Rapid and selective generation of H2S within mitochondria protects against cardiac ischemia-reperfusion injury

Mitochondria-targeted H2S donors are thought to protect against acute ischemia-reperfusion (IR) injury by releasing H2S that decreases oxidative damage. However, the rate of H2S release by current donors is too slow to be effective upon administration following reperfusion. To overcome this limitation here we develop a mitochondria-targeted agent, MitoPerSulf that very rapidly releases H2S within mitochondria. MitoPerSulf is quickly taken up by mitochondria, where it reacts with endogenous thiols to generate a persulfide intermediate that releases H2S. MitoPerSulf is acutely protective against cardiac IR injury in mice, due to the acute generation of H2S that inhibits respiration at cytochrome c oxidase thereby preventing mitochondrial superoxide production by lowering the membrane potential. Mitochondria-targeted agents that rapidly generate H2S are a new class of therapy for the acute treatment of IR injury.

The production of superoxide by the mitochondrial respiratory chain upon reperfusion of ischemic tissue is a key initiator of the oxidative damage that underlies IR injury [24][25][26]. Consequently, there is considerable interest in developing H 2 S-donors that protect against IR injury by decreasing mitochondrial oxidative damage [27][28][29][30]. Candidate protective mechanisms include free-radical scavenging by H 2 S [31][32][33][34] or via the reversible S-thiolation of protein cysteine residues to form a persulfide (r-SPSH) [35,36] that can prevent irreversible oxidative damage to cysteine residues and may enhance the protective activity of some proteins [37,38]. Alternatively, H 2 S is a reversible inhibitor of cytochrome c oxidase [39]. Thereby, H 2 S may lower the proton motive force, a major driver of mitochondrial superoxide production upon reperfusion following ischemia [24,25], but whether this contributes to its protection against IR injury is not known.
The mitochondria-targeted H 2 S donors AP39 and AP123 have also been developed [6,[40][41][42][43] (Fig. 1A-B). These compounds comprise the mitochondria-targeting lipophilic triphenylphosphonium (TPP) cation [44] coupled via a ten-carbon aliphatic linker to either an anethole dithiolethione moiety in AP39 [41,42] (Fig. 1A), or an hydroxythiobenzamide moiety for AP123 [45] (Fig. 1B). These AP39 and AP123 moieties spontaneously hydrolyse to release H 2 S [46][47][48][49][50]. Furthermore, the initial AP39 hydrolysis product RT01 hydrolyzes further to generate more H 2 S [43] (Fig. 1A). Due to the TPP component these molecules are rapidly concentrated several hundred-fold within mitochondria potentially leading to the local generation of H 2 S. These data were interpreted to suggest that the protective effects against IR injury of AP39, AP123 and RT01 are due to H 2 S release within mitochondria. However, to be effective, mitochondria-targeted H 2 S donors have to be taken up and deliver H 2 S rapidly and selectively within mitochondria during the first few minutes of reperfusion to counteract the oxidative damage caused by the burst of superoxide that occurs at the onset of reperfusion [24,25]. Thus, the time available clinically to reperfuse the ischemic tissue to treat heart attack or stroke is short. As rapid release of H 2 S in vivo within this timeframe was never confirmed [51], any acute protective effects of AP39 and AP123 against IR injury may be unrelated to H 2 S release.
Therefore, here we set out to develop a mitochondria-targeted agent that rapidly and selectively released H 2 S solely within mitochondria and could thus be administered upon reperfusion to prevent IR injury. Here we describe the development, assessment and mechanism of action of MitoPerSulf, a mitochondria-targeted molecule that rapidly releases H 2 S within mitochondria in vivo and is protective against cardiac IR injury when administered at reperfusion.

Design and synthesis of the rapid H 2 S releasing agent MitoPerSulf
To generate a molecule that rapidly and selectively releases H 2 S within mitochondria, we exploited the mitochondrial membrane potential-dependent accumulation of TPP cations, the chemistry of persulfides and the high mitochondrial concentration of protein and glutathione (GSH) thiols, which are particularly reactive due to the elevated matrix pH [52]. A mitochondria-targeted persulfide should react rapidly with intramitochondrial thiols to generate persulfides that react further with thiols to generate H 2 S and disulfides [10]. Due to its instability, we protected the persulfide by synthesizing it as a stable thioester with a benzoyl group, that will be rapidly removed by reacting with thiols within mitochondria. The rapid deprotection of the persulfide in vivo is essential for the timely generation of H 2 S. The persulfide benzoyl thioester enables this because the low pKa of the persulfide (~5.45) [53] makes it a good leaving group [54,55], as has been demonstrated previously [10]. To ensure rapid deprotection of the persulfide by thiol attack at the thioester carbonyl, rather than at the α-sulfur atom to form thiobenzoate and a mixed disulfide, we chose a penicillamine-based substituted tertiary persulfide that is sterically constrained at the α-sulfur atom [10]. By conjugating this moiety to a TPP cation via a five-carbon aliphatic linker we constructed a mitochondria-targeted penicillamine-based protected persulfide, Mito-PerSulf (Fig. 1C). The synthesis of MitoPerSulf involved modifying MitoNAP-SH, a late-stage intermediate used in the synthesis of MitoSNO [56] by converting it to a mixed disulfide with 2,2ʹ-dithiobis(benzothiazole) and then displacing the 2-mercaptobenzothiazole with thiobenzoic acid [10] (Fig. 1C). It is anticipated that the benzoyl thioester would first be rapidly cleaved by thiols within mitochondria, thus generating the unstable persulfide MitoNAP-SSH that should then transiently persulfidate mitochondrial thiols which then react further with other thiols to release H 2 S (Fig. 1D).

Activation of MitoPerSulf by glutathione in vitro
As GSH is the most abundant small molecule thiol within mitochondria, we assessed the activation of MitoPerSulf in vitro by reacting it with a 2-fold excess of GSH. This should be sufficient to activate Mito-PerSulf, while still allowing MitoNAP-SSH to persist for analysis (Fig. 2,  S2). We also used a 10-fold excess of GSH to better mimic the thiol concentration within mitochondria in vivo [57] (Fig. 2, S2). To trap the unstable thiol intermediates such as MitoNAP-SSH, we quenched the reaction with excess iodoacetamide (IAM) [58,59], followed by The mitochondria-targeting triphenylphosphonium group (TPP) leads to uptake of MitoPerSulf into mitochondria where the benzoyl thioester is cleaved by reaction with thiols to generate the unstable persulfide, MitoNAP-SSH that forms persulfides with mitochondrial thiols. These persulfides will then rapidly generate H 2 S and disulfides by reaction with other thiols. For simplicity, reactions with mitochondrial protein thiols are omitted and only reactions with GSH are shown. LC-MS/MS analysis to detect the carbamidomethylated (CAM) thiol adducts and other reaction products (Fig. S1). This analysis revealed the rapid formation of a benzoyl thioester of GSH that was complete within 1 min (GSCOPh; Fig. S2A). We also detected the uncapped persulfide MitoNAP-SSH as MitoNAP-SS-CAM, which was rapidly formed within 1 min and subsequently declined over time (Fig. S2B). These findings are consistent with the rapid activation of MitoPerSulf by thiols cleaving the benzoyl thioester to generate MitoNAP-SSH (Fig. 1D). Once formed, reaction of MitoNAP-SSH with other thiols (in this case GSH) could in principle occur at the α-sulfur to generate the disulfide MitoNAP-SSG with H 2 S release, or at the β-sulfur to generate MitoNAP-SH and glutathione persulfide (GSSH) (Fig. 2A). Formation of GSSH, detected as the GSS-CAM adduct, was rapidly generated in the presence of GSH and then declined over time (Fig. S2D), consistent with the initial formation of GSSH from MitoNAP-SSH that subsequently reacts with GSH to generate GSSG and H 2 S (Fig. 1D). The MitoNAP-SSG adduct also increased, albeit more slowly, over time (Fig. S2C), consistent with the subsequent disulfide exchange of MitoNAP-SH and GSSG. We also observed a slight increase in the MitoNAP-S-CAM adduct over time (Fig. S2E), while the GS-CAM adduct only decreased at the lower GSH concentration (Fig. S2F). The lag in formation of IAM adducts of GSSH relative to those of MitoNAP-SSH (Fig. S2G), upon reaction of MitoPerSulf with GSH are consistent with the early formation of MitoNAP-SSH, followed later by the formation of GSSH. Incubation of MitoPerSulf, with a 2-fold excess of GSH generated a little of the GSS-CAM adduct over time, measured as the GSS-CAM/GS-CAM ratio (Fig. S2H), but with a 10-fold excess of GSH there was no increase in GSS-CAM over time, consistent with the rapid reaction of GSSH with thiols. Only GS-CAM, and MitoNAP-S-CAM were observed when MitoNAP-SH was incubated with different concentrations of GSH (data not shown). The relative changes in all these species over time are shown in Fig. 2B and C. Together these data indicate that steric hindrance of the methyl groups prevents GSH reaction at the α-sulfur of MitoNAP-SSH, and that the main pathway is via attack of GSH on the β-sulfur ( Fig. 2A) [10].
Our hypothesis was that MitoNAP-SSH should react with thiols to generate free H 2 S. This was confirmed by assessing H 2 S diffusion through air to a lead acetate impregnated filter paper to form lead sulfide ( Fig. 2D and E). In contrast, the production of H 2 S by AP39, even in the presence of GSH, was negligible over this time scale ( Fig. 2D and E). Generation of H 2 S by MitoPerSulf in the presence of GSH was further demonstrated using an H 2 S electrode ( Fig. 2F and G). Again, the production of H 2 S by AP39 over this time scale was negligible, even in the presence of GSH ( Fig. 2F and G), consistent with its proposed mechanism as a slow-release H 2 S donor activated by hydrolysis [60]. Finally, we used the fluorescent probe WSP-5, in which a disulfide undergoes nucleophilic attack by HS − followed by cyclization to a fluorescent product [61]. Neither MitoPerSulf nor AP39 showed initial generation of H 2 S, but upon addition of GSH MitoPerSulf rapidly generated H 2 S, while AP39 did not (Fig. 2H).
The proposed reaction scheme for MitoPerSulf with thiols, illustrated using GSH, is shown (Fig. 3). In summary, the spontaneous production of H 2 S by MitoPerSulf and AP39 is very low, but in the presence of excess thiols, as occurs in vivo, MitoPerSulf rapidly generates H 2 S, while AP39 does not.

MitoPerSulf is taken up by mitochondria and cells rapidly forming H 2 S
To be an effective mitochondrial H 2 S-generating agent, MitoPerSulf has to be accumulated by mitochondria in response to the membrane potential (Δψ). Using a TPP-selective electrode we showed that Mito-PerSulf was accumulated by energized mitochondria and that the dissipation of Δψ with the uncoupler FCCP released the TPP-containing moiety (MitoNAP-SH) from the mitochondria (Fig. 4A). The Δψ-dependent uptake of MitoPerSulf by mitochondria was further confirmed by RP-HPLC analysis of mitochondria pelleted after incubation with MitoPerSulf (Fig. 4B). Only MitoNAP-SH was detected by HPLC following incubation of energized mitochondria with MitoPerSulf, consistent with reduction of MitoPerSulf to MitoNAP-SH by thiols within mitochondria (Fig. 4B). The effect of MitoPerSulf on respiration of isolated mitochondria showed that at high concentrations MitoPerSulf inhibited respiration, while the same concentration of MitoNAP-SH did not (Fig. S3A), suggesting that the effect of MitoPerSulf on respiration was most likely due to the generation of H 2 S, as is explored in detail later.
To investigate the generation of H 2 S within mitochondria, we next measured H 2 S release by MitoPerSulf when incubated with mitochondria in the presence of the fluorescent H 2 S sensor WSP-5 (Fig. 4C). This showed that when succinate was added to drive MitoPerSulf accumulation within mitochondria H 2 S production rapidly increased, but that addition of FCCP to prevent MitoPerSulf uptake blocked H 2 S generation. In contrast, AP39 did not generate H 2 S within mitochondria over this time scale. To examine whether MitoPerSulf can induce the formation of H 2 S within cells, we stably transfected mouse embryonic fibroblasts with a mitochondria-targeted version of the red fluorescent protein mScarlet and used the fluorescent H 2 S sensor SF7-AM, that tends to distribute evenly throughout the cell [62] (Fig. S3B). This showed the rapid and time-dependent formation of H 2 S from MitoPerSulf (Fig. S3B), but limited formation from AP39 over this time scale (Fig. 4D). Colocalization of the SF7-AM and mitochondrial matrix-targeted mScarlet signals showed that the H 2 S signal from MitoPerSulf was present in mitochondria ( Fig. 4D, inset), but also diffused throughout the cell (Fig. 4D). Together these data are consistent with rapid accumulation of MitoPerSulf within mitochondria where it generates H 2 S some of which may diffuse out to the rest of the cell.

MitoPerSulf metabolism within mitochondria
To analyze the interaction of MitoPerSulf with mitochondrial thiols we incubated isolated mitochondria with MitoPerSulf and then analyzed extracts by LC-MS/MS. This demonstrated the initial formation of the benzoylated GSH, GSCOPh, which then rapidly decreased (Fig. 4E). In order to increase the sensitivity of the LC-MS/MS detection for the lower amounts of MitoPerSulf metabolites being analyzed, we replaced IAM as the quenching reagent with IAM-TPP [63], an IAM derivative modified to incorporate a TPP cation. Trapping these species as X-CAM-TPP derivatives will introduce a fixed positive charge via the TPP moiety greatly enhancing detection sensitivity by MS (Fig. S1). Using this strategy, we demonstrated the initial formation of MitoNAP-SSH (detected as MitoNAP-SS-CAM-TPP) (Fig. 4F) and MitoNAP-SH (detected as MitoNAP-S-CAM-TPP) (Fig. 4G) within mitochondria. We also attempted to use IAM-TPP to detect GSSH (detected as GSS-CAM-TPP) within mitochondria incubated with MitoPerSulf, but the amounts detected were not significantly above baseline, consistent with the rapid metabolism of GSSH to H 2 S.
MitoNAP-SSH may also directly persulfidate protein thiols. To assess this possibility, we used recombinant Cofilin-1 protein in vitro, which contains 4 Cys residues (Fig. S4A), and is known to be persulfidated under certain conditions within cells [64]. We incubated Cofilin-1 with MitoPerSulf and GSH to generate MitoNAP-SSH and then assessed protein persulfidation by trapping with IAM, followed by trypsin digestion and LC-MS analysis to detect the persulfidated peptides (Fig. S4B). We detected two persulfidated peptides at Cys residues C39 and C139 in response to MitoPerSulf (Figs. S4C and D). We were not able to reliably detect persulfidation of cysteine residues C80 and C147. By comparing the relative amounts of the persulfidated Cys residues with those that were free to react with IAM we could estimate the extent of persulfidation as between 10 and 20% under these conditions (Figs. S4C and D). This suggests that MitoPerSulf can potentially lead to protein persulfidation. To assess if MitoPerSulf could lead to protein persulfidation within mitochondria, we incubated heart mitochondria with Mito-PerSulf under the same conditions as in Fig. 4 and then analyzed for protein persulfidation using a fluorescence tag switch method [38] followed by analysis of incorporated fluorescence after separation of proteins by SDS-PAGE. However, we did not find consistent increases in MitoPerSulf reacts with thiols (illustrated here solely with GSH) to rapidly form the persulfide, MitoNAP-SSH and GSCOPh. MitoNAP-SSH is further transformed by reacting with GSH to form GSSH and MitoNAP-SH. In the presence of excess GSH, GSSH forms H 2 S via formation of GSSG which can react with MitoNAP-SH to form the MitoNAP-SSG. fluorescent labelling of individual protein bands on the gels above control (Fig. S5). Furthermore, the negligible amounts of GSSH found when MitoPerSulf was incubated in vitro with excess GSH (Fig. 2E) and the lack of detection of GSSH within mitochondria incubated with MitoPerSulf make it likely that the majority of protein persulfides formed by MitoPerSulf are transient and react further to generate H 2 S. Together these data are consistent with the rapid but transient formation of persulfides from MitoPerSulf within mitochondria that rapidly react further with thiols to form H 2 S.

Distribution and cardioprotective effects of MitoPerSulf on acute IR injury in vivo
We next used an in vivo mouse model of cardiac IR injury to investigate the potential protective effects of MitoPerSulf. First, we analyzed the cardiac uptake of MitoPerSulf in vivo in mice following a bolus, intravenous tail vein injection of MitoPerSulf (0.2 mg/kg) with the tissue distribution analyzed by LC-MS/MS spectrometry. Tissues were reduced by addition of dithiothreitol (DTT) during extraction to convert any residual MitoPerSulf derivatives to MitoNAP-SH, thus data are reported as MitoNAP-SH content. As expected from similar TPP-based compounds [65], MitoPerSulf and any derivatives formed over this time scale were rapidly cleared from the plasma (Fig. S6A), leading to their rapid accumulation in the heart (Fig. S6B) as well as into the kidney and liver, with less penetration into the brain, followed by their gradual clearance from these tissues over time (Figs. S6B and C). Therefore, MitoPerSulf is taken up rapidly into the heart (Fig. S6B) following i.v. injection, making it suitable as a potential protective agent against cardiac IR injury for administration upon reperfusion.
Next, we assessed the protective effects of MitoPerSulf against Subsequently, samples were diluted (in cardiac IR injury by performing left anterior descending (LAD) coronary artery ligation in mice, followed by reperfusion and assessment of infarct size (Fig. 5A). Infusion of MitoPerSulf for 20 min starting 5 min before reperfusion resulted in a dose-dependent reduction of infarct size that reached a maximum at 10 μg/kg/min (Fig. S6D). Comparison of the most effective dose with the same concentration of MitoNAP-SH showed that MitoPerSulf was protective while MitoNAP-SH was not (Fig. 5B). As MitoNAP-SH is structurally very similar to MitoPerSulf and it is produced upon metabolism of MitoPerSulf within mitochondria this suggests that the protection against cardiac IR injury by MitoPerSulf is due to its rapid generation of H 2 S within mitochondria. Furthermore, as the uptake of MitoNAP-SH into mitochondria in vivo will be to a very similar extent as for MitoPerSulf, the protective effects of MitoPerSulf are not due to the disruption of mitochondrial function by the alkylTPP molecule. Acute protection against cardiac IR injury has been reported when AP39 is administered upon reperfusion [40,41,66]. We confirmed this protection here (Fig. 5C). The tacit assumption in these earlier publications was that the mode of action of AP39 was through H 2 S release in vivo, but this was not demonstrated. AP39 releases H 2 S far more slowly than MitoPerSulf within mitochondria (Fig. 4) making it unlikely that the protection against acute cardiac IR injury by AP39 is due to rapid H 2 S release and may instead be due to off-target effects. To explore this possibility, we made a chemically similar control version of AP39 that does not release H 2 S (Fig. 5D). AP39's reactive group is comprised of two planar highly conjugated rings capable of conjugation to each other at the oxygen atom of the ester. These rings are linked by a rotatable bond allowing other conformations (Fig. S7A). The planar 1,2-dithio-3-thione is weakly aromatic [67,68] and carbon and sulfur have very similar electronegativities. Therefore, we reasoned that a planar aromatic phenyl ring with the same number of heavy atoms would mimic its size, shape, and overall lipophilicity well (Fig. S7A). To confirm this the logP was calculated for the reactive head group of AP39 and the corresponding phenyl analogue using a consensus model built on Chemaxon and Klopman et al. [69] models using the PHYSPROP database (Fig. S7A). Calculating only the head group simplifies the calculation and avoids complications associated with the modelling of logPs of single ions [70,71]. The similarity of the logPs calculated for the head groups gave confidence that a control with the same TPP targeting group and alkyl linker would have similar physicochemical properties and thus uptake into mitochondria in vivo (Fig. S7B). RP-HPLC confirmed this similarity (Fig. S7C). The AP39 control compound was indeed as protective against cardiac IR injury as AP39 in the LAD model (Fig. 5C), further confirming that the protection afforded by AP39 is not due to the release of H 2 S, but to off-target effects, which may be due to accumulation of the hydrophobic alkylTPP molecule within mitochondria affecting organelle function. Of course, the slow release of H 2 S by AP39 may protect against tissue damage that occurs in the hours following reperfusion, but this was not explored here. In contrast, the protection against cardiac IR injury by MitoPerSulf, which rapidly releases H 2 S, but not by the chemically closely related compound MitoNAP-SH which does not release H 2 S, suggests that the rapid release of H 2 S within mitochondria in the heart is protective against IR injury.

Mechanism of protection by MitoPerSulf against acute cardiac IR injury
We next explored the mechanism of protection against cardiac IR injury by the rapid burst of H 2 S generation produced by MitoPerSulf within mitochondria. Mitochondrial superoxide production by reverse electron transport (RET) upon reperfusion is thought to initiate the damaging cycle that leads to tissue damage [24,25]. To explore whether H 2 S could alter this process, we investigated the effect of MitoPerSulf on superoxide production by RET in isolated mitochondria (Fig. 6). Addition of MitoPerSulf decreased respiration compared to control and this inhibitory effect of MitoPerSulf increased as the oxygen concentration diminished, thereby extending the time taken to remove all the oxygen from the incubation (Fig. 6A). In parallel, we measured the extent of superoxide production by RET through the generation of H 2 O 2 . In control mitochondria there was considerable H 2 O 2 generation that slowed as the oxygen level fell (Fig. 6B). Following anaerobiosis the fluorescence due to Resorufin decreased, due to its enzymatic reduction to dihydroresorufin upon anaerobic conditions [72], that is likely to disrupted by the presence of H 2 S [73]. In contrast, addition of MitoPerSulf greatly decreased H 2 O 2 generation, in parallel with its effect of on respiration (Fig. 6A). The control compound MitoNAP-SH had no effect on respiration (Fig. 6C), or on the generation of H 2 O 2 (Fig. 6D). Thus, the effect of MitoPerSulf on respiration and on the generation of H 2 O 2 is not due to any non-specific effects of the accumulation of the TPP cation on the mitochondria but instead is due to the generation of H 2 S within mitochondria. To investigate this further, we incubated mitochondria in the presence of H 2 S by adding Na 2 S, which had a very similar effect on mitochondrial respiration (Fig. 6E). The addition of H 2 S also slowed the rate of H 2 O 2 generation (Fig. 6F), compared to the control incubation. To better illustrate the effect of adding H 2 S on H 2 O 2 generation, we plotted the slope of the data in Fig. 6F against time, which showed that the rate of H 2 O 2 generation decreased immediately upon addition of H 2 S (Fig. 6G), while in contrast in the control incubation the rate of H 2 O 2 generation decreased gradually as the O 2 concentration decreased. These data suggest that the generation of H 2 S from MitoPerSulf within mitochondria disrupts respiration and thereby prevents mitochondrial superoxide production by RET.

Conclusions
The role of H 2 S donors as potential therapies has attracted considerable interest. In particular, it has been proposed that these donors could be used to prevent the damage associated with IR injury in heart attack and stroke by selective targeting to mitochondria. However, for the clinical treatment of IR injury it is necessary to add the protective agent upon reperfusion. While the targeting of compounds to mitochondria by conjugation to the lipophilic TPP cation is well established [44], the mitochondria-targeted H 2 S donors developed to date such as AP39 release H 2 S slowly, suggesting that any acute protective effects are not due to H 2 S release. Thus, the potential therapeutic utility of acute release of H 2 S within mitochondria remains unexplored. Here we addressed this by developing MitoPerSulf, a mitochondria-targeted H 2 S donor. We used a TPP cation to target MitoPerSulf to mitochondria in vivo, following intravenous administration. By adapting persulfide chemistry we were able to mask a reactive persulfide moiety that then rapidly releases H 2 S within mitochondria. This development opens the way for the development of further donors designed to rapidly release H 2 S within mitochondria.
Most importantly, we showed that MitoPerSulf was acutely protective in the in vivo LAD model of cardiac IR injury. In doing this, we utilized appropriate control compounds to show that the protective effects of MitoPerSulf were due to rapid H 2 S release and not to off-target effects of the mitochondria targeting TPP moiety. We also demonstrated, through the use of an appropriate control compound, that the reported protective effects of AP39 against IR injury were due to off-target effects resulting from the physicochemical properties of molecules that have a targeting TPP moiety linked by a long alkyl chain to a nonpolar biaryl system. Thus, for the first time we have demonstrated that the acute generation of H 2 S within mitochondria is a viable therapeutic strategy Fig. 6. Effect of MitoPerSulf on mitochondrial respiration and superoxide production by RET in vitro. Rat heart mitochondria RHM (1 mg protein/2 mL) were resuspended in KCl buffer in an Oroboros Oxygraph-2k system, and respiration and H 2 O 2 generation by RET were initiated by adding 10 mM succinate fol- against IR injury.
The mechanism of protection by acute H 2 S generation within mitochondria was also determined. H 2 S is well established to bind selectively and reversibly to cytochrome c oxidase and thereby to inhibit mitochondrial respiration. We showed that MitoPerSulf acted in this way by rapidly inhibiting respiration and that its inhibitory potency increased as the oxygen concentration decreased. This is consistent with the wellestablished competition between O 2 and H 2 S at cytochrome c oxidase. This inhibition of respiration will lower the mitochondrial protonmotive force and should thereby prevent the ability of mitochondrial complex I to generate superoxide by RET. We demonstrated this in isolated mitochondria with both MitoPerSulf and with pure H 2 S. Thus, we suggest that the protective effects of acute generation of H 2 S within mitochondria against IR injury is largely by preventing the burst of superoxide production by complex I upon reperfusion (Fig. 7). Even so, it is important to note that additional protective effects of H 2 S, such as by preventing overoxidation of protein thiols, are not excluded. The reversible inhibition of cytochrome c oxidase by H 2 S is similar to that by nitric oxide (NO) [56] and suggests that acute generation of NO within mitochondria may also be protective against IR injury by a similar mechanism. Indeed, in earlier work we developed a mitochon dria-targeted NO donor (MitoSNO) which was acutely protective against IR injury [56]. While we interpreted this as being due to the selective S-nitrosation of Cys 39 on complex I, thereby preventing RET, the degree of exposure of this Cys residue in vivo has been reassessed [63]. Thus, the protection against IR injury by MitoSNO may have been, at least in part, due to the reversible inhibition of cytochrome c oxidase decreasing respiration and thereby decreasing mitochondrial superoxide production at complex I upon RET.
In summary, we have developed the first approach to rapidly and selectively generate H 2 S within mitochondria in vivo. Using this approach, we were able to demonstrate that H 2 S is acutely protective against IR injury by reversibly inhibiting respiration at cytochrome oxidase and thereby preventing superoxide production at complex I.

Animals
All procedures were carried out in accordance with the UK Animals (Scientific Procedures) Act 1986 and the University of Cambridge Animal Welfare Policy. Procedures were approved to be carried out under the Project Licenses: 70/7963, 70/8238. Female Wistar rats, or male or female C57BL/6J mice (both Charles River Laboratories, UK) were maintained in pathogen-free facilities with ad libitum chow and water until being 8-20 weeks of age for experimental use.

Detection of H 2 S by amperometry
H 2 S release from MitoPerSulf and GSH was accessed by using H 2 S sensitive electrode. 5 mm fibre wire H 2 S microelectrode (WPI) was connected to Apollo 4000 free radical analyzer (WPI) and polarized overnight in 10 mM phosphate buffer saline (Gibco) under the 150 mV until reaching the stabile baseline. Amperometric traces in time were obtained by performing the reaction in 3 mL of 25 mM HEPES (pH 7.8) or 10 mM PBS buffer (pH 7.8) under constant and stable stirring and temperature in a multi-port reaction chamber (WPI). Reaction was performed by injecting the various concentration of GSH (0-1 mM) followed by injecting boluses of different concentrations of MitoPerSulf, MitoNAP-SH or AP39 (0-100 μM) from DMSO-based stock solutions into the reaction chamber. Results were obtained by measuring the difference of maximum signal obtained before and after the injections (p (A) max -p(A) min = Δp(A)) for each experimental condition. The H 2 S electrode was calibrated using 25 mM HEPES buffer (pH 7.8) and anaerobically prepared solutions of anhydrous and ultra-pure Na 2 S (Sigma Aldrich Product. Code. 407410) in Chelex-100 treated and argon-purged MiliQ dH 2 O prepared and used at the same day.

Detection of diffusible H 2 S by the lead acetate assay
Release of hydrogen sulfide in the gas phase was assessed using lead (II) acetate [75]. Lead (II) acetate-impregnated filter paper was prepared by soaking clean sheets of Whatman filter paper (# 3030-917) in 20 mM lead (II) acetate in dH 2 O for 20 min and drying them for 2 h at 50 • C. Upon drying, lead acetate impregnated paper was stored protected from light at room temperature in a dry and sealed glass container. In brief, 100 μL of reaction mixture containing 100 μM of MitoPerSulf, MitoNAP-SH or vehicle (EtOH) and different concentrations of GSH ranging from 0 to 1 mM in 25 mM HEPES buffer (pH 7.8) was placed in 96-well plate and covered with lead (II) acetate-impregnated filter paper leaving approximately 5 mm of head space between liquid phase and the filter paper. 96-well plate with samples was incubated at 50 • C in the oven for 2 h to allow efficient evaporation and accumulation of H 2 S in the head space of well plate and after the incubation the filter paper containing developed lead (II) sulfide spots was immediately scanned using bio scanner (HP) and analyzed by densitometry (ImageJ).

Generation of lentiviral particles and transduction of MEFs
MTS-Scarlet was amplified by PCR with specific oligonucleotides using pMTS_mScarlet_N1(Addgene; #85057) plasmid. This insert was introduced into the pWPXLd-IRES-HygroR lentiviral expression vector, modified versions of pWPXLd (Addgene; #12258), by restriction enzyme digestion with PmeI and BamHI and ligation with T4 DNA ligase (New England Biolabs). Lentiviral particles were generated in HEK293T packaging cells by co-transfection of the lentiviral expression vector with the packaging psPAX2 (Addgene; # 12260) and envelope pMD2.G (Addgene; # 12259) vectors with FuGENE HD (Promega) according to manufacturer's instructions. Mouse embryonic fibroblast cells (MEFs) were transduced with previously generated lentiviral particles with Polybrene (Merck, TR-1003) for 24 h. Transduced cells were then selected for resistance using hygromycin B (Roche, 10843555001) at 50 μg/mL.

Detection of H 2 S by fluorescent microscopy
MEFs stably expressing the fluorescent mitochondrial matrix red protein, MTS-mScarlet were grown in high glucose glutaMAX containing DMEM medium supplemented with 10 % FBS, 1 % Streptomycin-Penicillin solution and at 37 • C under the atmosphere of 5 % CO 2 . Upon reaching the 80 % confluency, cells were detached using 0.25 % trypsin and plated in glass bottom 35-mm high μ-Dish (ibidi, Germany) at 3 × 10 4 cells per dish. After attachment cells were stained with 2.5 μM SF7-AM in complete cell medium for 40 min in dark at 37 • C under the atmosphere of 5 % CO 2 . After staining, cells were washed three times with phenol red-free full DMEM and mounted on the microscope stage.
For some experiments cells were washed and imaged using phosphate buffer saline (PBS). 180 images per sample were obtained during the 900 s of live cell imaging (integration time: 5 s) at 37 o C and stimulation of H 2 S production was initiated by adding the boluses of 20 μM Mito-PerSulf or AP39 directly into dishes 10 s upon starting the time-lapse video recording. Fluorescence values were collected every 5 s for 15 min. Images were acquired using a 100x objective of the Nikon Eclipse Ti-E microscope, coupled to an Andor Dragonfly spinning disk confocal system equipped with an Andor Ixon camera, and 488 nm and 561 nm excitation lasers were used for SF7-AM and MTS-mScarlet, respectively. All images were postprocessed under the same parameters using ImageJ software (NIH) and for enhanced visualisation the original SF7-AM fluorescence was presented using the specific heat map projection of signal (ImageJ).

Mitochondria preparations
Rat liver and heart mitochondria (RLM and RHM respectively) were prepared by homogenization of heart tissue obtained from 10 to 12 weeks old Female Wistar rats (Charles River, UK) that were killed by stunning and cervical dislocation, in STEB buffer (250 mM sucrose, 5 mM Tris-HCl and 1 mM EGTA, pH 7.4). Following homogenization, mitochondria were isolated by differential centrifugation (2 x 2450 × g for 5 min, 2 x 9150 × g for 10 min at 4 • C). STE buffer was supplemented with 0.1% fatty acid-free BSA for isolation of RHM. Protein concentration was determined by the bicinchoninic acid (BCA) assay using BSA as a standard.

Mitochondrial uptake of MitoPerSulf
Mitochondrial uptake of MitoPerSulf was assessed using TPPselective electrode. The electrode was calibrated with five boluses of 1 μM MitoPerSulf followed by 1 mg/mL of RLM in KCl buffer (120 mM KCl, 10 mM HEPES, 1 mM EGTA, 1 mM MgCl 2 and 5 mM KH 2 PO 4 , pH 7.4). 5 mM succinate was then added to energize RLM after which, the H + /K + ionophore nigericin (0.5 μM) and the uncoupler carbonyl cyanide p-trifluoromethoxy-phenylhydrazone (FCCP, 0.5 μM) were added to maximize and collapse the mitochondrial membrane potential, respectively. The uptake was also analyzed with RP-HPLC analysis. RLM were incubated with 5 mM succinate and 4 μg/mL rotenone and 5 μM MitoPerSulf in KCl buffer ± 0.5 μM FCCP to collapse the membrane potential respectively. Compounds in the mitochondria were extracted from mitochondrial pellet with mixture of 20 % Acetonitrile/0.1 % TFA in water (v/v) after 3 min incubation period and detected by RP-HPLC as described above.

Mitochondrial respiration and superoxide production in vitro
Oxygen consumption and superoxide production were determined using the high resolution O2k oxygraph (Oroboros Instruments). Freshly isolated RHM (1 mg protein) were resuspended in 2 mL of KCL buffer supplemented with 17.6 U SOD, 8.76 U HRP, 12.5 μM Amplex Red and 3 μM BSA and oxygen consumption and superoxide production were induced by simultaneous addition of 10 mM of succinate in each chamber under the constant stirring and constant temperature (T = 37 • C). After 1 min, the indicated compounds and control (EtOH) were added and recording of amperometric and fluorescence changes was continued for 25 min. Obtained results of all measurements are presented as means ± s.e.m. of n = 3, repeated on 4 different occasions.

Pharmacokinetics analysis
200 μg/kg MitoPerSulf in 100 μL of saline was administered by tailvein injection in Wild-type male C57BL/6 mice. Tissues were collected after respective time periods, frozen in liquid nitrogen and then stored at − 80 • C. MitoPerSulf and its derivatives inside the tissues were reduced to MitoNAP-SH by addition of 0.3 M DTT during the procedure and then MitoNAP-SH, in homogenate was extracted with 0.1 % TFA/acetonitrile and its amount was analyzed by LC-MS/MS as described.

MS method development for detection of reaction metabolites
The mass spectrometric fragmentation patterns for reaction intermediates/metabolites of MitoPerSulf were determined in samples from in vitro kinetic experiments of MitoPerSulf in the presence of GSH. Samples were quenched with IAM or TPP-IAM and prepared as follows.  TPP-IAM quenching. For MS/MS analysis, a triple-quadrupole mass spectrometer was used (Waters Xevo TQ-S under positive ion mode: source spray voltage, 2.6 kV; ion source temperature, 150 • C). Nitrogen and argon were used as curtain and collision gas, respectively. For LC-MS/MS analyses the mass spectrometer was connected in series to an I-Class ACQUITY UPLC system (Waters). Samples were stored in an The following MS settings were used for the MRM detection of the individual compounds.

LC-MS/MS characterization of in vitro reaction products
To analyze the reaction in time, the reaction mixture of 100 μM of Both MeOH and CHCl 3 layers were removed from protein precipitates (protein disc in between two liquid phases) and the residual liquid was evaporated on air leaving the precipitated and labelled proteins at the bottom of the tubes. Precipitated proteins were dissolved in 50 μL of 50 mM ammonium bicarbonate buffer pH 7.8 containing 1 mM CaCl 2 and 12.5 ng/μL trypsin and digested overnight at 37 • C.
Peptides were resuspended in 3 % ACN, 0.1 % TFA buffer and portions were fractionated by liquid chromatography on a Biosphere C18 reversed-phase column, 75 μm inner diameter, 100 mm length (Nanoseparations, Nieukoop, Netherlands) in a Proxeon EASY-nLC II system using Buffer A (0.1 % formic acid, 2 % acetonitrile) and Buffer B (98% acetonitrile, 0.1 % formic acid) and a gradient of 2-35 % B over 84 min at a flow rate of 300 nL/min, followed by an increase in acetonitrile concentration to 90 % B over 5 min and re-equilibration with 2 % B within a total time of 102 min. The eluate was transferred in-line to a LTQ Orbitrap XL ETD mass spectrometer (Thermo Scientific, UK).
Peptides were analyzed by positive ion electrospray mass spectrometry in a data-dependent acquisition mode. Up to ten of the most abundant precursor ions with multiple charge states, were selected and fragmented by CID each second. The m/z values of precursor and up to 10 fragment ions were measured simultaneously in the Orbitrap (400-2000 m/z scan, resolution of 60 000) and ion-trap analyzers, respectively. A lock mass ion (polysiloxane, m/z = 445.1200) was used for internal MS calibration. For protein identification the fragment patterns were compared to the UniProt database using the Mascot search engine with Proteome Discoverer (v1.4) software (Thermo Scientific). Relative quantification was performed by comparing the peak area of XICs (extracted ion chromatograms) for the monoisotopic peak using Xcalibur software (Thermo Scientific).

Tag switch assay
Detection of protein persulfidation was performed by using the dimedone-based tag-switch method as reported previously [38] with modifications. In brief, 1 mg of RHM proteins were incubated in 2 mL of KCl buffer with 10 μM MitoPerSulf, MitoNAP or vehicle (EtOH) in the presence of 10 mM succinate and 4 μg/mL rotenone for 5 min at 37 • C.
Subsequently, pelleted mitochondria (1 min at 17 000 g) were resuspended in 50 μL of HENS buffer (50 mM HEPES, 1 mM EDTA, 0.1 mM Neocuproine, 1 % NP-40, 2% SDS and protease inhibitor cocktail, pH 7.4) supplemented with 5 mM 4-chloro-7-nitrobenzofurazan (NBF-Cl, from 500 stock solution in DMSO) and incubated at 37 • C for 30 min in dark. Proteins were retrieved using methanol/chloroform precipitation (H 2 O/MeOH/CHCl 3 : 4/4/1) and obtained protein pellets were resuspended using ultrasonication in 50 mM HEPES buffer (pH 7.4) containing 1 % SDS. Protein concentration was determined by BCA assay and 1 mg of protein were labelled with 25 μM of Daz-2-Cy5 alkyne click master mix [38] for 30 min at room temperature in the dark. After labelling, protein pellets obtained using methanol/chloroform were resuspended using ultrasonication in 50 mM HEPES buffer (pH 7.4) containing 1 % SDS and equal amount of protein (approximately 50 μg/sample) were resolved using standard Laemmli reducing 10% SDS PAGE. After electrophoresis, gel was fixed in the dark for 30 min, washed and equilibrate with dH 2 O and scanned using Typhoon FLA 9500 fluorescent scanner (Cy3 and Cy5 fluorescence was recorded using 473 and 635 nm filter sets). Obtained raw images were post processed using ImageJ software.

LAD ligation model
We used an open-chest, in situ mouse cardiac infarction model as recently described (Prag et al., 2022). Briefly, Wild-type male C57BL/6J mice (8-10 weeks of age; Charles River Laboratories, UK) were anethetized with sodium pentobarbital (70 mg per kg of body weight intraperitoneally (i.p.)), intubated endotracheally and ventilated with 3 cm H 2 O positive-end expiratory pressure. We monitored the adequacy of the anesthesia using corneal and withdrawal reflexes, and additional anesthesia was administered as needed throughout the experiment. We kept the ventilation frequency at 240 breaths per minute with a tidal volume between 125 μL and 150 μL. We performed a small thoracotomy, and the heart was exposed by stripping of the pericardium. All hearts underwent 30 min of regional ischemia by ligation of a main branch of the left coronary artery. We introduce MitoPerSulf or MitoNAP-SH (100 ng per kg body weight each) 10 min before reperfusion as a slow infusion intravenously into a tail vein over 20 min.
We assessed infarct size after 120 min of reperfusion using triphenyltetrazolium chloride (TTC) staining and expressed it as a percentage of the risk zone as described previously (Prag et al., 2022). For various experiments on treated tissues, we removed the left ventricle at various time points after reperfusion, as indicated in the corresponding Fig. legends.

Statistical analyses
Error bars represent the s.e.m. from at least three replicates unless otherwise stated. We quantified P values using Student's t-test or oneway ANOVA. Values of P < 0.05 was considered as statistically significant.

Declaration of competing interest
Authors have no conflict of interest to declare.

Data availability
Data will be made available on request.