Accelerating multiplexed profiling of protein-ligand interactions: high-throughput plate-based reactive cysteine profiling with minimal input

Summary

Chemoproteomics has made significant progress in investigating small molecule-protein interactions. However, the proteome-wide profiling of cysteine ligandability remains challenging to adapt for high-throughput applications, primarily due to a lack of platforms capable of achieving the desired depth using low input in 96- or 384-well plates. Here we introduce a revamped, plate-based platform which enables routine interrogation of either ~18,000 or ~24,000 reactive cysteines based on starting amounts of 10 or 20 μg, respectively. This represents a 5–10X reduction in input and 2–3X improved coverage. We applied the platform to screen 192 electrophiles in the native HEK293T proteome, mapping the ligandability of 38,450 reactive cysteines from 8,274 human proteins. We further applied the platform to characterize new cellular targets of established drugs, uncovering that ARS-1620, a KRAS^G12C inhibitor, binds to and inhibits an off-target adenosine kinase ADK. The platform represents a major step forward to high-throughput proteome-wide evaluation of reactive cysteines.

eTOC blurb

Yang et al. introduce a sample multiplexing-based high-throughput platform, TMT-ABPP, for proteome-wide reactive cysteine profiling, which offers a 2–3X improvement in cysteinome coverage while reducing input material by 5–10X compared to existing strategies. TMT-ABPP facilitates in-depth screening of cysteine ligandability using electrophile libraries and reveals off-targets of established compounds.

Graphical Abstract

Introduction

Proteins serve as primary targets for drug development to treat diseases. However, more than 80% of disease-relevant proteins pose a significant challenge in conventional drug development due to their lack of enzymatic domains and effective small-molecule modulators¹. Covalent inhibitors engaging nucleophilic cysteine residues have emerged as a viable modality to address the challenge, as exemplified by several successful drug development campaigns¹. In addition to its prevalence within active sites in diverse families of enzymes, cysteine plays a critical role in protein folding, enzymatic function, and response to reactive oxygen species (ROS)². As one of the most nucleophilic amino acids at physiological pH, cysteine has gained considerable attention in fragment-based drug discovery and chemical biology research^3–8.

Mass spectrometry-based chemoproteomics investigates protein-ligand interactions and has become an essential tool in profiling cysteine reactivity^3,6,8. Pioneered by Cravatt and coworkers, isoTOP-ABPP takes advantage of isotopically labeled cleavable tags and click chemistry to enable the assessment of cysteine reactivity and ligandability in various cellular states and electrophile perturbations^3,8. Realizing that isoTOP-ABPP requires large amount of starting material (500–1000 μg per electrophile^3,8) and can be improved further through incorporating tandem mass tag (TMT)-based higher sample multiplexing led to the development of the streamlined cysteine activity-based protein profiling (SLC-ABPP) to enable simultaneous measurement of 11 samples using substantially reduced sample input (50–100 μg per electrophile)⁶. In addition, a new desthiobiotin iodoacetamide (DBIA) probe was synthesized and used in the SLC-ABPP to label and enrich reactive cysteines in a sample⁶. Despite facilitating a number of chemoproteomic discoveries^9,10, SLC-ABPP is still not easily applied to large-scale library screening due to a number of challenges: 1) large material input prevents plate-based high-throughput workflows, 2) the coverage (~8,000 cysteines) is suboptimal compared to the number of detectable reactive cysteines from large scale studies within a single cell line^6,11, and 3) labeling 50–100 μg of protein lysates with TMT reagents for thousands of electrophiles can be cost-prohibitive.

Since SLC-ABPP, a number of technical innovations have been made, which we reason would greatly benefit TMT-based ABPP workflows. Our lab has pioneered the use of 18-plex TMT reagents to further increase the multiplexing capacity from 11 to 18^12,13. We have also developed a 96-well plate-based single-pot solid-phase-enhanced sample preparation (SP3)-TMT workflow for global proteomics¹⁴ using carboxylate-coated magnetic beads¹⁵. Notably, these beads have recently been demonstrated by Backus and colleagues to improve the recovery of biotinylated peptides and enhance chemoproteomics coverage when coupled with on-line high-field asymmetric waveform ion mobility spectrometry (FAIMS) separation^11,16. Inspired by these advances, we set out to revamp the entire sample multiplexing-based chemoproteomics platform, with the overarching goal of fully enabling deep high throughput profiling of reactive cysteines on small amount of material. Here, we present a new TMT-ABPP platform for plate-based, high-throughput proteome-wide profiling of reactive cysteines using as little as 10–20 μg native proteome input. In addition to reducing the input by 5–10X compared to the state-of-the-art SLC-ABPP⁶, the new platform routinely interrogates ~18,000 or ~24,000 reactive cysteines based on starting amounts of 10 or 20 μg, respectively, representing a 2–3X improvement in depth over SLC-ABPP. By upgrading from the 11-plex to the 18-plex TMT reagents, additional enhancements in throughput and reduction in variability are achieved. We demonstrate the TMT-ABPP platform by screening a library of 192 electrophiles, creating the largest coverage of reactive cysteinome within a single cell line. We further showcase that this TMT-ABPP platform can effectively reveal unknown cellular targets of well-studied compounds (e.g., ARS-1620, Sulfopin) and FDA-approved drugs (e.g., Ibrutinib, DMF) in three different cell lines. The new TMT-ABPP platform is fully geared toward screening libraries comprising thousands of electrophiles and deep profiling of the target landscape of covalent drugs.

Results

Next-generation TMT-ABPP platform

To achieve proteome-wide profiling of reactive cysteines using minimal input in 96-well plates, we set out to re-evaluate each step of the SLC-ABPP. Several major changes were made (Figure 1A), including 1) sample multiplexing through the use of 18-plex TMTpro reagents¹³, 2) incorporation of a modified magnetic bead-based SP3-TMT workflow^14,17–19 to accommodate all steps in 96-well plates, 3) the use of a streptavidin-agarose resin to capture reactive cysteines with 10X increased capacity relative to the previous workflow⁶ (Figure S2A, B), and 4) incorporation of online FAIMS gas-phase fractionation^11,20 (Figure S2C–E). Briefly, 10–20 μg of native cell lysates are treated with electrophiles bearing cysteine-reactive warheads in 96-well plates. The desthiobiotin iodoacetamide (DBIA) probe is then added to cap the remaining unliganded reactive cysteines and create an enrichable handle. Carboxylate-coated magnetic SP3 beads are adopted to remove excess reagents²¹. All subsequent steps, including protein alkylation, digestion, and TMT labeling are performed directly on the SP3 beads in plates, streamlining the entire process and obviating the need for tube-to-tube sample transfer. The TMT-labeled peptides from 18 samples are then combined into a single tube and subject to enrichment using the streptavidin-agarose resin. The enriched peptides are analyzed by LC-MS coupled with FAIMS gas-phase fractionation. With 10 μg input proteome per sample, we are able to extend the depth to ~18,000 cysteine sites within a single 3-hr analysis (Figure 1C). This is equivalent to 10 min. of analysis per sample or 11.2 min. per compound when two DMSO control samples are included (Figure 1B). If deeper coverage is desired, 20 μg input can be used to obtain more than 24,000 cysteine sites (Figure 1C). Quantitative results are calculated as competition ratios (CRs) between the DMSO control over compound treatment (Figure 1D). The introduction of the new TMT-ABPP platform yields several significant advantages: a 5–10X reduction in input requirement, 2–3X deeper coverage, and a greatly simplified workflow with all procedures conducted directly in plates.

Figure 1. TMT-ABPP for high-throughput sample multiplexing-based cysteinome profiling in 96 well plates.

A) Optimized workflow for TMT-ABPP integrating high-throughput sample preparation with TMT18plex-based sample multiplexing in 96-well plates. Steps performed in 96-well plates include treatment of lysate with compounds (green star), treatment with DBIA probe (red hexagon), trypsin digestion, and TMT labeling. After combing TMT-labeled samples into 18-plexes, biotinylated cysteine-containing peptides are enriched using streptavidin-agarose followed by mass spectrometry analysis. Competition ratios (CR) between DMSO control and compound treatment samples are measured for each cysteine.

B) TMT-ABPP significantly reduces instrument time for large-scale compound profiling (~11 min per compound).

C) TMT-ABPP achieves routine assessment of more than 18K or 24K reactive cysteines using as little as 10 or 20 μg of soluble cell lysate.

D) TMT-ABPP measures the competition ratio between vehicle and electrophile-treated samples to identify cystine-ligand interaction.

Figure 4. Electrophiles and protein structural environments.

A) Prediction-aware part-sphere exposure (pPSE) of all the cysteines with confident AlphaFold2 prediction quality (AlphaFold2 pLDDT>70³⁸) in the human proteome, reactive cysteines identified in the screening, and ligandable cysteines (CR>2). The pPSE value for cysteines is calculated by White et al.³⁷ using StructureMap as described in Bludau et al.³⁶.

B) Total and ligandable cysteines in representative protein domains.

C) CR values of cysteines treated with AC19 in EGF-domains are higher on average than cysteines not in the domain, suggesting AC19’s preference to cysteines in EGF-like domains.

D) AC19 and four analogs were used to study structure-activity relationship (SAR). The heatmap represents pairwise maximum common substructure-based Tanimoto coefficients⁴³.

E) CR values of cysteines in EGF-like domains treated with 5 structural analogs. These cysteines are almost exclusively ligandable by AC19, suggesting the azepanyl group plays an important role in SAR.

F) Examples of cysteines in the EGF-like domain that are only ligandable by AC19, but not any analogs. Bars represent mean CR ± s.d. (n=3).

Benchmarking TMT-ABPP using scout fragments

To benchmark the new TMT-ABPP platform, we treated 10 μg HEK293T cell lysates for 1 hr with varying concentrations (0 to 200 μM) of three well-studied highly reactive scout fragments³, KB02, KB03, and KB05 (Figure 2A). We generated technical duplicate samples to confirm their reproducibility. We profiled over 18,000 cysteines from ~6,000 proteins across 18 samples in a single-shot 3-hour LC-MS analysis (Figure 2B). In total, we profiled 22,483 cysteines from two replicates with 15,147 (67%) in common (Figure 2C), belonging to 6,411 proteins (Figure 2C). This represents a more than 2X deeper coverage compared to the largest dataset describing ligandability of these three fragments, at the same time a 10X reduction of input (10 μg vs. 100 μg)⁶. We observed dose-dependent CR values for thousands of cysteines treated with the three fragments (Figure 2E; Figure S1A), corroborating their broad reactivity^3,6. We define cysteines presenting a CR value>2 at 200 μM as ligandable, and as a result, 6,813 cysteines were liganded by at least one fragment (Figure S1C). We note that we used 200 μM of the fragment here instead of 500 μM as in previous reports^3,6, yet still generated more ligandable cysteines with CRs >2, highlighting the advantage of being able to assess 2X more sites. The quantification was highly reproducible between two replicate dose-response curves for ligandable cysteines, recording a median Pearson correlation (R) of >0.96 (Figure S1B). Within the fraction (1054; 16.5%) annotated as DrugBank²² targets (Figure 2D), 505 proteins possessed a cysteine that was liganded by the fragments (Figure 2D). The >2X improvement in depth allowed for an examination of ligand selectivity within the same protein. In addition to recapitulating that choloroacetamide-bearing KB02 and KB03 have higher potency than the acrylamide KB05 to the intrinsically hyperreactive C32⁸ of the glutathione S-transferase GSTO1³, we identified two additional reactive cysteines, C192 and C237, on GSTO1 and observed minimal CR changes with all three fragments (Figure 2F; Table S1). We also observed ligandable cysteines that had never been described previously (Figure 2G). These reactive cysteines present distinct ligandability profiles with the three fragments, indicating contributions from both the warhead and binding groups in conferring the selectivity. For example, KB02 is the most potent binder of C217 of the deubiquitinase UCHL5 whereas C3162 of COL6A3 favors KB05 and C30 of ATP5MK favors KB03 (Figure 2G). Strikingly, CysDB, a curated community-wide repository of human cysteine chemoproteomics data derived from nine high-coverage studies, has no cysteine reactivity data on COL6A3 and ATP5MK, indicating that they were never successfully profiled, likely due to limited depth²³. The ability to access 2X more cysteines and identify new ligandable sites and proteins while using only minimal input highlights the robustness and sensitivity of the TMT-ABPP platform.

Figure 2. Benchmarking low input TMT-ABPP with scout fragments.

A) Scout fragments KB02, KB03, and KB05 were used to profile cysteine ligandability in native HEK293T lysate. 10 μg of cell lysate was treated with each compound with concentration ranging from 10 μM to 200 μM for 1 hour and then processed using TMT-ABPP. Two replicates (2 ×18-plexes) were generated and analyzed.

B) Number of cysteines and proteins quantified in each replicate.

C) Overlap of cysteines and proteins between two replicates.

D) Among 6,393 proteins profiled with TMT-ABPP, 505 (7.9%) are considered as drug targets by DrugBank and have cysteines ligandable (CR>2) by one of the scout fragments, while the other 547 (8.6%) DrugBank target proteins have reactive cysteines, but they are not ligandable by scout fragments.

E) Heatmap of competition ratios (CRs) across all cysteines quantified in two replicates (n=22,492).

F) Scouts show selectivity to different cysteines on GSTO1. Points represent mean CRs from duplicate measurements.

G) Examples of newly identified and ligandable cysteines. These three cysteines were not identified in previous datasets profiling the fragments^23,59,60. COL6A3 and ATP5MK have not been identified as containing reactive cysteines previously²³. Fragments have different binding potencies and selectivities to the three cysteines. Points represent mean CRs of duplicate measurements.

Profiling electrophile fragment library by TMT-ABPP

Encouraged by the prospect of conducting efficient and cost-effective library screening in 96-well plates, we decided to employ TMT-ABPP to screen a library of 192 electrophiles encompassing 128 chloroacetamides and 64 acrylamides (Figure 3A). Here we set out to explore a real-world dataset created from low input starting amounts (20 μg proteome per sample). With 20 μg, two LC-MS analyses instead of a single shot were possible to boost cysteinome coverage. Inspired by the orthogonal separation provided by different compensation voltage (CV) settings on the FAIMS device^11,20, we included the orthogonality of different CV values and noted that two LC-MS analyses using 6 different CVs, grouped into two 3-CV sets (−35, −45, and −60; −30, −55, and −70) for two runs respectively, generated the best coverage (Figure S2C–E).

Figure 3. Screening of a library of 192 electrophiles in 96-well plates.

A) 192 electrophiles (128 chloroacetamides and 64 acrylamides) were screened using HEK293T cell lysate (20 μg lysate per compound) in 96-well plates. 12 × 18-plexes were generated.

B) On average, 24,205 reactive cysteines in 6,871 proteins were profiled in each 18-plex, resulting in a total of 5,228,402 measurements in 8,274 proteins.

C) Data completeness across 12 × 18-plexes. In total, 38,450 cysteines were quantified in at least one 18-plex, and 13,354 were quantified without any missing values.

D) Data completeness is correlated with protein copy number in HEK293T. Cysteines with higher data completeness are primarily on proteins with higher copy numbers, whereas cysteines showing lower data completeness are generally on low abundance proteins. HEK293T protein copy numbers were obtained from the OpenCell database²⁴.

E) Many enzymatic active cysteines can be liganded by covalent electrophiles. These enzymes belong to various families including glutathione transferase (e.g., GSTO1), dehydrogenase (e.g., ALDH1B1), E3 ubiquitin ligases (e.g., HERC2), and hydrolyses (e.g., BLMH).

F) Examples of ligandable enzymatic cysteines in HERC2, BLMH, ALDH6A1, and ALDH1B1. Color corresponds to competition ratio.

G) Three cyclic sulfone-containing chloroacetamides (CL41, CL71, and CL74) bind to C57 and C113 in PIN1.

H) Dose response of PIN1 C57 and C113 when treated with various concentrations of CL41, CL71, CL74, and Sulfopin. C113 is favored by these compounds and shows higher CR values.

To analyze the 192-compound library in HEK293T cell lysates required 12, 18-plex experiments with each compound at 50 μM. Per 18-plex, on average 24,205 reactive cysteines in 6,871 proteins were measured (Figure 3B; Table S2). In total, we profiled drug-cysteine interactions for 38,450 cysteines, with 23,108 having measurements in >50% of compound treatments and 13,354 with 100% data completeness (Figure 3C). This represents the single largest dataset in terms of number of reactive cysteines to date²³. We note this dataset was generated using a single cell line, whereas previous large cysteine datasets were usually generated using multiple distinct cell lines to improve coverage^6,11,23. As expected, the data completeness of each cysteine was, to some extent, correlated with its originating protein abundance (Figure 3D) when we referenced the OpenCell database which contains protein copy number information for HEK293T²⁴.

In the human proteome, many enzymes (e.g., deubiquitinase, ubiquitin ligase, phosphatase, oxidoreductase) utilize the thiol group on cysteines as active sites to catalyze biochemical processes^25–29. Covalent modification with electrophiles should ablate their enzymatic functions and provide a perturbation system to obtain functional insight²⁹. We examined 198 active site cysteines from a diverse range of enzymes (Figure 3E). Despite their intrinsic reactivity as active sites, these cysteines exhibited varying degrees of ligandability. Two of the most reactive active-site cysteines in the dataset belong to two members of the aldehyde dehydrogenase (ALDHs) family, C317 of ALDH6A1 and C319 of ALDH1B1. All ALDH family members preserve a conserved cysteine residue for their catalytic activity, and they are involved in a spectrum of diseases, such as atherosclerosis³⁰ and pancreatic³¹, colorectal³² and hepatocellular cancers³³. In fact, an earlier study demonstrated that pharmacological targeting of ALDH1B1 with small-molecule inhibitors has therapeutic potential to colorectal cancer treatment³². Both cysteines are primarily reactive to chloroacetamides (Figure 3F). C317 of ALDH6A1 was liganded by almost 90 chloroacetamides (CR>2), whereas C319 of ALDH1B1 was liganded by 2 acrylamides in addition to 57 chloroacetamides (Figure 3E, F). As another example, HERC2 belongs to the HECT (homologous to E6AP C-terminus) ubiquitin ligase family, which features an active-site cysteine (C4762) within the HECT domain responsible for accepting ubiquitin through a thioester bond and then transferring it to the substrate²⁷. In our dataset, approximately 20 chloroacetamides reacted with C4762 of HERC2 and exhibited CR values>2 (Figure 3E, F).

The electrophile library that we screened included 3 cyclic sulfone-containing chloroacetamides (Figure 3G). Cyclic sulfones have been found to function as important pharmacophores toward a range of targets in many therapeutic areas³⁴. In particular, early works have revealed that the cyclic sulfone moiety confers selectivity on a range of chloroacetamides toward C113 of the oncoprotein peptidyl-prolyl cis/trans isomerase PIN1³⁵, best exemplified by Sulfopin^4,6. Interestingly, we observed an additional cysteine, C57, of PIN1 (Figure 3G) that was also engaged by these compounds but eluded detection in all previous studies likely due to limited sensitivity^4,6. Considering both C57 and C113 are in the enzymatic PpiC domain, we speculate ligation of either would lead to alteration of the catalytic pocket and thereby its enzymatic activity. We further performed a dose-response study using CL41, CL71, CL74, and Sulfopin to evaluate their affinity toward the two cysteines. Despite presenting a similar trend when treated with the four electrophiles, C113 showed noticeably more ligation than C57 (Figure 3H), suggesting the binding preference. Sulfopin is the most potent among the four whereas CL41 is the least, elucidating the structure-activity relationship (SAR) potentially conferred by different R groups (neopentyl>cyclopentyl>ethyl>tolyl). In addition, these R groups presented various degrees of selectivity (Figure S3A, B), with Sulfopin being the most selective. Phenotypically, cells manifested different viability, and the most promiscuous CL41 caused cell death with the lowest LD₅₀ at 1.5 nM (Figure S3C).

Electrophile and protein structural environments

Local structural elements contribute to the reactivity and ligandability of cysteines. We used the deep screening data to investigate whether the new TMT-ABPP platform has any bias between exposed and buried cysteines. We chose to use the metric, prediction-aware part-sphere exposure (pPSE), as a measure of cysteine side chain accessibility. The pPSE is based on AlphaFold2-predicted protein structures, and it reflects the number of proximal α-carbons counted in a conical volume projecting 12 Å along the Cα-Cβ vector (or pseudo-vector in the case of glycine), with an internal angle of 70°. A high pPSE indicates a buried and less accessible residue, whereas a low pPSE suggests high accessibility^36,37. The pPSE values for all the cysteines in the UniProt human proteome have been calculated by White and colleagues³⁷. Here, we only included cysteines with confident AlphaFold2 prediction (pLDDT>70)³⁸ for comparison. We noted overlapping pPSE distributions among three groups: the DBIA-enriched reactive cysteines (n=27,308), ligandable cysteines (CR>2; n=2,746), and the entire human cysteinome (n=178,491; Figure 4A). We further classified cysteines according to their respective secondary structural contexts. Similarly, despite a slightly reduced efficacy of the screened electrophiles in liganding more buried cysteines in strands and unstructured regions, no obvious bias in the DBIA probe-enriched cysteines was observed. Overall, the result confirms the general applicability of the TMT-ABPP platform to probe cysteines possessing different intrinsic accessibility.

Protein domains are the fundamental elements of protein structure and often fold independently of the rest of the protein to form conserved tertiary structures, thus playing a critical role in determining characteristics in protein-ligand interactions³⁹. While previous studies have mostly focused on identifying ligands for individual cysteine sites, we reasoned with substantially improved cysteine coverage, including many more cysteines in different domains, we can further gain insight into how certain electrophiles may have preference to cysteines in a particular domain. We then mapped all cysteines into protein domains. Protein kinase domains usually contain redox-sensitive cysteines, which manifest as reactive cysteines (Figure 4B)⁴⁰. The prevalence of conserved cysteines in the kinase domain has been exploited in drug discovery. Key examples include Osimertinib targeting C797 of EGFR and Ibrutinib targeting C481 of BTK¹. More than 600 reactive cysteines in kinase domains were profiled, among which 54 were ligandable by at least one of the electrophiles (Figure 4B). The thioredoxin domain, widely existing in oxidoreductases, ranks second to the protein kinase domain in terms of the number of liganded cysteines. It modulates cellular homeostasis through a redox mechanism based on the reversible oxidation of two cysteine thiol groups to form a disulfide bond⁴⁰. The involvement of two redox-sensitive cysteines inevitably leads to their high reactivity, as 25 out of 88 cysteines in the thioredoxin domain were liganded (Figure 4B). Next, we asked whether certain electrophiles favor binding to cysteines in a particular domain. We noticed that AC19, 1-(4-(1H-1,2,3-triazol-1-yl)azepan-1-yl)prop-2-en-1-one, has enriched binding to cysteines in the EGF-like domain (Figure 4C). 77 cysteines in the EGF-like domain were profiled and they exhibited an average CR of 1.5 by AC19. The EGF-like domain, typically consisting of thirty to forty amino acids, is found in the sequence of epidermal growth factor (EGF) and in a conserved form across a wide range of proteins^41,42. Most occurrences of the EGF-like domain are found in the extracellular domain of membrane-bound receptors (e.g., LDLR, VLDLR, NOTCH2) or secreted proteins (e.g., EGF, TGF-α, Laminins). To dissect further the structure-activity relationship, we selected 4 structural analogs, 3 of which (AC82, AC123, and AC153) were not in the original 192 library, and assessed cysteine ligandability using them (Figure 4D). Interestingly, despite their structural similarity suggested by the Tanimoto coefficient⁴³ (Figure 4D), only AC19 displayed the preference to cysteines in EGF-like domains (Figure 4E; Table S3). For example, C434 of VLDLR, C1250 of NOTCH2, C4151 of LRP1, and C747 of JAG2 reside in their respective EGF-like domains and were all exclusively and significantly engaged by AC19 (Figure 4F). We therefore reasoned that the azepanyl group is the key in determining the SAR.

Identifying on/off targets of lead compounds and covalent drugs

We next asked whether the markedly improved sensitivity could lead to the identification of previously unknown cellular targets of well characterized compounds and covalent drugs. We chose 5 molecules (Ibrutinib, THZ1, ARS-1620, Sulfopin, and DMF) and included biological triplicate measurements in one 18-plex (Figure 5A). The experiment was performed in native lysates from three distinct cell types, namely HCC44 (lung adenocarcinoma), HEK293T (embryonic kidney), and SH-SY5Y (neuroblastoma), to evaluate any context-dependent target engagement. We treated the 20 μg cell lysates with individual compounds at 50 μM (except for DMF at 100 μM) for 1 hr. Taking advantage of the optimized FAIMS gas-phase fractionation (Figure S2C–E), we analyzed the three 18-plex samples (Figure 5A) with two LC-MS analyses, averaging 23,483 reactive cysteines in 6,708 proteins and totaling 33,341 cysteines in 8,328 proteins (Figure 5B; Table S4). We identified many cysteines that responded to the respective treatment. The KRAS^G12C mutation favors the active form of KRAS and results in abnormally high concentrations of GTP-bound KRAS, leading to hyperactivation of downstream oncogenic pathways and uncontrolled cell growth in cancers, such as non–small-cell lung cancer and colorectal cancer. ARS-1620⁵ is a potent covalent inhibitor of KRAS^G12C. Among the three cell lines, HCC44 possesses the G12C mutation, and the site was the most significantly engaged (Figure 5C). KRAS^G12C exhibited a CR of 2.7, suggesting an approximate 65% target engagement after 1-hr treatment. The incomplete engagement is likely attributed to the reduced KRAS cycling between GTP- and GDP-bound states in the lysate system, leaving a fraction of KRAS loaded with GTP at the 1-hour mark. The switch II binding pocket (S-IIP) is only accessible to ARS-1620 when KRAS is in the inactive GDP-bound conformation⁵. In addition to the designed target KRAS^G12C, several cysteines passed the threshold of CR value>2. C140 of ADK was commonly engaged in both HCC44 (CR=2.2) and HEK293T (CR=2.4) with statistical significance (p<0.01), and subthreshold in SH-SY5Y (CR=1.7) despite an equally significant p value (Figure 5C, D; Figure S4). ADK—adenosine kinase—catalyzes the phosphorylation of adenosine at the 5’ position in an ATP-dependent manner, and its dysfunction is involved in several pathologies, including diabetes, epilepsy, and cancer⁴⁴. Notably, this cysteine C140 is located within the active site of ADK⁴⁵, indicating the possibility of ARS-1620 acting as an ADK inhibitor. To test this hypothesis, we examined the activity of ADK under treatment with ARS-1620 (Figure 5E), along with an ADK inhibitor (the adenosine analogue A-134974) as the positive control⁴⁶. The assay is based on the use of inosine as a surrogate ADK substrate and a coupled reaction involving a highly active IMPDH (inosine monophosphate dehydrogenase) for a direct measurement of the inosine monophosphate (IMP) formed by ADK⁴⁷. Although ARS-1620 exhibited weaker inhibition compared to the A-134974, we observed a dose-dependent decrease in adenosine substrate conversion upon ARS-1620 treatment. We further docked ARS-1620 and A-134974 covalently or non-covalently with ADK (Figure 5F). ARS-1620 participates in a series of hydrogen bonds and π-stacking interactions with nearby residues Asn31, Ser82, Phe187, Thr190 and Phe218. These interactions have been observed in previous studies of potent ADK inhibitors, 5’-deoxy-5-iodotubercidin and an elongated alkynylpyrimidine ligand⁴⁵. The predicted binding pose for ARS-1620 at Cys140 of ADK shows high overlap with the predicted pose of A-134974 and the crystal structure of its analogue 5’-deoxy-5-iodotubercidin [PDB ID: 216A], corroborating the inhibitory effect of ARS-1620 (Figure 5G)⁴⁵.

Figure 5. Profiling of five model compounds using low input TMT-ABPP identifies new off-targets.

A) Five covalent compounds, including experimental molecules (ARS-1620, THZ1, Sulfopin) and FDA-approved drugs (Ibrutinib, dimethyl fumarate [DMF]), were used to profile their on- and off-targets in HCC44, HEK293T and SH-SY5Y lysates. 10 μg of cell lysate was treated with each compound (n=3 for each cell line) at a concentration of 100 μM (DMF) or 50 μM (other compounds) for 1 hour and then processed using TMT-ABPP. Control samples (n=3 for each cell line) were treated with DMSO. Three TMT18-plexes were generated and analyzed.

B) The overlap of cysteines and proteins quantified among experiments using different cell lysates. In total, 33,341 cysteines from 8,328 proteins were quantified.

C) Average CRs (n=3) of 22,251 quantified cysteines in HCC44 treated with 50 μM ARS-1620. Cysteines with CR values >2 are labeled. In addition to its designed target KRAS^G12C, ADK C140 and SETD1B C1007 also respond to ARS1620 treatment in HCC44.

D) The engagement of ADK C140 by ARS-1620 was reproduced in all three cell lines. Bars represent mean CR ± s.d. (n=3 for each cell line). **p ≤ 0.01, ***p ≤ 0.001.

E) ARS-1620 inhibits ADK activity and substrate conversion. Points represent mean absorbance intensity± s.d. (n=4), corresponding to ADK activity. A-134974 is a selective ADK inhibitor and was used as the positive control (n=3), and DMSO was used as a negative control (n=3).

F) Covalent docking prediction of ARS-1620 interactions to ADK (PDB ID: 2I6B). ADK residues shown in gray, ARS-1620 shown in teal, hydrogen bonds shown as yellow dashes and π-π interactions shown as black dashes.

G) Overlay of docked poses for ARS-1620 (teal), positive control A-134974 (purple) and crystal structure of 5’-deoxy-5iodotubercidin (gold, PDB ID: 2I6A). Cys140 of ADK shown as gray sticks.

H) The engagement of LYN C381 by Ibrutinib was reproduced in all three cell lines. Bars represent mean CR ± s.d. (n=3).

I) Ibrutinib inhibiting LYN kinase activity and ATP conversion with a IC₅₀ of 1 μM. Points represent average ATP conversion ± s.d. (n=3). Staurosporine (IC₅₀=69 nM) was used as the positive control.

In the same 18-plex experiment, we also profiled the Bruton’s tyrosine kinase (BTK) inhibitor Ibrutinib, an FDA-approved drug to treat chronic lymphocytic leukemia (CLL)⁴⁸. We consistently identified that Ibrutinib interacts with specific residue C381 within the kinase domain of the tyrosine-protein kinase LYN in all three proteomes (Figure 5H; Figure S4). This finding corroborates the initial study on Ibrutinib, which described its inhibitory effect on LYN despite not identifying the covalent interaction⁴⁸. We then further confirmed it using an in vitro kinase assay⁴⁹ measuring the conversion of ATP to ADP to validate the inhibitory effect of ibrutinib on LYN kinase (Figure 5I). The remarkable depth from 20 μg proteome input also led to the identification of potential cellular targets of THZ-1, Sulfopin, and DMF in addition to existing knowledge (Figure S4)^4,50,51.

Discussion

Since its inception, significant strides have been made in the field of ABPP, and it has greatly expanded its scope from solely focusing on enzymatic active sites to broadly mapping all reactive hotspots, enabling the discovery of valuable electrophile-based probes and drug candidates^1,3,8,29,52. Despite showing great potential for high-throughput unbiased drug discovery, proteome-wide profiling of thousands of compounds remains a formidable challenge, both technically and financially. As a result, current ABPP research is still limited to the characterization of a small number of electrophiles and a small fraction of the cysteinome^3,6,11,23. To truly enable proteome-wide discovery of protein-ligand interactions and comprehensive assessment of their affinity and selectivity entails substantial changes to existing approaches to build a platform amenable to plate-based sample processing using minimal input while substantially augmenting the coverage of cysteinome.

We generated a 96-well plate-based sample processing and a newly developed 18-plex TMT-based sample multiplexing strategy. As used, the 18-plex TMT reagents allow for a boost in throughput while adding a second DMSO control (two in total) to improve consistency¹³. We further added a number of major changes, primarily including a much higher capacity streptavidin-agarose resin for enriching reactive cysteine-containing peptides, a paramagnetic bead-based single-pot workflow, and an online gas-phase fractionation using the FAIMS device. As a result, the new TMT-ABPP platform enables routine profiling of reactive cysteines with 2–3X increase in depth (18,000 or 24,000 sites depending on the input) compared to a similar experiment using other existing methods. In addition, all steps are streamlined in 96-well plates. The process, spanning compound treatment to sample preparation for LC-MS analysis, can be efficiently completed in a time frame of just 2 days. The fact that all steps before pooling 18 TMT-labeled samples can be performed in the same plate and many plates can be processed in parallel makes it highly conducive for scaling up and automation. After TMT labeling, 18 samples are pooled into a single tube, consolidating the contents of an entire 96-well plate into just 6 multiplexed samples. We note that all the quantification was done using high-resolution MS2 (HRMS2) coupled with FAIMS. We previously demonstrated the FAIMS device can significantly reduce interference and improve quantitative accuracy in HRMS2-based quantification via gas-phase fractionation²⁰. Although MS3-based quantification still outperforms in accuracy⁵³, it comes at the cost of reduced number of measured peptides. We observed an average ~30% reduction in measured peptides²⁰, and we speculate the trend also translates into number of measured reactive cysteines. As the depth is critical to the initial proteome-wide screening, we reason that the FAIMS-MS2-based quantification achieves a balance between sensitivity and accuracy. Upon identifying a hit, more accurate and targeted quantification approaches, such as those developed by our lab^54,55, can be used to further establish fine structure-activity relationships when needed.

Two major applications can significantly benefit from the new TMT-ABPP platform described here. First, it can be applied to the screening of large electrophile libraries more efficiently and in a cost-effective way compared to the existing SLC-ABPP⁶. We showcased it by screening a library of 192 electrophiles. The dataset extends the number of reactive cysteines in HEK293T to nearly 39,000, surpassing any existing large datasets (Figure 3B, C)^6,11. It offers deep insight into protein-ligand interaction with many different perspectives. PIN1 is an attractive therapeutic target in cancer drug development. We identified an additional reactive cysteine, C57, alongside the known C113. We further revealed that the compounds capable of binding to C113 can also bind to C57, though with reduced potency. We note that exploring potential allosteric and steric interplay between the two ligation events on PIN1 is needed in the future to fully characterize the cysteine-ligand interactions. Through a comprehensive investigation utilizing 4,647,501 measurements of unique electrophile-cysteine pairs generated by our screening, we gained valuable insights into protein structural elements within the dataset (Figure 4). The TMT-ABPP platform provides structurally unbiased enrichment of reactive cysteines, as demonstrated by the pPSE values of all measured cysteines, which exhibit an overall distribution overlapping with that of the entire human cysteinome. We highlighted that the AC19, 1-(4-(1H-1,2,3-triazol-1-yl)azepan-1-yl)prop-2-en-1-one, favors binding to cysteines in the EGF-like domain. Based on this observation, we anticipate AC19 will provide a foundation for future probe development to assess function of proteins possessing the EGF-like domain. Second, we demonstrated that the TMT-ABPP platform can extensively identify cellular targets of existing covalent drugs. Better annotation of on/off targets is critical to improving efficacy and reducing undesired side effects. This can be achieved by exploring the dose-response relationship or/and context-dependent target engagement, and we have demonstrated both avenues. For example, we profiled 5 compounds in three different cell lines. The dataset consists of 33,341 reactive cysteines in 8,328 proteins (Figure 5B). Compared to a recent study¹¹ that recorded 34,225 cysteines by assaying 7 different cell lines coupled with different proteolytic digests and subcellular fractionation, this also represents a remarkable advance. The enhanced coverage translates into the identification of additional drug targets, such as the C140 of ADK for ARS-1620, and C381 of LYN for Ibrutinib. Given the indispensable role of LYN in CLL⁵⁶, we speculate the off-target and resulting inhibitory effect (Figure 5H, I) is indeed a synergistic mode-of-action in treating CLL. Further experiments are required to distinguish the roles of the electrophilic warhead from those of the ARS-1620 and Ibrutinib scaffolds in their respective off-target effects, as the potency of covalent inhibition is determined collectively by the initial non-covalent binding affinity conferred by the scaffold and the reactivity of the covalent warhead.

Looking beyond what we have showcased in this study, we further posit that the TMT-ABPP can be adopted in a number of other applications. Cysteine reactivity and ligandability are cellular context-dependent and changes can be induced by many internal and external factors such as cell cycle and oxidative stress. Therefore, interrogating reactive cysteines under different conditions may offer insight into cysteine-mediated biology⁷ and lead to identification of cellular state-specific protein-ligand interactions. At the molecular level, it has been shown that cysteine reactivity can be altered by protein phosphorylation on serine/threonine residues⁵⁷, and we speculate that similar interplays between cysteines and other post-translational modifications (PTMs) (e.g., acetylation, ubiquitination) also exist. In addition, many protein-protein interaction involve cysteines and they can be perturbed by site-specific liganding events⁵⁸. These specific studies can all benefit from enhanced coverage to resolve the fine interplay between cysteines and PTMs or protein-protein interactions. Moreover, we envision that the TMT-ABPP platform will be further improved with the next-generation mass spectrometry instruments, such as the Orbitrap Astral, and we will potentially be able to work with even lower input (1 μg native proteome) in 384-well plates.

In conclusion, TMT-ABPP integrates 18-plex sample multiplexing and a miniaturized workflow in 96-well plates to enable highly efficient electrophile screening using as little as 10 or 20 μg native proteome input while providing unparalleled depth of ~18,000 or ~24,000 reactive cysteines, respectively. We have here provided the detailed workflow as a resource to facilitate general chemoproteomics research and high-throughput screening. We showcased the capability and potential to enable future large-scale screening by performing an initial screen of an electrophile library consisting of 192 molecules. We also demonstrated the platform’s capability to identify new protein targets of well-studied covalent drugs and experimental molecules.

Limitations of the study

TMT-ABPP, while representing a major step forward, captures only a fraction of the extensive human cysteinome, which encompasses approximately 260,000 cysteines. We recognize that our current methodology falls short of providing a comprehensive evaluation. To overcome these limitations, forthcoming endeavors must integrate innovative chemical techniques, advanced mass spectrometric platforms, and sophisticated computational strategies. Moreover, the observed changes using the DBIA probe may arise from either competitive ligand binding or allosteric effects altering the reactivity of specific cysteines. The current TMT-ABPP workflow cannot distinguish between the two mechanisms, necessitating additional investigation when specific alterations are of interest.

Significance

Activity-based proteome profiling (ABPP) has recently become the state-of-the-art strategy to profile reactive cysteines and their ligandability with electrophilic compounds. Despite the latest advances, the ability to perform large-scale library screening using minimal starting material while achieving desired depth remains a formidable task due to a lack of platforms suitable for high throughput. By integrating a number of major changes, including i) an 18-plex TMT sample multiplexing strategy, ii) magnetic beads-based one-pot workflow, iii) a 10X higher capacity streptavidin resin, and iv) optimized mass spectrometry analyses, we have created a platform that enhances coverage by 2–3X and reduces input by 5–10X relative to previous state-of-the-art ABPP methods. This TMT-ABPP platform is fully amenable to 96-well plate-based sample manipulation and enables routine interrogation of ~18,000 or ~24,000 reactive cysteines using 10 or 20 μg proteome input, respectively. We demonstrated the TMT-ABPP platform by generating several proteome-wide profiling datasets of reactive cysteines in two major applications: 1) identification of new protein-ligand interactions to facilitate chemical biology research and drug discovery and 2) investigation of cellular targets for well-studied covalent drugs. This platform holds great potential for advancing our understanding of protein function and facilitating the discovery and characterization of new therapeutic targets using chemical biology approaches.

STAR Methods

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by Lead Contact, Steven P Gygi (steven_gygi@hms.harvard.edu).

Materials Availability

All unique/stable reagents generated in this study will be provided without restriction as long as stocks remain available and reasonable compensation is provided by the requestor to cover processing and shipment.

Data and Code Availability

• The mass spectrometry data have been deposited at the ProteomeXchange Consortium and are publicly available as of the date of publication. The accession number is PXD044402.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental Model and Study Participant Details

Cell Line and Cell Culture

Human cell lines HEK293T and SH-SY5Y were obtained from the American Type Culture Collection (ATCC), and HCC44 cells were purchased from DSMZ. HEK293T cells were cultured in high-glucose DMEM medium (Corning) supplemented with 10% FBS and 1% Penicillin/Streptomycin. HCC44 cells were cultured in RPMI-1640 medium (Corning) supplemented with 10% FBS and 1% Penicillin/Streptomycin. SH-SY5Y cells were cultured in MEM/F12 (1:1) medium (Corning) supplemented with 10% FBS and 1% Penicillin/Streptomycin. Cells were grown at 37°C in a humidified 5% CO₂ atmosphere.

Method Details

Compounds

All commercial compounds were prepared stock solution in DMSO.

Cell lysate preparation

Cells were grown in 15-cm dish to near confluent and washed twice with cold PBS. Cells were scraped, collected and stored at −80 °C until lysis. Frozen cells were resuspended with lysis buffer (PBS, pH 7.4, 0.1% NP-40) by syringe equipped with 21-gauge needle. The crude lysate was then subjected to EpiShear^™ Probe Sonicator (Active Motif) for further homogenization (5 min, 3-s on, 3-s off, 50% amp) on ice. Clear cell lysate containing soluble proteome was collected after centrifugation at 1400 g for 5 minutes. Protein concentration was measured by BCA assay.

Plate- and sample multiplexing-based reactive cysteine profiling

Cell lysate was diluted to 2 μg/μL with lysis buffer. 5–15 μL lysate, containing 10–30 μg protein respectively, was loaded into each well in 96-well plate. The following steps are used for 20 μg protein input. Amount of each reagent should be adjusted accordingly depending on the input. 5 μL of compound solution in lysis buffer was added to a final concentration of 50 μM and incubated for 1 hour. 5 μL of DBIA solution in lysis buffer was added to a concentration of 500 μM and incubated in dark for 1 hour. Addition of 3 μL SP3 beads (1:1 mixture of hydrophobic and hydrophilic type, 50 mg/mL, Cat. #45152105050250 and Cat. #65152105050250,) was followed by addition of 30 μL ~98% ethanol supplemented with 20 mM DTT. Lysate-bead mixture was incubated for 10 minutes with mild shaking before placing the plate on a magnetic stand to aspirate the supernatant. Beads were washed once with 80% ethanol and resuspended in 20 μL lysis buffer supplemented with 20 mM IAA and incubated in dark for 30 minutes with vigorous shaking. 40 μL ~98% ethanol supplemented with 20 mM DTT was added to the mixture. Mild shaking was performed before two washes using 80 % ethanol. The aqueous phase was then removed followed by adding 20 μL 200 mM EPPS buffer (pH 8.5) containing 0.2 μg Lys-C. After 3-hour incubation at room temperature, 10 μL EPPS buffer containing 0.2 μg trypsin was added and incubated with beads at 37 °C overnight.

TMT labeling and multiplexing

To the mixture of digested peptides and beads, 11 μL acetonitrile and 4 μL TMT (10 μg/μL) reagent were sequentially added, followed by gentle mixing for 60 minutes at room temperature. Reaction was quenched by adding 7 μL 5% hydroxyl amine. All samples containing peptide-beads mixture were combined and dried using a SpeedVac. The resulting multiplexed sample was resuspended in 10% formic acid and clear supernatant was separated from beads by magnetic stand. Clear supernatant was then desalted using a 100-mg Sep-Pak column and dried by SpeedVac.

Enrichment of cysteine peptides

Desalted TMT-labeled peptides were resuspended in 310 μL100 mM HEPES buffer (pH 7.4). Addition of pre-washed 50 μL Pierce^™ High Capacity Streptavidin Agarose (Cat. #20359) was followed by incubation of peptide-beads mixture at room temperature for 3 hours. Resulting mixture was then loaded on a Ultrafree-MC centrifugal filter (hydrophilic PTFE, 0.22 μm pore size) and centrifugated at 800 g for 30 seconds. Beads were washed sequentially with 300 μL 100 mM HEPES (pH 7.4) with 0.05% NP-40 twice, 350 μL 100 mM HEPES (pH 7.4) three times and 400 μL water once. Peptide were eluted sequentially by 1) incubation with elution buffer (80% acetonitrile, 0.1% formic acid) for 20 minutes at room temperature; 2) incubation with elution buffer for 10 minutes at room temperature; 2) incubation with elution buffer for 10 minutes at 72 °C. The combined eluent was dried in a SpeedVac and desalted via StageTip prior to LC-MS/MS analysis.

LC-FAIMS-MS/MS analysis

Samples were resuspended in LC-MS loading buffer (5% ACN and 5% FA) and loaded on a 100-μm capillary column packed with 30 cm of Accucore 150 resin (2.6 μm, 150Å; Thermo Fisher Scientific). Enriched cysteine samples were separated using a 180-min method on a Proxeon NanoLC-1200 UPLC system. Cysteine data were collected using a high-resolution MS2 method on an Orbitrap Eclipse mass spectrometer coupled with a FAIMS Pro device. Data were collected alternating between a set of three FAIMS compensation voltages (CVs). For the single-shot analysis starting with 10 μg lysate per TMT channel, only one set of CV values (−60, −45 and −35V) was used. For the double-shot analysis starting with 20 μg lysate per TMT channel, two sets of CV values (−60, −45 and −35V; −70, −55 and −30) were used. MS1 scans were collected in the Orbitrap with a resolution setting of 60 K, a mass range of 400–1600 m/z, an AGC at 100%, and a maximum injection time of 50 ms. MS2 scans were acquired in Top Speed mode with a cycle time of 1 s. Peptide precursors were selected and fragmented using HCD with a collision energy of 36. MS2 scans were collected in the Orbitrap with a resolution of 50K, a fixed scan range of 110–2000 m/z, and a 500% AGC with a maximum injection time of 86 ms. Dynamic exclusion was set to 120 s with a mass tolerance of ± 10 p.p.m.. The flowthrough were separated using a 60-min method and analyzed by FAIMS-MS/MS using CV set of −80, −60 and −40. Data dependent analysis were performed in same setting as analyzing cysteine samples except a dynamic exclusion time of 90 s.

Data analysis for cysteine identification, localization, and quantification

Raw files of cysteine profiling were searched using the open-source Comet search engine (ver. 2019.01.5)⁶¹ with the Uniprot human proteome database (downloaded 11/24/2021) with contaminants and reverse decoy sequences appended. Precursor error tolerance was 50 p.p.m. and fragment error tolerance was 0.02 Da. Static modifications include Cys carboxyamidomethylation (+57.0215) and TMTpro18 (+304.2071) on Lys side chains and peptide N-termini. Methionine oxidation (+15.9949) and DBIA-modification on cysteine residues (+239.1628) were allowed as variable modifications. Peptide spectral matches were filtered to a peptide false discovery rate (FDR) of <1% using linear discriminant analysis employing a target-decoy strategy^62,63. Resulting peptides were further filtered to obtain a 1% protein FDR at the entire dataset level (including all plexes in an experiment)⁶⁴. Cysteine-modified peptides were filtered for site localization using the AScorePro algorithm with a cutoff of 13 (P < 0.05) as previously described^65,66. Overlapping peptide sequences generated from different charge states, retention times and tryptic termini were grouped together into a single entry. A single quantitative value was reported, and only unique peptides were reported. Reporter ion intensities were adjusted to correct for impurities during synthesis of different TMT reagents according to the manufacturer’s specifications. For quantification of each MS2 spectrum, a total sum signal-to-noise of all reporter ions of 180 (TMTPro 18-plex) was required. Peptide quantitative values were normalized so that the sum of the signal for all proteins in each channel was equal to account for sample loading differences (column normalization).

ADK activity assay

The ADK activity assay was performed using the PRECICE^® ADK Assay Kit (NovoCIB) following the manufacturer’s protocol. Briefly, all cofactors (DTT and NAD), ADK kinase, IMPDH and ATP were reconstituted in 1X reaction buffer (diluted from 5X Reaction Buffer) to desire concentration, aliquoted and stored at −80°C until use. Aliquots of buffer and reagent were thawed and mixed to make final complete reaction buffer containing 100 mM Tris-HCl pH 8.5, 250 mM KCl, 10 mM MgCl₂, 2.5 mM NAD, 2.75 mM ATP, 20 mU/ml IMPDH, 2.2 mU/ml human recombinant ADK. To assay wells of clear round-bottom 96-well plate, 2 μL of vehicle (DMSO) or vehicle-dissolved compounds was added, followed by the addition of 200 μL complete reaction buffer. After a 1-minute agitation, the absorbance at 340 nm (A₃₄₀) of the assay wells was monitored by SpectraMax M5 microplate reader (Molecular Devices) with kinetic measurements done every 1 minute at 37°C for total 15 minutes as pre-incubation readout to check background signal. Then, 10 μL of 50 mM of Inosine was added to each well and another kinetic measurement of A₃₄₀ was done every 1 minute at 37°C for total 40 minutes to monitor the real-time ADK substrate conversion. ADK activity was calculated as relative A₃₄₀ to end-point value observed in vehicle control.

LYN kinase activity assay

The LYN kinase activity assay was performed using the LYN A Kinase Enzyme System (Promega) and the ADP-Glo^™ Assay (Promega) following the manufacturer’s protocol. Briefly, 1 μL of vehicle (DMSO) or compound solution in 1X reaction buffer (diluted from given 5X stock, 50 μM DTT and 2 mM MnCl₂) was added to white solid and flat bottom 384-well plate. Then, 2 μL of kinase solution containing 10 ng recombinant human LYN protein in 1X reaction buffer was added to each well, followed by brief centrifugation and mild shaking for 5 minutes. To start the enzymatic reaction, 2 μL of substrate-ATP mix containing 1 μg Poly(Glu4Tyr1) substrate and 10 μM ATP in 1X reaction buffer was added to each well. After brief centrifugation and mild shaking for 5 minutes, the plate was incubated at room temperature for 60 minutes. In parallel, a standard curve of ATP/ADP conversion was prepared and loaded to separate wells on the same plate. 5 μL of ADP-Glo Reagent was added to each well for 40-minute incubation at room temperature and then 10 μL of Kinase Detection Reagent was added for another 30-minute incubation. The final luminescence was read on SpectraMax M5 microplate reader (Molecular Devices) with integration time of 1s. A standard curve was generated by linear fitting of luminescence signal v.s. ATP/ADP conversion. The ATP conversion under treatment of vehicle or compound was calculated based on the standard curve.

Docking

Docking complexes were predicted using Schrodinger Maestro Version 13.1.141, release 2022–1^67–69. ADK was prepared using default settings in the protein preparation wizard with a single protein chain (chain A) and ligand. The non-covalent docking experiments were conducted using the PDB structure 2I6A⁴⁵ and covalent docking experiments were conducted using PDB structure 2I6B⁴⁵. Water was removed and the structures underwent minimalization using default settings. Ligands were prepared using default LigPrep settings and a maximum of 32 states were generated per ligand. A box size of 20 Å centered around residues 68, 123, 142, 201 and 297 was used for both experiments. The non-covalent docking was conducted in extra precision mode using default settings and covalent docking was conducted in pose prediction mode using Cys123 of the structure (Cys140 of the protein) as the reactive residue, Michael addition reaction SMARTS, a 3.5kcal/mol energy cutoff and a maximum output of 3 poses per ligand.

Cell Viability Assays

K562 cells were harvested and plated with 5×10³ cells in 100 μL media per well in 96-well plate. After overnight seeding, 25 μL media containing 5X dosing concentration of the compounds or vehicle was added to each well. After a 72-hour treatment at 37°C in a humidified 5% CO₂ atmosphere, 12.5 μL 10X resazurin solution (0.5 mg/mL) was added to each well. Then cells were incubated at 37°C overnight. The optical density was read at 570 nm and 600 nm by SpectraMax M5 microplate reader (Molecular Devices).

Quantification and Statistical Analysis

Calculation of competition ratio

The competition ratio (CR) was calculated as:

$$\mathit{\text{CR}}=\frac{\mathit{\text{TMT}}\mathit{\text{SN}}\mathit{\text{of}}\mathit{\text{vehicle}}(\mathit{\text{DMSO}})}{\mathit{\text{TMT}}\mathit{\text{SN}}\mathit{\text{of}}\mathit{\text{compound}}}$$

If vehicle-treatment was performed in multiple replicates, the averaged TMT SN of all vehicle replicates was used to calculate CR.

Calculation of Relative Cell Viability

The relative viability (RV) was measured and calculated as:

$$\mathit{\text{RV}}=\frac{117216\times O{D}_{570}of\mathit{\text{sample}}-80586\times O{D}_{600}of\mathit{\text{sample}}}{117216\times O{D}_{570}of\mathit{\text{vehicle}}-80586\times O{D}_{600}\mathit{\text{of}}\mathit{\text{vehicle}}}$$

Normalized data was graphed as bar graph or dot plot representing as mean of relative viability (n = 3) with ± SD as error bar.

Statistics

All statistical analyses were done using R (ver 4.3.1) in R Studio. Statistical significance was analyzed by performing unpaired two-tailed student t-test. P values were adjusted using the Benjamini–Hochberg method and filtered at q < 0.05 if needed.

Supplementary Material

3

Table S1. Cysteine competition ratios with scout fragments KB02, KB03, and KB05, related to Figure 2.

4

Table S2. Cysteine competition ratios from the screening of 192 electrophiles, related to Figure 3, 4

5

Table S3. Cysteine competition ratios with compound AC19 and its analogues, related to Figure 4

6

Table S4. Cysteine competition ratios from the target profiling of covalent drugs in 3 different cell lines, related to Figure 5.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant Proteins
KB02	Sigma-Aldrich	Cat# 912131-100MG; CAS: 57368-84-0
KB03	Sigma-Aldrich	Cat# 912654-100MG CAS: 790-75-0
KB05	Sigma-Aldrich	Cat# 911798-100MG CAS: 1956368-15-2
ARS-1620	Selleckchem	Cat# S8707
Ibrutinib	Selleckchem	Cat# S2680
THZ-1	Selleckchem	Cat# S7549
Dimethyl fumarate (DMF)	Sigma-Aldrich	Cat# PHR2118
Sulfopin	Selleckchem	Cat# S9782
Phosphate buffered saline (PBS)	Corning	Cat# 46–013-CM
NP-40	Sigma	Cat# 492016-100ML CAS: 9016-45-9
Dithiothreitol (DTT)	Sigma	Cat# 43815-1G CAS: 3483-12-3
Sera-Mag™ SpeedBead Carboxylate-Modified [E3] Magnetic Particles	Cytiva	Cat# 65152105050250
Sera-Mag™ SpeedBead Carboxylate-Modified [E7] Magnetic Particles	Cytiva	Cat# 45152105050250
Iodoacetamide (IAA)	Sigma	Cat# I1149-5G CAS: 144-48-9
Trypsin	Thermo Fisher	Cat# 90305
Lysyl Endopeptidase (Lys-C)	Wako	Cat# 129-02541
TMTpro 18-plex	Thermo Fisher	Cat# A52045
Customized covalent 192-fragment library (see Table S2)	Enamine	N/A
AC82	Enamine	Cat# EN300-2540835 CAS: 2179724-04-8
AC123	Enamine	Cat# EN300-6702314 CAS: 2224251-71-0
AC153	Enamine	Cat# EN300-6702221 CAS: 2187469-58-3
Critical commercial assays
PRECICE® Adenosine Kinase Assay Kit	NovoCIB SAS	Cat# K0507-01-1plate
LYN A Kinase Enzyme System	Promega	Cat# VA7486
ADP-Glo™ Assay	Promega	Cat# V6930
Deposited data
PRIDE	This paper	PXD: 044402
Experimental models: Cell lines
HEK293T	ATCC	Cat# CRL-3216
HCC44	DSMZ	Cat# ACC 534
SH-SY5Y	ATCC	Cat# CRL-2266
Software and algorithms
GraphPad Prism 9	GraphPad	https://www.graphpad.com/
R 4.3.1	R-Project	https://cran.rstudio.com/
Schrödinger Small Molecule Drug Discovery Platform 13.1.141	Schrödinger	https://www.schrodinger.com/

Highlights.

• Sample multiplexing-based high-throughput reactive cysteine profiling in 96-well plates

• Routine assay of ~18,000 or ~24,000 reactive cysteines using 10 or 20 μg input proteome

• Ligandability of 38,450 cysteines from 8,274 human proteins with 192 electrophiles

• Revelation of off-targets includes ADK for ARS-1620 and LYN for Ibrutinib