Rational Design of Reversible and Irreversible Cysteine Sulfenic Acid-Targeted Linear C-Nucleophiles

Abstract

Concerns about off-target effects has motivated the development of reversible covalent inhibition strategies for targeting cysteine. However, such strategies have not been reported for the unique cysteine oxoform, sulfenic acid. Herein, we have designed and identified linear C-nucleophiles that react selectively with cysteine sulfenic acid. The resulting thioether adducts exhibit reversibility ranging from minutes to days under reducing conditions, showing the feasibility of tuning C-nucleophile reactivity across a wide range of time scales.

Graphical Abstract

Oxidation of a protein cysteine thiol (Cys-SH) to sulfenic acid (Cys-SOH) by endogenous reactive oxygen species (ROS), including H₂O₂, is a reversible post-translational modification that plays a central role in regulating the activity, location, and function of a wide variety of proteins.¹ For example, Paulsen et al. reported that a critical active site cysteine (Cys797) in the epidermal growth factor receptor (EGFR) is S-sulfenylated by H₂O₂ induced by growth factor stimulation, which enhances its intrinsic tyrosine kinase activity.² Recently, we have shown that SIRT6 exerts redox control of HIF1A transcriptional activity via sulfenyl-mediated complex formation.^{3, 4} Aberrant Cys-SOH formation during oxidative stress or chronic inflammation has also been linked to human diseases, including cancer.^{5, 6}

The FDA approval of kinase inhibitors such as Afatinib and Ibrutinib targeting cysteine residues led to the re-emergence of covalent inhibition as a drug design strategy.⁷ Concerns about off-target modification has motivated the exploration of reversible covalent chemistry. Several such strategies have been reported, including bromomaleimide-based reagents that are labile under reducing conditions (Fig. S1A) and electron-deficient olefins, like acrylamides that rapidly and reversibly react with cysteine (Fig. S1B).^{8, 9} However, the aforementioned reversible covalent inhibitors are not capable of reacting with Cys-SOH under physiological conditions and cysteine oxidation is expected to impact their pharmacology. For this reason, we have proposed a complementary strategy in which the sulfenyl form of therapeutically important proteins could be exploited to develop a new class of irreversible and reversible covalent inhibitors targeting this unique cysteine oxoform.^{1, 2, 10–12}

Due to the electrophilic as well as nucleophilic character of the sulphur atom in Cys-SOH, detection methods exploiting both chemical features have been reported.^{1, 12} However, under physiological conditions, it is the electrophilic reactivity of Cys-SOH that dominates and the vast majority of probes for selective chemical detection are C-nucleophiles based on cyclic 1,3-carbonyl scaffolds, such as dimedone (R = R′ = Me, n = 1) (Fig. S1C). Such chemical probes have been widely employed for qualitative and quantitative study of the sulfenome. ^{1, 3, 13} Although dimedone and closely related probes are selective under aqueous conditions, they are collectively hampered by slow reaction kinetics when compared to competing biological fates of Cys-SOH (e.g., thiol-based reduction to form a disulfide or over-oxidation to sulfinic acid).¹

To overcome the issue of slow reactivity, we recently developed a mass spectrometry-based assay and dipeptide sulfenic acid model (1) that enabled us to screen a library of cyclic C-nucleophiles and identify novel compounds with >200-fold enhanced reactivity, compared to dimedone.¹² In general, cyclic C-nucleophiles form stable covalent adducts with Cys-SOH, a feature that renders such compounds useful for more comprehensive analyses of the sulfenome and as warheads for the development of irreversible covalent inhibitors that target Cys-SOH.^{2, 11} On the other hand, the irreversible nature of this reaction may increase the chances of forming permanent adducts with undesired proteins and precludes applications in which target release is favoured. Reversible covalent probes of Cys-SOH are yet unknown and represents a significant hurdle in the development of nucleophilic inhibitors.

Herein, we have employed our LC/MS assay to screen six different classes (3–8) of linear C-nucleophiles (Fig. 1). By contrast to cyclic 1,3-dicarbonyls, the keto-enol equilibrium is predominantly shifted towards keto form in linear 1,3-dicarbonyls (3).¹⁴ We reasoned that this shift towards keto tautomer should manifest as a difference in reactivity towards Cys-SOH and in the stability of the corresponding thioether adduct. Moreover, a systematic study of linear C-nucleophiles would also increase the diversity of chemical tools for use as probes and inhibitors of sulfenic acid-modified proteins.

Fig. 1.

Structure and reaction of various classes of linear C-nucleophiles with sulfenic acid 2.

Initial support for our hypothesis was obtained from the reactivity of 2,4-pentanedione (3a) towards sulfenic acid 2. Notably, 3a exhibited a 4:1 ratio of keto:enol tautomers and reacted with k_obs = 3.8 min⁻¹, almost 200-fold higher than its 5-membered cyclic analog (i.e., cyclopentane-1,3-dione, k_obs = 0.02 min⁻¹)¹². Next, we explored the effects of R₁ and R₂ substitution of 3 with different alkyl, aryl and electron-withdrawing groups (EWGs), such as –CO₂H, -CO₂Me, -CF₃. In general, while mono alkylated/arylated substrates 3b–c gave k_obs similar to 3a, diarylated analog 3d or analogs with EWGs in place of methyl, like 3e–j, reduced k_obs (Fig. S2). This decrease in activity can be attributed to the shift in keto-enol equilibrium towards the enol tautomer and enhanced carbanion stability at C-3. Moreover, substitution by highly EWGs, such as –CF₃, may lead to hydration of the carbonyl and complete loss of reactivity due an increase in C-3 carbanion pKa).¹⁵ Subsequently, for nucleophiles of structural class 4 and 5 we postulated that replacing one or both carbonyls with a carboxylate ester, amide or hydrazide functionality would increase the pK_a of the nucleophilic C-centre. Consistent with this hypothesis, we observed a marked decrease in reactive carbanion formation at physiological pH and a corresponding reduction in reactivity (Fig. S2).

An apparent trend in C-nucleophile reactivity towards Cys-SOH is that carbanion destabilization increases its reactivity, at least within a specific pKa range (i.e., 4–10).¹² Along these lines, we theorized that replacement of one carbonyl group with a sulfone with concomitant partial loss of resonance would also increase reactivity even further. For this reason, we evaluated linear C-nucleophiles in which one carbonyl was replaced with a sulfone functional group (6). Gratifyingly, a k_obs of 18.8 min⁻¹ or 25-fold increase relative to dimedone (k_obs = 0.8 min⁻¹) was observed for the reaction between 6d and sulfenic acid 2 (Fig. S2). Next, we evaluated the reactivity of molecules in which one or both carbonyls were replaced by an EWG such as nitro (-NO₂, 7) and cyano (-CN, 8). Given that the pK_as of 7a–c are predicted to be similar to 1,3-cyclohexanedione (~5), it was not too surprising that the nitro-functionalized nucleophiles exhibited similar k_obs with sulfenic acid 2. Fortunately, with predicted pK_a of ~10 and corresponding k_obs ≤ 20 min⁻¹, cyano-substituted nucleophiles 8a–h gave more promising results (Fig. S2). The above trends highlight the fine balance between keto-enol tautomerism and carbanion stability (i.e., more/less resonance forms) for linear C-nucleophile reactivity towards Cys-SOH.

In contrast to our study of cyclic C-nucleophiles where nucleophiles exhibiting > 200-fold enhancement in reactivity relative to dimedone were identified, the best C-nucleophile identified in initial screening of acyclic compounds exhibited a modest 25-fold rate enhancement compared to dimedone. Therefore, we decided to take a rational design approach in which structural elements of the most reactive linear C-nucleophiles were combined to produce “designer nucleophiles” with superior k_obs with Cys-SOH. To this end, we combined the phenylsulfonyl group of 6b (k_obs = 8.7 min⁻¹) and nitro group of 7b (k_obs = 0.5 min⁻¹) to create phenylsulfonylnitromethane (9). The reaction of 9 with sulfenic acid 2 gave a k_obs of 7.8 min⁻¹, which is similar to that of 6b. Although this initial compound did not afford better reaction kinetics, as discussed below, a nucleophile modelled on 9 did prove highly useful in the detection of protein Cys-SOH in cell lysates. In subsequent experiments, we combined the phenylsulfonyl moiety from 6b with nitrile moiety from malononitrile 8h (k_obs = 20.0 min⁻¹) to produce methyl- and phenylsulfonylacetonitrile (10a–b). Both 10a and 10b formed thioether adducts with sulfenic acid 2 with k_obs equal to 28.6 min⁻¹ and 78.0 min⁻¹, respectively (Fig. 2A). The findings were very encouraging and indicate the potential to design linear C-nucleophiles with favourable reactivity profiles.

Fig. 2.

(A) Strategy for rational design of linear C-nucleophiles exhibiting superior reactivity with Cys-SOH. (B) Structure and putative reaction sties of curcumin (11) and tofacitinib (12) towards sulfenic acid 2.

While scanning the literature for linear C-nucleophiles, we came across molecules such as the natural product, curcumin (11) and anti-rheumatoid arthritis drug, tofacitinib (12) (Fig. 2B). Curcumin is a diarylheptanoid with numerous biomedical properties.¹⁶ Of note, 11 possess an electrophilic as well as a nucleophilic reaction site. Although covalent bond formation between Cys-SH and the α,β-unsaturated ketone group of curcumin is logical and recently reported,¹⁷ we reasoned that the 1,3-dicarbonyl centre should also be able to react with Cys-SOH. Indeed, 11 successfully reacted with 2 exhibiting a k_obs of 1.9 min⁻¹. Moreover, as cyanoacetamido- based nucleophiles 8f–g reacted with sulfenic acid 2, we wondered whether 12 (containing a similar nucleophilic warhead) might also form a covalent bond with Cys-SOH. Consistent with our hypothesis, 12 effectively formed the expected thioether adduct with 2 (k_obs = 0.4 min⁻¹) (Fig. 2B).

In subsequent experiments, we examined the chemical selectivity of selected nucleophiles (3a, 4a, 6b, 8h, 9, 10b, 11, 12) in control experiments by profiling their reactivity with biofunctional groups such as lysine (amine), serine (alcohol), cysteine (thiol) and sulfinic acid (Scheme S3). Importantly, all acylic C-nucleophiles, with the exception of malononitrile 8h, gave no indication of cross-reactivity with other relevant functional groups. Next, we examined the reactivity (and selectivity) of linear C-nucleophiles (3a, 4a, 6b, 8h, 9, 10b, 11, 12) towards a protein Cys-SOH. For these studies, we used our well-established model, a Cys64Ser Cys82Ser variant of the thiol peroxidase, Gpx3.¹² Incubation of Gpx3-SOH (prepared in situ by treating 10 μM Gpx3-SH with 1.5 eq of H₂O₂) with the aforementioned linear C-nucleophiles, including curcumin 11 and tofacitinib 12, led to covalent modification of Gpx3-SOH. Control experiments clearly indicated that linear C-nucleophiles did not react with reduced Gpx3 (Scheme S4).

Given the differences among linear C-nucleophiles with respect to reactivity, pK_a, and carbanion stability it is possible that the resulting thioether adducts also display a range of stability under reducing conditions. As cell contains millimolar concentrations of glutathione (GSH) we tested the ability of this low-molecular weight thiol as well as dithiothreitol (DTT) and the phosphine-based reagent, TCEP to reduce the nascent covalent bond formed between 2 and representative members of our linear C-nucleophile library. Accordingly, thioethers 13 – 18 were prepared (Scheme S5) and examined for their stability under reducing conditions (Fig. 3 and Scheme S6). Thioether adducts 13 and 14 were unstable, ostensibly owing to carbonyl hydration, elimination of corresponding acetic acid unit, and net loss of a methylketo (–COMe) group (Scheme S7).¹⁵ More interesting, however, was the reversibility of many thioether bonds under reducing conditions to give the corresponding dipeptide-SH and linear C-nucleophile. To better characterize such reactions, pseudo 1^st order rate constants (k_obs) were determined for the reduction of all thioether adducts (Table S1). Interestingly, thioether adducts generated from our ‘designer nucleophiles’ exhibited distinct stability profiles. For example, thioether 15 displayed excellent stability under all reducing conditions, with k_obs ranging from 0.003 h⁻¹ to 0.005 h⁻¹ (t_1/2 = 6 – 10 days). By contrast, thioether 18 was the most unstable adduct with k_obs ranging from 1.9 h⁻¹ to 2.5 h⁻¹ (t_1/2 = 17 – 22 min), representing a 500-fold difference in stability (Fig. 3). All remaining thioether adducts gave stability profiles in between these two extremes (Table S1). Subsequently, we studied the stability of thioether adducts in the Gpx3-SOH model. Analogous to the above analysis with dipeptide sulfenic acid 2, the Gpx3-S-Nu adduct resulting from the reaction of Gpx3-SOH and 9 was the most stable and that of 10b was the least stable (Scheme S8).

Fig. 3.

Stability of thioethers 13 - 18 under reducing conditions.

Having characterized the proficiency and reversibility of linear C-nucleophiles with Cys-SOH in dipeptide- and protein-SOH model systems, we next investigated labelling of protein sulfenic acids in whole cell lysates. To facilitate detection, 9 and 10b were each modified to incorporate an alkyne handle (for ‘clicking’ with biotin azide and visualization by SDS-PAGE) without adversely affecting their reactivity profiles (Fig. 4A). For comparison, we chose cyclic C-nucleophile, DYn-2 (19), an alkyne-functionalized probe known for its ability to form stable thioether adducts with Cys-SOH.^2–4 Initial studies confirmed the reactivity and reversibility of NO₂-alkyne (20) and CN-alkyne (21) in the less complex Gpx3-SOH model. While all three nucleophiles formed a thioether adduct with Gpx3-SOH, the product of DYn-2 and 20 reactions were stable after GSH treatment, whereas the covalent product of 21 was completely reversible (Scheme S9). Subsequently, we probed HeLa cell lysate with DYn-2, NO₂-alkyne 20 and CN-alkyne 21 for protein S-sulfenylation (Fig. 4B). Compared to DYn-2, both NO₂-alkyne and CN-alkyne showed robust protein-SOH labelling at 100 μM probe concentration, an observation that is fully consistent with reaction rate constants obtained from sulfenic acid 2. Even more importantly, while CN-alkyne was more efficient than NO₂-alkyne at capturing the potein-SOH, labelling by 21 was totally reversible. On the other hand, protein-SOH labelling by 20 was robust and stable under reducing conditions.

Fig. 4.

Stability of protein-S-Nu thioether bonds under reducing conditions. (a) Alkyne-functionlized irreversible and reversible C-nucleophiles; (b) Gel-based reactivity profiling of irreversible and reversible C-nucleophiles in HeLa cell lysate.

We have shown that linear C-nucleophiles can be used for efficient and selective modification of Cys-SOH in peptides, proteins and lysates (k_obs = 0.05 – 129 min⁻¹). Moreover, we have studied the reversibility of the resulting thioether adducts and shown that linear C-nucleophiles, such as 21, form reversible covalent bonds. To the best of our knowledge, this work is the first to report the reversibility of such thioether adducts, a key feature that can be exploited in the development of small-molecule probes and inhibitors that target Cys-SOH within proteins. In another important aspect of this study, we have established parameters that must berigorously tested in the development of chemical probes for Cys-SOH. For example, we reveal the unstable nature of thioether adducts formed by β-ketoester-based probes ¹⁸, with direct implications for the use of these and related probes for Cys-SOH detection and proteomic analyses. Based on the reactivity and reversibility trends observed in our screen, we successfully designed two novel alkyne-functionalized linear C-nucleophile probes that react rapidly and selectively with Cys-SOH. Notably, we show that NO₂-alkyne 20 forms a stable adduct, whereas the bond formed by reaction with CN-alkyne 21 is reversible in reducing conditions.

Hitherto unknown, in the present study we also show the reactivity of the natural product, curcumin with Cys-SOH. Our observations, along with the knowledge that curcumin can also function as a Michael-acceptor with Cys-SH, highlight its dual reactivity and might account for the many intriguing biological activities reported for this natural product. Furthermore, we show that Tofacitinib, the active ingredient in Pfizer’s oral anti-inflammatory drug ‘Xeljanz’ prescribed for rheumatoid arthritis,¹⁹ reacts with Cys-SOH. Tofacitinib was initially reported as a selective inhibitor of JAK3 (Janus Kinase 3) in comparison to other kinases of this family, JAK1 and JAK2.²⁰ Sequence alignment of JAK family kinases demonstrates a high degree of homology in the active-site/inhibitor-binding pocket, suggesting that Tofacitinib should inhibit JAK1-3 with similar potency. Interestingly, a distinguishing feature of JAK3 is the presence of Cys₉₀₉ in the ‘hinge’ region, whereas JAK1/2 have a serine residue at that position.²¹ Since Cys₉₀₉ of JAK3 is located at the structural position corresponding to EGFR Cys797 ²² there is a high probability that this cysteine is redox sensitive and may become oxidized to sulfenic acid in cells. Such collective observations raise the possibility that Tofacitinib selectivity towards JAK3 relative to JAK1/2 may be due, at least in part, to covalent reversible interaction with Cys₉₀₉-SOH. More studies are required to evaluate this possibility and will be reported in due course.