Nucleophilic covalent ligand discovery for the cysteine redoxome

With an eye toward expanding chemistries used for covalent ligand discovery, we elaborated an umpolung strategy that exploits the ‘polarity reversal’ of sulfur when cysteine is oxidized to sulfenic acid, a widespread post-translational modification, for selective bioconjugation with C-nucleophiles. Here we present a global map of a human sulfenome that is susceptible to covalent modification by members of a nucleophilic fragment library. More than 500 liganded sulfenic acids were identified on proteins across diverse functional classes, and, of these, more than 80% were not targeted by electrophilic fragment analogs. We further show that members of our nucleophilic fragment library can impair functional protein–protein interactions involved in nuclear oncoprotein transport and DNA damage repair. Our findings reveal a vast expanse of ligandable sulfenic acids in the human proteome and highlight the utility of nucleophilic small molecules in the fragment-based covalent ligand discovery pipeline, presaging further opportunities using non-traditional chemistries for targeting proteins. Chemoproteomics reveals a vast expanse of ligandable cysteine sulfenic acids in the human proteome, highlighting the utility of nucleophilic small molecules in the fragment-based covalent ligand discovery pipeline.

Targeted covalent inhibitors 1 and degraders 2 have emerged as a promising therapeutic class.Covalent drugs have been successfully applied to key oncogenic driver proteins, including EGFR and mutant KRAS, historically considered to be 'undruggable' 3 .At the same time, most of the human proteome still lacks defined small-molecule probes, and expanding the protein landscape amenable to targeting by covalent chemistry remains an important goal.Existing methods use electrophile-functionalized fragments to target residues with nucleophilic side chains, namely cysteine but also lysine, tyrosine and aspartate 4 .
Among the amino acids targeted by residue-specific chemistry, cysteine is most nucleophilic and has, therefore, emerged as a predominant target in covalent drug discovery 3,4 .Most covalent inhibitors approved by the US Food and Drug Administration (FDA) as drugs or as candidates in clinical trials are designed to target a functionally important cysteine in its reduced thiol (-SH) form 5 .Examples include afatinib 6 and sotorasib 7 that target cysteines in epidermal growth factor receptor (EGFR) kinase and KRAS-G12C, respectively.When combined with chemoproteomic strategies, such as activity-based protein profiling (ABPP) 8 , thiol-directed electrophilic fragments have been successfully mined to expand the ligandable space in the human cysteinome 9,10 .
Cysteines are also subject to diverse redox chemistry in cells through reaction with exogenous and endogenous reactive oxygen species (ROS) and other biological oxidants 11 .Proteomic methods that site-specifically map the 'cysteine redoxome' or the inventory of oxidized cysteines have been instrumental in gaining a better understanding of redox signaling mechanisms 12 .Within the cysteine redoxome, S-sulfenation (-SOH) has emerged as a widespread oxidative modification, identified in thousands of proteins, with more than 50% of mapped sites having a stoichiometry ranging between 10% and 50% [13][14][15] .
Although not all sites of S-sulfenation have defined functional or structural roles, in general, toggling between the thiol and sulfenic acid state can operate as a regulatory switch similar, in many ways, to phosphorylation.For example, oxidation of an active site cysteine to sulfenic acid in EGFR increases its intrinsic tyrosine kinase activity 16,17 .On the other hand, enzymes that require a nucleophilic thiolate for Article https://doi.org/10.1038/s41589-023-01330-5acid-selective probe BTD 27,28 (Fig. 1a).This strategy is analogous to that used to assess nucleophilic reactivity of cysteine 29 and lysine 30 and is based on the principle that so-called 'hyper-reactive' sulfenic acids will be quantitively labeled at low BTD concentrations, whereas less reactive sites will exhibit concentration-dependent increases in probe labeling (Fig. 1b).
To test this concept, we prepared native lysates from MDA-MB-231 cells and incubated samples with BTD (0.5 mM or 5 mM).BTD-labeled proteomes were subsequently digested and clicked to isotopically differentiated biotin for enrichment, identification and quantification.Sulfenic acids with a heavy-to-light ratio (R H/L = R 10:1 ) below 2.0 were empirically designated 'hyper-reactive'; those with R 10:1 values higher than 2.0 but not greater than 5.0 were designated as 'moderately reactive'; and those with R 10:1 values higher than 5.0 were designated as sites having 'low reactivity'.Control experiments to minimize bias in quantification were performed in which lysates were treated with the same BTD concentration (that is, 5 mM versus 5 mM).High-confidence, quantifiable S-sulfenation sites were required to be detected in both 10:1 and 1:1 datasets with the later R value in the range of 0.67-1.5 (Extended Data Fig. 1).
In total, we were able to calculate R 10:1 for 622 cysteine S-sulfenation sites across 477 proteins (Fig. 1c and Supplementary Data 1).A subset of sulfenic acids (121% or 19.5%) were hyper-reactive, exhibiting dose-independent BTD reactivity (R 10:1 < 2.0); most (319% or 51.3%) were moderately reactive; and the remainder (182% or 29.3%) were characterized as having low reactivity (R 10:1 > 5.0).By examining flanking sequences of these sites, we found that the basic amino acid lysine and, to a lesser extent, arginine were significantly enriched at the −2 position for both hyper-reactive and moderately reactive sites.Charged residues were also existent in the consensus sequence motif of less reactive sites but were in positions comparatively further from cysteine (−5, −3 and +4; Extended Data Fig. 2).
Apropos to covalent targeting, the sulfur atom of sulfenic acid is distinguished from the thiol and other oxidation states by having moderate electrophilic reactivity 20 .The chemical reactivity of sulfenic acid toward electrophiles such as iodoacetamide is orders of magnitude slower than its thiol counterpart 21 .Chronic prolonged ROS production is associated with many diseases, including cancer, diabetes and Parkinson's disease, and is also central to the progression of inflammatory disorders 22 .Oxidized forms of cysteine can accumulate in proteins over time and affect the pharmacology of covalent drugs designed to target this residue in its reduced state.Consistent with the marked change in sulfur reactivity, chronic oxidative stress leads to the formation of sulfenated EGFR, and this subpopulation is refractory to afatinib treatment 16 .Similar observations have been made for covalent inhibitors that target KRAS-G12C 23 .
With an eye toward expanding the types of chemistry used in covalent ligand discovery beyond electrophilic fragments, we focused our efforts on an umpolung strategy that capitalizes on the 'polarity reversal' that occurs when cysteine is oxidized to sulfenic acid for selective bioconjugation with C-nucleophiles.Along these lines, prior studies in our laboratory have shown encouraging results, including the discovery of covalent inhibitors that target the redox-active cysteine in the active site of PTP1B 24,25 and the investigation of C-nucleophile 'warhead' reactivity through site-centric chemoproteomics 26 .Although these findings are important in the overall developmental process, understanding of nucleophile-sensitive sulfenic acids and global S-sulfenome ligandability remains at an early stage.
In this study, we adapted a quantitative chemoproteomic platform termed SulfenQ 14 to quantify intrinsic sulfenic acid reactivity and screened a library of nucleophilic small molecules against the S-sulfenome in a human cancer cell proteome.More than 500 liganded sulfenic acids were identified on proteins across diverse functional and structural classes, and, of these, more than 80% of liganded sites are unique and not targeted by electrophilic fragment analogs.Our findings establish nucleophilic covalent small molecules that target sulfenic acids in the cysteine redoxome as a valuable new resource for chemical probes and drug discovery, where ligandability in the proteome extends beyond cysteine and other nucleophilic side chains.

Quantitative profiling intrinsic S-sulfenate reactivity in the human proteome
In previous work, we described a ratiometric quantification method termed 'SulfenQ' that measures relative changes across the sulfenome in lysates or cells 14 .Building on this work, we first adapted this approach to quantify intrinsic electrophilic reactivity of cysteinyl sulfenic acids by dose-dependent proteome labeling with the 'clickable' sulfenic
Comparing these data with previously mapped sites of S-sulfination 32 indicates that oxidation to irreversible or, in some proteins, slowly reversible sulfinic acid is more likely to occur at sulfenic acids with at least moderately intrinsic reactivities (Fig. 1h).Next, we asked whether the intrinsic reactivity of the sulfur in S-sulfenated sites correlated to the intrinsic reactivity of cysteinyl thiols mapped in an earlier study 29 .No significant correlation was observed between these properties (r = 0.12; Fig. 1i).Although seemingly obvious, this analysis underscores the fundamental difference in chemical reactivity between cysteine and its closely related 'oxoform'; namely, the sulfur in hyper-reactive thiols is nucleophilic, whereas, in hyper-reactive sulfenic acids, it is electrophilic.For instance, active  Fig. 2 | Nucleophilic and electrophilic fragments with the same scaffold have distinct target profiles.a, General protocol for competitive chemoproteomic profiling of nucleophilic (sulfenic acid-reactive using SulfenQ) and electrophilic (thiol-reactive using QTRP) fragments.Competitive analysis of the MDA-MB-231 cell proteome pre-treated with fragment or DMSO vehicle (500 µM for 1 and 2, 2 mM for 3a and 3b), followed by a probe that is broadly reactive for thiols (100 µM IPM) or sulfenic acids (5 mM BTD).Sites were designated as liganded if the R H/L (R DMSO/Fragment ) value was ≥4 as measured by either QTRP or SulfenQ.b, Distribution of R H/L values of each fragment across the MDA-MB-231 sulfenome (3a and 3b) or cysteinome (1 and 2).Elements of each violin plot indicate the median (white dot) and the 1st and 3rd quartiles (lower-whisker and upperwhisker, respectively).c, Bar chart showing the number of thiol or sulfenic acid sites liganded by each fragment.The percentage of liganded sites is indicated on each bar.d, Venn diagrams showing the overlap of overall quantified cysteine sites in either the sulfenic acid or thiol forms (left) and fragment-liganded sites (right).e, Representative extracted XICs showing changes in probe-labeled peptides from proteins as indicated.The profiles for light-labeled and heavylabeled peptide are shown in red and blue, respectively.The average R H/L values calculated from biological duplicates are displayed below each XIC.f, Heat map showing representative fragment interactions for liganded cysteines in either sulfenic acid or thiol forms.Competition ratios for nucleophilic fragment 3b were obtained by SulfenQ, whereas those of electrophilic fragments 1 and 2 were obtained by QTRP in this study or by isoTOP ABPP from ref. 9. NA, not applicable.

Article
https://doi.org/10.1038/s41589-023-01330-5site cysteine in GSTO1 (C32) is hyper-reactive, whereas a sulfenic acid at this position is moderately reactive (Fig. 1j).Conversely, the active site cysteine in GAPDH (C247) has low reactivity, whereas its sulfenic acid form is moderately reactive (Fig. 1j).A significant corollary to these observations is that cysteine thiols identified as hyper-reactive are not necessarily redox-active in biological settings, unequivocally establishing the disparate reactivity of the cysteinome and sulfenome.

Comparison of cysteinome and sulfenome ligandability in the human proteome
Having established the unique intrinsic reactivity of the cysteinome and sulfenome, we next turned our attention to identifying sites that are susceptible to covalent modification by a small molecule having the same drug-like scaffold but functionalized with 'warheads' having nucleophilic or electrophilic reactivity in both sub-proteomes (Fig. 2a).To profile the cysteinome, we selected chloroacetamide and acryl amide warheads (1 and 2, Fig. 2b, inset), which have been widely used in fragment-based covalent ligand and drug discovery 33 .To map the sulfenome, we elected to use linear C-nucleophile 'warheads' rather than cyclic analogs such as dimedone to minimize differences in site-specific covalent modification that could result from gross structural differences between thiol-reactive and sulfenic acid-reactive 'warheads'.Consequently, we chose cyanoacetamide and nitroacetamide 'warheads' previously established by our group to undergo chemoselective reaction with sulfenic acid at moderate kinetic rates 34 (3a and 3b, Fig. 2b, inset, and Supplementary Figs. 1 and 2).Our choice of nucleophilic 'warheads' was also driven by our interest in exploring how covalent reversible and irreversible small molecules differ in their overall rates of ligandability.The thioether bond formed by the reaction of 3a or 3b with sulfenic acid differs in its stability; cyanoacetamides form a covalent bond that is reversible in the presence of reducing agents such as glutathione (GSH; t 1/2 ~2 h), whereas nitroacetamides form an irreversible covalent bond (t 1/2 > 24 h; Supplementary Fig. 3).Finally, it is also important to note that, although SN2 reaction between a thiol and a cyanoacetamide or a nitroacetamide can occur under organic conditions, -CN and -NO 2 are not good leaving groups under physiological conditions.No evidence for these reaction products, which are clearly distinguishable by mass spectrometry (MS), were observed in peptide or protein models [34][35][36] or in our chemoproteomic analysis of fragment reactivity detailed below.
Thiol-reactive and sulfenic acid-reactive 'warheads' were linked by amide bond formation to a commercially available 3,5-bis(trifluoromethyl)aniline fragment with validated ligandability throughout the cysteinome 9 .Covalent modification by electrophilic or nucleophilic fragments was measured using a quantitative chemoproteomic platform known as QTRP 37 (quantitative thiol reactivity profiling) or SulfenQ 13,14 , respectively, in a pseudo-competitive format (Fig. 2a).Native lysates prepared from MDA-MB-231 cells were treated with fragment or vehicle (500 µM 1 and 2, 2 mM 3a and 3b), followed by a probe that is broadly reactive for thiols or sulfenic acids (100 µM IPM or 5 mM BTD).The use of higher fragment and BTD concentrations compared to thiol-reactive counterparts was empirically determined to ensure sufficient analytic depth of the S-sulfenome in our chemoproteomic workflow 27 .We designated sites that are susceptible to covalent modification by nucleophilic or electrophilic small molecules, hereafter referred as to 'liganded sites' if they showed an R H/L (R DMSO/Fragment ) value of ≥4 as measured by either QTRP or SulfenQ.
Overall, we quantified >5,300 cysteinyl thiol (1 or 2) and >2,300 sulfenic acid (3a or 3b) sites from 3,856 proteins across all datasets (Fig. 2b and Supplementary Data 2).Among the quantifiable cysteines were 1,100 and 580 liganded sites, corresponding to a ligand rate (the percentage of liganded sites relative to total sites for each fragment) of 25.2% and 11.0% for 1 and 2, respectively (Fig. 2c and Extended Data Fig. 3).By contrast, only 24 sulfenic acids were identified as liganded by nucleophilic fragment, 3b, corresponding to a ligand rate of 1.3%, whereas 3a did not yield any detectable liganded sites (Fig. 2c).A fragment lacking the reactive warhead was also screened and showed negligible proteome reactivity (4, 0.05%; Supplementary Data 2), indicating that sites liganded by compounds 1, 2 or 3b most likely reflect a covalent interaction between the small molecule and target.We were initially discouraged by the lack of liganded sites identified by 3a; however, further analysis revealed that the reversible thioether bond was uniquely destabilized (t 1/2 < 0.1 h) due to the presence of two strong electron-withdrawing -CF 3 substituents.Although both thiol and sulfenic acid states were mapped onto more than 1,000 cysteines, only three of these sites were liganded by electrophilic and nucleophilic fragments (Fig. 2d).One such example is C217 of tRNA methyltransferase 61A (TRMT61A), which was liganded by 1, 2 and 3b (Fig. 2e).We also identified a small cohort of cysteines that were exclusively liganded by 3b (Fig. 2e,f), including C1592 of the small nuclear ribonucleoprotein U5 subunit 200 (SNRNP200), C398 of chaperonin containing T-complex protein1 subunit 3 (CCT3) and C335 of the heterogeneous nuclear riboprotein U (HNRNPU).Together, these findings indicate that the human sulfenome has a smaller but distinct ligandable space compared to the cysteinome.

Global S-sulfenome ligandability in the human proteome
Having established quantitative proteomic profiles that uncovered significant differences between cysteinome and sulfenome ligandability, we next turned our attention to measuring global S-sulfenate ligandability using our adapted SulfenQ workflow.With respect to library design, we prepared nucleophilic analogs of the electrophilic fragment library reported in 2016 (ref.9) (cyanoacetamides: 5a-30a, 35-41; nitroacetamides: 5b-30b and 42 and 43; Extended Data Fig. 4) and verified their reactivity using an established dipeptide sulfenic acid model (Supplementary Figs. 1 and 2).The preparation of cyanoacetamides from cyanoacetic acid was generally straightforward by EDC activation or in situ generation of the acid chloride and coupling to the corresponding amine.The synthesis of nitroacetamides was more challenging owing to the unstable nature of nitroacetic acid.Because standard amide coupling conditions could not be used for this series, several methods for nitroacetamide fragment preparation https://doi.org/10.1038/s41589-023-01330-5were used.As described in the Supplementary Note, some compounds were prepared by converting the nitroacetic acid to the acid chloride, followed by amide coupling.If this method was unsuccessful, we used bis(methylthio)-2-nitroethane as a masked carbonyl that could undergo addition/elimination with the desired amine.Nitroacetamide fragments that could not be prepared using either of these methods  were synthesized by acylation of the amine with bromoacetyl bromide to give the corresponding bromo amides, followed by installation of the nitro group using sodium iodide and silver nitrate.
With the nucleophilic fragments in hand, the 65-member library was screened against the sulfenome in a native lysate derived from MDA-MB-231 cells.In total, we quantified 8,235 BTD-labeled sites in 3,793 proteins (Supplementary Data 3).Among these, 4,845 (58.8%) sites were quantified in biological duplicates with a coefficient of variation (CV) less than 40%, underscoring the reproducibility of our method.On average, 1,958 BTD-labeled sites were quantified per dataset with median CV values ranging from 4.1% to 11.2%, and 6,132 (74.5%) cysteines were quantified in at least three fragment datasets (Extended Data Fig. 5).The nucleophilic fragment library exhibited noteworthy differences in target reactivity across the human sulfenome, with most showing liganded rates (also referred to as sulfenome reactivity) of less than 1.0%; only a single compound, 5b, had a significantly higher rate (20%; Fig. 3a,b).For example, C25 of the transport protein Sec61 gamma (SEC61G) was liganded by nitroacetamide 5b and not closely related analogs 3b or 9b (Fig. 3c).Likewise, C150 of tripeptidyl peptidase 2 (TPP2) was liganded by cyanoacetamide 5a and not closely related analogs 3a or 9a (Supplementary Data 3).
Next, we systematically investigated the relationship between structural features of recognition scaffolds and sulfenome reactivity.In general, nitroacetamide fragments with free NH groups displayed higher sulfenome reactivity than those without, whereas no apparent impact of molecular weight and CLogP was observed on liganded rates (Fig. 3d).Because the 'targeting' portion of the fragment had a minor impact on nucleophile reaction rate constants (Fig. 3e), the observed differences in ligandability are most likely attributed to binding energy imparted the unique scaffold.Similar observations were made for cyanoacetamide fragments (Extended Data Fig. 6a-c).
Irreversible nitroacetamide fragments exhibited higher liganded rates compared to reversible cyanoacetamide analogs (8.1% versus 1.7%; Fig. 3f and Extended Data Fig. 7).Among different scaffolds, m-trifluoromethyl o-methylphenyl 5a and 5b captured a comparatively large fraction of the sulfenome, yielding 61 and 248 liganded sites, respectively (Fig. 3g).Although 5a and 5b often targeted the same sites, the former was generally less potent (2.7% versus 21.8%; Fig. 3h), albeit with some interesting exceptions.Broadly reactive fragments such as 5a and 5b are commonly referred to 'scout' fragments 38 and provide privileged structures for evaluating ligandable sites.To independently corroborate the reactivity of fragments 5a and 5b, we synthesized the O-propargyl alkyne analogs 10a and 10b (Fig. 3i) and used them to probe the sulfenome.Both fragments were indeed broadly reactive, with 10b exhibiting greater proteome labeling than 10a (Fig. 3j).Control experiments also showed that protein labeling by 10a/10b was decreased but not completely abolished in the presence of an benzothiazine 'warhead' not functionalized with an alkyne known as benzyl-BTD 35 , suggesting likely differences in target preference conferred by the binding scaffold (Extended Data Fig. 8a).To place the phrase 'broadly reactive' across the sulfenome into greater context, roughly 3.7% of proteins with BTD-reactive sulfenic acids were liganded by 5a and 5b (Fig. 3k).Finally, we investigated whether the number of sites liganded by a scout fragment could be further increased in the context of a stressed proteome; native lysate was exposed to H 2 O 2 and then profiled using 5b in our chemoproteomic workflow.The number of sulfenic acid sites mapped by BTD and liganded by 5b were both increased under these conditions (Fig. 4a,b and Supplementary Data 4).Furthermore, H 2 O 2 treatment of lysate increased the extent of 5b target engagement, whereas many sites not liganded in the unstressed proteome became engaged by 5b (Fig. 4c).These data confirm the redox-dependent ligandability of a nucleophilic scout fragment, which increases with stress-induced cysteine oxidation, as expected.
Overall, the cyanoacetamide and nitroacetamide fragment libraries provided 524 liganded sites in 441 proteins, corresponding to 6.3% and 11.6% of the total mapped sites and proteins, respectively.This profile is distinct from that liganded by electrophilic fragments (Fig. 5a).Comparing the liganded dataset with intrinsic reactivity, hyper-reactive sulfenic acids exhibited lower ligandability than moderately reactive sulfenic acids (Fig. 5b).Another interesting feature to evaluate is modification stoichiometry as it relates to sulfenome ligandability.Although it is not possible to measure ligandability and modification stoichiometry (fraction or %SOH) in the same experiment, we were able to approximate this relationship using a recent dataset for a human sulfenome obtained from non-stressed cells (Fig. 5c).Comparing liganded sulfenic acids identified in this study to sites with available stoichiometry measurements, liganded sites are predicted to have a slightly lower occupancy (median value: 8.4% versus 14.5%).These collective observations provide further support for the notion that the fragment scaffold drives ligandability, along with a degree of covalent reversibility, as opposed to intrinsic electrophilic reactivity of proteomic sulfenic acids.

Nucleophilic covalent fragments affect the function of diverse proteins
The SulfenQ analyses herein constitute the largest human sulfenome dataset so far and, for the first time, reveals its potential for ligandability by nucleophilic fragments (Fig. 6a).Of these sites, a large fraction of proteins harboring sulfenic acids liganded by fragments in our nucleophilic library has not been recorded in the DrugBank 39 database (70.7%;Fig. 6b).In other words, nucleophilic small molecules targeted many proteins that currently have no known means of chemical interrogation.Of DrugBank database proteins with liganded sulfenic acids, 63.6% are characterized as enzymatic, whereas the non-DrugBank counterparts include a more diverse class of proteins (Fig. 6c).Furthermore, among all liganded sites identified in this study, only a small fraction is functionally annotated by UniProt (Fig. 6d).Nucleophilic fragments, therefore, represent unique chemical matter with tangible potential to illuminate new targets in the 'dark' proteome.https://doi.org/10.1038/s41589-023-01330-5 In subsequent studies, we assessed the ability of scout fragment 5b to modulate the biological function of proteins.Enzymes are a major class of drug targets and were highly enriched in proteins harboring sites liganded by nucleophilic fragments, particularly among the DrugBank (Fig. 6c).Covalent active site modification by fragment 5b is expected to irreversibly inhibit enzyme activity, and, to test this possibility, we selected three DrugBank enzymes targeted by nucleophilic ligands: GAPDH, GSTO1 and acetyl-CoA acyltransferase 1 (ACAT1).First, we conducted control experiments to establish that H 2 O 2 treatment increased covalent protein labeling by 10b, an alkyne-tagged 5b analog (Extended Data Fig. 8b); that protein labeling by 10b was decreased with increasing 5b competitor fragment; and that cysteine-to-serine mutation of the liganded sites in GAPDH, GSTO1 and ACAT1 diminished 10b labeling of variant proteins (Extended Data Fig. 8c,d).Having independently established these proteins as having active sites ligandable by 5b or 10b, we moved on to test the effect of ligation on catalytic activity.GAPDH inhibition by cysteine oxidation is well documented 40 and confirmed here by analyzing the effect of H 2 O on recombinant protein; oxidation of GSTO1 and ACAT1 active site cysteines similarly inhibited catalytic activity (Extended Data Fig. 9).H 2 O 2 -mediated inhibition of the enzymes was reversible and could be restored by treatment with dithiothreitol (DTT) and correlates to reduction of sulfenic acid to the active thiol form.On the other hand, incubation of fragment 5b with H 2 O 2 -treated enzymes led to the formation of the expected covalent thioether that was not reversible by DTT, and, therefore, catalytic activity could not be restored (Extended Data Fig. 9).Similar observations were made for nitroacetamide fragments 8b and 9b (Extended Data Fig. 9).
Encouraged by these biochemical findings, we next asked whether and how scout fragment 5b might modulate the function of enzymes in cells, particularly those that have not been recorded in the DrugBank.One such protein, peroxiredoxin-like 2A (PRXL2A), caught our attention because it harbors a 5b-liganded site (C85) within a redox-active CXXC motif (Fig. 6e).PRXL2A is an antioxidant enzyme that counteracts MAPK signaling, and the CXXC motif is critical for this inhibitory activity in cells 41 .We first confirmed that untagged 5b blocked 10b labeling of Flag epitope-tagged wild-type PRXL2A but not the C85S mutant in MDA-MB-231 cells (Extended Data Fig. 8e).Moreover, we found that treatment with fragment 5b indeed activated MAPK signaling, as shown by upregulated phosphorylation of p38 and ERK in a concentration-dependent manner (Fig. 6f).To establish that this effect is a direct result of C85 liganding by fragment 5b, we investigated the effect of C85S mutation in MDA-MB-231 cells transfected with a plasmid encoding wild-type or mutant PRXL2A.Notably, this mutation significantly attenuated the activation of MAPK signaling induced by fragment 5b (Fig. 6g).In addition, overexpression of wild-type PRXL2A, but not the C85S mutant, increased resistance to cell death induced by 5b (Fig. 6h), indicating that C85 in the redox-active CXXC motif of PRXL2A mediates, at least in part, the inhibitory activity of this fragment in cells.
We next turned our attention to liganded sites on proteins without enzymatic activity, among which is the oncoprotein heparin-binding growth factor (HDGF) 42 .Control experiments based on 5b/10b ligand competition and site-directed mutagenesis independently established C108 as the site in HDGF ligated by 5b (Extended Data Fig. 8f).HDGF C108 has been implicated in disulfide formation; however, it not known whether this site is redox regulated per se.We, therefore, investigated the effect of H 2 O 2 on HDGF sulfenation in MDA-MB-231 cells by expressing Flag-tagged wild-type HDGF and C108S mutant proteins.Sulfenic acid modification of HDGF in cells increased with exogenous H 2 O treatment, and C108S mutation abolished this effect, as expected (Fig. 6i).The overall level of HDGF sulfenic acid modification was increased in the C108S mutant compared to wild-type (Fig. 6i) and likely results from stabilized sulfenation at C12 in the absence of its resolving partner.
Given that HDGF can form a functional complex with nucleolin (NCL) protein 43 , we next asked whether the interaction between 5b and C108 site could perturb HDGF function through the NCL-protein kinase B (AKT) pathway 44 .Treatment of cells with 5b inhibited the uptake of HDGF into the nucleus (Fig. 6j) as well as its interaction with nuclear receptor NCL (Fig. 6k).AKT phosphorylation, a hallmark of pathway activation, decreased with 5b treatment in untransfected cells or in those overexpressing HDGF (Fig. 6l,m).HDGF C108S mutation attenuated the inhibitory effect of 5b on both AKT phosphorylation and the interaction between HDGF and NCL as well as their basal levels in the absence of stimulation (Fig. 6m).Furthermore, MDA-MB-231 cells overexpressing HDGF, but not the C108S mutant, exhibited an increase in cell survival (Fig. 6n).Together, these results suggest that the biological activity of nucleophilic fragment 5b in cells may be attributed, at least in part, to downregulation of the HDGF-NCL-AKT axis, specifically by targeting HDGF C108 in its sulfenic acid state.Our dataset offers many interesting leads to explore the importance of sulfenome ligandability by 'non-scout' fragment nucleophiles.Along these lines, we next examined the interaction of BRCA2 and CDKN1A-interacting DNA repair protein 45 (BCCIP) and compound 8b.We first confirmed the ligandability of C213 in BCCIP and the redox-sensitive nature of this liganded site (Extended Data Fig. 8g and Fig. 6o).BCCIP regulates DNA double-strand break-induced homologous recombinational repair (HRR) mainly through its interactions with two binding partners, BRCA2 and CDKN1A (generally referred to as p21).Considering that the liganded cysteine by 8b is localized in the p21-binding domain 45 , we next asked whether BCCIP function could be perturbed by this nucleophilic fragment in a p21-dependent manner.Indeed, 8b treatment inhibited the interaction of BCCIP with p21 in transfected HEK293T cells, whereas the C213S mutation ablated this effect (Fig. 6p).Furthermore, 8b significantly reduced the efficiency of HRR, mimicking the effect of siRNA-mediated knockdown of endogenous BCCIP (Fig. 6q).Conversely, 8b did not inhibit HRR in HEK293T cells expressing a C213S-BCCIP mutant (Fig. 6r).Together, these findings demonstrate that 8b targeting a functional sulfenic acid site in human BCCIP can modulate DNA damage repair, a process that has been implicated human diseases.

Discussion
Our findings reveal a vast expanse of ligandable sulfenic acids in the human proteome and highlight the utility of nucleophilic small molecules in the fragment-based covalent ligand discovery pipeline.The SulfenQ workflow, used in conjunction with the nucleophilic fragment library, enabled the identification of more sites of S-sulfenation in the human proteome than ever before.Of these, the great majority of liganded sites were identified in proteins with no known small-molecule probes, including proteins not found in the DrugBank or targeted by cysteine-reactive electrophilic fragments.Members of this nucleophilic fragment library were shown to irreversibly inhibit enzyme activity (GAPDH, GSTO1 and ACAT1) and modulate functional protein-protein interactions in cells (HDGF and its nuclear receptor; BCCIP and p21).An important question is whether there is value in targeting an oxidized, inactive form of an enzyme or protein.Here, we note that not all instances of cysteine oxidation are inhibitory, EGFR kinase being a prime example 17 .In cases where cysteine oxidation is inhibitory, targeting the enzyme or protein in this state would effectively prevent reactivation by biological reduction.For instance, stabilization of inactive oxidized PTP1B potentiates insulin signaling in a similar manner to inhibiting the catalytically active form of the enzyme 46 .Despite having the same fragment scaffold, nucleophilic small molecules liganded fewer sites than their electrophilic counterparts.The most likely explanation for this observation is differences in target abundance, as not all cysteinyl residues undergo redox reactions under physiological conditions.Also, the number of liganded sites identified in the cysteinome and sulfenome is dependent upon the redox environment.This point is illustrated by H 2 O 2 treatment of lysates, which increased targeting by scout fragment 5b.In addition to the identification of ligandable sites not accessed by electrophilic small molecules, targeting the sulfenome may also impart an additional layer of biochemical selectivity, as ligandability may be increased in pathological environments associated with increased oxidative stress.Similarly, the site occupancy or %SOH (and %SH) is also affected by changes in the redox environment.Site stoichiometry for post-translational modifications, including phosphorylation, can vary widely, and significant biological perturbations owing to sub-stoichiometric cysteine oxidation/modification are well known 47,48 .
Projecting forward, we envision several interesting pursuits to further expand the ligandability of the human sulfenome.Analogous to the landmark study reporting cysteine-targeted electrophilic fragments 9 , our library comprises fewer than 100 compounds.The chemical space of our libraries will undoubtedly benefit from the coupling of cyanoacetamide and nitroacetamide C-nucleophiles (and other linear and cyclic C-nucleophiles 15,26,34,35 ) to an expanded set of fragment amines and to fragments with different linkage types.Optimization of initial ligands into more advanced chemical probes may also be achieved through traditional structure-activity relationshipbased methods or by application of functionalized scout fragments for more facile screening 38 .Finally, because the sulfenome is context dependent, future studies will most certainly include global comparisons of ligandability across different cell/tissue types and redox states.
In summary, nucleophilic fragments that target sulfenic acid offer a complementary approach to electrophile-based covalent ligand discovery and expand access to proteins lacking small-molecule probes.The targeting of this oxidative cysteine modification and others, such as sulfinic acid or even disulfides, can be leveraged in future work to develop a new class of chemical probes and pharmaceuticals that exploit changes in cellular redox homeostasis.Liganded sites identified by nucleophilic fragments may also serve as the starting point for development of small molecules that are agnostic to changes in cysteine redox state.More generally, our approach underscores the important, broader effort to discover covalent ligands that target functional groups in proteins beyond cysteine and other nucleophilic residues, including cofactors 49 and non-redox-based post-translational modifications.

Chemicals
Chemical synthesis of the compounds, kinetic reactivity and thioether adduct stability was performed as indicated in the Supplementary Note.

Cell culture for chemoproteomics
MDA-MB-231 cells were grown to 90% confluence and washed with cold PBS (Thermo Fisher Scientific, C10010500), scraped with cold PBS and cell pellets were isolated by centrifugation (200g, 3 min, 4 °C).Cell pellets were either directly used or stored at −80 °C for further use.

Probe labeling and proteomic sample preparation
For analyzing intrinsic sulfenic acid reactivity in the human proteome, MDA-MB-231 cell pellets were lysed by sonication in four volumes of lysis buffer (50 mM HEPES pH 7.5, 150 mM NaCl and 1% IGEPAL) supplemented with 1× protease and phosphatase inhibitors (Bimake, B14012) and 200 U ml −1 catalase (Sigma-Aldrich, C1345).Cell lysates were divided into two parts and incubated with 0.5 mM or 5 mM BTD at 37 °C for 2 h with rotation, respectively.
For analyzing ligandability of fragments, lysates were incubated with the fragment or DMSO at 37 °C for 2 h with rotation.Fragments were made up as 50 mM stock solutions in DMSO and stored at −80 °C.Final concentrations were 500 µM for electrophilic fragments and 5 mM for nucleophilic fragments.For targeting the cysteinome, lysates were further incubated with 100 µM IPM (Kerafast, EVU111) at room temperature for 1 h.For targeting the sulfenome, lysates were further incubated with 5 mM BTD (empirically established to achieve adequate analytic depth of the S-sulfenome in our chemoproteomic workflow 27 ) at 37 °C for 1 h.After incubation with IPM or BTD, 10 mM TCEP (Sangon Biotech, A600974) was added at room temperature for 10 min, followed by alkylation with 40 mM IAM at room temperature for 30 min in the dark with rotation.
Protein pellets were resuspended with 50 mM ammonium bicarbonate (NH 4 HCO 3 , Sigma-Aldrich, A6141).Protein concentrations were measured by BCA (Tiangen, PA115) and adjusted to 2 mg ml −1 .The resulting samples (2 mg of protein per milliliter in 1-ml volume) were digested with sequencing-grade trypsin (Sigma-Aldrich, A6141) at a 1:50 (enzyme:substrate) ratio overnight at 37 °C.Tryptic digests were desalted with HLB extraction cartridges (Waters, 186000383) and evaporated to dryness.For quantitative analyses, the desalted tryptic digests were reconstituted in a solution containing 30% ACN (Sigma-Aldrich, A6141).CuAAC reaction was performed by addition of 1 mM azido-L-biotin (Kerafast, EVU102) or azido-H-biotin (1 µl of a 40 mM stock; Kerafast, EVU151), 10 mM sodium ascorbate (4 µl of a 100 mM stock, Sigma-Aldrich, A7631), 1 mM TBTA (1 µl of a 50 mM stock, Sigma-Aldrich, 678937) and 10 mM CuSO 4 (4 µl of a 100 mM stock, Sigma-Aldrich, 678937).Samples were allowed to react at room temperature for 2 h in the dark with rotation.The light and heavy Az-UV-biotin-labeled samples then were mixed equally together, purified by strong cation exchange (SCX) spin columns (The Nest Group, SMM HIL-SCX) and then enriched with streptavidin beads (GE, 17-5113-01) for 2 h at room temperature.Streptavidin beads were subsequently washed with 50 mM sodium acetate, 50 mM sodium acetate containing 2 M NaCl and water twice each with vortexing and/or rotation to remove non-specific binding peptides and then resuspended in 25 mM NH 4 HCO 3 , transferred to thin-walled borosilicate glass tubes and irradiated with 365 nm UV light (UVP, UVL-28 EL series) for 2 h at room temperature with magnetic stirring.The supernatant was collected, concentrated under vacuum and desalted with HLB extraction cartridges.The desalting peptides were evaporated to dryness and stored at −20 °C until liquid chromatography with tandem mass spectrometry (LC-MS/MS) analysis.

LC-MS/MS
LC-MS/MS analysis was performed on a Q Exactive Plus (Thermo Fisher Scientific) in line with an Easy-nLC1000 system (Thermo Fisher Scientific).Samples were reconstituted in 0.1% formic acid followed by centrifugation (16,000g for 10 min), and supernatants were pressure loaded onto a 2-cm microcapillary pre-column packed with C18 (3 mm, 120 Å, SunChrom).The pre-column was connected to a 12-cm, 150-mm-inner-diameter microcapillary analytical column packed with C18 (1.9 mm, 120 Å, Dr. Maisch) and equipped with a homemade electrospray emitter tip.The spray voltage was set to 2.0 kV, and the heated capillary temperature was set to 320 °C.The LC gradient consisted of 0 min, 7% B; 14 min, 10% B; 51 min, 20% B; 68 min, 30% B; and 69-75 min, 95% B (A = water and 0.1% formic acid; B = ACN and 0.1% formic acid) at a flow rate of 600 nl min −1 .Higher-energy collisional dissociation (HCD) MS/MS spectra were recorded in the data-dependent mode using a top 20 method.MS1 spectra were measured with a resolution of 70,000, an AGC target of 3 × 10 6 , a maximum injection time of 20 ms and a mass range from m/z 300 to 1,400.HCD MS/MS spectra were acquired with a resolution of 17,500, an AGC target of 1 × 10 6 , a maximum injection time of 60 ms, a 1.6-m/z isolation window and normalized collision energy of 30.Peptide m/z values that triggered MS/MS scans were dynamically excluded from further MS/MS scans for 18 s.

Peptide identification and quantification
Raw data files were searched against the Homo sapiens UniProt canonical database (20190924) using pFind Studio (version 3.1.5,http:// i.pfind.org/).The precursor ion mass and fragmentation tolerance were 10 ppm and 20 ppm, respectively, for the database search.A specific tryptic search was used with a maximum of three missed cleavages allowed.A maximum of three modifications were allowed per peptide.For all analyses, modifications of +15.9949Da (methionine oxidation, M) and +57.0214Da (IAM alkylation, C) were searched as variable modifications.For site-specific mapping of probe-modified sites, mass shifts of +252.122(C11H16O3N4) for IPM and +418.131(C19H22O5N4S1) for Article https://doi.org/10.1038/s41589-023-01330-5BTD were searched as variable modifications, respectively.No fixed modifications were searched for the sulfenome.A differential modification of 6.0201 Da on probe-derived modifications was used for all analyses.The false discovery rates (FDRs) at spectrum, peptide and protein level were less than 1%.Quantification of heavy to light ratios (R H/L ) was performed using pQuant as previously described 37 , which directly uses the raw files as the input.pQuant calculated R H/L values based on each identified MS scan with a 15-ppm-level m/z tolerance window and assigned an interference score (Int.Score) to each value from 0 to 1.The median values of probe-modified peptide ratios with σ less than or equal to 0.5 were considered to calculate site-level ratios.To obtain site-centric quantification data, output reports from pFind were further processed using in-house software written in the R programming language 37 .

Functional annotation of liganded cysteines
Functional annotation of liganded cysteines was performed using an in-house R script against UniProtKB/Swiss-Prot Protein Knowledge Database (release-2020_10).Relevant UniProt entries were mined for available functional annotations at the residue level, specifically for annotations regarding active sites, disulfides, metal binding sites and modified sites.Liganded proteins were queried against the DrugBank database (version 5.1) and partitioned into DrugBank and non-DrugBank proteins.

In situ fragment treatment and cell lysis
For HDGF and PRXL2A, MDA-MB-231 cells were grown until 80-90% confluency, washed once with PBS and treated with 5b prepared in serum-free medium as indicated concentrations at 37 °C for 2 h.For BCCIP, HEK293T cells were grown until 80-90% confluency, washed once with PBS and treated with 8b prepared in serum-free medium at the indicated concentrations at 37 °C for 12 h.Next, the media was removed, and cells were washed with PBS twice and then scrapped in NETN buffer (50 mM HEPES pH 7.5, 150 mM NaCl and 1% IGEPAL) supplemented with protease inhibitors.After centrifugation at 16,000g for 5 min at 4 °C, the supernatants were collected and adjusted to 2 mg ml −1 for subsequent analysis.
For HDGF and PRXL2A, transfection was performed by incubating 5 µg of plasmid and 10 µl of Xfect transfection reagent (Vazyme Biotech, T101-01) with MDA-MB-231 cells at 80% confluency on a six-well plate or 20 µg of plasmid and 40 µl of Xfect transfection reagent with MDA-MB-231 cells at 80% confluency on a 10-cm dish.Cells were cultured in DMEM supplemented with 10% FBS for another 48 h.For BCCIP, transfection was performed by incubating 5 µg of plasmid and 10 µl of Xfect transfection reagent (Vazyme Biotech, T101-01) with HEK293T cells at 80% confluency on a six-well plate or 15 µg of plasmid and 30 µl of Xfect transfection reagent with HEK293T cells at 80% confluency on a 10-cm dish.Cells were cultured in DMEM supplemented with 10% FBS for another 48 h.

Immunoprecipitation
For validation of liganded proteins by fragments 5b and 8b, protein lysates were treated with the corresponding alkyne-tagged probes, 10b and BTD, at 37 °C for 1 h.DMSO-treated samples were used as negative control.Probe-labeled proteomes were 'clicked' with azide-biotin and incubated with streptavidin beads at 4 °C overnight.Subsequently, the beads were collected, washed with lysis buffer, eluted with 2× SDS-PAGE protein loading buffer and analyzed by immunoblotting.
For analyzing sulfenation of Flag-tagged HDGF, cell pellets from MDA-MB-231 cells expressing Flag-tagged HDGF or the C108S mutant were treated with H 2 O 2 (0.5 mM, 10 min) or left untreated, washed three times with PBS and lysed in the presence of BTD (5 mM) on ice for 20 min in NETN buffer with protease inhibitor cocktail and catalase (200 U ml −1 ).For analyzing sulfenation of Flag-tagged BCCIP, cell pellets from HEK293T cells expressing Flag-tagged BCCIP or the C213S mutant were treated with H 2 O 2 (0.5 mM, 10 min) or left untreated, washed three times with PBS and lysed in the presence of BTD (5 mM) on ice for 20 min in NETN buffer with protease inhibitor cocktail with catalase (200 U ml −1 ).BTD-labeled proteomes were 'clicked' with azide-biotin.Next, samples were incubated with streptavidin beads at 4 °C overnight.Subsequently, the beads were collected, washed with lysis buffer, eluted with 2× SDS-PAGE protein loading buffer and analyzed by immunoblotting.
To analyze the HDGF-NCL interaction, MDA-MB-231 cells expressing Flag-tagged HDGF or the C108S mutant were washed twice with cold PBS and harvested.To analyze the BCCIP-p21 interaction, HEK293T cells expressing Flag-tagged BCCIP or the C213S mutant were washed twice with cold PBS and harvested.Cells were lysed in HEPES lysis buffer (50 mM HEPES, 150 mM NaCl and 1% (vol/vol) IGEPAL, pH 7.5) supplemented with protease inhibitors and catalase (200 U ml −1 ).Flag-tagged wild-type and mutant were immunoprecipitated from 1 mg of lysate with 20 µl of ANTI-FLAG M2 affinity gel (Sigma-Aldrich, A2220) by rotation.After incubation at 4 °C for 16 h, the agarose resin was collected and washed with lysis buffer, eluted with 2× SDS-PAGE protein loading buffer and analyzed by immunoblotting.

Western blotting
Proteins were resolved by 12% SDS-PAGE and transferred to polyvinylidene difluoride (PVDF).After transfer, the PVDF membrane was blocked with 5% milk or BSA (for phosphorylation detection) in TBST for 1 h at room temperature.The membrane was washed with three times for 5 min with TBST and incubated with indicated primary antibodies overnight at 4 °C.Next, the membrane was washed three times with TBST and incubated with the appropriate HRP-conjugated secondary antibody.The PVDF membrane was then washed with three times for 5 min with TBST, visualized on a Tanon 5200 scanner and analyzed in GelCap software version 5.6 with ECL chemiluminescence.

DNA repair assays
DNA repair of double-strand break by homologous recombination was assessed by a cell-based plasmid integration assay 50 .In brief, HEK293T cells were co-transfected with DR-GFP reporter and I-SceI expression plasmid (pCBA-I-SceI, obtained from Yali Chen from the Beijing Institute of Lifeomics).After 48 h, cells were harvested or treated as indicated and analyzed by flow cytometry (BD FACSVerse) to determine the extent of recombination induced by I-SceI digestion.In parallel, pCBA-I-SceI was co-transfected with pCherry plasmid (Clontech) to normalize for transfection efficiency. https://doi.org/10.1038/s41589-023-01330-5

Fragment binding assay
For in vitro assays, GSTO1, GAPDH or ACAT1 protein was treated with H 2 O 2 (2 eq) and then incubated with 0, 0.05, 0.1, 0.5 or 1 mM 5b at 37 °C for 1 h, followed by 10b (1 mM) at 37 °C for 1 h.10b-labeled proteins were 'clicked' with azide-biotin and analyzed by immunoblotting.For in cellulo assays, cells expressing Flag-tagged proteins or their mutants were lysed in the presence of 0.05, 0.1, 0.5 or 1 mM 5b or 8b at 37 °C for 1 h, followed by 1 mM 10b or BTD at 37 °C for an additional 1 h.Probe-labeled proteomes were 'clicked' with azide-biotin and captured with streptavidin beads at 4 °C overnight.Finally, the beads were collected, washed with lysis buffer, eluted with 2× SDS-PAGE protein loading buffer and analyzed by immunoblotting.

Immunofluorescence
MDA-MB-231 cells were cultured onto confocal dishes and treated with or without 50 µM 5b.After incubation at 37 °C for 2 h, cells were fixed in paraformaldehyde for 30 min at room temperature and permeabilized with 0.2% Triton X-100 for 30 min at 4 °C.Cells were blocked in 3% BSA in PBS, and fixed cells were then incubated with anti-HDGF at 4 °C overnight, followed by incubation with the appropriate fluorescent secondary antibodies.The nucleus was stained with DAPI (ZSGB-BIO, ZLI-9557, diluted at 1:1,000), and fluorescence was observed using a laser scanning microscope (Zeiss, LSM880 ELYRAS.1).

Enzyme activity assays
Recombinant proteins were reduced with 1 mM DTT for 30 min at 4 °C.DTT was then removed using Zeba Spin Desalting Columns equilibrated with PBS.DTT-free proteins were treated with H 2 O 2 (5 eq) at 37 °C for 1 h.Reactions were quenched by passage through a Zeba Spin Desalting Column.For ACAT1, 1 mM of 5a, 5b or 8b was added for 1 h at 37 °C.For GAPDH, 1 mM of 5a, 5b, 8b or 9b was added for 1 h at 37 °C.For GSTO1, 1 mM of 9b or 5b was added for 1 h at 37 °C.After incubation, 1 mM DTT was added, and the reaction incubated for 1 h.The enzymatic activity of ACAT1 was measured using the Fluorometric Acetyltransferase Activity Assay Kit (Abcam, ab204536).The assay was performed per the protocol with 2 µg of ACAT1 protein and 100 nM acetyl CoA (Sigma-Aldrich, 10101893001).The reaction mixture was incubated for 30 min at 37 °C with shaking at 120 r.p.m. and then terminated by the addition of isopropyl alcohol (50 µl).Detection solution (100 µl, supplied with the kit) was added and incubated for 10 min at room temperature.Fluorescence was measured at 380/520 excitation/emission (ex/em) on a SpectraMax microplate reader.The activity of GAPDH enzyme was quantified by the GAPDH Activity Assay Kit (Abcam, ab204732).In brief, 2 µg of GAPDH was used to measure GAPDH activity in 96-well plates.Absorbance was measured at 450 nm in kinetic mode on a SpectraMax microplate reader.The enzymatic activity of GSTO1 was measured using a fluorometric GST Activity Assay Kit (Abcam, ab65325).In brief, 2 µg of GSTO1 was used to measure enzymatic activity using 96-well plates according to the manufacturer's protocol.Fluorescence was measured at ex/em = 380/461 nm on a SpectraMax microplate reader.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Fig. 1 |
Fig.1| Quantitative profiling of intrinsic sulfenic acid reactivity in the human proteome.a, Chemical structures of a 'clickable' sulfenic acid-specific probe, BTD, and isotopically labeled azide-biotin reagents with a photocleavable linker (Az-UV-biotin).b, Adapted SulfenQ protocol for site-specific quantification of the intrinsic reactivity of cysteinyl sulfenic acids involves proteome labeling, click-chemistry-based incorporation of isotopically cleavable tags and protease digestion to provide probe-labeled peptides for MS analysis.Ratios for peptides were measured from MDA-MB-231 native cell lysates labeled with 0.5 mM or 5 mM BTD and are indicated as R 10:1 .c, Correlation of R 10:1 values with functional annotations from the UniProt database.Active sites, disulfide bonds or metal binding sites are shown in red, and all other quantified sulfenic acid sites are shown in black.A moving average line of functional annotated sites is shown in a dashed blue line.d, Pie chart showing the distribution of the number of probe-labeled sulfenic acid sites per protein.Active sites are denoted with red bars.e, Bar chart showing the number of hyper-reactive and all quantified sulfenic acid sites per protein for proteins found to contain at least one hyper-reactive site.f,g, Representative extracted ion chromatograms (XICs) showing changes in BTD-labeled peptides from known redox-sensitive proteins (f) and those bearing functionally important cysteines (g).Profiles for lightlabeled and heavy-labeled peptide are shown in red and blue, respectively.The average R 10:1 values calculated from four biological replicates are displayed below each XIC.h, Pie chart showing the percentage of previously identified S-sulfinated (-SO 2 H) sites across the intrinsic sulfenic acid reactivity ranges.i, Correlation between the intrinsic reactivity of reduced (-SH) and oxidized (-SOH) forms of the same cysteine residues across the MDA-MB-231 proteome.The former was retrieved from ref. 9.The latter was generated in this study by SulfenQ.j, Representative XICs showing changes in probe-labeled peptides bearing protein cysteines as indicated.Profiles for light-labeled and heavylabeled peptide are shown in red and blue, respectively.Left: IA-labeled peptides; average R SH 10:1 values are from ref. 9. Right: BTD-labeled peptides; average R SOH 10:1 values calculated from four biological replicates are displayed below each XIC.corr., correlation. Articlehttps://doi.org/10.1038/s41589-023-01330-5

Fig. 3 |
Fig. 3 | Global analysis of nucleophilic fragment-sulfenic acid interactions in a human proteome.a, Bar chart showing sulfenome reactivity values, referred to as liganded sulfenic acid rates, for nucleophilic fragments calculated as the percentage of all quantified sulfenic acid sites with R H/L values ≥4 for each fragment.MDA-MB-231 cell lysates were incubated with fragment (0.5 mM) or DMSO vehicle and then labeled with BTD (5 mM).Probe-labeled proteomes were processed and analyzed to obtain R H/L values for each profiled sulfenic acid site.b, Heat maps showing R H/L values (R H/L ≥ 4) of all sulfenic acid sites liganded by nitroacetamide (left) and cyanoacetamide (right) nucleophile fragments.ND, not detected.c, Representative extracted XICs showing changes in BTD-labeled peptides from SEC61G.The profiles for light-labeled and heavy-labeled peptide are shown in red and blue, respectively.The average R H/L values calculated from biological duplicates are displayed below each XIC.d, Scatter plots showing

Fig. 5 |
Fig. 5 | Sulfenome ligandability in the human proteome is unique.a, Comparison of ligandability of the MDA-MB-231 cysteinome and sulfenome.Venn diagram showing the overlap of IPM-profiled and BTD-profiled sites (left) and proteins (right).b, Bar charts showing total number (left) and percentage (right) of liganded sites per total number of sulfenic acids quantified across the indicated intrinsic reactivity ranges.c, Estimated stoichiometry of sulfenic acid modification at sites identified as liganded by nucleophilic fragments in the current study.The fraction of an individual cysteine in the sulfenic acid form is referred to as %SOH, and these values were retrieved from ref. 15.Bar chart shows the distribution of %SOH in sites liganded by nucleophilic fragments.

Fig. 6 |
Fig. 6 | Ligandability by nucleophilic fragments affects the function of diverse proteins.a,b, Pie charts showing the percentage of sulfenic acid sites and proteins liganded by nucleophilic fragments (a) or the percentage of liganded proteins found in DrugBank (b).c, Functional categorization of DrugBank and non-DrugBank proteins containing liganded sulfenic acid sites.d, Bar chart showing the functional categorization of liganded and unliganded sulfenic acid sites based on cysteine residue annotations in UniProt.e, Representative extracted XICs showing changes in probe-labeled peptides from proteins as indicated.Average R H/L values calculated from biological duplicates are displayed below each XIC.'p-' denotes phosphorylated.f,g, Western blots showing activation of MAPK signaling in MDA-MB-231 cells by 5b (f) and that 5b treatment (50 µM, 2 h) activates MAPK signaling in cells overexpressing PRXL2A but not the C85S mutant (g).h, Comparison of MDA-MB-231 cell survival rates expressing wild-type (WT) or C85S mutant PRXL2A upon 5b treatment.i, Western blots showing that H 2 O 2 treatment (0.5 mM, 10 min) leads to an increase of sulfenation on Flag-tagged HDGF but not the C108S mutant in MDA-MB-231 cells.j, Immunofluorescence images showing that 5b treatment (50 µM, 2 h) inhibits the uptake of HDGF into the nucleus.Scale bar, 5 µm.k-m, Immunoprecipitation-western blot showing that 5b

Extended Data Fig. 1 | 5 Extended Data Fig. 5 | 5 Extended Data Fig. 8 |
Global analysis of nucleophilic fragment-SOH interactions.(a) Rank plot showing R 10:1 (5 mM vs 0.5 mM, black) and R 1:1 (5 mM vs 5 mM, blue) values of BTD-labeled SOH sites across the MDA-MB-231 proteome.MDA-MB-231 cell lysates were labeled with BTD at concentration as indicated, and digested by trypsin.The resulting BTD-modified peptides were conjugated to light and heavy Az-UV-biotin, respectively, via click chemistry.The light and heavy labeled samples then were mixed equally in amount and subjected to streptavidin-based enrichment.After several washing steps, the modified peptides were selectively eluted from beads under 365 nm wavelength of UV light for LC-MS/MS-based proteomic analysis.(b) Line series plot showing the comparison of R H/L values obtained from common sites in the 10:1 and 1:1 dataset.(c) Representative XICs showing changes in BTD-labeled peptides from proteins as indicated.The profiles for light-and heavy-labeled peptide are shown in red and blue, respectively.The average R H/L values calculated from biological duplicates are displayed below each XIC.https://doi.org/10.1038/s41589-023-01330-Qualitycontrol of the BTD-based SulfenQ analyses.(a) Distribution of coefficient of variation values for SulfenQ analysis of each fragment.(b) Accumulating number of sulfenylated sites profiled by BTD across all datasets.(c) Frequency of quantification of all SOHs across all SulfenQ analyses performed with nucleophilic fragments.Extended Data Fig. 6 | Interactions between cyano-acetamide-based fragments and the MDA-MB-231 sulfenylome.(a) Scatter plots showing correlation of sulfenylome reactivity of each cyano-acetamide-based fragment with the corresponding structural features, including number of free NH (Left), molecular weight (MW, middle), and CLogP value (Right).(b) Scatter plots showing correlation of sulfenylome reactivity of each cyano-acetamide-based fragment with the corresponding kinetic parameters, including the observed binding constant (k obs , Left) and reaction rate constant (M -1 S -1 , Right).(c) Representative XICs showing 5a-induced changes in BTD-labeled peptides bearing protein cysteines as indicated.The profiles for light-and heavylabeled peptide are shown in red and blue, respectively.The average R H/L values calculated from biological duplicates are displayed below each XIC.https://doi.org/10.1038/s41589-023-01330-Validation of liganded proteins.(a) MDA-MB-231 cell lysates were pre-treated non-clickable BTD (BnBTD, 5 mM, 1 h), followed by treatment of 10a or 10b at the concentration of 5 mM for additional 1 h, clicked with azido biotin and analyzed by western blotting.(b) Recombinant, purified wild-type GSTO1, ACAT1 and GAPDH proteins were pretreated with H2O2 (100 µM, 10 min), followed by treatment with 500 µM 10b probe for 1 h.Protein samples were 'clicked' with azide-biotin and analyzed by western blotting.(c) Recombinant, purified wild-type GSTO1, ACAT1 and GAPDH proteins were treated with 5b as indicated concentrations followed by treatment with 10b probe, 'clicked' with azide-biotin and analyzed by western blotting.(d) Recombinant, purified wild-type and the corresponding cysteine-to-serine mutants as GSTO1, ACAT1 and GAPDH proteins were treated with 10b at the indicated concentration, clicked and analyzed by western blotting.(e,f) Validation of liganded proteins by 5b in cellulo.MDA-MB-231 cellsrecombinantly expressing wild-type and the corresponding cysteine-to-serine mutants as Flag epitope-tagged proteins (HDGF and PRXL2A) were treated with 5b as indicated concentrations followed by treatment with 10b probe.Protein samples were harvested, 'clicked' with azide-biotin, immunoprecipitated with streptavidin, eluted and separated with SDS-PAGE.Western blotting was performed using antibodies as indicated.Representative data from at least two independent experiments are shown.(g) HEK293T cells recombinantly expressing wild-type and the corresponding cysteine-to-serine mutant as Flag epitope-tagged BCCIP were treated with 8b as indicated concentrations followed by treatment with BTD probe (5 mM, 1 h).Protein samples were harvested, 'clicked' with azide-biotin, immunoprecipitated with streptavidin, eluted and separated with SDS-PAGE.Western blotting was performed using antibodies as indicated.Each experiment was repeated three times with similar results.ExtendedData Fig. 9 | Functional analysis of interactions between nucleophilic fragments and their liganded enzymes.(a) Gel blots showing the purity of each recombinant proteins as indicated.(b) Bar charts showing that the redox-dependent changes in enzymatic activities can be perturbed by the indicated nucleophilic fragments.Data are mean ± s.d.(representative data from biological triplicates are shown, two-sided Student's t-test, P < 0.05 was considered significant).Each experiment was repeated three times with similar results.(c) 3D protein structures of enzymes (PDB#:1EEM for GSTO1; PDB#:2F2S for ACAT1; PDB#:1U8F for GAPDH) were visualized with Pymol 2.1.1.Peptide containing liganded sites are highlighted in blue.Enzyme substrates or ligand are highlighted in red.(d) Representative XICs showing changes in BTD-labeled peptides from proteins as indicated.The profiles for light-and heavy-labeled peptide are shown in red and blue, respectively.The average R H/L values measured from biological duplicates are displayed below each XIC.

0.00% 1.3% 25.2% 11%
Venn diagram showing the comparison of the sulfenome obtained from MDA-MB-231 cells with or without H 2 O 2 treatment.MDA-MB-231 cells were treated with or without H 2 O 2 (0.5 mM, 15 min), followed by treatment of DMSO vehicle or 5b (2 mM, 1 h) and subsequent probe labeling with BTD (5 mM, 1 h).The resulting probe-labeled proteomes were processed and analyzed by LC-MS/MS.b, Rank plot of R H/L (DMSO versus 5b) values obtained from samples with or without H 2 O 2 treatment.c, Line series plots showing the comparison of the R H/L ratios obtained from samples with or without H 2 O 2 treatment.Left panel: liganded sites whose fragment-sulfenic acid interactions increased with H 2 O 2 treatment.Right panel: unliganded sites that were liganded after H 2 O 2 treatment.