The original isolation and characterization of cyclic GMP (cGMP) followed closely the discovery of cyclic AMP (cAMP) and the development of the “second messenger” hypothesis proposed by Sutherland and Rall in the late 1950's [6, 109, 110, 130–132]. Guanylyl cyclase synthetic activities as well as phosphodiesterase activities capable of catalyzing cGMP hydrolysis were subsequently identified in virtually all tissue sources examined, and their initial enzymatic and biochemical characterization undertaken [5, 60, 61, 71, 118, 142]. Observations concerning changes in cGMP levels in response to a wide variety of exogenous stimuli followed , and yet a clearly defined role for cGMP in a known intracellular signaling cascade has come only much more recently. Difficulties associated with the biochemical behavior of guanylyl cyclase in broken cell preparations, the specificity of cGMP action, relative in particular to cAMP, and the accurate identification and localization of cGMP target proteins in specific cell types have all presented hurdles to be overcome.
In an accompanying review by Hagiwara et al. the natriuretic peptide receptor family and its role in regulation of blood volume homeostasis is described. The natriuretic peptide receptors, along with the elucidation of the visual signal transduction cascade and recent developments in the nitric oxide signaling field have all fueled interest in the second messenger role of cGMP. While a central role for cGMP in a limited number of cellular signaling pathways is now accepted, its further widespread significance as an intracellular signaling molecule is implicit in the ubiquitous distribution of cGMP and the enzymes regulating its metabolism, the diverse nature of its target molecules, and observations concerning its interaction with other known signaling pathways. In this review, we will present an overview of our biochemical understanding of those molecules central to cGMP metabolism and action. Rather than attempt to discuss the numerous and varied signaling pathways and cell types in which a role for cGMP has been suggested, examples will be selected which serve to illustrate how we may integrate current biochemical information in the design and analysis of novel experimental approaches aimed at further defining the role of cGMP in differentiated cell function and growth.
GUANYLYL CYCLASES AND REGULATION OF CYCLIC GMP SYNTHESIS
Guanylyl cyclase exists in both soluble and membraneassociated forms, with members of each class co-existing within the same tissues or cell types in many instances (for review, see ). The primary structures now deduced for a number of guanylyl cyclase forms reveals a diverse receptor family exhibiting homology not only with adenylyl cyclase but, in the case of the membrane-associated forms of guanylyl cyclase, a number of protein kinase growth factor-linked receptors including platelet-derived growth factor, epidermal growth factor and the recently described Janus kinase (JAK) family (Fig. 1). Information concerning the presence of individual guanylyl cyclase isoforms within specific cell-types of interest and our ability to alter their synthetic activities under experimental conditions may be considered among our most direct points of access in defining cGMP-mediated intracellular events.
I. MEMBRANE GUANYLYL CYCLASE
Membrane-associated forms of guanylyl cyclase share a general structure shown in Figure 1. Comprised of an extracellular peptide-binding site, a single putative transmembrane domain, and an intracellular component containing both the cyclase catalytic region as well as an additional domain exhibiting homology with protein kinases (kinase homology domain), members of this guanylyl cyclase sub-family play a central role in a number of physiological responses, including visual signal transduction, regulation of blood volume and electrolyte balance, and sperm motility. To date, cDNAs representing four mammalian membrane guanylyl cyclases have been cloned, based on sequence homology with previously cloned cDNAs for highly homologous sea urchin forms [125, 136]. Two of the mammalian forms, GC-A and GC-B, have been demonstrated to function as cell surface receptors for the natriuretic peptides, ANP, BNP, and CNP [27, 28, 119]. A third form, GC-C, is characterized on the basis of its ability to be specifically activated by the low molecular weight peptides (STa's) produced by pathogenic E. coli, the agents responsible for one form of secretory diarrhea . More recently, an endogenous mammalian peptide, termed guanylin, has been identified as a probable physiological ligand for GC-C [35, 121]. A fourth member of the membrane guanylyl cyclase family has been cloned from human retina, although an extracellular ligand for this presumed receptor has yet to be identified .
Our current understanding of the array of peptide ligands and their specificity towards the individual members of the membrane guanylyl cyclase receptor family has been recently reviewed elsewhere [41, 49]. As new members of this receptor family are identified using molecular cloning techniques, the identification and characterization of their specific peptide ligands will represent a fundamental first step in understanding their differential regulation. Recent studies examining the ligand binding and catalytic properties of members of this receptor family suggest, however, potentially complex patterns of intracellular regulation may also exist which incorporate features specific to individual membrane guanylyl cyclase isoforms. An understanding of these additional regulatory features will enhance our ability to further address the role of this receptor family in cell function and in our consideration of their potential as targets for pharmacological intervention.
1. Ligand-induced desensitization of membrane guanylyl cyclase
A distinguishing feature of the peptide binding and activation properties common to all membrane guanylyl cyclases studied to date is a pattern of homologous desensitization in response to ligand occupancy. The molecular basis of desensitization of a membrane guanylyl cyclase was first examined by several groups using sperm cells from the sea urchin Arbacia punctulata [9, 10, 133, 143]. The guanylyl cyclase present in sperm membranes exists as a phosphoprotein, and may be activated by the appropriate ligand in membrane preparations in which care has been taken to retain the receptor in the phosphorylated state [143, 144]. Activation by peptide is transient, however, and the subsequent desensitization correlated with dephosphorylation of the receptor protein. This is in contrast to observations in other membrane receptor systems in which desensitization is coupled to an increase in phosphorylation [31,127]. Using purified guanylyl cyclase, the phosphorylation state was shown to influence not only the absolute activity of the enzyme but also the degree of cooperativity between catalytic and GTP-binding sites .
This model of desensitization coupled to dephosphorylation has been extended to the mammalian cyclases by Potter and Garbers . Using a mammalian 293 cell line in which the rat atrial natriuertic peptide receptor, GC-A, was stably expressed, a pattern of ligand-mediated desensitization linked to dephosphorlyation was demonstrated. Incubation of membranes with protein phosphatase 2A mimicked the mobility shift and desensitization produced by extended ligand occupancy, an effect which was blocked by okadaic acid, an inhibitor of this phosphatase. These same authors have recently demonstrated that protein kinase C-dependent desensitization of GC-A may also occur via dephosphorylation, although interestingly, peptide mapping experiments reveal that the specific phosphorylation sites may differ between protein kinase C-dependent and -independent pathways . This observation suggests a complex regulatory mechanism for GC-A activity in response to homologous and heterologous signals, in which the bound and unbound conformations of the receptor provide alternative substrates for phosphorylation.
2. Adenine nucleotides and the role of the kinase homology domain
The kinase homology domain present in all known membrane guanylyl cyclase receptors has thus far not been demonstrated to possess any intrinsic kinase activity. Rather, a key regulatory role for this domain is suggested in the observed enhancement of ligand-induced stimulation of catalytic activity in response to adenine nucleotides and in phosphorylation-dependent mechanisms of desensitization (for a recent review, see ).
The kinase homology domain of membrane guanylyl cyclase shares a number of conserved amino acids seen within the catalytic domains of active protein kinase forms. A notable exception is the replacement of an invariant Asp present in active kinases with various amino acids. Interestingly, members of the JAK protein kinase family also exhibit a substitution in this position within the catalytically-inactive domain 2 [148, 149]. In the case of GC-A, ATP markedly enhances, but is not absolutely required for hormone stimulation. This effect is mimicked by non-hydrolyzable ATP analogs, indicating that protein kinase activity, whether intrinsic or extrinsic, is not necessary for this effect. Although ATP binding at a site within the kinase homology domain has not been demonstrated directly, this region has been implicated in this effect in various studies [49, 138]. The regulation of membrane guanylyl cyclase forms by adenine nucleotides may thus correspond functionally to the regulation of adenylyl cyclase by guanine nucleotides and G proteins.
The role of the kinase domain in mediating phosphorylation-dependent effects on enzyme activity is unclear. Deletion of the kinase homology domain of GC-A eliminates peptide signaling, and results in a dephosphorylated form of the protein [27, 81]. Studies currently in progress to identify phosphorylation sites in a purified guanylyl cyclase from the sea urchin H. pulcherrimus may provide further details concerning the location of the residues mediating this effect (N. Suzuki, personal communication).
The role of adenine nucleotides in the regulation of another membrane guanylyl cyclase form, the heat-stable enterotoxin receptor, GC-C, has been extensively examined. Like GC-A, GC-C exhibits ligand-induced desensitization in response to exposure to, in this case, the E. coli-derived peptide, STa. Adenine nucleotides also enhance the hormone responsiveness of GC-C, however the mechanisms through which these phenomena occur appear to be fundamentally different compared with GC-A [137–139].
In contrast to GC-A, there is little evidence in support of a role for dephosphorylation in the ligand-mediated desensitization of GC-C. Immunopurified GC-C from 293 cells which had been preincubated in the presence of [32P]-orthophosphate showed no incorporation of radioactive phosphate, and yet exhibited desensitization in the absence of contaminating phosphatases .
Results derived from radiation inactivation experiments performed using both intestinal brush border membrane vesicles and a stably-transfected mammalian 293 cell line expressing recombinant rat GC-C indicated a target for STa-stimulated GC-C which corresponds in size to a homodimeric receptor . Using measurements based on hydrodynamic parameters, however, these same authors conclude that GC-C is present as a trimer under native, unstimulated conditions. A model in which STa binding promotes formation of the active dimer within a larger receptor complex is proposed. Adenine nucleotides may bring about the observed increase in hormone responsiveness by stabilizing the active, homodimeric cyclase form within this complex [138, 139].
These authors suggest that, although the ATP effect may be exerted through an interaction at the protein kinase-like domain in both GC-A and GC-C, the mechanism of the effect is fundamentally different. The relatively low degree of identity noted between the protein kinase-like domains of GC-A and GC-C and the inability of this domain from GC-C to functionally replace its counterpart in a chimeric GC-A receptor may also reflect this difference [81, 120].
Like GC-C, GC-A likely exists as an oligomeric complex in the absence of ligand, a shared feature which distinguishes both these membrane guanylyl cyclase forms from a number of other peptide hormone receptors [29, 80, 90, 139]. Differences between GC-A and GC-C are again noted, however, in terms of these properties. GC-A has been suggested to exist as a tetramer in the absence of ligand , in further contrast to the trimeric form proposed for GC-C discussed above. The formation of heteromeric membrane guanylyl cyclase forms has also been proposed, although the physiological significance of such complexes is unclear .
II. CYTOPLASMIC GUANYLYL CYCLASE
Although soluble guanylyl cyclase activity had been previously observed in broken cell preparations from various tissues, it was not until the late 1970's that agents capable of specifically activating this enzyme were identified. Sodium azide, sodium nitroprusside, and nitric oxide (NO) were shown to activate soluble guanylyl cyclase activity, resulting in increases of several hundred-fold in the levels of cGMP [14, 99]. This effect was presumed to occur through their interaction with the heme prosthetic group originally shown to be present in a cytosolic guanylyl cyclase purified from bovine lung . Along with cyclooxygenase and glyceral-dehyde-3-phosphate, cytoplasmic guanylyl cyclase is a likely candidate receptor for mediating many of the physiological actions of NO [98, 115].
A number of lines of evidence suggest cGMP may mediate some or all of the actions of NO in known target tissues. NO or its synthetic precursors activate cytosolic guanylyl cyclase, with increases in cGMP correlating with physiological responses in most studies [7, 70]. In addition, the NO effect on smooth muscle relaxation has been mimicked using membrane permeable cGMP analogs .
Despite the evidence linking NO with elevations in cGMP, few studies have directly addressed the participation of soluble guanylyl cyclase in a nitric oxide induced physiological response. Rather, inhibitors of nitric oxide synthase or agents such as methylene blue or hemoglobin, which inactivate NO itself have been employed [98, 115]. Cytosolic forms of guanylyl cyclase exist as heterodimers composed of two structurally homologous α and β subunits . To date, two α forms (α1 and α2) and two β forms (β1 and β2) have been identified by cDNA cloning. Coexpression of both subunits appears necessary for formation of NO-sensitive soluble guanylyl cyclase activity [19, 62]. A recent study by Yuen et al.  took advantage of these biochemical observations to directly examine the role of soluble guanylyl cyclase in sodium nitroprusside induced elevations in cGMP using dominant-negative subunits. When transfected into a rat insulinoma cell line, α1 subunits containing point mutations within the catalytic domain blocked sodium nitroprus-side-stimulated guanylyl cyclase activity. Mutant α subunits were, nevertheless able to dimerize with wild-type β subunits, as demonstrated by immunoblotting using subunit-specific antisera. Interestingly, the cell type in which this study was performed also contains an endogenous membrane guanylyl cyclase, GC-C. Transfection of the dominant-negative soluble α1 subunit failed to impair STa-stimulated increases in cGMP content.
The approach used in this study provides the basis for directly examining the role of soluble guanylyl cyclase in physiologically-relevant signaling pathways. Studies aimed at developing dominant-negative forms of membrane guanylyl cyclases are currently in progress and should prove similarly useful in examining signaling pathways involving these receptors .
INTRACELLULAR TARGETS FOR CYCLIC GMP
In contrast to cAMP, for which the vast majority of its physiological actions occur through activation of cAMP-dependent protein kinase, cGMP interacts with three classes of intracellular target proteins: cGMP-dependent protein kinases, cGMP-regulated cyclic nucleotide phosphodiesterases, and cyclic nucleotide-gated ion channels (Fig. 2). Within a given cell, any or all of these targets may potentially exist and participate in signaling events. Overlap between cAMP- and cGMP-dependent pathways can occur at the level of both their receptor proteins as well as in the substrates for phosphorylation by their respective kinases, and interactions with other second messenger systems, adding further to the challenge of defining physiologically relevant pathways. Nevertheless, our appreciation of the biochemical features of cGMP target proteins has grown substantially in recent years, permitting the development of more specific experimental approaches.
I. CYCLIC GMP-DEPENDENT PROTEIN KINASE
Cyclic GMP-dependent protein kinases (cGMP kinase) are serine/threonine kinases, and are generally divided into two classes, termed Type I and II. Type I forms exist as homodimers composed of 78 kDa subunits. Two highly homologous Type I isoforms have been identified (α and β), which differ only within their N-terminal regions, and likely arise as alternatively-spliced mRNA products from a single gene . Type I isoforms have been identified in a number of tissues in which cGMP-mediated regulatory pathways are known to occur. Type Iα is found predominantly in smooth muscle cells from a number of tissue sources, while Type Iβ is found in highest concentrations in cerebellar Purkinje cells, smooth muscle cells, and platelets . Type I cGMP kinase has been most often linked to pathways controlling intracellular Ca2+ concentrations.
In contrast to the Type I isoforms, Type II cGMP kinase appears to exist as a monomeric protein with an apparent molecular weight of 86 kDa, and has been localized predominantly within the intestinal epithelial brush border [36, 72]. While Type I isoforms are generally found in soluble fractions prepared from cell extracts, Type II cGMP kinase is associated with the intestinal brush border membrane, where it likely functions in the control of intestinal Cl− absorption and secretion .
Cyclic GMP kinases are regulated by cGMP in vitro with Ka values in the range of 0.05–0.5 μM, depending upon the assay conditions. Cyclic AMP is also capable of activating cGMP kinases, at Ka values of roughly 5–50 μM [34, 40]. Site-directed mutagenesis studies performed using cGMP and cAMP kinase have confirmed the importance of an invariant threonine residue present within the cyclic nucleotide binding site of cGMP selective target proteins [4, 123, 145]. Conversion of the alanine found in the corresponding position in the cAMP kinase R subunit slow site to a threonine residue was sufficient to reduce the dissociation of [3H]-cGMP to a rate even lower than that seen for the slow site of cGMP kinase itself . The threonine hydroxyl group at this location is predicted to hydrogen bond with the guanine 2-amino group of cGMP .
Like many other kinases, cGMP kinases are autophos-phorylated, in this case at sites within the N-terminal region . Autophosphorylation would appear to increase the susceptibility of cGMP kinases to activation by cAMP, by increasing both the rate of cGMP dissociation as well as the affinity for cAMP at the allosteric site [48, 67, 83]. As discussed below, this change in the affinity of cGMP kinase for cAMP appears to be of physiological significance, particularly as concerns the Type Iβ isoform, which has a somewhat lower affinity for cyclic GMP compared to the α form [88, 151].
1. Kinase-selective cyclic nucleotide analogs
The apparent lack of specificity seen for cyclic nucleotide-mediated activation of cGMP- and cAMP kinases presents a problem in attempting to design experimental approaches capable of distinguishing between these alternate signaling pathways. Results obtained using methods in which potentially non-physiologic concentrations of either native cyclic nucleotide are achieved will clearly be suspect. The development of potent highly selective cyclic nucleotide analogs suitable for use in intact cell systems provides a pharmacological approach that circumvents this potential hazard while also providing selectivity in terms of alternative pathways of action for cGMP itself.
An ideal cyclic nucleotide analog for use in experiments with intact cells will exhibit a high degree of selectivity for cGMP kinase versus cAMP kinase, exhibit a low degree of cross-reactivity towards other cGMP target proteins such as phosphodiesterases and ion channels, will be of sufficient lipophilicity so as to readily cross cell membranes, and be resistant to degradation, particularly phosphodiesterase-mediated hydrolysis, such that working intracellular concentrations will be maintained for prolonged periods, while not competing for the hydrolysis of endogenous cyclic nucleotides. In Table 1 , a number of cyclic nucleotide analogs are compared in terms of these properties using results obtained from in vitro experiments with purified target enzymes. These analogs represent only a small subset of compounds among many which were previously designed for use in defining the substrate binding properties of cAMP- and cGMP-dependent protein kinases [34, 104].
Properties of cyclic nucleotides and derivatives
The 8-bromo- and N6,O2-dibutyryl- (not listed) derivatives of cAMP and cGMP have historically been among the most widely used analogs for studies examining cyclic nucleotide-mediated signaling pathways in intact cells, and yet these compounds are less than ideal. As can be seen in Table 1, 8-bromo-cAMP, while relatively selective for cAMP kinase, is readily hydrolyzed by various PDE isozymes. 8-bromo-cGMP, is somewhat more resistant to hydrolysis, however both compounds posses relatively low degrees of lipophilicity, making them potentially less potent in intact cell experiments. N6,O2-dibutyryl derivatives of cAMP and cGMP are not among the more potent activators of their respective kinases, and require metabolism and release of free butyrate to induce activation, a metabolic event which will tend to vary from cell to cell.
Newer, commercially-available derivatives have now been developed which possess properties closer to the ideal. Among activators of cAMP kinase, Sp-5,6-DCl-cBiMPS (Sp-5,6-dichloro-1-β-D-ribofuranosyl-benzimidazole-3′, 5′-mono-phosphate) is both potent and selective, is resistant to hydrolysis by various PDE isozymes, and is highly lipophilic. Among selective activators of cGMP kinase, 8-pCPT-cGMP (8-p-chlorophenylthio-cGMP) displays the best combination of properties. These 2 compounds have been used in recent studies with intact platelets to examine the relative contributions of cAMP- and cGMP-dependent pathways in the inhibition of platelet activation [42, 52, 103].
Cyclic nucleotide derivatives with inhibitory actions on cAMP and cGMP kinases have also been developed based on stereoisomers in which the equatorial exocyclic oxygen of the cyclic phosphate is replaced with a sulfur atom. These phosphorothioate derivatives are highly resistant to phosphodiesterase-mediated hydrolysis, although (Rp)-cGMPS is less than ideal both in terms of kinase specificity and in its degree of lipophilicity. Addition of the p-chlorophenylthio-moiety, however, leads to an improvement in both of these properties, and (Rp)-8-pCPT-cGMPS has been shown to antagonize the activation of cGMP kinase by 8-pCPT-cGMP in a recent study in intact platelets .
Hidaka and colleagues  have developed a number of selective chemical inhibitors of protein kinases, termed the H series, however no data obtained from studies using these compounds in intact cells is currently available. In the case of cAMP kinase, use of the highly potent heat-stable inhibitor protein, PKI, represents another possible experimental means of inhibiting cAMP-mediated signaling. Unfortunately, a corresponding inhibitory peptide for cGMP kinase has not been identified, and at any rate, such agents require direct access to cAMP kinase and are thus not well suited for use in most intact cell experiments. An alternative experimental approach designed to isolate a role for cGMP kinase in mediating the inhibitory effect of cGMP on intracellular calcium (ICa2+) in rat cardiac myocytes involved the use of a partially-proteolyzed, constitutively active cGMP kinase . Partial trypsinization of purified cGMP kinase removes the first 77 amino acids of the amino-terminal “dimerization domain”, a region of the protein which mediates formation of the holoenzyme, contains the autophos-phorylation sites, determines cooperativity between the cGMP binding sites, as well as containing an autoinhibitory domain. Removal of this region, which is characterized structurally by a leucine zipper motif similar to those believed to mediate dimerization in other proteins, yields a truncated kinase monomer which does not require cGMP for activity . When perfused into rat ventricular myocytes, low concentrations (10–100 nM) of this fragment reduced cAMP-stimulated ICa2+ to a degree comparable to that which could be achieved using 8-bromo-cGMP or cGMP itself .
2. Physiological significance of kinase cross-activation
Having discussed strategies aimed at isolating cGMP and cAMP signaling pathways, it should be noted that there is evidence to suggest that activation of cGMP kinase by cAMP may, in fact, be of physiological significance. Endogenous intracellular levels of cAMP are typically estimated to be roughly 10-fold greater than those of cGMP, and as such might be sufficiently high to favor interaction with cGMP kinases. In cultured rat aortic smooth muscle cells, cAMP has been suggested to reduce intracellular Ca2+ levels at least in part via activation of cGMP kinase . The inverse situation, that is cross-activation of cAMP kinase by cGMP, may also occur in vivo. In T84 cells, a human colon carcinoma cell line which contains high amounts of the heat-stable enterotoxin receptor form of guanylyl cyclase, GC-C, but little or no Type I cGMP kinase, activation by STa produces substantial increases in cGMP and Cl− secretion, and an increase in the cAMP kinase ratio. Cyclic GMP analogs with high selectivity towards cGMP kinase failed to mimic this effect . The contribution of a cGMP-inhibited phosphodiesterase, known to exist in cells of the distal colon and participate in elevating cAMP levels in response to cGMP in that tissue, was not, however examined in these cells . The observed failure of cGMP kinase selective analogs to increase Cl− secretion is perplexing in light of the observed presence of the membrane-associated Type II cGMP kinase in these cells .
3. Phosphorylation substrate selectivity of cGMP kinase
The majority of known protein substrates for cGMP kinase are also subject to phosphorylation by other kinases, notably cAMP kinase. Relatively few proteins have been identified which appear to be exclusive cGMP kinase substrates. Colbran et al.  have shown that a bovine lung cGMP-binding cGMP-specific phosphodiesterase is a relatively selective substrate for cGMP kinase compared to cAMP kinase, and have proposed the role of a phenylalanine residue located in proximity to the phosphorylation site as exerting a negative effect on cAMP kinase action. Specific sites within histone H1 and H2B, as well as a 23 kDa protein isolated from cerebellum appear to be somewhat selective substrates for cGMP kinase-mediated phosphorylation [2, 54, 64, 155]. Like most other studies comparing the patterns of phosphorylation by these and other kinases, however, this work was performed in vitro using purified proteins or peptides, and thus may not accurately reflect events in intact cells. Indeed, little comparative data concerning the specificity of cGMP- versus cAMP-dependent phosphorylation is available from studies performed in intact cells or tissues.
A recent report by Butt  examined the pattern of phosphorylation of a focal adhesion vasodilator-stimulated protein (VASP) in vitro and in intact human platelets. VASP is a major substrate for both cGMP- and cAMP-dependent protein kinases in platelets as well as other circulating cells of the cardiovascular system. VASP phosphorylation in intact platelets occurs in response to agents which elevate cAMP (e.g. Prostaglandin E1) or cGMP (e.g. nitro-vasodilators), or by administration of selective membrane-permeable cyclic nucleotide analogs including 8-pCPTcG and Sp-5,6-DCl-cBiMPS. VASP phosphorylation correlates well with the inhibition of platelet aggregation produced by these agents [20, 51, 57, 58, 102, 116, 141]. In in vitro assays with purified VASP, three phosphorylation sites were identified as overlapping targets for cGMP- and cAMP-dependent phosphorylation. Among these sites, two serine residues were rapidly phosphorylated by either kinase, while the third site, a threonine residue, was phosphorylated at a much slower rate. Differences were noted in vitro in the rates of [32P]-incorporation at the two serine residues by cGMP kinase versus cAMP kinase, using purified VASP or synthetic peptide substrates. In fact, the site more readily phosphorylated by cGMP kinase appears to represent one of the best and most selective substrates for this kinase identified to date. As predicted by previous studies examining the substrate specificities of cAMP and cGMP kinase using peptide substrates, this site is more basic than that preferred by cAMP kinase . Interestingly, when the kinetics of cGMP kinase- and cAMP kinase-mediated phosphorylation at these sites were compared in immunopreciptated VASP from intact platelets treated with cAMP- or cGMP-elevating agents, the rates of [32P]-incorporation by cGMP kinase at the two serine sites appeared similar, highlighting the potential difficulties associated with predicting patterns of in vivo phosphorylation based on data obtained in vitro.
4. Differential localization of cGMP target proteins
A number of explanations may account for the observed differences in phosphorylation kinetics seen when comparing in vitro VASP phosphorylation with patterns seen in intact platelets. Differences in the intracellular concentrations of cyclic nucleotides, their respective kinases and various substrates could all be envisioned to produce alterations in the qualitative and quantitative patterns of phosphorylation. In an immunocytochemical study using cyclic nucleotide-specific antisera, the distribution of cGMP and cAMP immediately after agonist stimulation in intact cells was examined . Patterns of intracellular distribution varied with the stimulatory agonist used. Sodium nitroprusside, for instance, produced a diffuse widespread cytoplasmic staining pattern, in contrast to atrial natriuretic peptide, which produced intense staining in the plasma membrane region at early time points following administration. In the latter case, target proteins associated with the cell membrane would be exposed to higher cGMP concentrations within a more rapid time frame. This observation may provide an explanation for the observed failure of sodium nitroprusside to mimic the effects of atrial natriuretic peptide in certain cell types.
In terms of cGMP target proteins and phosphorylation substrates, there is evidence to suggest that localization may occur via direct interactions with cytoskeletal proteins in some cell types. Wyatt et al.  have proposed a model in which a Ca2+ signal is required to bring about the transient co-localization of cGMP kinase and vimentin in the neutrophil, leading to chemotaxis and degranulation through an unknown mechanism. In neutrophils, cGMP appears to participate in chemotaxis and degranulation through activation of cGMP kinase. Under physiological circumstances, elevations in cGMP likely occur through NO formation secondary to increases in intracellular Ca2+ and activation of NO synthase . A Ca2+ signal also appears necessary to permit the cGMP kinase-mediated phosphorylation of the intermediate filament cytoskeletal protein vimentin. VASP, the phosphorylation target described above in platelets, appears to be associated with the cytoskeleton and has also been shown to bind actin . It is tempting to speculate concerning a possibly analagous situation involving membrane localization of cGMP kinase in the platelet.
The concept of differential localization of cGMP target proteins, phosphorylation substrates, as well as of guanylyl cyclases and intracellular cGMP concentrations is attractive, as it allows for a high degree of plasticity in the mechanisms through which specificity of a cGMP signal can be achieved.
II. CYCLIC GMP-REGULATED PHOSPHODIESTERASES
Cyclic nucleotide phosphodiesterases (PDEs) provide the major means through which intracellular cGMP and cAMP signals are attenuated, by conversion to their respective 5′-monophosphates. A central regulatory role for PDEs in intracellular signal transduction is implicit not only in the nature of their catalytic reaction, however, but also in the observed regulation of various PDE isozymes by known intracellular effectors including Ca2+/calmodulin, G proteins, phosphorylation, and cGMP itself. In addition, differences in substrate specificity and affinity among PDE forms are now well established. The complement of PDE isozymes within a given cell type will thus partially determine not only the magnitude and duration of an intracellular cGMP signal, but also provide a junction linking cGMP with cAMP and other intracellular signaling pathways. During the last several years, direct protein sequencing and molecular cloning methods have led to a dramatic increase in phosphodiesterase sequence information, with more than 30 different isozymes, representing at least 7 gene families, currently recognized . Most of the PDE isozymes comprising the various families shown in Table 2 share a common conserved catalytic domain of approximately 270 amino acids located in the carboxyl terminus of the enzyme [11, 25]. Isozymes within a family display a high degree of sequence identity (60–70%) while also exhibiting common patterns of regulation and substrate specificity.
Cyclic nucleotide phosphodiesterase (PDE) families
PDE forms known to be directly regulated by cGMP include the PDE2 and PDE3 families. The PDE2 or cGMP-stimulated phosphodiesterases catalyze the hydrolysis of both cAMP and cGMP with approximately equal affinities (Km's of 10–30 μM) and similar maximal rates . In addition to serving as a catalytic substrate, cGMP is also a positive allosteric regulator of PDE2 forms through interaction at a high affinity regulatory site [128, 129]. To date, two PDE2 variants have been identified and appear to arise as alternatively spliced mRNA forms from a single gene . The two forms differ in terms of their association with the particulate cell fraction, with the membrane-associated form predominating in cells of the central nervous system. In other tissues, such as the heart, both forms are expressed at approximately equal ratios. In general, PDE2 forms are widely distributed, with highest concentrations occurring within the glomerulosa cells of the adrenal cortex . As described further below, a role for cGMP-mediated increases in PDE2 activity leading to decreases in intracellular cAMP concentrations has been demonstrated in several physiological processes including responses to atrial natriuretic peptide in PC12 and primary bovine adrenal glomerulosa cells, and inhibition of cAMP-stimulated increases in ICa2+ in cardiomyocytes [63, 93, 147].
The PDE3 family represents isozymes which specifically hydrolyze cAMP, with Km's in the sub-micromolar range . PDE3 activity is inhibited by cGMP, although it remains unclear as to whether this effect occurs through interaction of cGMP at a unique allosteric site or by direct competition for cAMP binding within the catalytic domain. In contrast to the antagonistic effect of cGMP on cAMP levels when acting through the PDE2 isozymes described above, cGMP's actions in cells in which PDE3 isozymes predominate would be predicted to result in elevations in cAMP. PDE3 isozymes are known to be activated by phosphorylation and mediate, at least in part, many of the metabolic effects of insulin on carbohydrate and lipid metabolism . The insulin effect is likely indirect, since phosphorylation of PDE3 occurs at serine/threonine residues, rather than at tyrosine as would occur by a direct action of the insulin receptor tyrosine kinase. As described further below, PDE3 forms appear also to play a central role in the control of cardiac contractility.
Among the remaining PDE gene families listed are forms which specifically hydrolyze cGMP including the multi-subunit, membrane-associated photoreceptor PDE (PDE6) and cGMP-specific cGMP-binding (PDE5) families. The latter represent something of an enigma, for while they bind cGMP at high affinity non-catalytic sites which exhibit properties similar to those of the PDE2 forms, cGMP binding has not been demonstrated to alter the kinetic properties of the enzyme's hydrolytic activity . Finally, the Ca2+/calmodulin-dependent PDE family (PDE1), which hydrolyze both cGMP and cAMP, provide an obvious link between Ca2+ and cyclic nucleotide signaling pathways. PDE1 isozymes have been shown to be regulated by phosphorylation, which leads to a decreased affinity for Ca2+/calmodulin .
1. cGMP-stimulated reductions in cAMP via PDE2 activation
Amino acid sequence comparisons suggest that the putative allosteric cGMP-binding domain present in PDE2 (cGMP-stimulated) isozymes is structurally distinct from the more closely related domains identified in cGMP protein kinase and cGMP-gated ion channels [74, 126]. This fact provides an explanation for the observed failure of some analogs to mimic the effects of cGMP in intact cell experiments. Despite the substantial increases in cGMP seen in adrenal glomerulosa cells in response to ANP, the failure of several groups to mimic ANP's inhibitory effect on aldoster-one biosynthesis using 8-bromo-cGMP, a known activator of cGMP kinase, prompted the suggestion that ANP's action in these cells was not mediated by cGMP. The observation that glomerulosa cells contain high concentrations of a cGMP-stimulated (PDE2) isozyme suggested an alternative mechanism, however, in which the cGMP signal would attenuate aldosterone biosynthesis through reductions in cAMP resulting from PDE2 activation. 8-bromo-cGMP is, in fact, a poor activator of PDE2, and as a result would not be expected to mimic the cGMP signal in such a pathway. Evidence in support of a PDE2-mediated mechanism of cGMP action in glomerulosa cells was obtained using cAMP-derivatives which displayed differential rates of hydrolysis in in vitro assays using immunopurified PDE2 obtained from these cells. A strong correlation between the degree to which the stimulatory effect of a given analog on aldosterone biosynthesis could be inhibited by ANP, and its relative susceptibility to hydrolysis by PDE2 was shown .
The differences in specificity seen for the cGMP binding sites in PDE2 and cGMP kinase should, it follows, allow for the design of analogs which selectively activate PDE2. One commercially-available analog, caged cGMP (1-[2 nitrophenylethyl]-cGMP), was shown to produce partial activation of immunopurified PDE2, and was able to mimic the inhibitory effect of ANP on aldosterone biosynthesis in isolated primary adrenal cells . Interestingly, this compound is marketed as an “inactive” cGMP analog which, when exposed to ultraviolet light, yields cGMP. Such a description clearly presumes a role for cGMP kinase in mediating cGMP action.
2. Multiple cGMP-regulated PDE forms in a single cell type
ANP-mediated increases in cGMP linked to reductions in cAMP have also been demonstrated in PC12 cells as well as cardiac myocytes [89, 147]. The latter provides an interesting example of how multiple cGMP-regulated PDE isozymes may act in a coordinate fashion and in concert with other cGMP target proteins within a single tissue or cell type. The opposing effects of cAMP and cGMP on the inotropic response of the heart have been well documented, and are believed to converge at the level of the L-type calcium current, which correlates well with cardiac contractility . In cardiac myocytes, elevations in cAMP in response to a number of hormones including β-adrenergic agonists, histamine, and glucagon, produce an inotropic response through phosphorylation of the L-type calcium channel or a closely associated regulatory protein, leading to an increased mean probability of channel opening and elevations in ICa2+. Cyclic AMP also elicits a positive chronotropic response, through a novel phosphorylation-independent activation of sino-atrial node pacemaker cells .
In the heart, the negative inotropic effects of cGMP may occur through three possible intracellular targets. Cardiomyocytes from various species have been shown to contain both cGMP-stimulated (PDE2) and cGMP-inhibited (PDE3) phosphodiesterase isozymes, as well as cGMP kinase. Data obtained from patch-clamp experiments with isolated cells obtained from several species indicate a role for each of these three cGMP target proteins in transduction of the cGMP signal, with surprising differences in their relative contributions noted between species [44, 63, 84, 97, 105, 140]. In frog heart, elevations in cGMP produced by agents such as acetylcholine, adenosine, or ANP are believed to antagonize cAMP-stimulated inotropic responses via activation of cGMP-stimulated (PDE2) activity and resulting decreases in cAMP, through a mechanism analagous to that described for the inhibition of aldosterone biosynthesis by ANP in adrenal glomerulosa cells. Evidence for this pathway includes results showing that increases in ICa2+ in response to intracellular perfusion of non-hydrolyzable cAMP analogs were not antagonized by hormone effectors, and that 8-bromo-cGMP, a potent activator of cGMP kinase, had no effect on ICa2+ [44, 63]. A previously known adenosine deaminase inhibitor, EHNA (erythro-9-[2-hydroxy-3-nonyl]-adenine) has recently been identified as a selective inhibitor of PDE2. When perfused into isolated frog ventricular myocytes, EHNA antagonized the effects of cGMP on ICa2+ .
In contrast to the frog, results obtained from similarly performed experiments using guinea pig cardiomyocytes suggest a central role for cGMP kinase in mediating the cGMP signal [84, 97]. In this case, intracellular perfusion with 8-bromo-cGMP did produce reductions in cAMP-stimulated ICa2+. Evidence for the role of cGMP kinase in guinea pig cardiomyocytes was also provided by the previously described study in which a constitutively active, partially proteolyzed cGMP kinase monomer was perfused intracellularly, leading to reversal of the cAMP-elevated ICa2+ .
Finally, evidence for the occurrence of a cGMP-inhibited PDE (PDE3) in cardiac myocytes has come from studies using chemical inhibitors of these isozymes . Perhaps more than is true for any other known cGMP target protein, specific chemical inhibitors, some of which are currently under investigation as potential therapeutic agents for the treatment of congestive heart failure, are available for PDE3 forms . In examining the effect of two such agents, milrinone and indolidan [LY195115], in frog ventricular myocytes, an increase in ICa2+ was noted when cells were first partially stimulated by agents capable of elevating cAMP . This finding might suggest that cGMP may increase, rather than decrease, ICa2+, however intracellular perfusion of cGMP only elicited reductions in ICa2+, probably because the activation of PDE2 overwhelms the effect of inhibition of PDE3. In a recent study using isolated human atrial myocytes, Kirstein and co-workers  examined the effect of the NO-donor SIN-1 (3-morpholino-sydnonimine) on ICa2+. Extracellular application of SIN-1 elicited a potent stimulation of basal ICa2+, but this effect did not occur when ICa2+ was first elevated using milrinone. These authors conclude that NO elicits elevations in ICa2+ in these cells via inhibition of the cGMP-inhibited PDE3.
Although the relationship between the PDE2 and PDE3 forms under normal physiologic conditions is still not clear, differences in their Km values for cAMP hydrolysis suggest that their relative function may depend on the activation state of the cell, and that a cGMP signal will be differentially transduced depending upon the instantaneous cAMP levels in the cardiomyocyte.
III. CYCLIC GMP-GATED ION CHANNELS
Among the newest and most rapidly expanding areas of cGMP signaling research concerns the identification and characterization of cyclic nucleotide-gated ion channels which bind and are activated by cGMP or cAMP in a phosphorylation-independent manner. The original member of this relatively new class of ion channels was identified in rod photoreceptors and found to mediate the direct effect of cGMP on what was eventually shown to represent the lightsensitive conductance in these cells [43, 78, 152]. A distinct but closely related photoreceptor channel has been identified in retinal cones, which play a predominant role in visual perception under conditions of bright light [30, 65]. The rod and cone photoreceptor channel forms have been shown to be derived from distinct genes .
The photoreceptor cGMP-gated channel was purified, characterized in reconstituted systems, and subsequently cloned [33, 75]. The deduced amino acid sequence reveals six transmembrane spanning domains, similar to other Na+ and K+ channels, and a single cGMP binding site per monomer in the region of the carboxy terminus. As previously described, a clear homology exists between this domain and the corresponding sites in cGMP kinase, however the photoreceptor channel exhibits a markedly lower affinity for cGMP. Under physiological conditions, high intracellular cGMP concentrations (∼5–10 μM) within the rod cell maintain the channel in the open state. This relationship allows for the rapid response times required in this signaling pathway, since an immediate dissociation of cGMP and ensuing channel closure will occur following reductions in intracellular cGMP concentrations resulting from lightstimulated, phosphodiesterase-mediated hydrolysis.
The role of this channel in vertebrate photoreceptor signal transduction has been reviewed extensively (for a recent review, see ). In terms of its ion permeability characteristics, the photoreceptor channel is relatively nonselective, and actually exhibits a higher permeability to divalent cations, notably Ca2+, than to monovalent cations. Under physiological conditions, however, the dark current through the open channel is carried predominantly by Na+, which is present in the extracellular medium at a concentration approximately 100 times greater than that of Ca2+. Upon exposure to light and channel closure, Ca2+ entry stops, but efflux via the Na+/Ca2+ exchanger continues, leading to an eventual decline in intracellular Ca2+. This phenomenon provides the basis for a feedback mechanism linking reductions in Ca2+ with stimulation of the membrane form of guanylyl cyclase present in these cells by recently described Ca2+-sensitive retinal guanylyl cyclase activating proteins [39, 56, 106]. At least 2 apparently distinct proteins have been reported, with molecular weights of 20- (p20) and 24-kDa (p24). p 20 has been cloned and exhibits homology with other Ca2+-binding proteins. Both p20 and p24 activate retinal guanylyl cyclase in low Ca2+ and appear to be retinal-specific. Increases in intracellular cGMP occurring through this mechanism antagonize light-stimulated phosphodiesterase-mediated hydrolysis, thus down-regulating the sensitivity of the rod cell to light and providing a mechanism for light adaptation. The identification of Ca2+-sensitive guanylyl cyclase activating proteins adds yet another dimension to the interdependence seen between cGMP and Ca2+ in a number of signaling pathways. In a recent report, Hsu and Molday  have also demonstrated a direct interaction between calmodulin and the photoreceptor channel. In a reconstituted system, calmodulin A increased the Km for cGMP-dependent efflux of Ca2+, consistent with a previous report showing that calmodulin reduces the affinity of the channel for cGMP in ROS vesicle preparations. This mechanism provides a second branch in the Ca2+-dependent photoadaptive process  (Fig. 3).
Following the identification of the photoreceptor channel, a second member of the cyclic nucleotide-gated ion channel family was identified in olfactory epithelium [37, 91, 100]. In this sensory pathway, cAMP rather than cGMP is believed to function as the primary second messenger, since numerous odorant receptors have been shown to be coupled to elevations in intracellular cAMP through G-protein-mediated increases in adenylyl cyclase activity [18, 113]. In a pathway analagous to that described for the photoreceptor cell, cAMP opens this channel through a direct interaction, leading to membrane depolarization. The olfactory cyclic nucleotide-gated channel exhibits similar ion permeability characteristics to those seen for the photoreceptor channel, however in this case an increase in intracellular Ca2+ is believed to be coupled to adaptation following prolonged stimulation . This channel is activated at lower cyclic nucleotide concentrations than those required by the photo-receptor channel and, interestingly, cGMP is still a somewhat more potent activator than the presumed physiological ligand, cAMP . A signaling role for cGMP is suggested by results indicating that some odorant receptors are coupled to increases in inositol trisphosphate (IP3), and that in olfactory cilia both IP3 and cAMP may elicit increases in intracellular Ca2+ through activation of non-specific cation channels. This increase in Ca2+ would be predicted to lead to increases in cGMP by activation of NO synthase [17, 112].
Both the photoreceptor and olfactory cyclic nucleotidegated channels are believed to exist as heteromeric complexes composed of the primary channel polypeptide in combination with additional subunits which impart specific properties to the functional channel [16, 26]. In the case of the photoreceptor form, this subunit confers high affinity inhibition by L-cis-diltiazem to the channel oligomer . The reported interaction of calmodulin with the photoreceptor channel discussed above appears to result from binding to the 240 kDa protein subunit (subunit 2 or β) of the channel complex .
Recently, a role for cyclic nucleotide-gated ion channels in non-sensory tissues has been implied by the identification of a third member of this receptor family, cloned independently by two groups, from bovine kidney and testis [13,146]. The deduced amino acid sequence for this channel reveals approximately 60% amino acid identity with the photoreceptor and olfactory forms, however its electrophysiological properties more closely resemble those of the photoreceptor channel . Northern blotting and PCR analysis reveal the presence of this channel in heart atrial and ventricular tissue, as well as in the retina, where it was specifically localized to the cone photoreceptor using immunocytochemical means [13, 146]. Both the photoreceptor and olfactory channel sub-types have also been identified in non-sensory tissues. The olfactory channel has been shown to occur in aorta , while PCR analysis reveals that the photoreceptor channel sub-type is present in low concentrations in bovine kidney, consistent with previous observations in rat kidney and mouse cortical collecting duct cells [1, 13].
Although specific functional roles for cyclic nucleotidegated channels remain to be defined in other tissues, as in the photoreceptor and olfactory epithelial cell, they provide the basis for a second messenger-mediated pathway controlling Ca2+ influx. The cGMP-gated channel cloned by Weyand et al.  from testis was localized in sperm membranes in functional assays and by immunoblot analysis. Although well documented in invertebrate models, the role of guanylyl cyclase and cGMP in vertebrate sperm function is less clearly established . It is nonetheless tempting to speculate concerning a possible role for cGMP signaling through interaction with a cyclic nucleotide gated channel in a sensory pathway regulating vertebrate sperm chemotaxis in fertilization.
The presence of a cGMP-gated channel in the renal tubular epithelium provides the basis for a model for atrial natriuretic peptide's (ANP) diuretic effect in the kidney. Although ANP increases Na+ and Cl− transport in the renal tubule, these cells do not contain significant quantities of cGMP kinase or the ANP receptor guanylyl cyclase [59, 73]. Circulating ANP would, however, be predicted to increase cGMP production in vascular epithelial tissue which could be released and filtered into the tubular lumen. In this scenario, an ANP-mediated increase in intracellular Ca2+ would lead to enhanced Na+ and Cl− transport, and diuresis .
The signaling functions of cGMP are undoubtedly widespread and closely interwoven with the workings of other intracellular second messenger systems. In comparison to the cAMP signaling system, which is characterized by the built-in diversity afforded by G-protein-receptor coupling, cGMP signaling is distinguished by the variety of its target proteins and in its direct links to other intracellular effector pathways. As illustrated by the examples cited in this review, in many instances multiple cGMP target proteins and guanylyl cyclase forms may exist within a given cell type. In addition, a high degree of overlap is observed between cAMP and cGMP, both in terms of kinase activation and phosphorylation targets, and as competing substrates for phosphodiesterase-mediated hydrolysis and inactivation. While this complexity presents a challenge to those attempting to define a role for cGMP in specific signaling cascades, our enhanced appreciation of the molecules central to cGMP metabolism and action provides the key to the design and accurate interpretation of biochemical approaches capable of further uncovering the varied roles of cGMP in cell growth and function.
The author extends his sincere thanks to Drs. Joseph A. Beavo, Rodolphe Fischmeister, Elke Butt, and Norio Suzuki for their assistance in the preparation of this manuscript.