High throughput glutathione and Nrf2 assays to assess chemical and biological reactivity of cysteine-reactive compounds

Abstract

Compounds that inhibit their target through covalent binding offer a number of unique advantages as potential therapeutic agents. They can achieve high ligand efficiency, and thus high potency, which can translate into a reduced drug dosage. In addition, covalent binding can result in a long duration of action of the compound and thus less frequent drug dosing. Despite these advantages, there are several safety concerns about this class of compounds because of their inherent reactivity and potential for non-specific binding to cellular proteins. The primary aim of this study was to establish the thiol reactivity of acrylamides and other cysteine-reactive groups by measuring reactivity against various cysteine-containing cellular components. Compounds were incubated with glutathione, bovine serum albumen (BSA) or human liver microsomes (HLM), and the reduction in thiol levels was quantitated using Ellman's reagent. In addition, the ability of compounds to induce Nrf2 activity in a reporter gene assay was used as a functional readout of compound reactivity. The assays were validated using known thiol-reactive compounds and Nrf2 inducers such as sulforaphane (SFP) and 1,2-dithiole-3-thione (D3T). Our results demonstrate that acrylamides possess low reactivity unless activated by electron-withdrawing N-substituents. Analogous propynamides and vinyl sulphones were more reactive than acrylamides with respect to glutathione activity but were less reactive in the Nrf2 assay. Large drug-like molecules, including irreversible kinase inhibitors targeting BTK and EGFR, were far less reactive than their lower molecular weight counterparts, suggesting that the presence of an electrophile is not sufficient to predict thiol reactivity. These studies demonstrate the utility of multiple thiol sources in addition to glutathione when assessing the thiol reactivity of covalent inhibitors.

---

Introduction

Designed covalent inhibitor drugs are electrophilic and react with nucleophiles such as the cysteine, lysine or histidine side chains of proteins.^1,2 As a class, they offer several unique advantages as a drug discovery approach. First, covalent drugs can achieve high ligand efficiency, defined as the free energy of binding per heavy atom, as a result of their non-equilibrium binding mechanism in which the covalently bound form is highly favored. This can translate into highly potent inhibitors, and ultimately, a reduced drug dosage. Competition with endogenous substrates is limited and in the case of kinase inhibitors, for example, this is advantageous since competition with the high levels of cellular ATP is often a major challenge for non-covalent compounds. The interaction of covalent irreversible inhibitors with their targets is time-dependent, and results in complete inactivation of the target.³ This is an advantage compared to non-covalent inhibitors, where the degree of target inhibition can vary. In addition, covalent binding results in a long duration of action of the inhibitor, especially if the target protein has a long half-life, since recovery of target activity is dependent on new protein synthesis. Practically, this translates into less frequent drug dosing.

Despite these advantages, design of covalent inhibitors often has been avoided because of the possible safety liability of these compounds due to their potential to react non-selectively, and covalently bind to off-target proteins and/or DNA, which could result in in vivo toxicity. The ability to assess potential off-target reactivity with biologically relevant nucleophiles such as glutathione is important to establishing confidence in the safety profile of the covalent inhibitor.⁴ Glutathione, which plays an important role in protection of cells from oxidative stress, is present in mM concentrations in cells. Glutathione binding of a compound results in compound excretion into the bile, and thus serves as an important route of elimination. In the assessment of drug candidates, conjugation of a compound to glutathione in buffer at physiological pH can be used as a biochemical assay to rank-order compounds based on reaction rates. Mass spectrometry or NMR approaches can be used to quantify glutathione adduct formation and compound half-life. This approach has the advantage of being quantitative, but it usually is not amenable to large numbers of compounds. In addition, this in vitro assay may not translate directly to compound reactivity with glutathione in blood or liver tissue.

Besides glutathione, cells contain additional relevant biological nucleophiles that can bind and act as cellular sensors for electrophilic compounds. The Nrf2 transcription factor is one such protein which functions as a key regulator in the detoxification pathway of electrophiles.^5,6 Nrf2 binds to the antioxidant response element (ARE) that is upstream of a large number of enzymes, including glutathione-S-transferase, heme oxidase, superoxide dismutase, and NADPH quinone oxidoreductase, all of which are involved in protecting cells from injury by electrophiles or oxidative stress.⁷ In non-stressed cells, Nrf2 is bound to Keap1, an adaptor protein for Nrf2 ubiquitination. This complex is located in the cytosol, but under conditions of stress or following electrophile binding, Keap1 is released and Nrf2 translocates to the nucleus, initiating gene transcription. Human Keap1 contains 25 cysteine residues, and mapping studies have identified a number of specific residues that are critical for compound induction of Nrf2 transcriptional activity.⁸ These results suggest that Keap1 is a direct electrophile sensor for cysteine-reactive compounds. In fact, a number of biologically important compounds activate this pathway including the active metabolite of acetaminophen, N-acetyl-p-benzoquinoneimine (NAPQI), dexamethasone 21-mesylate, prostaglandin J2 (15-deoxy-delta (12,14) prostaglandin J2), iodoacetamide and dinitrochlorobenzene.⁸ Compounds that cause skin sensitization act via formation of a covalent adduct with endogenous skin proteins, which results in an immune response.⁹ A reporter assay that detects the activation of Nrf2 in human keratinocytes, called KeratinoSens assay, has been utilized as a high throughput biosensor assay for compounds that cause skin sensitization.^10,11

In our drug discovery efforts to identify potent and safe covalent irreversible therapeutic agents, we have utilized a high throughput, plate-based assay using Ellman's reagent¹² to determine the reactivity of cysteine-reactive compounds with glutathione. In addition to this assessment of direct chemical reactivity, the safety liability of compounds was explored further using an Nrf2 reporter assay to assess the potential for compound reactivity with this biologically relevant nucleophile.

Experimental

Materials

Test compounds were purchased from Sigma-Aldrich (St. Louis, MO) or obtained from the Pfizer chemical bank (Groton, CT). Compounds were tested in dose–response fashion in concentrations from 300 μM to 150 nM. These concentrations cover a range of optimal compound solubility, and allow comparison of the results to other in vitro assays.¹³L-Glutathione in its reduced form and bovine serum albumin were purchased from Sigma-Aldrich. Human liver microsomes (catalog# 452156) were obtained from BD Biosciences (San Jose, CA).

Cell culture

The HEK-293 cell line was obtained from American Type Culture Collection (ATCC; Manassas, VA). HEK-293 cells were grown in Eagles Minimum Essential Medium (ATCC) supplemented with 10% (v/v) fetal bovine serum and 100 units ml⁻¹ penicillin and 100 μg ml⁻¹ streptomycin. Cells were maintained at 37 °C, 5% CO₂, 95% humidity.

Transduction and stable cell line selection

ARE (antioxidant response element) Cignal lentiviral construct and transducing reagents were purchased from SABiosciences (Frederick, MD). Cells were plated in 12-well plates at 2.5 × 10⁵ cells per well. Transductions were performed according to manufacturer's directions, using 20 μl of lentiviral particles and 8 μM concentration of Sureentry (SABiosciences) transfection reagent. Stable cell lines were selected using 1 μg ml⁻¹ puromycin (Sigma, St. Louis, MO). Single cells were isolated from Polyclonal cell lines using a FACS Vantage Cell Sorter (BD, Franklin Lakes, NJ), and expanded.

Nrf2 assay

Transduced cells were plated in 384-well plates at 2000 cells per well. After overnight incubation, the medium was replaced with fresh growth medium containing various concentrations of test compound (300 μM–150 nM). Cells were incubated for an additional 24 hours, after which luciferase activity and cell viability were determined using Steady-Glo Luciferase Assay System (Promega, Madison, WI) and Cell Titer-Glo® (Promega, Madison, WI), respectively. Luminescence was measured on an EnVision 2103 Multilabel Reader (Perkin-Elmer, Waltham, MA).

Ellman's reagent assay

Stock solutions of the test compounds, prepared in DMSO, were diluted to 600 μM in 400 mM Tris, pH 8.5. Reduced glutathione (400 μM) and bovine serum albumin (BSA) (448 μM) solutions were prepared fresh in 400 mM Tris, pH 8.5. Human liver microsomes were diluted to 2 mg ml⁻¹ in 400 mM Tris, pH 8.5. 5-5′-Dithiobis-2-nitrobenzoic acid (DTNB, Ellman's reagent) was prepared fresh in methanol at a concentration of 20 mM and kept away from light.

In order to choose a suitable concentration of glutathione to use for the Ellman's reagent assay, a standard curve for glutathione was set up in a 96-well plate by adding 50 μl of glutathione at various concentrations (0–500 μM) in 400 mM Tris, pH 8.5 to 50 μl of 400 mM Tris, pH 8.5 followed by the addition of 100 μl of 0.5 mM DTNB in 400 mM Tris, 1 mM EDTA, pH 10. The absorbance of TNB was read at 412 nm in a Spectramax plate reader and was found to be proportional to the glutathione concentration (data not shown). N-Ethylmaleimide (NEM) was then used as a positive control to determine its reactivity with glutathione at three glutathione : NEM ratios. We found that the smaller the glutathione : NEM ratio, the greater the sensitivity of detection of glutathione that had reacted with NEM (data not shown). Hence, we chose to use a glutathione concentration of 400 μM for subsequent testing of compounds so that the assay had a good sensitivity of detection as well as a large absorbance window (A₄₁₂ = 0.6). A similar experimental setup was performed with BSA as well as with human liver microsomes to choose a suitable concentration of these two thiol sources for the assay.

Test compounds (50 μl) were incubated with 400 μM glutathione (50 μl) in 96-well plates at room temperature. For the experiment in Fig. 4, glutathione was replaced with either 448 μM BSA (50 μl) or 2 mg ml⁻¹ human liver microsomes (HLM) (50 μl). DMSO, at a concentration of 2% (v/v), in 400 mM Tris, pH 8.5 was used as the vehicle. Reactions were carried out at room temperature and DTNB was used to determine the thiol concentration present in the glutathione, BSA and human liver microsomes assays at T₀, T₆₀ and T₁₈₀ minutes. Thiols react with DTNB to form a mixed disulfide and 2-nitro-5-thiobenzoic acid (TNB), and the concentration of TNB can be measured spectrophotometrically. DTNB was diluted to a final concentration of 0.5 mM in 400 mM Tris, 1 mM EDTA, pH 10, and 100 μl added to each well. The absorbance of TNB was read at 412 nm in a SpectraMax plate reader (Sunnyvale, CA).

Results

Electrophilic reactivity

The potential safety liability of the test set was examined in assays that assessed the electrophilic reactivity of the compounds. Direct chemical reactivity with glutathione was measured with Ellman's reagent¹² with some modifications, while an Nrf2 reporter assay was used to assess compound electrophilicity within a more biologically-relevant cellular context. In addition, a subset of compounds was assessed in the Ellman's reagent assay for direct chemical reactivity against solutions of BSA and HLM. Fig. 1 shows the compounds tested and the diverse chemical space they cover. Compounds 1–8 were used in the development and validation of the Ellman's assay and Nrf2 assay, and the test set of compounds focused on a variety of Michael acceptors, some of which explored the impact of N-substitution of acrylamide (9–17). These results were compared to the activity of well-known thiophiles and marketed covalent drugs (18–25).

Fig. 1 Structure of the test compounds.

The Ellman's reagent assay was chosen for glutathione reactivity because it is a high throughput assay with a spectrometry-based readout that allows for straightforward testing of large numbers of compounds. Compounds (300 μM) were incubated with 200 μM glutathione at pH 7.4 over a 5 hour time course and the fraction of glutathione remaining was measured with DTNB. As the ultimate aim of the study was to establish the thiol reactivity of Michael acceptors, the first 5 compounds were chosen to aid in assay development and explore the assay sensitivity. N-Ethylmaleimide (NEM) 1 is a reactive electrophile which rapidly quenched almost all glutathione within 30 minutes (Fig. 2A). Only weak activity was observed for the other 4 compounds. In order to increase the concentration of reactive thiol anion, the pH of the assay buffer was increased to 8.4 and enhanced rates of reaction were observed for compounds 2–5 (Fig. 2B). All subsequent assays were conducted at this higher pH.

Fig. 2 Compound reactivity with glutathione. Compounds 1, 2, 4, and 5 (300 μM) were incubated with 200 μM glutathione at pH 7.4 (A) or pH 8.5 (B), and after various times of incubation at 37 °C, the fraction of glutathione remaining was measured as described in Methods.

N-Phenylacrylamide 2 decreased the level of glutathione by 28% after 3 hours incubation whereas the benzyl analogue 3 possessed no glutathione reactivity (Table 1). The cyanamide 4 was relatively inactive whereas the alkyne 5 reduced glutathione levels such that only 33% remained at the 3 hour time point.

Table 1 Glutathione and Nrf2 reactivity assays^a

Compound	Fraction of GSH remaining after 1 h	Fraction of GSH remaining after 3 h	NRF2 dose ≥2-fold	Max. fold induction
a All results are shown as the mean of triplicate determinations in 2–3 separate experiments. Compounds (300 μM with the exception of 18 and 19 which were tested at 100 μM) were tested in the Ellman's assay and the fraction of glutathione remaining after incubation for 1 and 3 hours was measured. Compounds were tested for induction of Nrf2 reporter activity, and the lowest concentration (μM) at which >2-fold induction of activity was observed, and the maximum induction over the compound concentration-response curve are reported. NE: Nrf2: no effect, or <2-fold induction; Ellman's: no effect or >0.97. Tox: cytotoxicity was observed at all tested concentrations.
1 (NEM)	0.02	0.02 Remaining Glu a	Tox	Tox
2 (Phenyl acrylamide)	0.88	0.72	150	2.4
3 (Benzyl acrylamide)	NE	NE	NE	NE
4	0.94	0.92	NE	NE
5	0.61	0.33	NE	NE
6 (TBHQ)	0.14	0.14	0.15	9.1
7 (Sulforaphane)	0.32	0.24	0.15	2.2
8 (D3T)	0.76	0.57	9.4	8.0
9	0.69	0.40	NE	NE
10	0.84	0.67	38	7.5
11	0.68	0.45	Tox	Tox
12	0.07	0.04	NE	NE
13	0.94	0.80	300	2.8
14	0.25	0.05	0.59	2.6
15	NE	0.88	NE	NE
16	NE	NE	NE	NE
17	NE	NE	NE	NE
18 (Canertinib)	NE	NE	NE	NE
19 (Ibrutinib)	NE	NE	NE	NE
20 (Omeprazole)	0.40	0.27	2.3	10.5
21 (Pantoprazole) (3antop(panotprazole))	0.93	0.86	NE	1.9
22 (Penicillin)	NE	NE	NE	NE
23 (Oxazolone)	NE	0.96	0.15	2.6
24 (Indole-3-carbinol)	0.97	0.96	2.3	4.9
25 (Benzofuroxan)	0.52	0.29	9.4	6.8

---

Tetra-butyl hydroquinone (TBHQ) 6, sulforaphane 7 and 1,2-dithiole-3-thione (D3T) 8 served as positive controls for the Nrf2 assay. Compounds were tested at concentrations from 300 μM to 100 nM, and the lowest concentration resulting in a ≥2-fold induction of luciferase activity, as well as the maximum fold induction, were measured (Table 1). To normalize for cytotoxic effects of the compounds, cell viability was measured and only effects at viability ≥80% were considered significant. TBHQ caused an approximate 3-fold induction at 150 nM, with a maximum induction of 9-fold at 19 μM. D3T treatment resulted in a maximum 8-fold induction at 150 μM (Fig. 3).

Fig. 3 Nrf2 induction by TBHQ (A) and D3T (B). HEK-293 cells were treated with various concentrations of the compounds for 24 hours and Nrf2 beta lactamase activity and cell viability were measured as described in Methods.

Consistent with the thiol sensitivity of the Nrf2 assay, TBHQ and sulforaphane also showed marked effects in the glutathione assay, decreasing glutathione levels 76–86% at the 3 hour time point. Despite the strong induction in the Nrf2 assay by D3T, this compound showed less reactivity in the direct glutathione assay, decreasing glutathione levels by 43% at 3 h. Out of the preliminary set of Michael acceptors, NEM was cytotoxic to the cells at all concentrations and only phenyl acrylamide 2 caused a >2 fold induction but at a high concentration, 150 μM. The vinyl sulfone 9 showed significant reactivity with glutathione, but had no effect in the Nrf2 assay, while the alkyne 10 was reactive in both assays, resulting in a 7.5-fold induction of luciferase activity.

The structure-activity relationship associated with the activity of the acrylamide 2 was explored further with compounds 11–19. Compounds 11 and 12 are pyridine analogues of 2 and were expected to have greater activity owing to the electron-withdrawing nature of the pyridine. Consistent with this hypothesis, glutathione levels were decreased by 55% and 96% by compounds 11 and 12, respectively. Compound 12 differs from 11 by the addition of a pyridone ring, which itself may be an electrophile and thus react with glutathione. Both compounds 11 and 12 were cytotoxic in the HEK-293 cells at all concentrations tested and thus effects on the reporter assay could not be measured. The heteroaromatic compound 13 was not cytotoxic and induced Nrf2 activity by 2.8-fold at 300 μM yet had little effect on glutathione levels. Compound 14 caused an almost complete depletion of glutathione levels by 3 h, and also showed potent stimulation of Nrf2 activity, with a 2.6-fold increase in reporter activity at <1 μM.

Compounds 15–19 represent more ‘drug-like’ examples that were included in order to understand the impact on reactivity when the cysteine-reactive group is attached to a larger structural motif. Compound 15 showed only a low level of reactivity with glutathione (88% remaining at 3 h), and induced Nrf2 activity by only 1.6 fold. Although both compounds 16 and compound 10 contain the same alkyne Michael acceptor, the structurally larger 16 did not react with glutathione nor did it have any effect in the Nrf2 assay. Similarly, compound 17 contains a vinyl sulfone as does compound 9, yet 17 had no effect in either assay. These results show that the presence of the Michael acceptor alone is not sufficient to predict thiol reactivity, particularly at the concentrations used in these assays. The kinase inhibitors, canertinib 18 and the racemic form of ibrutinib, 19, are acrylamide-containing compounds. Canertinib and ibrutinib have been tested in clinical trials. These compounds inhibit their target kinases by binding to a key cysteine residue in the active site of the kinase. Canertinib inhibits the EGFR kinase while ibrutinib is a BTK inhibitor.^14,15 For these compounds, binding is covalent, resulting in irreversible inhibition of the targets. Except for 15, none of the compounds in this group reacted with glutathione, nor did they have any effect in the Nrf2 assay. These results demonstrate that although the kinase inhibitors are potent inhibitors and show in vivo efficacy against their respective targets, they are not promiscuous with respect to thiol reactivity, as evidenced by their lack of reactivity with glutathione, as well as in the Nrf2 assay.

Compounds 21–25 were included to appreciate the full range of values that could be obtained from these assays and also to understand applicability of the assay with respect to the detection of potent electrophiles or known Nrf2 inducers. Omeprazole 20 and pantoprazole 21 are proton pump inhibitors that target the gastric H⁺, K⁺-ATPase.¹⁶ They are activated to thiophilic cations under acidic conditions which react with cysteines of the ATPase and covalently inhibit the enzyme. Omeprazole incubation with glutathione for 3 h resulted in a 73% decrease in glutathione levels, and in the Nrf2 assay, caused a 2.8-fold increase in luciferase activity at 2.8 μM. In contrast, the structurally related pantoprazole showed little reactivity with glutathione and only induced Nrf2 activity by 1.9-fold at 300 μM. This difference in reactivity may be explained by the presence of the electron-withdrawing fluorine in pantoprazole which decreases the tendency of the benzimidazole to protonate, thus slowing the formation of the active rearrangement product relative to the more basic omeprazole. This would result in a lower reactivity with glutathione, as observed.

Treatment in humans with the beta-lactam antibiotic penicillin 22 has been shown to result in drug hypersensitivity owing to the tendency of penicillin to bind covalently to proteins, leading to an allergic reaction. The beta lactam ring is able to acylate the nucleophilic Lys, Cys or His side chains in proteins but penicillin did not react with glutathione, nor did it induce Nrf2 activity. These results are consistent with previous studies that have demonstrated that the class of beta-lactam antibiotics only react at a slow rate with cysteine thiols and that they are more likely to acylate lysine residues.¹⁷

2-Phenyl oxazolone 23 was chosen as it is a potent skin sensitizer yet, similarly to the beta-lactam antibiotics, studies suggest it is more likely to react with lysine rather than cysteine.¹⁸ In fact, 23 was not active in the Ellman's reagent assay but did promote a 2–3-fold induction of Nrf2 (Table 1). Similar induction was reported in the AREc32 breast carcinoma cell line, which carries a luciferase reporter gene under the control of eight copies of the antioxidant response element (ARE).¹¹ The inconsistency between Nrf2 induction and reactivity in the Ellman's assay may reflect the pH difference between the 2 assays. At pH 8.4, direct adduct reaction with glutathione is slow. Under the lower physiological pH conditions of the Nrf2 assay, the enol ether could hydrolyze and release a thiol-reactive aldehyde which may contribute to the observed induction.¹⁸ The 2-isopropyl analogue of 23 also was tested but was inactive in both assays (results not shown), further supporting hydrolysis of the enol ether as a requirement for Nrf2 reactivity.

Similarly, I3C (24) was not active in the Ellman's reagent assay, yet was a strong inducer of Nrf2. This supports the assumption that I3C is likely to undergo acid-catalysed transformation to a reactive unsaturated imine,¹⁹ a process which would be disfavored in the alkaline conditions of the modified Ellman's assay. Finally, benzofuroxan 25 is a well-known direct electrophilic thiophile²⁰ and was active in the Ellman's assay as well as inducing Nrf2 activity by approximately 7-fold.

Reactivity with BSA and HLM

Drugs that are administered orally can bind to plasma proteins, and also can undergo hepatic transformation and elimination. Thus, electrophilic compounds not only can react with glutathione, but they also potentially can bind covalently to proteins in blood and tissues. Glutathione is often thought to be the quantitatively dominant intracellular thiol. However, studies by Requejo and co-workers demonstrated that the concentration of exposed protein thiols in liver and heart tissue is significantly higher than that of cellular glutathione.²¹ Therefore, it was of interest to compare reactivity of the electrophilic compounds with BSA and HLM. The thiol reactivity of compounds 1, 2, 5, 10–12, and 25 were tested with BSA, or with HLM and the amount of remaining thiol in the solution was measured (Fig. 4). Compounds 1, 12 and 25 were most reactive with glutathione. NEM (1) was highly reactive with BSA and slightly less so with HLM (75% depletion). For compounds 12 and 25, the reactivity with BSA and HLM was considerably less than with the glutathione solution, showing that glutathione reactivity did not translate directly into reactivity with other biological thiols. Compound 5 was unusual in that the rate of thiol depletion in HLMs was similar to that of the glutathione solution, yet no activity was seen in BSA nor in Nrf2. Compounds 2, 10, 11 and 19 showed little to no reactivity with BSA and HLM.

Fig. 4 Compounds (300 μM) were incubated with glutathione, BSA or HLM as described in Methods, and after 3 hours, the fraction of remaining thiol was measured with Ellman's reagent as described in Methods.

Discussion

Irreversible inhibition through the covalent binding of a compound to a target offers the potential for high levels of potency and selectivity which can translate into low doses for therapeutic efficacy. Such covalent modification also can result in a long duration of action, depending on the target half-life, since re-synthesis of the drug target is required to restore activity.³ These attributes contribute to the pharmacological advantages of irreversible inhibitors compared to reversible inhibitors and, as of 2011, there were 39 marketed covalent drugs.¹ The covalent drugs clopidogrel, lansoprazole and esomeprazole were three of the top-selling drugs in the US in 2009, and all 3 are designated as blockbuster drugs. Omeprazole (20) and pantoprazole (21) are also in this ‘prazole’ class. Aspirin, another covalent drug, is the most widely used drug in the world. These examples demonstrate that drugs that act via a covalent mechanism of inhibition are not by definition unsafe.

These drugs cover a wide range of indications as well as mechanisms. Approximately one-third of the marketed covalent drugs are anti-infectives, with the remainder spread between oncology, gastrointestinal, central nervous system, and other areas. Covalent drugs are utilized for both acute and chronic indications, such as infection and acid reflux for example, respectively.¹ In general, the irreversible strategy utilizes an electrophilic functional group within the inhibitor that can covalently bind to the nucleophilic amino acids serine, lysine or cysteine in the target protein. However, the exact mechanism of inhibition varies.^2,3 For example, clopidogrel is a prodrug which targets the P2Y12 platelet receptor. It is activated by P450 enzymes and acts by forming a disulfide bridge with the platelet ADP receptor.²² The beta-lactam antibiotics undergo ring opening and covalently modify Lys, Cys or His side chains, and inactivate penicillin-binding proteins. Finasteride, used for prostate hypertrophy, is a mechanism-based inhibitor that reacts with the NADPH cofactor to produce a substrate analog that binds to and inhibits the target 5-alpha-reductase with nM affinity.²³ Although several compounds are in clinical trials, there are no marketed covalent drugs which act as direct Michael acceptors, resulting in inhibition of their target.

Despite the advantages of a covalent approach, safety considerations have been a concern. Although target binding is likely to be relatively specific, covalent inhibitors are inherently reactive and therefore non-specific or off-target reactivity may result in toxicity. Non-specific binding to proteins can result in hapten formation, as in the case of the beta-lactams, with subsequent antibody and hypersensitivity reactions.²⁴ Random, covalent binding to cellular macromolecules and proteins may cause direct tissue injury, although identifying the exact molecular species responsible for the toxicity can be challenging. For example, isoniazid, given in the treatment of tuberculosis, is a covalent inhibitor of enol-acyl carrier protein reductase. Isoniazid has been reported to cause liver injury including fulminant liver failure. However, the metabolite of isoniazid that is proposed to cause liver injury is different from the chemical species that is generated for covalent adduction.² Overall, the challenge is to find the optimal balance between reactivity and selectivity. To date, no screening assays exist that can predict these toxicities and thus, despite the therapeutic advantages of covalent drugs, development of this class lags behind that of reversible drugs.

At physiological pH, the thiol functional group is mostly protonated and is essentially unreactive to mild or weak electrophiles. The protonated thiol is reported to be over 1000-fold less reactive than its deprotonated counterpart, the thiolate anion.²⁵ Furthermore, the thiol in glutathione has a pK_a of 9.4²⁶ and is expected to react at physiological pH with only reactive radicals or highly electrophilic species such as NEM. Our results were consistent with this general lack of reactivity at pH 7.4, and even at pH 8.4, only half the compounds tested reacted with glutathione. In these assays, in order to maintain the linearity of the Ellman's reagent assay, 200 μM glutathione was the maximum concentration utilized. Thus the compound : glutathione ratio was approximately 1 : 1, which may limit reactivity to only highly electrophilic compounds. Higher ratios of glutathione to compound may be necessary to drive reactivity in this assay. The pK_a of many proteinous thiols may be up to 4 orders of magnitude lower, resulting in higher reactivity, compared to glutathione. Thus, compounds were tested for reactivity with BSA and HLM and also in the cellular Nrf2 reporter assay as a further means of determining their overall thiol reactivity.

In the Nrf2 assays, several of the compounds were cytotoxic and thus their ability to induce Nrf2 activity could not be assessed. The mechanism(s) underlying the toxicity are not known, but thiol reactivity could be a contributing factor, since compounds 1 and 12 showed high reactivity with glutathione and also were toxic. For the remaining compounds, reactivity with glutathione and in the Nrf2 assay generally were aligned, although the fold induction of Nrf2 reporter activity could not be predicted from glutathione reactivity. For example, the oxazolone 23, which showed little reactivity with glutathione, induced Nrf2 activity at a low concentration. Compound 14 only activated Nrf2 activity by 2.6-fold, yet it depleted glutathione levels by 95%. These results suggest that the factors important for conferring thiol reactivity in a cellular context, as in the Nrf2 assay, may be different than the direct chemical reactivity in a glutathione solution.

Compound reactivity with BSA or HLM was similar to glutathione reactivity, although the degree of thiol depletion was less for the BSA and HLM solutions. Compound incubations with HLM in the presence or absence of NADPH has been used to measure the contribution of metabolic activation in compound thiol reactivity.⁴ In our assays, we used BSA or HLM (without the addition of NADPH) to provide a measure of reactivity with cellular proteins where compound permeability into cells was not required.

Our results demonstrate that N-alkyl acrylamides, such as 3 and 19, are essentially unreactive across the assays used in this study. These results are consistent with a recent report of novel transglutaminase 2 irreversible inhibitors containing N-alkyl acrylamide which also showed no reactivity with glutathione over a 120 h incubation.²⁷ Acrylamides can be activated, however, by the presence of suitable aromatic groups, including a phenyl ring (as in 2) or furthermore by electron-withdrawing heterocycles (as in 11–15). For low-molecular weight analogues, reactivity can be expressed as follows: propynamides ≅ vinyl sulphones > N-aryl-acrylamides > cyanamides > N-alkyl-acrylamides. Such reactivity, however, does not necessarily translate to more complex, drug-like compounds and this is demonstrated by compounds 15–17 which were inactive in the Ellman's assay, despite containing electron withdrawing N-substituents. Such loss in reactivity could be due to steric issues at the low concentrations applied in the assay. Similarly, when within series that use the same irreversible group, the rate of Nrf2 induction also may be influenced by size and/or steric issues and this is supported by the alkynes 5 and 10, both of which are glutathione reactive but the larger derivative 5 does not induce Nrf2. Attempts to model computationally the reactivity of acrylamides was hampered by the small dataset of glutathione-reactive examples. Nevertheless, the order of reactivity for acrylamides 2, 11, 12 and 14 correlated very well with the order derived from their values of the energy of the Lowest Occupied Molecular Orbital (LUMO). For these cases, the a lower LUMO energy value was associated with greater glutathione reactivity (data not shown). In contrast, acrylamide 13 had the lowest LUMO energy but was the least reactive in the Ellman's assay. Further studies are ongoing to expand the dataset and provide more information so that a more robust model can be developed.

The high target selectivity and specificity associated with a covalent strategy makes it particularly attractive for the drug discovery of kinase inhibitors.²⁸ Historically, two of the major challenges in this field are selectivity, owing to the high degree of conserved amino acid sequences across the kinome, and the mM levels of intracellular ATP against which reversible inhibitors must compete. Some of the >500 mammalian protein kinases, however, contain a specific cysteine in or near the ATP binding pocket that can be utilized for irreversible binding.²⁹ These have been classified into 4 subgroups based on the relative location of the cysteine to the kinase ATP binding pocket.³⁰ Group 3 consists of 10 kinases, each with a cysteine positioned at the end of the C-lobe α-helix. The ErbB family of kinases are in this group, and potent compounds targeting EGFR and ErbB2 were the first covalent kinase inhibitors to be reported.³¹ Fry et al. showed that the acrylamide-containing PD-168393 bound to Cys 773 in EGFR with nM potency, inactivating the kinase.³² At least 10 covalent inhibitors directed against this family of kinases have been reported, and 4 of the compounds are undergoing clinical development.²⁹

Canertinib (18,CI-1033) is an irreversible EGFR, ErbB2, and ErbB4 inhibitor that is efficacious in animal models of cancer.³³ It is an acrylamide-containing compound that did not react with GSH nor did it induce Nrf2 activity (Table 1). It binds with nM affinity to its kinase targets, and co-crystalization studies demonstrate that the acrylamide binds to the Cys 773 of EGFR (and the analogous Cys 784 and Cys 778 in ErbB2 and ErbB4, respectively).³⁴ It shows good selectivity against other kinases, although results from extensive kinase testing have not been reported.³³ Canertinib was tested in Phase I and Phase II clinical trials and ultimately was dropped from further clinical development due to lack of efficacy and adverse side effects.³⁵ In the Phase I trials, the most common side effects were diarrhea (47%), rash (55%) and nausea (38%). Diarrhea and rash also are associated with treatment with erlotinib, gefitinib and lapatinib, and are known class effects associated with anti-cancer agents targeting ErbB kinase family members. Unlike canertinib, these other agents are reversible inhibitors, suggesting that those toxicities associated with canertinib may not be due solely to its activity as an irreversible inhibitor. However, at the highest doses in the Phase 1 study, >560 mg, 2/19 patients had grade 3 hypersensitivity, which could be the result of a hapten immune response.³⁶ In two Phase II studies totaling 303 patients, diarrhea, nausea and rash again were the most common adverse events, and no hypersensitivity effects were reported.^37,38

The TEC family of kinases (Bmx, Btk, Itk, Blk, TEC and Txk) also are included in the Group 3 cysteine kinases, and several groups have reported potent and selective covalent inhibitors of Btk with efficacy in animal models of disease.^15,39,40 Compound 19 was unreactive in both the glutathione and Nrf2 assays (Table 1) and this compound is the racemic form of Pharmacyclics PCI-32765. PCI-32765 binds to Cys 481 of Btk and inhibits the enzyme with an IC₅₀ of 0.5 nM. This compounds shows overall kinome selectivity although kinases with a cysteine homologous to Cys 481 are also targets of PCI-32765 and reported IC₅₀ values against EGFR, TEC and Itk are 12 nM, 1 nM and 12 nM, respectively, suggesting that biological effects resulting from inhibiting these other homologous kinases are possible.^28,41,42 PCI-32765 has been tested in clinical trials of non-Hodgkin's lymphoma and chronic lymphocytic leukemia and has been well tolerated with only grade 1 adverse events reported in one Phase 1 study with 56 patients. In a phase 1b/II trial in patients with relapsing/refractory chronic lymphocytic leukemia, the most frequently reported adverse events were grade 1 or 2 diarrhea, fatigue and nausea. Grade 3 adverse events occurred in 21% of the patients and were due to infectious complications. Response rates have been quite promising and progression-free survival at 6 months was 90% for patients receiving 840 mg PCI-32765, indicating efficacy and good tolerability.^15,42–44 This compound has a short half-life in humans (2–3 h) yet remains covalently bound to Btk for at least 24 h.⁴¹ This combination of brief exposure, thus limiting the duration of off-target effects, coupled with the duration of Btk inhibition may contribute to the safety and efficacy of this compound.

Overall, the results to date with the irreversible kinase inhibitors indicate that although the compounds contain reactive Michael acceptors that result in target inhibition, the compounds were unreactive in the assays described here, and furthermore, this reactivity does not appear to translate directly into adverse events in human clinical trials.

Conclusions

These studies demonstrate the utility of multiple assays for determining the glutathione reactivity and off-target thiol reactivity of cysteine-reactive compounds. In particular, testing compounds in the Nrf2 assay provides a means of assessing thiol reactivity in a cellular context. The results show that thiol promiscuity is a balance between the electronic properties of the covalent group and the size of the molecule. In the case of the covalent drugs, lack of reactivity with these additional biological thiol sensors is consistent with a high degree of selectivity for their biological targets.

Acknowledgements

We would like to thank Pfizer colleagues Drs Adam Gilbert, Nigel Greene and William Pennie, and Dr Denton Hoyer (Yale) for discussion and providing valuable feedback.

References

1. J. Singh, R. C. Petter, T. A. Baillie and A. Whitty, Nat. Rev. Drug Discovery, 2011, 10, 307.
2. M. H. Potashman and M. E. Duggan, J. Med. Chem., 2009, 52, 1231.
3. D. S. Johnson, E. Weerapana and B. F. Cravatt, Future Med. Chem., 2010, 2, 949.
4. A. S. Kalgutkar and M. T. Didiuk, Chem. Biodiversity, 2009, 6, 2115.
5. K. Itoh, J. Mimura and M. Yamamoto, Antioxid. Redox Signaling, 2010, 13, 1665.
6. K. R. Sekhar, G. Rachakonda and M. L. Freeman, Toxicol. Appl. Pharmacol., 2010, 244, 21.
7. H. Motohashi and M. Yamamoto, Trends Mol. Med., 2004, 10, 549.
8. M. Kobayashi, L. Li, N. Iwamoto, Y. Nakajima-Takagi, H. Kaneko, Y. Nakayama, M. Eguchi, Y. Wada, Y. Kumagai and M. Yamamoto, Mol. Cell. Biol., 2009, 29, 493.
9. A. Natsch, Toxicol. Sci., 2010, 113, 284.
10. R. Emter, G. Ellis and A. Natsch, Toxicol. Appl. Pharmacol., 2010, 245, 281.
11. A. Natsch and R. Emter, Toxicol. Sci., 2008, 102, 110.
12. G. L. Ellman, Arch. Biochem. Biophys., 1959, 82, 70.
13. N. Greene, M. D. Aleo, S. Louise-May, D. A. Price and Y. Will, Bioorg. Med. Chem. Lett., 2010, 20, 5308.
14. W. J. Slichenmyer, W. L. Elliott and D. W. Fry, Semin. Oncol., 2001, 28, 80.
15. T. Robak and E. Robak, Expert Opin. Investig. Drugs, 2012, 21, 921.
16. J. M. Shin and G. Sachs, Biochem. Pharmacol., 2004, 68, 2117.
17. A. Ariza, D. Garzon, D. R. Abanades, V. de los Rios, G. Vistoli, M. J. Torres, M. Carini, G. Aldini and D. Perez-Sala, J. Proteomics, 2012, 77, 504.
18. A. Natsch, H. Gfeller, F. Kuhn, T. Granier and D. W. Roberts, Chem. Res. Toxicol., 2010, 23, 1913.
19. G. L. Skiles and G. S. Yost, Chem. Res. Toxicol., 1996, 9, 291.
20. F. Willenbrock and K. Brocklehurst, Biochem. J., 1986, 238, 103.
21. R. Requejo, T. R. Hurd, N. J. Costa and M. P. Murphy, FEBS J., 2010, 277, 1465.
22. D. J. Angiolillo, Drugs, 2012, 72, 2087.
23. J. E. Drury, L. Di Costanzo, T. M. Penning and D. W. Christianson, J. Biol. Chem., 2009, 284, 19786.
24. D. A. Khan and R. Solensky, J. Allergy Clin. Immunol., 2010, 125, S126.
25. R. A. Bednar, Biochemistry, 1990, 29, 3684.
26. S. G. Tajc, B. S. Tolbert, R. Basavappa and B. L. Miller, J. Am. Chem. Soc., 2004, 126, 10508.
27. M. E. Prime, O. A. Andersen, J. J. Barker, M. A. Brooks, R. K. Cheng, I. Toogood-Johnson, S. M. Courtney, F. A. Brookfield, C. J. Yarnold, R. W. Marston, P. D. Johnson, S. F. Johnsen, J. J. Palfrey, D. Vaidya, S. Erfan, O. Ichihara, B. Felicetti, S. Palan, A. Pedret-Dunn, S. Schaertl, I. Sternberger, A. Ebneth, A. Scheel, D. Winkler, L. Toledo-Sherman, M. Beconi, D. Macdonald, I. Munoz-Sanjuan, C. Dominguez and J. Wityak, J. Med. Chem., 2012, 55, 1021.
28. Q. Liu, Y. Sabnis, Z. Zhao, T. Zhang, S. J. Buhrlage, L. H. Jones and N. S. Gray, Chem. Biol., 2013, 20, 146.
29. T. Barf and A. Kaptein, J. Med. Chem., 2012, 55, 6243.
30. J. Zhang, P. L. Yang and N. S. Gray, Nat. Rev. Cancer, 2009, 9, 28.
31. J. Singh, E. M. Dobrusin, D. W. Fry, T. Haske, A. Whitty and D. J. McNamara, J. Med. Chem., 1997, 40, 1130.
32. D. W. Fry, A. J. Bridges, W. A. Denny, A. Doherty, K. D. Greis, J. L. Hicks, K. E. Hook, P. R. Keller, W. R. Leopold, J. A. Loo, D. J. McNamara, J. M. Nelson, V. Sherwood, J. B. Smaill, S. Trumpp-Kallmeyer and E. M. Dobrusin, Proc. Natl. Acad. Sci. U. S. A., 1998, 95, 12022.
33. L. F. Allen, I. A. Eiseman, D. W. Fry and P. F. Lenehan, Semin. Oncol., 2003, 30, 65.
34. C. Carmi, M. Mor, P. G. Petronini and R. R. Alfieri, Biochem. Pharmacol., 2012, 84, 1388.
35. A. Ocana and E. Amir, Cancer Treat. Rev., 2009, 35, 685.
36. R. G. Zinner, J. Nemunaitis, I. Eiseman, H. J. Shin, S. C. Olson, J. Christensen, X. Huang, P. F. Lenehan, N. J. Donato and D. M. Shin, Clin. Cancer Res., 2007, 13, 3006.
37. S. Campos, O. Hamid, M. V. Seiden, A. Oza, M. Plante, R. K. Potkul, P. F. Lenehan, E. P. Kaldjian, M. L. Varterasian, C. Jordan, C. Charbonneau and H. Hirte, J. Clin. Oncol., 2005, 23, 5597.
38. O. Rixe, S. X. Franco, D. A. Yardley, S. R. Johnston, M. Martin, B. K. Arun, S. P. Letrent and H. S. Rugo, Cancer Chemother. Pharmacol., 2009, 64, 1139.
39. K. H. Kim, A. Maderna, M. E. Schnute, M. Hegen, S. Mohan, J. Miyashiro, L. Lin, E. Li, S. Keegan, J. Lussier, C. Wrocklage, C. L. Nickerson-Nutter, A. J. Wittwer, H. Soutter, N. Caspers, S. Han, R. Kurumbail, K. Dunussi-Joannopoulos, J. Douhan 3rd and A. Wissner, Bioorg. Med. Chem. Lett., 2011, 21, 6258.
40. Z. Pan, H. Scheerens, S. J. Li, B. E. Schultz, P. A. Sprengeler, L. C. Burrill, R. V. Mendonca, M. D. Sweeney, K. C. Scott, P. G. Grothaus, D. A. Jeffery, J. M. Spoerke, L. A. Honigberg, P. R. Young, S. A. Dalrymple and J. T. Palmer, ChemMedChem, 2007, 2, 58.
41. J. A. Burger and J. J. Buggy, Leuk. Lymphoma, 2013 DOI:10.3109/10428194.2013.777837.
42. L. A. Honigberg, A. M. Smith, M. Sirisawad, E. Verner, D. Loury, B. Chang, S. Li, Z. Pan, D. H. Thamm, R. A. Miller and J. J. Buggy, Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 13075.
43. S. O'Brien, J. A. B and S. E. Coutre, Ann. Oncol., 2011, 22, 123.
44. S. O'Brien, J. A. Burger, K. A. Blum, R. R. Furman, S. E. Coutre, J. Sharman, I. W. Flinn, B. Grant, N. A. Heerema, A. J. Johnson, T. Navarro, E. Holmgren, E. Hedrick and J. C. Byrd, ASH Annual Meeting Abstracts, 2011, 118, 983.

---