Robust proteome profiling of cysteine-reactive fragments using label-free chemoproteomics

Biggs, George S.; Cawood, Emma E.; Vuorinen, Aini; McCarthy, William J.; Wilders, Harry; Riziotis, Ioannis G.; van der Zouwen, Antonie J.; Pettinger, Jonathan; Nightingale, Luke; Chen, Peiling; Powell, Andrew J.; House, David; Boulton, Simon J.; Skehel, J. Mark; Rittinger, Katrin; Bush, Jacob T.

doi:10.1038/s41467-024-55057-5

Download PDF

Article
Open access
Published: 02 January 2025

Robust proteome profiling of cysteine-reactive fragments using label-free chemoproteomics

Nature Communications volume 16, Article number: 73 (2025) Cite this article

15k Accesses
10 Citations
25 Altmetric
Metrics details

Subjects

This article has been updated

Abstract

Identifying pharmacological probes for human proteins represents a key opportunity to accelerate the discovery of new therapeutics. High-content screening approaches to expand the ligandable proteome offer the potential to expedite the discovery of novel chemical probes to study protein function. Screening libraries of reactive fragments by chemoproteomics offers a compelling approach to ligand discovery, however, optimising sample throughput, proteomic depth, and data reproducibility remains a key challenge. We report a versatile, label-free quantification proteomics platform for competitive profiling of cysteine-reactive fragments against the native proteome. This high-throughput platform combines SP4 plate-based sample preparation with rapid chromatographic gradients. Data-independent acquisition performed on a Bruker timsTOF Pro 2 consistently identified ~23,000 cysteine sites per run, with a total of ~32,000 cysteine sites profiled in HEK293T and Jurkat lysate. Crucially, this depth in cysteinome coverage is met with high data completeness, enabling robust identification of liganded proteins. In this study, 80 reactive fragments were screened in two cell lines identifying >400 ligand-protein interactions. Hits were validated through concentration-response experiments and the platform was utilised for hit expansion and live cell experiments. This label-free platform represents a significant step forward in high-throughput proteomics to evaluate ligandability of cysteines across the human proteome.

Assigning functionality to cysteines by base editing of cancer dependency genes

Article 02 October 2023

Proteome-wide ligandability maps of drugs with diverse cysteine-reactive chemotypes

Article Open access 26 May 2025

Reimagining high-throughput profiling of reactive cysteines for cell-based screening of large electrophile libraries

Article 04 January 2021

Introduction

Small molecule probes offer powerful tools for the study of biological systems and can serve as starting points for the development of therapeutics¹. The vast majority of human proteins lack such chemical tools, which hinders our ability to explore the function of these proteins in the context of health and disease^2,3. The development of high-quality ligands across the human proteome is now recognised as a key objective to enable the functional studies of biological systems and to accelerate identification of therapeutic opportunities^4,5,6.

Traditional methods for ligand and tool discovery, such as high-throughput screening of small molecules with individual proteins, are resource intensive, prohibiting their utility in expanding the liganded proteome. Furthermore, the study of purified and often truncated forms of proteins is a poor reflection of the interactions made by the full-length proteins in their native cellular environment. Novel methods to enhance the throughput, scope, and accessibility of ligand discovery, particularly in the context of the native proteome, will be essential for realising the ambition of the research community to discover a chemical probe for every expressed protein^2,3,7.

Mass spectrometry (MS)-based chemoproteomics methods are emerging as powerful approaches to expand the ligandable proteome, enabling the identification of small molecule-protein interactions in a cellular context. Small molecule ligands that act via an irreversible covalent mechanism are particularly suited to these studies, as the interactions are retained through sample preparation and mass spectrometry analysis, allowing for the robust detection of binding events from complex mixtures^8,9,10. The covalent bond can also provide an increase in potency due to prolonged target engagement, leading to the discovery of probe molecules from relatively low molecular weight chemotypes. The benefits of covalent fragment-based ligand discovery can then be exploited by using modestly sized libraries (~10²–10³) of low molecular weight fragments (<300 Da), while still efficiently covering chemical space^11,12,13.

Chemoproteomics screening of covalent small molecules was pioneered using activity-based probes (ABPs) for the development of inhibitors against specific enzyme families, e.g., serine hydrolases and de-ubiquitinases^14,15,16,17. Screening compounds in competition with ABPs enables excellent sensitivity and coverage within the targeted protein family, but does not inform on the proteome-wide activity of these small molecules. Recent studies have developed chemoproteomics methods with expanded proteome coverage by using family-agnostic probes for nucleophilic residues, e.g., cysteine and lysine^18,19,20,21. In particular, hyperreactive iodoacetamide probes that enable enrichment and quantification of cysteine-containing peptides have been employed for competitive cysteinome profiling of electrophilic fragments by chemoproteomics^22,23.

Key challenges still exist in the development of chemoproteomics platforms for competitive profiling of large compound libraries. Sample preparation and analysis must have sufficient throughput to profile entire libraries of covalent compounds, and the analytical technique must deliver sufficient sensitivity to detect a significant portion of the expressed proteome. Additionally, excellent reproducibility and data completeness is required to enable robust hit calling and the generation of full matrix datasets, which are suited to the implementation of machine learning models to drive iterative library design. To date, methods have employed long MS-proteomics acquisition times (~3-h chromatographic gradients) and data-dependent acquisition (DDA) mass spectrometry. Improvements in throughput have been achieved by multiplexing using tandem mass tag (TMT) labelling, however these batch-based DDA analyses can give low data completeness and poor reproducibility when combining datasets, which, in addition to the costs associated with TMT reagents, limits their accessibility for large library profiling^24,25,26.

Here, we present a high-throughput label-free quantification chemoproteomics (HT-LFQ) platform for profiling covalent libraries against the cysteinome. The method offers sensitivity and cysteinome coverage that compares favourably with reported methods to date, and offers improved reproducibility and data completeness. Sample preparation is performed using a 96-well plate-based workflow, consisting of a low-cost protein clean-up method with no requirement for isotopic labelling. Sample analysis employs label-free quantification (LFQ) and data independent acquisition (DIA), which yields excellent sensitivity and reproducibility of peptide detection. Importantly, the fragmentation and analysis of all precursors in DIA affords improved data completeness between experiments, which contrasts with more variable peptide identification in DDA^27,28,29. Using our HT-LFQ chemoproteomics platform, we screen a library of 80 chloroacetamide fragments against the cysteinome in two cell lines (HEK293T and Jurkat), and identify ligands for over 400 cysteine sites, including a number of protein families of high interest to drug discovery. The high-throughput nature and reproducibility of our platform allows ready access to further hit characterisation, including concentration-response studies and interrogation of structure-activity relationships. Collectively, we demonstrate that our HT-LFQ platform represents a powerful methodology to enable efficient discovery of chemical tools for proteins from biological samples.

Results

A high-throughput, label-free chemoproteomics workflow

To develop a high-throughput label-free quantification DIA chemoproteomics platform capable of screening compound libraries against the cysteinome, we employed a competitive profiling strategy using a previously reported hyperreactive iodoacetamide desthiobiotin (IA-DTB) probe¹⁸. Cell lysates were treated with cysteine-reactive fragments, followed by treatment with IA-DTB to enrich cysteine-containing peptides. MS-based quantification of enriched peptides and comparison of these peptide intensities between fragment-treated and DMSO control samples enabled calculation of the fragment engagement of each cysteine, reported as a competition ratio (CR).

We developed a sample preparation protocol to ensure reproducible recovery of cysteine-containing peptides from cell lysates in 96-well plate-based format, without the need for isotopic labelling (Fig. 1a). Following treatment of lysates with IA-DTB, a plate-based sample clean-up was performed, employing solvent precipitation on glass beads followed by an on-bead tryptic digestion. This precipitation approach allows for consistent protein recovery alongside efficient removal of excess small molecules and detergents³⁰. Finally, digested and desthiobiotin-modified peptides were captured on high-capacity neutravidin resin and recovered using mildly acidic aqueous/organic mixtures. In this workflow, samples remain in 96-well plate format from compound treatment through to mass spectrometer injection, and a single experimentalist can readily prepare 384 samples (4 plates) in 2–3 days. The plate-based and label-free nature of this sample preparation method provides the throughput and reproducibility required for efficient screening of large libraries of covalent fragments.

**Fig. 1: Workflow and cysteinome coverage from our HT-LFQ chemoproteomic method.**

Liquid chromatography-mass spectrometry (LC-MS) data was acquired using an Evosep One coupled to a Bruker timsTOF Pro 2. On this mass spectrometer, cysteinome coverage is maximised by separating peptides based on four properties (ion mobility, retention time, mass-to-charge ratio and intensity) and employing DIA with parallel accumulation-serial fragmentation (PASEF) for improved precursor identification and quantification (Fig. 1b)³¹. Initially, a hybrid (DDA/DIA) spectral library of IA-DTB-modified precursors was built utilising off-line peptide fractionation and longer chromatographic gradients. Subsequently, treated samples were analysed using short (21-min/60 samples per day) chromatographic gradients with DIA methods, and raw data were searched against the hybrid reference library. Data analysis was performed by comparison of the peptide intensities in treated and control samples, along with calculation of key statistics and data filtering (as described in the experimental methods) to identify the most robust and selective interactions.

HT-LFQ chemoproteomics yields deep, reproducible peptide detection

To probe the reproducibility and coverage of our platform, we prepared 16 control (DMSO) samples from HEK293T and Jurkat lysates (32 samples in total). From these samples, we identified ~35,000 cysteine-containing peptides using 21-min chromatographic gradients (Fig. 1c and Supplementary Fig. 1a). On average, ~23,000 cysteine-containing peptides were detected in each replicate, which matches the detection depth previously achieved with IA-DTB probes using 3-h gradients and isotopic labelling strategies (Supplementary Fig. 1b)^18,32. Within each cell line, we observed high data completeness, with a median overlap of 82% between the peptides detected in any two replicates, and with two-thirds of all peptides being detected in over 75% of samples (Fig. 1d and Supplementary Fig. 1c, d). This high level of data completeness is a key advantage of label-free/DIA over TMT/DDA proteomics, and is essential for competitive profiling where inconsistent peptide detection can hinder comparison between control and treatment conditions²⁹. Furthermore, excellent reproducibility of the peptide intensities was observed, with a median Pearson correlation between replicates of 0.96 (Supplementary Fig. 1e), and a median coefficient of variation of 24.8% (Supplementary Fig. 1f, g). Comparison of the data acquired using HEK293T and Jurkat lysates revealed that the majority of detected peptides were observed in both cell lines, however inclusion of both cell lines allowed us to increase cysteine coverage by 10–15%, highlighting the potential to expand cysteinome coverage using cells of different biological origin (Supplementary Fig. 2).

Features of the detected cysteines and proteins

The ~32,000 cysteine sites detected by our HT-LFQ platform come from over 8000 proteins, representing ~40% of the proteins in the human proteome and 12% of all cysteine residues in the proteome. Various features of these detected cysteines and proteins were evaluated to assess their biological relevance and to rationalise factors that might affect detectability.

We rationalised that protein abundance and solvent accessibility may impact cysteine coverage. Protein abundance was approximated by the detection of proteins in global proteomics analysis of HEK293T cells, which revealed 80% of IA-DTB detected peptides arise from these abundant proteins, and cover ~30% of all cysteines in these proteins. Other features that are likely to lead to non-detection of cysteines include peptide physicochemical properties, involvement in disulfide bonds, and post-translational modifications (Fig. 1e and Supplementary Fig. 3)³³. We additionally observed ~5000 cysteine sites from proteins we do not detect via global proteomics, likely due to the reduced complexity of chemoproteomics samples following the enrichment step.

The solvent accessibility of the cysteine residues detected in our platform was evaluated using previously reported prediction-aware part-sphere exposure (pPSE) values as a measure of solvent exposure³⁴. We observed that the distribution of pPSE values for cysteines detected by our platform matches the distribution of the whole cysteinome (Supplementary Fig. 4), confirming that the probe is not biased towards the modification of highly exposed cysteine residues, consistent with previous analyses of IA-DTB coverage³². This is an important observation, as many functionally important cysteine residues lie within pockets or regions that are not fully solvent accessible. Engagement of more buried cysteine residues is expected due to protein dynamics, and, in some instances, modification of a single cysteine residue may lead to partial protein unfolding that increases the accessibility of additional residues.

To evaluate our coverage of the proteome with respect to protein function and prior knowledge of tractability, we referenced protein annotations from the ‘Illuminating the Druggable Genome’ initiative, which classifies proteins into one of four target development levels: ‘Tclin’, proteins that are already the targets of approved drugs; ‘Tchem’, proteins that have known small molecule ligands; ‘Tbio’, proteins that lack chemical tools but have well-studied biology; and ‘Tdark’, proteins for which very little information is known^3,7,35. Using this protein classification, we detect ~7000 proteins in the Tbio and Tdark categories, highlighting the opportunity to identify probes for previously unliganded proteins (Fig. 1f). Furthermore, we see good coverage of proteins from families of strong interest to pharmaceutical drug development, such as kinases, as well as proteins from underrepresented families, such as transcription factors and epigenetic proteins³⁶. Similarly, the families detected reveal good representation from nuclear proteins and enzymes. The observed under-representation of membrane-bound proteins (e.g., ion channels and GPCRs) is expected given their low solubility under the cell lysis conditions employed. Taken together, this analysis confirms that our HT-LFQ chemoproteomics platform offers the opportunity for largely unbiased profiling of a significant proportion of the proteome, including many proteins that currently lack chemical tools or small molecule drugs.

Reactive fragment screening by HT-LFQ chemoproteomics

To test the applicability of the HT-LFQ platform for library screening, 80 chloroacetamide-functionalised fragments were screened against both the HEK293T and Jurkat proteomes. The library was designed to cover diverse, fragment-like chemical space (molecular weights 160–320 Da; hydrogen bond donors/acceptors ≤ 3), and included a range of physicochemical properties and molecular shapes (Fig. 2a, Supplementary Fig. 5 and Supplementary Table 1)^37,38. We also included a degree of compound similarity to allow interrogation of structure-activity relationships.

**Fig. 2: The 80-compound screen performed in HEK293T and Jurkat lysate using HT-LFQ chemoproteomics.**

The 80-member library was screened in both HEK293T and Jurkat lysates at 50 μM following incubation for 1 h. Peptide intensities measured in fragment-treated samples (n = 4) were compared to the DMSO control samples (n = 16) and reported as competition ratios (\({\rm{CR}}={{\rm{Intensity}}}_{{\rm{DMSO}}}/{{\rm{Intensity}}}_{{\rm{compound}}}\)). In total, this resulted in a dataset of almost 5 million CRs and associated p-values, describing the interaction of the 80 fragments with the >30,000 cysteine-containing peptides detected across the two cell lines (Fig. 2b and Supplementary Fig. 6a). To focus on the most robust liganding events in each screen, we performed strict peptide filtering and have defined liganded peptides as those with statistically significant competition of at least 50% (log₂(CR) ≥ 1.0, -log₁₀(p-value) ≥ 1.3). The filters we have applied are described in detail in the experimental methods and the results on peptide numbers in Supplementary Table 2. From these 80 compounds, we detected a total of 742 unique liganding events for 438 cysteine sites from 413 proteins (Supplementary Data 1). On average, five liganded cysteines were identified per compound (Supplementary Fig. 6b). Six compounds (PP23, PP219, PP57, PP225, PP147, PP207) showed increased reactivity in both HEK293T and Jurkat lysate, each liganding at least 35 cysteine sites. These compounds are expected to be more promiscuous due to the presence of certain functional groups for example, electron withdrawing anilines (e.g., PP219) and α-substituents that disrupt the planarity of the amide bond (e.g., PP207).

We identified a number of ligands for proteins that already have chemical probes, however, the majority of these proteins (>80%) currently lack chemical matter to probe cellular function (i.e., Tbio or Tdark target development level) (Fig. 2c). The proteins liganded also come from a number of protein families of therapeutic value. We ligand 22 kinases; as a protein family, kinases have drawn significant interest for the development of targeted covalent inhibitors capable of disrupting cell signalling pathways (Fig. 2c)^39,40. In addition to kinases, we liganded over 130 other enzymes, including 17 E3 ligases; these interactions could prompt the development of novel hetero-bifunctional molecules capable of inducing protein proximity between an E3 ligase and a protein of interest for targeted protein degradation^41,42.

Liganded cysteine sites were further classified according to their presence in protein pockets. To determine if a cysteine residue is located within a compound-accessible pocket, we applied the program Fpocket to the predicted monomer structure of the respective protein, as obtained from AlphaFold Database v4^43,44,45. Across the proteome, 40% of cysteines are located in (or within 1.5 Å of) a protein pocket (Fig. 2d). When applied to our liganded sites, the percentage of cysteine residues located in pockets increases to 49%, which is further increased to 55% when considering only liganded sites of high occupancy (log₂(CR) ≥ 2.0). This enrichment indicates that non-covalent fragment recognition is a key contributing factor to protein engagement. However, the observation of liganding events in regions that lack detectable pockets in their structure, or in regions of protein disorder, highlights the ability of covalent ligands to target regions of proteins that have been traditionally challenging to target with non-covalent molecules.

The strength and selectivity of the strongest interactions we detected is summarised by the heatmaps in Fig. 2e (HEK293T) and Supplementary Fig. 6c (Jurkat). Several cysteine sites were strongly liganded by multiple fragments, indicating enhanced reactivity and/or high ligandability of these residues. Indeed, the four most frequently liganded cysteine sites are nucleophilic active site residues – ACAT1 Cys126, ALDH6A1 Cys317, NIT1 Cys203 and NIT2 Cys153 (liganded by 12–40% of the fragments) (Fig. 2f)^46,47. Such interactions highlight tractable opportunities for covalent inhibition of enzyme activity. While these reactive cysteine residues are engaged by many compounds, some compounds show high selectivity for these sites, including PP187 for ACAT1 and PP173 for ALDH6A1 (indicated by arrows in Fig. 2e; associated volcano plots shown in Supplementary Fig. 6d) highlighting the opportunity to develop selective ligands for highly reactive residues.

To investigate trends in the observed compound-protein pairings, hierarchical clustering was performed, grouping compounds based on their molecular fingerprints (Morgan) and proteins based on their competition ratios across the library⁴⁸. This clustering approach highlighted similar binding profiles between structurally similar compounds and proteins. Notably, NIT1 and NIT2 proteins showed near-identical binding profiles to the fragment library, in particular to hindered tertiary chloroacetamides, consistent with the homology of the active sites of these proteins (Fig. 2g). The selectivity of this enzyme family for these compounds demonstrates the potential to identify molecular chemotypes that can be developed towards activity-based protein probes⁴⁷.

Of particular interest were instances where specific interactions were observed between non-hyperreactive cysteines and non-promiscuous fragments⁴⁹. We determined the selectivity of each detected interaction by calculating the difference between the CR for a given interaction and the mean of the five strongest interactions that were detected both for that cysteine site and compound, respectively. By this metric, the top five most selective interactions that were detected in both HEK293T and Jurkat lysate were: MOB4 (Cys134) with PP48, MKLN1 (Cys82) with PP156, VCP (Cys522) with PP183, TPMT (Cys70) with PP222, and the active site residue of GSTO1 (Cys32) with PP1 (highlighted by boxes in the heatmaps in Fig. 2e; volcano plots shown in Fig. 3)⁵⁰. The interaction between GSTO1 Cys32 and PP1 was deprioritised for follow up, as it has been liganded by covalent fragments in multiple other datasets⁵⁰.

**Fig. 3: Top four specific protein-fragment interactions detected in the initial screen.**

Interestingly, previously reported TPMT and VCP inhibitors bear strong resemblance to the ligands we have identified. Thiopurine S-methyltransferase (TPMT) activity is inhibited by a range of non-covalent benzoic acid derivatives (e.g., 5-aminosalicylic acid, 5-ASA), and our hit fragment, PP222, contains the same benzoic acid core (Fig. 3a)⁵¹. The targeted cysteine, Cys70, lies within a buried substrate-binding cavity of this enzyme, with a high density of basic residues nearby that can engage acidic small molecules. VCP is a homohexameric ATPase that has been targeted for the treatment of multiple diseases including acute myeloid leukemia. Covalent inhibitors have been identified that target Cys522, which lies in one of the nucleotide binding pockets, to inhibit enzyme activity and cell growth^52,53,54. One of these inhibitors, FL-18, bears structural similarity to the 5-6-fused heterocyclic ring system of PP183 (Fig. 3b).

Both MOB4 and MKLN1 belong to the Tbio target development category and therefore have no known chemical tools to probe cellular function. The globular adaptor protein, MOB4, is a key component of STRIPAK (striatin-interacting phosphatase and kinase) complexes that play key roles in regulating diverse cellular functions, including cell cycle control and motility^55,56. MOB4 performs a scaffolding function in STRIPAK complexes, and Cys134 lies adjacent to two protein-protein interaction interfaces (Fig. 3c). Therefore, PP48 offers a route to a tool molecule to modulate complex formation or dissociation, and thus STRIPAK function^57,58. MKLN1 (muskelin) is part of the CTLH complex, a multi-subunit RING E3 ubiquitin ligase⁵⁹. PP156 binds to Cys82 which lies adjacent to a proposed self-association interface in muskelin (Fig. 3d) and could be used to probe the functional relevance of this interface⁶⁰. Intriguingly, neither the MOB4 or MKLN1 cysteine sites are located in pockets according to Fpocket analysis of monomeric protein structures, and thus these ligands may bind to pockets formed in multimeric protein complexes.

Proteome-wide concentration-response analysis

The throughput of the HT-LFQ chemoproteomics platform enables fragment screening at multiple concentrations to accurately assess the potency and selectivity of liganding events. We selected eight compounds for screening in HEK293T cell lysate via 10-point concentration-response (0.4–200 μM, quadruplicate measurements). Four of these compounds were those that were identified to form highly selective interactions in the initial screen: PP183 (VCP), PP222 (TPMT), PP156 (MKLN1), and PP48 (MOB4). This compound set was supplemented with four additional fragments that varied in their overall promiscuity: PP207 (high promiscuity), PP156 (medium promiscuity), PP152 (low promiscuity), and PP216 (low promiscuity) (Fig. 4a).

**Fig. 4: Concentration-response chemoproteomics experiment.**

From this experiment, we identified 761 liganding events at the highest concentration of 200 μM, tenfold greater than the number of interactions detected at 50 μM (81 liganding events) (Fig. 4b). The overall promiscuity of the compounds screened broadly matched the initial screen, with a significant range in the number of cysteine sites engaged by different compounds. The high promiscuity of compounds such as PP207 (which bound 426 cysteine sites at 200 μM) highlight its utility as a ‘scout-like’ fragment to identify cysteine residues across the proteome amenable to covalent modification⁶¹. For the lowly reactive compounds PP152 (39 sites) and PP216 (11 sites), screening at a high concentration can be used to identify selective interactions that can be optimised. These data also illustrate how the screening concentration used can affect the apparent selectivity of a compound (e.g., PP207 at 100 µM and 6.25 µM) (Fig. 4c), highlighting the value of methods that offer sufficient throughput and flexibility in experimental design to allow screening at multiple concentrations.

Screening at multiple concentrations allows for curve fitting to obtain half-maximal target engagement (TE₅₀) values for each fragment-cysteine interaction. After filtering, based on potency and the quality of curve fitting, a heatmap was generated to enable visualisation of concentration-dependent binding events, which showed strong agreement with the interactions detected in the initial single-shot screen (Fig. 4d, e and Supplementary Fig. 7). The four compound-protein pairs of interest were confirmed with the following pTE₅₀ (i.e., -log₁₀(TE₅₀)) values: MOB4-PP48 pTE₅₀ = 5.4 ± 0.1, VCP-PP183 pTE₅₀ = 4.9 ± 0.1, TPMT-PP222 pTE₅₀ = 5.7 ± 0.4, and MKLN1-PP156 pTE₅₀ = 5.2 ± 0.2. None of these four cysteine sites were liganded by any other fragment at any concentration, highlighting the specificity of the cysteine site for the respective fragment (Fig. 5a), and no other peptides from these proteins showed a concentration-dependent change in intensity upon fragment treatment (Supplementary Fig. 8). Each compound had no more than three off-targets within ΔpTE₅₀ ≤ 0.5, including sites commonly bound by chloroacetamide fragments, such as ALDH6A1 Cys317 and ATP6V1A Cys138 (Fig. 5b)⁵⁰.

**Fig. 5: The selectivity and potency of prioritised protein-fragment interactions.**

HT-LFQ chemoproteomics for hit expansion and live cell treatments

PP48, which bound to MOB4 Cys134, as well as to the active site cysteines NIT1 Cys203 and NIT2 Cys153, was selected for hit expansion to explore how our chemoproteomics platform could be used to drive structure-activity relationships (SAR). The non-covalent core of PP48 contains three molecular features: a diazepane ring, an amide linker, and a substituted aromatic ring. Binding profiles from structurally-similar compounds that were tested in the initial library screen suggested an important role for the terminal aromatic ring in MOB4 binding (Fig. 2g). To further explore SAR around PP48, we designed seven additional analogues (Supplementary Fig. 9a), varying the nature of the linker, the substituents on the aromatic system, and the structure of the diazepane ring. Each analogue (PP48a-g) was screened in five-point concentration-response (3–50 μM) in HEK293T lysate.

This screen highlighted divergent SAR and opportunities to drive compound selectivity. We identified several useful control compounds: PP48a, PP48d and PP48e (Fig. 5c and Supplementary Fig. 9b). Compared to PP48, PP48a (chloro to methoxy) and PP48e (amide to urea linker and loss of aromatic substituent) both showed similar engagement to the primary off-targets NIT1 and NIT2, but showed either reduced or completely abolished binding to MOB4. Conversely, for PP48d, where the amide carbonyl is removed, engagement of NIT1 and NIT2 is significantly reduced compared to MOB4. These compounds could therefore act as useful controls in deconvoluting the effects of on-target and off-target binding in functional assays.

The aforementioned screening and concentration-response experiments were all performed in cell lysates to maximise sample throughput. However, any hits identified through such screens have most applicability in functional and phenotypic experiments performed in live cells. Therefore, live HEK293T cells were treated with PP48, PP48a, PP48d and PP48e at 25 µM for 2 h, followed by cell lysis, IA-DTB treatment, and quantification of modified peptides (Fig. 5c). Importantly, we confirmed the binding of PP48 to MOB4 Cys134, NIT1 Cys203 and NIT2 Cys153, with additional off-targets also identified (e.g., SORD Cys45), potentially due to the increased incubation time employed for the live cell treatment (1 h vs 2 h) or increased temperature (room temperature vs 37 °C) (Supplementary Fig. 9c). The selectivity profile of the control compounds, PP48a, PP48d and PP48e, also showed concordance between lysate and cell treatment. These results highlight that profiling of compounds in lysates is an effective method to improve throughput and simplify workflows while being an effective surrogate for measurement of interactions formed in live cells⁶².

Discussion

Identifying chemical tools for the unliganded proteome is essential to accelerate the exploration of protein function and the discovery of therapeutic opportunities. Various initiatives, such as ‘Target 2035’, have been established with the aim to identify pharmacological modulators for every protein in the human proteome^7,35. Achieving this ambition requires the development of high-throughput platforms to screen molecules in complex cellular environments and identify novel protein-ligand interactions. Platforms that combine competitive profiling of covalent fragments with chemoproteomics have drawn interest from academia and industry, allowing screens to be performed in cells and lysates, and providing a quantitative readout of fragment-protein engagement.

We have developed a label-free chemoproteomics platform for screening reactive fragments across the proteome by competition with a hyper-reactive IA-DTB probe. We have moved away from recently reported DDA and TMT labelling-based platforms to overcome limitations of these methods associated with incomplete datasets and batch effects. By employing LFQ and DIA, our platform offers high analytical throughput, deep cysteinome coverage, and high sample reproducibility and data completeness, whilst avoiding the need for costly isotopic labelling reagents.

Our LFQ based chemoproteomics platform allows for comparison of peptide quantities between an unlimited number of samples. This offers excellent flexibility in experimental design, facilitating screening of large libraries at multiple concentrations. Performing DIA ensures every peptide in the sample is fragmented regardless of abundance, maintaining data reproducibility and completeness. Furthermore, an extra level of peptide separation was performed via trapped ion-mobility using PASEF (on a Bruker timsTOF Pro 2), which provides unrivalled cysteinome depth for DIA-based cysteine profiling (~23,000 cysteines identified per run; ~30,000 per experiment)²⁹. Finally, diaPASEF enabled short (21-min) chromatographic gradients without compromising identification depth³¹. A remaining limitation in chemoproteomics studies is the overall cysteinome coverage. Our HT-LFQ chemoproteomics platform achieves coverage of ~40% of the proteome but less than 15% of the entire cysteinome. From our analysis, protein abundance, lysis conditions and peptide properties were key factors in determining cysteine detection, and so we anticipate that performing screens in different cell lines with varied protein expression profiles maybe beneficial in improving cysteine coverage.

We evaluated the platform by profiling a reactive fragment library of 80 chloroacetamides in quadruplicate in two cell lines, producing a robust dataset of liganding interactions across the cysteinome. The screen identified 438 liganded cysteine sites from 413 proteins, including over 300 proteins from the Tbio and Tdark target development categories, for which no chemical tools exist^3,7,35. While examination of the location of liganded cysteines highlighted an enrichment of cysteines near pockets, many liganded cysteines were in regions predicted to be disordered and many others may lie adjacent to pockets that only develop upon the formation of protein complexes⁶³. Together, these observations highlight the value of screening compounds against proteins in an endogenous setting, and the potential for covalent compounds to ligand proteins previously considered to be undruggable. The majority of the interactions displayed good selectivity (with each compound engaging only ~5 cysteines, on average, at 50 μM), representing promising starting points for the development of chemical probes. Furthermore, screening of hit fragments across a concentration gradient produced rich datasets of concentration-response curves for every detected cysteine residue, allowing for prioritisation of compounds for further development based on potency and proteome selectivity.

The throughput and flexibility of the platform facilitates the exploration of structure-activity relationships by chemoproteomics. We explored the selectivity and potency of structural analogues of PP48, which engaged the adaptor protein MOB4 as well as cysteine residues in NIT1 and NIT2. SAR analogues showed differential binding profiles, highlighting opportunities to improve selectivity for MOB4 or NIT1/2. While the majority of screening was performed in cell lysates to facilitate plate-based workflows, a number of SAR compounds were profiled in live cells, validating the interactions detected in lysate-based experiments. The simultaneous quantification of cellular on-target potency as well as proteome-wide off-target binding is a key benefit for chemoproteomics screening of covalent compounds and has the potential to streamline early drug discovery efforts. This information is particularly challenging to obtain when screening non-covalent compounds in cellular assays or screening against purified proteins.

Looking forward, we anticipate application of this robust screening platform to profile larger compound libraries against the native proteome. The resulting full matrix datasets will offer opportunities for machine learning approaches to predict ligandability and drive iterative library design towards selective chemical probes, and thus expand the liganded proteome.

Methods

Compounds

All chloroacetamide fragments were purchased from Enamine (catalogue numbers provided in Supplementary Table 1) and stored either as solids at −20 °C or as DMSO stocks at −80 °C. Iodoacetamide desthiobiotin (IA-DTB) was synthesised according to literature reports¹⁸. The product was purified via silica gel chromatography (ISCO 80 g RediSep Gold column, 0–20% methanol in dichloromethane, 60 ml/min flow rate, dry loaded as silica gel powder). Fresh stock was prepared from solid at a concentration of 50 mM in DMSO and used immediately.

Cell culture

HEK293T cells were maintained at 37 °C, 5% CO₂ in Dulbecco’s modified eagle medium (Gibco, 41966-029) supplemented with 10% (v/v) fetal bovine serum (Gibco, 10270-106) and 1% v/v penicillin–streptomycin-glutamine (Gibco, 10-378-016). Jurkat cells were maintained at 37 °C, 5% CO₂ in RPMI-1640 GlutaMAX (Sigma Aldrich, R8758) supplemented with 10% v/v fetal bovine serum (Gibco, 10270-106) and 1% v/v penicillin-streptomycin (Sigma Aldrich, P4333).

Preparation of chemoproteomics samples

Lysate compound treatment

Cell pellets (pre-washed twice with PBS) were suspended in RIPA lysis buffer (150 mM NaCl, 1.0% IGEPAL CA-630 (Sigma Aldrich, I8896), 0.5% sodium deoxycholate (Sigma Aldrich, D6750), 0.1% SDS (Fisher Scientific, 10607633), 50 mM HEPES, pH 8.0) containing protease inhibitor cocktail (Sigma Aldrich, P1860). Samples were sonicated with a Branson probe sonicator (3–5 × 2 s pulses, 10% amplitude) and filtered through a 0.22 µM filter. The protein concentration of the lysate was determined using a Pierce™ BCA assay kit (Thermo Scientific, 23227) according to manufacturer’s instructions. Lysate was diluted to the desired concentration and used on the day of lysis.

For cell lysate treatment, compounds were first prepared at an appropriate concentration in DMSO (2 μL) in a 96-well plate. To this, cell lysate (200 or 500 µg protein) was added to each well to reach a final volume of 200 μL and the plates were then shaken on a Thermomixer (room temperature (rt), 600 rpm, 1 h). Experiments were performed with technical replicates, where multiple identical samples were prepared in parallel.

Live cell compound treatment

HEK293T cells were cultured in 10 cm dishes to 90% confluency. Each plate was treated with a compound (25 μM, final DMSO concentration of 0.4% v/v) or DMSO alone for 2 h at 37 °C. Media was removed and cells were washed three times with PBS. Cell lysis and protein quantification methods were performed as reported above, with samples clarified by centrifugation (16,000 rcf, 10 min, 4 °C) instead of filtration. Replicate samples were prepared from cells that had been grown separately for one doubling time. Compound treatment and subsequent sample preparation on these replicate samples was performed in parallel.

IA-DTB treatment

IA-DTB (500 µM; fresh DMSO stock) was added to wells containing compound-treated cell lysates on 96-well plates (rt, 600 rpm, 1 h; performed in the dark). Following treatment, samples were reduced with dithiothreitol (Thermo Scientific, R0861) (5 mM, rt, 600 rpm, 30 min) and alkylated with iodoacetamide (Thermo Scientific, 122270250) (10 mM, rt, 600 rpm, 30 min; performed in the dark).

Glass bead slurry preparation

To prepare stock glass bead mixtures, glass spheres (Supelco, 440345) were first suspended at a concentration of 100 mg/mL in ultrapure water (LC/MS grade)³⁰. The resulting slurry was then vortexed and centrifuged (1500 rcf, 5 min, 4 °C), and the buoyant beads were gently aspirated to leave a glass bead pellet. This process was repeated twice with ultrapure water and twice with acetonitrile. After the final wash, the bead pellet was resuspended in acetonitrile (making the solution up the same volume as the original 100 mg/mL solution in water), and a final concentration of 50 mg/mL was assumed, as half of the beads are typically lost during the washing procedure. The beads were stored at 4 °C until needed.

Glass bead assisted sample clean up and digestion

Glass beads from the pre-prepared stock solution were diluted in acetonitrile (6.25 mg/mL) and this bead slurry was dispensed into MultiScreen deep filter plates (Merck Millipore, MDRLN0410) (800 µL/well). IA-DTB-treated cell lysates were transferred to the bead slurry and agitated (600 rpm, 5 min), inducing protein precipitation. The plates were centrifuged (1500 rcf, 2 min) to remove the supernatant and washed with 80% v/v ethanol (1500 rcf, 2 min, 3×). Proteins were resolubilised in HEPES (50 mM, pH 8.5, 250 µL/well) containing Pierce™ Trypsin Protease (Thermo Scientific, 90059) (1:100 enzyme/protein ratio) and digested overnight (rt, 800 rpm). Peptides were recovered into collection plates through centrifugation (1500 rcf, 5 min) and subsequent washes of the glass beads with HEPES (50 mM, pH 8.5, 2 × 75 µL/well).

Enrichment

Peptide solutions were transferred to a sealed Microlute plate (Porvair Sciences, 240002). Pierce™ High Capacity NeutrAvidin™ Agarose resin (Thermo Scientific, 29204) was prepared by washing with HEPES (50 mM, pH 8.5, 3×), and then dispensed (50 μL of slurry) into each well of the sealed Microlute plate and agitated with the peptides (800 rpm, 2 h). The drain cap seal was removed and the plate was centrifuged (700 rcf, 1 min) to remove the supernatant. Beads in each well were washed with 0.1% SDS in HEPES (50 mM, pH 8.5; 850 µL/well, 3×), HEPES (50 mM, pH 8.5, 850 µL/well, 3×), and finally ultrapure water (850 µL/well, 3×). Enriched peptides were eluted from the neutravidin resin with 1:1 acetonitrile/water, containing 0.1% formic acid (200 µL/well, 700 rcf, 1 min). This elution step was repeated (100 µL/well, 2×). The collection plate was frozen and samples were dried using a Speedvac at 4 °C. Plates were stored at −80 °C.

Desalting

C18 Nest desalting plates (The Nest Group Inc, HNS S18V) were conditioned with acetonitrile (300 µL/well) and centrifuged (50 rcf, 1 min). Plates were equilibrated with ultrapure water/0.1% trifluoroacetic acid (300 µL/well, 2×) and centrifuged (500 rcf, 5 min). Samples were re-dissolved in ultrapure water/0.1% trifluoroacetic acid (200 µL/well) on a Thermomixer (rt, 600 rpm, 5 min), then loaded into the prepared C18 NEST plate(s) and centrifuged (500 rcf, 5 min). Samples were washed with ultrapure water/0.1% trifluoroacetic acid (200 µL/well, 2×). Peptides were eluted using 1:1 acetonitrile/water with 0.1% trifluroacetic acid (150 µL/well, 2×) by centrifugation (500 rcf, 5 min). The collection plate was frozen and samples were dried using a Speedvac at 4 °C. Plates were stored at −80 °C.

Preparation of global proteomics samples

Sample preparation and peptide recovery

HEK293T cell pellets were lysed in RIPA lysis buffer according to the lysis procedure reported above for chemoproteomics samples. Lysate was prepared at 1 µg/µL and diluted further to 0.5 µg/µL with S-Trap lysis buffer (10% SDS (Fisher Scientific, 10607633), 100 mM triethylammonium bicarbonate (Sigma Aldrich, T7408), TEAB, pH 7.55). Samples were reduced and alkylated as reported above. Samples were then acidified with 10 µL of 12% phosphoric acid (Supelco, PX1000). S-Trap binding buffer (750 µL, 90% methanol, 100 mM, TEAB, pH 7.1) was added to each sample and gently agitated. Acidified mixtures were transferred into wells of a 96-well S-Trap™ plate (Protifi, C02-96well-1) and centrifuged (1500 rcf, 2 min). Captured protein was washed with S-Trap binding buffer (1500 rcf, 2 min, 3×) and digested with S-Trap digestion buffer (125 µL, 50 mM TEAB) containing Pierce™ Trypsin Protease (1:25 enzyme/protein ratio) (47 °C, 1 h). Following digestion, S-Trap digestion buffer (80 µL) was added and peptides were collected into a 96 well plate by centrifugation (1500 rcf, 2 min). Each well was washed with 80 µL of 0.1% aqueous formic acid, followed by 80 µL of 50% aqueous acetonitrile containing 0.1% formic acid. The collection plate was frozen and samples were dried using a Speedvac at 4 °C. Plates were stored at −80 °C.

Liquid chromatography-mass spectrometry

High pH off-line fractionation

To generate a library of IA-DTB modified peptides, high pH off-line fractionation was performed. High pH off-line fractionation was either performed with an XBridge BEH C18 XP column (2.5 µm × 3 mm × 130 mm, Waters, 186006710) coupled to an UltiMate 3000 HPLC system, or with a Pierce™ High pH Reversed-Phase Peptide Fractionation Kit (Thermo Scientific, 84868). For the off-line fractionation on the XBridge BEH C18 column a 60-min acetonitrile gradient (1–35% acetonitrile) was performed at a flow rate of 200 µL/min using the following buffers: 10 mM ammonium hydroxide pH 10 (buffer A), and 90% acetonitrile, 10% 100 mM ammonium hydroxide pH 10 (buffer B). 24 fractions were consolidated and dried using a Speedvac at 4 °C. For the off-line fractionation kit, samples were prepared according to manufacturer’s instructions. These samples were dried using a Speedvac at 4 °C.

Evotip sample loading

Prepared samples were re-dissolved in Optima™ water/0.1% formic acid (100 µL/well) on a Thermomixer (rt, 600 rpm, 5 min). Evotips (Evosep, EV2001 or EV2011) were conditioned according to manufacturer’s instructions. Approximately 200 ng (the desired loading) of each peptide mixture spiked with indexed retention time (iRT) peptides (Biognosys, Ki-3002-1) was loaded onto a conditioned Evotip and queued on an Evosep One liquid chromatography system with the pre-defined 30 and 60 SPD (samples per day) methods and the corresponding Evosep Performance columns (Evosep, EV1109 and EV1137). Mobile phases A and B were 0.1% (v/v) formic acid in water and 0.1% (v/v) formic acid in acetonitrile, respectively.

Mass spectrometry: general

The Evosep One was coupled online to a hybrid (trapped ion mobility spectrometry) TIMS quadrupole TOF (time of flight) mass spectrometer (Bruker timsTOF Pro 2) via a captive spray nano-electrospray ion source. All chemoproteomics samples were analysed with an ion mobility range from 1/K0 = 1.638 to 0.6 Vs cm⁻² and global proteomics samples were analysed with an ion mobility range from 1/K0 = 1.6 to 0.6 Vs cm⁻². Equal ion accumulation time and ramp times were applied in the dual TIMS analyser of 100 ms each. Mass spectra were recorded from 100–1700 m/z. The ion mobility dimension was calibrated regularly using all three ions from an Agilent electrospray ionisation LC/MS tuning mix (m/z, 1/K0: 622.0289, 0.9848 Vs cm⁻²; 922.0097, 1.1895 Vs cm⁻²; and 1221.9906, 1.3820 Vs cm⁻²). When operating the mass spectrometer in ddaPASEF mode, 10 PASEF/MS-MS scans were used per topN acquisition cycle. Singly charged precursors were excluded by their position in the m/z-ion mobility plane, and precursors that reached a target value of 20,000 arbitrary units were dynamically excluded for 0.4 min. When operating the mass spectrometer in diaPASEF mode, 8 diaPASEF scans per TIMS-MS scan were used, giving a duty cycle of 0.96 seconds. For chemoproteomics DIA analysis, variable ion mobility windows were used with fixed mass windows of 25 m/z and with a mass range of 400–1000 m/z (Supplementary Table 3). For global proteomics DIA analysis, recent developments in window optimisation were incorporated to allow for variable ion mobility windows and variable mass windows, between a mass range of 262.18–1398.68 m/z (Supplementary Table 4)⁶⁴.

Chemoproteomics data analysis

Mass spectrometry raw files for chemoproteomics were analysed using Spectronaut (Biognosys; version 16).

Generation of a hybrid library of IA-DTB modified peptides

In total, 72 mass spectrometry files were used to generate a hybrid DDA/DIA library of IA-DTB modified peptides from both HEK293T and Jurkat cell lysates. This library was generated in Spectronaut using the search algorithm Pulsar. Peptide lengths of 7–52 amino acids and with up to two miscleavages were permitted. The following variable modifications were applied: oxidation (methionine, +15.99 Da), IA-DTB (cysteine, +296.18 Da), acetylation (N-terminus, +42.01 Da), and carbamidomethylation (cysteine, +57.02 Da). All searches were performed against three FASTA files that contained the canonical UniProt human protein sequences, common contaminants⁶⁵, and iRT fusion peptides, respectively.

Data analysis for compound screening

For analysis of diaPASEF files, standard Biognosys settings were used with minor modifications. In brief, a precursor Q-value cutoff of <0.01 was used with an experiment wide protein Q-value cutoff of <0.01 and a probability cut off for PTM localisation of >0.75. Subsequent analysis was performed using Python. Quantification was performed on the precursor level and intensities associated with equivalent peptides (i.e., only differed in charge state or methionine oxidation state) were summed together to give a single peptide-level intensity. The ability of a compound to complete with the IA-DTB probe was quantified through competition ratios (\({\rm{CR}}={{\rm{Intensity}}}_{{\rm{DMSO}}}/{{\rm{Intensity}}}_{{\rm{compound}}}\)) and the significance of the difference between control and compound-treatment samples was calculated using Welch’s t-test. The following criteria were used to identify binding events: mean log₂(CR) ≥ 1 and -log₁₀(p-value) ≥ 1.3. In addition, the peptide was required to be robustly detected without any confounding factors: the peptide must have been detected in at least two of the compound-treated replicates and in ≥90% of all samples in the experiment; the peptide must have a coefficient of variation (CV) of ≤40% in the control samples; the peptide must only have a single DTB modification and, if multiple cleavage forms of the same peptide exist in the dataset, then only the most abundant of these peptides is considered.

In cases where data was acquired at multiple compound concentrations, the mean intensity of each peptide (percent relative to the DMSO control) at varying concentrations (log₁₀-transformed) of each compound was fit to a 4-parameter logistic function: \({{\rm{E}}}_{\inf }+\frac{{{\rm{E}}}_{0}-{{\rm{E}}}_{\inf }}{1+{10}^{n}({\log }_{10}({{\rm{TE}}}_{50})-x)}\), where E₀ is the relative peptide intensity when no compound is present (typically the top plateau), E_inf is the relative peptide intensity when infinite compound is present (typically the bottom plateau), n is the slope, and TE₅₀ is the relative peptide intensity at the midpoint of the curve. Fitting was performed in Python using lmfit⁶⁶, with the following bounds: 60 ≤ E₀ ≤ 140; E_inf ≥ 0; −50 ≤ n ≤ 0; and log₁₀(TE₅₀) was varied up to 3 log units outside the concentration range tested. The reported errors for best fit values are estimated standard errors calculated by lmfit.

Global proteomics data analysis

Raw mass spectrometry data files were analysed using Spectronaut (version 18) with directDIA. The following search parameters were used for directDIA: peptide lengths of 7–52 amino acids with up to two miscleavages were permitted, with one fixed modification (carbamidomethylation; cysteine, +57.02 Da) and the following variable modifications: oxidation (methionine, +15.99 Da) and acetylation (N-terminus, +42.01 Da). All searches were performed against three FASTA files that contained the canonical UniProt human protein sequences, common contaminants⁶⁵, and iRT fusion peptides, respectively.

Protein and residue annotations

All proteins in the human proteome and their sequences (‘one sequence per gene’) were obtained from UniProt. The location of cysteine residues within sequences was extracted and annotation of whether these cysteines lie within MS-detectable tryptic sequences was determined through in silico trypsin digestion with the following rules: cleavage after lysine or arginine as long as the next residue is not proline, and permitting peptides with a length of 7–40 amino acids.

Annotations concerning the Illuminating the Druggable Genome (IDG) protein families and target development levels were obtained from the Pharos database. Residue-specific annotations for post-translational modifications and disulfide bonds were obtained from UniProt⁶⁷. Residue-specific pPSE values were obtained from published data³⁴.

Chemical and physical properties of the 80-member chloroacetamide library

Physiochemical properties of the reactive fragments were calculated using LiveDesign (Schrodinger Suite 2023-2). Bemis-Murcko frameworks were assigned manually^68,69.

Principal moment of inertia (PMI) values were calculated using Molecular Operating Environment (MOE: Chemical Computing Group, version: 2019.0101)⁷⁰. A three-dimensional model was first generated for each compound from the SMILES string, by performing a conformational search with the following parameters: force field – MMFF94x; method – stochastic; rejection limit – 200; iteration limit – 10000; amide bond rotation allowed; unconstrained double bond rotation allowed; chair conformations not enforced; not refined with quantum mechanics; root mean squared deviation limit – 0.15; conformation limit – 1. Normalised principal moment of inertia ratios (NPRs) were then calculated from the resulting PMI values.

Molecular similarity was quantified using Morgan fingerprints (radius = 2, bits = 1024) and Tanimoto similarity scores, calculated using RDKit (2022.09.5)⁷¹. Hierarchical clustering of compounds based on molecular similarity was performed using SciPy with the Ward variance minimisation algorithm⁷².

Fpocket analysis

To identify which cysteines are located within ligandable pockets, the program Fpocket was applied on monomeric, three-dimensional protein models, as predicted by AlphaFold2^43,44. Fpocket detects impressions on the protein surface by rolling a series of sphere probes (alpha spheres) with sizes spanning over a specified range of radii. If an alpha sphere touches three atoms of the protein simultaneously, it is placed at that position. By default, a pocket is reported if it contains at least 35 alpha spheres. The range of radii used in our case was 3.0–5.0 Å. We defined a cysteine residue as being located in a pocket if its thiol atom was within 1.5 Å of the closest alpha sphere. The parameters of these calculations were defined empirically, after examining several examples of cysteine liganding events.

Figure preparation

Figure 1a, b was created using image templates from BioRender.com under the institutional license belonging to the Francis Crick Institute (https://BioRender.com/m32r739).

Reporting summary

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

Data availability

The raw mass spectrometry proteomics files and database search results have been deposited at the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository with data set identifiers PXD054105, PXD054127 and PXD054145⁷³. Source Data is provided with this paper as a Source Data file. Source data are provided with this paper.

Change history

23 January 2025
In the version of the article initially published, in the “HT-LFQ chemoproteomics for hit expansion and live cell treatments” section, “diazepane” originally appeared as “diazepine” and has now been corrected in the HTML and PDF versions of the article.

References

Garbaccio, R. M. & Parmee, E. R. The impact of chemical probes in drug discovery: a pharmaceutical industry perspective. Cell Chem. Biol. 23, 10–17 (2016).
PubMed MATH CAS Google Scholar
Müller, S. et al. Target 2035 – update on the quest for a probe for every protein. RSC Med. Chem. 13, 13–21 (2022).
PubMed MATH Google Scholar
Oprea, T. I. et al. Unexplored therapeutic opportunities in the human genome. Nat. Rev. Drug Discov. 17, 317–332 (2018).
PubMed PubMed Central MATH CAS Google Scholar
Blagg, J. & Workman, P. Choose and use your chemical probe wisely to explore cancer biology. Cancer Cell 32, 9–25 (2017).
PubMed PubMed Central MATH CAS Google Scholar
Arrowsmith, C. H. et al. The promise and peril of chemical probes. Nat. Chem. Biol. 11, 536–541 (2015).
PubMed PubMed Central MATH CAS Google Scholar
Forrest, I. & Parker, C. G. Proteome-wide fragment-based ligand and target discovery. Isr. J. Chem. 63, 1–13 (2023).
MATH Google Scholar
Nguyen, D. T. et al. Pharos: collating protein information to shed light on the druggable genome. Nucleic Acids Res. 45, D995–D1002 (2017).
PubMed CAS Google Scholar
Johansson, H. et al. Fragment-based covalent ligand screening enables rapid discovery of inhibitors for the RBR E3 ubiquitin ligase HOIP. J. Am. Chem. Soc. 141, 2703–2712 (2019).
PubMed PubMed Central MATH CAS Google Scholar
Boike, L., Henning, N. J. & Nomura, D. K. Advances in covalent drug discovery. Nat. Rev. Drug Discov. 21, 881–898 (2022).
PubMed PubMed Central MATH CAS Google Scholar
McCarthy, W. J., van der Zouwen, A. J., Bush, J. T. & Rittinger, K. Covalent fragment-based drug discovery for target tractability. Curr. Opin. Struct. Biol. 86, 102809 (2024).
PubMed CAS Google Scholar
Erlanson, D. A., Fesik, S. W., Hubbard, R. E., Jahnke, W. & Jhoti, H. Twenty years on: the impact of fragments on drug discovery. Nat. Rev. Drug Discov. 15, 605–619 (2016).
PubMed CAS Google Scholar
Grant, E. K. et al. A photoaffinity-based fragment-screening platform for efficient identification of protein ligands. Angew. Chem. Int. Ed. 59, 21096–21105 (2020).
MATH CAS Google Scholar
Aatkar, A. et al. Efficient ligand discovery using sulfur(VI) fluoride reactive fragments. ACS Chem. Biol. 18, 1926–1937 (2023).
PubMed PubMed Central MATH CAS Google Scholar
Bachovchin, D. A. & Cravatt, B. F. The pharmacological landscape and therapeutic potential of serine hydrolases. Nat. Rev. Drug Discov. 11, 52–68 (2012).
PubMed PubMed Central CAS Google Scholar
Cookson, R. et al. A chemoproteomic platform for selective deubiquitinase inhibitor discovery. Cell Rep. Phys. Sci. 4, 101636 (2023).
MATH CAS Google Scholar
Chan, W. C. et al. Accelerating inhibitor discovery for deubiquitinating enzymes. Nat. Commun. 14, 1–13 (2023).
ADS MATH CAS Google Scholar
Hewings, D. S. et al. Reactive-site-centric chemoproteomics identifies a distinct class of deubiquitinase enzymes. Nat. Commun. 9, 1162 (2018).
ADS PubMed PubMed Central MATH Google Scholar
Kuljanin, M. et al. Reimagining high-throughput profiling of reactive cysteines for cell-based screening of large electrophile libraries. Nat. Biotechnol. 39, 630–641 (2021).
ADS PubMed PubMed Central CAS Google Scholar
Yan, T. et al. SP3‐faims chemoproteomics for high‐coverage profiling of the human cysteinome. ChemBioChem 22, 1841–1851 (2021).
PubMed PubMed Central MATH CAS Google Scholar
Vinogradova, E. V. et al. An activity-guided map of electrophile-cysteine interactions in primary human T Cells. Cell 182, 1009–1026.e29 (2020).
PubMed PubMed Central MATH CAS Google Scholar
Backus, K. M. et al. Proteome-wide covalent ligand discovery in native biological systems. Nature 534, 570–574 (2016).
ADS PubMed PubMed Central MATH CAS Google Scholar
Maurais, A. J. & Weerapana, E. Reactive-cysteine profiling for drug discovery. Curr. Opin. Chem. Biol. 50, 29–36 (2019).
PubMed PubMed Central MATH CAS Google Scholar
Wu, S. et al. Cysteinome: the first comprehensive database for proteins with targetable cysteine and their covalent inhibitors. Biochem. Biophys. Res. Commun. 478, 1268–1273 (2016).
PubMed MATH CAS Google Scholar
Pappireddi, N., Martin, L. & Wühr, M. A review on quantitative multiplexed proteomics. ChemBioChem 20, 1210–1224 (2019).
PubMed PubMed Central MATH CAS Google Scholar
O’Connell, J. D., Paulo, J. A., O’Brien, J. J. & Gygi, S. P. Proteome-wide evaluation of two common protein quantification methods. J. Proteome Res. 17, 1934–1942 (2018).
PubMed PubMed Central MATH Google Scholar
Li, J. et al. TMTpro reagents: a set of isobaric labeling mass tags enables simultaneous proteome-wide measurements across 16 samples. Nat. Methods 17, 399–404 (2020).
PubMed PubMed Central MATH CAS Google Scholar
Krasny, L. & Huang, P. H. Data-independent acquisition mass spectrometry (DIA-MS) for proteomic applications in oncology. Mol. Omics 17, 29–42 (2021).
PubMed MATH CAS Google Scholar
Guzman, U. H. et al. Ultra-fast label-free quantification and comprehensive proteome coverage with narrow-window data-independent acquisition. Nat. Biotechnol. https://doi.org/10.1038/s41587-023-02099-7 (2024).
Yang, F., Jia, G., Guo, J., Liu, Y. & Wang, C. Quantitative chemoproteomic profiling with data-independent acquisition-based mass spectrometry. J. Am. Chem. Soc. 144, 901–911 (2022).
PubMed MATH CAS Google Scholar
Johnston, H. E. et al. Solvent precipitation SP3 (SP4) enhances recovery for proteomics sample preparation without magnetic beads. Anal. Chem. 94, 10320–10328 (2022).
PubMed PubMed Central MATH CAS Google Scholar
Meier, F. et al. diaPASEF: parallel accumulation–serial fragmentation combined with data-independent acquisition. Nat. Methods 17, 1229–1236 (2020).
PubMed MATH CAS Google Scholar
Yang, K. et al. Accelerating multiplexed profiling of protein-ligand interactions: High-throughput plate-based reactive cysteine profiling with minimal input. Cell Chem. Biol. 31, 565–576.e4 (2023).
PubMed MATH Google Scholar
Li, Y. F., Arnold, R. J., Tang, H. & Radivojac, P. The importance of peptide detectability for protein identification, quantification, and experiment design in MS/MS proteomics. J. Proteome Res. 9, 6288–6297 (2010).
PubMed PubMed Central MATH CAS Google Scholar
White, M. E. H., Gil, J. & Tate, E. W. Proteome-wide structural analysis identifies warhead- and coverage-specific biases in cysteine-focused chemoproteomics. Cell Chem. Biol. 30, 828–838.e4 (2023).
PubMed CAS Google Scholar
Kelleher, K. J. et al. Pharos 2023: an integrated resource for the understudied human proteome. Nucleic Acids Res. 51, D1405–D1416 (2023).
PubMed MATH Google Scholar
Reinecke, M. et al. Chemical proteomics reveals the target landscape of 1000 kinase inhibitors. Nat. Chem. Biol. 20, 577–585 (2024).
PubMed MATH CAS Google Scholar
Keeley, A., Petri, L., Ábrányi-Balogh, P. & Keserű, G. M. Covalent fragment libraries in drug discovery. Drug Discov. Today 25, 983–996 (2020).
PubMed CAS Google Scholar
Keserű, G. M. et al. Design principles for fragment libraries: maximizing the value of learnings from pharma fragment-based drug discovery (FBDD) programs for use in academia. J. Med. Chem. 59, 8189–8206 (2016).
PubMed MATH Google Scholar
Lu, X., Smaill, J. B., Patterson, A. V. & Ding, K. Discovery of cysteine-targeting covalent protein kinase inhibitors. J. Med. Chem. 65, 58–83 (2022).
PubMed CAS Google Scholar
Pan, Z. et al. Discovery of selective irreversible inhibitors for Bruton’s tyrosine kinase. ChemMedChem 2, 58–61 (2007).
PubMed MATH CAS Google Scholar
Grimster, N. P. Covalent PROTACs: the best of both worlds? RSC Med. Chem. 12, 1452–1458 (2021).
PubMed PubMed Central MATH CAS Google Scholar
Kennedy, C., McPhie, K. & Rittinger, K. Targeting the ubiquitin system by fragment-based drug discovery. Front. Mol. Biosci. 9, 1019636 (2022).
Le Guilloux, V., Schmidtke, P. & Tuffery, P. Fpocket: an open source platform for ligand pocket detection. BMC Bioinformatics 10, 168 (2009).
PubMed PubMed Central MATH Google Scholar
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
ADS PubMed PubMed Central MATH CAS Google Scholar
Varadi, M. et al. AlphaFold protein structure database: massively expanding the structural coverage of protein-sequence space with high-accuracy models. Nucleic Acids Res. 50, D439–D444 (2022).
PubMed MATH CAS Google Scholar
Haapalainen, A. M. et al. Crystallographic and kinetic studies of human mitochondrial acetoacetyl-CoA thiolase: the importance of potassium and chloride ions for its structure and function. Biochemistry 46, 4305–4321 (2007).
PubMed MATH CAS Google Scholar
Peracchi, A. et al. Nit1 is a metabolite repair enzyme that hydrolyzes deaminated glutathione. Proc. Natl. Acad. Sci. USA 114, E3233–E3242 (2017).
PubMed PubMed Central CAS Google Scholar
Rogers, D. & Hahn, M. Extended-connectivity fingerprints. J. Chem. Inf. Model. 50, 742–754 (2010).
PubMed MATH CAS Google Scholar
Hallenbeck, K. K., Turner, D. M., Renslo, A. R. & Arkin, M. R. Targeting non-catalytic cysteine residues through structure-guided drug discovery. Curr. Top. Med. Chem. 17, 4–15 (2017).
PubMed PubMed Central CAS Google Scholar
Boatner, L. M., Palafox, M. F., Schweppe, D. K. & Backus, K. M. CysDB: a human cysteine database based on experimental quantitative chemoproteomics. Cell Chem. Biol. 30, 683–698.e3 (2023).
PubMed PubMed Central CAS Google Scholar
Szumlanski, C. & Weinshilboum, R. Sulphasalazine inhibition of thiopurine methyltransferase: possible mechanism for interaction with 6‐mercaptopurine and azathioprine. Br. J. Clin. Pharmacol. 39, 456–459 (1995).
PubMed PubMed Central CAS Google Scholar
Magnaghi, P. et al. Covalent and allosteric inhibitors of the ATPase VCP/p97 induce cancer cell death. Nat. Chem. Biol. 9, 548–559 (2013).
PubMed MATH CAS Google Scholar
Ding, R. et al. Discovery of Irreversible p97 Inhibitors. J. Med. Chem. 62, 2814–2829 (2019).
PubMed MATH CAS Google Scholar
Ye, Z. et al. A targeted covalent inhibitor of p97 with proteome-wide selectivity. Acta Pharm. Sin. B 12, 982–989 (2022).
PubMed MATH CAS Google Scholar
Shi, Z., Jiao, S. & Zhou, Z. STRIPAK complexes in cell signaling and cancer. Oncogene 35, 4549–4557 (2016).
PubMed MATH CAS Google Scholar
Duhart, J. C. & Raftery, L. A. Mob family proteins: regulatory partners in hippo and hippo-like intracellular signaling pathways. Front. Cell Dev. Biol. 8, 1–22 (2020).
MATH Google Scholar
Sijbesma, E. et al. Site-directed fragment-based screening for the discovery of protein-protein interaction stabilizers. J. Am. Chem. Soc. 141, 3524–3531 (2019).
PubMed MATH CAS Google Scholar
Cawood, E. E. et al. Modulation of amyloidogenic protein self-assembly using tethered small molecules. J. Am. Chem. Soc. 142, 20845–20854 (2020).
PubMed PubMed Central MATH CAS Google Scholar
Maitland, M. E. R., Lajoie, G. A., Shaw, G. S. & Schild-Poulter, C. Structural and functional insights into GID/CTLH E3 ligase complexes. Int. J. Mol. Sci. 23, 5863 (2022).
Delto, C. F. et al. The LisH motif of muskelin is crucial for oligomerization and governs intracellular localization. Structure 23, 364–373 (2015).
PubMed CAS Google Scholar
Bar-Peled, L. et al. Chemical proteomics identifies druggable vulnerabilities in a genetically defined cancer. Cell 171, 696–709.e23 (2017).
PubMed PubMed Central MATH CAS Google Scholar
Baltgalvis, K. A. et al. Chemoproteomic discovery of a covalent allosteric inhibitor of WRN helicase. Nature 629, 435–442 (2024).
ADS PubMed CAS Google Scholar
Ogasawara, D. et al. Chemical tools to expand the ligandable proteome: diversity-oriented synthesis-based photoreactive stereoprobes. Cell Chem. Biol. https://doi.org/10.1016/j.chembiol.2024.10.005 (2024).
Skowronek, P. et al. Rapid and in-depth coverage of the (Phospho-)Proteome with deep libraries and optimal window design for dia-PASEF. Mol. Cell. Proteomics 21, 100279 (2022).
PubMed PubMed Central MATH CAS Google Scholar
Frankenfield, A. M., Ni, J., Ahmed, M. & Hao, L. Protein contaminants matter: building universal protein contaminant libraries for DDA and DIA proteomics. J. Proteome Res. 21, 2104–2113 (2022).
PubMed PubMed Central CAS Google Scholar
Newville, M., Stensitzki, T., Allen, D. B. & Ingargiola, A. LMFIT: non-linear least-square minimization and curve-fitting for Python. Preprint at https://doi.org/10.5281/zenodo.11813 (2015).
The Uniprot Consortium. UniProt: the universal protein knowledgebase in 2023. Nucleic Acids Res. 51, D523–D531 (2023).
Google Scholar
Bemis, G. W. & Murcko, M. A. The properties of known drugs. 1. molecular frameworks. J. Med. Chem. 39, 2887–2893 (1996).
PubMed MATH CAS Google Scholar
Giordanetto, F., Jin, C., Willmore, L., Feher, M. & Shaw, D. E. Fragment hits: what do they look like and how do they bind? J. Med. Chem. 62, 3381–3394 (2019).
PubMed PubMed Central CAS Google Scholar
Wirth, M. & Sauer, W. H. B. Bioactive molecules: perfectly shaped for their target? Mol. Inform. 30, 677–688 (2011).
PubMed MATH CAS Google Scholar
RDKit: open-source cheminformatics. Preprint at https://doi.org/10.5281/zenodo.7671152.
Virtanen, P. et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nat. Methods 17, 261–272 (2020).
PubMed PubMed Central MATH CAS Google Scholar
Perez-Riverol, Y. et al. The PRIDE database resources in 2022: a hub for mass spectrometry-based proteomics evidences. Nucleic Acids Res. 50, D543–D552 (2022).
PubMed CAS Google Scholar
Wu, H. et al. Structural basis of allele variation of human thiopurine-S-methyltransferase. Proteins 67, 198–208 (2007).
PubMed PubMed Central MATH CAS Google Scholar
Banerjee, S. et al. 2.3 Å resolution cryo-EM structure of human p97 and mechanism of allosteric inhibition. Science 351, 871–875 (2016).
ADS PubMed PubMed Central MATH CAS Google Scholar
Jeong, B. -C. et al. Cryo-EM structure of the Hippo signaling integrator human STRIPAK. Nat. Struct. Mol. Biol. 28, 290–299 (2021).
PubMed PubMed Central MATH CAS Google Scholar

Download references

Acknowledgements

We thank Joanna Kirkpatrick, Toby Baker, and both the Cell Services STP and the Proteomics STP at the Francis Crick Institute for support with this manuscript. We are grateful to Prof. Nick Tomkinson and the University of Strathclyde for their enthusiastic support of a secondment for Harry Wilders into the Crick-GSK LinkLabs during the course of his PhD. This work was supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK (CC2075, CC2000) to K.R. and S.B., the UK Medical Research Council (CC2075, CC2000) to K.R. and S.B., and the Wellcome Trust (CC2075, CC2000) to K.R. and S.B. This project has been funded by a Prosperity Partnership grant from the Engineering and Physical Sciences Research Council (EPSRC), EP/V038028/1 to A.P., D.H., S.B., J.M.S., K.R. and J.B. We also thank GSK for its commitment to support fundamental discovery research through the initial establishment of the Crick-GSK LinkLabs partnership and its contribution to the EPSRC Prosperity Partnership grant.

Funding

Open Access funding provided by The Francis Crick Institute.

Author information

These authors contributed equally: George S. Biggs, Emma E. Cawood.

Authors and Affiliations

Crick-GSK Biomedical LinkLabs, GSK, Gunnels Wood Road, Stevenage, Hertfordshire, UK
George S. Biggs, Emma E. Cawood, Aini Vuorinen, Harry Wilders, Ioannis G. Riziotis, Jonathan Pettinger, Andrew J. Powell, David House & Jacob T. Bush
Molecular Structure of Cell Signalling Laboratory, The Francis Crick Institute, London, UK
George S. Biggs, Aini Vuorinen, William J. McCarthy, Antonie J. van der Zouwen & Katrin Rittinger
Proteomics Science Technology Platform, The Francis Crick Institute, London, UK
George S. Biggs, Aini Vuorinen & J. Mark Skehel
DSB Repair Metabolism Laboratory, The Francis Crick Institute, London, UK
Emma E. Cawood & Simon J. Boulton
University of Strathclyde, Pure and Applied Chemistry, Glasgow, UK
Harry Wilders
Software Engineering and AI, The Francis Crick Institute, London, UK
Ioannis G. Riziotis & Luke Nightingale
GSK Chemical Biology, GSK, Collegeville, PA, USA
Peiling Chen

Authors

George S. Biggs
View author publications
Search author on:PubMed Google Scholar
Emma E. Cawood
View author publications
Search author on:PubMed Google Scholar
Aini Vuorinen
View author publications
Search author on:PubMed Google Scholar
William J. McCarthy
View author publications
Search author on:PubMed Google Scholar
Harry Wilders
View author publications
Search author on:PubMed Google Scholar
Ioannis G. Riziotis
View author publications
Search author on:PubMed Google Scholar
Antonie J. van der Zouwen
View author publications
Search author on:PubMed Google Scholar
Jonathan Pettinger
View author publications
Search author on:PubMed Google Scholar
Luke Nightingale
View author publications
Search author on:PubMed Google Scholar
Peiling Chen
View author publications
Search author on:PubMed Google Scholar
Andrew J. Powell
View author publications
Search author on:PubMed Google Scholar
David House
View author publications
Search author on:PubMed Google Scholar
Simon J. Boulton
View author publications
Search author on:PubMed Google Scholar
J. Mark Skehel
View author publications
Search author on:PubMed Google Scholar
Katrin Rittinger
View author publications
Search author on:PubMed Google Scholar
Jacob T. Bush
View author publications
Search author on:PubMed Google Scholar

Contributions

G.B. carried out method development, proteomics experiments, and initial data analysis; E.C. carried out all downstream data analysis and visualisation; A.V. assisted in initial method development and sample preparation; W.M.C., H.W., A.v.d.Z. and J.P. assisted in compound selection and annotation; I.R. performed Fpocket analysis; L.N. developed additional scripts for data analysis; P.C. synthesised chemical probes; A.P., D.H., S.B and J.M.S. assisted in project supervision; G.B., E.C., K.R. and J.B. designed the project and wrote the paper with input from all authors. All authors have seen and approved the manuscript.

Corresponding authors

Correspondence to Katrin Rittinger or Jacob T. Bush.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Description of Additional Supplementary Information

Supplementary Data 1

Reporting Summary

Transparent Peer Review file

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Biggs, G.S., Cawood, E.E., Vuorinen, A. et al. Robust proteome profiling of cysteine-reactive fragments using label-free chemoproteomics. Nat Commun 16, 73 (2025). https://doi.org/10.1038/s41467-024-55057-5

Download citation

Received: 26 July 2024
Accepted: 28 November 2024
Published: 02 January 2025
DOI: https://doi.org/10.1038/s41467-024-55057-5

This article is cited by

Enantioselective OTUD7B fragment discovery through chemoproteomics screening and high-throughput optimisation
- Aini Vuorinen
- Cassandra R. Kennedy
- Katrin Rittinger
Communications Chemistry (2025)