Scalable Thiol Reactivity Profiling Identifies Azetidinyl Oxadiazoles as Cysteine-targeting Electrophiles

Abstract

Cysteine reactive groups are a mainstay in the design of covalent drugs and probe molecules, yet only a handful of electrophiles are routinely used to target this amino acid. Here, we report the development of scalable thiol reactivity (STRP), a method which enables the facile interrogation of large chemical libraries for intrinsic reactivity with cysteine. High throughput screening using STRP identified the azetidinyl oxadiazole as a moiety that selectively reacts with cysteine through a ring opening-based mechanism, capable of covalently engaging cysteine residues broadly across the human proteome. We show the utility of this reactive group with the discovery of an azetidinyl oxadiazole containing small molecule that augments the catalytic activity of the deubiquitinase UCHL1 in vitro and in cells by covalently modifying a cysteine distal to its enzymatic active site. This study adds a novel cysteine targeting group to the electrophilic lexicon and provides robust methodology to rapidly surveil libraries for reactivity with cysteine.

Graphical Abstract:

Introduction:

Chemical moieties that react with cysteine serve essential roles as drugs, probe molecules, cross-linkers, and conjugation agents, given the intrinsic nucleophilicity of cysteine relative to other protein coding amino acids.¹ While becoming ever more critical to contemporary pharmacology, the use of cysteine reactive groups within FDA approved drugs, for example, has been largely restricted to acrylamides with only a handful of exceptions.² When developing covalent small molecule inhibitors, high affinity non-covalent binders are typically first optimized by medicinal chemistry and then only with complex structural information in hand are a limited number of electrophilic groups appended to endow these inhibitors with covalency. In recent years, unbiased approaches using chemical proteomics have uncovered an alternative route to covalent ligand discovery by broadly screening cellular proteomes using diverse electrophilic scout fragments.^{3, 4} Such ‘reactivity first’ type approaches have both broadened the scope of the druggable proteome and inverted the order of operations by which chemical inhibitors are discovered. Collectively, these studies have suggested that an augmented repertoire of reactive groups targeting cysteine might enable greater access to less readily drugged sites in the proteome with molecules possessing altered selectivity, stability, and reactivity.

One potential source for new cysteine reactive groups might be the chemical matter lying ‘dormant’ in large chemical libraries. Indeed, several studies have serendipitously identified latent and unexpected covalent reactivities using high throughput screening and downstream target identification.^{5, 6} Despite this potential opportunity, methods to directly survey large chemical libraries (>10³ molecules) for reactivity with cysteine are lacking. Often these approaches involve LC-MS-based or colorimetric assays, which are limited in their scalability and sensitivity.^{7, 8} Previously, we used Nature’s electrophile sensor, the KEAP1-NRF2 pathway in mammalian cells, to identify 2-sulfonyl pyridines as a previously unappreciated cysteine reactive group, capable of being optimized as selective inhibitors to adenosine deaminase.⁹ While this approach allows for scalable discovery of cysteine reactive groups in the context of the living cell, it can only identify electrophiles that react with the various ‘cysteine sensor’ residues of KEAP1.¹⁰

We reasoned that a potentially superior approach to identifying new electrophilic groups might involve assaying the reactivity of a given small molecule library member with cysteine directly in solution. Accordingly, we report here the development of scalable thiol reactivity profiling, a methodology inspired by the historical synthesis of luciferin that assesses the cysteine reactivity of a small molecule in vitro. We use this method to screen >10,000 small molecules, ultimately identifying the azetidinyl oxadiazole as a novel, cysteine selective reactive group that broadly reacts with cysteines across multiple protein families in the proteome and elicits modulatory effects on modified proteins.

Results and Discussion:

Development of Scalable Thiol Reactivity Profiling.

We first sought inspiration in firefly luciferin, a molecule that has been used extensively in numerous scalable biological assays for its capacity to luminesce when oxidized in the presence of its cognate enzyme luciferase.¹¹ Emil White and colleagues reported the first chemical synthesis of luciferin and its analogs in the 1960s in which a single step reaction involving the cyclization of D- or L-cysteine with 2-cyano-6-hydroxy benzothiazole (CHBT throughout) yields the enantiomeric mixture of D/L-luciferin (Figure 1A).^{12, 13} The capacity of this reaction to spontaneously proceed in water has been exploited in recent years, as others have used solutions of CHBT and luciferase to quantify the levels of free cysteine in cells and tissues by recording the levels of produced luminescence (Figure 1B).^14–16

Figure 1. A competitive assay for detecting the reactivity of small molecules with cysteine, inspired by the historical synthesis of luciferin.

A) The first synthetic route to luciferin and its analogs described in the literature by Emil White and colleagues. B) Work from Ohimya using CHBT to detect L-cysteine levels in situ. C) Scheme depicting this work. A competitive assay based on the spontaneous formation of D-luciferin in vitro. Small molecules are first incubated with D-cysteine, either allowing D-cysteine to form D-luciferin or a non-productive covalent adduct with cysteine. Addition of luciferase allows the detection of cysteine in the case of non-reactive compounds or decreased luminescence in the case of reactive small molecules.

Given the capacity of cysteine to spontaneously form luciferin in the presence of CHBT, we reasoned that a stepwise, competitive screening assay could be generated to report on cysteine reactivity using these reagents. In this scenario, a small molecule in a screening well would first be exposed to a solution of D-cysteine for a set period, yielding either no product or a cysteine adduct, in the case of a cysteine reactive compound. Next, a solution of CHBT would be dispensed to each well, allowing the formation of D-luciferin, if no adduct had depleted the supply of reactive cysteine. Dispensing a solution of luciferase followed by luminescence recording on a plate reader would allow one to determine if the molecule within the screening well had reacted with cysteine or not, depending on the luminescence value recorded (Figure 1C).

With this framework, we optimized the Scalable Thiol Reactivity Profiling (STRP) assay for use in high throughput screening (Figure 2A). This assay first involves the 3-hour incubation of D-cysteine (2.5 µM) in 20 µL of TRIS buffer with a given small molecule (1.5–250 µM) delivered by pintool transfer (100 nL). A molecular excess of CHBT in 20 µL of buffer is then dispensed using an automated liquid handler followed by another 30-minute incubation. The addition of luciferase, ATP, and MgCl₂ followed by shaking and luminescence measurement on a standard plate reader enables the discrimination of thiol reactive compounds (low signal) from unreactive ones (high signal). Importantly, were able to determine that a broad range of D-cysteine (1–10,000 µM), CHBT (1–1000 µM), and luciferin (0.1–1,000 µM) could be accommodated in the assay conditions, ultimately yielding a dynamic range of luminescence signal of >10⁴-fold relative to background (Figure S1A–C). The optimized conditions were chosen such that the concentration of D-cysteine could be stoichiometrically competed with typical concentrations of screening compound (1–10 µM), although this data suggests that the STRP assay could be used in various settings with high signal to noise.

Figure 2. Scalable thiol reactivity profiling (STRP) reports on the cysteine reactivity of electrophilic compounds in miniaturized format.

A) Schematic depicting the steps and timing of the miniaturized Scalable Thiol Reactivity Profiling (STRP) assay. B) Structures and summary of activities of inhibitory potencies of a training set of known cysteine-reactive small molecule fragments in the STRP (C) and control (D) assays (n=3 technical replicates; mean and s.e.m.).

We additionally developed a control assay to determine if a given hit compound might be decreasing signal through mechanisms other than by reacting with cysteine. This control assay involves the initial formation of D-luciferin by combining D-cysteine with CHBT, followed compound dispensing, luciferase exposure, and then luminescence recording (Figure S1D). Compounds that decrease signal in the control assay are likely inhibitors of luciferase or are non-specific modulators of the assay (e.g., aggregators).

We next sought to understand if the STRP assay could accurately and dose dependently measure the cysteine reactivity of a ‘training set’ of twelve diverse, electrophilic small molecules that have been previously reported to react with cysteine in the literature. This compound set included methylsulfonyl benzothiazole¹⁷ (MSBT, 1; IC₅₀= 19.9 µM), 4-Chloro-7-nitrobenzofurazan¹⁸ (NBD-Cl, 2 IC₅₀ = 8.3 µM), dimethyl fumarate¹⁹ (DMF, 3; IC₅₀= 13.4 µM); monomethyl fumarate¹⁹ (MMF, 4; IC₅₀ = 15.4 µM), n-ethyl maleimide²⁰ (NEM, 5; IC₅₀ = 6.3 µM), ethyl vinyl sulfone²¹ (EVS, 6; IC₅₀ = 13.6 µM), phenylmethylsulfonyl fluoride²² (PMSF, 7; IC₅₀ = 50.5 µM), 3-(Chloromethyl)-4-ethyl-4H-1,2,4-triazole (CMET, 8; IC₅₀ = 31.4 µM), chloroacetamide²³ (CA, 9; IC₅₀ = 114.7 µM), iodoacetamide²⁴ (IA, 10; IC₅₀ = 20.7 µM), acrylamide²³ (11, IC₅₀ = 54.5 µM), and methylglyoxal⁵ (MGX, 12, IC₅₀ = 85 µM); of which all were found to inhibit STRP assay luminescence signal concentration dependently (Figure 2B,C). Notably, this assay could accurately predict which of two similar molecules with established reactivities was more reactive. For example, dimethyl fumarate was found to be more potent than monomethyl fumarate (13.4 vs. 15.4 µM respectively), and iodoacetamide was found to be considerably more potent than chloroacetamide (20.7 vs 114.7 µM, respectively). Importantly, only two of these molecules, 1 and 2, were found to show any inhibitory activity in the control assay at concentrations less than 250 µM, and the IC₅₀s (cIC50s = 129 and 211 µM respectively) were considerably higher than those of obtained with STRP assay (Figure 2D). Collectively, these data suggest that the STRP assay can report on the cysteine reactivity of diverse electrophilic chemicals across a broad range of concentrations.

High throughput screening identifies azetidinyl oxadiazoles as cysteine reactive groups.

As proof of concept that STRP can identify novel cysteine reactive groups, we next engaged in a high throughput screening campaign, surveying a commercial library of 10,561 small molecules largely devoid of common electrophiles (e.g., acrylamides, chloroacetamides, etc.) for SRTP inhibitory activity (Figure 3A, Supplementary Data Table 1). Strikingly, we found that among the top 100 scoring hits, 11 bore azetidines within their structures, prompting us to inquire whether this moiety might possess a previously unappreciated reactivity towards cysteine. We obtained fresh material for 3 related hits, compounds 13, 14, and 15, which all bear azetidinyl oxadiazoles (Figure 3B). We confirmed dose responsive inhibitory activity for these compounds (13, IC₅₀ = 113; 14, IC₅₀ = 102; 15, IC₅₀ =45), all of which do not display inhibitory activity in the control STRP assay (IC₅₀ > 250 µM; Figure 3C, D).

Figure 3. High throughput screening with STRP identifies azetidines as cysteine selective reactive groups.

A) Z scores of luminescence signal from a high throughput screen of 10,561 diverse small molecules assayed for STRP inhibitory activity with azetidine containing hit compounds colored in peach. Structures and summary of activities (B), dose responsive inhibitory activity in the STRP (C) and control (D) assays of three commercially available azetidine containing hit compounds (13, 14, 15; n=3 technical replicates; mean and s.e.m.). E) Scheme depicting the steps to form the ring opened, Boc protected cysteine adduct of 15 in vitro. F) ¹H NMR spectra of 15 (bottom, gray) and the ring opened cysteine adduct of 15 (top, green). Protons noted on structures to the right as colored circles correspond to circles noted on spectra (left). G) Extracted ion chromatograms (EICs) for the adducted, ring opened species with each of the protected nucleophilic amino acids. Structures of each predicted adduct are depicted in Figure S2.

We next sought to understand the mechanism by which the azetidnyl oxadiazole might react with cysteine to form a covalent adduct. From model in vitro reactions with 15 and N-Acetyl-L-cysteine methyl ester (Ac-Cys-OMe) in phosphate buffer (pH 8), we found strong evidence for the formation of opening of the azetidine ring via cysteine attack at C2. We were able to trap the ring opened product of 15 with Ac-Cys-OMe via Boc protection (Figure 3E), enabling larger scale reactions, purification, and confirmation of the structure of the purified product by NMR (Figure 3F and Supporting Information). We additionally confirmed that among putatively nucleophilic amino acids besides cysteine (lysine, serine, threonine, histidine, and tyrosine), only cysteine showed robust evidence of product formation from in vitro reactions with these protected amino acids with 15 (Figure 3G, Figure S2).

We then sought to understand the stability of the formed adduct of 15 with Ac-Cys-OMe. We were able to isolate the ring opened adduct of 15 and Ac-Cys-OMe without the need for Boc protection, consistent with the idea that the adduct is not immediately reversibly covalent (Figure S3A). We then monitored its stability over the course of 24 hours at 37 °C in phosphate buffered saline. The half-life was found to be 9.56 hours, as determined measuring the integrated peaks derived from UPLC-UV analysis (Figure 3B, C). We take this data to indicate that the azetidinyl oxadiazole is not strictly a reversibly covalent molecule but results in an adduct which displays some level of intrinsic instability in solution.

We next performed a structure activity relationship study to determine, what, if any, other azetidine containing molecules might also possess reactivity to cysteine. From a library of 24 additional commercially available C3 substituted azetidine fragments, we found that only 3 retained cysteine reactivity (IC₅₀ <250 µM), as determined by inhibitory activity in the STRP assay (Figure S4). Notably, 39 also contains an oxadiazole like 13, 14, and 15 and exhibited similar potency in the STRP assay (IC₅₀ = 102 µM; Figure S4). Methoxyphenyl substituted 17 and pyrrolidine substituted 35 also displayed inhibitory activity in this assay (Figure S4).

To gain a potential mechanistic explanation for the increased reactivity of oxadiazole substituted azetidines, we performed density functional theory (DFT)-based calculations comparing 15 to 19, an imidazole substituted azetidine with no SRTP inhibitory activity. We propose this reaction proceeds through thiolate attack at the carbon of position 2 of the azetidine ring through a concerted S_N2-like reaction (Figure S5). The transition state of 19 with ethanethiol was found to be of a lower energetic barrier (27.72 kcal/mol) relative to that of 15 (32.41 kcal/mol). Decreased energy required to achieve the transition state was attributed to the potential of the azetidine of 19 to make favorable H-bonding interactions with the oxadiazole nitrogen, as well as the natural bond orbital (NBO) analysis of the oxadiazole 15 suggested a relatively favorable NBO charge on the azetidine carbon (−0.15 for 15 vs. −0.171 for 19).

Azetidnyl oxadiazoles broadly engage cysteines throughout the proteome.

We next sought to understand if the azetidinyl oxadiazole group could covalently ligand cysteines across the proteome in live cells. Accordingly, we synthesized 41, a derivative of 16 bearing an alkyne moiety for use in affinity tagging and enrichment studies (Figure 4A). 41 retained similar inhibitory potency in the STRP assay (34 µM vs. 45 µM for 16) and did not interfere in the control STRP assay (cIC > 250 µM; Figure 4B). Notably, we found that exposure of HEK293T cells to increasing concentrations of 41 (1–1000 µM) for one hour resulted in the dose dependent labeling of many observable bands across a broad molecular weight range (Figure S6A, B) after lysates were subjected to click reaction-based conjugation with rhodamine azide. Additionally, we found that the vast majority of rhodamine labeled bands could be competed away when iodoacetamide (10 mM) was administered to cells 1 hour before exposure to 41 (Figure 4C, Figure S6C), indicative that 41 likely only labels cysteine residues in live cells.

Figure 4. Alkyne probe 41 covalently engages diverse protein families across the cysteine proteome.

A) Structure and summary of activities of 41. B) STRP assay and control assay of the indicated concentrations of alkyne probe 41 (n=3 technical replicates; mean and s.e.m.).. C) Rhodamine based fluorescence scan of SDS-PAGE resolved proteomes from HEK293T cells exposed to the indicated concentrations of 41 for 1 hour with or without pre-treatment with iodoacetamide (IA, 10 mM) for 1 hour. D) Fold enrichment vs −log₁₀P values for covalently modified proteins identified by MS/MS-based proteomics after streptavidin enrichment from HEK293T cells exposed to 41 (1 mM) for 1 hour (n=3 biological replicates) E) Fraction of identified targets that have been liganded according to DrugBank. F) Fraction of identified targets associated with the indicated protein functions. G) −log₁₀P values of the top GO term associations for statistically enriched proteins. H) Structures of scout fragments KB02 and MMNP. I, J) Scatter plot of isoTOP-ABPP liganding ratios of cysteine containing peptides from HEK293T cells engaged by treatment with 41 (1 mM), KB02 (500 µM), and MMNO (500 µM) for 1 hour. Inset Venn diagram depicts overlap of liganded peptides with KB02 or MMNP selective peptides in peach, 41 selective peptides in blue, and shared cysteine containing peptides in green (n=3 biological replicates per condition).

We next performed chemical proteomic enrichment studies to identify the proteins modified by 41. Here, HEK293T cells were exposed to 41 (1 mM) for one hour and lysates subjected to click chemistry reactions to affix biotin azide to labeled proteins. Streptavidin enrichment in denaturing conditions coupled to shotgun AP-MS/MS proteomic analysis identified 546 proteins that were statistically enriched in these conditions (Figure 2D). Most of these protein targets have not been drugged (81%), as determined by DrugBank annotations (Figure 4E).²⁵ Likewise, these liganded proteins correspond to a broad array of protein functional classes including transporters, enzymes, chaperones, and others (Figure 4F). These protein targets conformed to several protein classes as assessed by GO term enrichment, including RNA binding proteins, ubiquitin conjugation machinery, and isopeptide hydrolases (Figure 4G). These classes were represented among the most enriched proteins, which included E3 ligases, cysteine protein hydrolases, and RNA binders (Figure 4D). We chose to focus our attention on two highly enriched E3 ligases, UBR7 (ubiquitin protein ligase E3 component n-recognin 7) and UBE2O (ubiquitin conjugating enzyme E2 O), and two highly enriched deubiquitinases, USP11 (ubiquitin specific peptidase 11) and UCHL1 (ubiquitin c-terminal hydrolase L1). From experiments in which these four proteins were overexpressed as FLAG-tagged transgenes in HEK293T cells, we found that a modest concentration of 41 (20 µM) resulted in the robust labeling of all four transgenes after one hour of exposure, as determined by rhodamine positivity of anti-FLAG immunoprecipitated material (Figure S6D), confirming the accuracy of our AP-MS chemoproteomic profiling. Collectively, these data suggest that 41 and likely the azetidinyl oxadiazole can covalently label a substantial portion of the cysteine proteome, targeting many proteins that have yet to be drugged with small molecule ligands.

We further sought to understand if the azetidinyl oxadiazole might covalently engage a distinct set of cysteines across the proteome relative to other cysteine reactive groups. Accordingly, we performed a modified isoTOP-ABPP experiment, a quantitative LC-MS/MS based technique for measuring occupancy of cysteines across the proteome (Figure S7A). Treating HEK293T cells with 41 (1 mM), KB02 (a chloroacetamide that has been frequently used to map ligandability proteome wide, 500 µM), and MMNP (4,6-Dimethyl-2-(methylsulfinyl)-3-nitropyridine, an SNAr-based warhead previously shown to access a divergent swath of the cysteine proteome, 500 µM), we found that we could accurately map liganding of 14,834 cysteine containing peptides proteome wide (Figure 4H). 41 liganded considerably fewer peptides than the other scout fragments, as 235 were liganded as compared to the 1707 of KB02 and the 1669 of MMNP (Figure S7B). Of these, 63 peptides were shared between 41 and KB02 and 62 peptides were shared between 41 and MMNP (Figure 4I, J). These data suggest that the azetidnyl oxadiazole indeed engages a largely unique set of cysteines in the proteome with considerably detuned reactivity relative to other electrophilic reactive groups.

STRP accurately predicts the reactivity of azetidinyl oxadiazoles and other cysteine reactive electrophiles.

With the full repertoire of used probes in hand, we next sought to confirm that the decreased observed reactivity of the azetidinyl oxadiazole in the STRP and cellular labeling assays was consistent with an overall decrease in reactivity with cysteine in vitro. Accordingly, we established UPLC-UV-based methodology to assess reactivity of a given electrophile with protected cysteine (Ac-Cys-OMe) in vitro (see Methods). We compared azetidinyl oxadiazoles 15 and 41 to compounds 1 and 2, the only two compounds from our training set that retained good stability and had good chromophores for tracking reaction rate. As anticipated these two compounds showed considerably longer half-lives in the presence of Ac-Cys-OMe (346 minutes for 15, 9.52 minutes for 41) than 1 (0.81 minutes) or 2 (0.002 minutes). Importantly these half-lives correlate well with IC₅₀ values obtained from the STRP assay (R² > 0.9; Figure S8). These data indeed confirm that the azetidinyl group is indeed less reactive than other chemotypes and that the STRP assay can accurately measure compound reactivity.

Discovery of an activator of UCHL1.

Lastly, we sought to demonstrate the utility of the identified reactive group by characterizing the mechanism by which an azetidinyl oxadiazole containing compound might modulate the function of a covalently modified protein. We therefore focused our efforts on the interaction of 41 with UCHL1, a thiol peptidase that cleaves C-terminally conjugated ubiquitin molecules from client proteins central to the maintenance of neuronal cellular populations, where it is most highly expressed.²⁶ The nucleophilicity of the active site cysteine has enabled the discovery of several covalent inhibitors of the catalytic activity of UCHL1 in the literature.^{27, 28}

We first confirmed 41 covalently labels UCHL1 in a dose dependent manner from experiments in which FLAG tagged transgene was overexpressed in HEK293T cells (Figure S9A). To our surprise, 41 was not found to label the nucleophilic active site cysteine (C90), but instead labeled C152, a residue distal to the active site. Experiments in which C90 or C125 of the FLAG tagged transgene were mutated to serine or alanine confirmed that only labeling was abrogated in the context of C152 mutation (Figure 5A, Figure S9B). We additionally confirmed the nature of this labeling event, as we could detect the ring opened tryptic peptide fragment adduct containing C152 by MS/MS (Figure 5B, Figure S9C) from cells overexpressing FLAG tagged UCHL1 treated with 41 (20 µM). We lastly traced the kinetics of labeling of recombinant protein in vitro, with 41 labeling recombinant UCHL1 fully within 3 hours at room temperature (Figure S9D).

Figure 5. Covalent modification of C152 by 41 allosterically augments UCHL1 activity.

A) Rhodamine fluorescence scan and Coomassie staining of anti-FLAG immunoprecipitated content of the indicated FLAG tagged UCHL1 transgenes overexpressed in HEK293T cells and exposed to 41 (20 µM) for 1 hour. B) MS/MS spectra of UCHL1 peptide containing modified C152 from anti-FLAG immunoprecipitated material from HEK293T cells expressing UCHL1-FLAG and exposed to 41 (20 µM) for 1 hour. C) Image of apo crystal structure of UCHL1 (PDB: 2ETL) indicating the catalytic triad H161, D176, and C90 along with C152 (green) within the flexible crossover active site loop (gray). D) Activity of recombinant UCHL1 exposed to the indicated concentrations of 41 for one hour. E) Activity of anti-FLAG immunoprecipitated UCHL1 from HEK293T cells exposed to 41 (10 µM) for 1 hour. Anti-HA Western blotting for ubiquitinated anti-FLAG immunoprecipitated SMAD2 (F) and ALK5 (G) after exposure of HEK293T cells expressing UCHL1 to 41 (100 µM) for 1 hour. H) Relative enzymatic activity of recombinant WT or C152S UCHL1 in the presence or absence of active site inhibitor IA (5 mM; n=3). I) Anti-HA Western blotting for ubiquitinated anti-FLAG immunoprecipitated SMAD2 in cells expressing WT or C152S UCHL1.

UCHL1 bears one of the most complex protein folds yet discovered for a eukaryotic protein with five internal backbone crossings forming a Gordian knot.²⁶ In addition, UCHL1 possesses a ‘crossover loop’ which, upon its movement, enables access to the active site for protein substrates (Figure 5C). C152 occupies a site near the middle of the active site crossover loop, suggesting that its modification might modulate the catalytic activity of the protein. Indeed, previous work has shown that C152 can be modified by the endogenous lipophilic electrophilic compound, 5-deoxy-Δ12,14-prostaglandin J2 (15dPGJ2), a modification which, in neuronal cells, promotes aggregation of the protein and cellular death.²⁹ Accordingly, we sought to understand how modification of C152 by 41 might also modulate the enzymatic activity of UCHL1. We first established an enzymatic assay measuring the enzymatic ability of UCHL1 to cleave a fluorogenic ubiquitin substrate, and confirmed the assay was highly sensitive to known inhibitors, including 6RK71, iodoacetamide (IA), and IMP-1710 (Figure S9E).

To our surprise, we found that treating recombinant preparations of UCHL1 with increasing concentrations of 41 increased the rate of reaction by nearly two-fold (EC₅₀ = 6.6 µM; Figure 5D). We additionally found that FLAG tagged protein isolated from HEK293T cells labeled with 41 (10 µM) in situ additionally displayed increased rates of reaction, albeit to a lesser degree (Figure 5E). UCHL1 has been shown, at least in part, to promote cellular survival by deubiquitinating components of the TGF-beta signaling pathway, SMAD2 and ALK5, resulting in augmented signaling through this pathway.³⁰ We found that treatment with 41 in HEK293T cells increased the capacity of myc tagged UCHL1 to further deubiquitinate both FLAG tagged SMAD2 and ALK5 as assessed by anti-HA (measuring ubiquitination) Western blotting of FLAG immunoprecipitated material (Figure 5F, G). These data indicate that covalent modification of C152 by 41 functionally increases the isopeptidase activity of UCHL1 in vitro and in cells.

Lastly, we sought to understand if we might recapitulate the effect of modifying C152 with 41 using genetic means. Interestingly, the simple C152S substitution was found to increase enzymatic activity relative to the WT enzyme in an in vitro assay using recombinant preparations of protein (Figure 5H), the enzymatic activity of which could be fully inhibited by treatment with active site modifier IA. We additionally demonstrated that introduction of myc tagged UCHL1 bearing the C152S transgene could more readily deubiquitinate FLAG SMAD2 in HEK293T cells (Figure 5I), largely mimicking the activity of covalent modification by 41. These data suggest that C152 may be a highly sensitive switch, regulating activity of UCHL1. However, how modification by 41 or by genetic means can result in increased activity of this enzyme will necessarily be the work of complex future mechanistic studies.

Conclusions:

Here, we have described a simple and scalable assay for measuring the intrinsic reactivity of small molecules with cysteine. The STRP assay enables the unbiased evaluation of cysteine reactivity in solution, as STRP accurately reports on the covalent reactivity of known cysteine reactive electrophiles across several reaction types (S_NAr, Michael addition, etc.). Additionally, the STRP assay can produce measurable luminescence signal across several orders of magnitude of reactant concentrations, suggesting that even weakly cysteine reactive molecules might be able to be identified in an appropriately tuned assay condition. The work described here was performed in 384-well format using typical equipment found in an academic laboratory; however, given the robustness of this assay, it is highly likely that it could be further miniaturized (e.g., 1536-well format) to screen larger libraries (>10⁶ compounds) with automated screening equipment.

From screening a moderately sized library of ~10,000 diverse small molecules for STRP inhibitory activity, we identified that azetidines, namely azetidinyl oxadiazoles, possess previously unreported reactivity with cysteine. This ring opening reaction proceeds through a S_N2-like mechanism, one we believe to be accelerated by intramolecular interactions between the azetidine nitrogen and the oxadiazole. Interestingly, azetidines have been lauded in the medicinal chemistry literature for their physical properties, namely their stability in biological systems and their ability to induce rigidity into small molecules without adding significant molecular weight.³² In contrast, this work suggests that certain azetidines might pose a potential cysteine reactivity risk and should be profiled appropriately before their introduction into a medicinal chemistry campaign. While we have shown that oxadiazole substituted azetidines possess increased reactivity relative to other molecules evaluated here, there are likely an expanded repertoire of azetidines that possess good reactivity with cysteine. Indeed, the SAR campaign in this work was more limited, and future mechanistic studies will be required to determine the scope of reactive azetidines that might be harnessed for use as cysteine targeting reactive groups.

We performed chemical proteomics with an alkyne derivatized azetidinyl oxadiazole, 41, discovering that hundreds of proteins in live cells across various protein classes could be covalently modified by this chemotype. This result indicates that this novel reactive group may provide an unprecedented capacity to target cysteines in certain protein classes or domains, as most proteins identified here have not been drugged before. Likewise, we have shown that 41 ligands a largely orthogonal set of cysteines compared to standard scout fragments. We demonstrated the utility of this reactive group by studying the effect of 41 on modifying UCHL1, a deubiquitinase that is essential to neuronal survival. We found that 41 does not modify the presumably more nucleophilic active site cysteine, C90, but instead modifies C152, a residue on a flexible loop covering the active site, movement of which is required for access of ubiquitin substrates.^{26 15, 33, 34}. One such mutation, E7A, results in early onset neurodegeneration by decreasing catalytic activity of UCHL1.³³ This glutamate makes key H-bonding contacts with residues Q151 and R153 (on either side of C152) to stabilize the crossover active site loop when bound to ubiquitin. As noted, C152 has also been the site of covalent modification by endogenous electrophilic chemicals and has been reported to be susceptible to s-nitrosylation.³⁵ Given these observations, it is conceivable that C152 may act as a sensor switch to control the level of activity of UCHL1 and that modification of C152 by 41 may promote a more constitutively active state of the protein. Clearly, future work will be required to fully delineate the role of C152 in the regulating the catalysis of UCHL1. Nevertheless, 41 will likely serve as a useful mechanistic tool to aid in these efforts. Likewise, it will be interesting to understand if molecules like 41 can be optimized by medicinal chemistry to increase UCHL1 activity in the context of neurodegenerative disease. Projecting forward, we anticipate the azetidinyl oxadiazole, and, more broadly, the azetidine, will provide a novel chemotype from which to design new covalent inhibitors and to target the cysteine proteome more effectively.