanes 4 and 5). However, in H 2 O 2 -treated cells, there is only a slight shift in VASP migration, and this shift is seen in cells transfected with wild-type or C43S-mutant PKGI␣, indicating that it is not due to Cys 43 oxidation (Fig. 1B, compare lanes 1 and 3 and lanes 4 and 6). These results strongly suggest that Cys 43 disulfide bond formation does not activate PKGI␣ in cells.

Oxidation of PKGI␣ Cys 43 does not increase kinase activity in vitro
To directly examine whether disulfide formation at Cys 43 increases kinase activity, we performed in vitro kinase assays. We used anti-Flag affinity beads to purify Flag-tagged PKGI␣ from transiently transfected 293T cells; this one-step purification protocol allowed us to quickly isolate highly purified fulllength wild-type and C43S-mutant PKGI␣ ( Fig. 2A). The purified kinases were incubated for 1 h in 15 mM DTT or allowed to oxidize by exposure to air in the absence of DTT. Immediately prior to measuring kinase activity, aliquots were removed for analysis by non-reducing SDS-PAGE/immunoblotting (as seen in Fig. 2B); under these conditions wild-type PKGI␣ was ϳ3.8% oxidized in the presence of DTT and ϳ59.7% oxidized in the absence of reducing agent. As expected, C43S-mutant PKGI␣ did not form a disulfide bond and ran as a monomer under both reducing and oxidizing conditions. Kinase assays performed using the small peptide Glasstide as a substrate revealed that oxidation slightly increased the basal activity of both wild-type and C43S-mutant PKGI␣ (Fig. 2C); basal activity of wild-type PKGI␣ increased from 0.137 to 0.206 pmol/min/g, whereas activity of the C43S-mutant kinase increased from 0.170 to 0.214 pmol/min/g (maximum kinase activity in the presence of cGMP was ϳ6.2 pmol/min/g under all conditions). Identical results were found using Kemptide as a substrate (supplemental Fig. S1). It was recently reported that PKGI␣ Cys 43 disulfide bond formation increased in vitro PKGI␣ activity toward histone H1 but not RhoA (16). However, in contrast to those results, we found that Cys 43 oxidation did not increase PKGI␣ activity toward histone H1 (Fig. 2D); in fact, we did not even observe the slight oxidation-induced increase in basal activity of wild-type and mutant enzyme seen with small peptide substrates. Thus, under our conditions, oxidation of Cys 43 did not directly activate PKGI␣ in vitro.

PKGI␣ substrate affinity is not altered by oxidation at Cys 43
PKGI␣ Cys 43 oxidation has been reported to change the enzyme kinetics (12). In previous studies, Burgoyne et al. (12) found that in the absence of cGMP, oxidation decreased the K m for Glasstide from 247 to 37 M but had no apparent effect on V max , whereas in the presence of cGMP, oxidation lowered the K m for Glasstide from 289 to 89 M and lowered V max by ϳ60% compared with the reduced enzyme. Because we found that oxidation did not directly activate PKGI␣, we next examined whether oxidation altered PKGI␣ affinity for peptide substrates. Using purified wild-type and C43S-mutant PKGI␣, we performed in vitro kinase assays under the same oxidizing or reducing conditions described above. In the presence of cGMP, we found that oxidation/reduction caused no difference in the K m or V max for Glasstide in wild-type or C43S-mutant PKGI␣ (Fig. 3, A and B). In the absence of cGMP, oxidation slightly increased V max in both wild-type and C43S-mutant PKGI␣ (note the different scales in Fig. 3). This slight increase in basal activity was similar to what was seen in Fig. 2C and was not due to disulfide bond formation at Cys 43 , because it was the same in wild-type and C43S-mutant enzyme. The complete set of K m and V max values are listed in Table 1. Fig. 3E shows the amount of reduced versus Cys 43 -cross-linked PKGI␣ for each reaction condition.

PKGI␣ C43S has a reduced affinity for cGMP
We next checked whether oxidation altered the response of PKGI␣ to cGMP. Using wild-type and C43S-mutant PKGI␣ under oxidizing and reducing conditions, we performed in vitro kinase assays in the presence of increasing amounts of cGMP (Fig. 4). We found that oxidation of wild-type PKGI␣ had no significant effect on the K a for cGMP: K a ϭ 0.074 Ϯ 0.003 and 0.072 Ϯ 0.004 M for the reduced and oxidized wild-type enzyme, respectively. However, unexpectedly, we found that the C43S mutation caused PKGI␣ to require a higher amount of cGMP to half-maximally activate the kinase: K a ϭ 0.373 Ϯ 0.017 and 0.337 Ϯ 0.015 M for the reduced and oxidized mutant enzyme, respectively. Thus, whereas reduction/oxidation had no effect on cGMP-induced activation of wild-type or C43S-mutant PKGI␣, the C43S mutation caused PKGI␣ to be less sensitive to activation by cGMP.
Taken together, our results demonstrate that PKGI␣ is not directly activated by Cys 43 oxidation, and Cys 43 oxidation does not change the kinetic properties of the enzyme toward substrates. However, the C43S mutation causes an ϳ5-fold decrease in cGMP affinity of the enzyme. This loss in cGMP sensitivity could explain the reported phenotype of the C43S knock-in mouse, which appears to have a loss of PKGI␣ function.

Discussion
Since its discovery in the 1970s, PKG has been extensively studied; however, some of its biochemical properties are still incompletely defined. Recent structural studies have demonstrated unique features that play important roles in regulating cGMP-induced kinase activity, including the structural basis for cGMP selectivity and novel interchain contacts that regulate kinase activation (10,20,21). Another recently proposed mechanism for PKGI␣ regulation was direct activation by oxidation-induced disulfide bond formation at Cys 43 . How the formation of a disulfide bond at Cys 43 activates the kinase was not determined, and investigating the mechanism was the starting point for our current study.
In contrast to previous reports, we found that oxidation of PKGI␣ at Cys 43 does not directly activate PKGI␣. In addition, Cys 43 oxidation had no effect on substrate affinity or reaction velocity using Glasstide as a substrate, and oxidation did not lead to increased histone H1 phosphorylation. Although it is unclear why our results differ from earlier studies, one factor may be that the original studies were performed using PKGI␣ purchased from a commercial vendor and that the kinase could only be stimulated 1.5-3-fold by cGMP (12). This suggests that the kinase was proteolytically degraded, leading to a largely cGMP-independent kinase activity, and the poor quality of the enzyme may have affected the biochemical assays. In a subsequent paper by Prysyazhna et al. (16), PKGI␣ was purified using a cAMP-agarose affinity column, followed by cAMP elution and dialysis to remove cAMP. This kinase could be stimulated ϳ8-fold by cGMP as determined by Western blotting performed with phospho-specific antibodies to detect RhoA and histone H1 phosphorylation (see Fig. 6A in Ref. 16). In our experience, it is very difficult to fully remove cAMP during PKG purification, especially from PKGI␣ (10,20,22), and in general, Western blotting is semi-quantitative and is not an accurate method to measure kinase activity.
It could be argued that the N-terminal Flag tag on our PKGI␣ constructs affected kinase activity; however, we note that the kinase used in the present studies was purified intact, had a low basal activity, and could be stimulated 20 -50-fold by cGMP. In addition, the Flag-tagged wild-type kinase had a K a for cGMP of 72-74 M, which is consistent with previous results measuring the K a of untagged PKGI␣ in cell lysates (23,24). Our V max and K m values for Glasstide differ from those obtained by Glass and Krebs (K m ϭ 28.8 M and V max ϭ 20 pmol/min/g (25)); the different values may reflect the variation in kinase assay conditions, because their assays were performed using 2 mM Mg 2ϩ , whereas we used 10 mM Mg 2ϩ . In addition, Glass and Krebs had found that for small peptide substrates, PKGI␣ activity peaked at 2 mM Mg 2ϩ and rapidly decreased as free Mg 2ϩ levels increased, whereas for large substrates like histone, activity steadily increased as free Mg 2ϩ reached 75 mM (26). In comparison with our assays, previous in vitro kinase assays determining   (12,16). We should also point out that the transfected kinases used for our in-cell assays were without epitope tag (Fig. 1), demonstrating that oxidation did not increase the activity of untagged PKGI␣ in cells. Therefore, we do not feel that the N-terminal Flag tag affected our results. We found that C43S-mutant PKGI␣ has an ϳ5-fold lower affinity for cGMP compared with the wild-type enzyme. This is an important finding, because mice homozygous for C43S-mutant PKGI␣ have phenotypes consistent with a loss of PKGI␣ function (17)(18)(19). These phenotypes include hypertension, insensitivity to nitroglycerin-induced vasodilation, and protection from septic shock; they have been interpreted to be the result of defective redox-sensing properties normally ascribed to Cys 43 disulfide bond formation. However, mice null for the ␤ 1 subunit of soluble guanylate cyclase are also resistant to nitroglycerin-induced vasodilation (27), indicating that the canonical NO-cGMP-PKG pathway is important and that oxidation sensing by Cys 43 disulfide formation is not the main mechanism for nitroglycerin-induced vasodilation. Although our current studies have not looked at PKGI␣ oxidation in mice, our results strongly suggest that defective signaling in PKGI␣ C43S mice is due at least in part to an increased K a for cGMP rather than a loss of redox-induced activation of the enzyme.
It is possible that loss of Cys 43 oxidation might also affect other aspects of PKGI␣ signaling; for example, oxidation of PKGI␣ may alter its interaction with specific interacting proteins, which have been shown to bind to the PKGI␣ leucine zipper (5,7,28). Indeed, oxidation has been shown to increase the in vitro interaction between PKGI␣ and two of its interacting proteins, RhoA and MYPT1 (12). In addition, a recent paper by Nakamura et al. (15) demonstrated that the C43S mutation appeared to alter PKGI␣ subcellular localization in cardiac myocytes, which suggests a change in association of with interacting/anchoring proteins in cells, but the structural basis for oxidation-induced changes in these interactions was not examined. We are currently pursuing these studies.
How does the C43S mutation lead to decreased cGMP affinity? Since PKGI␣ and PKGI␤ were first purified, it has been known that their different N termini cause the two isoforms to have different K a values for cGMP, even though the sequences of the cyclic-nucleotide-binding pockets are identical (23). We have previously used hydrogen/deuterium-exchange mass spectrometry to study the PKGI␤ regulatory domain and found that, in the presence of the N-terminal LZ and autoinhibitory subdomains, hydrogen/deuterium exchange was increased throughout the cGMP-binding pockets (29). The increased conformational dynamics correlate with increased cGMP affinity. Because small molecule ligands are thought to stabilize preexisting protein conformations (30,31), we reasoned that the N terminus shifted the ensemble of conformations that PKGI␤ adopts in solution, such that the cyclic nucleotide-binding pockets spend more time in conformations that resemble the cGMP-bound forms. Therefore, it is possible that the C43S mutation, which lies at the end of the LZ, causes a change in the conformational dynamics of the nucleotide-binding pockets, which leads to lower cGMP affinity. In a previous analysis of the PKGI␣ LZ, the C43S mutation lowered the T m for thermal denaturation from Ͼ108 to 81.4°C under oxidizing conditions and from 93.0 to 83.3°C under reducing conditions (32). Although these melting temperatures are not physiological, thermal denaturation measures melting of the entire LZ, and the lower T m in the C43S-mutant samples may reflect structural destabilization of the region surrounding Cys 43 , which could occur in the full-length protein under physiological conditions.
In conclusion, we found that disulfide formation at Cys 43 does not directly activate PKGI␣ in vitro or in intact cells. In addition, "redox dead" C43S-mutant PKGI␣ has a higher K a for cGMP, and this decreased cGMP affinity may at least partially explain the loss-of-function PKGI␣ phenotype observed in the C43S knock-in mice. Our results also highlight the general fact that mutant enzymes should be carefully characterized biochemically before cellular or physiological inferences are made.

Cell culture and transfection
293T cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum at 37°C in a 5% CO 2 atmosphere. The cells were transfected using Lipofectamine 2000 according to the manufacturer's instructions (Life Technologies).

RhoA and VASP phosphorylation in 293T cells and Western blotting
293T cells were split into 12-well cluster dishes such that they would be 90 -95% confluent 18 h later, at the time of transfection. The cells were transfected with expression vectors for wild-type or C43S-mutant PKGI␣ and RhoA or VASP as indicated in the figure legends. The next day, the wells were treated for 60 min with 250 M 8-CPT-cGMP, 100 M H 2 O 2 , or vehicle as indicated. The medium was aspirated, and the cells were directly lysed by adding non-reducing SDS sample buffer (60 mM Tris-HCl, pH 6.8, 2% SDS, 0.01% bromphenol blue, and 100 mM maleimide (to prevent oxidation during sample processing)). Cell lysates were transferred to microcentrifuge tubes and sonicated 2 ϫ 20 s at 1-watt power. The proteins were separated by SDS-PAGE, transferred to Immobilon-P, blocked in 5% nonfat dry milk in TBS, and blotted with the indicated antibodies. The blots were developed using SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific).

Protein expression and purification
The cells were split into a 6-well cluster dish and transfected with Flag-tagged wild-type or C43S PKGI␣ (three wells each). Approximately 20 h post-transfection, the cells were scraped in PBS and lysed in buffer A (PBS, 0.1% Nonidet P-40) containing 1ϫ protease inhibitor mixture (Calbiochem), and lysates were cleared by centrifugation (16,000 ϫ g, 10 min at 4°C). Cleared lysates were incubated with 20 l of anti-Flag M2 affinity gel (Sigma) for 1 h at 4°C with constant mixing. The beads were washed 2 ϫ 200 l of buffer A, 2 ϫ 200 l of PBS with 500 mM NaCl, and 2 ϫ 200 l of PBS. Bound proteins were eluted with 4 ϫ 10 l of elution buffer (PBS with 100 g/ml Flag peptide). For each elution step, the beads were incubated with buffer for 5 min on ice. The four eluates for each protein were pooled. Proteins were quantified by SDS-PAGE/Coomassie staining using BSA standards on the same gel. The gels were scanned, and quantification was performed using ImageJ.

Kinase oxidation/reduction and in vitro kinase assays
Flag-tagged wild-type and C43S PKGI␣ purified from transiently transfected 293T cells were diluted to ϳ1 ng/l in kinase dilution buffer (10 mM potassium phosphate, pH 7.0, 1 mM EDTA, and 0.1% BSA). For reduced samples, the dilution buffer contained 15 mM DTT. 10 l of diluted kinase was added to 5 l of 3ϫ kinase reaction mix (120 mM HEPES, pH 7.4, 1.5 mM Glasstide, 30 mM MgCl 2 , 150 M ATP, 180 Ci/ml [␥-32 P]ATP, and Ϯ 30 M cGMP). The reactions were performed for 1.5 min at 30°C and stopped by spotting on P81 phosphocellulose paper. Unincorporated [␥-32 P]ATP was removed by washing P81 paper four times for 5 min with 2 liters of 0.452% o-phosphoric acid. 32 P incorporation was measured by liquid scintillation counting. In some experiments, 1.5 mM Glasstide was replaced by 1.56 mg/ml Kemptide or 3 g of histone H1. The reactions with histone H1 were run for 8 min. For experiments examining enzyme kinetics, reactions were performed Ϯ 10 M cGMP with increasing amounts of Glasstide (0.005-1.0 M) or with 1.5 mM Glasstide in the presence of increasing concentrations of cGMP (0.003-10 M).

Data analysis
The data were analyzed using GraphPad Prism 7. The V max and K m values were measured using non-linear fit Michaelis-Menten analysis, and cGMP K a values were determined by plotting [agonist] versus normalized response with variable slope.
Author contributions-D. E. C. conceived the project. H. K., S. Z., and D. E. C. performed the experiments. R. B. P. and D. E. C. analyzed the data and wrote the paper.