Reactive chemistry for covalent probe and therapeutic development

Abstract

Bioactive small molecules that form covalent bonds with a target protein are important tools for basic research and can be highly effective drugs. This review highlights reactive groups found in a collection of thiophilic and oxophilic drugs that mediate pharmacological activity through a covalent mechanism of action. We describe the application of advanced proteomic and bioanalytical methodologies for assessing selectivity of these covalent agents to guide and inspire the search for additional electrophiles suitable for covalent probe and therapeutic development. While the emphasis is on chemistry for modifying catalytic serine, threonine or cysteine residues, we devote a substantial fraction of the review to a collection of exploratory reactive groups of understudied residues on proteins.

Getting attached – embracing covalent bonds for chemical probe and therapeutic discovery

Covalent drugs perturb protein function by forming a specific bond between the ligand and an amino acid residue on a target protein. Several important medicines produce a therapeutic response through a covalent mechanism of action (MOA)^1,2. For example, the non-steroidal anti-inflammatory drug (NSAID) aspirin remains one of the most widely used medication worldwide. In the United States (US) alone, an astounding 40,000 metric tons (120 billion tablets) are consumed annually³. Other covalent drugs include the proton pump inhibitor omeprazole and the P2Y purinergic receptor 12 inhibitor clopidogrel that together accounted for ~70 million prescriptions in the US in 2019 (clincalc.com/DrugStats). The mounting clinical data is changing our view of covalent MOA from a perceived safety liability to a pharmacological strategy with distinct advantages⁴. While many of the approved covalent drugs were discovered serendipitously, recent examples of targeted covalent inhibitors (TCIs) support a path toward rational design^5–7.

The renewed interest in covalent drugs is motivated by distinct features of this class including the ability to achieve non-equilibrium blockade of a target resulting in high biochemical efficiency⁸, pharmacological activity that can outlast drug pharmacokinetics⁹, and access to challenging binding pockets on protein targets traditionally perceived as ‘undruggable’¹⁰. Covalent attachment is not intended to be a general solution but can be highly effective for applications when complete inactivation of a target is advantageous⁴. The beneficial features of covalent inhibitors (and drugs) are fully realized, however, when selectivity is optimized for covalent modification of a target protein while minimizing off-target protein-adduct formation. The latter scenario can lead to toxicity and potential idiosyncratic adverse drug reactions (see Glossary)¹¹.

The need for selectivity when developing and deploying covalent compounds is clear. Advances in chemical proteomic and bioanalytical technologies has enabled evaluation and medicinal chemistry optimization of covalent small molecule selectivity on a proteomic scale^12,13. This review aims to highlight reactive groups successfully implemented in thiophilic and oxophilic drugs that operate through a covalent MOA (irreversible and covalent reversible). When available, we describe the strategies employed to understand selectivity and MOA of electrophiles used in covalent drugs to offer insights to design principles for future inhibitor development. While the emphasis is on chemistry for modifying catalytic serine/threonine or cysteine residues, largely owing to the clinical success of targeting these nucleophilic residues^5,14, we devote a substantial fraction of the review to a collection of reactive groups with promising activity against less explored residues on proteins.

Thiophilic Drugs

The landmark approval of sotorasib (Lumakras^™) by the FDA in 2021 as a drug for treating non-small cell lung cancer highlights decades of research towards mutant-selective KRAS inhibition¹⁵. Mutations at glycine 12 (G12) impair guanosine triphosphate (GTP) hydrolysis and results in constitutive activation of KRAS¹⁵. The mutation of G12 to a cysteine (KRAS^G12C) is prevalent in cancer (~13% of lung adenocarcinoma; ~50% of mutant KRAS is p.G12C)¹⁵. The mutated cysteine site is located in proximity to an allosteric pocket under the switch II loop region (S-IIP) of KRAS¹⁶. Cysteine-reactive S-IIP KRAS^G12C inhibitors^17–19 and drugs^20,21 were developed to engage this transient pocket and covalently react with the guanosine diphosphate (GDP)-bound state of KRAS^G12C to trap this oncogene in its inactive conformation¹⁵ (Figure 1A). Notably, the KRAS^G12C covalent inhibitors are inactive against wild-type KRAS because of the absence of the reactive mutant cysteine site²⁰. Treatment with KRAS^G12C covalent inhibitors results in substantial depletion of active KRAS-GTP and inhibition of downstream RAS signaling in vitro^18,22 and in vivo¹⁹.

Figure 1. Thiophilic drugs and (seleno)cysteine reactive electrophiles.

(A) Progression of SAR studies, leading to sotorasib and the co-crystal structure of the sotorasib-Cys12 adduct on KRAS^G12C (PDB ID: 6OIM). (B) Chemical structure of ibrutinib (Imbruvica^®) and the co-crystal structure showing the ibrutinib-Cys481 adduct in BTK (PDB ID: 5P9J). (C) BODIPY (PCI-33380) and alkynyl (Probe 4) derivatives of ibrutinib as probes for ABPP (See Box 1). (D) Masked nitrile oxide electrophile reported by Schreiber et al. and the proposed MOA for selective labeling of selenocysteine.

Shokat et al. initiated the search for covalent inhibitors of KRAS^G12C in 2013 using disulfide tethering to detect novel allosteric pockets and identify vinyl sulfonamides and acrylamide inhibitors of KRAS^G12C (Ref. [16]). Medicinal chemistry optimization further enhanced potency of initial hit compounds to produce ARS-853 (Figure 1A), an acrylamide-based covalent modifier (via Michael addition) of KRAS^G12C that reduced the level of GTP-bound KRAS^G12C by greater than 95% in KRAS^G12C-mutant lung cancer cells^18,22. To determine cellular activity of ARS-853 with KRAS^G12C, Liu et al. developed a liquid chromatography-tandem mass spectrometry (LC-MS/MS) peptide assay that measures loss of G12C-containing tryptic peptide abundances as a readout of target engagement¹⁸. Proteome-wide selectivity of ARS-853 was assessed by chemical proteomics using a general cysteine-reactive probe, iodoacetamide-desthiobiotin (Figure 2; see Box 1 for an introduction to activity-based protein profiling (ABPP) methodology) to identify KRAS^G12C as the most potent target along with a handful of off-targets across >1,500 detected proteins (~2,700 cysteine sites)¹⁸. Subsequent proteomics combined with structure-activity relationship (SAR) resulted in the potent KRAS^G12C inhibitor ARS-1620 that bound more strongly due to noncovalent orientation from its quinazoline core and a biphenyl moiety locked as an (S)-atropisomer¹⁹, which inspired and ultimately led to the development of AMG 510 or sotorasib²⁰ (Figure 1A).

Figure 2. Schematic of activity-based protein profiling.

(ABPP) using gel electrophoresis (A) or LC-MS/MS detection (B). (C) Structures of reactive groups, multiplexing strategies, and types of ABPs used for ABPP analysis. SILAC: stable-isotope labeling using amino acids in cell culture; CuAAC: copper-catalyzed azide-alkyne [3+2]-cycloaddition; LC-MS/MS: liquid chromatography-tandem mass spectrometry. Samples analyzed are typically derived from studies of compound/probe treatments in lysates, cells, or animals. Clinical samples can also be subjected to this approach. See Box 1 for the supporting text to introduce ABPP methodology.

Box 1. Introduction to activity-based protein profiling (ABPP).

ABPP serves as a powerful methodology for functional analysis of proteins in complex biological systems⁵¹. Selective labeling of active enzymes and proteins is achieved principally through detection mediated by covalent activity-based probes (ABPs, Figure 2). The reactive group (also referred to as the electrophile) of the ABP facilitates covalent modification of a nucleophilic amino acid site(s) on the target protein. The recognition (binding/affinity) element, which can also include a linker for separating the reactive- and reporter-group, is designed to direct ABPs to proteins of interest and can be used for class-specific chemical proteomics.

ABP labeling of proteins was originally developed to target the active sites of enzymes through covalent modification of a catalytic serine/threonine⁵⁰ or cysteine residue¹¹³. The expansion of ABPP investigations to non-catalytic residues, and thus extending the scope of functional sites on proteins addressable by chemical proteomics, was facilitated by identification of electrophiles with desirable reactivity and amino acid chemoselectivity. Reactivity ABPs have been developed for diverse residues including cysteine¹¹⁴, lysine^13,110,111, methionine¹¹⁵, aspartate/glutamate^116–118, and tyrosine¹¹⁹. A description of different classes of ABPs and their applications are described in the following comprehensive review [51].

The competitive assay format measures the ability of candidate electrophilic compounds (fragments, TCIs, natural products) to block probe labeling⁵¹. In-gel fluorescence assays can be rapidly deployed for parallel evaluation of potency and selectivity to identify SAR for medicinal chemistry optimization. Biotin or desthiobiotin tagged ABPs facilitate enrichment by avidin affinity chromatography followed by LC-MS/MS identification of ABP-modified proteins. Protease digestion prior to affinity chromatography enriches for ABP-modified peptides derived from target proteins to quantify (when combined with stable-isotope labeling using amino acids in cell culture (SILAC)⁵¹ or tandem mass tag (TMT) isotopic labeling¹²) the site-specific activity of covalent compounds. The highly multiplexed capabilities of TMT labeling has increased the throughput of LC-MS/MS ABPP studies to enable screening of large electrophilic compound libraries¹².

An alkyne group can be embedded into the scaffold of covalent inhibitors with minimal impact on physicochemical/steric properties to facilitate direct proteome-wide assessment of selectivity irrespective of protein class or function. Inclusion of an alkyne handle on the probe permits click chemistry, i.e., copper(I)-catalyzed azide-alkyne [3+2] cycloaddition (CuAAC¹²⁰) derived from the Huisgen reaction. Thus, an azide-fluorophore or azide-biotin can be conjugated to an alkyne-modified clickable analog ex vivo to allow an efficient and modular biorthogonal reaction for detecting proteins modified by covalent inhibitors from treated cells and animals⁵¹.

Several FDA-approved drugs for cancer treatment rely on covalent modification of non-catalytic cysteines in kinases^14,23, such as ibrutinib (Imbruvica^®, Figure 1B), afatinib (Gilotrif^®), dacomitinib (Vizimpro^®), osimertinib (Tagrisso^®), and neratinib (Nerlynx^®). A covalent strategy provides several advantages over traditional reversible, ATP-competitive kinase inhibitors including prolonged and more productive target binding in vivo by minimizing competition with high intracellular ATP concentrations^24,25. The variable number, location and degree of conservation of kinase active site cysteines can also be exploited for selectivity²⁵. Selectivity profiling of ibrutinib was determined by developing a fluorescently tagged analog (Figure 1C, PCI-33380) through attachment of a Bodipy fluorophore directly to the acrylamide group to evaluate target engagement and selectivity in cellular and animal experiments²⁶. Although PCI-33380 was shown to engage only Bruton’s tyrosine kinase (BTK) in B-cell proteomes, a separate study developed a clickable ibrutinib analog that incorporated an alkyne group (Probe 4, Figure 1C) as a latent reporter tag for minimal impact (as opposed to the bulky Bodipy dye) on structure and activity²⁷. Importantly, the acrylamide on the clickable ibrutinib probe was not directly modified; reactivity and the resulting selectivity assessment can be impacted if the reactive group is altered. These selectivity profiling studies revealed several off-targets (~14 in total) in cellular probe labeling studies, which is not surprising given the nucleophilic nature of the thiolate side chain of cysteine found on kinases and other proteins. See Box 1 for additional details on employing clickable analogs for direct evaluation of selectivity by chemical proteomics.

The prevalence of the acrylamide reactive group in kinase drugs is likely due to a combination of its small size to access protein pockets, hydrogen bonding capability, and its tempered reactivity compared with other cysteine-reactive electrophiles²⁸. Chemical proteomic studies, however, have shown that even highly optimized acrylamide kinase drugs exhibit off-target activity when evaluated in live cells and animal models^27,29. Medicinal chemistry optimization combined with comprehensive selectivity profiling (e.g. using ABPP, Box 1 and Figure 2) is a path for mitigating off-target activity of acrylamide compounds³⁰. The development of electrophiles with tunable cysteine reactivity offers additional opportunities for achieving selectivity³¹. The discovery of masked nitrile-oxide electrophiles (Figure 1D) revealed a prodrug strategy for selective modification of the active site selenocysteine of the lipid hydroperoxide reductase GPX4³². Further exploration of masked electrophiles (including nitroisoxazoles and α-nitroketoximes), metabolically labile electrophilic groups³³, or covalent reversible scaffolds will provide additional opportunities in the future for targeting cysteine and the more nucleophilic selenocysteine counterpart^34,35.

Oxophilic Drugs

The slow, but reversible, bond formation between boron and a nucleophilic oxygen (Figure 3A) has been exploited in medicinal chemistry for over 40 years³⁶. Boron-containing moieties, however, can also contribute to non-covalent interactions. In an intriguing case, Raines et al. performed a SAR on the anilinosulfonyl portion of darunavir (Prezista^®), an FDA-approved drug for HIV, and identified the boronic acid derivative was a more potent inhibitor of HIV-1 protease, likely due to the low-barrier hydrogen bond between the boronic acid and Asp30 (Figure 3B). Importantly, the lack of any covalent bond between the boronic acid and HIV-1 protease in a co-crystal structure³⁷ (PDB ID: 6C8X) suggests that boronic acids can function as isosteres of anilino groups. There are currently five FDA-approved drugs (Figure 3C and D) available to treat human ailments ranging from cancer, e.g., bortezomib³⁸ and ixazomib citrate³⁹, to bacterial or fungal infection, e.g., vaborbactam⁴⁰, tavaborole⁴¹, and crisaborole⁴², the latter of which were approved within the past six years and represent a novel class of cyclic boron acid-based therapeutics⁴³. Whereas boronic acids rely on boron’s oxophilicity, which can be tuned by adjusting the electronics of substituent(s) on boron, cyclic boron acid-moieties have the added benefit of ring-strain release upon formation of the boron-oxygen bond between serine, threonine, cis-diols (Figure 3E), or activated water (Figure 3F)^44,45. The success of boron-containing compounds for developing therapeutic agents has incited further exploration of this compound class as exemplified by the suite of boronic acid- and cyclic boronic acid-containing drugs and their biological targets⁴⁶. The reversible bond formed between boron and oxygen can limit the efficacy of boron-containing therapeutics as they can be outcompeted by more potent, natural ligands, an issue that may be resolved by installation of an electrophile capable of irreversible modification of the target amino acid site.

Figure 3. Oxophilic drugs and mechanisms of reactivity.

(A) General reaction between phenylboronic acid and water. (B) Phenylboronic acid derivative of darunavir and a schematic representation of the low-barrier hydrogen bond with Asp30 in HIV-1 protease. (C) FDA-approved drugs containing boron. (D) Co-crystal structure of ixazomib-Thr1 adduct in human 20S proteasome (PDB ID: 5LF7). (E) Co-crystal structure of tavaborole-cAMP adduct in leucyl tRNA (tRNA^Leu) synthetase (PDB ID: 2V0C). (F) Co-crystal structure of crisaborole derivative an2898 adducted to an activated water molecule (H₂O*), which is more nucleophilic due to coordination of the oxygen in water with Mg²⁺ and Zn²⁺, in phosphodiesterase 4 (PDE4; PDB ID: 3O0J). Purple sphere: Zn²⁺; Red sphere: H₂O; Yellow sphere: Mg²⁺. (G) Skeletal drawing of orlistat and the suggested mechanism for covalent inhibition of fatty acid synthase (FASN).

One of the oldest and most widely used covalent modifier of the human proteome is acetylsalicylic acid, the active ingredient in the over-the-counter NSAID drug Aspirin^®, which acetylates COX-1 (S530) and COX-2 (S516)⁴⁷. Aspirin^® is an electrophilic fragment and selectivity profiling using chemical proteomics and alkyne-modified analogs revealed hundreds of additional modified protein targets and broad reactivity against other nucleophilic residues including cysteine, lysine, and tyrosine^47,48. Another oxophilic drug, rivastigmine contains an aryl carbamate group for carbamoylation of the active site serine of acetylcholinesterase (AChE) that effectively (because of slow turnover of adduct) leads to irreversible inhibition of acetylcholinesterase (AChE) for treatment of Alzheimer’s disease^49,50.

Besides tuning electronics, reactivity of covalent inhibitors can be enhanced by ring-strain release via ring-opening reactions. Orlistat is a β-lactone that mediates pharmacological activity through inhibition of pancreatic lipases in the gastrointestinal tract to reduce fat absorption for treatment of obesity⁵⁰. Additional studies have revealed that orlistat targets additional SHs including ABHD12⁵¹ and fatty acid synthase (FASN)⁵². Orlistat binds irreversibly to human FASN through covalent modification of the catalytic serine (S2308). The resulting ester adduct formed between S2308 and the ring-opened β-lactone in the acyl-enzyme intermediate permits packing of the hexanoyl tail of orlistat against H2481 (Figure 3G). This interaction may prevent activation of water for nucleophilic attack to hydrolyze the FASN-orlistat intermediate, which is likely to occur given sufficient time⁵². β-Lactamase inhibitors, such as the FDA-approved antibiotics clavulanic acid, sulbactam, and tazobactam, operate in a similar fashion as orlistat. Ring opening of the β-lactam by a catalytic serine is responsible for inhibition, which is at first surprising due to the canonical stability of amides; however, four-membered rings are relatively unstable and ring-opening, as well as recyclization⁵³, is facilitated by the residence time of drugs in an active site as well as the increased nucleophilicity of a catalytic serine or cysteine⁵⁴.

Exploratory reactive groups for covalent ligand development

A common feature of TCIs that have progressed to approved drugs is the presence of an optimized recognition element that facilitates a specific, non-covalent orientation within the binding site for selective covalent modification, although this is not always the case, e.g. dimethyl fumarate (Tecfidera^®) and Aspirin^®. A testable path toward selective modification of an intended amino acid site is the combination of a tempered electrophile with a high affinity binding element to achieve high effective concentrations in a protein site. The selection of an appropriate reactive group is important to avoid general alkylation or acylation of the most accessible (or nucleophilic) sites. We focus our discussion on reactive groups that have led to development of selective inhibitors or show promising activity (and tunability) against less explored amino acid residues. Additional electrophiles for covalent probe development can be found in published reviews^55,56 or studies^13,57.

Serine-reactive groups

The success of covalent drugs (e.g. rivastigmine) that modify the catalytic serine to inactivate serine hydrolases (SHs) implicated in human diseases led to a broader search for electrophiles suitable for development of TCIs⁵⁰. SHs are a large and diverse family of enzymes that constitute approximately 1% of the human proteome (~240 human members), which can be further divided into serine proteases and metabolic serine hydrolases^58,59. SHs carry out numerous biochemical functions in mammalian physiology and disease by serving as lipases, amidases, esterases, and proteases. SHs adopt an α,β-hydrolase fold and the enhanced nucleophilicity of the base-activated catalytic serine (as part of the canonical serine-histidine-aspartate catalytic triad) makes this residue susceptible to covalent modification by electrophiles⁵⁰.

Screening O-aryl carbamates against SHs in mouse brain proteomes by competitive ABPP (see Box 1) resulted in discovery of lead monoacylglycerol lipase (MAGL) inhibitors⁶⁰. MAGL is responsible for hydrolyzing the principal endocannabinoid lipid, 2-arachidonoylglycerol (2-AG)⁶¹. Through several rounds of ABPP-guided medicinal chemistry, a highly potent MAGL inhibitor JZL-184 (Figure 4A) was developed and confirmed to selectively engage MAGL in vivo by gel- and LC-MS/MS-based ABPP⁶⁰. Application of JZL-184 as a chemical probe helped elucidate the role of MAGL in inflammation, pain sensation, anxiety, memory, and cancer⁶². The JZL-184 carbamate scaffold was expanded to N-hydroxysuccinimidyl (NHS)⁶³ and O-hexafluoroisopropyl (HIFP) carbamates⁶⁴ to improve proteome-wide selectivity and drug-like properties. These efforts ultimately led to the clinical candidate ABX-1431 that showed high potency, selectivity, and oral bioavailability⁶⁵ (Figure 4A).

Figure 4. Oxophilic reactive groups in drugs and chemical probes.

(A) Progression of the MAGL inhibitor JZL-184 to HFIP-carbamates KML-29 and ABX-1431; the latter analog is currently in clinical evaluation for treatment of neurological disorders. (B) Left panel: general reaction scheme between a 1,2,3-triazole-urea and a nucleophilic alcohol (found on serine side chain group) that forms a Michaelis-Menten complex followed by release of the triazole to produce a carbamate adduct group (AG). Right panel: trend for tuning the reactivity of heterocyclic-ureas for developing serine hydrolase inhibitors.

SHs can also be irreversibly inactivated by 1,2,3-triazole ureas (Figure 4B)⁶⁶. This electrophile generally shows higher reactivity than carbamate counterparts but retains an appropriate degree of chemoselectivity to avoid cross-reactivity outside of the SH class⁶⁶. Importantly, the reactivity of 1,2,3-triazole ureas is sensitive to the position of the nitrogen(s) in the heterocyclic leaving group (LG)⁶⁷ as well as functional modifications that can alter the acidity⁶⁶ (Figure 4B). Selective 1,2,3-triazole ureas inhibitors of SHs have been reported with several compounds displaying high potency in cells (<1 nM) and animals (<1 mg kg⁻¹)^66,68. The range of SHs amenable to covalent inactivation by 1,2,3- and 1,2,4-triazole ureas is broad and includes peptidases (APEH⁶⁶), lipases (PAFAH2⁶⁶, ABHD6⁶⁹, DAGLα⁷⁰, DAGLβ⁷¹, ABHD2⁷²), poorly characterized enzymes (ABHD11⁶⁶, LYPLAL1⁷³), as well as SH targets in Staphylococcus aureus⁷⁴ and Toxoplasma gondii⁷⁵. Recently, N-acyl pyrazole ureas and carbamates were explored for development of SH inhibitors and found to exhibit tunable reactivity through functional group modifications on the acyl- and pyrazole-group resulting in identification of potent, covalent inhibitors of the anandamide hydrolase FAAH⁷⁶.

Sulfur(VI)-fluoride exchange (SuFEx) chemistry for targeting diverse residues

Sharpless and coworkers identified the sulfonyl fluoride (−SO₂F) group as a tempered electrophile that reacts when the fluoride ion is properly activated or solvated⁷⁷. In contrast with other sulfonyl halides, the unique stability-reactivity features of the SO₂F group, including e.g. resistance to reduction, thermodynamic stability, and exclusive reaction at sulfur, is important for its ability to facilitate reaction with protein sites through sulfur(VI) fluoride exchange (SuFEx, Figure 5A)⁷⁷. The additional requirement of hydrogen bonding-assisted release of the fluoride in the microenvironment of a protein site limits acid-base chemistry in physiological conditions, thereby permitting the binding element of sulfonyl fluorides to noncovalently interact with the target protein(s) with sufficient residence time for selective SuFEx reaction⁷⁷. Covalent reaction with sulfonyl fluorides (−SO2F) and fluorosulfates (−OSO₂F) likely occur through nucleophilic substitution reactions with addition–elimination and direct substitution pathways as potential mechanisms⁷⁷.

Figure 5. Representative examples of pharmacologically relevant sulfur(VI)-electrophiles.

(A) Sulfur-fluoride exchange (SuFEx) chemistry including sulfonyl fluorides, aryl fluorosulfates, and sulfonimidoyl fluorides for synthetic and medicinal chemistry. (B) Sulfonyl fluoride XO44 as a broadly reactive probe for kinome ABPP studies. (C) Scheme showing selective reactivity against orthogonal residues on DcpS by deploying a fluorosulfate (FS-p1) or sulfonyl fluoride (SF-p1) electrophile. (D) General reaction scheme between a sulfonyl-triazole and a tyrosine, i.e., sulfur(VI)-triazole exchange (SuTEx), the reactivity of which can be tuned by adjusting the electronics on the adduct group and/or the leaving group (LG). (E) Structure of TH211 and KY-26, both of which are general SuTEx probes of the kinome, as well as HHS-0701, an inhibitor of prostaglandin reductase 2 (PTGR2).

Since the introduction of SuFEx in 2014, various applications have been reported using this click chemistry^78–90. Taunton et al. performed a SAR study on a reported pyrimidine 3-aminopyrazole scaffold and identified XO44 as a covalent probe suitable for kinome ABPP studies in live cells through covalent modification of lysine residues in ATP binding pockets (Figure 5B)⁹¹. XO44 was applied in a competitive ABPP analysis of dasatinib to identify several unknown and previously disclosed kinase targets of this tyrosine kinase drug used for treating chronic myeloid leukemia⁹¹. Sharpless et al. screened SuFEx fragment compounds to identify a lead fluorosulfate compound that inhibits human neutrophil elastase with sub-micromolar potency through covalent modification of the catalytic serine (Ser195)⁹².

Arylfluorosulfates were shown to exhibit less reactivity compared to aryl sulfonyl fluorides when administered to live cells. Simple arylfluorosulfates selectively reacted with tyrosine residues, specifically in the fatty acid binding site of intracellular lipid binding proteins (iLBPs)⁹³. Mutagenesis experiments revealed the removal of either arginine in the Arg111-Arg132-Tyr134 carboxylic acid binding module that is conserved amongst the iLBP family reduces the reactivity of Tyr134. The cationic charges of either arginines next to Tyr134 are speculated to perturb the pK_a of the phenol and enhance nucleophilicity⁹³. Similarly, Jones et al. designed arylfluorosulfates that modified a non-catalytic serine in the binding site of the mRNA decapping scavenger enzyme DcpS that was identified by LC-MS/MS as the dehydroalanine species from β-elimination⁹⁴. Interestingly, the aryl sulfonyl fluoride congener modified Tyr113 or Tyr143 in the active site of DcpS⁹⁵ (Figure 5C).

Tyrosine-reactive groups

The phenol side chain group of tyrosine residues on proteins can serve as a nucleophilic site for covalent modification^96,97. The abundance of tyrosine is comparable with cysteine (~2–3%⁹⁸), which makes this residue attractive for developing covalent ligands with opportunities for protein- and site-selectivity. The phenol of tyrosine is also distinct in its amphiphilic nature that facilitates its location in hydrophilic and hydrophobic protein pockets⁹⁹. In addition to its catalytic function¹⁰⁰, tyrosines are prominent sites for phosphorylation¹⁰¹ and other post-translational modifications (PTMs)¹⁰², which offers additional opportunities for perturbing non-catalytic functions using tyrosine-reactive ligands.

As described in the previous section, SuFEx molecules have been shown to react with tyrosine sites on proteins^93,95. Our group recently introduced sulfonyl-triazoles as a new reactive group for preferential modification of tyrosines in catalytic and non-catalytic sites of proteins through sulfur-triazole exchange (SuTEx) chemistry (Figure 5D)¹⁰³. SuTEx shares common features with SuFEx. The deployment of a heterocyclic LG, however, offers an additional site for functional group modifications that enables covalent probe development. A dramatic increase in reactivity was observed for sulfonyl probes containing a triazolide compared with a fluoride LG, which could be due to additional stabilization from resonance in the former LG¹⁰³. The enhanced reactivity facilitated development of first generation SuTEx probes that were stable in aqueous solvents and could broadly modify thousands of tyrosine sites in lysates and live cells. The ability to globally profile tyrosine sites across the human proteome facilitated identification of a subset of tyrosines that displayed enhanced nucleophilicity (deemed “hyper-reactive” tyrosines) in functional domains of diverse proteins.

The reactivity of SuTEx compounds is tunable. The electrophilicity can be tempered in a graded fashion by selection of functional group modifications with differential electronic (and likely steric) impact. The location of functional group modification (adduct- vs leaving-group) can also alter reactivity of resulting sulfonyl-triazoles (Figure 5D). The “tunability” feature of SuTEx has enabled fragment-based ligand discovery in proteomes¹⁰⁴ and live cells¹⁰⁵. More recent studies have tested the inclusion of elaborated binding elements on the adduct- (KY-26¹⁰⁶, HHS-0701¹⁰⁵) and leaving-group (TH211¹⁰⁷) to guide SuTEx TCI development (Figure 5E). Additional methods for tyrosine bioconjugation have been reported⁹⁷ including a recently described bioorthogonal chemistry strategy using photoredox catalysis¹⁰⁸.

Lysine-reactive groups

Selective covalent binding to lysine is more difficult than cysteine due to the sheer abundance of lysine residues (~650,000) in the human proteome¹⁰⁹. The reduced nucleophilicity of the lysine amino side chain requires a suitable microenvironment (nearby residues for perturbation of pKa or high effective molarity from reversible binding affinity) for effective protein modification. Various amine-reactive probes have been explored and discussed in a comprehensive review¹⁰⁹. In a recent chemoproteomic evaluation by the Cravatt group, ~180 aminophilic probes were screened against >14,000 lysine sites identified in the proteome to establish an important resource of lysine-reactive groups to facilitate chemical probe development¹³.

These studies expanded on previous ABPP investigations of covalent compounds with activated esters^110,111, N,N′-diacylpyrazolecarboxamidines¹¹⁰, SuFEx^83,91, iminoboronates¹¹², and other reactive groups¹³ to assess greater than 30 distinct aminophilic chemotypes in the human proteome. The authors identified many lysines in structurally and functionally diverse proteins that are preferentially modified (i.e. liganded) by the aminophilic reactive groups tested. The authors deployed broadly-reactive dicarboxaldehyde fragments to assess ligandability of lysines across the proteomes. Among the protein targets identified, these ‘scout’ fragments engaged conserved lysines in the interferon-induced RNA-binding proteins such as IFIT1 (K151), IFIT3 (K148) and IFIT5 (K150) involved in suppressing viral replication. These ‘scout’ fragments provide a scaffold for probe development by facilitating rapid evaluation of protein-ligand interactions. While advantageous, electrophilic fragments such as dicarboxaldehydes are not selective for a particular protein and thus require additional medicinal chemistry optimization for TCI development. Cyanomethyl acyl sulfonamides were identified as lead compounds that showed clear SAR for blockade of IFIT protein members, with lysine site specificity in some examples, and their respective RNA–protein interaction functions¹³.

Concluding remarks and future perspectives

In this review, we describe reactive groups that are found in a collection of thiophilic and oxophilic covalent drugs. Many electrophiles found in drugs are directed towards catalytic serine, threonine, and cysteine sites because of the importance of these residues in protein function and their accessibility in structurally defined pockets. The discovery of a transient allosteric pocket in the S-IIP region of KRAS^G12C provided a key site for covalent binding to the mutant but not wild-type protein¹⁵. An interesting avenue for future exploration is leveraging existing proteomic methodologies to systematically identify cryptic functional pockets on proteins and proteoforms (including mutant proteins). The intrinsic nucleophilicity of cysteines on proteins necessitate exploration of orthogonal reactive chemistry for covalent binding to underexplored residues such as lysine and tyrosine. An emerging trend in the field that will help address this question is a shift towards screening of covalent compound libraries using multiplexed LC-MS methods for broad identification of electrophiles and their respective binding (chemo)selectivity^12,13,57.

In summary, a covalent MOA offers exciting opportunities for developing inhibitors or antagonists of challenging protein targets. To realize the benefits of this compound class, we must complement ongoing and future efforts to develop reactive chemistry for all potential nucleophilic sites on proteins with careful understanding of selectivity and MOA (see Outstanding Questions).

Outstanding questions.

• How can we leverage existing proteomic technologies to systematically identify unknown or cryptic functional pockets on proteins (e.g. the S-IIP pocket of KRAS^G12C)?

• What is the most effective strategy for identifying electrophiles with minimal cysteine binding and enriched activity against underexplored and less nucleophilic residues?

• How can we diversify existing covalent compound libraries to more effectively survey chemical and biochemical space?

• What is the best approach for addressing off-target binding activity of covalent compounds that are not readily detected (e.g. reversible interactions) by current proteomic screening methods?

Highlights.

• Covalent drugs offer distinct advantages including high biochemical efficiency via non-equilibrium blockade, pharmacological activity that can outlast drug pharmacokinetics, and pharmacological access to ‘undruggable’ protein targets.

• While many covalent drugs were discovered serendipitously, recent examples of targeted covalent inhibitors (TCIs) in clinical use or testing support a path towards rational development.

• Advances in synthetic chemistry, proteomic screening, and bioanalytical instrumentation has enabled evaluation and medicinal chemistry optimization of covalent small molecule selectivity at an unprecedented scale.

• The development of reactive groups for modifying protein sites beyond catalytic serine, threonine, or cysteine residues offers great potential for expanding the scope of covalent probe and therapeutic discovery.

Glossary

Allosteric pocket

A shallow pocket on a protein that is not in the active site and is not directly associated with enzymatic activity but can regulate metabolic and signaling function after a small molecule or protein binds.

Bioorthogonal chemistry

Chemical reactions designed to study biomolecules by permitting selective modifications without interfering with complex biological processes in living systems.

Chemical proteomics

The design and implementation of small molecule probes for functional investigations of the biochemistry and biology of reactive amino acids in proteins using gel- or mass spectrometry-based techniques.

Click chemistry

Robust, high-yielding, and operationally simple chemistry amenable to various substrates that is often utilized in bioorthogonal chemistry to study the function of biomolecules.

Idiosyncratic adverse drug reactions

Unpredictable side-effects observed by a small subset of patients that are inexplicable with regards to the known mechanism(s) of action of the consumed drug.