Improved tag-switch method reveals that thioredoxin acts as depersulfidase and controls the intracellular levels of protein persulfidation

H2S signals via protein persulfidation. To be regulatory the modification will have to be reversible. Using a new method for persulfide detection, we discover this missing link and show that thioredoxin system acts as depersulfidase in vivo.


Introduction
Oxidative posttranslational modications (oxPTM) of protein cysteine residues are important for regulation of diverse cellular functions. 1 From S-nitrosation 2,3 and S-glutathionylation 4,5 to S-sulfenylation, 1,6 these modications have been linked to signalling by reactive nitrogen and oxygen species. The recent emergence of a new signalling molecule, hydrogen sulde (H 2 S), has provided insights into another oxPTM of cysteine, persuldation (also called S-sulydration). 7-10 Thus far, persuldation of a handful of proteins has been demonstrated to be important for the regulation of blood pressure (K ATP channel), 11 cell senescence (Keap-1), 12 and cell apoptosis (GAPDH and NF-kB). 7,13 Additionally persuldation of parkin increased its activity, suggesting that pharmacological H 2 S donors might be benecial for the treatment of Parkinson's disease 14 and possibly other neurodegenerative diseases.
Intracellular protein disulde and S-glutathionylation levels are controlled by the thioredoxin (Trx) system. [15][16][17] The enzymatic system, consisting of Trx, thioredoxin reductase (TrxR) and NADPH, represents the main disulde reductase system in cells. Trx also regulates S-nitrosation. 18,19 Numerous studies have shown that Trx acts as a denitrosylase and is also involved in selective trans-nitrosation (i.e. the transfer of an "NO + " moiety from an S-nitrosothiol to another cysteine residue). [20][21][22] The latter is best illustrated by Trx-mediated caspase 3 S-nitrosation, which prevents apoptosis. 20 To exert a regulatory function similar to that of phosphorylation/dephosphorylation or S-nitrosation/denitrosation, S-per-suldation levels must be enzymatically regulated. 9,10 Due to the meta-stability of persuldes and their higher nucleophilicity (and therefore reactivity) compared to thiols, it is challenging to study their biochemistry. The lack of selective tools for labelling persuldes additionally complicates these studies. 23, 24 We have recently developed a tag-switch method for selective persulde labelling 25,26 and assessed alternative mechanisms for persul-de formation, suggesting that persuldes could be formed from sulfenylated cysteine residues. 27 Some evidence already exists to suggest that Trx can cleave persulde from the active site of mercaptopyruvate sulfur transferase (MST) leading to H 2 S release. 28,29 To asses the global role of Trx system on protein persuldation, we herein developed an improved tag-switch assay and showed that the Trx system reacts approximately an order of magnitude faster with persuldes than their disulde analogs, and is the major regulator of protein persulde levels in the cells.

Results and discussion
To assess the potential role of Trx in persulde reduction, we rst used the low molecular weight (LMW) persulde analogue of S-nitrosopenicillamine (NAP-SSH) 30 (Fig. 1A). NAP-SSH spontaneously decomposes when added in the buffer, as some of us demonstrated previously. 30 The reaction was monitored by time-resolved mass spectrometry. NAP-SSH is metastable and its decay was monitored by the disappearance of the parent ion peak (at m/z 240.0347; calculated 240.0359 Fig. 1B and S1 in ESI †), and the appearance of peaks with m/z 208.0623 (calculated 208.0638) and 225.0883 (calculated 225.0904) corresponding to [NAPSH + H] + and [NAPSH + NH 4 ] + , respectively ( Fig. 1C and S1 in ESI †). The half-time for the spontaneous decay of 50 mM NAP-SSH in ammonium carbonate buffer pH 7.38 was estimated to be 2.7 AE 0.3 min and this is in good agreement with previous chemical studies on its stability. 30 The    Fig. 1D and E, respectively. Complete oxidation is demonstrated by the leward m/z 2 shi (Fig. 1F). Two reaction mechanisms are possible to explain the reaction: (i) transfer of the sulfane sulfur from NAP-SSH to the nucleophilic cysteine of Trx leads to the transient formation of Trx-SSH followed by the displacement of the sulde anion by the resolving cysteine and formation of a disulde bond, or (ii) a nucleophilic attack of one of the Trx cysteines to the sulfane sulfur of NAP-SSH with immediate elimination of H 2 S and formation of Trx-NAPSH disulde complex, followed by the displacement of NAPSH by the resolving cysteine and formation of a disulde bond. Either mechanism could result in H 2 S generation. Therefore we examined H 2 S production from 100 mM NAP-SSH treated with 10 mM Trx using an H 2 Sselective electrode. While a H 2 S-specic current was not detected with either Trx or NAP-SSH alone (the latter decomposes spontaneously to give a mixture of tri, tetra and pentasuldes of NAPSH), 30 the addition of NAP-SSH to Trxcontaining buffer induced an immediate electrode response ( Fig. S3 in ESI †) indicative of H 2 S formation. Although Trx showed activity towards LMW persuldes, protein-bound persuldes are expected to represent a larger sulfane sulfur pool. Therefore, we tested the reaction of Trx with protein persuldes. We recently reported preparation of human serum albumin persulde (HSA-SSH) and described its reactivity. 27 Addition of Trx (E. coli) to a solution containing 20 mM HSA-SSH led to a concentration-dependent increase in the initial rate of H 2 S production, as monitored by an H 2 S-sensitive electrode ( Fig. 2A), conrming our hypothesis that Trx reduces protein persuldes in addition to LMW persuldes. Using an initial rate approach, a second order rate constant for the reaction of Trx with HSA-SSH was estimated to be 4.1 AE 0.8 Â 10 3 M À1 s À1 (Fig. 2B). MS analysis of the reaction mixture containing 20 mM Trx and 40 mM HSA-SSH revealed complete oxidation of Trx ( Fig. S4 in ESI †) as observed with NAP-SSH (Fig. 1F).
In cells, the disulde bond in the active site of Trx is reduced by TrxR, which uses NADPH as an electron source. The catalytic cycling of Trx between the oxidized and reduced states was observed when 2 mM HSA-SSH was added to a solution containing substoichiometric Trx (0.5 mM) in the presence of TrxR and NADPH (0.007 and 250 mM), which led to a 95 AE 3% recovery of H 2 S from HSA-SSH (Fig. 2C).
We next assessed the kinetic parameters for the depersuldation of HSA-SSH by the Trx/TrxR system. The dependence of the initial rate of NADPH consumption on the concentration of HSA-SSH exhibited Michaelis-Menten-like kinetic behaviour (Fig. 2D) yielding the apparent K m and V max parameters shown in Table 1. The apparent k cat /K m value is similar to that reported for insulin reduction by Trx, 31 which is one of the best substrates for the enzyme, indicating that the Trx/TrxR system can also readily cleave protein persuldes.
A recent study indicated the existence of high levels of intracellular and circulating LMW cysteine and glutathione persuldes and suggested that uncatalyzed trans-persuldation reactions between LMW persuldes and cysteine residues lead to protein persuldation. 32 This study also showed that the H 2 Sproducing enzymes, g-cystathionase (CSE) and cystathionine bsynthase (CBS), produce large amounts of cysteine persulde (Cys-SSH/Cys-SS À ) from cystine (CysSSCys) (Fig. 2E). We therefore tested the reactivity of Trx towards Cys-SSH. Recombinant human CSE was incubated with CysSSCys at 37 C for 15 min before separating the reaction mixture from the protein.
Approximately 50% of the substrate was converted to product during this time as assessed using Ellman's reagent. Trans-persuldation kinetics was initially monitored by spectro-uorimetry 31 measuring the tryptophan uorescence changes caused by Trx oxidation (Fig. 2F). The dependence of the k obs of the uorescence on the concentration of either CySSCys or the CySSCys/CysSS À mixture, yielded rate constants for the oxidation of Trx by CySSCys and CysSS À of 5 AE 1 Â 10 2 M À1 s À1 and 4.5 AE 0.1 Â 10 3 M À1 s À1 , respectively (Table 1, Fig. S5 in ESI †). When the Trx/TrxR/NADPH system was used, Michaelis-Menten-like kinetic behaviour was observed ( Fig. 2G) with the CysSS À /CysSSCys mixture yielding $2-fold higher V max than CySSCys alone. Knowing the actual concentrations of both CysSS À and CysSSCys in the reaction mixture, we used a competitive two-substrate equation 33 to estimate the kinetic parameters for CysSSH (Table 1). This analysis revealed that the Table 1 Kinetic parameters for the Trx/TrxR/NADPH-catalysed depersulfidase reaction. 1 mM Trx, 0.01 mM TrxR and 250 mM NADPH were used for the reaction. Kinetic parameters for CysSS À from the mixture of CysSS À and CysSSCys were derived using the equation given by Pocklington and Jeffery, 1969 (ref. 33) Substrate  k cat /K m for CySS À (3.9 Â 10 5 M À1 min À1 ) is almost an order of magnitude greater than for CysSSCys (0.5 Â 10 5 M À1 min À1 ). It is important to note that CysSSH, like NAP-SSH, is intrinsically unstable in solution and it equilibrates to give a mixture of cysteine polysuldes. Therefore, the kinetic values reported here for CysSSH are likely to be an underestimation.
One of the main problems for studying cellular persuldation is the lack of selective tools for detecting protein persul-des. Several methods have been described to date including the tag-switch assay. 7,13,25 This latter method is based on the strategy of forming a mixed disulde using methylsulfonyl benzothiazole (MSBT), an aromatic thiol-blocking reagent. 25 The resulting mixed disulde is more reactive towards the biotin-tagged cyanoacetic acid nucleophile than towards cysteine disuldes present in proteins (Fig. 3A). The method shows selectivity for persuldes without detectable cross-reactivity with other post-translational modications of cysteine that have been tested. 25,26 The original assay used a biotinylated cyanoacetic acid tag, which requires Western-blot transfer and streptavidin or antibodies for visualization. 25,26 To increase sensitivity, we synthesized two new cyanoacetic acid derivatives with the uorescent BODIPY moiety (CN-BOT) or the Cy3-dye (CN-Cy3) (Fig. 3A, Scheme S1-S3 in ESI, † for details see Experimental section in ESI †). Both new tags labelled HSA-SSH, yielding uorescent products ( Fig. S6 in ESI †). Preliminary studies using cell lysates showed that the CN-Cy3 tag gave better resolution of labelled bands and higher uorescence intensity than the CN-BOT tag. However, when xed cells were used, CN-Cy3 proved difficult to wash out (visible colouring of the washing solution was noticed even 1 week aer washing). Therefore CN-BOT was used for labelling cells for microscopy and CN-Cy3 for the labelling in cell lysates.
The increase in sensitivity could be best illustrated by comparing CN-Biotin with CN-Cy3 labelling, as shown in Fig. S7 in ESI. † We used human erythrocytes and homogenates of Drosophila melanogaster heads. Although persuldation was detected by Western blot in the samples labelled by CN-Biotin, stronger signals as well as higher numbers of protein bands were detected in samples labelled by CN-Cy3, demonstrating signicant improvement of the sensitivity (Fig. S7 in ESI †).
Brain, heart, kidney and lung tissue extracts from 3C57BL/6 CSE +/+ and CSE À/À mice 34 were also labelled with CN-Cy3. Lower signal intensity indicative of lower total protein persuldation was observed in all tissue extracts from CSE À/À mice conrming the method's selectivity, with the effects being the strongest in lungs and heart and only modest in brain, which is consistent with the predominant roles of CBS and mercaptropyruvate sulfur transferase (MST) in brain (Fig. 3B).
Protein persuldation in xed cells was visualized by confocal microscopy using the tag-switch assay and the CN-BOT reagent. As a negative control, cells were only treated with CN-BOT without the initial MSBT blocking step, and as a positive control, cells were pre-treated with H 2 S (1 h with 100 mM Na 2 S). The signal intensity was barely detectable when CN-BOT was used alone, and 20-fold greater laser intensity was needed to observe a signal ( Fig. S8A and B in ESI †). Treatment with Na 2 S led to a substantial increase in uorescence intensity. Further validation of the CN-BOT tag-switch assay was obtained from experiments in which the CSE inhibitor propargylglycine (PG) 25 and CBS inhibitor aminooxyacetic acid (AOAA) were used. 35 Both inhibitors decreased the signal intensity in bovine arterial To further validate the microscopy results, we transfected mitochondria with RFP fused to the leader sequence of E1 alpha pyruvate dehydrogenase (CellLight®, Thermo Fischer Scientic) and treated the cells with AP39, a novel mitochondria targeted H 2 S donor. 37-39 Based on co-localization studies, a signicant portion of persuldation appeared to be localized in mitochondria in the presence or absence of 100 nM AP39 (Fig. 4A and B). AP39 induced a strong increase in intracellular persuldation, $2-fold higher than observed with 1 mM D-cysteine (Fig. 4C), suggesting that this compound can be a powerful pharmacological tool for modulating mitochondrial and intracellular H 2 S-mediated persuldation.
H 2 S cannot react directly with cysteine thiols, and a plausible mechanism for persulde formation is the reaction of H 2 S with sulfenic acids or with disuldes. 1, 10 We recently characterized the kinetics of these two reactions although a priori, the reaction of H 2 S with disuldes is less likely to be important in the cellular milieu, which is reducing. 27 Cells treated with H 2 O 2 showed increased persuldation, relative to control cells, and that could have been formed via either mechanism ( Fig. 5A and B). To distinguish between these mechanistic possibilities, we treated cells with the thiol oxidants, diamide (which induces disulde bond formation) and H 2 O 2 (which induces sulfenic acids in addition to disul-des). While H 2 O 2 -treatment of SH-SY5Y cells resulted in a signicant increase in uorescence intensity indicative of protein persuldation, treatment with diamide led to a slight decrease in intensity compared to untreated controls ( Fig. 5A and B), conrming that H 2 S-induced persulde formation would be a consequence of its reaction with sulfenic acids and not with protein disuldes.
During optimization of the labelling protocol, we observed that persuldation levels in cell lysates obtained using native conditions (HEN buffer pH 7.4, 1% protease inhibitor cocktail, 0 C), decreased with incubation/extraction time ( Fig. 6A and S10 in ESI †). We hypothesized that the decrease in persuldation levels could be due to the activity of the Trx/TrxR/NADPH system during sample preparation. To test this hypothesis, we added 2 mM auranon, an inhibitor of the Trx/TrxR/NADPH system, [40][41][42] to the native lysis buffer, which led to preservation of the persuldation levels (Fig. 6A). Furthermore, addition of Trx/TrxR/NADPH to the cell lysate led to immediate H 2 S release, as detected by the H 2 S electrode (Fig. 6B).
Next, we assessed how inhibition of the Trx/TrxR system affected intracellular protein persuldes levels. Cells (SH-SY5Y and BAE) were treated with auranon (2 mM, 1 h) and then labelled with CN-BOT. Incubation with auranon led to 1.8 AE

(BAE cells) and 2.1 AE 0.3 (SH-SY5Y cells) fold increases in
uorescence intensity ( Fig. 6C and F) compared to control, which was enhanced further by combining auranon with H 2 S. An overall increase in intracellular persuldation was also observed in cell lysates (Fig. 6D, E, G and H) and the quantication of in-gel uorescence signal matched nicely with the data observed by uorescence microscopy.
Finally, we used an isotope dilution method mass spectrometry for quantication of the effect of auranon inhibition of TrxR on intracellular persulde levels. This method is based on the reaction of triphenylphosphines with sulfane sulfur (Fig. S10 in ESI †). 43 BAEC lysates were prepared with probe 1 being added directly to the lysis buffer together with probe 2 (Fig. S10 in ESI †). Following protein precipitation, acetonitrile extracts were analysed by MS (Fig. S10 in ESI †). As observed using the tag-switch assay, sulfane sulfur levels increased in cells treated with auranon with a total value of 9 AE 2 pmol mg À1 protein in control cells and 22 AE 5 pmol mg À1 protein in auranon-treated cells (Fig. 7A).
The increased Trx levels is associated with certain disease states, such as rheumatoid arthritis, hepatitis C or HIV-1 infections. [44][45][46] HIV-1 patients with high viral load have increased levels of circulatory Trx. 44, 45 We used plasma samples from HIV-1 patients with high viral load and compared their sulfane sulfur levels with those from HIV-1 patients treated with antiretroviral therapy (ART) that contain no detectable viral load. Signicantly lower total sulfane sulfur levels were detected in patients with high viral load (and therefore high circulatory Trx levels) 44,45 compared to ART-treated patients, conrming the role of Trx as a putative depersuldase (Fig. 7B). Furthermore, there is a positive correlation between the viral load and the level of circulatory sulfane sulfur levels: the higher the viral load the lower the sulfane sulfur levels (Fig. S11 in ESI †). Although the actual role of lower persulde levels in those HIV-1 patients awaits further investigation, this is the rst in vivo evidence that Trx acts as depersuldase.

Conclusion
In conclusion, our work identies Trx as a major regulator of intracellular persuldation. Surprisingly, Trx showed almost an order of magnitude higher reactivity towards persuldes than towards disulde analogs. The high reactivity of Trx towards LMW persuldes raises questions about the unusually high levels of LMW persuldes (>100 mM) recently reported. 32 The identication of Trx as a depersuldase provides a new perspective on persuldation as an oxPTM of cysteine by revealing an enzymatic mechanism for reversing this posttranslational modication.
The improved tag-switch assay for facile persulde labelling and subsequent in-gel detection or visualization by microscopy led to the observation that a signicant portion of persuldation is apparently localized in mitochondria. The use of the mitochondria targeted H 2 S donor, or the MST substrate Dcysteine, dramatically increased intracellular levels of persuldation implying an important role for MST and mitochondriaderived H 2 S in protein persuldation. Under oxidative stress conditions, formation of sulfenic acids is triggered which can react with H 2 S to form protein persuldes. These conditions also favour inhibition of the Trx system further prolonging the half-life of protein persuldes. The actual consequences of those steps remain to be elucidated. On one hand such a mechanism could be protective, as overoxidized persuldated cysteine residues (P-SSO 3 2À ) 23,25 could be reduced by Trx system (similar like in the case of 3 0 -phosphoadenosine-5 0 -phosphosulfate reductase). 47 Overoxidation of cysteine residues to form P-SO 3 2À , however, remains irreversible. On the other hand, excessive persuldation could be inhibitory for some enzymes and prove detrimental in certain disease states. The improved tag-switch method for persulde detection and herein reported observation that Trx regulates persulde levels provide a good starting point for the future studies which will answer those questions and elucidate the actual mechanisms of H 2 S signalling.

Preparation of cysteine persulde
Recombinant human CSE (polymorphic variant S403) was expressed and puried as described previously. 49 Twenty mg of CSE was added to 100 mM HEPES pH 7.4 containing 1 mM Lcystine. The mixture was incubated at 37 C for 15 min and the enzyme was separated by ultraltration (centrifugation at 14 000 Â g at 4 C for 15 min) using a centrifugal lter with a molecular weight cutoff of 10 kDa (Carl Roth, Karlsruhe, Germany). Cysteine persulde in the ltrate was quantied with the DTNB method, using an extinction coefficient 3 412 of 14 150 M À1 cm À1 . Concentrations of remaining unreacted cystine were calculated accordingly. Cysteine persulde solutions were used immediately aer ltration and prepared freshly before each experiment.

Preparation of HSA-SSH
Persulde of human serum albumin was prepared and its concentration quantied as reported previously. 27 Enzyme kinetics of human Trx with cysteine persulde and HSA-SSH Kinetics of the reaction of human Trx with cysteine persulde was performed measuring NADPH oxidation on HP 8452A diode array spectrophotometer. As unreacted cystine could not be separated from cysteine persulde, additional control measurements with respective cystine concentrations under otherwise identical conditions were performed. 1 mM human Trx, 10 nM TrxR and 250 mM NADPH were used in all measurements with various concentrations of the substrates. The absorption maximum of NADPH at 340 nm was monitored over a course of 5 min in intervals of 5 s. The initial rate of A 340 decrease was tted linearly using Origin® analysis soware. Consumption of NADPH over time was determined using an NADPH extinction coefficient 3 340 of 6020 M À1 cm À1 . Rates of NADPH oxidation were plotted against the concentration of the mixture of cysteine persulde and cystine or cystine alone, respectively. The data obtained were tted using Michaelis-Menten equation in Origin® analysis soware. An analogous set of experiments was performed using the persulde derivative of HSA.

Kinetics of the direct reaction of human Trx with cysteine persulde and NAP-SSH
Kinetics of the reaction of thioredoxin with cysteine persude was monitored on FP-8200 spectrouorometer (Jasco, Germany) using an excitation wavelength of 280 nm and a maximal emission of 345 nm. In two separate sets of experiments either human or E. coli Trx (for comparison) were used. Concentration of the enzyme was kept at 1 mM while substrate concentrations ranged between 10 mM and 50 mM. Given pseudo rst-order conditions, observed rate constants k obs were obtained by tting the decrease in emission at 345 nm at a given cysteine persul-de or cystine concentration using a rst order exponential decay t in Origin® analysis soware. The k obs were plotted against (CysSS À + CysSSCys) or CysSSCys concentrations, and linearized. For comparison, an analogous set of measurements was performed using NAP-SSH and E. coli Trx.

H 2 S detection
Kinetics of H 2 S release was followed amperometrically using a selective H 2 S electrode connected to Free Radical Analyzer (World Precision Instruments). Experiments were performed in 96-well plate in 200 mL volume. Electrode was calibrated using standard Na 2 S solutions.

Spectral characteristics of free and protein-bound uorescent dyes
To compare spectral characteristics of free CN-BOT and free CN-Cy3 with those of their HSA-bound counterparts, absorption and emission spectra of those species were recorded. Namely, the labelling was performed by blocking 200 mM HSA-SSH and 200 mM HSA-SH (as a control) with 10 mM MSBT for 1 h at 37 C. Aer purication from excess blocking reagent by methanol/chloroform precipitation, samples were split in two aliquots and treated with 0.5 mM of CN-Cy3 and CN-BOT for 1 h at 37 C, respectively.
Excess free dye was removed by three steps of ultraltration (centrifugation at 14 000 Â g for 15 min) using a centrifugal lter with a molecular weight cut-off of 3 kDa (Carl Roth, Karlsruhe, Germany). For this, the solution, which was concentrated about 10-fold aer the individual centrifugation steps, was diluted to its original volume and the centrifugation was repeated. Aer the last centrifugation step, a methanol/chloroform precipitation was performed and the obtained protein pellets were re-dissolved in PBS. Absorption spectra were recorded at a HP 8452A diode array spectrophotometer, and uorescence spectra were at a FP-8200 spectrophotometer (Jasco, Germany).
Mice CSE +/À and CSE À/À mice were generated and characterized earlier (Ishii et al., 2010). In this study, CSE +/À males and females were bred to obtain CSE +/+ , CSE +/À and CSE À/À littermates. The Berlin Animal Review Board approved all protocols that were conducted to American Physiological Society standards.
Homogenates were centrifuged at 10 000 Â g for 20 min at 4 C and clear supernatant was further incubated on ice for 1 h. 200 mL of supernatant samples were precipitated with water/methanol/chloroform mixture (v/v/v: 4/4/1) and centrifuged at 14 000 Â g for 20 min at 4 C. Obtained protein precipitates were dried under vacuum at 4 C and stored at À80 C. For improved tagswitch assay equal amounts of sample proteins (50 mg) were incubated with 100 mM Cy3-CN in 50 mM HEPES (pH 7.4) supplemented with 1.5% SDS, in dark, at 37 C for 2 h, mixed with 4Â non-reducing sample buffer (Bio-Rad, USA) and incubated for 30 min at 50 C in dark. Samples were resolved on 10% SDS polyacrylamide gels in dark and immediately aer short xation and washing gels were scanned using ChemiDoc MP uorescent imager (Bio-Rad, USA). Obtained images were semi-quantied using ImageJ soware (NIH, USA).

In situ uorescence detection of intracellular persuldation
Cells were grown in m-dishes (35 mm, high) obtained from Ibidi® following manufacturer's instructions. The treatments with inhibitors, Na 2 S and AP39 were performed over 1 h. Aer treatments the cells were washed twice with warm sterile PBS. Fixation was carried out by incubation with ice-cold methanol at À30 C for 20 min and subsequent permeabilization with icecold acetone at À30 C for 5 min.

Mitochondrial visualization and co-localization studies
For co-localization studies, cells were transfected with CellLight® Mitochondria-RFP, BacMam 2.0 obtained from ThermoFisher Scientic® following the manufacturer's instructions. It is a fusion construct of the leader sequence of E1 alpha pyruvate dehydrogenase and TagRFP, providing specic targeting of cellular mitochondria-RFP through transient transfection with the insect virus baculovirus. Aer transfection and following the expression of mitochondria-RFP overnight, the cells were treated with AP39, D-Cys and auranon for 1 h at 37 C. Cells were xed with 4% formaldehyde for 45 min, washed with PBS and incubated with 50 mM HEPES containing Triton (1%) and MSBT (10 mM) at room temperature overnight. Aer washing with PBS (3Â), the xed cells were incubated with CN-BOT (25 mM) in PBS for 1 h at 37 C. The cells were then washed 5Â with PBS and stained with DAPI following manufacturer's recommendation.

Detection of protein persuldation in cell lysates
Cells were grown in T-75 cell culture asks. Aer treatments, the cells were washed twice with warm sterile PBS. Lysis was performed by addition of 800 mL HEN buffer (50 mM HEPES, 0.1 mM EDTA, pH 7.4) containing SDS (1.5%), NP-40 (1%), protease inhibitor cocktail (1%) and MSBT (10 mM) to a T-75 ask. Cells were incubated for 10 min on ice with occasional scraping of the ask surface with a cell scraper. The lysates were transferred to microcentrifugation tubes and incubated at 37 C for 1 h. Protein precipitation was performed using H 2 O/CHCl 3 /MeOH (4/1/4) precipitation, with subsequent vigorous mixing and centrifugation (20 000 Â g, 20 min, 4 C). The supernatant was discarded and the precipitate was dried. The pellet was resuspended in 300 mL of 50 mM HEPES containing 3% SDS. Lysates were stained with CN-Cy3 (60 mM) for 1 h at 37 C. SDS-PAGE was run under non-reducing conditions and the gels were recorded on a ChemiDocTM MP System with the Cy3 setup.
Isotope dilution mass spectrometry for the quantication of sulfane sulfur The reagents were synthesized as described previously. 40 1 mM probe 1 and 30 mM probe 2 (nal concentrations) were added into the HEN lysis buffer which also contained 1.5% SDS and 1% protease inhibitors. 200 mL of this mixture was added directly into the ask and cells scraped from the ask. Cell lysates were then incubated 1 h at 37 C with occasional vigorous vortexing. Protein concentrations were determined in the samples and proteins were then precipitated by the addition of acidied acetonitrile. Samples were vortexed and centrifuged, and the supernatants were taken for the MS analysis.

HIV patients
11 HIV-1-infected patients were included in the study. The study was approved by the Ethics Committee of the Medical Faculty of the Friedrich-Alexander-Universität Erlangen-Nürnberg and the patients gave informed consent. The patient characteristics are shown in the Table S1. † Five patients were on antiretroviral treatment with a suppressed viral load of <20 copies per ml and a median CD4 count of 364/ml (range 309-955). Six patients were untreated with a median HIV-1-viremia of 10 700 copies per ml (range 2600-5 400 000) and a median CD4 count of 436 (range 143-684). Sulfane sulfur levels were expressed as pmol mg À1 of plasma protein.

Conflict of interest
MW, MEW and the University of Exeter have patent applications relating to the therapeutic and agricultural use of H 2 S donors including AP39.

Author contribution
RW, CO, JM and MRF performed most of the experiments. SW synthesized and characterized CN-BOT. AM synthesized and characterized CN-Cy3. IA, MG and II provided CSE À/À mice and helped with tag-switch method evaluation in animal tissue. EG synthesized NAP-SSH. MX provided isotope-dilution MS reagents and helped with data planning and analysis. EGH and TH collected plasma samples from HIV patients and performed data analysis. RT, MW and MEW synthesized AP39 and helped in data analysis. PKY and RB puried recombinant human CSE and helped with data analysis and manuscript writing. All authors contributed in manuscript writing. MRF designed the study, analysed the data and wrote the manuscript.