AzidoTMT Enables Direct Enrichment and Highly Multiplexed Quantitation of Proteome-Wide Functional Residues

Abstract

Recent advances in targeted covalent inhibitors have aroused significant interest for their potential in drug development for difficult therapeutic targets. Proteome-wide profiling of functional residues is an integral step of covalent drug discovery aimed at defining actionable sites and evaluating compound selectivity in cells. A classical workflow for this purpose is called IsoTOP-ABPP, which employs an activity-based probe and two isotopically labeled azide-TEV-biotin tags to mark, enrich, and quantify proteome from two samples. Here we report a novel isobaric 11plex-AzidoTMT reagent and a new workflow, named AT-MAPP, that significantly expands multiplexing power as compared to the original isoTOP-ABPP. We demonstrate its application in identifying cysteine on- and off-targets using a KRAS G12C covalent inhibitor ARS-1620. However, changes in some of these hits can be explained by modulation at the protein and post-translational levels. Thus, it would be crucial to interrogate site-level bona fide changes in concurrence to proteome-level changes for corroboration. In addition, we perform a multiplexed covalent fragment screening using four acrylamide-based compounds as a proof-of-concept. This study identifies a diverse set of liganded cysteine residues in a compound-dependent manner with an average hit rate of 0.07% in intact cell. Lastly, we screened 20 sulfonyl fluoride-based compounds to demonstrate that the AT-MAPP assay is flexible for noncysteine functional residues such as tyrosine and lysine. Overall, we envision that 11plex-AzidoTMT will be a useful addition to the current toolbox for activity-based protein profiling and covalent drug development.

Introduction

There is a growing interest in developing targeted covalent inhibitors (TCIs) for proteins that are difficult to target by conventional drug formats. This has been partially stimulated by published studies showing that a vast number of nucleophilic residues in the proteome can be liganded by covalent fragments or structurally elaborated compounds.¹⁻⁴ In these studies, high-fidelity interactions between covalent compounds and target proteins are achieved because reactions are carried out in the native cellular environment. However, deconvoluting small-molecule–protein interactions among the highly complex cellular proteome remains challenging. A number of chemical proteomic approaches have been developed to identify proteins that interact with a small molecule.⁵⁻¹⁴ In particular, proteome-wide activity-based target profiling using a broadly reactive probe such as iodoacetamide alkyne, combined with competitive labeling by compounds of interest, has become a predominant choice for covalent inhibitors. Covalent modification of target sites by the compounds results in subsequently reduced labeling by the reactive probe and loss of probe signals measured by mass spectrometry. This provides an indirect evidence of covalent compound labeling at specific residues on targets within the proteome.

Classic activity-based probes are applied using a workflow that integrates three key elements: a probe containing a covalent reactive group, an affinity handle for enrichment of probe-labeled proteins, and a quantitative element. A classical workflow with these components is called isoTOP-ABPP (isotopic Tandem Orthogonal Proteolysis-Activity-Based Protein Profiling, Figure 1A)¹⁵ The biorthogonal copper(I)-catalyzed alkyne–azide cycloaddition (CuAAC) or “click reaction”¹⁶ is integrated in the workflow to couple the activity-based chemical probe (alkyne-tagged) with a biotin affinity enrichment handle (azide-tagged). The clickable probe makes it versatile for multiple types of residue targets, including cysteine, lysine, and tyrosine, by simply swapping the probe for the desired target specificity.^4,17,15 This method has been successfully applied to profile cellular targets for covalent inhibitors¹⁸⁻²² and to monitor posttranslational modifications.^23,24,3 The quantitative element in isoTOP-ABPP is based on heavy and light isotopically labeled TEV tags and introduced during click reaction immediately after probe labeling. This allows heavy and light clicked samples to be mixed as a pair and processed concurrently to minimize sample-to-sample variation. However, this limits the method to only a binary comparison. For a study consisting of multiple conditions and replicates, several isoTOP-ABPP experiments are needed. This poses a challenge in experimental throughput and introduces variability across larger groups of experimental conditions due, in part, to missing values from one or more conditions in the aggregated results. Alternatively, two multiplexed workflows using desthiobiotin iodoacetamide and tandem mass tags (TMT), named TMT-ABPP and SLC-TMT (Figure 1A), have recently been reported to study proteome-wide ligandable cysteines and cysteine covalent inhibitors.^25,26 TMT employs a set of isobaric tags to label and quantify up to 11 or 18 (TMTpro version) samples via unique reporter ions produced during peptide fragmentation in a mass spectrometer.²⁷ Unlike isoTOP-ABPP, TMT labeling in the SLC-TMT type workflow is applied to peptides at the later stage after enrichment, which introduces a source of variation during sample preparation such as protein precipitation, protein digestion, etc., and a cost in throughput.

Figure 1.

Synthesis and characterization of 11plex-AzidoTMT. (A) A comparison of key components (reactive probe, affinity handle, and quantitative element) in isoTOP-ABPP, TMT-ABPP, and AT-MAPP workflows. (B) One-step synthesis of 11plex-AzidoTMT by mixing 11plex-TMT with an excess of 3-azido-1-propanamine for an hour at room temperature. Each 11plex-AzidoTMT carries 13C and 15N stable isotopes, which are fully elaborated in Figure S1A. (C) AzidoTMT is purified and confirmed by high-resolution mass spectrometry based on intact molecular weight. (D) Determination of AzidoTMT:Alkyne reaction stoichiometry by titrating AzidoTMT against a fixed amount of iodoacetamide-alkyne and monitoring the triazole product after CuAAC reaction. (E) A general scheme of AT-MAPP workflow based on AzidoTM. A red asterisk indicates a reactive warhead.

IsoTOP-ABPP and SLC-TMT approaches both have their unique advantages and limitations. We consider an ideal workflow as being flexible to various probes, having multiplexing capability, and possessing the ability to adhere quantitation tags to target sites in the earliest stage of the workflow. This leads us to keep the core concept of isoTOP-ABPP but seek an alternative azide that can carry highly multiplexed isobaric tags (Figure 1A). Amine-reactive TMT is an activated NHS ester that can be easily coupled to primary amines. We expect that a compound with both amino and azido functional groups can be readily transformed to a multiplexed reagent. Based on this, we developed a novel 11-plex azide-functionalized TMT reagent (AzidoTMT, Figure 1B and Figure S1A) through a one-pot high-yield reaction using commercially available starting materials. AzidoTMT can be coupled to any alkyne-derivatized chemical probe via click reaction so that the quantitative element is directly applied to probe-labeled sites just after cell lysis. After combining and digesting all labeled samples, peptides of interest can be enriched via anti-TMT immunoaffinity precipitation, identified, and quantified in the mass spectrometer. This workflow is termed AT-MAPP (AzidoTMT-based Multiplexed Activity-based Protein Profiling), which differentiates from isoTOP-ABPP by the number of multiplexing and differentiates from SLC-TMT by the labeling strategy. We applied this, together with an iodoacetamide-alkyne (IA-Alkyne) probe, to profile proteome-wide selectivity of ARS-1620, a KRAS G12C covalent inhibitor, and to screen protein targets liganded by covalent fragment compounds in cells. In addition, we demonstrate the versatility of this reagent to study other reactive residues when a sulfonyl fluoride probe DAS1 is used instead of IA-Alkyne.

Experimental Procedures

AzidoTMT Synthesis, Purification, and Characterization

Each TMT tag from a 10plex-TMT set (126–131N) in 0.8 mg unit size was dissolved in 40 μL of anhydrous acetonitrile immediately before use. TMT-131C in 5 mg unit size was dissolved in 250 μL of anhydrous acetonitrile, and a 40-μL aliquot was used. Twenty microliters of 1 mol/L 3-azido-1-propanamine in DMSO stock was added to each TMT solution and incubated for 1 h at room temperature. The reaction was quenched by adding 3 mL of 0.1% trifluoroacetic acid (TFA) in water. AzidoTMT was purified by a C18 column with the following steps: (1) Load crude product on 3 cm³ C18 Sep-Pak columns (Waters, Milford, MA). (2) Wash twice with 4% acetonitrile and 0.1% TFA in water, 1.5 mL each. (3) Elute twice with 25% acetonitrile and 0.1% TFA in water, 1.5 mL each. An equal but small aliquot (0.5%) of Sep-Pak eluents from each AzidoTMT tag was mixed, diluted, and acidified with 0.1% TFA and analyzed in an LTQ Orbitrap Fusion Tribid mass spectrometer (Thermo Fisher Scientific, San Jose, CA) with the following parameters: nanoLC at 0.5 μL/min over an Acquity UPLC BEH130 C18 column (1.7 μm, 130 Å, 100 μm × 100 mm); LC gradient from 0 to 40% solvent B (2% water/98% ACN/0.1% FA) in 7 min. Full MS and MS/MS spectra were acquired in the Orbitrap at 60 000 and 50 000 resolution, respectively. Precursor fragmentation was performed using HCD at 55% normalization collision energy. TMT reporter ions derived from the precursor 330.2432 Da were extracted and quantified.

The remaining AzidoTMT eluents were lyophilized and reconstituted in DMSO with volume adjustments according to their respective TMT reporter ion intensities. A small aliquot of the stock from each tag was mixed, diluted, and acidified with 0.1% TFA and directly infused in an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific, San Jose, CA). Full MS was acquired from 100 to 500 Da mass range in the Orbitrap at 15 000 resolution for peak confirmation. To determine AzidoTMT concentration, a fixed amount of IA-Alkyne (100 μL of 100 μM stock in PBS buffer (pH = 7.4)) was titrated with 11plex-AzidoTMT (0, 0.25, 0.5, 1.0, 1.0, 1.5, 1.5, 2.0, 2.0, 2.5, and 3.0 μL of stocks from AzidoTMT-126 to AzidoTMT-131C, respectively). Briefly, to each of the 11 aliquots of IA-Alkyne, the following reagents were added in the stated order: one of the AzidoTMT tags in the above-mentioned volume, premixed BTTAA and CuSO₄ solution (50× aqueous stock; final concentration 2 and 1 mM, respectively), sodium ascorbate (100× aqueous stock; final concentration 2 mM). All aliquots were incubated in RT for 60 min and then quenched by EDTA (2 mM final) before being combined into one tube. The mixture was diluted and acidified with 0.1% TFA and analyzed in the LTQ Orbitrap Fusion Tribid mass spectrometer as described. TMT reporter ions derived from the precursor 298.1240 Da (observed mass corresponds to doubly charged IAalkyne-AzidoTMT click reaction product) were extracted and quantified.

Cell Preparation, Compound Treatment, and Reactive Probe Labeling

HCC1171 cells were grown in 3D ultralow attachment (ULA) plates (Corning, #3814) under standard culture conditions (RPMI medium supplemented with 2% glutamine and 10% FBS) at 40K cells/mL. Cells were grown overnight and treated the next day with ARS-1620 at 0.1, 1, and 10 μM or with DMSO for 4 h at 37 °C. After treatment, cells were harvested by scraping and washed by PBS buffer before lysis. To each 10 million cells, 0.3 mL of chilled PBS (pH = 7.4) was added, followed by probe sonication for 30 s. Lysate was cleared by centrifugation at 20 000g for 5 min at 4 °C. Protein concentration was measured by Bradford assay and adjusted to 2 mg/mL. Each sample was further labeled by 100 μM IA-Alkyne for 1 h in dark.

A2058 cells were maintained at 37 °C and 5% (v/v) CO₂ in Dulbecco’s modified Eagle’s medium with 10% (v/v) fetal bovine serum and 4 mM l-glutamine. Cells were either treated with compound A1, A2, A3, or A4 at 500 μM for 2 h (in-cell samples) or lysed in PBS with probe sonication, adjusted to 2 mg/mL, and incubated with compound A1, A2, A3, or A4 at 500 μM for 1 h (in-lysate samples). IA-Alkyne probe labeling was performed as described above. For the cysteine thiol reactivity profiling, untreated A2058 cells were lysed in PBS and titrated with DMSO and 10 concentrations of iodoacetamide ranging from 0.051 to 1000 μM in threefold steps for 30 min incubation in the dark. Upon completion, each sample was further labeled by 100 μM IA-Alkyne for 1 h in the dark.

For a sulfonyl fluoride-based probe and sulfonyl fluoride fragments (S1–S20), 2 mg/mL A2058 protein lysates were aliquoted and treated with 100 μM of final stock of DMSO or 100 μM of one of the 20 sulfonyl fragments followed by 1 h incubation in the dark. Afterward, the DAS1 probe was added at a 100 μM final concentration and incubated for 1 h at room temperature in the dark. Excess probe was further quenched with a 150 μM final concentration of l-lysine monohydrochloride.

AzidoTMT Click Reaction, Protein Digestion, Fractionation, and Enrichment

To each of the 11 probe-labeled protein lysates in an experiment set, one unique AzidoTMT tag was added based on the labeling scheme described in Supplemental Table 1. Click reaction was carried out by adding premixed BTTAA and CuSO₄ solution (final concentration 2 and 1 mM, respectively) and sodium ascorbate (final concentration 2 mM) to the lysates, vortexed, and incubated for 1 h at room temperature. Upon completion, reactions were quenched with 2 mM EDTA, and all samples were combined. Methanol/chloroform precipitation was used to remove excessive reagent. In brief, to 1 vol of combined sample, 4 vol of methanol, 1 vol of chloroform, and 3 vol of water were added sequentially with vortexing between each addition. The mixture was centrifuged at 20 000g for 2 min, and the top liquid layer was carefully removed. An additional 4 vol of methanol was added, vortexed, and centrifuged again at 20 000g for 2 min. All liquid was carefully removed from the precipitated protein pellet and discarded. The pellet was air-dried for 30 min and immediately resuspended in 1 mL of freshly prepared 8 M urea in 20 mM HEPES buffer (pH = 8.0) with brief sonication.

Resuspended samples were reduced with 4.5 mM dithiothreitol for 20 min at 60 °C and alkylated with 11 mM iodoacetamide for 15 min in the dark. Samples were diluted with 3 mL of 20 mM HEPES (pH = 8.0). Trypsin (1:40 enzyme-to-substrate ratio) was added and incubated at room temperature overnight with end-to-end rotation. The next day, extra trypsin (1:200 enzyme-to-substrate ratio) was added to digest for an additional 2 h. Digests were then acidified with 20% TFA at a 1:20 (v/v) ratio and centrifuged to remove any debris. Cleared digests were desalted using a 3 cm³ Sep-Pak column containing 200 mg of C18 absorbent (Waters, Milford, MA). Dried and resuspended eluents were fractionated using offline high-pH reversed phase as previously described.⁶ Forty-eight fractions were collected evenly over the elution profile. Every sixth fraction was pooled into one, which resulted in a set of six distinct groups after pooling.

Pooled fractions were dried and reconstituted in TBS (25 mM Tris, 150 mM NaCl, pH 7.2) for anti-TMT immunoaffinity enrichment. For each fraction, a 200 μL slurry of Immobilized Anti-TMT Antibody Resin was used. The resin was prewashed according to the manufacture’s protocol, followed by incubating with peptides for 2 h at room temperature with gentle shaking. Additional washing steps were applied in the following order: 100 μL of 2 M urea in TBS (repeat four times), 100 μL of TBS (repeat four times), and 100 μL of dd water (repeat three times). Peptides were eluted from the resin using two rounds of 200 μL of 50% acetonitrile with 0.5% TFA for 10 and 1 min each. Eluents were combined, dried, and further cleaned by a C18 STAGE tip before LC-MS/MS analysis.

Proteome-wide Cysteine Analysis using IodoTMT and SLC-TMT

IodoTMT and SLC-TMT workflows were performed as described previously.^25,28 Briefly, HCC1171 lysates at 2 mg/mL in PBS were alkylated with 100 μM IodoTMT or DBIA in the dark for 1 h. To quench excess reagent, DTT was added and allowed to incubate at 60 °C for 20 min. Then 11 mM IAM was added with additional 20 min incubation in the dark. For IodoTMT samples, samples were combined and protein precipitation by methanol chloroform precipitation was performed before digestion with Promega trypsin at a 1:40 ratio. Peptides were desalted and subjected to anti-TMT immunoaffinity enrichment, staged tip clean up, and injection onto the mass spectrometer. For SLC-TMT, methanol chloroform precipitation, trypsin digestion, and TMT labeling were performed before samples were combined. Afterward, streptavidin enrichment was performed with 50% prepared slurry rotating at 4 °C for 2 h. Multiple washes were performed with PBS, 0.1% SDS PBS, and HPLC grade water followed by elution with 50% ACN/0.1% TFA twice. Samples were dried in a speedvac, cleaned by C18 stage tip, and ready for LC-MS³ injection.

LC-MS/MS Sample Analysis, Peptide Identification, and Quantification

Data were collected on Orbitrap Fusion Tribrid Mass Spectrometers (Thermo Fisher Scientific, San Jose, CA). Peptides were loaded onto a New Objective PicoFrit Acquity BEH130 Å C18 column (1.7 μM, 100 μM × 250 mm) in Solvent A (98% water, 2% acetonitrile, 0.1% formic acid) with a flow rate of 0.7 μL/min in Solvent A (98% water/2% acetonitrile/0.1% formic acid), separated at a flow rate of 0.5 μL/min with a linear gradient of 2–35% solvent B (98% acetonitrile/2% water/0.1% formic acid) over 158 min, and sprayed into the mass spectrometer via a Nanospray Flex-Ion Source (Thermo Scientific) at a voltage of 1.9 kV. Full MS scans were collected at 120 000 resolution in the Orbitrap across a range from 350 to 1600 m/z, with an automatic gain control (AGC) target of 2 × 10⁵ and a maximum injection time of 50 ms. The 10 most abundant MS2 ions were selected using 0.7 Da isolation width and an AGC target of 5 × 10³ with a maximum injection time of 100 ms. Ions were fragmented with a CID energy of 35 and analyzed in an ion trap. MS3 spectra were acquired in the Orbitrap by isolating eight MS2 fragment ions in synchronized precursor selection (SPS) mode with a higher collision dissociation (HCD) fragmentation energy setting of 55, an AGC of 1 × 10⁵, a maximum injection time of 150 ms, an isolation width of 1.4 Da, and a resolution of 50 000.

MS/MS spectra were searched using Comet (2017.01 rev. 1) against UniProt DB (2017_08) prefiltered with taxonomy “9606”, appended with common contaminating proteins, and concatenated by all decoy sequences. Search parameters included trypsin cleavage with an allowance of up to two missed cleavage events, a precursor ion tolerance of 50 ppm, and a fragment ion tolerance of 1.0005 Da. Searches included variable modifications of methionine oxidation (+15.9949 Da) and static carbamidomethylation modification of cysteine (+57.0215 Da). For IA-Alkyne modification, additional variable modifications of cysteine probe adduct (+466.3219 Da) were included, and for DAS1 modification, additional variable modification of lysine/arginine/serine/threonine/tyrosine probe adduct (+606.3157 Da) was included. Peptide spectra matches (PSMs) were filtered with a false discovery rate (FDR) of 1% at the peptide level using linear discrimination²⁹ and filtered again for all probe-modified sequences. Modification sites were localized using the Ascore algorithm.³⁰ Due to many possible modification residues, an additional Ascore filter of greater than or equal to 15 was applied to the DAS1 set to only retain confidently localized sites (>90% success rate). TMT reporter ions produced by the AzidoTMT tags were quantified with an in-house software package known as Mojave³¹ by calculating the highest peak within 20 ppm of theoretical reporter mass windows and correcting for isotope purities. Quantified PSMs were filtered by total TMT reporter ion intensity greater than 30 000 and isolation specificity greater than 0.5. All raw mass spectrometric data files have been deposited in MassIVE under the following accession number: MSV000086546.

Global Protein Expression Profiling for the ARS-1620 Treated Samples

Lysate proteins were reduced with 5 mM DTT and alkylated with 15 mM iodoacetamide. Approximately 75 μg from each sample was precipitated using methanol/chloroform. The precipitated protein was washed two times with 100% methanol. Precipitated protein was digested with LysC (Wako Chemicals USA, Richmond, VA) at a 1:25 protease-to-protein ratio at 25 °C for 12 h. Following the LysC digestion, trypsin (Promega, Madison, WI) was added at a 1:50 protease-to-protein ratio. Trypsin digestion was carried out at 37 °C for 8 h. The peptides generated from the protease digestion were labeled in 500 mM EPPS pH 8.0 with TMT 11-plex reagent (ThermoFisher Scientific, Waltham, MA) at a 8:1 TMT-to-peptide (mg/mg) ratio, incubating at 25 °C for 3 h. TMT reactions were quenched with 0.5% hydroxylamine, and samples were acidified with TFA and combined into a “final mix”. The final mix peptides were desalted on 50 mg Waters tC18 SepPak cartridges (Waters Corporation, Milford, MA) and dried by centrifugal evaporation. Approximately 115 μg of peptide mix was subjected to orthogonal basic-pH reverse phase fractionation on a 3 × 100 mm column packed with 1.9 μm Poroshell C18 material (Agilent, Santa Clara, CA), utilizing a 45 min linear gradient from 8% buffer A (5% acetonitrile in 10 mM ammonium bicarbonate, pH 8) to 35% buffer B (acetonitrile in 10 mM ammonium bicarbonate, pH 8) at a flow rate of 0.4 mL/min. Ninety-six fractions were consolidated into 24 samples, acidified with formic acid, and vacuum-dried. The samples were resuspended in 5% formic acid, desalted on StageTips packed with Empore C18 material (3M, Maplewood, MN), and vacuum-dried. Peptides were reconstituted in 5% formic acid, 5% acetonitrile for LC-MS/MS/MS analysis. All mass spectra were acquired on an Orbitrap Fusion Lumos coupled to an EASY nanoLC-1200 (ThermoFisher) liquid chromatography system. Approximately 2 μg of peptides was loaded on a 75 μm capillary column packed in-house with Sepax GP-C18 resin (1.8 μm, 150 Å, Sepax) to a final length of 35 cm. Peptides for total protein analysis were separated using a 90 min linear gradient from 5% to 22% acetonitrile in 0.1% formic acid. The mass spectrometer was operated in a data-dependent mode. The scan sequence began with FTMS1 spectra (resolution = 120 000; mass range of 350–1400 m/z; max injection time of 50 ms; AGC target of 1e6; dynamic exclusion for 60 s with a ±10 ppm window). The 10 most intense precursor ions were selected for ITMS2 analysis via collisional-induced dissociation (CID) in the ion trap (normalized collision energy (NCE) = 35; max injection time = 35 ms; isolation window of 0.7 Da; AGC target of 8e3). Following ITMS2 acquisition, a synchronous-precursor-selection (SPS) MS3 method was enabled to select eight MS2 product ions for high-energy collisional-induced dissociation (HCD) with analysis in the Orbitrap (NCE = 55; resolution = 50 000; max injection time = 86 ms; AGC target of 8e4; isolation window at 1.2 Da for +2 m/z, 1.0 Da for +3 m/z, or 0.8 Da for +4 to +6 m/z).

Data Analysis

For each PSM, probe-modified residue numbers were extracted from the primary protein reference within UniProt. Quantified PSMs were normalized based on global median abundance among all PSMs and then summarized to the site level using MSstats_3.14.1.³² Fold changes were determined among predefined treatment groups, and p-values were computed in MSstats as previously described. p-Values were further adjusted for multiple comparisons using the Benjamini–Hochberg method. The dose-dependent alkylation curve and EC50 of ARS-1620 were computed in Prism 8 using a variable-slope least-squares fit for the log(inhibitor) vs normalized response. The Wilcoxon rank sum test was performed with two-sided comparison using the built-in package “stats” in R (3.5.0). Iodoacetamide titration curve fitting was performed in R (3.5.0) using the Self-Starting Nls Four-Parameter Logistic Model (SSfpl function from the default “stats” package). To correct for systematic variation prior to modeling, the data set was normalized by a sigmoid curve fitted from median abundances at each distinct concentration. No attempt at curve fitting was made if the abundance in the highest concentration was >50% of the abundance in the DMSO control. If a sigmoid curve was modeled, an RC50 was calculated based on the curve to compute IAA concentration corresponding to 50% abundance level versus DMSO control. Visual interactive data analyses were performed using TIBCO Spotfire v 7.8 and can be accessed via these links for the studies described in the manuscript: ARS1620: https://sld-acs-sf.sf.perkinelmercloud.com/spotfire/wp/analysis?file=/Guest/1-ATMAPP-ARS1620&elqTrackId=69315abac08e47c2874bee1870c436b3&elqaid=9716&elqat=2; covalent fragments: https://sld-acs-sf.sf.perkinelmercloud.com/spotfire/wp/analysis?file=/Guest/2-ATMAPP-Fragments&elqTrackId=bcdece13b2fc4146b0023d38b8ac6f1b&elqaid=9716&elqat=2.

Covalent Docking of the Sulfonyl Fluoride Fragment to NOQ1

In silico simulation of a fragment docked onto a protein through covalent bonding was performed using the ADFRSuite program³³ (version 1.0rc1) according to the developers’ instructions. The three-dimensional structure of the desired compound was prepared in Avogadro³⁴ (version 1.2.0) from its SMILES string. The structure of target protein NQO1 was downloaded from the RCSB PDB database (PDB id: 1d4a). The ligand and protein were then modified to make a docking point at residue Y128 (equivalent to Y129 in the Uniprot sequence). Docking was simulated in a grid box of 18 × 18 × 18 Å centered on the geometric center of the residue with 50 independent searches each using up to 2 500 000 evaluations of the scoring function. The output ligand file was analyzed with the target protein in Pymol (version 2.5.1).

Results

Design, Synthesis, and Characterization of 11plex-AzidoTMT

We envision that a multiplexed azido derivative may be used to enable hypermultiplexing with click reactions. Commercially available TMT reagents consist of an amine-reactive N-hydroxysuccinimide (NHS) ester group, a mass balancer group, and a mass reporter. While the mass balancer and mass reporter groups are critical for quantitation, the amine-reactive group is amenable for further derivatization. We identified a small bifunctional compound, 3-azido-1-propanamine (APA), with a primary amino group for NHS coupling and an azido group for click reaction. The synthetic reaction was carried out in microscale. Individual units (0.8 mg) of commercially available 11-plexed TMT were mixed with a molar excess of APA reagent for 1 h to yield 11-plexed AzidoTMT (Figure 1B). Crude product was purified on a C18 Sep-pak column to remove excess starting material and byproducts (Figure S1B). Characterization of AzidoTMT focused primarily on mass measurement and relative quantification by high-resolution mass spectrometry. The desired product was confirmed by directly infusing into a mass spectrometer Figure 1C) with the resulting reagent mass matching the theoretical mass with 0.6 mDa mass error (1.8 ppm). Due to the low starting quantity and spontaneous hydrolysis of the NHS ester group during storage and reaction, the yield for each of the 11-plexed products was not measured. Instead, the AzidoTMT concentration was adjusted based on the TMT reporter ion signal to normalize abundance among all 11-plex products (Figure S1C). Reaction stoichiometry between AzidoTMT and an alkyne was monitored by titrating AzidoTMT against a fixed amount (100 μL of 100 μM solution) of IA-Alkyne. Upon copper-catalyzed alkyne–azide cycloaddition, AzidoTMT was successfully conjugated to the IA-Alkyne probe as monitored by LC-MS (Figure S1D). The triazole product was quantified by TMT reporter ions fragmented from the conjugated molecule (Figure 1D). At ≥1 μL of supplied AzidoTMT, a plateau was reached, indicating a stoichiometric reaction and maximal yield. When this stoichiometric ratio (i.e., 1 μL of AzidoTMT stock for every 100 μL of 100 μM alkyne probe) was applied to whole cell extracts, 95% of IA-Alkyne probe-labeled peptides were identified as the triazole form after tryptic digestion (data not shown). A possible limitation to click chemistry conjugation is the fragmentation of the azido TMT reagent leading to competition and decrease of the reporter fragmentation. Our analysis of the formation of the oxonium ion indicates there its contribution is <1% of the total MS3 TMT reporter ion signal and MS2 fragment ion signal for the majority of the observed spectra (Figure S1E).³⁵ Next, we integrated AzidoTMT into the classical IsoTOP-ABPP workflow and conceived the AT-MAPP (AzidoTMT-based Multiplexed Activity-based Protein Profiling) workflow as illustrated (Figure 1E). Owing to the direct installation of the TMT moiety to target sites via alkyne probe and click reaction, clicked products can be enriched from the proteome complex using immobilized anti-TMT antibody²⁸ without needing a biotin or desthiobiotin affinity tag. This approach minimizes the overall size of the tag added to peptides. Having synthesized the 11plex-AzidoTMT set, we next sought to demonstrate its utility in chemical proteomic applications.

Benchmarking AzidoTMT with a Previously Reported KRAS G12C Covalent Inhibitor ARS-1620 and Comparison with Other Cysteine-Profiling Workflows

IsoTOP-ABPP and its variants have heretofore been the primary assays used to profile proteome-wide target selectivity for covalent inhibitors.^{36,37,19,38,39} By extension, we posited that AT-MAPP would facilitate compound testing via multiple concentrations, time points, and replicates simultaneously. We applied our approach to profile proteome-wide selectivity of the previously reported covalent inhibitor, ARS-1620.³⁷ We first compared AT-MAPP to other cysteinome-profiling workflows, including iodoTMT²⁸ and SLC-TMT,²⁵ under the equivalent condition. The HCC1171 human lung cancer cell line harboring the KRAS G12C mutation was treated with DMSO or 10 μM ARS-1620, lysed, and processed with one of the three workflows, with three technical replicates of DMSO or ARS-1620 in each. Among all, SLC-TMT had the best enrichment specificity (98%) as compared to AT-MAPP (59%) and iodoTMT (50%) and therefore had the greatest number of probe-modified cysteine sites quantified. However, AT-MAPP and iodoTMT outperformed SLC-TMT in quantitation accuracy as demonstrated by the coefficient of variances (Figure 2A; Supplemental Table 2), underscoring the advantage of combining samples early at the protein level. Next we applied the AT-MAPP workflow to monitor the proteome-wide cysteine profile with a dose titration of ARS-1620 in the HCC1171 (Figure 2B; Supplemental Table 1). In brief, treated cells were lysed, incubated with 100 μM of IA-Alkyne probe, clicked with 11plex-AzidoTMT, combined, and trypsinized. After offline high-pH reversed phase fractionation, probe-labeled cysteine peptides were captured by immunoprecipitation with the anti-TMT antibody and analyzed in a mass spectrometer. From one SPS-MS3 acquisition totaling 18 h, a total of 18 490 cysteine sites from 6124 proteins were identified at less than 1% false discovery rate and quantified across all 11 samples. TMT intensities among replicates were highly reproducible, with median coefficient of variances (CV) of 6%, 7%, 5%, and 6% for the DMSO, 0.1, 1, and 10 μM conditions (Figure S2A).

Figure 2.

Benchmarking AzidoTMT with a previously reported KRAS G12C covalent inhibitor ARS-1620. (A) Violin plot of CVs for identified and quantified sites across workflows AT-MAPP, iodoTMT, and SLC-TMT. (B) A schematic overview of the experiment setup to profile ARS-1620 cysteine targets using AT-MAPP. Live cells are pretreated with ARS-1620, lysed, and labeled by IA-Alkyne probe. Probes are further clicked with AzidoTMT, digested, fractionated, enriched, and analyzed using mass spectrometry. (C) A volcano plot is shown to highlight key changes from 10 M ARS-1620 as compared to the DMSO control. Each dot represents a unique cysteine site and is colored based on total protein change (or in grey if not quantified). Each treatment condition is quantified from three replicates. The fold change and p-value are calculated by the “msstats” package in R. (D) Dose-response curves for mutant KRAS C12 residue and MYC C300 residue from AT-MAPP. Each condition is represented by two or three replicates, and a curve is fitted using nonlinear regression. Western blot analysis confirms KRAS covalent modification. Alkylated KRAS shows as a distinct band due to slower migration as compared to unmodified KRAS on SDS-PAGE. (E) Representative examples of quantified cysteine sites with different classes of regulation patterns upon inhibitor treatment. The left and right panels display proteins that do not change while site levels decrease or increase respectively. The middle panel highlights cysteine sites that change in accordance with the total protein-level change.

TMT intensities were then summarized to site level to allow comprehensive assessment of compound behavior in cells (Figure 2C; Supplemental Table 2). As expected, Cys12 of the KRAS G12C mutant was alkylated in a concentration-dependent manner by ARS-1620 (Figure 2D). The abundance of the unmodified Cys12 peptide was reduced by as much as 91% (10 μM vs DMSO log2 ratio = −3.51, adjusted p-value = 1.8 × 10^–5) when treated with 10 μM of compound for 4 h, in agreement with the Western blot data (Figure 2D). Based on the dose-response curve, the cellular EC50 of ARS-1620 for alkylating KRAS G12C in HCC1171 cells was 0.87 μM (95% CI: 0.67–1.1 μM). Additional probe-labeled, cysteine-containing peptides showed dose-dependent reduction in compound-treated samples as compared to the control, including a previously reported off-target FAM213A C85. At the highest dose, 13 sites showed significant reduction (log2FC ← 1, adjusted p-value <0.01). It is important to note that although a reduction in cysteine labeling typically is ascribed to competitive alkylation by covalent compounds, other factors can contribute as well. One such factor would be protein downregulation either through inhibited synthesis or inducible degradation. Interestingly, the MYC peptide spanning C300 (MYC C300) was probe labeled and displayed dose-dependent decrease with ARS-1620 (Figure 2D). MYC is an important proto-oncogene but also an intrinsically disordered protein generally considered to be undruggable.^40,41 A recent study reported a covalent ligand targeting the C171 site of MYC.⁴² However, the MYC C300 site was not previously known for covalent ligand modification nor for cysteine probe labeling. We confirmed identification of the alkylated C300 peptide from endogenous samples using a synthetic peptide with identical residue sequence and cysteine modification (Figure S2B). This decrease in MYC C300 peptide observed in cysteine profiling can be explained by a decrease at the protein level, as follow-up Western blotting analysis showed a dose-dependent decrease in total MYC level too (Figure S2C). It has been reported that RAS regulates MYC protein levels both transcriptionally and post-transcriptionally,⁴³⁻⁴⁵ while KRAS inhibition by ARS-1620 substantially downregulates the MYC mRNA level.³⁷ Inhibition of the RAS effector pathway downstream MEK1/2 by U0126 was shown to suppress MYC S62 phosphorylation and destabilize MYC protein.⁴⁶ Despite insufficient evidence to demonstrate direct site modification by ARS-1620, our observation suggests a solvent-exposed probe-labeled residue C300 could be an actionable site on MYC for future efforts aiming to covalently target MYC.

In live cells treated with covalent inhibitors, cysteine alkylation profiles are complicated by a number of biological and analytical factors. For example, covalent inhibitors can alter transcription, translation, posttranslational modifications, and protein stability. For this reason, we acquired protein-level changes for a total of 7763 proteins from the same set of samples and integrated these into the analysis. Indeed, several off-target transcription factor hits such as JUN C99 and HES1 C128 changed in parallel with their corresponding protein levels (Figure 2E). Although we cannot rule out possible inducible degradation consequent to direct covalent inhibitor binding, these hits should be interpreted with caution when nominating liganded residues. While most of the proteome-wide cysteine-profiling studies for KRAS G12C inhibitors have focused on the sites that decrease upon treatment,³⁶⁻³⁹ little attention has been paid to the upregulated sites. A few important proteins in the RAS-RAF-MAPK signaling cascade and downstream effectors showed upregulated cysteine peptide abundance. Surprisingly, some are not directly regulated through protein expression (Figure 2C and E). Among them, the most significantly upregulated site was CIC C1420 (10 μM vs DMSO log2 ratio = 4.37, adjusted p-value = 3.6 × 10^–4). CIC (Capicua) is a transcriptional repressor and is negatively regulated by active MAPK signaling via phosphorylation events.⁴⁷⁻⁵¹ The CIC total protein level did not change within the treatment time frame (10 μM vs DMSO log2 ratio = 0.12). This provides an example of posttranslational modification affecting cysteine peptide quantitation. Similar observation was also made from BRAF C748 and RPS6KA1 C223, which showed increased cysteine alkylation but little total protein change.

Deep Proteome-Wide Cysteine Profiling Identifies Covalent Fragment Targets

Having validated AT-MAPP to identify known covalent inhibitor targets, we next applied the workflow to screen covalent fragments and elucidate their targets in cells. We selected A2058, a melanoma cancer cell line, and four covalent fragments with an acrylamide warhead as our model system in this study (Figure S3A and B). Thus, this setup allowed all samples to be profiled together using 11plex-AzidoTMT. This approach was executed with compound treatment in cell as well as in lysate form. After treatment in-cell, lysates were collected and processed using the AT-MAPP workflow. Conversely, the in-lysate workflow consisted of compound treatment after protein lysate was obtained. A total of 16 089 cysteine sites were identified and quantified in-cell, and 19 605 sites in-lysate (Figure 3B), both of which presented a unique subset only identified in one study. In general, both studies shared a similar set of liganded sites, yet their magnitude of modification differed, probably due to a combination of cell permeability and local protein concentration in-cell, among many other factors (Figure 3C). For instance, RIF1_C880 was only liganded in-cell whereas TEC_C449 was more ligandable in-lysate as compared to in-cell, highlighting that there might be intrinsic factors that account for a fragment’s reactivity. A fragment’s reactivity in a buffer may not reflect its intrinsic reactivity in a cellular environment and affects site ligandability when tested. This may be one of many factors that should be considered when performing fragment-based screening. When treated in intact cell, compounds A1, A2, A3, and A4 yield 15, 0, 15, and 12 hits whose abundance was decreased by more than 50%, respectively, accounting for 0.07% of total sites on average per compound. Among the four fragments that were screened, at least two compounds share sites: DDIT4 C140, FAM213A C85, MTMR12 C152, and VAT1 C86, and all other hits were unique to one of the four fragments, indicating a diverse target space among these covalent fragments.

Figure 3.

Deep proteome-wide cysteine profiling identifies covalent fragment targets. (A) Schematic representation of experiment design for cysteine reactivity profiling. Dose-titration curves from reactivity profiling for sites liganded by covalent fragments. (B) A Venn diagram representing the total and overlapping quantified cysteine sites between in-cell and in-lysate covalent fragment targets using AT-MAPP. (C) A comprehensive list of cysteine sites with their thiol reactivities in the native environment and functional annotations. Sites shown in red are annotated as active site or redox-active disulde. (D) A heatmap of liganded sites with fold changes (log2 FC) among compounds A1, A2, A3, and A4 both in-cell and in-lysate. (E) Comparisons of cysteine reactivity between cysteines liganded by covalent fragments and all other identied cysteine sites. The p-value is calculated using the Wilcoxon rank sum test. **: p < 0.0001; *: p < 0.01; n.s.: not significant.

Covalent inhibitor target engagement is primarily determined by four factors: noncovalent scaffold binding affinity, warhead electrophile reactivity, target residue intrinsic reactivity, and warhead positioning geometry. We were curious if the cysteine sites modified by covalent fragments were primarily driven by their intrinsic thiol reactivity. Proteome-wide cysteine reactivity was assessed by a highly reactive alkylating agent iodoacetamide (IAM). A2058 cell lysates were treated with IAM at 10 concentrations from 0.051 to 1000 μM in threefold increments. Free cysteine not reacted with iodoacetamide was further labeled with IA-Alkyne probe, clicked with 11plex-AzidoTMT, and processed for LC-MS/MS analysis (Figure 3A). This experiment yielded the most comprehensive collection of in situ cysteine reactivity data by far with full titration curves and site-specific RC50s (the required concentration for an alkylating agent to reduce the free thiol level by half at a given cysteine site). We compared the cysteine coverage of our study to that found in the CysDB⁵² and determined there are 2908 cysteine sites not found in the CysDB (Supplemental Table 2). Similar to a previous report,¹⁵ proteome-wide cysteine thiol reactivity demonstrates a large variation among all identified sites. Cysteine sites with low RC50s (i.e., high reactivity) included functionally important cysteines such as catalytic centers and redox-active disulfides (Figure 3D). Interestingly many of these functionally relevant cysteine sites were highly enriched hits in the fragment-screening experiments, in particular for compounds A3 and A4 (Figure 3E; Figure S3B). This suggests that more reactive sites are more ligandable by covalent fragments. However, none of the most reactive sites (i.e., RC50 < 10^–6 M) were liganded by these compounds, suggesting the cysteine reactivity is not the only driver of ligandability. As demonstrated by the diversity in sites liganded by compounds A1–A4, there is a substantial contribution from noncovalent scaffold-binding affinity that is driving target selectivity, even for fragment compounds that have a simple chemical structure, underscoring the importance of fragment-binding affinity to achieve targeted protein modification.

Mapping Proteome-wide Noncysteine Hotspots with Sulfonyl Fluoride-Based Compounds and AzidoTMT

A unique advantage in the classical isoTOP-ABPP and our AT-MAPP assay is that the azido group can react with any alkyne-based probe via click reaction. This expands the application of AzidoTMT beyond cysteine profiling to potentially any residue for which an alkyne-functionalized chemical probe is available. A sulfonyl fluoride probe (DAS1) was recently reported to label a wide range of nucleophilic ligandable hotspots, including tyrosine, lysine, serine, and threonine.⁵³ We tested the sulfonyl fluoride probe after initially labeling with 20 in-house-generated fragment compounds (S1–S20) in lysate to search for hotspots (Figure S4). A2058 cell lysates were labeled with DMSO in triplicate, and each of the 20 compounds in duplicates and samples were analyzed using five AzidoTMT-11plex sets. Due to lower probe reactivity, 507 probe-labeled sites (or site combinations, i.e. multiple probe-modified tyrosine, lysine, threonine, or serine residues located on one peptide) were identified after applying an Ascore cutoff of 15 to localize modification sites. Probe modification sites were primarily localized to lysine and tyrosine residues, accounting for 43% and 48%, respectively, with a minor fraction to serine (5%) and threonine (4%) (Figure 4A). Among them, 18 sites were liganded by at least one of the fragments (Figure 4B; Supplemental Table 2). Not all liganded sites were identified and quantified for all 20 compounds. Fragment screening revealed hits showing selective labeling on physiologically relevant residues. Compound S8 specifically and potently modified tyrosine 129 on the protein NAD(P)H:quinone oxidoreductase 1 (NQO1). Tyrosine 129 has been shown to be an integral part of forming confirmations that allow the opening and closing of the binding site for NQO1.⁵⁴ Computer-generated docking simulation highlights the compound fitting into a pocket and possibly making hydrogen bonds with surrounding residues (Figure 4C). With tyrosine 129 forming polar contacts with NQO1 substrates such as FAD,⁵⁵ NQO1 has been indicated to be overexpressed in many solid tumors, thus making it an attractive target for cancer drug development.⁵⁶ While much remains to be done to leverage these reactive hotspots, this data set provides a proof-of-concept demonstrating application of AzidoTMT toward other alkyne-functionalized chemical probes.

Figure 4.

Mapping proteome-wide hotspots with sulfonyl fluoride-based compounds. (A) A summary of the frequency of residues modified in this study by DAS1, a probe with a sulfonyl fluoride warhead for targeting nucleophilic residues. (B) A heat map profiling various liganded sites identified from 20 sulfonyl fluoride-based fragments using the AT-MAPP workflow where each compound’s ligandability was assessed by its competition to the DAS1 probe. (C) In silico covalent docking of compound S8 on NADPH dehydrogenase 1 protein (PDB:1d4a) at the Y129 site indicating its location in a prime pocket for binding.

Discussion

It remains challenging to target certain highly prevalent oncogenes such as KRAS and MYC. Recent KRAS G12C inhibitor data from both preclinical and clinical studies reignited broad interest in development of targeted covalent inhibitors for these and other “undruggable” targets.^36,57,37 Notably, chemical proteomics has been an integral tool for discovering ligandable hotspots and evaluating targeted covalent inhibitor selectivity, among which isoTOP-ABPP is a prime workflow. To expand isoTOP-ABPP beyond binary comparisons, we generated a multiplexed reagent called AzidoTMT and integrated that into the standard workflow to form a new and improved workflow named AT-MAPP (AzidoTMT-based Multiplexed Activity-based Protein Profiling). Chemical reaction to synthesize 11plex-AzidoTMT is an easy one-pot, one-step robust reaction between an azido-functionalized primary amine and TMT (an activated NHS ester), both of which are commercially available. The azido group can be conjugated to any alkyne-functionalized probe via click reaction, making it a versatile reagent with broad applications. Once clicked, all samples can be immediately mixed, eliminating intersample variations from preparation steps that follow, including protein precipitation and proteolytic digestion. Another practical benefit with combining all samples into one at the protein level is simplification of liquid-handling operations. Using AzidoTMT for activity-based protein profiling is cost-effective. With a TMT set in a 0.8 mg unit size, theoretically it can generate 2.3 μmol of AzidoTMT per tag. For a typical cysteine-profiling experiment using 100 μL of 100 μM IA-Alkyne (= 0.01 μmol) per condition, 2.3 μmol of AzidoTMT is enough for more than 100 experiments when applied in a twofold excess. Even with partial hydrolysis and loss from purification steps, from our experience, AzidoTMT generated from one unit of TMT reagent is enough for >50 experiments.

One main advantage of AzidoTMT is multiplexing. With 11plex TMT tags, complex experimental designs involving multiple cells, compounds, time points, doses, or replicates can fit into one plexed set for simultaneous sample processing and analysis.⁵⁸⁻⁶² This eliminates the need to run multiple binary isoTOP-ABPP experiments. Importantly, TMT-based quantitation has no missing data across channels, which is a significant advantage over concatenating binary quantification events from multiple runs. Here we applied AzidoTMT to three studies. Each set consistently yields over 10 000 unique sites fully quantified across all experimental conditions, which has a much deeper coverage than isoTOP-ABPP experiments. Excellent quantitation accuracy can be achieved, with median CVs of less than 10% across quantified sites following median normalization of the TMT data. We recommend acquiring TMT reporter ion data with SPS-MS3 to reduce the impact of ratio compression, a phenomenon inherently associated with isobaric quantitation.^63,64 TMT quantitation accuracy can be further improved by real-time search and intelligent fragment peak picking, as recently reported.^65,66 This would be especially beneficial for AzidoTMT-labeled peptides because theoretically only half of peptide fragments ions carry TMT reporter tag for quantitation. Focusing on TMT-tagged, peptide-specific fragments for MS3 would further improve the signal-to-noise and accuracy of product reporter ions.

Target deconvolution for covalent inhibitors and assessing their proteome-wide selectivity represents an important application of proteome-wide cysteine profiling using IsoTOP-ABPP or other assays built upon the same concept. In such a model, reduction in the mass spectrometric signal of a probe-tagged cysteine peptide is often interpreted as competitive inhibitor labeling on the same residue. The corresponding site is then inferred as a covalent inhibitor target. However, covalent inhibitors can affect proteome in ways other than direct labeling, especially when live cells are treated. For instance, inhibition of mutant K-Ras suppresses PI3K and MAPK signaling cascades, resulting in remarkable changes in protein turnover and posttranslational modifications.³⁷ These changes can propagate to cysteine peptides and show up as putative “hits”. Two approaches may be taken to reconcile a target list with higher fidelity. One is to convert the covalent inhibitor into a probe by installing an affinity handle and use it for affinity capture of direct targets.^5,19,39 When such a probe is not available or does not completely recapitulate the entire target profile like the parent molecule, the second approach is to interpret proteome-wide cysteine-profiling hits with additional data from protein and post-translational modification level changes. In-depth cysteine profiling followed by comprehensive proteomic analysis (global proteome, global phosphoproteome, etc.) can give insights into a covalent inhibitor’s mechanism of action and target selectivity.

In our study, we observed an upregulated peptide bearing C1420 residue from the protein CIC. Interestingly, a previous phosphoproteomics study reported increased mono- (S1419) or di- (S1418, S1422) phosphorylation upon KRAS inhibitor treatment on the same CIC tryptic peptide sequence.⁶⁷ Researching our cysteine-profiling data with additional S/T/Y phosphorylation as variable modification did not yield direct evidence for phosphorylated peptides containing the same alkylated C1420. Therefore, we can only speculate that when mutant KRAS is constitutively active, the major CIC tryptic peptide bearing C1420 and five serine/threonine residues is hyperphosphorylated. However, upon KRAS inhibition, the major population shifts to a hypophosphorylated form, either through protein turnover or dephosphorylation. This interpretation is consistent with the increase of alkylated C1420 identified in the unphosphorylated form as seen in our study and the increase of mono- and diphosphorylation in the same peptide sequence as reported previously. KRAS inhibition has profound effects on the global phosphorylation landscape directly through MAPK and PI3K signaling and indirectly through their corresponding feedback mechanisms.^68,67 These can, in turn, affect cysteine alkylation readouts if phosphorylation alters the composition of phosphorylated and unphosphorylated peptide forms bearing a cysteine site or the ability to generate the Cys-modified peptide by changing cysteine solvent accessibility for probe labeling or altering proteolysis efficiency.⁶⁹ Besides CIC C1420, additional hits including BRAF C748 and RPS6KA1 C223 showed increased cysteine alkylation but little total protein change. Indeed, KRAS signaling regulates BRAF S750 and RPS6KA1 S221 sites, both of which are colocalized with the respective cysteine sites on tryptic peptides.⁷⁰⁻⁷² Because CIC phosphorylation is directly associated with MAPK activity, this allows evaluation of direct G12C alkylation and its downstream effect in the same assay.

In summary, using the newly developed reagent AzidoTMT and the corresponding workflow AT-MAPP, we’ve been able to identify covalent inhibitor targets, screen covalent fragments, profile cysteine thiol reactivity, and map proteome-wide lysine hotspots. The AzidoTMT-based workflow can be further extended to additional residues such as functional tyrosine sites using tyrosine-selective sulfonyl-triazole probes.⁴ Because quantitative tags are directly attached to the site of interest, we envision that studies focusing on site-specific changes will benefit the most. Overall, this novel reagent represents a significant advancement in the chemical proteomic toolbox to support the future needs of covalent inhibitor drug development.

Significance

Recently, targeted covalent inhibitors (TCI) bring a new avenue to drug conventionally “undruggable” targets. Chemical proteomics workflow such as isoTOP-ABPP is an integral part of drug discovery and development to discover ligandable hotspots and study TCI selectivity. Classical isoTOP-ABPP is limited to binary comparison of two samples in parallel. SLC-TMT on the other hand allows multiplexed analysis but has only been applied to desthiobiotin iodoacetamide for cysteine labeling thus far. In order to expand the workflow to accommodate a more complex experiment setup while still maintaining the flexibility in pairing with alkyne-based probes, we developed a novel reagent, AzidoTMT, for highly multiplexed activity-based protein profiling (termed AT-MAPP workflow). This reagent set consists of 11 isobaric tags and can be synthesized easily from commercial products through one-step reaction. We demonstrate its seamless integration to multiple types of studies to investigate covalent inhibitor selectivity, screen covalent fragment targets, profile cysteine thiol reactivity, and map proteome-wide noncysteine hotspots. Benchmarking with ARS-1620 identifies expected target K-Ras G12C and off-targets reported elsewhere, but some hits can be explained by protein-level change. Correlating covalent fragment screening hits with their cysteine reactivity profile reveals a strategy to perform target-based covalent fragment screening. We envision that AzidoTMT is a versatile reagent with broad application and can make a significant contribution to covalent drug development.

Glossary

Abbreviations

AT-MAPP

AzidoTMT-based multiplexed activity-based protein profiling

ABPP

activity-based protein profiling

IsoTOP-ABPP

isotopic tandem orthogonal proteolysis

CuAAC

copper(I)-catalyzed alkyne–azide cycloaddition

SLC-TMT

streamlined ABPP of reactive cysteines

TFA

trifluoroacetic acid

PBS

phosphate-buffered saline

DTT

dithiothreitol

IA-Alkyne

iodoacetamide-alkyne

SMILES

simplified molecular-input line-entry system

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jproteome.2c00703.

• Supplementary figures including AzidoTMT reagent synthesis and qualifications, benchmarking AzidoTMT for selectivity, cysteine reactivity profiling with dose-titration curves, chemical structures of fragment compounds, and source material and software versions (PDF)

• Detailed TMT-labeling scheme for all experiments (Table S1) (XLSX)

• All quantitative proteomic results summarized to site or protein level (Table S2) (ZIP)

• All peptide-spectra-matches data with site localization and raw TMT reporter intensities (Table S3) (ZIP)

The authors declare the following competing financial interest(s): All authors are current or former employees and shareholders of Genentech, a member of the Roche Group.