Biogenesis of reactive sulfur species for signaling by hydrogen sulfide oxidation pathways

Abstract

The chemical species involved in H₂S signaling remain elusive despite the profound and pleiotropic physiological effects elicited by this molecule. The dominant candidate mechanism for sulfide signaling is persulfidation of target proteins. However, the relatively poor reactivity of H₂S toward oxidized thiols, such as disulfides, the low concentration of disulfides in the reducing milieu of the cell and the low steady-state concentration of H₂S raise questions about the plausibility of persulfide formation via reaction between an oxidized thiol and a sulfide anion or a reduced thiol and oxidized hydrogen disulfide. In contrast, sulfide oxidation pathways, considered to be primarily mechanisms for disposing of excess sulfide, generate a series of reactive sulfur species, including persulfides, polysulfides and thiosulfate, that could modify target proteins. We posit that sulfide oxidation pathways mediate sulfide signaling and that sulfurtransferases ensure target specificity.

Sulfide chemistry is intimately interwoven with the emergence of life on this planet. It is posited that the interaction between sulfide spewing from alkaline submarine hydrothermal vents and the acidic iron-containing waters of the Hadean ocean led to formation of catalytic colloidal iron-sulfur membranes that promoted the synthesis of the first organic compounds¹. Fossils of sulfur-metabolizing microbes dating back almost 3.5 billion years provide some of the earliest evidence for a sulfur-based chemolithoautotrophic lifestyle². Indeed, the composition of sulfur isotopes in biogenic sedimentary sulfides provides a useful record of biospheric oxygenation and the coevolution of life and the environment³. Sulfide continued to influence evolution in later eons, and it is postulated that the presence of a metastable sulfidic oceanic zone limited metazoan colonization of the continental shelf⁴, while oceanic sulfide toxicity and hydrogen sulfide (H₂S) emissions into the atmosphere have been implicated as important drivers of the Permian-Triassic mass extinction⁵. As sulfur is the sixth most abundant element in the microbial biomass⁶, its metabolism is critically important to the global biogeochemical sulfur cycle.

What is the key to the versatility of sulfur in metabolism? The answer lies in its ability to cycle through a variety of biologically relevant oxidation states ranging from −2, as in H₂S, to +6, as in sulfate (SO₄²⁻) (Fig. 1). The higher valence states of sulfur are obtained through oxidation of sulfide or thiols to compounds such as thiyl radical (RS^•, −1, where the number refers to the formal oxidation state of the sulfur atom(s)), hydropersulfide (RS-S⁻, −1, −1), disulfide (RSSR, −1, −1) and sulfenic acid (RSOH, 0). In parallel to the usage for reactive oxygen and nitrogen species, the term ‘reactive sulfur species’ (RSS) is used here to refer collectively to reactive sulfur chemotypes (both organic and inorganic) that, under physiological conditions, can react with, oxidize or reduce other molecules^7–9. Oxidized sulfur species can be catenated as in hydropolysulfides (RS-S-S⁻, −1, 0, −1), have interchalcogen bonds as in sulfite (SO₃²⁻, +4) and sulfate (SO₄²⁻, +6), or have both as in thiosulfate (S₂O₃²⁻, −1, +5). For simplicity, hydropersulfide and hydropolysulfides will be referred to as persulfides and polysulfides, respectively.

Figure 1. Structures of some biologically relevant RSS chemotypes.

The use of the term ‘sulfane sulfur’ in the literature is sometimes confusing. Here, the IUPAC nomenclature for sulfane sulfur—sulfur-bonded sulfur—is used. The red (−2), green (−1), yellow (0), blue (+4), purple (+5) and orange (+6) rectangles are used to designate the valence states of sulfur as specified. In molecules containing catenated sulfurs, n ≥ 2 and R ≠ H.

The cellular capacity for redox signaling via different site-specific cysteine modifications allows an oxidant signal to be transmuted into a biological response. However, unlike with reactive oxygen and nitrogen species, whose roles in cellular signaling and in varied physiological processes have been extensively studied, the role of RSS in redox homeostasis, mammalian metabolism and signaling has only recently begun to attract attention^7–10. Central to the growing interest in RSS biochemistry is the realization both of the mammalian capacity to produce and oxidize H₂S and of the ability of H₂S, or a downstream product, to elicit varied and profound physiological effects¹¹. Some important missing pieces in this emerging story are the mechanisms by which sulfur signals are conveyed and whether H₂S or an RSS derived from it (or both) constitute the signaling molecule(s). In this article, we propose that the canonical mitochondrial¹² and the newly described heme-dependent¹³ sulfide oxidation pathways are important sources of RSS and that enzyme-catalyzed transpersulfidation reactions are an important mechanism for ensuring target specificity of reversible post-translational persulfide modification.

Properties of H₂S

To evaluate the varied roles ascribed to H₂S, such as its reactivity toward amino acid side chains to effect protein modification and its antioxidant properties, it is important to first examine its chemical attributes. A colorless gas with the smell of rotten eggs, H₂S is highly soluble in water (~80 mM at 37 °C) and can permeate lipid bilayers with facility^14,15. H₂S is a weak diprotic acid with pK_a1 and pK_a2 values of 7 and 12–15 (ref. 16), respectively, and thus exists predominantly (72%) as the sulfide anion at the physiological pH of 7.4. By comparison, the pK_a values for cysteine and glutathione, two small-molecule thiols, are 8.3 and 8.8, respectively. H₂S is used in this article to refer collectively to all three species that exist in solution: the diprotonated (H₂S), monoanion (HS⁻) and dianion (S²⁻) forms.

H₂S is a relatively stable sulfur species with a half-life on the order of minutes¹⁰. H₂S shows low reactivity toward disulfides such as glutathione disulfide (GSSG) and cystine, and its rate of disappearance in buffer at pH 7.4 is comparable in the absence or presence of these disulfides¹⁷. H₂S shows slightly higher reactivity toward the disulfide cystamine¹⁷. Careful studies on the kinetics of H₂S reaction with disulfides versus cysteine sulfenic acid or cysteine nitrosothiols are needed to evaluate their kinetic competence in the context of persulfide modification of proteins. A mixture of unstable intermediates and reaction products are formed during the uncatalyzed oxidation of H₂S in aerobic solutions, including elemental sulfur (S₈), polysulfanes (S_n²⁻, n = 2–7), sulfite and thiosulfate¹⁸.

The bimolecular rate constants for the reaction of HS⁻ with two-electron oxidants such as hydrogen peroxide (H₂O₂, 0.73 M⁻¹ s⁻¹), peroxynitrite (4.8 × 10³ M⁻¹ s⁻¹) and hypochlorite (8 × 10⁷ M⁻¹ s⁻¹) at pH 7.4 and 37 °C reveal a wide range of reactivity (Table 1) and are comparable to the rate constants reported for cysteine and glutathione¹⁹. However, the pH-independent rate constants for HS⁻, a measure of its intrinsic reactivity, are lower by 3-fold (with hypochlorous acid), 6-fold (with peroxynitrous acid) and 25-fold (with H₂O₂) as compared to cysteine and still lower than for glutathione. This difference in intrinsic reactivity is likely due to the absence of an inductive effect on sulfide and to the lower Brønsted basicity of HS⁻ as compared to the thiolates of cysteine and glutathione¹⁹. The modest bimolecular rate constant for the uncatalyzed oxidation of H₂S by H₂O₂, and the low intracellular concentration of H₂S (discussed below), make it unlikely that this reaction is physiologically significant, unless methods exist for concentrating H₂S and oxidants (for example, H₂O₂ or peroxynitrite) in certain cellular locales.

Table 1.

Reactivity of H₂S and its oxidation products

Oxidant	k (M−1 s−1)a	Temperature (°C)
H2S + O2•− → H2O2 + S•−	2.0 × 102	37
HS− + H2O2 → HSOH + OH−	0.73	37
HS− + ONOOH → HSOH + NO2−	4.8 × 103	37
HS− + HOCl → HSCl + OH−	8.0 × 107	37
HSOH + HS−→ HSSH + OH−	1.0 × 105	37
HS− + S•− → HSS•2−	5.4 × 109	37
HSS•2− → HS− + S•−	5.3 × 105 s−1 b	37
HSS•2− + O2 → HSS− + O2•−	4.0 × 108	37

^aH₂S, which is fully reduced, shows a spectrum of reactivity with oxidants ranging from sluggish (with H₂O₂) to high (with HOCl). The initial products are either one (S^•−) or two electron (for example, HSOH) more oxidized than H₂S and represent reactive sulfur species, which can undergo further reactions. The rate constants for the reaction of GSH and cysteine, where available, show that they are very similar to those for H₂S at the same pH and temperature. The values in this table are from ref. 19 and denote bimolecular rate constants, with the one exception noted.

^bThe rate constant for this unimolecular reaction has units of s⁻¹.

In comparison to glutathione and cysteine, which function as redox buffers in the intracellular compartment, H₂S is predicted to be a minor antioxidant. The standard two-electron redox potential for the H₂S/S⁰ couple is estimated to be −280 mV (versus the normal hydrogen electrode) at pH 7.0 (based on the value of +140 mV reported in acidic solution)⁷ and is similar to those for the glutathione disulfide/glutathione (E°′ = −262 mV) and the cystine/cysteine (E°′ = −245 mV) redox couples²⁰. However, because the intracellular steady-state concentration of H₂S (~10–30 nM)^17,21 in mammals is several orders of magnitude lower than those of cysteine (~0.1–1 mM)²² or glutathione (1–10 mM)²², it is unlikely that H₂S has a quantitatively significant role in oxidant defense except perhaps in specific cellular niches.

The sulfhydryl radical (HS^•) produced by one-electron oxidation of sulfide is a strong oxidant and, once generated, can be involved in a series of radical chain reactions (Table 1). The standard reduction potential for the SH^•/HS⁻ redox couple is +920 mV²³. HS^• is reactive and can react with HS⁻ to form the hydrodisulfide radical anion (HSSH^•−), which can further react with O₂ to generate HSS⁻ and O₂^•−. Reaction of the HS^• with O₂ gives the sulfur dioxide radical anion (SO₂^•−)²³.

Two mechanisms that have been discussed in the context of sulfide-based signaling are persulfidation forming protein persulfides and interaction of sulfide with metalloproteins and are described below.

Metal sulfide complexation

The role of metals as sulfide carriers has been known for a while. For example, ferric ion in hemoglobin Hb1 in Lucina pectinata²⁴, a clam found in sulfide-rich mangroves, and zinc ions in with sulfide-oxidizing bacteria. H₂S binds to the open coordination site in ferric hemoglobin to generate low-spin six-coordinate ferric sulfide heme, in equilibrium with the ferrous sulfide radical tautomer (Fig. 2a). The fate of ferric sulfide hemoglobin is determined by various factors including sulfide concentration and the dielectric constant of the distal heme pocket, with a polar environment favoring heme reduction concomitant with sulfide oxidation²⁶. Sulfide can also react with ferrous-oxy heme, leading, in a series of steps, to sulfur addition to the porphyrin ring, forming sulfhemoglobin²⁷. The latter is an irreversible modification also seen with myoglobin²⁸ and is therefore unlikely to be useful in a signaling context.

Figure 2. Models for H₂S interaction with heme proteins.

(a) Binding of sulfide anion to ferric heme results in the formation of ferric sulfide complex, which, depending on the polarity of the distal heme pocket, could lead to heme iron reduction and formation of the sulfur radical. Depending on the heme protein, either H₂S or HS⁻ can bind ferric heme. (b) In principle, H₂S can also bind ferrous hemes, as seen with model porphyrinate complexes.

In principle, sulfide can bind to an open metal-coordination site or displace a weak ligand, such as water. In fact, the primary basis of H₂S toxicity is via metal coordination, that is through its binding to the binuclear Cu–heme iron complex in cytochrome c oxidase, resulting in reversible inhibition of this terminal station in the electron transfer chain where O is reduced to water²⁹. Sulfide anion coordination to model ferrous iron porphyrinate complexes (Fig. 2b) has been characterized³⁰. The corresponding ferric compounds are rapidly reduced by sulfide, suggesting that heme groups in polar protein environments might be susceptible to similar chemistry following sulfide binding. It has been speculated that the binding of sulfide to neuroglobin, a heme protein relatively abundant in neurons and retina, plays a protective role under conditions of oxygen deprivation, such as stroke, when sulfide concentrations reportedly increase³¹. The physiological relevance of sulfide binding to neuroglobin and to other heme proteins in a protective or signaling capacity remains to be assessed.

In order for sulfide coordination to a metallocenter to serve in a signaling pathway leading to a cellular response, it must elicit an effector function either within the same protein or with a downstream partner protein. A specific receptor for sulfide is not known, unlike with •NO, for which soluble guanylate cyclase serves as a heme protein sensor, activating production of the second messenger cGMP³². It remains to be seen whether sulfide binding can modulate the functions of zinc, mononuclear iron, copper or heme-containing metalloproteins in the context of signal transduction.

Protein modification by persulfidation

Persulfidation of proteins

Cavallini and co-workers demonstrated in 1970 that persulfides form upon addition of sodium sulfide to proteins containing disulfide bonds, such as insulin and RNase³³. Studies by Massey and co-workers on xanthine oxidase and aldehyde oxidase^34,35 are often cited as evidence for the importance of stabilized persulfides to enzymatic activity. However, spectroscopic and structural studies have since demonstrated that the labile sulfur in these enzymes is a sulfido ligand to molybdenum and not a persulfide^36–38. Persulfidation at a single cysteine residue in superoxide dismutase was established by mass spectrometry and shown to stabilize the enzyme against oxidation-induced aggregation without affecting catalytic activity³⁹. The opposite effect, inhibition following persulfidation, has been observed with a number of erythrocytic enzymes, including glucose 6-phosphate dehydrogenase and adenylate kinase⁴⁰, and with protein tyrosine phosphatase 1B (PTP1B)⁴¹. In PTP1B, the active site cysteine was specifically modified, lending confidence about the potential relevance of the persulfide modification as a regulatory mechanism. In vivo, persulfidation reportedly enhances the catalytic activity of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and the polymerization of actin⁴². Persulfidation of parkin, an E3 ubiquitin ligase, increases its activity⁴³, and this modification is depleted in brain of individuals with Parkinson’s disease.

Despite the handful of examples, whether persulfidation of proteins by H₂S is important in a signaling context and by what mechanism this modification is introduced remain open questions. As H₂S cannot react directly with thiols and its rate of auto-oxidation might be too slow to be biologically relevant (Table 1)¹⁶, reactions involving the addition of protein thiolates to either HSOH or HSSH to give the persulfide modification (Fig. 3a) might not be physiologically relevant. In contrast, proteins with reactive cysteines that have become oxidized might be susceptible to persulfidation by sulfide. A subset of all protein cysteines are reactive, and these generally have low pK_a values. Reactive cysteines are often found in protein microenvironments with a positive electrostatic potential that stabilizes the thiolate anion⁴⁴. Their reactivity makes them susceptible to oxidative modifications including disulfide formation, S-glutathiolation, sulfenic acid formation or S-nitrosylation that lead to altered protein structure and/or function (Fig. 3b). In proteins containing a disulfide bond, a nucleophilic attack by the sulfide anion would yield the persulfide modification, although its lifetime could be short owing to elimination by the resolving cysteine. Considering the low concentration of disulfides in the reducing milieu of the cell and the modest reactivity of sulfide toward disulfides¹⁷, the physiological relevance of this persulfidation mechanism remains to be assessed. In the absence of disulfide bonds, an alternative route to cysteine activation for sulfide attack is necessary; possibilities include glutathiolated cysteine (Cys-SSG), cysteine sulfenic acid (Cys-SOH) and S-nitrosylated cysteine^45,46. The mechanisms for persulfidation of proteins shown in Figure 3b potentially connect sulfide-based signaling to pathways involving oxidative or nitrosative stress. Persulfidation is a reversible modification that can be removed by reducing agents such as glutathione⁴⁷, proteins such as thioredoxin⁴⁸ or glutaredoxin or via re-formation of a disulfide bond initiated by nucleophilic attack by a resolving cysteine (Fig. 3c).

Figure 3. Potential mechanisms for persulfidation and the resolution of this modification.

(a) H₂S is relatively stable and is oxidized slowly by H₂O₂ to give HSOH. Either HSOH or HSSH formed from HSOH in the presence of a second equivalent of H₂S can be attacked by a reactive cysteine on a protein to generate the persulfide modification. The bimolecular rate constants for HSOH and HSSH formation in solution at pH 7.4 and 37 °C are noted. (b) Persulfidation can result from the nucleophilic attack of a sulfide anion on an oxidized protein thiol (such as disulfide, mixed disulfide, cysteine sulfenic acid or S-nitrosylated cysteine). (c) Persulfide modifications on proteins are reversible and, unless sequestered, labile. They can be removed via persulfide interchange reactions involving glutathione (GSH), thioredoxin (Trx) or a cysteine on the same or a different protein. In all cases, the product is ultimately H₂S, formed upon reduction of the persulfide moiety by either a second mole of GSH or the NADPH–thioredoxin reductase system.

An obvious limitation of the solution-based persulfidation mechanisms described in Figure 3a,b is their lack of specificity. In principle any reactive cysteine on a protein could attack oxidized sulfide, or reactive cysteines that have themselves become oxidized can be attacked by sulfide. Furthermore, multiple cysteines can be modified on the same protein, as reported for serine and homoserine dehydratases, alcohol dehydrogenase and 3-hydroxybutyrate dehydrogenase^49,50. Unregulated persulfidation would lead to undesired changes in protein function and could be averted by the low steady-state concentration of sulfide and/or, as discussed later, by enzyme-catalyzed transpersulfidation⁵¹.

Protein modification by polysulfides

An alternative hypothesis for sulfide-based post-translational protein modification is that H₂S is not the sulfurating species at all; rather, the S-sulfurating agent is an oxidation product of sulfide—that is, a polysulfide—that is more reactive than H₂S. In fact, it has been argued that the varied effects of H₂S can be attributed to polysulfide species that contaminate commercially available sulfide salts⁵², and polysulfides show substantially lower IC₅₀ values than H₂S when the direct comparison is made⁵³. Polysulfanes, such as elemental sulfur, can sulfurate cysteine thiols^49,50. However, the limited solubility of elemental sulfur and the relative instability of polysulfides pose obvious challenges for their handling within the cell. Furthermore, the potential source of elemental sulfur and/or polysulfide species in mammals is unclear. In bacteria, polysulfides and elemental sulfane sulfur can be produced during sulfide oxidation by sulfide quinone oxidoreductase (SQR)⁵⁴. In contrast, mammalian SQRs do not catalyze multiple rounds of H₂S oxidation and polymerization before releasing catenated sulfur products into solution. Instead, the oxidized sulfur is transferred to a small-molecule acceptor such as glutathione in each round of the catalytic cycle, forming a persulfide (rather than polysulfide) product^12,47. The concentration of free polysulfide is reportedly ~25 nmol/g of wet mouse brain tissue⁵³ and ~0.3 μM in mouse blood⁵⁵.

Polysulfides can modulate protein functions such as activation of transient receptor potential (TRP) A1 channels⁵³, inhibition of the lipid phosphatase PTEN⁵² and activation of the transcription factor Nrf2 by modification of Keap1 (ref. 56). The reactivity of polysulfides raises questions about how specificity in protein modification is achieved, as in principle a reactive cysteine on a protein can attack at multiple sites along the polysulfide chain, generating a heterogeneous mix of products (Fig. 4a). Furthermore, the resultant product mixture can undergo further reactions with either the same or a different protein or with small-molecule thiols, forming sulfur-bridged homo- or heterodimeric species, respectively (Fig. 4b). Although the effects of polysulfide exposure on cellular responses (for example, Ca²⁺ influx by TRPA1 channel activation)⁵³ have been reported, the nature of the resulting modification has not been directly established. With PTEN, exposure to polysulfides resulted in an intramolecular disulfide bond that was identified by mass spectrometric analysis⁵², indicating loss of the initial per- or polysulfide modification, probably by the mechanism shown in Figure 4b. Polysulfide exposure of cells resulted in homo- and heterodimerization of Keap1 (ref. 56). An inter- or intramolecular disulfide link could be a signature of an initial and transient polysulfide modification.

Figure 4. Potential complexity associated with uncatalyzed protein modification by polysulfides.

(a) A reactive thiol on a protein (dark blue sphere) could attack one of several sulfurs in a polysulfide chain generating a series of protein modifications (1–5) and eliminating H₂S or polysulfides with varying number of sulfur atoms, which are omitted for clarity of presentation. This reaction complexity could be averted in an enzyme-catalyzed persulfidation reaction in which nucleophilic attack on a specific sulfur atom in the polysulfide chain is promoted. (b) Further reactions with persulfide- or polysulfide-modified proteins with a cysteine on the same or a different (light blue sphere) protein (or a small-molecule thiol) could result in homo- or heterodimerization in which the protein subunits are linked via bridging sulfur atoms. Alternatively, a vicinal cysteine residue on the modified protein could give rise to an intramolecular sulfur linkage.

Transpersulfidation of proteins

A number of sulfur-containing cofactors and modified thionucleosides obtain their sulfur atom by persulfide transfer⁵⁷—that is, via transpersulfidation. The pK_a of persulfides is ~1–2 units lower than for the corresponding thiols⁵⁸, and the terminal sulfur can act as a nucleophile or an electrophile in the unprotonated and protonated states, respectively⁵⁹. Persulfides are reactive and are inherently unstable and disproportionate to RSH and S° in solution. In light of their dual reactivity and instability, the report that cellular and tissue concentrations of small-molecule persulfides are of the same order of magnitude as for thiols such cysteine⁶⁰ is surprising. Sulfurtransferases are found in archaea, bacteria and eukaryotes and contain the rhodanese homology domain fold with an α/β topology (Fig. 5a), which can be present in a single copy, in tandem repeats or fused with other proteins⁶¹. Rhodanese homology domain proteins have consensus sequence motifs at the N and C termini and a catalytic cysteine embedded in an active site loop. The rhodanese homology domain proteins are often found encoded in multiple copies in genomes and can function as persulfide carriers, as in the biosynthesis of 4-thiouridine⁶² and molybdopterin⁶³. A sulfur relay system, in which a persulfide group is shuttled from a cysteine desulfurase through a succession of carrier proteins, is employed for synthesis of the 2-thiouridine modification in tRNA^64,65.

Figure 5. Rhodanese homology domain proteins and persulfide relay system for transpersulfidation.

(a) Structures of three rhodanese homology domain proteins in which the active C-terminal domain is shown in red (top, S. cerevisiae putative single-domain thiosulfate sulfurtransferase (TST) YOR285, PDB 3D1P), green (middle, bovine rhodanese, PDB 1RHD) or blue (bottom, human 3-mercaptopyruvate sulfurtransferase (MST), PDB 4JGT), and the inactive N-terminal domain, when present, is shown in yellow. The active site cysteines are in ball representation, and a persulfide intermediate is seen in the rhodanese structure. (b) Sulfurtransferases such MST and TSTs (for example, rhodanese and TSTD1) and sulfide quinone oxidoreductase (SQR) accept sulfur atoms from their respective substrates (3-mercaptopyruvate for MST, thiosulfate or GSSH for TSTs and H₂S for SQR) and form a persulfide intermediate. The outer sulfur can be transferred to a protein or small-molecule acceptor, resulting in a persulfide product. (c) Specificity in transpersulfidation can be achieved by enzyme-catalyzed transfer of the persulfide group. In the example shown here, either GSSH or thiosulfate transfers the outer sulfur to TST, forming a persulfide intermediate, which can be transferred to protein targets (1, 2) or to an intermediate carrier such as glutaredoxin (Grx, 3), which, in turn, transfers the persulfide to a target protein (4). Because two-cysteine-containing thiroredoxins and Grxs have vicinal resolving cysteines, the resulting persulfide modification on them is expected to be short-lived. In contrast, in carriers with a single active site cysteine, such as a subset of Grxs, the lifetime of the persulfide modification will be longer.

In mammals, sulfurtransferases involved in sulfide synthesis and oxidation form active site persulfides (Fig. 5b). These include mercaptopyruvate sulfurtransferase, which transfers the sulfur from 3-mercaptopyruvate to an active site cysteine, forming a persulfide intermediate^48,66. Human rhodanese uses GSSH as a sulfur donor, forming an active site persulfide⁴⁷. It can also employ thiosulfate as a sulfur donor, albeit less efficiently than GSSH by a factor of >2 ×10⁵-fold⁴⁷. A third human rhodanese domain–containing sulfurtransferase, TSTD1, utilizes thiosulfate as a sulfur donor, although the kinetics of this reaction have only been characterized at an alkaline pH⁶⁷. Whereas the majority of sulfurtransferases are found in the cytoplasm, rhodanese is localized to mitochondria and mercaptopyruvate sulfurtransferase is found in both the cytoplasm and mitochondria. Finally, the mitochondrial enzyme SQR, involved in the sulfide oxidation pathway (discussed below), forms a persulfide intermediate^47,68. Enzymes involved in H₂S biogenesis, cystathionine β-synthase and γ-cystathionase, can synthesize cysteine persulfide from cystine⁶⁰.

The persulfide intermediate formed in these enzymes can be subsequently transferred to a protein such as glutaredoxin or to a small-molecule acceptor such as cysteine, hypotaurine or glutathione^{33,47,48,66,68}. In principle, sulfurtransferases could modify target proteins either directly or indirectly via a secondary carrier, which in turn transfers the persulfide to a client protein (Fig. 5c). The use of secondary carriers such as glutaredoxin, thioredoxin or other rhodanese homology domain proteins would permit a signaling pathway to fan out from a single persulfide generating enzyme and also achieve targeting specificity. Persulfides transferred to two-cysteine thioredoxins and glutaredoxins are likely to be short-lived owing to the presence of resolving cysteines, setting up a competition between H₂S elimination (Fig. 3c) and transpersulfidation. Evidence in support of transpersulfidation as a mechanism for sulfur transfer include data from studies of rhodanese, which reportedly mediates persulfidation of malate dehydrogenase in vitro, resulting in increased activity⁶⁹, and from the Marina arenicola SQR, which transfers persulfide to thioredoxin⁷⁰, although direct evidence for this transfer was not provided.

Mitochondrial H₂S oxidation pathway: a source of RSS?

The potential for the mitochondrial sulfide oxidation pathway present in most mammalian cells to serve as a source of RSS, and therefore, as a mediator of sulfide-based signaling, has not been appreciated. Instead, the pathway is generally viewed as one that exists to convert H₂S to a mixture of innocuous products, thiosulfate and sulfate (Fig. 6a), which are subsequently eliminated. The first step in the sulfide oxidation pathway is catalyzed by the flavoprotein SQR, which catalyzes a two-electron oxidation of H₂S and transiently forms an enzyme-bound persulfide intermediate, which is then transferred to a small-molecule acceptor. The electrons released in this reaction are passed via ubiquinone to complex III, making sulfide an inorganic substrate for mammalian oxidative phosphorylation⁷¹. The glutathione persulfide (GSSH) product of the reaction^12,47 is a substrate for the nonheme iron–containing persulfide dioxygenase⁷² and for rhodanese^12,47. Sulfite oxidase, present in the intermitochondrial membrane space, oxidizes sulfite to sulfate. Organization of the sulfide oxidation pathway can vary in different archaea, bacteria and plants, variously including components that are homologous to the mammalian enzymes or different from them^73–75.

Figure 6. Sulfide oxidation pathways.

(a) The canonical sulfide oxidation pathway found in most tissues resides in the mitochondrion and involves four enzymes. In the first step of the pathway, sulfide quinone oxidoreductase (SQR) oxidizes sulfide to persulfide, which is transferred from the active site of SQR to a small molecule acceptor, such as glutathione (GSH). The glutathione persulfide (GSSH) product can be oxidized by persulfide dioxygenase (PDO) to sulfite or can be used in a sulfurtransferase reaction catalyzed by rhodanese (Rhd) in the presence of sulfite, to form thiosulfate. In the final step, sulfite is oxidized to sulfate by sulfite oxidase (SO). (b) Heme-dependent sulfide oxidation pathway. An alternative pathway for sulfide oxidation involves ferric heme–dependent conversion of H₂S to a mixture of thiosulfate and polysulfides. This newly discovered mechanism has been established for human hemoglobin and could be an activity of other heme proteins as well. For clarity, the fate of the H₂S sulfur atom is highlighted in red and other reactants and reaction stoichiometries are omitted.

At least four RSS are produced during mitochondrial sulfide oxidation. The first is the SQR-bound persulfide, which in principle can be transferred to another protein or to any of a number of small-molecule acceptors^47,68. SQR is anchored to the inner membrane with its catalytic domain protruding into the matrix. Its ability to donate sulfane sulfur to protein acceptors in the mitochondrial membrane or in the matrix needs to be assessed. GSSH is the second RSS produced in the sulfide oxidation pathway and is predicted to be the predominant product of the human SQR reaction at physiologically relevant concentrations of small-molecule acceptors⁴⁷. GSSH is reactive and can transfer its sulfane sulfur group to a target protein (Fig. 5b) as demonstrated with human rhodanese, which subsequently transfers the sulfane sulfur to sulfite, forming thiosulfate (Fig. 6a)^12,47.

Thiosulfate is the third RSS product in the sulfide oxidation pathway and, although relatively stable in solution, is a substrate for thiosulfate:glutathione sulfurtransferases, such as rhodanese⁴⁷ and TSTD1 (ref. 67). Its solution stability combined with its ability to serve as a source of sulfane sulfur makes thiosulfate a potentially ideal ‘Trojan horse’ RSS that, upon activation by a specific sulfurtransferase, can mobilize its terminal sulfur for transfer. The role of thiosulfate in sulfide-based signaling needs to be assessed, particularly in light of recent reports of its therapeutic potential in acute lung injury⁷⁶ and hypertensive cardiac disease⁷⁷. The archaeon Metallosphaera cuprina utilizes tetrathionate as an energy source and, interestingly, transfers the thiosulfate group between cysteines in proteins involved in dissimilatory sulfur metabolism⁷⁸. Although an enzyme that catalyzes the oxidative condensation of 2 mol of thiosulfate to form tetrathionate has not been described in mammals, such an activity, if present, would open up yet another RSS-based post-translational modification.

Sulfite, the fourth RSS produced during sulfide oxidation, is readily oxidized by enzymes such as myeloperoxidase, prostaglandin H synthase and eosinophil peroxidase generating highly reactive products including the sulfite radical anion (^•SO₃⁻) and the peroxymonosulfate (⁻O₃SOO^•) and sulfate (SO₄^•−) radical anions⁷⁹. It is likely that cells limit the damaging potential of sulfite by efficiently oxidizing it to sulfate by means of sulfite oxidase, or through its utilization by other enzymes, such as rhodanese. The toxicity of sulfite is consistent with the fact that free serum sulfite concentrations are below detection limits⁸⁰, while free tissue sulfite concentrations have not been reported to our knowledge. Interestingly, the antimicrobial and antioxidant properties of sulfite have led to its use as a preservative in the food and pharmaceutical industries, and sulfite sensitivity is observed in a subset of asthmatics⁸¹. Sulfite reductase found in archaea, bacteria and plants represents an alternate route for sulfite removal^82,83.

As a source of RSS, the sulfide oxidation pathway bears similarity to the mitochondrial electron transfer pathway, a quantitatively significant source of ROS⁸⁴. A role for mitochondrial ROS has been demonstrated in bacterial clearance by macrophages⁸⁵, inflammasome activation⁸⁶ and induction of a cellular hormetic response⁸⁷. Mechanisms of dynamic regulation of ROS and RSS production are poorly understood, and it is not known whether an increase in the mitochondrial proton motive force, known to intensify ROS production⁸⁸, affects RSS output.

Metal-catalyzed H₂S oxidation: another source of RSS?

Oxidation of sulfide to thiosulfate by hemoglobin⁸⁹ and ferritin⁹⁰ has been reported, and protection against sulfide toxicity by induced methemoglobinemia has been known for many years⁹¹. However, until recently, the mechanism of heme-dependent sulfide oxidation remained a mystery. Numbering ~5 trillion per liter of blood, red blood cells are expected to influence sulfide homeostasis in the circulation, particularly in light of their capacity to generate H₂S⁹². Recently, catalytic sulfide oxidation by human methemoglobin has been shown to yield thiosulfate and metal-bound polysulfide¹³ (Fig. 6b). The factors governing partitioning of the initially formed ferric sulfide hemoglobin between thiosulfate and polysulfide formation are not presently known. The postulated chemical mechanism of sulfide oxygenation is complex, and thiosulfate production is predicted to predominate when the flux of H₂S is low and oxygen tension is high. It is likely that other heme-containing proteins such as myoglobin and cytochrome P₄₅₀ are capable of catalyzing similar sulfide oxidation chemistry in other tissues. Although heme protein–dependent formation of polysulfides describes a route for their biogenesis in mammalian cells, it also raises questions regarding polysulfide management via sequestration or transfer given the inherent reactivity of the product.

Sulfide rapidly reduces the ferric forms of cytochrome c oxidase and cytochrome c. The product of cytochrome c oxidase reduction under aerobic conditions has not been characterized but is predicted to be elemental sulfur^93,94. Sulfide binds to the heme iron in myeloperoxidase⁹⁵ and lactoperoxidase⁹⁶. Although it has been observed that re-formation of the iron–sulfide complex of myeloperoxidase from compound I is oxygen dependent, possible sulfide oxidation products were not characterized⁹⁵. In addition to its reversible inhibition of myeloperoxidase, it is speculated that sulfide could modulate inflammation⁹⁵. Sulfide coordinates to copper in superoxide dismutase, which converts it to elemental sulfur⁹⁷. Careful characterization of sulfide affinity and reaction kinetics, along with product analysis, during catalytic sulfide oxidation by these and other metalloproteins is needed to assess their physiological relevance to RSS production and H₂S clearance.

Future directions

Still in its infancy, the field of H₂S chemical biology is ripe for investigation, with many more questions open than answered. At the most fundamental level is the issue of whether H₂S or an RSS derived from it is pertinent to sulfide-based signaling or whether, as with superoxide and H₂O₂ in ROS signaling, multiple species are involved. Furthermore, the chemical nature of the interaction of H₂S or RSS with other signaling molecules, particularly ^•NO, needs to be addressed in light of the growing evidence for their crosstalk^55,98,99. Our hypothesis that sulfide oxidation pathways are a source of RSS with signaling potential and our discovery of a novel heme-dependent catalytic sulfide oxidation pathway in red blood cells¹³ raise questions about the fate of the resulting RSS and the involvement of other heme proteins in RSS generation in other tissues. Finally, studies aimed at unraveling cellular strategies for regulating RSS production, sequestration and mobilization will be critical for understanding sulfide signaling and crosstalk with other signaling pathways.