The biology of cyclic GMP-dependent protein kinases.

cGKs belong to the family of serine/threonine kinases and are present in a variety of eukaryotes ranging from the unicellular organism Paramecium to Homo sapiens (1, 2). Mammals have two cGK genes, prkg1 and prkg2, that encode cGKI and cGKII. The N terminus (the first 90–100 residues) of cGKI is encoded by two alternatively spliced exons that produce the isoforms cGKI and cGKI . The enzymes have a rod-like structure and are activated at submicromolar to micromolar concentrations of cGMP (3, 4). They are composed of three functional domains: an N-terminal (A) domain, a regulatory (R) domain, and a catalytic (C) domain (for details see Refs. 1 and 2). The regulatory domain contains two tandem cGMP-binding sites that interact allosterically and bind cGMP with high and low affinity. Occupation of both binding sites induces a large change in secondary structure (5) to yield a more elongated molecule (6, 7). The catalytic domain contains the MgATPand peptide-binding pockets. Binding of cGMP to both sites in the R domain releases the inhibition of the catalytic center by the N-terminal autoinhibitory/pseudosubstrate site and allows the phosphorylation of serine/threonine residues in target proteins and in the N-terminal autophosphorylation site. Activation of heterophosphorylation may be preceded by autophosphorylation. Autophosphorylation increases the spontaneous activity of cGKI and cGKII (8–11) and is initiated by the binding of low cGMP concentrations to the high affinity site of cGKI (12, 13). In addition to controlling activation and inhibition of the catalytic center, the N terminus has two other functions: dimerization, cGKs are homodimers that are held together by a leucine zipper present in the N terminus, and targeting, the enzymes are targeted to different subcellular localizations by their N termini.

The smooth muscle IP 3 receptor type 1 coprecipitates with cGKI␤ and a 125-135-kDa protein identified as IRAG (31). IRAG is expressed together with cGKI␤ in smooth muscle, platelets, and some neurons (14). It is located at the endoplasmic reticulum membrane and is preferentially phosphorylated by cGKI␤ at Ser-683 and Ser-696 (bovine sequence) (32,33). cGKI␤-dependent phosphorylation of IRAG inhibits IP 3 -induced Ca 2ϩ release in intact and permeabilized cells. The IRAG sequence between amino acids 152 and 184 interacts specifically with the leucine zipper of cGKI␤ (33). IRAG has a coiled-coil domain that binds in vivo with the IP 3 receptor type I (32). Deletion of exon 12 that codes for the Nterminal part of the coiled-coil domain results in a hypomorphic IRAG allele (32). 8-Br-cGMP-dependent relaxation of muscarinic and ␣-adrenergic receptor-induced contraction of colon and aorta muscle strips, respectively, is abolished in these mice. In agreement, 8-Br-cGMP does not attenuate noradrenaline-induced Ca 2ϩ transients in isolated aortic smooth muscle cells of IRAG mutants (32). However, 8-Br-cGMP-induced relaxation of K ϩ -induced smooth muscle contraction is unchanged in the IRAG mutants. In contrast, 8-Br-cGMP is unable to affect Ca 2ϩ release and contraction in smooth muscles from cGKI-deficient mice. These finding show for the first time that cGKI has multiple targets in vivo that contribute to smooth muscle relaxation (Fig. 1).
An additional mechanism that lowers [Ca 2ϩ ] i is the direct phosphorylation of Ca 2ϩ -activated maxi-K ϩ (BK Ca ) channels by cGKI. Phosphorylation increased channel opening at constant [Ca 2ϩ ] i (34,35). The cGKI isozyme that phosphorylates BK Ca channels is not known. Opening of BK Ca channels hyperpolarizes the membrane and closes a number of channels, including L-type calcium channels, thereby reducing Ca 2ϩ influx. This mechanism contributes to the regulation of vascular tone, as shown in wild type and cGKIdeficient mice (22). The cGKI-dependent regulation of BK Ca channels depended on specific splice variants (36) and may be mediated indirectly by cGKI-dependent activation of an associated protein phosphatase 2A (37,38).
Regulation of BK Ca channels cannot account for the ability of 8-Br-cGMP to relax K ϩ -contracted smooth muscle strips of the IRAG mutant (32), because a change in BK Ca channel activity does not affect the membrane potential under these conditions. An alternative target for cGKI is MLCP because smooth muscle tone is decreased by dephosphorylation of the RLC (39). MLCP is a trimer comprising a 110 -130-kDa regulatory MBS also identified as myosin phosphatase targeting subunit (MYPT), a 37-kDa catalytic subunit and a 20-kDa protein of unknown function (40). Several studies have shown that the inhibition of MLCP activity can be linked to increased Ca 2ϩ sensitivity of smooth muscle contraction. Involved in this regulatory pathway are Rho kinase, arachidonic acid, and protein kinase C and its substrate CPI-17 (40). Phosphorylation of MBS at Thr-696 (human sequence) by Rho kinase or MYPT1 kinase inhibits the activity of MLCP allowing phosphorylation of RLC and contraction at constant [Ca 2ϩ ] i (25,40). MBS is phosphorylated by cGKI␣ at several sites (41,42). cGKI␣ is targeted specifically by its leucine zipper to MBS (41). cGKI␣-dependent phosphorylation of MBS at Ser-695 inhibits the phosphorylation at Thr-696 and thereby the decrease in MLCP activity (43). This mechanism would allow a reduction in RLC phosphorylation and relaxation at constant [Ca 2ϩ ] i .
The individual contribution of the described cGKI-dependent mechanisms regulating smooth muscle tone might vary in different tissues. Deletion of the cGKI gene and of exon 12 of the IRAG gene suggests that phosphorylation of IRAG, BK Ca channels, and MBS occurs in the same cell. However, it has been shown that modulation of the Ca 2ϩ sensitivity is important in smooth muscles from the small intestine, whereas regulation of the [Ca 2ϩ ] i appears to be more important in vascular smooth muscle. In general, these findings clearly establish that NO signals through cGKI␣ and cGKI␤ in smooth muscle and that these isozymes use different targets to affect smooth muscle tone.
Regulation of Smooth Muscle Proliferation-Migration, proliferation, and dedifferentiation of vascular smooth muscle cells are considered to be essential events in the development of atherosclerosis and vascular restenosis. Many groups have reported that NO, atrio-natriuretic peptide, and membrane-permeable cGMP analogs prevent proliferation, migration, and dedifferentiation of vascular smooth muscle cells (44,45). Recent evidence indicates that cGKI modulates gene expression through the extracellular signal-regulated kinase/mitogen-activated protein kinase pathway either by stimulation or by its inhibition (46,47). Differential phosphorylation of VASP might be critical in these studies. VASP, a member of the Ena/Mena/Vasp family, is an actin-binding protein that is localized to the focal adhesion and cell-to-cell contacts in many cells (48). VASP is phosphorylated by protein kinase C, cAMP kinase, and cGKI at two sites with different preference. Recently, it was reported that expression of VASP and phosphorylation of Ser-157 increased proliferation whereas phosphorylation of Ser-239 decreased proliferation (49). Surprisingly, serum as used in cell culture studies stimulates phosphorylation of Ser-157 of VASP through PKC (50), an effect of serum not noticed so far. In a mouse model of ischemic vessel growth, the presence of cGKI was essential for vessel growth (51). In a mouse model of atherosclerosis it was found (52) that smooth muscle-specific deletion of cGKI retarded the development of atherosclerotic plaques and smooth muscle proliferation in vivo. In this study (52), it was observed that stimulation of cGKI increased smooth muscle cell growth in vivo, whereas NO inhibited growth independent of cGKI. It is therefore likely that in vivo cGKI stimulates smooth muscle growth and that Minireview: cGMP Protein Kinases 2 the existing controversy will wane as soon as better controlled experiments are initiated.
Platelet Aggregation-In most cases, aggregation of platelets is initiated at areas where the endothelial cell layer has been destroyed. Endothelial cells release prostacyclin and NO, which prevent platelet adhesion and aggregation. These factors raise cAMP and cGMP levels in platelets and thereby inhibit clot formation. Platelets have a high concentration of cGKI␤ that is activated in response to NO and has an anti-aggregatory function (53)(54)(55). Under specific conditions, cGKI may also promote platelet activation (56), perhaps by facilitating the release of ADP (57). However, NO or cell-permeable cGMP analogs did not inhibit aggregation of cGKI-deficient platelets, whereas aggregation was prevented by cAMP-elevating agents (53). An intact platelet NO/cGKI signaling pathway is essential to prevent platelet aggregation after ischemia in vivo (53). Platelets contain two well established cGKI substrates, VASP and IRAG. Deletion of the VASP gene in mice did not grossly affect platelet aggregation (58,59), but considerably affected the interaction of platelets with the endothelium in vivo (60). Activation of cGKI inhibited the release of Ca 2ϩ from IP 3 -sensitive stores in wild type and VASP-deficient platelets to similar extents (58). Platelets express IRAG (14). Furthermore, platelets from the IRAG mutant mice (32, 61) have a severe defect in the cGMP-mediated prevention of aggregation, 2 indicating that IRAG is an essential component of this pathway.

Roles of cGKII
Our knowledge about the function of cGKII is at a very preliminary state. Only three areas have been identified where cGKII plays a role: secretion, bone growth, and circadian rhythmicity.
Secretion-Intestinal Cl Ϫ /fluid secretion is increased by substances that stimulate cAMP (cholera toxin) or cGMP (guanylin, STa) levels. cGKII is highly expressed in the apical membrane of the enterocytes of the small intestine. The mucosa of cGKII-deficient mice responded normally to cAMP analogues, whereas STainduced electrogenic anion secretion was blocked. Active cGKII phosphorylated cystic fibrosis transmembrane conductance regulator and increased Cl Ϫ and water secretion. These results established that cGKII is essential for guanylin/STa-dependent secretion in the small intestine (62,63).
NO/cGMP were reported to modulate renin release in the kidney. Deletion of the cGKI gene had no effect on renin release in isolated kidneys or kidney cells, whereas the inhibitory effect of NO on renin release was blunted in cGKII-deficient mice (64). Blood pressure was not affected in cGKIIϪ/Ϫ mice. This is interesting because it was reported that activation of cGKII increased the secretion of aldosterone in rat adrenal cortical cells (65). It is possible that the opposite effect of cGKII on two blood pressureelevating factors cancelled each other in the knock-out animals.
Bone Growth-The particulate GCs, GC-A and GC-B, are expressed abundantly in mouse tibial epiphysis and vertebrae (66). Cultivation of mouse tibias in the presence of 1 M brain-natriuretic peptide, a ligand for GC-A and GC-B, induced a significant increase in total bone length. Transgenic mice overexpressing BNP exhibited skeletal overgrowth which was restricted to those bones that grow by endochondral ossification. cGKI and -II are expressed in the growth zone of bones (62). The deletion of cGKI has no apparent effect on the growth of the skeleton (20). In contrast, cGKII-deficient mice are dwarfs with 16 -30% shorter limbs. cGKII is essential for the CNP/cGMP-mediated endochondral ossification (67) and regulates autonomous bone growth.
Circadian Rhythmicity-The suprachiasmatic nucleus (SCN) harbors the circadian clock pacemaker that runs on a close to 24-h time scale and is reset to the external time period by multiple pathways (68). Both cGKs are expressed in the SCN. Studies with cGKIIϪ/Ϫ mice showed that cGKII influenced the phase shift of the circadian clock at the onset of the wheel running activity (69). The involvement of cGKII in the regulation of the circadian clock was confirmed (70) in the isolated rat SCN, although in that system cGKII was required for night to day progression of the clock (71).

Behavior and cGKs
cGKs and the Mammalian Brain-Although cGKII is expressed in many neurons, deletion of the gene resulted only in mild neurological effects such as a moderate enhanced anxiety-like behavior and a hyposensitivity to acute alcohol intake (18). Although the neuronal expression of cGKI is rather restricted, a number of important phenotypes are associated with tissue-specific deletion of cGKI (for an extensive discussion see Ref. 72). cGKI␣ expressed in sensory neurons of dorsal root ganglia is necessary for the correct guidance of sensory axons during embryonic development (73). Recent evidence suggests that cGKI␣ is also involved in some forms of pain perception (74). cGKI␤ is involved in hippocampal long term potentiation, a paradigm for learning, as shown in cell culture (75) and in hippocampus-specific cGKI knock-out mice (76). The biological significance of this finding is unclear because the cGKI mutants showed no learning defect in several tests (76). The target(s) of cGKI in hippocampal and dorsal root neurons are unknown.
Cerebellar Purkinje neurons express high levels of cGKI␣ and the G-substrate, a peptide specifically phosphorylated by cGKI. The phosphorylated G-protein is an inhibitor of protein phosphatase 1/2A (77,78). It has been hypothesized that inhibition of the protein phosphatases allows phosphorylation of the AMPA receptor by other kinases and its internalization leading to long term depression. Indeed, coincidence of an increase in [Ca 2ϩ ] i and cGKI␣ activity is necessary for induction of long term depression (79) and cerebellar learning (80).
Food-searching Behavior and cGKs-cGKs were also identified in a large number of invertebrates. Drosophila melanogaster has two cGK genes, dg1 and dg2 (81). dg2 encodes a protein kinase that is related to the mammalian cGKI gene (1,82). Analysis of naturally occurring Drosophila variants in food-searching behavior indicated that a high activity of the dg2 kinase is a major determinant of rover versus sitter behavior (83). A similar situation has been identified in honey bees. A cGKI-like kinase activity is upregulated when the young bees change from hive work to foraging (84). The opposite change has been found in Caenorhabditis elegans. In these animals, long distance roaming for food is associated with a decreased cGK activity (85). Interestingly, the cGK gene involved in this behavior is more related to mammalian cGKII than to cGKI (82) pointing to the possibility that increased cGKII activity is related to a sedentary life style in vertebrates. The above studies unequivocally demonstrate that cGKs are involved not only in cardiovascular physiology but also regulate complex central nervous processes and that even subtle changes in cGK activities can lead to naturally occurring behavioral variants.