Wittig reagents for chemoselective sulfenic acid ligation enables global site 1 stoichiometry analysis and redox-controlled mitochondrial targeting 2 3


 Triphenylphosphonium ylides, known as Wittig reagents, are one of the most commonly used tools in synthetic chemistry. Despite their considerable versatility, Wittig reagents have not yet been explored for their utility in biological applications. Here, we introduce a new chemoselective ligation reaction that harnesses the reactivity of Wittig reagents and the unique chemical properties of sulfenic acid, a pivotal post-translational cysteine modification in redox biology. The reaction, which generates a covalent bond between the ylide nucleophilic α-carbon and electrophilic γ-sulfur is highly selective, rapid, and affords robust labeling under a range of biocompatible reaction conditions, including in living cells. We highlight the broad utility of this conjugation method to enable site-specific proteome-wide stoichiometry analysis of S-sulfenylation, visualize redox-dependent changes in mitochondrial cysteine oxidation, and redox-triggered TPP generation for controlled delivery of small molecules to mitochondria.

| Repurposing triphenylphosphonium ylides as probes for electrophilic sulfur in proteins. a, Wittig reagents can be illustrated as the phosphorane or ylide form. With anion-stabilizing groups (ASG), Wittig reagents act as water-compatible carbon-based nucleophiles. b, Sulfenic acids are formed via two major pathways: direct oxidation or hydrolysis of polarized sulfur species. c, Biocompatible reaction between sulfenic acids and stabilized Wittig reagents. d, Determination of S-sulfenylation site occupancy. After proteomic workflow, labeled protein thiols (top) and sulfenic acids (bottom) yielded isotopomers. e, Taking advantage of the TPP group installed after reaction with the Wittig reagent enables direct or redox-triggered cargo delivery to mitochondria. 50 protonated and deprotonated states are accessible at physiological pH 8 . If stabilized by the 51 microenvironment, the thiol-sulfenic acid (or sulfenate) pair can operate as a switch triggered by redox 52 changes, regulating protein function, structure, and localization 9 . Alternatively, the electrophilic sulfur in 53 sulfenic acid may condense with a protein or low-molecular-weight thiol to form a disulfide or, under 54 conditions of excess oxidative stress, may be oxidized further to sulfinic and sulfonic acids. In either 55 scenarioas stabilized or transient intermediate -sulfenic acids are central modifications in the domain 56 of biological redox-regulation 10,11 ( Supplementary Fig. 1b). Furthermore, our group 8,12-16 and others 17,18 57 have shown that the electrophilic character of sulfur in sulfenic acid is chemically distinct from protein 58 electrophiles, including modifications to amino or thiol functional groups, making this species an ideal 59 candidate to examine for reactivity with Wittig reagents. 60 61 With the unexamined potential of nucleophilic Wittig reagents in biological applications and the 62 advancement of our understanding of electrophilic sulfenic acid modifications in biology and 63 Next, we interrogated the selectivity of WYne probes 8-10 against related redox cysteine modifications. 107 WYne probes did not exhibit cross-reactivity with glutathione, glutathione disulfide, S-nitrosoglutathione, 108 cysteine sulfinic acid or glutathione sulfonic acid ( Fig. 2d and Supplementary Fig. 6). WYneC underwent 109 Figure 2 | WYne probe reactivity with sulfenic acid in complex biological settings. a, Physical property and kinetic profile of the WYne probes. b, Base-promoted hydrolysis facilitated cleavage of the TPP group from WYne probes. c, A small molecule sulfenic acid model (dipeptide-SOH) was used to kinetically evaluate WYne probes. d, Reaction of WYne probes and various sulfur species in pH 7.4 aqueousorganic buffer. Owing to rapid kinetics, the reaction between WYneC and dipeptide-SOH (R-SOH) was performed at pH 4.9 and the rate constant extrapolated to pH 7.4 (see Supplementary Methods). Reactivity with other biological sulfur species was not observed. e, Glutathione peroxidase 3 (Gpx3, from S. cerevisiae) as a model to study protein sulfenic acids. f, Intact protein MS analyses suggested that WYne probes exclusively targeted the sulfenic acid form of Gpx3 and not the thiol, sulfinic acid or disulfide forms. g, In-gel fluorescence detection of Gpx3 sulfenic acid. h, In situ labeling of A549 cells with WYne probes showed time and dose-dependence.
UV-cleavable-biotin reagents (1:1) via CuAAC. Light and heavy biotinylated peptides were mixed 152 equally, captured on streptavidin beads, and photoreleased for MS-shotgun proteomics for identification 153 and quantification. Probe-labeled peptides covalently conjugated to light or heavy tags yield an isotopic 154 signature in which only peptide assignments with a light/heavy ratio close to 1.0 are recognized as true 155 identifications. Surprisingly, our initial database search targeting intact modifications derived from WYne 156 probes yielded only 474, 147, and 0 probe-labeled sites for WYneC, WYneO, and WYneN, respectively 157 (Fig. 3b,c and Supplementary Table 1), whereas many unassigned MS1 peaks with the isotopic signature 158 (RL/H≈1.0) were identified in the raw data. Hence, we conducted a blind search as previously described 22 , 159 which revealed that all three WYne probes underwent a loss of the TPP moiety, likely caused by base-160 promoted hydrolysis during tryptic digestion (Fig. 3d). The cleaved products exhibited different patterns 161 of MS/MS fragmentation compared to the corresponding intact modification derived ( Supplementary Fig.  162 11) and the intensity of the former was dramatically higher than the latter (Fig. 3b). Fortuitously, WYneN-163 derived cysteine modifications were quantitatively transformed into the TPP-cleaved products, providing 164 a much higher yield of identified sulfenic acid sites compared to those obtained with WYneC or WYneO 165 as well as two previously reported sulfenic acid probes, DYn-2 and BTD (Fig. 3c). Accordingly, we 166 focused subsequent proteomic validation studies on WYneN. 167

168
To benchmark reaction with the S-sulfenylome in situ, intact A549 cells were labeled with WYneN and 169 processed with the aforementioned chemoproteomic workflow. The BTD probe was used as control, as 170 it previously provided the largest S-sulfenylome dataset to date 15,21 . Under identical conditions (500 μM, 171 unexpected, we observed that lower concentrations of WYneN correlated with a higher percentage of 177 functionally important cysteine residues (e.g., active site, disulfide, metal binding) mapped across sites 178 (Fig. 3f). This finding suggests that high WYneN reactivity exhibited by individual sulfenic acid 179 modifications positively associates with functional importance. We also investigated the selectivity of 180 WYneN on a proteomic scale using the Open-pFind algorithm to re-analyze the in situ dataset of WYneN 181 labeling by a targeted search of all polar amino acids 23 . WYneN predominately labeled cysteine residues 182 at both a site-(90.0%, Supplementary Fig. 12a) and spectral-level (99.8%, Supplementary Fig. 12b). with WYneN (Fig. 3g). Probe-tagged proteomes with and without oxidant treatment were digested into 195 tryptic peptides, conjugated with light and heavy Az-UV-biotin reagents, respectively, and processed as 196 described above. In this workflow, the light to heavy ratio calculated for each WYneN-labeled cysteine 197 residue provides a measure of its relative level in H2O2-treated samples versus unoxidized control 198 samples. In total, we identified and quantified 2,234 probe-modified sites on 1,633 proteins, including 199 numerous functionally important cysteine residues (Supplementary Fig. 13a and Supplementary Table  200 3). Of these, 9.9% quantified sites showed ≥1.5-fold dynamic changes after H2O2 treatment. Positively 201 regulated S-sulfenylation sites likely indicate protein-stabilized sulfenic acids, while negatively regulated 202 S-sulfenylation sites suggests overoxidation. For example, S-sulfenylation of the active sites of PRDX6 203 (C47) and ASAH1 (C43) decreased upon H2O2 treatment, indicative of hyperoxidation to sulfinic and 204 sulfonic acid states ( Supplementary Fig. 13b). Moreover, S-sulfenylation of the surface-exposed cysteine 205 residue C152 in GAPDH was negatively regulated at a ratio of 0.88, in contrast, the buried cysteine C247 206 showed a 2.44-fold sulfenylation under H2O2-induced stress ( Supplementary Fig. 13b). Likewise, three 207 cysteine residues C90, C152 and C220 of ubiquitin carboxy-terminal hydrolase L1 (UCHL1) were all 208 identified as S-sulfenylation sites, and the most-buried, non-catalytic C220 showed a 2.6-fold increase of 209 modification under stress ( Supplementary Fig. 13b). This residue was previously found for UCHL1 to 210 promote the assembly of mTOR complex 2 and phosphorylation of the pro-survival kinase AKT and is a 211 known S-nitrosylation site as well as a potential alkylation site 24-26 . These findings reinforce the concept 212 of dynamic protein S-sulfenylation in cells and offer hypotheses to explain how non-catalytic cysteines 213 may affect enzymatic functions via redox regulation. 214 215 Upon deeper analysis of these data, we noted that WYneN detected sulfenic acid modification of the key 216 nucleophilic cysteine within redox-active CXXC motifs 27 (Fig. 3h). The "attacking" cysteines in this 217 sequence are susceptible to rapid thiol-disulfide exchange with the "resolving" cysteine that precludes 218 chemical ligation by less efficient dimedone-based probes. By contrast, WYneN enabled dynamic 219 quantification of sulfenic acid modification at the "attacking" cysteines of many CXXC-containing proteins, 220 including C32 of thioredoxin-1 (TRX-1), C23 of glutaredoxin-1 (GRX-1), C397 of protein disulfide-221 isomerase (PDIA1), and C12 of antioxidant 1 copper chaperone (ATOX-1) (Fig. 3i). The ability of WYneN 222 to effectively trap and label the sulfenic acid state of TRX-1, GRX-1 and ATOX-1 was further validated 223 by intact MS and in-gel fluorescence analyses of these recombinant proteins (Fig. 3j,k). These findings 224 reveal the CXXC motif in thiol oxidoreductases as heretofore unrecognized direct targets of oxidation. 225 setting stage for redox biology applications enabled by this biocompatible and highly selective chemistry. 227 228 Proteome-wide analysis of cysteine sulfenic acid site stoichiometry. Although methods have been 229 developed for the relative quantification of sulfenic acid post-translational modifications, it has not been 230 possible to quantify stoichiometry (also referred to as site occupancy) at a global level in cells. Defining 231 the fraction of proteins that contain sulfenic acid at a given site is essential step toward understanding 232  Table 4). Consistent with 247 the often-transient nature of sulfenic acid in cells, the %SOH values for the majority of the cysteinome 248 (73%) were calculated to be lower than 30%, with an average of 21.1% and a median of 14.5% (Fig. 4b). 249 We also found that multiple cysteines on the same protein had significantly different %SOH values. For 250 example, five sulfenylated cysteines were mapped onto NDUFS1, a core subunit of the mitochondrial 251 complex I, with %SOH values ranging from 13.4% to 74.1% (Fig. 4b), among which were three metal 252 binding sites (C64, C78 and C92). In another case, %SOH value of C152 on GAPDH was found to be 253 much higher than that of C247 (Fig. 4b), in accordance with our previous finding based on spectral 254 counting of sulfenylated peptides bearing these two sites 13 . Also of interest, we measured %SOH values 255 for many protein tyrosine phosphatases (PTPs), which contain a conserved [I/V]HCSXGXGR[S/T]G motif 256 in their active site (Fig. 4c). The invariant cysteine is essential for catalysis and can be negatively 257 regulated by oxidation 29 . Such active sites in the majority of PTPs showed a higher %SOH value than 258 the median value for overall sites (Fig. 4b,c). For example, we detected 20.0% S-sulfenylation of PTPN1 259 disulfide bond formation compared to 2-Cys peroxiredoxins 31 . 264

265
In order to investigate the relationship between %SOH and functionality, we retrieved information about 266 cysteine residues with annotated functions from the UniProt knowledge database 32 . In general, modified 267 cysteines (mainly through S-nitrosylation) tended to be less S-sulfenylated than other annotated or 268 unannotated cysteines (Fig. 4e), as different types of cysteine modifications compete for the same site, 269 thereby diminishing the %SOH. By contrast, active-site cysteines were distributed more broadly in the 270 range of 30-60% S-sulfenylation, compared to those with other functional annotations or without 271 annotation (Fig. 4e). In addition to PTPs, other classes of enzymes in which the active site cysteine is 272 highly prone to oxidation and known to be redox regulated, such as ubiquitinating and deubiquitinating 273 enzymes, were identified 33 (Supplementary Table 4). We also examined the cellular localization and 274 Gene Ontology (GO) classification of the proteins with cysteines within the different ranges of %SOH 275 (Fig. 4f,g). Major oxidant-generating cellular compartments, including the peroxisome, endoplasmic 276 reticulum and mitochondrion, were distinguished as having more highly S-sulfenylated proteins (%SOH 277 ≥ 60) as compared to other compartments (Fig. 4f). Likewise, protein cysteines involved in two key 278 oxidative pathways, oxidative phosphorylation and fatty acid oxidation, exhibited significantly higher 279 overall %SOH relative to other metabolic pathways, including the TCA cycle and 280 glycolysis/gluconeogenesis pathways (Fig. 4g). Going forward, quantification of %SOH using the 13 C5 281 WYneN probe for selective ligation will greatly assist in prioritizing sites for functional analyses and 282 defining mechanistic models of redox-dependent protein regulation.  = 4). f, WYneN10 disrupted mitochondrial respiration in A549 cells to a greater extent than other WYne probes. OCR, oxygen consumption rate. Error bars represent ± s.e.m of biological replicates (n = 6). P values were calculated using a two-tailed, unpaired t-test. ns, not significant, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. pH. Despite its lower abundance, the deprotonated form of WYneN remains highly reactive with sulfenic 295 acid. Given the pH gradient of approximately one unit between the mitochondrial matrix (pH = 8) and the 296 cytosol (pH = 7) 36 , we hypothesized that WYneN would partition to the mitochondria, become more 297 deprotonated, and preferentially label protein sulfenic acids in this compartment (Fig. 5a). To test this 298 idea, we functionalized WYneN with a BODIPY tag for fluorescence visualization, but live cell imaging 299 suggested a poor localization of fluorescence (R < 0.3). This finding is consistent with the lack of 300 compartmental bias observed in our chemoproteomic studies and could be explained by the ionic nature 301 of WYneN, which decreases its inherent ability to penetrate phospholipid membranes. To tackle this 302 issue, we synthesized a WYneN derivative with a 10-carbon aliphatic linker, WYneN10 (11, Fig. 5b). indicating probe accumulation in this organelle ( Fig. 4c and Supplementary Fig. 14). In addition, 306 mitochondrial uncouplers such as carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP) and 307 antimycin A dissipated mitochondrial membrane potential, and suppressed the mitochondrial staining 308 pattern ( Supplementary Fig. 15). Next, we tested if BDP-WYneN10 could detect redox-dependent 309 changes in mitochondrial cysteine oxidation -using sulfenic acid content as an indicator -under 310 conditions of exogenous oxidative stress. Live A549 cells were imaged after treatment of BDP-WYneN10 311 and H2O2. Indeed, mitochondrial BODIPY fluorescence intensity increased concomitant with oxidant 312 concentration (Fig. 5d,e and Supplementary Fig. 16) that can be attributed to increased sulfenic acid 313 modification of mitochondrial proteins concomitant with covalent reaction of BDP-WYneN10. To further 314 characterize the effect of WYne probes on mitochondria function, we performed a mitochondrial stress 315 test and monitored the oxygen consumption rate (OCR) of A549 cells. Of all WYne probes, only lipophilic, 316 cationic WYneN10 disrupted mitochondrial respiration with respect to spare respiratory capacity, maximal 317 respiration and ATP production ( Fig. 5f and Supplementary Fig. 17). Overall, these data indicate the 318 elevated pKa inherent to the amide-functionalized lipophilic Wittig reagent, WYneN10, and its reactivity 319 with sulfenic acid, can be exploited to visualize changes in mitochondrial cysteine oxidation. 320 Redox-triggered in situ TPP generation for mitochondrial delivery. To further showcase the 322 biocompatible, chemoselective reaction between sulfenic acids and Wittig reagents, we considered that 323 knowledge, such a controllable system has not been reported, whilst the closest precedence to this 327 concept are some commercially available MitoTracker probes that are readily oxidized during basal 328 cellular respiration to become cationic and fluorescent. In contrast to existing mitochondrial targeting 329 strategies that are limited to constitutive, uncontrolled delivery 35 , our approach provides a reaction-based 330 switch to sequester non-cationic substrates for mitochondrial enrichment that can respond to cellular 331 redox changes. 332

333
To investigate the feasibility of this concept, we utilized sulfoxides as caged precursors to small molecule 334 sulfenic acids (Table 2). This decaging reaction typically requires heat but can take place at ambient 335 temperatures when the β-carbon is connected to strong electron withdrawing groups (EWGs) 37,38 . To 336 (HOCl)a powerful oxidant produced by biological systems as an immune response -to convert them 338 to sulfoxides. The reaction was hindered when a methyl group was present at the α-position and 339 decreased further with dimethyl substitution. Strength of the EWG also played an essential role. For 340 example, β-hydrogen acidity was greatly enhanced by benzenesulfonyl groups, resulting in faster 341 elimination compared to compounds with ester substituents. This series of compounds offered us a 342 broad range of rates for the decaging of sulfenic acids, from 10 min to 6 h for reaction to finish. Next, we 343 employed Wittig reagents 3-5 to capture the nascent sulfenic acids and furnish TPP derivatives 15-17. 344 The best yield was obtained from 14 with highly nucleophilic Wittig reagents 3 and 4, while cages 12 and 345 13 with faster release rates gave moderate to poor yield (Table 2). Live cell images of HeLa cells co-346 treated with MitoTracker and 15 or 16 indicated that the TPP-linked compounds successfully 347 accumulated in mitochondria, as opposed to a disulfide control (Fig. 6b). Ester derivative 16 exhibited a 348 higher degree of mitochondrial enrichment (Pearson's correlation coefficient R = 0.80) relative to the 349 ketone derivative 15 (R = 0.55), owing to the higher pKa of the ester 39 , which leads to a greater degree 350 of protonation at physiological pH ( Supplementary Fig. 18). Enzymatically generated HOCl produced by 351 the myeloperoxidase (MPO) enzyme system, also provided the desired product 16 (Fig. 6c and  352 Supplementary Fig. 19). This concept was also evaluated in situ, where redox-caged sulfenic acid To date, site-specific chemoproteomic profiling of protein sulfenic acids in intact cells has been predicated 384 almost exclusively on DYn-2, a 1,3-cyclohexadione-based probe, which is quite selective, but hampered 385 by slow reaction kinetics (Supplementary Fig. 1c). In this regard, a very large amount of protein materials 386 (30-40 mg) is typically required 20 , thereby greatly precluding many physiologically relevant applications. 387 Notably, as compared to our previous analysis using DYn-2 13 , 10-fold fewer starting materials were 388 needed for our chemoproteomic study with WYneN that achieves even a higher coverage of the 389 sulfenylome. Strikingly, among ~1,000 newly discovered in situ sulfenylated sites were the "attacking" 390 cysteines within CXXC motifs that are prone to rapid thiol-disulfide exchange, further demonstrating 391 excellent kinetics of WYne reagents. Moreover, of interest, our dose-dependent labeling experiment (Fig.  392 3f) reveals that the reactivity of WYneN towards sulfenic acids positively associates with functional 393 importance (e.g., active site, disulfide, metal binding). Indeed, S-sulfenylation, as an electrophilic 394 modification of the cysteine proteome, can effectively inverse the polarity of the nucleophilic sulfur atom 395 and render it more reactive. Therefore, our reaction-based approach for profiling this electrophilic 396 cysteinome directly and reliably discovers functional, redox regulated cysteines. 397 398 Quantification of %SOH may also be essential to assess potential of drug resistance for those covalent 399 inhibitors targeting free thiols. In addition, stoichiometric quantification of %SOH on a particular cysteine 400 site offers another interesting aspect to identify functional nodes through which endogenous reactive 401 oxygen species exerts regulatory functions. Existing proteomic approaches often report percent cysteine 402 modification but rely on differential alkyation-reduction strategies to indirectly assess the total reversible 403 oxidation, lumping all S-modifications together, in contrast with direct measurement of distinct oxoforms 404 in the cysteinome. Using 13 C5 WYneN and IPM, we report here the first measurement of %SOH in 405 cysteines on a proteome-wide scale in cells. Notably, many cysteine residues that play major regulatory 406 roles in redox signaling are sulfenylated at a relatively low extent, including those in 2-Cys peroxiredoxins. 407 In concert with this observation, a recent report demonstrates that a redox switch with low percent 408 oxidation is indeed critical for the physiological roles of Akt 42 . Thus, we reason that, like other post-409 translational modifications, such as phosphorylation 43  An oxidant-triggered chemical switch for mitochondrial delivery has many advantages including improved 430 target selectivity and decreased cargo toxicity. By tuning the substituents on the redox-caged substrate, 431 we achieved a wide range of temporal responses for both acute release and prolonged actions. Beyond 432 a proof-of-concept fluorescent cargo, we implemented an antioxidant-based mitochondrial delivery 433 system to counteract the effect of endogenous HOCl. Since HOCl is mainly secreted in myeloperoxidase 434 (MPO)-rich immune cells, such as neutrophils or macrophages, it causes inflammatory damage to 435 surrounding healthy cells. In this context, the HOCl-triggered antioxidant may alleviate mitochondrial 436 oxidative damage caused by inflammation. This redox-triggered strategy opens door to more selective 437 therapies that target cancer, aging and degenerative diseases, which are often associated with elevated 438 oxidative stress. It could also improve the efficacy of established mitochondria-targeting approaches with 439 payloads such as antioxidants, apoptosis-inducing drugs, imaging molecules, and genetic materials. The WYneN adduct is sensitive to prolonged incubation under strong reducing and basic conditions, so 444 appropriate care must be taken during analyses. At present, the Wittig-based redox-switch is most easily 445 triggered by HOCl. The ability to efficiently react with H2O2 would further enhance the application of our 446 controlled delivery mechanism. WYneN10 is well poised for imaging and detection of mitochondrial 447 protein sulfenic acids, but its hydrophobic linker limits application in quantitative MS. These and other 448 areas are interesting directions for further studies. 449

450
In summary, our findings demonstrate that Wittig reagents, together with a unique form of electrophilic 451 sulfur found in proteins constitute a new class of highly selective and biocompatible reactions. Projecting 452 forward, we envision several exciting areas wherein this reaction can be applied that should help address 453 fundamental questions about redox modification in the human cysteinome and, more broadly, the sulfenic 454 acid moiety as a target for covalent drugs and chemical ligation handle for mitochondrial targeting. 455