Covalent Fragment Screening Using the Quantitative Irreversible Tethering Assay

Summary

The quantitative irreversible tethering (qIT) assay is a fluorescence-based method of identifying covalent fragments that selectively engage with a target protein. Here, a detailed protocol is outlined to describe the qIT assay, with example results included.

Abstract

Compounds that form covalent bonds with specific target proteins offer a variety of advantages as chemical probes and therapeutic agents. Most commonly, mildly reactive, electrophilic small molecules are employed to form covalent bonds with select cysteine side chains in specific proteins. Electrophile-first approaches of ligand discovery, whereby a library of electrophilic small molecules are screened against a protein target, have become popular as they avoid the need for time-consuming downstream installation of an electrophilic warhead. Such screening is complicated, however, as electrophilic ligands can exhibit a wide range of different rates of spontaneous reaction with cysteines. Quantitative-irreversible tethering (qIT) offers a fluorescence-based method for hit identification and development that normalizes data for these differences in intrinsic compound reactivity. Rates of reaction of individual compounds with a target protein are determined and compared to compound reactivity with the unstructured tripeptide glutathione (this being a proxy for spontaneous compound reaction), enabling the identification of compounds that preferentially react with the protein of interest. This methodology has been successfully applied to identify selective covalent fragments against several drug targets, including SARS-CoV-2 main protease, cyclin-dependent kinase 2, and RAP27A. Here, we demonstrate the application of qIT to a target protein to generate a quantitative and robust data set, allowing prioritization of hit ligands for future development.

Introduction

Cysteine-reactive targeted covalent ligands (molecules that exert a biological effect by reacting with a select cysteine residue on a specific protein) have become increasingly popular as chemical tools and therapeutics over recent decades¹. These typically consist of a mildly reactive electrophile (such as an acrylamide, chloroacetamide, or vinyl sulfone) appended to a ligand moiety with a high affinity for a protein target². Ideally, ligand-protein interaction proceeds via a two-step mechanism involving initial non-covalent ligand binding, followed by irreversible covalent bond formation between the target cysteine and electrophile³.

Several early targeted covalent ligands, including Bruton's Tyrosine Kinase inhibitor Ibrutinib, were developed by furnishing a previously identified reversible ligand with an electrophilic handle⁴. However, this approach is not always successful, and other attempts have found electrophile installation to require significant optimization⁵. Instead, electrophile-first approaches, particularly using electrophilic fragments, have become more popular in recent years. These typically involve incubating a library of electrophiles with a protein and then selecting compounds with an enhanced rate of reaction towards a cysteine, which is assumed to be induced by productive non-covalent interactions between the ligand and binding pocket⁶. Notably, these methods were successfully applied during the development of the KRAS (G12C) inhibitor Sotorasib¹.

Electrophile-first screening poses several unique challenges when compared to traditional non-covalent ligand screening, principally due to the high variability in reactivity within and between warhead classes⁷. Most cysteine-reactive electrophiles will, over time, react with all solvent-exposed cysteine residues, regardless of any selective interaction for the target protein. Ligand occupancy will, therefore, increase over time for almost all electrophiles, and without controlling for reactivity, hit selection will often be biased towards more reactive fragments that engage a target more rapidly⁸.

Moreover, electrophile-first screening methods often use mass spectrometry to detect protein-ligand adduct formation, which is limited in throughput⁹. Regular reaction sampling is required over an extended time period (hours to days) to properly characterize reaction kinetics for a library of compounds with varied electrophilicity¹⁰. A screen of 300 fragments sampled at 10-time points will, therefore, require 3000 mass spectra to be performed, exceeding the throughput of most protein mass spectrometry systems. Higher-throughput methods of monitoring protein-ligand adduct formation are, therefore, desirable.

The approach outlined here, the quantitate irreversible tethering (qIT) assay, is an electrophile-first screening method that controls intrinsic reactivity and utilizes a simple fluorescence readout⁷. Protein, or glutathione (an unstructured, thiol-containing tripeptide used as a control for intrinsic reactivity) are incubated with a library of electrophilic fragments, and regular samples taken for quenching into a solution of CPM (7-Diethylamino-3-(4'-Maleimidylphenyl)-4-Methylcoumarin, a fluorogenic thiol-quantification agent). Adduct formation between the cysteine and electrophilic fragment reduces the amount of thiol available for reaction with CPM, reducing the resulting fluorescence on quenching. CPM fluorescence thus acts as a proxy for unreacted thiol, allowing the kinetics of thiol reaction with each electrophilic fragment to be determined. Electrophilic fragments with significantly enhanced protein reactivity over glutathione are selected as hits and prioritized for further development (Figure 1).

Figure 1: Schematic of qIT. A schematic overview of the qIT assay. (created using BioRender). Please click here to view a larger version of this figure.

Hit selection using REF values controls for intrinsic reactivity and ensures that fragments are only selected as hits if they form productive non-covalent interactions with a protein target. The simple fluorescence readout also means the qIT assay is highly scalable and does not require mass spectrometry-based methods of reaction monitoring.

Here, a comprehensive procedure for the qIT assay is outlined, with example results included to illustrate assay performance and hit selection.

Protocol

1. Fragment library construction

1. Design a library of electrophilic fragments and either synthesize or purchase them. Electrophilic fragment libraries typically consist of mildly reactive electrophiles (such as acrylamides, chloroacetamides, and vinyl sulfonamides) appended to a diverse array of chemical scaffolds. Electrophilic fragment libraries are widely available from commercial suppliers, and more detail about their design and construction is reviewed elsewhere^5,11.
2. Dissolve each compound in DMSO to a concentration of 50 mM, then dispense a different compound into each well of columns 3 to 22 of a 384-well reservoir plate, termed the Fragment Library Plate.
   1. Dispense DMSO into each well of columns 1, 2, 23, and 24 for the positive and negative controls. Plate layout can be varied depending on the number of compounds being screened (Figure 2). The volume added to the library plate will depend on the number of screens to be performed, considering approximately 1.8 μL from each well is required per qIT screen. Prepare this Fragment Library Plate in advance and store it at 4 °C.

Figure 2: Plate layout. An overview of qIT assay 384-well plate layout. Please click here to view a larger version of this figure.

2. Assay setup (Figure 3)

1. Prepare the following solutions: recombinant protein solution (target protein in any buffer), qIT assay buffer (150 mM NaCl, 25 mM Hepes (pH 8)), qIT assay quench buffer (150 mM NaCl, 25 mM Hepes (pH 7.5)), and compound stock solutions (electrophilic fragment (50 mM) in DMSO).
   NOTE: A pH of 8 is used in the reaction buffer to accelerate the reaction between the target cysteine residue and fragments, allowing the rate of reaction to be measured even for extremely mildly reactive fragments. A slightly lower pH of 7.5 is used in the quench buffer as the thioether bond formed between the maleimide of CPM and the cysteine thiol is prone to hydrolysis at more basic pHs (for example, previous work found cysteine-maleimide conjugates are far more stable at pH 7.3 than pH 8¹²).
2. Remove the DMSO fragment stock solutions from storage and leave them at room temperature for a minimum of 2 h to allow compound stocks to thaw.
3. Degas qIT assay buffer by bubbling with argon gas for approximately 30 min.
4. Thoroughly exchange the target protein solution into degassed qIT assay buffer. Crucially, reducing agents such as DTT and TCEP must be thoroughly removed to concentrations < 100 nM as they will interfere with thiol quantification. This step can be performed via a repetitive series of buffer exchanges, as described below.
   1. Dilute protein solution in excess of degassed qIT assay buffer in a spin filter unit, concentrate using centrifugation (4000 x g, 4 °C, 15 min) and repeat. Typically, 4 rounds of 1 in 15 dilutions are required to sufficiently remove DTT (1 mM) from 1 mL of recombinant protein stock solution, although this step can be adjusted depending on the reducing agent concentration in the original recombinant protein solution.
      NOTE: Alternatively, dialyzing the target protein solution against a large excess of degassed qIT assay buffer can achieve the same outcome.
5. Dilute the target protein solution (produced in step 2.1) to a protein concentration of 15 μM in degassed qIT assay buffer. Approximately 9 mL of target protein solution is required to fill a 384-well plate with sufficient dead volume.
6. Prepare a 1.5% (v/v) suspension of TCEP-agarose in degassed qIT assay buffer. Approximately 9 mL of TCEP-agarose suspension is required per screening plate.
7. Dispense 20 μL of TCEP-agarose solution into every well of two 384-well plates.
8. Prepare a 15 μM glutathione stock in degassed qIT assay buffer. This should be prepared immediately before plating as glutathione rapidly oxidizes in solution without reducing agent.
9. Dispense 20 μL of qIT assay buffer into columns 1 and 2 of both 384-well plates.
10. Dispense 20 μL of 15 μM protein or glutathione solution (in accordance with the intended plate) into columns 3-24 of the 384-well plate. These are termed the Protein Reaction Plate and the Glutathione Reaction Plate. Leave to reduce at RT for 1 h.
11. Dispense 58.2 μL of degassed qIT assay buffer into each well of a 384-well plate, then transfer 1.8 μL from each corresponding well of the Fragment Library Plate to create a Compound Dilution Plate, where each compound is diluted to a concentration of 1.5 mM in qIT assay buffer plus 3% DMSO.
    NOTE: This step is made easier with a 384-pipetting station, which allows compound stocks to be transferred for every well in one operation.
12. Following 1 h incubation of protein and glutathione with TCEP-agarose, transfer 20 μL of compound solution from each well of the Compound Dilution Plate into both the protein and glutathione reaction plates to begin the assay. This step is also made easier with a 384-pipetting station. The assay steps are summarized in Figure 3.
    NOTE: The fragment concentration can be varied as required, and qIT assays can be repeated for the hits at lower concentrations, if desired, to confirm the protein affinity is not an artifact of the high fragment concentration. The relatively high fragment concentration is used to elevate the k_pro and k_GSH to ensure that accurate rates are determinable for mildly reactive compounds (which can display t_1/2(GSH) > 24 h under the assay conditions). Moreover, although a fragment concentration of 500 µM is high, high µM to low mM concentrations have been used in fragment screening, for example¹³.
13. Start a timer - the assay has begun, and quenches should be performed at regular time points in accordance with the section below.

Figure 3: qIT assay summary. A summary illustrating the key steps of qIT assay reaction well setup. (created using BioRender). Please click here to view a larger version of this figure.

3. CPM quenches

NOTE: CPM quenches are performed to quantify the amount of unreacted thiol remaining in the protein and glutathione reaction plates. These should be performed at discrete time points (for example, the following time points are typically used: 7, 15, 30, 60, 120, 240, 360, 1220, and 1440 min) to allow reaction kinetics to be monitored.

1. Prepare CPM stock solutions (500 μM) in DMSO, dispense them into 111.2 µL aliquots in microcentrifuge tubes, and store them at -20 °C. These CPM stocks can be prepared in advance, and 1 aliquot will be required per quench time point.
2. Prepare CPM quench solution (1.4 μM) by thawing a frozen CPM stock solution (500 μM) and adding it to 40 mL qIT assay quench buffer.
3. Dispense 27 µL of CPM quench solution into each well of two 384-well plates-two are required for quenching both the protein and glutathione reaction plates. These are termed CPM-Quench Plates, prepare them a maximum of 30 min before use.
4. Centrifuge the protein reaction plate at 200 x g for 3 min to ensure TCEP-agarose is pelleted.
5. At pre-determined time points, transfer a 3 µL of sample from every well of the Protein Reaction Plate to a pre-filled-CPM Quench Plate. This step is also simplified using a 384-pipetting station.
6. Mix the CPM Quench Plate by aspirating and dispensing 3 x 25 µL. It is important to remove the sample from the top of the well to avoid aspirating the TCEP-agarose at the bottom, which will interfere with CPM quenching.
7. Repeat steps 3.4-3.6 with the Glutathione Reaction Plate.
8. Incubate the CPM Quench Plates (prepared in 3.4-3.6) at room temperature for 1 h.
9. Measure the fluorescence intensity of the CPM quench plate (excitation at 384 nm and emission at 470 nm).

4. Data analysis and hit selection

1. For each quench, calculate Z' as a measure of assay quality. Any CPM quenches with Z' < 0.5 should be disregarded.
   Z' = 1 - ((3σ_{positive control} + 3σ_{negative control}) / (µ_{positive control} - µ_{negative control}))
   Where σ = fluorescence intensity standard deviation and µ = fluorescence intensity mean
2. Calculate mean fluorescence values for the negative and positive control wells (columns 1 and 2, and 23 and 24, respectively).
3. Normalize each reaction well to the average fluorescence values of the positive and negative controls:
   % modified = 100 x (compound well fluorescence - negative control fluorescence) / (positive control fluorescence - negative control fluorescence)
4. Separately plot % modified versus time for each compound reacting with glutathione and with target protein.
5. Fit normalized % modified versus time data to a one-phase exponential decay model for both the reaction with protein and with glutathione using the equation:
   % Modified = ae ^{- kt + c}
   Where a = the range of % modified values, k = rate of reaction, t = time, c = the % modified value at infinite time. Curve fitting and parameter determination can be performed using commercial software (such as Graphpad).
6. Use the rate at which fragments react with protein (k_pro) and glutathione (k_GSH) to calculate the Rate Enhancement Factor (REF) value for each fragment:
   REF = k_pro / k_GSH
7. Use the REF values to select hit fragments. This can be performed by setting a REF threshold (e.g., REF > 3), an arbitrary hit rate (e.g., top 2%), or by selecting fragments with REF 3 standard deviations greater than the geometric mean.

Results

The qIT assay is used here to identify covalent fragment hits for an unnamed, single cysteine protein. MgCl₂ (10 mM) was added to the reaction buffers as the target protein is known to bind to Mg²⁺ ions. The screening was performed in triplicate, and the rate at which each fragment reacted with both the target protein (k_pro) and glutathione (k_GSH) determined. Z' values were greater than 0.5 for all quench plates and averaged 0.75 for protein quenches and 0.79 for glutathione quenches, indicating excellent assay performance. REF values were then calculated for each fragment, and hits selected where REF > 3.

The representative results (Figure 4), provide an example of hit selection using the qIT assay. Compound 1 reacted at a 6.0-fold greater rate with the cysteine residue on the target protein than with glutathione and, using a criterion of REF > 3, was selected as a hit (Figure 4B). This rate acceleration indicates the presence of a templating effect, whereby non-covalent interactions position 1 in the correct orientation for the cysteine on the target protein to react with the chloroacetamide warhead. In contrast, 2 is an example of a non-hit fragment as the reaction between 2 and the target cysteine residue is not accelerated relative to the reaction between 2 and glutathione. The lack of rate acceleration indicates there are not productive non-covalent interactions positioning 2 in the correct orientation to react with the target cysteine residue.

Figure 4: Representative results. The results show protein and glutathione reaction monitoring by qIT for (A) A high intrinsic reactivity, a possible false-positive hit fragment, and (B)a hit fragment. Please click here to view a larger version of this figure.

Discussion

The qIT assay provides an effective and convenient platform for screening cysteine-reactive electrophilic fragments against a target protein. However, there are critical considerations that may hamper assay performance if not adequately accounted for. Most notably, target proteins must only include a single surface-exposed cysteine to be suitable for qIT screening, as other cysteine residues will react with CPM and interfere with fluorescence measurements. In previous cases where target proteins contain multiple surface-exposed cysteines (such as for the SARS-CoV-2 main protease, the small GTPase RAP27A, and Cyclin-dependent kinase 2), screening constructs with interfering residues removed have previously been used successfully. However, the effectiveness of this strategy will vary depending on the protein^10,14,15.

Additionally, as the qIT assay measures the decrease in thiol available to react with CPM over time, any factor that reduces the proportion of available thiols, such as thiol oxidation or protein aggregation, may hamper assay performance. Buffer degassing and the use of an immobilized reducing agent (TCEP-agarose) ensure thiol oxidation is slow, and any remaining oxidation should be mitigated by normalizing fluorescence measurements to that of the positive control wells. However, problems can arise when an uneven distribution of TCEP-agarose between wells causes non-uniform thiol oxidation. Consistent TCEP-agarose plating is therefore essential for assay performance, which can be challenging as TCEP-agarose particles sink in pipetting reservoirs.

Protein aggregation can also be a problem, as the assay conditions (24 h incubation at room temperature) are relatively harsh. Troubleshooting strategies such as fusing target proteins to solubilizing tags, altering reaction buffers, or performing the assay at lower temperatures can reduce aggregation and may solve protein stability issues. Certain fragments may also induce protein aggregation, which again reduces the proportion of thiol available to react with CPM and creates false positive hits. Orthogonal hit validation using intact-protein mass spectrometry is an effective method of detecting these cases, as false positive hits will create a small or non-existent population of labeled species when incubated with a target protein.

When these considerations are controlled for, the qIT assay has some clear advantages over other covalent fragment screening methods. Hit selection using REF ensures that hits are only selected if they form productive non-covalent interactions with the target protein. Fragment electrophilicity varies over large ranges, both between different warheads and within warhead classes; for example, previous work has found several hundred-fold variations in the rate of glutathione reactivity between acrylamides¹⁰. This wide range of electrophilicity can make it challenging to distinguish fragments with high k_pro values due to a templating effect from those which are simply highly reactive, which can lead to highly reactive fragments being selected as false positive hits. For example, 2 reacted with the target cysteine residue more rapidly than 1 and, without considering intrinsic reactivity, may be selected as a hit. However, 2 does not react with the target cysteine residue at a greater rate than with glutathione, indicating the greater k_pro of 2 relative to 1 is due to greater fragment intrinsic reactivity, and not a templating effect.

The simple fluorescence readout used by the qIT assay to quantify the remaining thiol is also highly advantageous. Other methods of covalent fragment screening use intact-protein mass spectrometry to determine the rate at which fragments react with a target protein and LC-MS to determine the rate at which glutathione reacts with fragments - both of which require specialized equipment and lack throughput^8,16. The convenient fluorescence readout makes the qIT assay highly scalable and requires only commonly available laboratory equipment.

Combined, these innovations create a robust and scalable method of covalent fragment screening, which negates the need for mass-spectrometry equipment.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Argon Gas	BOC	I1149
Corning 384-well, low-flange, flat-bottom assay plates	Corning	3575
CPM	Invitrogen	D346
DMSO	Sigma-Aldrich	D8418
Immobilized TCEP-agarose	Pierce	77712
Micro plate reader	CLARIOstar
PCR Plate Seals	BioRad	MSA5001
Reduced L-glutathione	Sigma-Aldrich	G4251
Sodium Chloride	Sigma-Aldrich	S9888
Sodium Hepes	Sigma-Aldrich	RES6007H-A7