Hydrogen sulfide as a gasotransmitter


  • Moataz M. Gadalla,

    1. Department of Pharmacology and Molecular Sciences, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author
  • Solomon H. Snyder

    1. Department of Pharmacology and Molecular Sciences, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    2. The Solomon H. Snyder Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    3. Department of Psychiatry and Behavioral Sciences, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
    Search for more papers by this author

Address correspondence and reprint requests to Solomon H. Snyder, The Solomon H. Snyder Department of Neuroscience, Johns Hopkins University School of Medicine, 725 North Wolfe Street, Baltimore, MD 21205-2105, USA. E-mail: ssnyder@jhmi.edu


J. Neurochem. (2010) 10.1111/j.1471-4159.2010.06580.x


Nitric oxide (NO) and carbon monoxide (CO) are well established as messenger molecules throughout the body, gasotransmitters, based on striking alterations in mice lacking the appropriate biosynthetic enzymes. Hydrogen sulfide (H2S) is even more chemically reactive, but until recently there was little definitive evidence for its physiologic formation. Cystathionine β-synthase (EC, and cystathionine γ-lyase (CSE; EC, also known as cystathionine, can generate H2S from cyst(e)ine. Very recent studies with mice lacking these enzymes have established that CSE is responsible for H2S formation in the periphery, while in the brain cystathionine β-synthase is the biosynthetic enzyme. Endothelial-derived relaxing factor activity is reduced 80% in the mesenteric artery of mice with deletion of CSE, establishing H2S as a major physiologic endothelial-derived relaxing factor. H2S appears to signal predominantly by S-sulfhydrating cysteines in its target proteins, analogous to S-nitrosylation by NO. Whereas S-nitrosylation typically inhibits enzymes, S-sulfhydration activates them. S-nitrosylation basally affects 1–2% of its target proteins, while 10–25% of H2S target proteins are S-sulfhydrated. In summary, H2S appears to be a physiologic gasotransmitter of comparable importance to NO and carbon monoxide.

Abbreviations used:



large-conductance calcium-activated potassium channels


cystathionine β-synthase


carbon monoxide


cystathionine γ-lyase


endothelial-derived hyperpolarizing factor


endothelial-derived relaxing factor


endothelial nitric oxide synthase


glyceraldehyde 3-phosphate dehydrogenase


hydrogen sulfide


heme oxygenase


inducible nitric oxide synthase


ATP-sensitive potassium channels


long-term potentiation


neuronal nitric oxide synthase


nitric oxide


nitric oxide synthase




phosphatidylinositol (4,5)-bisphosphate


pyridoxal 5′-phosphate


S-adenosyl methionine


soluble guanylyl cyclase

The notion that gases can serve as messenger molecules stems largely from research indicating that nitric oxide (NO) is a physiologic vasodilator and mediates the tumoricidal/bactericidal actions of macrophages (reviewed in Moncada et al. 1991). Subsequently, NO was established as a neurotransmitter/neuromodulator in the brain and peripheral nervous system (Bredt and Snyder 1989, 1990; Bredt et al. 1990, 1991a,b, 1992; Burnett et al. 1992; Nelson et al. 1995). Soon thereafter, evidence accumulated establishing carbon monoxide (CO) as physiologically generated and mediating non-adrenergic non-cholinergic neurotransmission in the intestine as well as neural activity in the brain (Verma et al. 1993; Zakhary et al. 1997; Xue et al. 2000; Boehning et al. 2004). Both of these gaseous molecules are well accepted as gasotransmitters; a term which, as used here, does not necessarily imply that the gaseous molecule is a neurotransmitter but rather that it transmits information between cells in various parts of the body.

It was easy to accept that NO and CO are physiologically relevant, once the biosynthesis of both substances was established from reasonably well characterized enzymes. In the case of NO, three isoforms of NO synthase (NOS; EC, derived from three distinct genes, convert arginine to NO and citrulline, with neuronal NOS (nNOS) highly localized to the brain and peripheral nerves as well as a few non-neural tissues, endothelial NOS (eNOS) generating NO that regulates blood vessels, and inducible NOS (iNOS) occurring ubiquitously throughout the body, but with highest densities in inflammatory cells such as macrophages. nNOS and eNOS are constitutive enzymes activated by calcium-calmodulin which explains their rapid augmentation in response to depolarizing events (Bredt and Snyder 1989). By contrast, iNOS is inducible, largely in response to inflammatory stimulation, and is not notably influenced by calcium (Cho et al. 1992; Lowenstein et al. 1992, 1993). Mice with targeted deletion of the three enzymes lose the capacity to generate NO in the relevant target organs (Huang et al. 1993, 1995; MacMicking et al. 1995; Wei et al. 1995; Shesely et al. 1996; Son et al. 1996; Morishita et al. 2005).

Carbon monoxide has long been known to be formed by two isoforms of heme oxygenase (HO) which derive from distinct genes (Maines 1988). HO-1 is a markedly inducible enzyme whose formation is stimulated by diverse stressors, including heme, and is abundant in liver, kidney and spleen; organs responsible for degradation and heme catabolism of aged red blood cells (Poss and Tonegawa 1997). By contrast, HO-2, localized to neurons in the brain and the endothelial layer of blood vessels, is constitutive and activated by calcium-calmodulin, much like nNOS and eNOS (Verma et al. 1993; Zakhary et al. 1996; Boehning et al. 2004). Although HO-2 is constitutive, glucocorticoids (Weber et al. 1994; Raju et al. 1997) and opiates (Li and Clark 2000; Panahian and Maines 2001) have been shown to increase HO-2 expression. HO-1 was first identified in aging red blood cells where it degrades the heme ring of hemoglobin generating biliverdin, which is rapidly reduced by biliverdin reductase to bilirubin. When the heme ring is cleaved at the α-meso carbon bridge, the one carbon fragment is liberated as CO by oxidation, a process that was well documented but largely overlooked by biologists until appreciation of NO led to demonstration that CO is also a gasotransmitter. Recently, mitochondrial soluble adenyl cyclase was found to be regulated by carbon dioxide/bicarbonate, indicating that carbon dioxide too might be a gasotransmitter (Acin-Perez et al. 2009).

Awareness of hydrogen sulfide (H2S) precedes by centuries the appreciation of NO and CO. It was referred to as aer hepaticus (hepatic air) by alchemists (Myers 2007). In 1777 Carl Wilhelm Scheele was the first chemist to prepare and characterize H2S, describing it as ‘sulfuretted hydrogen,’ in Chemische Abhandlung von der Luft und dem Feuer (Chemical Treatise on Air and Fire). H2S is odoriferous at concentrations less than 1 ppm, causes headaches at 4 ppm and is lethal at high levels (Reiffenstein et al. 1992). It is about five times more potent as a toxin than CO, acting largely by inhibiting cytochrome c oxidase (Lloyd 2006). All of us possess abundant levels of H2S in our gut derived predominantly from bacteria that can form H2S by the reduction of sulfate as well as the decomposition of sulfur containing amino acids such as cysteine and methionine, sulfated polysaccharides and sulfur containing lipids. Actions upon the gut of bacterially generated H2S are of some interest (Lloyd 2006). However, most biomedical researchers would be more disposed toward investigating a substance generated by mammalian enzymes under physiologic circumstances. Several pathways for the physiologic formation of H2S have been widely discussed and inhibitors of these enzymes influence H2S levels. However, none of the inhibitors have been extraordinarily potent or selective. Woody Allen apocryphally commented, ‘Ninety percent of life is showing up.’ In the absence of definitive evidence for the physiologic formation and function of H2S, the world of biomedical science would not be persuaded of a physiologic role for H2S. Very recently, deletion of a putative biosynthetic enzyme for H2S, cystathionine γ-lyase (CSE; EC, also known as cystathionase, was shown to deplete endogenous H2S levels and to markedly alter vasorelaxation and blood pressure (Yang et al. 2008). Hence, H2S now warrants inclusion in the family of gasotransmitters.


The two principal enzymes proposed as a physiologic sources of H2S both metabolize cystathionine. Cystathionine is well established as an intermediate in various cycles involving sulfur-containing amino acids but has not had a prominent role in biomedical research. It is formed by the enzyme cystathionine β-synthase (CBS; EC, which condenses homocysteine with serine to generate the thiol ether cystathionine (Fig. 1a). In the condensation, the hydroxyl group of serine is replaced with the thiolate of homocysteine. The gene of human CBS is localized to chromosome 21 at 21q22.3 (Münke et al. 1988). In human and rat CBS exists primarily as a homotetramer with a subunit molecular weight of 63 kDa. Each subunit also binds the cofactors pyridoxal 5′-phosphate (PLP), S-adenosyl methionine (SAM) and heme (Miles and Kraus 2004; Banerjee and Zou 2005). The heme appears to be a redox sensor, while SAM is an allosteric activator of the enzyme. The C-terminal portion of CBS contains a tandem repeat of two ‘CBS domains’ which appear to act as inhibitors of enzymatic function, as their deletion activates CBS (Kery et al. 1998; Shan and Kruger 1998). The CBS domains have been proposed to act as energy sensors (Scott et al. 2004).

Figure 1.

 (a) The classically described roles of CBS and CSE in sulfur metabolism. CBS condenses homocysteine with serine to generate the thiol ether cystathionine. CSE hydrolyzes cystathionine into cysteine, α-ketobutyrate and ammonia. (b) H2S producing reactions catalyzed by CBS and CSE. CBS catalyzes the β-replacement reaction of cysteine (Cy–SH) with a variety of thiols (R–SH) to generate H2S and the corresponding thiol ether (R–S–Cy). CSE catalyzes the β-disulfide elimination reaction of cystine (Cy–S–S–Cy), this is followed by a reaction with a variety of thiols, to generate H2S and the corresponding disulfide (R–S–S–Cy).

Recently CBS has been shown to be sumoylated at lysine 211 in the ‘CBS domain’ (Kabil et al. 2006). Sumoylation often elicits nuclear localization of proteins and may explain the substantial levels of CBS in the nucleus. Sumoylation inhibits the catalytic activity of CBS (Agrawal and Banerjee 2008). Interestingly, CBS physiologically binds huntingtin, the protein mutated in Huntington’s Disease (Boutell et al. 1998). Huntingtin itself is also sumoylated which enhances the neurotoxicity of mutant huntingtin (Steffan et al. 2004; Subramaniam et al. 2009).

Heme binds to the N-terminal portion of CBS comprising about 70 amino acids. In its ferrous state, this heme binds both CO and NO (Taoka and Banerjee 2001). CO binds with higher affinity, with a Ki of about 5.6 μM, while NO (Ki∼ 360 μM) is only about two percent as potent so that its binding probably is not physiologically relevant (Taoka et al. 1999; Taoka and Banerjee 2001). CO inhibits CBS activity. The interaction of CO with CBS is analogous to its interaction with heme in the transcription factor neuronal PAS (Per, Amt, Sim) domain protein 2 wherein CO disrupts the DNA binding activity of neuronal PAS domain protein 2 (Dioum et al. 2002). The potent influence of CO upon CBS raises the possibility of cross-talk between CO and H2S as messenger molecules.

S-adenosyl methionine activates CBS several fold by binding to the CBS domain in the carboxyl terminus of the enzyme (Kery et al. 1998; Shan and Kruger 1998). Thus, truncated CBS, lacking the C-terminus, displays fivefold greater catalytic activity than the native enzyme and is no longer stimulated by SAM (Taoka et al. 1999). The biologic rationale for activation of CBS by SAM is unclear. One possibility is that the CBS domain is an energy-sensing domain. This notion is based on findings that AMP-activated protein kinase binds CBS at its CBS domain (Scott et al. 2004). One wonders whether SAM regulation of CBS reflects some sort of reciprocal link between signaling by H2S and signaling by SAM’s methylation of multiple targets.

Cystathionine β-synthase can catalyze H2S formation from cysteine through a β-replacement reaction with a variety of thiols (Braunstein et al. 1971; Porter et al. 1974) (Fig. 1b). This is coupled with the formation of the corresponding thiol ether. CBS levels are relatively high in the brain where it is postulated to be the physiologic source of H2S (Abe and Kimura 1996). Using both cysteine and homocysteine as co-substrates simultaneously, the Vmax of H2S production for human CBS is 22–40 fold higher than for cysteine alone (Singh et al. 2009). In this reaction, the Km values for cysteine and homocysteine are 6.8 mM and 3.2 mM, respectively. Accordingly, homocysteine might be a preferred co-substrate for H2S generation. In determining whether CBS physiologically generates H2S, many investigators have relied upon the inhibitors, hydroxylamine and amino-oxyacetate (Abe and Kimura 1996). These do inhibit the generation of H2S from cysteine in brain homogenates, but both are general inhibitors of all PLP-dependent enzymes.

Cystathionine γ-lyase can also form H2S from cyst(e)ine (Cavallini et al. 1962a,b; Szczepkowski and Wood 1967) (Fig. 1b), though the classical function of CSE is to hydrolyze cystathionine into cysteine with ammonia and α-ketobutyrate as byproducts (Fig. 1a). The enzyme converts cystine to thiocysteine, pyruvate and ammonia, in a β-disulphide elimination reaction, with the thiocysteine then reacting with cysteine or other thiols to produce H2S and cystine or the corresponding disulfide (Fig. 1b). In most peripheral tissues CSE levels are much higher than those of CBS, while in the brain, CBS predominates (Yang et al. 2008; Mustafa et al. 2009a,b; Abe and Kimura 1996).

Cystathionine γ-lyase inhibitors have been employed to examine the enzyme’s role in generating H2S physiologically. The two principal inhibitors utilized are dl-propargylglycine (PAG) (Abeles and Walsh 1973; Washtien and Abeles 1977) and β-cyano-l-alanine (β-CNA) (Pfeffer and Ressler 1967). They influence other enzymes such as cystathionine γ-synthetase (EC (Marcotte and Walsh 1975), methionine γ-lyase (EC (Johnston et al. 1979), aminotranferases (Marcotte and Walsh 1975; Tanase and Morino 1976; Alston et al. 1980; Burnett et al. 1980) and d-amino acid oxidase (EC (Horiike et al. 1975; Marcotte and Walsh 1976). Thus, one must be cautious in interpreting results utilizing such agents. However, it is of interest that PAG and β-CNA do suppress H2S production by the liver and kidney but not by the brain; fitting with other evidence that CBS is the predominant source of H2S in brain tissue (Abe and Kimura 1996).

Like CBS, CSE is a PLP-dependent enzyme. If CSE were to generate H2S as a physiologic signaling molecule, one might expect it to be influenced by signaling systems such as calcium. Indeed, CSE is selectively activated by calcium-calmodulin similar to the activation of eNOS, nNOS and HO-2 (Yang et al. 2008).

Definitive evidence that CSE is a physiologic source for H2S comes from experiments employing CSE knockout mice (Yang et al. 2008). H2S levels in aorta and heart of homozygous CSE knockout mice are reduced by about 80% with a 50% reduction in heterozygous knockouts. Serum H2S levels in homozygous and heterozygous CSE knockouts are reduced 50% and 20%, respectively. The residual H2S in mutant serum may reflect non-enzymatic reduction of elemental sulfur to H2S or H2S generated from other tissues by CBS. The studies with CSE knockouts establish that H2S is a product of normal mammalian physiology.

Hydrogen sulfide is presumed to exist in an ionized form in most tissues as HS. Kimura and associates (Ishigami et al. 2009; Shibuya et al. 2009) have characterized a form of H2S which they refer to as ‘bound sulfur.’ This material presumably arises when the sulfur of H2S is incorporated into proteins, bound to other sulfur atoms to form persulfides. Presumably this bound sulfur releases H2S under reducing conditions. These authors showed that the bound H2S was not depleted in CBS knockout mouse brain (Ishigami et al. 2009). It was possible to generate this H2S pool from cysteine by the coordinate actions of two enzymes, 3-mercaptopyruvate sulfurtransferase (EC and cysteine aminotransferase (EC The physiologic significance of this pool of sulfur is unclear. Definitive evidence awaits studies with deletion of the postulated enzymes utilizing techniques such as RNA interference or mutant mice.

Signaling mechanisms

Signaling by NO was first characterized in terms of its relaxation of blood vessels. NO binds with high affinity to heme in the active site of soluble guanylyl cyclase (sGC), altering the enzyme’s conformation and enhancing its catalytic activity. Generated cyclic GMP then leads to smooth muscle relaxation through activation of cyclic GMP-dependent protein kinase which results in protein phosphorylation, a decrease in cytosolic calcium, and dephosphorylation of the myosin light chain. CO also activates sGC but is substantially less potent than NO. Its potency is dramatically increased in the presence of certain agents such as 3-(5′-hydroxymethyl-2′-furyl)-1-benzylindazole, a benzyl indazole derivative (Friebe et al. 1996). Conceivably, conformational alterations such as those elicited in the enzyme by 3-(5′-hydroxymethyl-2′-furyl)-1-benzylindazole occur in intact organisms and lead to enhanced and physiologic potency of CO in vivo. Such a view would be consonant with direct evidence that cyclic GMP levels in various tissues are markedly depleted in HO-2 knockout mice (Zakhary et al. 1997; Watkins et al. 2004).

Hydrogen sulfide also binds with high affinity to heme. However, it does not appear to physiologically stimulate sGC (Abe and Kimura 1996). Moreover, the ability of H2S to relax blood vessels is not impaired in the presence of inhibitors of sGC (Zhao et al. 2001).

If H2S does not act through sGC, how does it signal? A clue comes from NO, which can S-nitrosylate cysteines of various proteins (Stamler et al. 1992a,b; Stamler et al. 1997). Because both NO and the thiol groups of cysteines are chemically reactive, armchair chemistry would predict nitrosylation of cysteines in proteins (Fig. 2). Stamler and associates (Jia et al. 1996; Xu et al. 1998; Mannick et al. 1999) showed such modification for a wide range of proteins. Demonstration of physiologic nitrosylation of numerous proteins under basal conditions by endogenously generated NO was rendered feasible by development of the biotin switch assay (Jaffrey and Snyder 2001; Jaffrey et al. 2001). In this procedure free thiols are blocked by the sulfhydryl-reactive compound, methyl methane thiolsulfonate; the nitrosylated thiols are then exposed by treatment with ascorbate, labeled with biotin, coupled to streptavidin, and nitrosylated proteins are then separated by gel electrophoresis. A substantial number of proteins are basally nitrosylated, including glyceraldehyde 3-phosphate dehydrogenase (GAPDH; EC, glycogen phosphorylase (EC, creatine kinase (EC, sodium/potassium adenosine triphosphatase (EC, NMDA-glutamate receptor, β-tubulin and actin. Nitrosylation of these and other proteins is abolished in nNOS knockout mouse brain (Jaffrey et al. 2001).

Figure 2.

 A model protein with some of the possible states of the cysteine thiol groups. From the top to the bottom, a free thiol (–SH), an S-nitrosylated thiol (–SNO), an S-sulfhydrated thiol (hydropersulfide) (–SSH) and a disulfide is shown.

In the absence of ascorbate some proteins were still labeled; indicating that in addition to S-nitrosylation, ascorbate-dependent labeling, there was another thiol modification of cysteine that was labeled independent of ascorbate. Mass spectrometric analysis indicated that the labeling reflects S-sulfhydration, attachment of an additional sulfur to the thiol (–SH) groups of cysteines yielding a hydropersulfide (–SSH) moiety (Mustafa et al. 2009b) (Fig. 2). This is not to be confused with S-thiolation or S-thionylation, in which a protein thiol forms a mixed disulfide with a small-molecular weight thiol such as glutathione or cysteine (Thomas et al. 1995). S-thiolation blocks the protein thiol rendering it non-reactive, whereas S-sulfhydration yields an –SSH moiety which has enhanced chemical reactivity.

Numerous proteins, such as β-tubulin, actin, and GAPDH, are basally sulfhydrated. For most proteins, especially GAPDH in the liver, sulfhydration is substantially more prevalent than nitrosylation. Sulfhydration is abolished in CSE knockout mouse liver, but is unaffected in livers of nNOS, eNOS and iNOS knockouts. Sulfhydration occurs at physiologic levels of l-cysteine with maximal stimulation of GAPDH, β-tubulin and actin at about 0.6–1 mM l-cysteine, comparable to its physiologic concentrations in the liver.

Nitrosylation of most enzymes and receptors inhibits their activity. This fits with the importance of cysteine thiols for activities of many proteins and nitrosylation masking the critical reactive thiol groups. By contrast, sulfhydration merely changes an –SH to an –SSH which would enhance chemical reactivity and might even afford greater access to targets. Indeed, whereas nitrosylation of GAPDH abolishes its catalytic activity (Hara et al. 2005), H2S elicits a sevenfold increase in GAPDH activity (Fig. 3a; Mustafa et al. 2009b). Dithiothreitol (DTT) reverses GAPDH activation by H2S (Fig. 3b), and H2S fails to increase the activity of C150S mutant GAPDH (Fig. 3c), consistent with the H2S augmentation of GAPDH activity occurring via sulfhydration at C150 (Mustafa et al. 2009b). H2S increases the Vmax of GAPDH with no effect on Km (Fig. 3d; Mustafa et al. 2009b). Activation of GAPDH by H2S enzymatically generated from l-cysteine by CSE is observed in HEK293 cells transfected with CSE (Fig. 3e; Mustafa et al. 2009b). Similarly, sulfhydration directly enhances actin polymerization with no effect on its depolymerization (Mustafa et al. 2009b).

Figure 3.

 (a) Sulfhydration physiologically increases the catalytic activity of GAPDH. GAPDH activity assay in vitro at 37°C with increasing sodium hydrogen sulfide (NaHS) levels. NaHS dose-dependently activates GAPDH. (b) DTT (1 mM) reverses GAPDH activation by 10 μM NaHS in vitro. All results are mean ± SEM. **< 0.01. (c) Wild-type versus C150S mutant GAPDH activity in vitro with 15 μM NaHS. Wild-type (wt) but not C150S GAPDH is activated by NaHS. All results are mean ± SEM. **< 0.01. (d) GAPDH activity with increasing substrate, glyceraldehyde 3-phosphate (G3P), levels with or without 10 μM NaHS. NaHS increases overall Vmax without affecting Km (∼0.8 mM). (e) GAPDH activity in HEK293 cells transfected with nothing, or plasmids endcoding wild-type CSE, or catalytically inactive CSE and incubated with increasing concentrations of l-cysteine in the media for 1 h at 37°C. GAPDH is activated in a dose-dependent manner in the presence of wild-type CSE. (f) In vivo GAPDH activity from wild-type versus CSE−/− liver. CSE−/− mice show decreased GAPDH activity (n = 6 animals). All results are mean ± SEM. *< 0.05. Reproduced with permission from Mustafa et al. (2009b).

Sulfhydration is a prominent post-translational modification with 10–25% of endogenous GAPDH, β-tubulin and actin basally sulfhydrated (Mustafa et al. 2009b). By contrast, physiologic nitrosylation levels affects only 1–2% of target proteins (Jaffrey et al. 2001). The physiologic relevance of sulfhydration is evident in the reduction of GAPDH activity by about 25–30% in livers of CSE knockout mice despite normal levels of GAPDH protein (Fig. 3f; Mustafa et al. 2009b). This finding corresponds reasonably well with the extent of activation elicited by H2S and the proportion of total GAPDH which is sulfhydrated.

The fact that a very large number, perhaps the majority, of proteins are basally sulfhydrated and that sulfhydration alters protein function, suggests that sulfhydration is an important physiologic signal.

Physiologic roles of H2S

Blood vessels

The best known physiologic role for NO is as endothelial-derived relaxing factor (EDRF). EDRF activity was defined by the classic studies of Furchgott (Furchgott and Zawadzki 1980). Whereas norepinephrine constricts blood vessels by directly contracting the smooth muscle, Furchgott showed that the vasorelaxant action of acetylcholine is lost when the endothelial layer of blood vessels is removed. A substance with the properties of NO was released by endothelial tissue, and NO’s actions fit with the properties of EDRF. With the development of eNOS knockout mice, direct verification of the NO-EDRF hypothesis was possible. eNOS knockouts display elevated blood pressure and diminished EDRF activity in some vascular beds (Huang et al. 1995). CO also behaves like an EDRF. Like eNOS, HO-2 is localized to the endothelial layer of blood vessels whose endothelial-dependent relaxation is blocked by HO inhibitors (Zakhary et al. 1996).

Hydrogen sulfide has long been known to relax blood vessels (Zhao et al. 2001). Direct evidence bearing upon a potential EDRF activity for H2S awaited investigations employing CSE knockout mice (Yang et al. 2008). These mice develop age-dependent hypertension peaking at 12 weeks of age with blood pressures 18 mm Hg greater than control mice (Fig. 4a), similar to the hypertension of eNOS knockouts (Huang et al. 1995; Yang et al. 2008). Interestingly, the hypertension of CSE knockouts is age dependent. Blood pressure of heterozygotes resembles that of homozygotes at early ages, but by 10 weeks of age the homozygous mice display levels 10 mm Hg greater than the heterozygotes (Fig. 4a). The age-dependent hypertension parallels the ontogeny of CSE which attains peak levels 3 weeks after birth (Ishii et al. 2004).

Figure 4.

 (a) Age-dependent hypertensive phenotype of CSE male knockout mice. The hypertensive phenotype peaks at 12 weeks of age with blood pressures 18 mm Hg greater than wild-type control mice (+/+). Blood pressure of heterozygotes (−/+) resembles that of homozygous knockouts (−/−) at early ages, but by 10 weeks of age the homozygous knockout mice display levels 10 mm Hg greater than the heterozygotes (= 12). (b) Immunohistochemical localization of CSE to the endothelium of arterial blood vessels (black arrows) in wild-type mice. The signal is abolished in CSE knockout mice. (c) The contactile effects of phenylephrine on the mesenteric artery is the same in wild-type, heterozygous and homozygous knockout mice (= 15). (d) Methacholine relaxation of the mesenteric artery is reduced by about 80% in homozygous CSE knockout vessels and about 50% in heterozygotes (= 15). All results are means ± SEM. *< 0.05 vs. wild-type; #< 0.05 vs. heterozygote. Reproduced with permission from Yang et al. (2008).

Hydrogen sulfide satisfies the principal properties of an EDRF (Yang et al. 2008). It is selectively localized to the endothelial layer of blood vessels (Fig. 4b). In CSE knockout mesenteric arteries the contractile effects of phenylephrine (Fig. 4c), exerted upon α-adrenoceptors of vascular muscle, and the direct relaxing effects of NO donors are the same as in wild-type animals. H2S more potently relaxes mesenteric arteries of CSE knockouts than wild-type, indicating super-sensitivity associated with decreased endogenous H2S. By contrast, methacholine relaxation of the mesenteric artery is reduced by about 80% in homozygous CSE knockout vessels and about 50% in heterozygotes (Fig. 4d). The methacholine relaxation reflects EDRF activity, as it is abolished by removal of the endothelium.

Thus, most EDRF activity of the mesenteric artery can be attributed to H2S. Muscarinic cholinergic treatment of blood vessels activates eNOS to produce NO. Similarly, methacholine treatment of endothelial cells triples H2S levels which are abolished by depletion of CSE utilizing RNA interference.

If the great majority of mouse mesenteric artery EDRF activity is attributable to H2S, what is the role of NO? NO is well established as an EDRF in numerous vascular beds, but EDRF activity in many vessels is only partially diminished by NOS inhibitors and in eNOS knockouts (Brandes et al. 2000; Félétou and Vanhoutte 2007). EDRF activity attributable to NO is most prominent in large vessels such as the aorta, while in resistance vessels that regulate blood pressure more directly, NO’s effects are less evident (Brandes et al. 2000). Differences among diverse vascular beds and species variations may account for discrepant observations. Determining the relative roles of NO, CO and H2S in mediating physiologic EDRF activity will require side-by-side comparisons of HO-2, eNOS and CSE knockout mice as well as studies in multiple species.

How does H2S relax blood vessels? NO is well established to act by stimulating sGC. CO does elevate cyclic GMP levels. However, endogenous CO-induced vasodilation occurs via a cyclic GMP-independent mechanism (Naik and Walker 2003). It appears likely that CO acts via the large-conductance calcium-activated potassium channels (BKCa). Thus, inhibitors of BKCa channels block endogenous CO-elicited vasodilation (Naik and Walker 2003). Moreover, HO inhibitors reduce BKCa channel activity in several vascular beds (Kaide et al. 2001; Zhang et al. 2001; Li et al. 2008). Inhibitors of sGC do not influence CO-induced BKCa channel activation (Kaide et al. 2001; Xi et al. 2004). Interestingly, the actions of CO on BKCa may involve binding to heme, analogous to NO binding to heme in sGC. Thus, the α-subunit of BKCa contains a heme-binding pocket, and binding of heme to the channel inhibits its activity, CO binds to channel-associated heme to elicit channel activation (Jaggar et al. 2005).

A major component of EDRF activity involves hyperpolarization, a phenomenon that is not elicited by sGC. Thus, to fully explicate EDRF, investigators have sought an endothelial-derived hyperpolarizing factor (EDHF). Compounds postulated to mediate EDHF activity include prostacyclin generated from arachidonic acid by cyclo-oxygenase (EC, epoxyeicosatrenoic acids generated from arachidonic acid by cytochrome P450 epoxygenase (EC, hydrogen peroxide, potassium ions, C-type natriuretic peptide, electrical coupling through myoendothelial junctions mediated by connexins, and NO itself (reviewed in Bellian et al. 2008; Luksha et al. 2009). For none of these substances has definitive evidence been provided employing genetic mutant animals provided.

In mouse mesenteric artery and aorta, inhibition of eNOS and cyclooxygenase reduces cholinergic EDRF activity only about 20% (Mustafa et al. unpublished observation). The remaining 80% of cholinergic relaxation reflects pronounced hyperpolarization with resting membrane potentials approximating the potassium equilibrium potential. This hyperpolarization is virtually abolished in CSE homozygous knockout mice.

Endothelial-derived hyperpolarizing factor activity reflects opening of potassium channels (Bellian et al. 2008; Luksha et al. 2009). The vasorelaxant effects of H2S are blocked by inhibitors of the ATP-sensitive potassium channel (KATP) (Zhao et al. 2001; Zhao and Wang 2002; Cheng et al. 2004). Glibenclamide, a potent and selective inhibitor of KATP, reduces cholinergic hyperpolarization of the mesenteric artery smooth muscle cells by about 70% (Mustafa et al. unpublished observation). By contrast, glibenclamide doesn’t affect relaxation elicited by NO donors.

How does H2S stimulate KATP? KATP possesses nine cysteines with C43, that lies close to the surface, selectively influenced by oxidative insults. KATP is sulfhydrated with the sulfhydration abolished by mutations of C43 (Mustafa et al. unpublished observation). Thus, H2S vasorelaxation reflects hyperpolarization mediated by the opening of KATP channels via their sulfhydration at C43. KATP is physiologically activated by binding of the phospholipid phosphatidylinositol (4,5)-bisphosphate (PIP2) (Baukrowitz et al. 1998; Shyng and Nichols 1998). PIP2 binding to KATP is abolished in cells lacking CSE or containing catalytically inactive enzyme, and H2S donors markedly stimulate PIP2-KATP binding (Mustafa et al. unpublished observation). The PIP2-KATP binding involves the sulfhydrated C43, as binding is markedly reduced in KATP-C43S mutants.

As physiologic vasodilation is thought to be determined largely by EDHF, the evidence that EDHF activity is predominantly determined by H2S fits with a major role for H2S as an EDRF/EDHF.


There is abundant literature on potential roles of H2S in inflammation. Some studies indicate that endogenous H2S is anti-inflammatory. Thus, one of the earliest events in inflammation is adherence of leukocytes to vascular endothelium and their subsequent migration into underlying tissue. The CSE inhibitor β-CNA markedly increases leukocyte-endothelial adherence as well as carrageenan-induced leukocyte infiltration and paw edema (Zanardo et al. 2006). H2S donors display anti-inflammatory effects, inhibiting leukocyte-endothelium bonding and reducing carrageenan-induced paw edema. H2S donors reduce visceral pain in a colorectal distension model (Distrutti et al. 2006a,b) and diminish colitis in rats (Fiorucci et al. 2007).

By contrast, some studies indicate a pro-inflammatory action of H2S. H2S levels and CSE expression are increased in several models of inflammation, and the CSE inhibitor PAG reduces inflammation in some of these models (Mok et al. 2004; Li et al. 2005; Bhatia et al. 2005a,b; Collin et al. 2005). In rodent sepsis, H2S increases levels of substance P in the lung (Zhang et al. 2007). Also, H2S induces the formation of pro-inflammatory cytokines and chemokines by up-regulating nuclear factor kappa-light-chain-enhancer of activated B cells (Zhi et al. 2007).

Despite discrepancies, the evidence that H2S is anti-inflammatory is sufficient that efforts are under way to attack inflammatory diseases with H2S releasing drugs. For instance, diclofenac derivatives that release H2S have been developed for use as anti-inflammatory drugs (reviewed in Wallace 2007). An H2S-releasing mesalamine derivative, ATB-429, displays analgesic and anti-inflammatory effects and has been effective in models of inflammatory bowel disease (Distrutti et al. 2006a,b).

Because of its chemical activity and abundant production from bacteria in the colon, there has been speculation that bacterially generated H2S mediates the pathophysiology of ulcerative colitis (Pitcher and Cummings 1996). Short-chain fatty acids, especially butyrate, are thought to be important in maintaining normal colonic mucosal function (Cummings 1981). Butyrate oxidation provides about 70% of colonic energy whereas the small intestine preferentially utilizes glucose and glutamine (Watford et al. 1979; Roediger 1980, 1982; Ardawi and Newsholme 1985; Cummings et al. 1987). H2S donors interfere with colonic butyrate metabolism (Christl et al. 1996). It is conceivable that the therapeutic effects of 5-aminosalicylate in ulcerative colitis reflect influences upon H2S, as patients treated with the drug display substantially decreased levels of sulfide in their feces (Pitcher et al. 1995).

Nervous system

The journey to establishing a neural role for any substance commences with ascertaining its localization. In the 1960s histochemical fluorescent techniques that visualize biogenic amines such as serotonin, dopamine and norepinephrine, permitted mapping their neuronal pathways with major functional insights (Carlsson 1987). Immunohistochemistry for a wide range of neuropeptides and neurotransmitter related enzymes established these substances as neurotransmitter candidates (Jones and Hartman 1978). Selective neuronal localizations of nNOS (Bredt et al. 1990) and HO-2 (Verma et al. 1993) have helped to characterize neurotransmitter properties for NO and CO, respectively. For H2S, one would hope to localize the biosynthetic enzymes by immunohistochemistry. Relatively little investigation has yet been reported. Szurszewski and colleagues (Linden et al. 2008) conducted immunohistochemical studies of both CSE and CBS. For CSE, neuronal localizations were evident in the myenteric plexus of neurons in the small intestine suggesting that like NO and CO, H2S might be a non-adrenergic non-cholinergic neurotransmitter. In the brain, where CSE levels are low, localizations were predominantly in white matter. CBS immunohistochemistry in the brain also revealed prominent white matter localizations with negligible neuronal staining. However, caution is warranted in interpreting these findings. The publication did not display western blots to clarify whether the antibody reacted with substances other than CBS or CSE. A principal control was pre-absorption with the immunizing antigen which does not rule out non-specific staining. Further studies employing CBS and CSE knockout mice as controls would be useful.

Influences of H2S upon neuronal activity in the brain have been explored extensively by Kimura and colleagues (Kimura et al. 2005). This group noted that physiologic concentrations of H2S enhance long-term potentiation (LTP). Sodium hydrogen sulfide (NaHS) applications and weak tetanic stimulation of rat hippocampal slices alone did not elicit LTP, while the simultaneous application of both led to robust LTP (Abe and Kimura 1996). The effect of H2S on LTP was abolished by NMDA antagonists. Interestingly, NO and CO also induce LTP, but do so even when NMDA receptors are blocked (Zhuo et al. 1993). NMDA receptors possess reactive cysteines and are known to be nitrosylated with resulting channel blockade (Lei et al. 1992; Choi et al. 2000). Conceivably H2S regulates NMDA transmission by sulfhydrating NMDA receptors.

Besides its actions upon neurons, H2S also appears to influence astrocytes (Nagai et al. 2004). H2S donors elicit calcium waves in astrocytes and increase intracellular levels of calcium. The increased intracellular calcium occurs rapidly following H2S exposure and decays slowly, whereas the oscillations of calcium decay rapidly. Effects of H2S donors are evident both in primary cultures of astrocytes and in glia within hippocampal slices. The increased intracellular calcium in astrocytes following H2S administration reflects calcium entry, as it is suppressed in calcium-free media and is associated with a direct influx of calcium similar to that elicited by calcium ionophores. The type of calcium channel involved has not yet been established.

Hydrogen sulfide may also serve as a neuroprotectant. Glutamate neurotoxicity in brain cultures involves, at least in part, inhibition of cystine uptake (Tan et al. 2001). The cystine/glutamate anti-porter couples influx of cystine with efflux of glutamate. This process is blocked by high concentrations of exogenous glutamate which are cytotoxic via a process designated oxytosis (Tan et al. 2001). How does H2S act in this model? Glutamate reduces levels of intracellular glutathione, and H2S increases them both in untreated and in glutamate-exposed preparations (Kimura and Kimura 2004). In support of this model, buthionine sulfoximine (Griffith 1982), which inhibits γ-glutmaylcysteine synthase (EC, a rate limiting enzyme in glutathioine biosynthesis, prevents the H2S-elicited stimulation of glutathione levels and cell survival. H2S elicits augmented glutathione by stimulating cystine entry into cells, reversing the inhibition of cystine transport by glutamate (Kimura and Kimura 2004).

Interestingly, the first recognized sign of CBS deficiency in humans is mental retardation (Mudd et al. 1999). CBS deficient patients also suffer from seizures, abnormal electroencephalograms, extrapyramidal disturbances and psychiatric disorders (Mudd et al. 1985; Abott et al. 1987). The role of H2S in these disturbances is yet to be examined. Another interesting observation is that CBS is enriched in the brains of Down’s patients (Ichinohe et al. 2005). This is not surprising since the CBS gene is located on chromosome 21. However, the role of CBS and H2S in the mental retardation found in Down syndrome is also yet to be examined.

The future

Because H2S is a chemically reactive substance with toxic actions, its influences upon various tissues have been well characterized for many decades. However, translating pharmacologic effects into evidence for endogenous, physiologic function is a major challenge. Direct evidence that H2S is physiologically generated by the enzymes CSE and CBS is very recent. Mice with targeted deletion of these two enzymes have been valuable tools in this endeavor, but many basic studies remain to be carried out. Localizing CBS and CSE immunohistochemically in all organs of the body, especially the brain, is a seemingly simple minded task but of immense importance. Phenotypic characterization of the CBS and CSE mutant mice is critical. Using the mice to establish roles for H2S in nervous system function should be reasonably straightforward. Behavioral analysis, monitoring neurotransmission in various pathways, exploring synaptic plasticity in models such as LTP and long-term depression, are all approaches that are today the bread and butter of neuroscience. Regardless of what is found in the future, it is likely that H2S will join NO and CO as an important gasotransmitter. In the vascular system, evidence is strong for a major role of H2S as a physiologic vasodilator. S-sulfhydration as an important mode of post-translational protein modification is established. As H2S is generated physiologically in almost all organs of the body, it is likely that functions in diverse tissues, especially the nervous system, will emerge in the not-too-distant future.


This work was supported by the National Institutes of Health Medical Scientist Training Program Award (T32 GM007309) to M.M.G., and US Public Health Service Grants (MH018501 and DA000226) and Research Scientist Award (DA00074) to S.H.S.