Cysteine Redox Sensor in PKGIa Enables Oxidant-Induced Activation

See allHide authors and affiliations

Science  07 Sep 2007:
Vol. 317, Issue 5843, pp. 1393-1397
DOI: 10.1126/science.1144318


Changes in the concentration of oxidants in cells can regulate biochemical signaling mechanisms that control cell function. We have found that guanosine 3′,5′-monophosphate (cGMP)–dependent protein kinase (PKG) functions directly as a redox sensor. The Iα isoform, PKGIα, formed an interprotein disulfide linking its two subunits in cells exposed to exogenous hydrogen peroxide. This oxidation directly activated the kinase in vitro, and in rat cells and tissues. The affinity of the kinase for substrates it phosphorylates was enhanced by disulfide formation. This oxidation-induced activation represents an alternate mechanism for regulation along with the classical activation involving nitric oxide and cGMP. This mechanism underlies cGMP-independent vasorelaxation in response to oxidants in the cardiovascular system and provides a molecular explantion for how hydrogen peroxide can operate as an endothelium-derived hyperpolarizing factor.

Oxidant molecules can cause cellular damage, dysfunction, and disease, but also play crucial roles in homeostatic maintenance of healthy cells and tissues (13). The modification of proteins by oxidant species with a coupled alteration in function allows cells to sense oxidants and, therefore, to influence biological responses (46). Cysteinyl thiols in proteins can undergo posttranslational modifications in the presence of oxidants that are important initiators of redox signaling (711). Here we report that guanosine 3′,5′-monophosphate (cyclic GMP or cGMP)–dependent protein kinase (PKG), specifically the Iα isoform, is redox-sensitive and that oxidation directly activates the kinase. Oxidative stress causes interprotein disulfide bond formation between two cysteine 42 (Cys42) residues on adjacent chains in the PKGIα homodimer complex, rendering the kinase catalytically active, independently of cGMP. Disulfide-linked enzyme has increased affinity for substrate, whereas activation by cGMP causes an increase in the maximum velocity of the enzyme-catalyzed reaction (Vmax). Consistent with disulfide-mediated activation of PKGIα, hydrogen peroxide (H2O2) induces vasorelaxation of the coronary vasculature, consistent with its known function as an endothelium-derived hyperpolarizing factor (EDHF).

We found that the regulatory (RI) subunits of adenosine 3′,5′-monophosphate (cyclic AMP)– dependent protein kinase form interprotein disulfides during exposure to H2O2 (12), which leads to activation of the enzyme (13). In addition, we showed that PKGIα also forms an interprotein disulfide during oxidative stress, which was reversed by the reducing agent 2-mercaptoethanol (Fig. 1). PKG contributes to the regulation of fundamental biological processes, including growth and development, gene expression, nociception, learning, behavior, synaptic plasticity, and sexual function (14). In the cardiovascular system, PKG regulates blood pressure, excitation-contraction coupling, and platelet aggregation and, furthermore, has roles in diseases such as atherosclerosis, abnormal cardiac and vascular remodeling, and heart failure (15).

Fig. 1.

Disulfide dimerization of PKGIα in response to H2O2. Quantitative analysis (standard error bars shown) of immunoblots (see fig. S1A), showing the proportion of monomeric and dimerized kinase after exposure to various concentrations of H2O2.

PKG forms a parallel aligned homodimer, with each subunit containing both catalytic and regulatory domains. The two subunits are held together under nondenaturing conditions by a leucine zipper in the N-terminal regulatory domain, and the Cys42 residues in aligned PKG molecules are in close proximity (16). PKGI has two splice-variant isoforms, Iα and Iβ, differing at their N termini in ∼100 amino acids. The redox active Cys42 is unique to the Iα isoform. This was thought to be a structurally important constitutive disulfide (17), but our results suggest this is not the case, and reflects the susceptibility of this kinase to artifactual disulfide oxidation when tissue is homogenized in air. By including the thiol-alkylating agent maleimide in our preparation buffers (18), we trapped PKGIα in its in vivo redox state, revealing enhanced disulfide formation in tissue subjected to pro-oxidizing conditions (Fig. 1 and fig. S1A). The biochemical mapping of the interchain disulfide between Cys42 residues on adjacent chains (17) is consistent with a molecular model of the N-terminal 58 amino acids of the kinase, on the basis of nuclear magnetic resonance analysis under reducing conditions (16) (fig. S2). PKGIα disulfide formation occurs in tissue during H2O2 treatment or in vitro (without reducing agents) on exposure to air, consistent with an initial “priming” sulfenation (-SOH) of one Cys42 before subsequent reduction by the other on the parallel chain. Cys42 is surrounded by basic residues, which promote ionization to the thiolate anion, rendering the cysteines reactive, which, together with proximity of ∼8 Å (fig. S2) between interchain thiols, explains their susceptibility to disulfide formation.

We assessed whether nascent disulfide formation induced by H2O2 had a functional correlation. We found H2O2 treatment caused vasorelaxation in the heart, evidenced by a lowering of the coronary perfusion pressure during constant flow. Nitric oxide (NO) mediates vasodilation by binding to and activating soluble guanylate cyclase (sGC), producing cGMP which binds and activates PKG (1921). H2O2 treatment induced time-dependent vasorelaxation (Fig. 2A), comparable to that caused by NO. H2O2-induced PKGIα disulfide formation [measured by immunoblot (Fig. 2A and fig. S1B)] correlated well with relaxation [correlation coefficient (r) = 0.95, P < 0.0001]; the half-time (t1/2) for each was ∼1 min. We reasoned that PKGIα disulfide formation may directly contribute to vasorelaxation, and this could be independent of the NO-cGMP signaling pathway. To test this hypothesis, we conducted experiments using rat thoracic aortic vessels with intact endothelium (Fig. 2B). Vasorelaxation and PKG disulfide formation were observed with H2O2 (0.01 mM to 10 mM). The NO donor, N-[4-[1-(3-aminopropyl)-2-hydroxy-2-nitrosohydrazino]butyl]-1,3-propanediamine (spermine NONOate, 0.1 μM to 0.1 mM) also induced relaxation (but not disulfide formation), which was significantly attenuated by pharmacological inhibition of sGC with 1H-[1,2,4]oxadiazolo[4,3-a]quinoxalin-1-one (ODQ, 5 μM) or PKG with Rp-8-bromo-guanosine 3′,5′-monophosphothioate (Rp-8-bromo-cGMPS, 100 μM). In contrast, H2O2-induced relaxation was inhibited by Rp-8-bromo-cGMPS, but not ODQ. Again H2O2-mediated relaxation correlated with the extent of PKGIα oxidation (r = 0.80, P < 0.0001). However, higher concentrations of oxidant were required to produce relaxation or disulfide formation in these vessels compared with isolated hearts. The exact reason for this is unclear, but probably relates to differences in endogenous oxidant generation, peroxidase enzymes, and variable half-life and compartmentalization of oxidants. Aortic rings have a lower basal amount of PKGIα disulfide dimer than hearts, which may reflect enhanced antioxidant status, possibly explaining why higher, albeit physiological, concentrations of H2O2 concentrations are required to relax these preparations. These studies indicate that the H2O2-mediated relaxation is independent of the sGC-cGMP pathway. The existence of other EDHFs is indicated by experiments showing that vasorelaxation occurs independently of the established NO and prostacyclin pathways. In some vascular beds, H2O2 functions as an EDHF (22). Our observations provide a molecular explanation for how H2O2 mediates signaling via PKG without increases in cellular concentrations of cGMP. Insulin also caused disulfide oxidation in PKGIα, which is consistent with its ability to generate superoxide and H2O2 (5) and non–endothelium-dependent relaxation (23). However, the extent of insulin-mediated oxidation cannot explain the vasorelaxation induced by this hormone in our studies.

Fig. 2.

(A) Effect of H2O2 on vasorelaxation and PKGIα dimerization (which occur correlatively). Proportion of PKGIα in dimerized form (left panel) or vasorelaxation (right panel) after treatment of isolated perfused rat hearts for the indicated times (standard errors shown, r = 0.95, P < 0.0001). Effect of NO donor S-nitroso-N-acetylpenicillamine (SNAP) is shown in the dotted line. (B) Vasorelaxation in isolated thoracic aortic rings induced by NO (top left), H2O2 (top middle), or insulin (bottom left) and dimerization of PKG by H2O2 (top right) or insulin (bottom right). Shown are the effects of inhibition by sGC (ODQ, 5 μM) or PKG (Rp-8-bromo-cGMPS, 100 μM) on NO or H2O2-mediated relaxation. *P < 0.05 compared with control. The degree of relaxation with the extent of disulfide oxidation correlated (r = 0.80, P < 0.0001).

We further substantiated disulfide formation as a direct activator of recombinant PKGIα in in vitro kinase assays with the PKG substrate peptide Glasstide. cGMP treatment increased the Vmax of the kinase by 45 ± 14%, but had little effect on its Michaelis constant (Km) for substrate (Fig. 3A). In contrast, disulfide oxidation had little effect on Vmax, but decreased the enzyme's Km from 247 to 36 μM(P < 0.05). The increased affinity induced by oxidation was reversible, as dithiothreitol (DTT) treatment of the activated disulfide kinase returned activity to basal level (fig. S4A). Disulfide activation is specific to the Iα isoform, as the Iβ splice variant was not redox modulated (fig. S4B). We also determined the binding constant (Kd) of wild-type PKGIα or the mutant with Ser substituted for Cys42 (Cys42Ser or C42S mutant) for its physiological substrate RhoA (Fig. 3B). Oxidation enhanced the wild-type kinase's substrate affinity 13-fold (4.6 to 0.36 μM), but had no effect when the redox cysteine was replaced with serine. In subcellular fractionation studies from hearts (Fig. 3C), H2O2 induced translocation of disulfide PKGIα from the cytosol to the membrane and myofilament fractions (compartments where principal PKG substrates are found), consistent with the oxidation-induced increase in substrate affinity. Subcellular translocation of other kinases as a result of posttranslational modifications that enhance their substrate affinity have been reported (24). Our observations concur with a study showing PKGIα activation by metal ion–mediated oxidation (25).

Fig. 3.

(A) Effect of oxidation and cGMP on enzyme kinetics. Purified PKGIα and [32P]ATP were used in in vitro kinase assays with Glasstide substrate. PKG with (right) and without (left) cGMP. (B) Effect of oxidation on the affinity (Kd) of PKG for RhoA substrate protein. (Left) Wild-type reduced or oxidized; (right) mutant reduced or oxidized. (C) Translocation of disulfide PKGIα from the cytosol to membrane or myofilament-nuclear fractions in isolated hearts treated with the indicated concentrations of H2O2.

To test the role of oxidation in direct activation of PKGIα, we overexpressed FLAG-tagged wild-type or redox-insensitive Cys42Ser mutants in A10 smooth muscle cells. Overexpressed wild-type PKGIα, like endogenous enzyme, formed disulfide bonds following H2O2; however, the Cys42Ser mutant was unaffected (Fig. 4A). The principle mechanisms by which PKG increases vasorelaxation involve the phosphorylation of the large-conductance, Ca2+-activated K+ (BKCa) channel (ultimately decreasing intracellular concentration of free calcium), as well as activation of the myosin phosphatase complex (14). Myosin phosphatase (MYPT1) dephosphorylates the myosin light chain (MLC) to decrease smooth muscle myofilament sensitivity. We observed H2O2-induced dephosphorylation of MLC in cells expressing wild-type PKGIα (Fig. 4A). Furthermore, cells expressing redox-insensitive PKGIα demonstrated no detectable dephosphorylation of MLC. Confocal imaging of FLAG-tagged PKGIα showed that wild-type kinase relocalized in A10 cells after H2O2 treatment, but the Cys42Ser mutant did not (Fig. 4A). Wild-type or Cys42Ser mutant PKGIα was affinity-purified using 8-(2-aminoethyl) thioguanosine 3′,5′-monophosphate immobilized on agarose (8-AET-cGMP-agarose) from A10 cells treated with or without H2O2 (Fig. 4B). Affinity-purified PKGIα migrated as a disulfide dimer, shown by analysis using non-reducing SDS–polyacrylamide gel electrophoresis of cells exposed to H2O2, whereas the Cys42Ser mutant remained monomeric. The PKG substrates BKCa channel and MYPT1 copurified with disulfideoxidized wild-type kinase formed in response to H2O2, but mutant PKGIα did not form these associations. Disulfide formation of PKGIα was also induced by insulin (Fig. 4C), a hormone that is known to increase H2O2 concentration through stimulation of superoxide-generating oxidases (5).

Fig. 4.

(A) Effect of H2O2 on PKG substrate phosphorylation in A10 smooth muscle cells overexpressing wild-type or Cys42Ser (C42S) mutant PKGIα. (Left) FLAG-tagged Cys42Ser mutant PKGIα was overexpressed in A10 cells. D, disulfide dimer; M, reduced monomer. (Middle) FLAG-tagged wild-type PKGIα formed a disulfide bond in response to H2O2 and decreased MLC phosphorylation. *P <0.05 compared with control. Both of these H2O2-mediated events were absent in cells overexpressing Cys42Ser mutant PKGIα. (Right) Immunofluorescence was used to localize overexpressed PKG (stained with an anti-FLAG antibody) in A10 cells treated with or without H2O2. H2O2-treatment induced a cellular relocalization of wild-type PKGIα (note arrows), but this did not occur with the mutant kinase. (B) Affinity of wild-type or Cys42Ser PKGIα from transfected A10 cells treated with or without H2O2 for endogenous substrates. D, disulfide dimer; M, reduced monomer. Disulfide oxidized kinase associates with PKG substrates, the BKCa channel and MYPT1; whereas mutant Cys42 does not. Quantitative analyses of these data are shown in fig. S3B. (C) Effect of insulin on PKGIα disulfide formation in A10 cells.

We have described a pathway and mechanism by which the oxidant H2O2 directly activates PKGIα (fig. S5). All PKG isoforms have cGMP-binding cassettes (amino acids 102 to 240 in Iα), that indirectly couple kinase activity to cellular NO abundance. The amino acid sequence immediately N-terminal to these cGMP-binding sites in PKGI almost completely defines the sequence difference between α and β isoforms. The α isoform appears to have dual modes of activation, by NO-cGMP and by thiol oxidants, such as H2O2. These two sensors are side by side in the protein, seemingly optimally placed to operate as an allosteric trigger. This couples structural alterations in the kinase, either by disulfide oxidation or cGMP binding, to catalytic activity. The Cys42 redoxsensorin PKGIα is highly conserved throughout vertebrates. Consequently, this redox control of PKGIα may be a generic regulatory mechanism.

Supporting Online Material

Materials and Methods

Figs. S1 to S5


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article