A case of mistaken identity: are reactive oxygen species actually reactive sulfide species?

Abstract

Stepwise one-electron reduction of oxygen to water produces reactive oxygen species (ROS) that are chemically and biochemically similar to reactive sulfide species (RSS) derived from one-electron oxidations of hydrogen sulfide to elemental sulfur. Both ROS and RSS are endogenously generated and signal via protein thiols. Given the similarities between ROS and RSS, we wondered whether extant methods for measuring the former would also detect the latter. Here, we compared ROS to RSS sensitivity of five common ROS methods: redox-sensitive green fluorescent protein (roGFP), 2′, 7′-dihydrodichlorofluorescein, MitoSox Red, Amplex Red, and amperometric electrodes. All methods detected RSS and were as, or more, sensitive to RSS than to ROS. roGFP, arguably the “gold standard” for ROS measurement, was more than 200-fold more sensitive to the mixed polysulfide H₂S_n (n = 1–8) than to H₂O₂. These findings suggest that RSS may be far more prevalent in intracellular signaling than previously appreciated and that the contribution of ROS may be overestimated. This conclusion is further supported by the observation that estimated daily sulfur metabolism and ROS production are approximately equal and the fact that both RSS and antioxidant mechanisms have been present since the origin of life, nearly 4 billion years ago, long before the rise in environmental oxygen 600 million years ago. Although ROS are assumed to be the most biologically relevant oxidants, our results question this paradigm. We also anticipate our findings will direct attention toward development of novel and clinically relevant anti-(RSS)-oxidants.

redox modifications of organic sulfur by reactive oxygen species (ROS) control protein structure, activity, and trafficking and are important signaling mechanisms controlling cellular homeostasis (3, 8, 17, 51, 56, 66). This “Redox Code” has been said to have been “…richly elaborated in an oxygen-dependent life, where activation/deactivation cycles involving O₂ and H₂O₂ contribute to spatiotemporal organization for differentiation, development, and adaptation to the environment” (22). The code is a natural extension of the Ox-Tox hypothesis (29, 52), in which the rise in atmospheric oxygen ∼600 million years ago is believed to have necessitated sophisticated defenses against oxygen toxicity. Presumably, this was subsequently adapted for regulatory functions.

ROS are produced by sequential one-electron reduction of O₂ producing superoxide (superoxide radical anion; O₂^·−), hydrogen peroxide (H₂O₂), hydroxyl radical (HO^·), and water (Eq. 1):

$${\text{O}}_2\left(+{\text{e}}^{-}\right)\to {{\text{O}}_2}^{·-}\left(+{\text{e}}^{-}\right)\to {\text{H}}_2{\text{O}}_2\left(+{\text{e}}^{-}\right)\to {\text{HO}}^{·}\left(+{\text{e}}^{-}\right)\to {\text{H}}_2\text{O}$$ (1)

H₂O₂ is the most logical candidate for ROS signaling because of its stability, intracellular and transcellular mobility, and its specificity for protein thiols (17), although a case can also be made for O₂^·− (66). The presumed ubiquity and importance of ROS signaling and potential for pathophysiological consequences have understandably fostered the development of a variety of analytical methods to measure and track them (13, 58, 63, 66).

However, it seems unlikely to us that all this came about only in the last one-eighth of all (and half of eukaryotic) evolution. Many redox signaling mechanisms and antioxidant systems present in modern-day organisms can be traced back nearly 4 billion years to the appearance, in an anoxic environment, of the last universal common ancestor (4, 28, 35, 43, 48, 55, 68). ROS would have been scarce during this period and most likely for the ensuing 3.2 billion years of evolution because the oceans remained anoxic or severely hypoxic and became more sulfidic until the dramatic oxygenation 600 million years ago (49, 50). We propose that redox stress and signaling were not a response to ROS, but was necessitated by another stressor, reactive sulfide species (RSS). RSS were not only inextricably tied to the origin of life, but it is becoming increasingly apparent that they have persisted up to the present day as integral components of homeostasis (6, 45, 62).

The products of sequential one-electron oxidation of hydrogen sulfide (H₂S) chemically mirrors those of O₂ reduction producing, in order, the thiyl radical (HS·), hydrogen persulfide (H₂S₂), persulfide radical (H₂S₂·; “supersulfide”), and molecular sulfur (S₂ or S_n, where n = 2–8; Refs. 46 and 47) This is shown in Eq. 2, although the reaction is shown in reverse, and the H is adjusted irrespective of pK to enable better comparison of the sulfur species with the oxygen species in Eq. 1.

$${\text{S}}_2\leftarrow \left(-{\text{e}}^{-}\right){{\text{S}}_2}^{·-}\leftarrow \left(-{\text{e}}^{-}\right){\text{H}}_2{\text{S}}_2\leftarrow \left(-{\text{e}}^{-}\right)\text{HS}·\leftarrow \left(-{\text{e}}^{-}\right){\text{H}}_2\text{S}$$ (2)

Similarities between ROS and RSS are striking; 1) both ROS and RSS are produced endogenously in the cytosol and, perhaps more importantly, in mitochondrion; 2) estimated tissue production of ROS and RSS is similar; 3) H₂O₂ and H₂S₂ are the most probable signaling moieties of ROS and RSS, respectively; and 4) both H₂O₂ and H₂S₂ signal through redox-sensitive protein cysteines and activate common effectors (see discussion). However, unlike H₂O, H₂S is also a reactive species or can act as one. In direct comparisons, RSS appear more efficacious than ROS in inactivating phosphatase and tensin homolog (PTEN; Ref 16), and preliminary studies have shown that RSS may be mistaken for ROS when using amperometric electrodes (DeLeon ER and Olson KR, unpublished data) or redox-sensitive green fluorescent protein (roGFP; Ref 16).

Given the relative nonspecificity of many ROS probes toward other ROS (13, 63) and the chemical and biochemical similarities between ROS and RSS, we hypothesized that methods for measuring ROS would also detect RSS. To this end, we examined the responses of “classical” ROS indicators, roGFP, 2′,7′-dihydrodichlorofluorescein (DCF), Amplex Red, MitoSox Red, and amperometric H₂O₂ sensors to ROS (H₂O₂ and O₂^·−), and to potential RSS, including H₂S, mixed polysulfides (H₂S_n; n = 2–8), and specific polysulfides (H₂S₂, H₂S₃, H₂S₄).

MATERIALS AND METHODS

Chemicals.

SSP4 was generously provided by Dr. Ming Xian, Washington State University. DCF, MitoSox Red, Amplex Red, and LB Broth (Miller) were purchased from ThermoFisher Scientific (Grand Island, NY). Na₂S₂, Na₂S₃, and Na₂S₄ were purchased from Dojindo Laboratories, Kumamoto, Japan. All other chemicals were purchased from Sigma-Aldrich (St. Louis, MO). Na₂S was used to produce H₂S.

roGFP.

A plasmid containing a roGFP2 sequence, an ampicillin resistance sequence, and a lac operon promoter sequence was obtained from the University of Oregon (Eugene, OR). The plasmid was then transfected into BL21D3 Escherichia coli cells. The success of the transfection was verified via purification and spectroscopy.

To purify the roGFP protein, three colonies were chosen at random and grown in LB Broth (Miller) in 100 μM ampicillin to an optimum optical density of 0.800 and absorbance units of 600 nm. The cells were then pelleted using an Avanti J-30I centrifuge at 6,000 g for 6 min. The pellet was then weighed and resuspended in a 5× volume. This was left for 20 min in lysis buffer, and then sonicated to ensure complete cell lysis. The cell lysate was then centrifuged at 50,000 g for 30 min, and the supernatant was filtered through a 0.8-μm, then 0.45-μm syringe filter. Ni-NTA agarose solution was then added to the filtered supernatant, and after 2 h, the supernatant was passed through a Thermo Scientific 5-ml polypropylene column, washed, and eluted.

On the basis of initial studies, the protein appeared to be nearly completely oxidized once purified. To obtain a partially reduced protein, the roGFP was pretreated with 1 mM dithiothreitol (DTT) for 20 min and then dialyzed overnight with two rinses in 100% N₂ sparged Sorenson's (phosphate) buffer at a pH of 7 using Spectra/Por membrane tubing with 3.5-kDa molecular weight cutoff. This yielded protein that was 25–40% oxidized. Purified protein was diluted 1:50, and a volume of 200 μl of diluted protein was then added to each well of a 96-well plate; the final protein concentration was ∼20 μM. Samples were read on a plate reader (see General protocol: roGFP, MitoSox Red, Amplex Red, and DCF) at excitation wavelengths of 405 and 488 nm and an emission wavelength of 510 nm. The fraction oxidation was determined by ratiometric analysis and comparisons to roGFP, which was completely oxidized with 10 mM H₂O₂ and completely reduced with 30 mM DTT, as described previously (38).

DCF.

H₂DCF-DA (2′,7′-dichlorodihydrofluorescein diacetate) is a nonfluorescent lyophilic ester that is readily taken up by cells and deesterified to the nonfluorescent alcohol, H₂DCF. In our experiments, which were carried out in buffer, we used the deacetlyated form, H₂DCF. When H₂DCF is oxidized to DCF, it fluoresces upon exposure to blue light (26). DCF reportedly responds indirectly to H₂O₂ or O₂^·− after their oxidation to HO^· (26). DCF is rapidly oxidized by heme in hemoglobin, myoglobin, and cytochrome c to fluorescent fluorescein independent of ROS (44). Ten micromolar DCF was used with 500/525 nm excitation/emission (e/m) wavelength. Samples were processed as described in General protocol: roGFP, MitoSox Red, Amplex Red, and DCF.

Amplex Red.

Amplex Red reacts with H₂O₂ in the presence of horseradish peroxidase to produce the fluorescent resorufin. High levels of H₂O₂ become inhibitory as they can directly oxidize resorufin (53). We used 50 μM Amplex Red with 565/585 e/m and followed the general protocol.

MitoSox Red.

MitoSox Red is a mitochondrial-targeted fluorescent dye that shows high reactivity but questionable specificity for superoxide and is susceptible to photo-oxidation and oxidation by heme proteins and by H₂O₂ in the presence of iron or copper (13, 69). We used 15 μM MitoSox Red with 510/580 e/m and followed the general protocol, as described below.

General protocol: roGFP, MitoSox Red, Amplex Red, and DCF.

In a typical experiment, roGFP, MitoSox Red, Amplex Red, or DCF was aliquoted into black 96-well plates containing buffer in a darkened room. A black cover with a Parafilm liner was placed over the plates to further minimize photobleaching and reduce H₂S volatization. The edge of the plate was also wrapped with Parafilm. Baseline fluorescence was measured on a SpectraMax M5e plate reader (Molecular Devices, Sunnyvale, CA) every 10 min over 30 min (t = −30, −20, and −10 min). The plates were then uncovered; the experimental compounds (ROS and RSS) were added, the plate was resealed, and fluorescence was recorded for an additional 100 min (0–90 min), with sampling at 10-min intervals. To reduce well-to-well variation, the samples were normalized to the baseline fluorescence at t = −10 min. All experiments were done at room temperature in a darkened room. A minimum of three replicates were obtained over several days. Hypoxia experiments were performed with the plate reader in a model 856-HYPO hypoxia chamber (Plas Labs, Lansing, MI) under 100% N₂, which lowered the ambient O₂ to less than 0.35% (Po₂ ∼5 mmHg). Under normal barometric conditions (∼747 ± 2 mmHg), this produced an O₂ concentration less than 3.8 μM in the buffer.

Polysulfide production from H₂S.

As described in results, we consistently observed that H₂S appeared to oxidize the ROS-sensitive fluorescent dyes. However, this is theoretically impossible because the H₂S sulfur is completely reduced, and, therefore, cannot act as an oxidant (9, 60). It has also been reported that the Na₂S salt used to produce H₂S is contaminated with oxidized sulfur, much in the form of polysulfides (16). To determine whether the Na₂S contained oxidized polysulfides, or if H₂S produced polysulfides in solution, we monitored polysulfide production from H₂S with the polysulfide-selective fluorescent compound SSP4 (50 μM). These experiments were conducted in both normoxia and in the hypoxia chamber under 100% argon, which reduced O₂ to <0.35% (Po₂ ∼5 mmHg).

Amperometric sensors.

ISO-HPO-2 and ISO-NOP amperometric H₂O₂ and NO sensors were purchased from World Precision Instruments (WPI) (Sarasota, FL). They are designed for tissue culture with 2-mm diameter replaceable membrane sleeves and a reported detection limit of <100 nM (ISO-HPO-2) and 1 nM (ISO-NOP). The sensors were connected to WPI Apollo 4000 or TBR 4100 free radical analyzers, and the data were archived on a laptop PC with software provided by the manufacturer and exported into Microsoft Excel. The H₂S amperometric sensor was constructed in-house, as described previously (65). This sensor has a sensitivity of 14 nM H₂S gas (∼100 nM total sulfide). The sensors were calibrated daily with fresh standards.

Reaction chambers with three side ports for H₂S, H₂O₂, and O₂ sensors and a 1-cm wide by 2-cm deep central well for samples were purchased from WPI (NOCHM-4). A polycarbonate stopper with a Viton o-ring accommodated the NO sensor, and an additional hole in the stopper permitted venting the headspace air when the stopper was lowered into the chamber and provided an access port for sample injection with a Hamilton microliter syringe. The chamber was placed on a magnetic stirrer and stirred with a Teflon micro stir bar. All experiments were done at room temperature.

Data analysis.

The data were analyzed and graphed using QuatroPro (Corel, Ottawa ON, Canada) and SigmaPlot 13.0 (Systat Software, San Jose, CA). Statistical significance was determined using SigmaPlot 13.0, one-way ANOVA, and Holm-Sidak multiple comparisons for roGFP dose-response curves and comparisons between H₂O₂, KO₂, H₂S, and H₂S_n, or Students' t-test for comparing normoxia to hypoxia or iron chelation with DTPA. Results are given as means ± SE; significance was assumed when P ≤ 0.05.

RESULTS

roGFP.

Redox-sensitive green fluorescent protein (roGFP) is arguably the gold standard ROS indicator (58). The degree (fraction) of roGFP oxidation is typically compared with roGFP that is completely oxidized with a strong oxidant, such as H₂O₂ or reduced with a reductant, such as dithiothreitol (DTT).

In our experiments, prereduced roGFP was dose dependently oxidized by ROS, H₂O₂, and O₂^·− (KO₂) and also oxidized by RSS, H₂S, H₂S_n (K₂S_n), H₂S₂, H₂S₃, and H₂S₄ in normoxia (Figs. 1 and 2A). With the exception of H₂S, RSS were better oxidants of roGFP than ROS; H₂S_n was 216 times, H₂S₂ and H₂S₃ were 25 times, and H₂S₄ was 70 times more efficacious than H₂O₂ (Table 1). The increase in roGFP oxidation as S increases is likely why H₂S_n is the most potent. Although theoretically H₂S cannot oxidize roGFP because the sulfur is completely reduced (9, 60), it clearly reacts with roGFP. This process appears to be independent of the presence of oxygen, as H₂S-mediated roGFP oxidation was also observed in anoxia (Fig. 2B), and under these conditions, the efficacy of H₂S was surprisingly increased. This suggests there is direct chemical interaction between H₂S and roGFP cysteines.

Fig. 1.

Redox-sensitive green fluorescent protein (roGFP) is dose-dependently oxidized by reactive oxygen species (ROS), H₂O₂ and O₂^·− (KO₂) and by reactive sulfide species (RSS) H₂S and H₂S_n. Cumulative dose-dependent responses to ROS and RSS (including H₂S₂, H₂S₃, and H₂S₄; see also Fig. 2A) clearly show that RSS are more potent oxidants of roGFP, although H₂S_n reduces roGFP at concentrations >300 μM. roGFP alone is shown in black, and symbol colors progress with concentration along the spectrum from cyan to red. Values are expressed as means ± SE; n = 3 replicates of each concentration. Fraction oxidized × 100 = percent oxidized.

Fig. 2.

A: roGFP oxidation by polysulfides, H₂S₂, H₂S₃, and H₂S₄ in normoxia increases as S_n increases. B: roGFP oxidation by H₂O₂, H₂S, and H₂S_n in hypoxia is shown. The effects of H₂O₂ and H₂S_n are similar to that in normoxia (compare with Fig. 1), whereas H₂S is more efficacious in hypoxia. Symbols are the same as those used in Fig. 1. Values are expressed as means ± SE; n = 3 replicates of each concentration.

Table 1.

Effective concentration for half-maximal oxidation (EC₅₀) of redox sensitive green fluorescent protein (roGFP) by reactive oxygen (ROS) and reactive sulfide (RSS) species

	EC50, μM
H2O2	953 ± 23.3
O2·−	202 ± 14.9
H2S	2669 ± 369.3
H2Sn	4.4 ± 0.5
H2S2	39.9 ± 0.9
H2S3	41.0 ± 4.5
H2S4	14.4 ± 0.7

Values are expressed as means ± SE (n = 4 replicates H₂S; three replicates all others). Each half-maximal effective concentration (EC₅₀) is significantly different from all others (P < 0.001) except H₂O₂ vs. H₂S (P = 0.01), H₂S₃ vs. H₂S₄ (P < 0.004) and H₂S₂ vs. H₂S₃ (not different).

DCF.

Both ROS and RSS produced dose-dependent increases in 10 μM DCF fluorescence (Figs. 3, 4, 5, and 6). In normoxia, high concentrations (1 and 3 mM) of H₂S produced the most rapid increase in DCF fluorescence, whereas the effects of H₂O₂ and H₂S_n appeared somewhat sigmoidal, which suggests there are multi-step processes. Fluorescence produced by 3 mM H₂S at 90 min was significantly (P < 0.001) greater than that produced by either H₂O₂ or H₂S_n at the same time period. High variability of the 3 mM KO₂ response precluded any significant differences between it and the other ROS or RSS. At 100 μM, the H₂O₂ effects were significantly (P < 0.001) greater than that produced by KO₂ or RSS, and H₂S effects were greater than those of H₂S_n but not KO₂. Maximal fluorescence of pure polysulfides increased as the number of sulfur atoms increased from H₂S₂ to H₂S₄ (Fig. 7).

Fig. 3.

2′, 7′-dihydrodichlorofluorescein (DCF) oxidation (expressed as relative fluorescence units) produced by H₂O₂ from 1 μM to 1 mM (A) and in an expanded scale from 1 μM to 100 μM (B) in normoxia (21% O₂), hypoxia (<0.4% O₂), the iron chelator diethylene triamine pentaacetic acid (DTPA, 50 μM), and hypoxia plus DTPA. Hypoxia with or without DTPA increased H₂O₂-mediated DCF fluorescence, whereas DCF was ineffective. Symbols are the same as those in Fig. 1. Values are expressed as means ± SE; n = 6 replicates of each concentration, An asterisk (*) indicates expanded scale.

Fig. 4.

DCF oxidation (expressed as relative fluorescence units) produced by KO₂ (O₂^·−) from 1 μM to 1 mM (A) and in an expanded scale from 1 μM to 100 μM (B) in normoxia (21% O₂), hypoxia (<0.4% O₂), the iron chelator DTPA (50 μM), and hypoxia plus DTPA. DTPA produced the greatest increases in 1 and 3 mM KO₂-mediated DCF fluorescence, whereas DCF fluorescence was not increased by KO₂ at concentrations less than 300 μM. Symbols are the same as in Fig. 1. Values are expressed as means ± SE; n = 4 replicates of each concentration. An asterisk (*) indicates expanded scale.

Fig. 5.

DCF oxidation (expressed as relative fluorescence units) produced by H₂S from 1 μM to 1 mM (A) and in an expanded scale from 1 μM to 100 μM (B) in normoxia (21% O₂), hypoxia (<0.4% O₂), the iron chelator DTPA (50 μM), and hypoxia plus DTPA. Hypoxia with or without DTPA increased H₂S-mediated DCF fluorescence, whereas DCF was ineffective. Symbols are the same as in Fig. 1. Values are expressed as means ± SE; n = 6 replicates of each concentration. An asterisk (*) indicates expanded scale.

Fig. 6.

DCF oxidation (expressed as relative fluorescence units) produced by H₂S_n from 1 μM to 1 mM (A) and in an expanded scale from 1 μM to 100 μM (B) in normoxia (21% O₂), hypoxia (<0.4% O₂), the iron chelator DTPA (50 μM), and hypoxia plus DTPA. Hypoxia with or without DTPA increased H₂S_n-mediated DCF fluorescence, whereas DCF was ineffective. Symbols are the same as in Fig. 1. Values are expressed as means ± SE; n = 6 replicates of each concentration. An asterisk (*) indicates expanded scale.

Fig. 7.

DCF fluorescence is increased by all polysulfides (H₂S₂, H₂S₃, and H₂S₄) at high S concentrations (left) and expanded scales (right) show that it is readily oxidized by polysulfides as low as 1 μM. DCF was slightly more sensitive to H₂S₃ than to other polysulfides. Values are expressed as means ± SE; n = 3 replicates of each concentration.

Hypoxia (Po₂ ∼5 mmHg) for 90 min significantly (P < 0.001) increased ROS and RSS fluorescence at 1 mM and all but KO₂ at 300 μM (Figs. 3–6). Hypoxia also appeared to increase the rate of fluorescence development produced by 3 mM ROS and RSS. Hypoxia doubled 100 μM H₂O₂, H₂S, and H₂S₂ fluorescence but did not affect KO₂ fluorescence. Hypoxia also doubled spontaneous DCF fluorescence, so it was not apparent whether the increase in ROS and RSS fluorescence was just proportional to the increase in DCF fluorescence or specific for the reactive species. It is clear, however, that hypoxia did not diminish either ROS or RSS effects. The iron chelator DTPA (diethylene triamine pentaacetic acid; 50 μM) did not significantly alter response profiles of 10 μM DCF to either ROS or RSS in normoxia or hypoxia. This suggests that a Fenton-type reaction (26) was not involved. DCF alone exhibited a slight increase in oxidation over the 120-min experimental period, most likely due to photosensitivity.

Amplex Red.

As expected, H₂O₂ and KO₂ increased Amplex Red fluorescence (not shown), although it was autoinhibited by 1 mM H₂O₂ due to resorfin oxidation, as reported previously (53). Conversely, 1 mM H₂S, H₂S_n, H₂S₂ H₂S₃, and H₂S₄ were only slightly stimulatory, and the latter four were slightly inhibitory at higher concentrations (not shown). In normoxia, the addition of a one-electron oxidant (FeCl₃, 500 μM) did not affect H₂O₂ fluorescence, slightly increased KO₂ fluorescence, but greatly increased H₂S and H₂S_n fluorescence (Fig. 8A). The increase in fluorescence from 1 mM H₂S in the presence of FeCl₃ appeared sigmoidal, suggesting a chain reaction. Other one-electron oxidants, hemin (500 μM) and CuCl₂ (500 μM), had qualitatively similar effects as FeCl₃ (not shown), suggesting that RSS radicals may be more reactive with Amplex Red. Unlike FeCl₃ and CuCl₂, hemin appeared to reverse 1 mM H₂O₂ inactivation of Amplex Red (not shown).

Fig. 8.

Oxidation of Amplex Red in 500 μM FeCl₃ (A) and MitoSox Red (B) by reactive oxygen species (ROS), H₂O₂, and O₂^·− (KO₂), and reactive sulfide species (RSS), H₂S, and (H₂S_n). RSS were less efficacious oxidants of Amplex Red than ROS. H₂O₂ and H₂S were equipotent MitoSox oxidants, although the response to H₂S was faster. MitoSox Red was not oxidized by either O₂^·− or H₂S_n, and the latter decreased fluorescence intensity. Symbols as in Fig. 1. Values are expressed as means ± SE; n = 3 replicates of each concentration.

MitoSox Red.

H₂O₂ and H₂S were essentially equipotent, whereas KO₂ was considerably less efficacious, and high concentrations of H₂S_n were slightly inhibitory (Fig. 8B). H₂S₂ and H₂S₄, but not H₂S₃ increased Mitosox Red fluorescence at low concentrations, whereas all appeared slightly inhibitory at higher concentrations (not shown).

Polysulfide production from H₂S.

When H₂S was added to wells containing the polysulfide-sensitive fluprophore SSP4, there was a steady increase in fluorescence irrespective of oxygen concentration (Fig. 9). The addition of similar concentrations of H₂S_n to SSP4 in either normoxia or hypoxia produced a rapid increase in fluorescence that plateaued within 20–25 min and was ∼27 times greater than that produced by H₂S (Fig. 9). This suggests that H₂S slowly forms polysulfides when in solution and that this process is independent of oxygen or requires very little of it.

Fig. 9.

SSP4 fluorescence as an indication of polysulfide production from H₂S (top) in well plates in normoxia and hypoxia compared with SSP4 fluorescence produced by equivalent concentrations of H₂S_n (bottom). Polysulfides were slowly, but continuously, formed from H₂S irrespective of oxygen concentration. Symbols are the same as in Fig. 1. Values are expressed as means ± SE; n = 3 replicates of each concentration.

Amperometric sensors.

H₂O₂ amperometric sensors are generally designed to measure H₂O₂ release from cells or tissues into media. H₂O₂ diffuses across a presumably selective polymeric membrane and is oxidized in a specific electrolyte at the manufacturer's recommended polarizing voltage of 450 mV. As expected, the H₂O₂ sensor produced a nearly linear response to H₂O₂. The amperometric sensor also responded to H₂S and polysulfides (H₂S₂, H₂S₃, and H₂S₄) with even greater sensitivity (Fig. 10, A–C). Although the responses to both H₂O₂ and sulfides tended to decrease at higher concentrations, the H₂O₂ sensors were nearly 30 times more sensitive to H₂S than to H₂O₂, three times more sensitive to H₂S₂ than to H₂O₂, and twice as sensitive to H₂S₃ and H₂S₄. The decrease in response as S_n increases could be due to the increase in percent ionization of the polysulfide, which would likely impair diffusion across the sensor membrane. Because both pKa₁ and pKa₂ decrease as the number of sulfur atoms increases (25), at pH 7.0, the percent H₂S_n/HS_n−/S_n2− decreases from 50/50/∼0 when n = 1 (H₂S) to 11.3/83.2/5.6 when n = 2 and 1.3/32.8/65.9 when n = 4. Nevertheless, not only do H₂O₂ sensors fail to discriminate between peroxide and all sulfur species examined, they are considerably more sensitive to the latter.

Fig. 10.

Amperometric H₂O₂ sensors are nearly 30 times more sensitive to H₂S and 2–3 times more sensitive to H₂S₂ H₂S₃ and H₂S₄ than to H₂O₂. A: real-time traces of H₂O₂ sensor responses to 0.5–10 μM H₂S vs. 20–125 μM H₂O₂. Calibration curves for H₂S, H₂S₂, H₂S₃, H₂S₄, and H₂O₂ at full (B) and expanded (C) scale. D: calibration curves of NO amperometric sensors to NO, H₂S, and H₂O₂. NO sensors were more sensitive to H₂O₂ and nearly 80 times more sensitive to H₂S than to NO. Values are expressed at means ± SE; n = 3 replicates of each concentration.

We also examined the cross-sensitivity of H₂S, O₂, and NO amperometric sensors to H₂S, O₂, H₂O₂, and the nitric oxide donor S-nitroso-N-acetylpenicillamine. H₂S and O₂ sensors were selective for their respective gases and did not exhibit appreciable cross sensitivity, whereas the NO sensor was considerably more sensitive to H₂O₂ than to NO and nearly 80 times more sensitive to H₂S than to NO (Fig. 10D). Although not the focus of the present study, these results also call into question how often RSS are mistaken for reactive nitrogen species.

DISCUSSION

There is an ever-increasing appreciation of the chemical similarities between ROS and RSS and their ability to activate common effectors via identical mechanisms. However, little is known about whether these similarities extend to their analysis. We examined the sensitivity to RSS of five methods commonly used to detect ROS. These experiments were performed in cell-free media to minimize cross-reactivity with other biological oxidants and to eliminate the effects of enzyme-catalyzed reactions. Our experiments show that methods commonly used to measure ROS cannot distinguish ROS from RSS and in some instances may be considerably more sensitive to the latter. Collectively, these results suggest that the extent of ROS signaling in biological systems may be overestimated and that RSS may be more physiologically relevant than currently appreciated.

roGFP, arguably the gold standard ROS indicator (58), can be targeted to specific intracellular compartments (64) and has been studied extensively. Fluorescence absorbance of roGFP changes when an inserted disulfide bridge is oxidized or reduced, allowing ratiometric calculation of redox state that, although it is essentially independent of pH and many other interferences (58), clearly is sensitive to RSS. It is likely that the inserted cysteines are the target of both ROS and RSS, as they are in many other proteins (9, 31, 34, 37, 39, 41, 56). Our findings not only question whether ROS or RSS are being measured in cells, but they also suggest that RSS may be more efficacious than ROS as regulators of redox-sensitive protein cysteines. Indeed, this has already been demonstrated for PTEN (16).

Lack of specificity of the other fluorescent indicators used in this study, especially cross reactivity with other oxygen and nitrogen radicals is well known (11, 24, 36, 54, 63, 67, 69). Studies on cross reactivity of these indicators with RSS have been limited.

Eghbal et al. (7) observed that H₂S (0.5 mM) increased H₂DCF-DA fluorescence of isolated rat hepatocytes. H₂S (1.9 mM) also increased H₂DCF-DA and MitoSOX Red fluorescence in erythrocytes from the sulfide-tolerant annelid, Glycera dibranchiata (23). These results were interpreted as a sulfide-induced increase in oxidative stress and O₂^·− production. Only Julian et al. (23) examined the effects of H₂S in cell-free buffer and observed that while 0.29, 0.73, and 1.9 mM H₂S produced significant increases in H₂DCF-DA fluorescence, this was very small (1/50) compared with the increase in erythrocytes. Our experiments show that DCF, MitoSox Red, and Amplex Red also respond to RSS in cell-free buffer which, with the exception of Amplex Red, eliminates an enzyme-catalyzed production of ROS from RSS. However, we could not eliminate the possibility that the fluprophore, or horseradish peroxidase employed in the Amplex Red assay, could also catalyze ROS production (11, 24, 36, 54, 63, 67, 69). Amplex Red was shown to be auto-oxidized by 0.5–1.5 mM GSH, but not by GSSG (61). Both SOD and catalase prevented GSH oxidation suggestive of a O₂^·− intermediate (61). Although we found Amplex Red was less sensitive to RSS than to ROS, the latter clearly did respond to RSS in the presence of one-electron oxidants.

To examine these reactions further, we hypothesized that if sulfide increases ROS, then by lowering O₂ availability, i.e., hypoxia, or removing iron impurities with DTPA, we should reduce ROS production and, thereby, decrease RSS activation of the fluorescent indicators. To this end, we compared the effects of normoxia (∼225 μM O₂) and hypoxia (<3.8 μM O₂) on RSS oxidation of roGFP and DCF and the effects of DTPA on DCF oxidation in both normoxia and hypoxia. Contrary to what we expected, hypoxia actually increased RSS oxidation potency in roGFP and either increased or did not change RSS oxidation of DCF. Although we cannot rule out some ROS production in hypoxia, our results suggest that the effects of RSS are not due to increases in ROS production but a direct effect of RSS. Even if ROS were generated from RSS and only the former were measured in our experiments, the ROS would be merely reporters of RSS and not indicators of primary ROS production.

Estimation of ROS and RSS production from cysteine metabolism.

A comparison of the estimated rates of ROS and RSS production in humans shows that these values are surprisingly similar. Oxygen consumption in a typical adult male is around 250 ml/per min (2), which is equivalent to 360 liters per day. At 22.4 l/mol, this is equivalent to 16 mol of diatomic oxygen (O₂) per day. Mitochondrial ROS production in rat muscle is 0.35 percent of O₂ consumption at rest and decreases to 0.01% during heavy exercise (14). This is equivalent to between 1 and 56 mmol ROS per day; at an average of 0.1% ROS/mol O₂, this is 16 mmol/day. The average intake of sulfur amino acids (S-AA) is 26 mmol/day, and S-AA from protein turnover adds another 70 mmol/day, ∼90% (88 mmol/day) of which is used for protein synthesis (19, 20). Of the remaining 10 mmol, approximately half are desulfurated (59), generating 5 mmol H₂S/day. [Although gut flora produce considerable H₂S, up to 40 μM in the colon, it is effectively oxidized by the epithelium and is not an appreciable source of reduced sulfur (10, 33)]. Keeping in mind that the methods used to measure ROS production may actually be measuring RSS production, the above values may be considerably closer, and RSS production may well exceed ROS production.

ROS and RSS formation and metabolism.

ROS and RSS also share similar sites of production and metabolism. ROS are produced endogenously in the cytosol and perhaps, more importantly, in the mitochondria (3, 8, 14, 17, 30, 51, 56, 57, 66); as many as 10 different sites may produce ROS in mitochondria (1). Although production and metabolism of H₂S are being extensively investigated, RSS production, metabolism, and signaling mechanisms are only beginning to receive attention. RSS can be produced in both cytosolic and mitochondrial compartments via a variety of mechanisms that can be independent of the presence of oxygen (reviewed in Refs. 9, 40, 42).

One-electron oxidation of thiols (RS), H₂S, or their persulfides produces the thiyl (RS·⁻), sulfhydryl (HS·⁻), or their respective persulfide radicals (RS₂·⁻, HS₂·⁻). These reactions are generally catalyzed by heme iron or ferric (Fe³⁺) or cupric (Cu²⁺) ions. As with oxygen radicals, the sulfur radicals are very reactive and can generate a variety of downstream effects, including production of other radicals (e.g., RS⁻ + HS·⁻ → RS₂·⁻), or in combination with each other, generating inorganic and organic polysulfides (e.g., 2HS·⁻ → H₂S₂). Persulfides and polysulfides can also be produced by two-electron oxidation of organic sulfur (cysteine) and transferred to a protein thiol (e.g., 3-mercaptopyruvate sulfur transferase, 3MST), forming 3MST persulfide (3MST-S-S). The hydrogen persulfide, H₂S₃, has recently been shown to be formed from 3MST persulfide in neuronal cells, and it may exist in micromolar concentrations (27). H₂S can also reduce a disulfide bond (RSSR′ + H2S → RSSH + R'S). The best example of this is the initial step in H₂S metabolism, where H₂S reduces a disulfide bridge on the mitochondrial enzyme sulfur quinone oxidoreductase (SQR). The resulting sulfane sulfur is then transferred to either glutathione-forming GSSH (32) or sulfite-forming thiosulfate (S₂O₃²⁻; Ref. 21); or, after repetitive cycles, it forms a progressively longer polysulfide between Cys160 and Cys356 on SQR (5). Once formed, the sulfane sulfur of a persulfide or a polysulfide is very mobile in that it can be transferred to other thiols, including persulfides and polysulfides (e.g., 2HS₂⁻ → HS₃⁻ + HS⁻, 2HS₃⁻ → HS₄⁻ + HS₂⁻), and longer-chain inorganic polysulfides can readily undergo homolytic dissociation (HS₄⁻ → 2HS₂⁻). In addition, the disulfide, cystine, can be transported into cells and metabolized by cytosolic enzymes cystathionine β-synthase and cystathionine γ-lyase to cysteine and glutathionine polysulfides, Cys-S_n and GS_n (where n = 1–4); these polysulfides may exist in the cell in excess of 100 μM (18), although it should be noted that these compounds are unstable, and confirmation of cellular values will require more reliable standards. This further increases the complexity of sulfide signaling and provides additional evidence that RSS are more abundant than ROS in cells.

It is clear from these studies that H₂S, H₂S_n, and H₂S_2–4 have effects similar to those produced by commonly accepted oxidants in five different methods routinely employed to identify ROS. How they do this is not as clear. The RSS may react directly with the fluorescent dye, or it may generate ROS, which, in turn, reacts with dye. This is further complicated by H₂S because it is in its most reduced form (9, 60) and theoretically cannot act as an oxidant, although our experiments suggest it may spontaneously oxidize to a polysulfide in both normoxia and hypoxia. How this occurs even in low oxygen remains to be determined. In addition, it is generally thought that polysulfides become good reductants as the number of sulfur atoms increases (9, 42, 47). Obviously, there is considerable chemistry to be solved to understand how RSS interact/react with these varied indicators. Irrespective of whether the RSS act as direct oxidants, generate ROS that now serve as reporters of RSS, or excite by other mechanisms, the important point is that all of the methods examined in this paper cannot distinguish ROS from RSS. Thus, it is quite likely that many reports of ROS production are actually measuring RSS production.

Actual reports of cellular RSS concentrations are rare, but preliminary studies provide concentrations that equal or exceed ROS (18, 27). This is further supported by our calculations, above, of sulfur and oxygen turnover. In addition, RSS are more chemically and biochemically versatile than ROS. In cells H₂S₂ can act as a reductant or oxidant (9, 42, 47), whereas the high pKa of H₂O₂ (>11) restricts it to an oxidant except under extreme (and nonphysiological?) circumstances. With the exception of dismutation, ROS become inexorably more reduced in cells, whereas RSS can be reversibly oxidized and reduced and interact with far more effector systems. ROS are either monoatomic or diatomic oxygen, whereas polysulfides can contain numerous sulfur atoms, the latter of which are usually assumed to be up to eight, but far longer chains are found in some bacteria (12, 15) and could be present in metazoans as well. All of these sulfane sulfur molecules are potentially RSS. Clearly, these attributes of RSS, along with their extensive involvement in evolution, suggest that they deserve considerably more attention and that the relative biological importance of ROS and RSS needs to be reassessed.

Perspectives and Significance

Sulfur, sulfide, and related species have been important in biological systems since the origin of life, far longer than oxygen. Formation of sulfur-based redox-type reactive species undoubtedly accompanied the origin of life, and this necessitated appropriate defenses, as well as providing opportunities to use these molecules in specialized signaling pathways. The chemical similarities between reactive sulfide species (RSS) and reactive oxygen species (ROS) readily lend themselves to the idea that with the appearance of oxygen 600 million years ago, there would be a likely transition from the former to the latter, if the former were ever considered in the first place. In vitro studies conducted at unphysiological oxygen (i.e., room air is 21% O₂, Po₂ ∼140–160 mmHg, and O₂ concentration can range from 175 to 260 μM between 20°C and 37°C) only confound the issue. Are these assumptions valid? Can 3.2 billion years of evolution be discarded? Our study, which shows that common analytical methods cannot distinguish between ROS and RSS, suggests that perhaps we need to take a broader perspective of cellular oxidants and antioxidants. Perhaps evolution laid the groundwork for redox regulation that has yet to be discovered.

GRANTS

This research was supported by National Science Foundation Grant No. IOS 1446310 (to K. R. Olson) and National Science Foundation Predoctoral Fellowship No. DGE 1313583 (to E. R. DeLeon).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Author contributions: E.R.D., Y.G., and K.R.O. conception and design of research; E.R.D., Y.G., E.H., M.A., N.A., A.D., and S.P. performed experiments; E.R.D., Y.G., E.H., M.A., N.A., A.D., S.P., and K.R.O. analyzed data; E.R.D., E.H., M.A., N.A., A.D., and K.R.O. interpreted results of experiments; E.R.D. and K.R.O. edited and revised manuscript; E.R.D., Y.G., E.H., M.A., N.A., A.D., S.P., and K.R.O. approved final version of manuscript; K.R.O. prepared figures; K.R.O. drafted manuscript.