Nitric Oxide and Related Substances as Neural Messengers

Solomon H. Snyder and Ted M. Dawson

The history of neurotransmission is full of surprises. A reasonable person would assume that the brain could make do with few, perhaps only two, neurotransmitters—one excitatory and one inhibitory. For most of the 20th century this appeared to be the case, because between the 1920s (when acetylcholine was appreciated) and the late 1960s only a handful of molecules were accepted as neurotransmitters, specifically biogenic amines and amino acids. Research on opiate receptors and enkephalins spurred interest into peptides, and, within a few years, up to 50 or more neuropeptides had been characterized. Though differing markedly in many properties, amines, amino acids, and peptides follow closely the conventional neurotransmitter dogma. They are stored in synaptic vesicles which release their contents by exocytosis involving fusion with the plasma membrane and expulsion. They diffuse to closely adjacent cells, where they interact reversibly with membrane protein receptors which influence cellular events through the mediation of intracellular second messenger molecules or by influencing ion permeation. Inactivation either by enzymes or by reuptake pumps is also a crucial factor in regulating the duration of transmitter action.

The discovery of nitric oxide (NO) as a neurotransmitter has radically altered our thinking about synaptic transmission. Being a labile, free radical gas (though in most biological situations NO is in solution), NO is not stored in synaptic vesicles. Instead it is synthesized as needed by NO synthase (NOS) from its precursor L-arginine. Rather than exocytosis, NO simply diffuses from nerve terminals. It does not react with receptors but diffuses into adjacent cells. In place of reversible interactions with targets, NO forms covalent linkages to a multiplicity of targets which may be enzymes, such as guanylyl cyclase (GC) or other protein or nonprotein targets. Inactivation of NO presumably involves diffusion away from targets as well as covalent linkages to an assortment of small or large molecules such as superoxide and diverse proteins.

While the neurosciences have been markedly influenced by new insights into NO in the nervous system, NO was first appreciated in mammalian systems associated with inflammatory responses and blood vessel reactivity. In the early 1980s, studies of nitrosamines as carcinogens led to the demonstration that endogenous nitrates can be produced, because germ-free rats excrete large amounts of nitrates as do humans, whose excretion rises markedly during infections (54). Clever detective work led to the finding that the nitrates in the urine arise from macrophages through oxidation of the guanidine nitrogen of L-arginine, giving rise to L-citrulline and a reactive substance subsequently shown to be NO. The ability of macrophages to kill tumor cells and fungi depended upon external arginine, whose effects were blocked by arginine derivatives which also blocked the formation of nitrite, leading to identification of NO as the active substance (48).

A role of NO in blood vessels derives from work in the 1970s implicating NO as the active metabolite of nitroglycerin and other organic nitrates in dilating blood vessels by stimulating cGMP formation through activation of GC (1). Meanwhile, Furchgott and Zawadzki (23) had shown that blood vessel relaxation in response to acetylcholine and other substances requires the endothelial lining which releases a labile substance that diffuses to the adjacent smooth muscle. The active agent was identified as NO (35, 52).

A role in the brain for NO first came from observations that brain cells in culture stimulated by excitatory amino acids release a substance with the properties of NO (25, 26). A definitive involvement of NO was demonstrated by the ability of NOS inhibitors, such as nitroarginine and methyl arginine, to block the pronounced stimulation of cGMP in brain slices that is elicited by the excitatory transmitter glutamate acting at N-methyL-D-aspartate (NMDA) subtype receptors (4, 5, 15).

Because NO cannot be stored by conventional means nor inactivated after synaptic release, its biosynthesis constitutes the only means for regulating NO levels. Not surprisingly, NOS is one of the most regulated enzymes in biology. NOS oxidizes the guanidine group of L-arginine in a process that consumes five electrons and results in the formation of NO with stoichiometric formation of L-citrulline (Fig. 1). Initial efforts to purify the enzyme were unsuccessful because of a rapid loss of enzyme activity upon purification. The discovery that calmodulin is required for NOS activity in the brain permitted a simple purification of brain NOS to homogeneity (3). Based on this scheme, other groups purified brain NOS; and macrophage and endothelial NOS proteins were purified as well (4, 5, 15, 49). Molecular cloning of the cDNA for brain, endothelial, macrophage, and nonmacrophage inducible forms of NOS has considerably clarified NOS function (5, 15, 45, 49) (Fig. 2). The structure of NOS as well as biochemical features elucidated in numerous studies reveal a remarkable multiplicity of regulatory mechanisms.

Oxidative enzymes generally employ an electron donor. NOS is unprecedented in employing five. Cloned NOS displays recognition sites for NADPH, flavin mononucleotide (FMN), and flavin adenine dinucleotide (FAD). Direct biochemical analysis shows FAD and FMN bound stoichiometrically to NOS (5, 49). The only other mammalian enzyme that possesses recognition sites for both FMN and FAD as well as NADPH is cytochrome P-450 reductase (CPR) (Fig. 2). CPR is the electron donor for the liver's drug-metabolizing cytochrome P-450 enzymes. The carboxyl half of NOS displays about 60% amino acid identity to CPR. Presumably, early in evolution, CPR donated electrons for NOS, and at some point a fusion between CPR and NOS took place. Indeed, when the N-terminal and C-terminal halves of NOS are expressed separately and mixed together, one obtains NOS catalytic activity (D. S. Bredt and S. H. Snyder, unpublished observations). NOS also utilizes tetrahydrobiopterin as an electron-transferring cofactor (5, 49). Recently, several groups showed that NOS contains bound heme which reacts with CO to form a species absorbing at 450 nm, indicating that NOS itself is a cytochrome P-450 enzyme (5, 49). It is likely that the mechanism of electron transfer is similar to that of the P-450 enzymes—namely, that NADPH reduces FAD, which reduces FMN, which, in turn, transfers electrons to the ferric heme promoting the interaction with molecular oxygen. The exact role of tetrahydrobiopterin is not clear but probably involves a stabilization of the enzyme (5, 49).

NOS enzymes can be discriminated as inducible or constitutive. The brain and endothelial forms are constitutive in that stimuli for NO formation do not typically result in new enzyme protein synthesis. Instead, in the brain a stimulus (such as glutamate) acting at NMDA receptors triggers Ca2+ influx which binds to calmodulin, thereby activating NOS. This mode of activation explains the ability of glutamate neurotransmission to stimulate NO formation in a matter of seconds. In blood vessels, acetylcholine acting at muscarinic receptors on endothelial cells activates the phosphoinositide cycle to generate Ca2+, which stimulates NOS. Thus, constitutive NOS accounts for the role of NO in mediating rapid events such as neurotransmission and blood vessel dilatation. New synthesis of constitutive brain NOS can take place, because neuronal damage in the spinal cord is associated with the appearance of NADPH-diaphorase staining or newly immunoreactive NOS neurons and the induction of mRNA for brain NOS in dorsal root ganglia (5, 15). New synthesis of endothelial NOS in the brain also occurs following middle cerebral artery occlusion (72).

The inducible NOS of macrophages and nonmacrophage sources is not stimulated by Ca2+. Surprisingly, inducible NOS enzymes possess calmodulin recognition sites (Fig. 2). Nathan and colleagues (10) have shown that calmodulin is very tightly bound to inducible NOS, with the binding unaffected by Ca2+, whereas calmodulin cannot bind to neuronal NOS unless Ca2+ is present. The fact that calmodulin binds so tightly to inducible NOS that it can be considered an enzyme subunit accounts for the resistance of inducible NOS to Ca2+ activation (54).

Under normal circumstances, macrophages possess no detectable NOS protein. Stimuli such as interferon-g and lipopolysaccharide (LPS) elicit new NOS protein synthesis over 2–4 hr, mediating the NO responses to inflammatory stimuli. It was first thought that macrophages contained the only form of inducible NOS. Following endotoxin treatment, inducible NOS activity has been demonstrated in a great diversity of animal tissues lacking macrophages (54). The hepatocyte inducible NOS which has been recently cloned (27) might represent the prototype for nonmacrophage inducible NOS. Conceivably the ubiquitous distribution of this form of inducible NOS reflects a primitive sort of immune response. The simplicity of the NO system might have sufficed to repel invading microorganisms early in evolution.

NOS can also be regulated by phosphorylation. Consensus sequences for phosphorylation by cAMP-dependent protein kinase are evident in neuronal and endothelial NOS and hepatic inducible NOS (Fig. 2). These are not as obvious in the macrophage form of NOS. Consensus sites for phosphorylation by other kinases have not been characterized in detail. However, biochemical studies indicate that neuronal NOS can be phosphorylated by cAMP-dependent protein kinase, protein kinase C, cGMP-dependent protein kinase, and Ca2+/calmodulin-dependent protein kinase (5, 15). Phosphorylation by all of these enzymes decreases enzyme catalytic activity (5, 15; J. L. Dinerman, J. P. Steiner, T. M. Dawson, and S. H. Snyder, in preparation). This provides for multiple levels of enzyme regulation. For instance, Ca2+-calmodulin can directly activate the enzyme and, by phosphorylation through Ca2+/calmodulin-dependent protein kinase, inhibit enzyme activity. Ca2+, together with lipids, also activates protein kinase C, whose actions would also inhibit NOS. NO stimulates GC to form cGMP, which, via cGMP-dependent protein kinase, can inhibit NOS.

For inducible NOS one would expect the regulatory region of the gene to determine the rate of synthesis of enzyme protein. Characterization of the promoter region of the gene for macrophage inducible NOS (macNOS) reveals a pattern for complex regulation (44, 70). There appear to be two distinct regulatory regions upstream of the TATA box, which is 30 base pairs upstream of the transcription start site. One of these, region 1, lies about 50–200 base pairs upstream of the start site. Region 1 contains LPS-related response elements such as the binding site for NF-IL6 and the KB binding site for NFKB, indicating that this region regulates the LPS-induced expression of macNOS. Region 2, which is about 900–1000 bases upstream of the start site, does not itself directly regulate NOS expression, but provides a 10-fold increase above the 75-fold increase in NOS expression provided by region 1. Region 2 contains motifs for interferon-g-related transcription factors and thus is presumably responsible for interferon-g-mediated regulation. In sum, LPS- and interferon-g-responsive elements occur in two distinct regulatory genes: LPS directly stimulates macNOS expression, whereas interferon-g acts only in the presence of LPS.

This unique organization of gene enhancers may explain important aspects of inflammation. In sepsis, LPS is released from gram-negative bacterial cell walls and circulates throughout the body to stimulate inflammatory responses. By contrast, interferon-g is released locally and serves to augment inflammatory responses in specific cell populations close to its release. LPS alone stimulates macrophages only to a limited extent. Interferon-g elaborated by infiltrating lymphocytes can prime the macrophages for a maximal response to LPS. Thus maximal production of NO is restricted to those cells needed to kill the invader, thereby minimizing damage to adjacent tissue.

The NOS proteins are fairly large proteins. Neuronal NOS, the largest, has a molecular weight of 160 kD and occurs as a dimer. Endothelial and the inducible NOS enzymes are in the range of 130 kD and also function as dimers. Neuronal and macrophage NOS have been characterized largely as soluble proteins, though subcellular fractionation reveals a substantial amount of particulate neuronal NOS which is not readily solubilized by high salt concentrations (D. S. Bredt and S. H. Snyder, unpublished observations) and a particulate neuronal NOS has been purified from rat cerebellum (32). Endothelial NOS is predominantly particulate (56). Molecular cloning of endothelial NOS reveals no obvious transmembrane-spanning regions (5, 15, 45, 49). However, there is a consensus motif for N-terminal myristoylation, whose deletion in mutagenesis experiments renders NOS soluble (9). Moreover, [3H]myristate is directly incorporated into endothelial NOS (9). Insertion of the myristoyl group in the plasma membrane presumably accounts for the enzyme's particulate location.

With any neurotransmitter, major insight into function comes with information about localization. Purification of neuronal NOS permitted the development of antibodies for immunohistochemical staining (4). Throughout the brain, neuronal NOS occurs only in neurons. In many areas such as the cerebral cortex, hippocampus, and corpus striatum, NOS neurons comprise only about 2% of all the cells. They are scattered in no obvious pattern and display morphologic properties of medium-to-large aspiny neurons. In the hippocampus, none of the pyramidal cells contain NOS, but granule cells of the dentate gyrus have abundant NOS. In the corpus striatum, NOS occurs in both the cell bodies and terminals of the medium aspiny neurons. In most areas, NOS-containing cells are prominent, whereas in the islands of Callejae, NOS staining is confined to a dense fiber bundle.

In striking contrast to the pattern in the cerebral cortex, in the cerebellum NOS occurs in a high proportion of certain cell types. For instance, NOS is abundant in all granule cells and all basket cells, but in no Purkinje cells. This pattern explains how glutamate influences cGMP in the cerebellum. Endogenous cGMP is selectively concentrated in Purkinje cells, which receive input from terminals of granule and basket cells. Granule and basket cells possess NMDA receptors. Presumably, stimulation of the NMDA receptors on basket and granule cells triggers formation of NO which diffuses to Purkinje cells to activate GC.

While GC is clearly a target for NO in the cerebellum, this link may not be universal throughout the brain. If NO transmission occurred exclusively through GC and if all the GC in the brain were associated with NO transmission, then GC and NOS localizations should be closely similar. However, they differ markedly, indicating that NO may act through other targets than GC and/or GC may be the target for other transmitters besides NO.

Many, if not all, neurons in the brain contain more than one neurotransmitter. There does not appear to be a specific pattern for NOS. Thus, in the cerebellum NOS occurs in the glutamate-containing granule cells as well as in the gamma-aminobutyric acid (GABA)-containing basket cells. Many of the cerebral cortical NOS neurons also contain GABA. In the corpus striatum all NOS neurons stain for somatostatin and neuropeptide Y, but in areas such as the pedunculopontine nucleus of the brainstem, NOS neurons lack somatostatin and neuropeptide Y but stain for choline acetyltransferase (13).

What are the normal functions of these NOS neurons? One answer is that NO is responsible for cGMP generation. This answer leads to another question, namely, What is the role of cGMP? Though cGMP has been studied in the brain for well over 30 years, its exact functions remain obscure.

NO appears to influence neurotransmitter release. In several model systems, NOS inhibitors such as nitroarginine block the release of neurotransmitters (15)). In brain synaptosomes the release of neurotransmitter evoked by stimulation of NMDA receptors is blocked by nitroarginine (33), whereas release elicited by potassium depolarization is not affected (33). Presumably, glutamate acts at NMDA receptors on NOS terminals to stimulate the formation of NO, which diffuses to adjacent terminals to enhance neurotransmitter release, so that blockade of NO formation inhibits release. Potassium depolarization will release transmitter from all terminals so that any effect of NO would be masked.

PC12 cells, which develop neuronal properties in the presence of nerve growth factor, provide a valuable system linking NO to transmitter release. Rogers and colleagues (59, 60) showed that the release of acetylcholine in response to depolarization is markedly enhanced after 8 days of nerve growth factor application. NOS staining and NOS catalytic activity, which are absent in untreated PC12 cells, do not appear until 8 days, coincident with marked enhancement of neurotransmitter release. Release of both acetylcholine and dopamine from the cells is blocked by NOS inhibitors and reversed by excess L-arginine (33).

Direct evidence for specific neurotransmitter functions of NO comes from studies in the peripheral autonomic nervous system. NOS neurons occur in the myenteric plexus throughout the gastrointestinal pathway (2, 13). Depolarization of myenteric plexus neurons is associated with relaxation of the smooth muscle associated with peristalsis. The blockade of this process by NOS inhibitors indicates that NO is the transmitter (5, 52, 54).

In blood vessels, besides localizations in the endothelium, NOS occurs in autonomic nerves in the outer, adventitial layers of various large blood vessels (2, 55). In the cerebral cortex and the retina these neurons derive from cells in the sphenopalatine ganglia at the base of the skull (55). Approximately 40% of the NOS neurons contain the neuropeptide vasoactive intestinal polypeptide (VIP) (55). NOS neurons are prominent in penile tissue, specifically the pelvic plexus and its axonal processes that form the cavernous nerve as well as the nerve plexus in the adventitia of the deep cavernosal arteries and the sinusoids in the periphery of the corpora cavernosa (8). Electrical stimulation of the cavernous nerve in intact rats produces prominent penile erection which is blocked by low doses of intravenously administered NOS inhibitors (8). Nerve-stimulation-induced relaxation of isolated corpus cavernosum strips is also blocked by NOS inhibitors (58). These findings establish that NO is the transmitter of these nerves which regulate penile erection.

In the adrenal gland, NOS occurs in discrete ganglion cells and fibers in the medulla (2, 13). Splanchnic nerve stimulation augments both blood flow and catecholamine secretion from the adrenal medulla, with nitroarginine blocking the blood flow but not catecholamine secretion (7). NOS is also prominent in fibers and terminals in the posterior pituitary gland (2, 13), but its relation to function has not yet been established.

NO has been implicated in long-term potentiation (LTP) in the hippocampus. Nitroarginine application to hippocampal slices blocks LTP formation. Injection of nitroarginine into pyramidal cells of the hippocampus also inhibits LTP, suggesting that NO might act as a retrograde messenger for LTP passing from pyramidal cells to Schaffer collateral terminals (4, 5, 15). However, neuronal NOS is not yet demonstrable in these pyramidal cells (4).

While NO mediates normal synaptic transmission, excess levels of NO may be neurotoxic. Evidence has accumulated for a number of years that glutamate released in excess, acting via NMDA receptors, mediates neurotoxicity in the focal ischemia of vascular stroke (11) which is blocked by NMDA antagonists (50). Glutamate neurotoxicity may also contribute to dysfunction in neurodegenerative diseases such as Alzheimer's and Huntington's diseases. Because glutamate, via NMDA receptors, stimulates NO formation, one might expect excess NMDA receptor stimulation to destroy NOS neurons. Surprisingly, NOS neurons are resistant to NMDA neurotoxicity (17). This conclusion derives from the demonstration that NOS neurons are identical to those that stain for NADPH-diaphorase (13, 34). Diaphorase staining reflects a blue precipitate obtained with tetrazolium dyes in the presence of NADPH (5, 15). Numerous studies have demonstrated that diaphorase-staining neurons are notably resistant to destruction in Huntington's and Alzheimer's diseases, in vascular stroke, and in NMDA neurotoxicity (15). We were struck with the close similarity in localizations of diaphorase- and NOS-staining neurons. Transfection of neuronal NOS cDNA into human kidney 293 cells lacking NOS or diaphorase results in staining of the cells for diaphorase and NOS in exactly the same proportions as neurons in the brain (13). Because diaphorase derives from any NADPH oxidative activity, most diaphorase in brain homogenates is unrelated to NOS, but a discrete portion represents NOS (34).

If NMDA stimulates NOS neurons to make NO, but these cells are themselves resistant to neurotoxicity, could the released NO damage other cells? Exposure of cerebral cortical cultures to NMDA kills 60–90% of neurons, with NOS-diaphorase cells being undamaged (17, 37). Treatment with nitroarginine or other NOS inhibitors or removal of arginine from the media block this neurotoxicity (17, 18). The toxicity is also prevented by flavoprotein and calmodulin inhibitors. Superoxide dismutase attenuates neurotoxicity. Because this enzyme removes superoxide which interacts with NO to form the toxic radical peroxynitrite, NO presumably kills via peroxynitrite. NO has been implicated in NMDA neurotoxicity in a variety of models, including hippocampal slices, striatal slices, and several culture systems (5, 15). Others have failed to show that NO is involved in NMDA neurotoxicity (15), and NO may be neuroprotective (38). NO may exert both neurodestruction and neuroprotection, depending on its oxidation–reduction status (40), with NO- being neurodestructive and NO+ being neuroprotective (40). If NMDA neurotoxicity is responsible for neuronal damage in vascular stroke, then NOS inhibitors should be neuroprotective. Administration of low doses of nitroarginine blocks neural damage following middle cerebral artery occlusion in mice, rats, and cats (15). High doses of NOS inhibitors exacerbate the damage following occlusion of the middle cerebral artery (15!popup(ch60ref15)), presumably through decreased cerebral blood flow.

Why are NOS neurons resistant to NMDA toxicity? One possibility arises from findings of Michel et al. (51) regarding translocation of NOS in endothelial cells. Phosphorylation of NOS translocates the enzyme from membrane to soluble fractions. Because phosphorylated NOS is catalytically inactive, NO will not be generated within the cytoplasm. Instead, catalytically active, nonphosphorylated NOS is localized to the plasma membrane, where it presumably generates NO that is released into the extracellular environment. While neuronal NOS has been thought to be predominantly soluble, about 50% of NOS activity in brain homogenates is particulate and cannot be solubilized even with strong salt treatment (D. S. Bredt and S. H. Snyder, unpublished observations). Thus, in neurons as well as blood vessels the active form of NOS may be the unphosphorylated enzyme localized to the plasma membrane to release NOS to the exterior. Presumably, NO is never released in the interior of NOS cells, which accordingly are resistant to NO damage.

NO can mediate other forms of neurotoxicity. The pathophysiology of acquired immunodeficiency syndrome (AIDS) dementia has been a puzzle, because little human immunodeficiency virus (HIV) virus is detected in neurons in the brain. Instead, the gp120 coat protein appears to mediate some of the toxicity. Extremely low, picomolar concentrations of gp120 kill neurons in primary cortical cultures (6, 21). The killing is absolutely dependent upon the presence of glutamate acting through NMDA receptors (19, 41). We showed that this toxicity requires NO, being absent in arginine-free medium and blocked by various inhibitors of NOS (19). The gp120 neurotoxicity also requires the presence of macrophages and/or astrocytes (39). These cells produce cytokines and arachidonic acid metabolites which can potentiate NMDA receptor currents. Conceivably, gp120 elicits release of arachidonic acid metabolites and cytokines from macrophages and glia, which synergize with glutamate to activate NMDA receptors, in turn triggering the formation of NO (Fig. 3).

Neuroprotection can derive from indirect means of NOS inhibition. Gangliosides are neuroprotective in animal models of neural damage and in patients with spinal cord injury (46, 62). Based on observations that gangliosides bind calmodulin (30, 31), we showed that a series of gangliosides inhibit NOS activity and prevent glutamate toxicity in neuronal cultures with potencies closely paralleling their affinities for calmodulin and their ability to inhibit NOS (14).

Immunosuppressants such as FK506 and cyclosporin A bind to small soluble receptor proteins. The drug–receptor complex in turn binds to the Ca2+ activated phosphatase calcineurin to inhibit calcineurin activity (42). Thus, treatment with these immunosuppressant drugs leads to accumulation of phosphorylated substrates of calcineurin (64). We observed that NOS is a calcineurin substrate and that phosphorylated NOS levels are enhanced by FK506 and cyclosporin A (16). Because phosphorylated NOS is catalytically inactive, treatment with the immunosuppressants should be equivalent to treatment with NOS inhibitors. Indeed, both FK506 and cyclosporin A block NMDA neurotoxicity in cortical cultures at very low concentrations (Fig. 4). The neuroprotective effect of FK-506 may have clinical relevance. FK506 and cyclosporin A have been employed extensively in organ transplant surgery. FK506 penetrates readily into the brain, whereas cyclosporin A does not. In a study of liver transplantation, 7 of 14 patients receiving cyclosporin A showed global cerebral ischemia, whereas none of the 14 patients receiving FK506 showed such alterations (43). Furthermore, cyclosporin A reduces infarct volume following middle cerebral artery infarction in rats (61).

While activation of GC accounts for some of the synaptic activities of NO, it cannot account for neurotoxicity, because inhibitors of GC do not block neurotoxicity (15, 17). Moreover, 8-bromo cGMP, which penetrates readily into cells, is not neurotoxic. Identifying the molecular target for neurotoxicity is difficult, because when a cell is killed, all of its biochemistry deteriorates. Numerous candidates for the toxic actions of NO have been investigated.

NO activates GC by binding to iron in the heme which is at the active site of the enzyme, altering the enzyme's conformation to activate it. NO can bind to nonheme iron in numerous enzymes such as NADH-ubiquinone oxidoreductase, NADH:succinate oxidoreductase, and cis-aconitase, all iron–sulfur enzymes (54). NO can bind to the iron in ferritin, an iron storage protein, thereby liberating the iron, which could cause lipid peroxidation (54). NO also binds to the nonheme iron of ribonucleotide reductase to inhibit DNA synthesis (54). Its ability to bind iron enables NO to influence iron metabolism. Iron metabolism is regulated post-transcriptionally by specific mRNA–protein interactions between iron-regulatory factor (IRF) and iron-responsive elements (IREs) which occur in the untranslated regions of the mRNA transcripts for the erythroid form of 5-aminolevulinate synthase, the transferrin receptor and ferritin (36, 53). Weiss et al. (68) recently showed that NO formed by macrophages augments the IRE binding activity of IRF which causes translational repression of IRE containing messenger RNA.

NO can stimulate the S-nitrosylation of numerous proteins (63). NO also stimulates the auto-ADP-ribosylation of glyceraldehyde-3-phosphate dehydrogenase (5, 15). This ADP-ribosylation takes place at the cysteine which is at the active site of the enzyme, hence inhibiting its activity and potentially depressing glycolysis.

How is one to ascertain which of these actions is responsible for neurotoxicity? Recent studies provide evidence that DNA damage is central to NO neurotoxicity (71). NO, like other free radicals, can damage DNA by base deamination (69). DNA damage stimulates the activity of poly(ADP ribose) synthetase (PARS). PARS is a nuclear enzyme which utilizes NAD as a substrate to catalyze the attachment of 50–100 ADP-ribose units to nuclear proteins such as histones and, most prominently, to PARS itself (20). In brain homogenates incubated with [32P]NAD, NO stimulates the poly-ADP-ribosylation of PARS (71). NMDA neurotoxicity in cortical cultures is blocked by a series of PARS inhibitors with potencies closely paralleling their potencies in inhibiting PARS. These observations implicate the following series of events in NO neurotoxicity (Fig. 5). NO damages DNA to activate PARS. Massive activation of PARS depletes the cell of NAD and ATP, because four high-energy phosphate bonds, the equivalent of four molecules of ATP, and one of NAD are consumed in the activation of PARS and the regeneration of NAD, respectively. Considering that PARS is a particularly abundant protein and that catalytic activity involves the addition of up to 100 ADP-ribose units to a single protein molecule, it is not surprising that depletion of energy sources takes place when PARS is activated. The resulting cell death accordingly can be blocked by PARS inhibitors. With lesser degrees of DNA damage, PARS activation is thought to facilitate DNA repair (24) (see Functional Brain-Imaging Studies in Schizophrenia, Neuropsychological Assessment of Patients with Alzheimers's Disease, The Treatment of Tardive Dyskinesias, and Prion Diseases).

The dramatic properties of NO suggested that it may not be the only gaseous, labile small molecule transmitter. CO is an additional candidate. There are several resemblances between NO and CO disposition. Electrons for NO synthesis are donated by a CPR-like activity of NOS. CPR itself donates electrons to heme oxygenase, the enzyme that makes CO. Heme oxygenase cleaves the heme ring into CO and biliverdin, which is rapidly reduced to bilirubin. Also like NO, CO can bind to the iron in heme, accounting for CO's lethality because hemoglobin can no longer deliver oxygen to tissues. CO can bind to the heme in GC to activate cGMP formation (47). In cultures of olfactory neurons, we showed that CO is responsible for maintaining endogenous cGMP levels, because potent, selective inhibitors of heme oxygenase deplete cGMP, whereas nitroarginine is ineffective (67). Also, as with NOS, heme oxygenase displays discrete localizations. Our in situ hybridization studies demonstrated selective localizations with high levels in the pyramidal cells of the hippocampus and dentate gyrus (67). In the cerebellum, high concentrations are evident in the granule and Purkinje cells layers, while the pontine nuclei are also heavily labeled. Large densities of heme oxygenase 2 (HO2) are also observed in the piriform cortex, tenia tecta, olfactory tubercle, and islands of Callejae. Highest densities in the brain are in the neurons of the olfactory epithelium and in the neuronal and granule cell layers of the olfactory bulb. Somewhat similar localizations are evident in immunohistochemical studies of enzyme protein (22) (J. L. Dinerman and S. H. Snyder, unpublished observations).

Like NO, there are two systems for CO generation. An inducible heme oxygenase (HO1) is responsible for the destruction of heme in aging red blood cells (12, 66). New HO1 protein is formed in response to heme as well as numerous oxidative stressors. In contrast, the constitutive form of the enzyme, HO2, is not inducible. HO1 is concentrated in peripheral tissues such as the spleen and liver, while particularly high densities of HO2 occur in the brain (66, 67).

Though only recently identified in the brain, CO already has been implicated in various functions. Because HO2 is concentrated in hippocampal pyramidal cells, CO might be a candidate for the retrograde messenger of LTP. Zinc protoporphyrin IX (ZnPP-9) blocks the induction of LTP in the CA1 region of hippocampal slices (73). Application of CO to slices produces a long-lasting increase in the size of evoked synaptic potentials when applied at the same time as a weak tetanic stimulation (65).

Just as NO mediates certain glutamate actions at NMDA receptors, CO may be responsible for glutamate effects via metabotropic receptors. Glaum and Miller (28) showed that metabotropic receptor activation in the solitary tract nucleus of the brainstem regulates a specific channel conductance through a cGMP mechanism. NO synthase inhibitors fail to alter the conductance, but ZnPP-9 and related agents block effects of receptor stimulation in proportion to their potencies as inhibitors of heme oxygenase.

CO also appears to be involved in the regulation of carotid body sensory activity. The chemosensors of the carotid body are regulated by molecular oxygen and inhibited by CO (29). HO2 activity is readily demonstrable in the carotid body (J. L. Dinerman, N. R. Prabhakar, and S. H. Snyder, unpublished observations), and ZnPP-9 markedly enhances the chemosensory discharge of the carotid body (57).

In a remarkably brief period of time, NO and CO have been recognized as putative neurotransmitters. Their unexpected properties have revolutionized thinking about criteria for a chemical's candidacy as a neurotransmitter and about how synaptic transmission takes place. The involvement of NO and CO in several important areas of neuronal function suggests that agents influencing the disposition of NO and CO may have therapeutic relevance. Whether other gases and free radicals will join NO and CO as transmitters is an open question.

published 2000