Covalent allosteric modulation: an emerging strategy for GPCRs drug discovery

Abstract

Designing covalent allosteric modulators brings new opportunities to the field of drug discovery towards G-protein-coupled receptors (GPCRs). Targeting an allosteric binding pocket can allow a modulator to have protein subtype selectivity and low drug resistance. Utilizing covalent warheads further enables the modulator to increase the binding potency and extend the duration of action. This review starts with GPCR allosteric modulation to discuss the structural biology of allosteric binding pockets, the different types of allosteric modulators, as well as the advantages of employing allosteric modulation. This is followed by a discussion on covalent modulators to clarify how covalent ligands can benefit the receptor modulation and to illustrate moieties that can commonly be used as covalent warheads. Finally, case studies are presented on designing class A, B, and C GPCR covalent allosteric modulators to demonstrate successful stories on combining allosteric modulation and covalent binding. Limitations and future perspectives are also covered.

Graphical Abstract

1. INTRODUCTION

Drug discovery and development has been enhanced by recent scientific advances in computation, biophysics, and high-throughput screening (HTS) technology. This allows for the development of small molecule drugs with exceptional selectivity towards target proteins which were previously thought to be “undruggable”. Small molecules acting as allosteric modulators (AM) bound to the allosteric site display better pharmacological advantages over orthosteric ligands. A successful allosteric modulator will have capability of binding to an allosteric site and amending the conformation of the orthosteric binding site of targeting biomolecule.¹ However, it is important to state that many small molecules bind to the allosteric sites without disturbing the binding properties of the orthosteric sites. Thus, it is a key aspect of designing effective allosteric modulators which bind to allosteric sites as well as inducing a conformational change in the biomolecules.

With the great contribution of the development of various approaches to design allosteric modulators, we could advance in the understanding of allosteric regulatory mechanism and individual protein functions.^2–4 The development of covalent drugs and allosteric modulators has been increased incredibly and exponentially in recent years (Fig. 1).^{5, 6} Despite the potential rewards of allosteric therapeutics, it is still not easy to target to an allosteric binding site with small molecules and finding allosteric modulators has proved to be a major challenge.⁷ Unfortunately, it is not a simple matter to locate and illustrate new allosteric sites. Most allosteric modulators have been discovered serendipitously by High Throughput Screening (HTS) experiments.⁸ Even with the advance of computational approaches to predict allosteric interactions and monitor potential allosteric modulators in recent years, there are still limitations associated with the design and discovery of allosteric modulators. Figure 2 shows a brief timeline of covalent drug discovery since 1890. The concept of the covalent drugs/inhibitors that can contribute to improving human health is becoming more broadly appreciated in recent years.^9–13 It has been actively discussed that several properties of covalent drugs could be advantageous under certain events compared to their non-covalent equivalents.¹⁰ In this perspective, we herein covered the latest advances on the covalent allosteric discovery with the focus on G-protein-coupled receptors (GPCRs).

Figure 1.

Number of publications obtained from the search of the term “Covalent drug” and “Allosteric modulator” in SciFinder®.

Figure 2.

Timeline of covalent drug discovery.

2. GPCRs ALLOSTERIC MODULATION

GPCRs are membrane receptors constituted with seven transmembrane (TM) domains. Over 40% of available drugs on the market in this day and age are targeting GPCRs as GPCRs are widely involved in diversified physiological systems.¹⁴ To date, there are more than 800 receptors in the GPCR superfamily.¹⁵ On the basis of sequence homology, these receptors can be grouped into four classes, A, B, C, and F (Frizzled). The traditional methods for GPCRs drug discovery have the focus on targeting the orthosteric binding pocket, which results in corresponding agonists, antagonists, inverse agonist, etc. for specific proteins. Over the past years, distinct binding sites (allosteric binding pockets) have been recognized and reported with the advances in GPCRs conformational studies.¹⁶ Developing ligands targeting allosteric sites to modulate pharmacological functions brings new opportunities to the field of drug discovery. In this section, the GPCRs allosteric modulation will be specified in the following three aspects: GPCRs structural biology with the focus on locations of allosteric sites; the different types of modulators and their distinctive functions; and the benefits of GPCRs allosteric modulators.

2.1. GPCRs allosteric binding sites

Allosteric modulators interact with GPCRs at binding pockets that are topographically distinctive from the orthosteric binding pockets. Even with a universal seven transmembrane domain, receptor-ligand binding mode can vary between GPCR classes. Under the classic pattern of structural architecture for a class A GPCR, the orthosteric binding pocket is formed by the transmembrane helixes while the extracellular loops and the N-terminus of the peptide chain define the allosteric binding domain. The pattern for a class B GPCR is opposite that usually the N-terminus and the extracellular loops hold the orthosteric ligand while the cavity among the seven helixes carry the allosteric modulator. Meanwhile, the class C GPCRs like metabotropic glutamate receptors are often dimers containing a cysteine-rich domain and venus flytrap domain at the extracellular N-terminus for the orthosteric interactions while the allosteric modulator may occupy the TM domain for the modulation.¹⁷ The vast development of X-ray Crystallography and the Cryo-Electron Microscopy in the past decades dramatically advanced the studies of GPCR structural biology. With the increasing number of GPCR-allosteric modulator complexes been resolved, the understanding towards the position of possible allosteric pockets has been furthered.

For class A GPCRs, the reported potential allosteric binding pockets are throughout the seven TM domain from the extracellular loops to the intracellular end. As shown in Figure 3a, five allosteric binding pockets (ABP) are summarized based on the published crystal or Cryo-EM receptor-ligand complexes. ABP1 and ABP2 are cavities located inside the helixes and flexible loops, while ABP3, ABP4, and ABP5 are pockets attached to the outer surface of the helixes. Andrew C. Kruse et al. published their crystallization result of both the orthosteric agonist Iperoxo and the allosteric modulator LY2119620 with the human M2 muscarinic acetylcholine receptor (Fig. 3b).¹⁸ The position of the orthosteric binding pocket is conserved and mainly defined by D103 and Y104 on TM3, W400, Y403, and N404 on TM6, and Y426 on TM7. The hydrogen bond interaction can be observed between the Iperoxo and the asparagine N404. The allosteric modulator LY2119620 binds to the cavity formed by extracellular loops and the N-terminus directly above the orthosteric ligand. The modulator interacts with the extracellular vestibule extensively and causes a slight additional contraction of the protein.¹⁸ Critical residues that involved in pocket formation are Y80 on TM2, E172 and Y177 on ECL2, N410 on TM6, N419 on ECL3, and W422 and Y426 on TM7. Hydrogen bonds are expected between the LY2119620 and tyrosine Y80, asparagine N410, and asparagine N419.

Figure 3.

Illustrated allosteric binding pockets for class A GPCRs. a. The positions of five reported allosteric binding pockets throughout the seven transmembrane helixes. The helixes are shown in gray and the allosteric binding pockets are shown in salmon (ABP1), pale green (ABP2), light blue (ABP3), lemon (ABP4), and light pink (ABP5). b. The receptor-ligands complex for ACM2, Iperoxo, and LY2119620. c. The receptor-ligands complex for CCR2, BMS681, and CCR2-RA-[R]. d. The receptor-ligands complex for CB1, CP55940, and ORG27569. e. The receptor-ligands complex for GPR40, MK8666, and AP8. f. The receptor-ligand complex for PAR2 and AZ3451. Ligands are shown in magenta and key residues are shown in green.

Yi Zheng et al. reported the co-crystal structure of both the orthosteric inhibitor BMS-681 and the allosteric antagonist CCR2-RA-[R] in human CC chemokine receptor 2 (Fig. 3c).¹⁹ The CCR2-RA-[R] blocks the conformational change of the receptor and further prevents the formation of the G-protein-binding interface. CCR2-RA-[R] sterically blocks Y305 on TM7 and restricts R138 on TM3 to orientate the G-protein interface, which prevents the target to form the active conformation.¹⁹ Again, the position of the orthosteric binding pocket is located among the transmembrane helixes facing the extracellular space. The residues, Y49 on TM1, W98 on TM2, Y120 on TM3, Q288, and T292 on TM7, are pivotal for the BMS-681 binding. Among these residues, tyrosine Y49, glutamine Q288, and threonine T292 can form hydrogen bonds with the inhibitor. The allosteric binding pocket that CCR2-RA-[R] occupies is settled at the intracellular ends of TM1–3 and TM5–7. Through the mutagenesis study, the critical residues of V244 on TM6, Y305 on TM7, and K311 and F312 on TM8 are established. It is a highly-enclosed pocket with balanced hydrophobic and hydrophilic characteristics. Lysine K311 and phenylalanine F312 are observed to form hydrogen bonds with the modulator.

Zhenhua Shao et al. demonstrated another mode of allosteric modulation with the solved crystal structure of cannabinoid receptor 1 with the orthosteric agonist CP55940 and the negative allosteric modulator ORG27569 (Fig. 3d).²⁰ In this structure, the conserved orthosteric binding pocket holds the agonist while the modulator binds to an extrahelical site within the inner leaflet of the membrane. This extrahelical site is usually known for the GPCR cholesterol interactions. Serine S173 on TM2, lysine K192 on TM3, isoleucine I267 on ECL2, and serine S383 on TM7 are vital for agonist binding and formation of hydrogen bonds with CP55940 to hold the ligand. The allosteric site is mainly defined by residues on TM2 and TM4, which created a largely hydrophobic environment for the modulator. Key residues involved are H154, G157, S158, and V161 on TM2, and F237, C238, W241, T242, and I245 on TM4. ORG27569 prevents the conformational change of side chains that are critical in receptor activation through the contact with F237 on TM4 in its inactive conformation. Since F237 helps to position F155 which further interacts TM6, ORG27569 may contribute to the maintenance of the off dual toggle switch (F200 on TM3 and W356 on TM6) in an indirect manner.²⁰

Jun Lu et al. published the crystal structure of human GPR40 in complex with both the partial agonist MK-8666 and an Ago Positive Allosteric Modulator (AgoPAM) AP8 (Fig. 3e).²¹ A novel lipid-facing AgoPAM-binding pocket outside the transmembrane helixes is revealed. The conserved orthosteric site holds the partial agonist and faces the extracellular space. Tyrosine Y91 on TM3, arginine R183 on TM5, and tyrosine Y240 on TM6 can have the formation of hydrogen bonds with the MK compound. The allosteric site locates at the extrahelical region formed by TM2, TM3, TM4, and TM5. Hydrogen bonds between AP8 and Tyrosine Y44 on TM2 and Serine S123 on TM4 are observed. Through the comparison with the binary complex of GPR40-MK8666, the allosteric modulator induced conformational change which involves rearrangements of transmembrane helices 4 and 5 (TM4 and TM5) and transition of the intracellular loop 2 (ICL2) into a short helix.^{21, 22}

Robert K. Y. Cheng et al. reported the crystal structure of protease-activated receptor 2 (PAR2) in complex with the AZ3451 (Fig. 3f).²³ The AZ3451 was reported both as antagonist and allosteric modulator^{23, 24} that binds to an allosteric site outside the helical bundle, which is remote to the conserved orthosteric pocket. Authors also proposed that the required structural rearrangements can be prevented by blocking this site. Their study exposed another possible allosteric pocket for class A GPCRs. This pocket is in general defined by L123 on TM2, F154, A157, and C161 on TM3, and W199, L203, and Y210 on TM4. Hydrogen bond interaction is established between the modulator and the tyrosine Y210 on TM4. Robert K. Y. Cheng et al. discussed the potential mechanism of action for this modulator. Similar with other extrahelical modulators, AZ3451 may restrict the inter-helical conformational rearrangement which is essential for agonist induced receptor activation.²³

For class B GPCRs, two reported allosteric binding pockets are summarized here for potential receptor-modulator interactions (Fig. 4a). Kaspar Hollenstein et al. reported the crystal structure of the transmembrane domain of the human corticotropin-releasing factor receptor type 1 (CRFR1) in complex with the antagonist CP-376395 (Fig. 4b).²⁵ The pocket is circled by TM3, TM4, and TM5. F203 and M206 on TM3, V279, L280, N283, F284, L287, and I290 on TM5, and T316, L319, L320, L323, G324, and Y327 on TM6 create the cavity to hold the small molecule. Specifically, asparagine N283 formed the hydrogen bond interaction with GP-376395. The cytoplasmic half of TM3, TM5, and TM6 are tethered after the binding of GP-376395, which possibly prevents the receptor to transfer into an active conformation.²⁵ Ali Jazayeri et al. published their work on the crystallization of human GCGR in complex with the antagonist MK-0893 (Fig. 4c).²⁶ In their work, authors concluded that this is a novel allosteric binding pocket for class B GPCRs. The binding pocket is found to be outside the helix bundles in a position between TM6 and TM7. The co-crystalized compound expands further into the lipid bilayer. L329 on TM5, F345, R346, A348, K349, L352, and T353 on TM6, and L399, L403, N404, and K405 on TM7 are involved in forming the binding site. Strong hydrophilic interactions can be noticed. The amide group of the MK-0893 forms hydrogen bonds with lysine K349 and Serine S350 on TM6. The carboxyl group stretches further to create polar contacts with arginine R346 on TM6, asparagine N404 and lysine K405 on TM7. The TM6 is held in the inactive state as the MK-0893 functions as a clamp in the pocket, which further restrains the essential conformational changes of the target for G-protein coupling.²⁶

Figure 4.

Illustrated allosteric binding pockets for class B and class C GPCRs. a. The positions of two reported allosteric binding pockets for class B GPCRs. The helixes are shown in gray and the allosteric binding pockets are shown in salmon (ABP1) and palegreen (ABP2). b. The receptor-ligands complex for CRFR1 and CP376395. c. The receptor-ligands complex for GCGR and MK0893. d. The positions of the reported allosteric binding pocket for class C GPCRs. e. The receptor-ligands complex for MGLUR5 and Mavoglurant. Ligands are shown in magenta and key residues are shown in green.

For Class C GPCRs, the allosteric modulators may occupy the TM domain for the modulation (Fig. 4d). Andrew S. Dore et al. illustrated how the negative allosteric modulator Mavoglurant interacts with the metabotropic glutamate receptor 5 using the crystal structure of the receptor-ligand complex (Fig. 4e).²⁷ The allosteric binding pocket is defined by residues from TM2, TM3, TM5, TM6 and TM7. Mavoglurant fits inside the pocket with the alkyne linker traversing a narrow channel between Y659, S809, Val806, and P655. Three hydrogen bonds could be formed between mavoglurant and N747 on TM5, and S805 and S809 on TM7. The saturated bicyclic ring system sits within a mainly hydrophobic pocket created by residues in TM5, TM6, and TM7.

2.2. Types of GPCRs allosteric modulators

Orthosteric binding pockets are usually binding sites for endogenous ligands and have historically been the major focus for conventional drug discovery. According to the down-stream physiological responses triggered through binding, the orthosteric ligands can be categorized into agonists, partial agonists, antagonists, inverse agonists, etc (Fig. 5a). A full agonist can dramatically increase the G protein signaling responses, while a partial agonist may only increase those responses to a certain degree. An antagonist often blocks the receptor and prevents other molecules from activating the target. On the other hand, an inverse agonist can bind the receptor and further decease the down-stream signaling.²⁸

Figure 5.

(a) The categorization of orthosteric ligands, (b) The categorization of allosteric modulators.

Allosteric modulation is a different story. Generally speaking, the allosteric modulators usually will not directly trigger the G protein signaling responses but instead influence the orthosteric regulations (Fig. 5b).^{29, 30} If a modulator can increase the affinity or efficacy of an orthosteric agonist, it is termed as a positive allosteric modulator (PAM). In contrast, if a modulator can decrease the affinity or efficacy of an orthosteric ligand, it is termed as a negative allosteric modulator (NAM). Also, if one modulator can only occupy the allosteric binding pocket but has no influence on orthosteric binding/response, it is termed as a silent allosteric modulator (SAM). In recent years, the concept of the Ago Allosteric Modulator was proposed to describe the modulator that has the ability to show both allosteric and orthosteric activities.^{14, 31} For example, Chiori Yabuki et al. reported that the fasiglifam (Fig. 6) can function as an ago-allosteric modulator on free fatty acid receptor 1 (FFAR1)/GPR40 with a partial agonistic activity.³² The Ca²⁺ influx and insulin secretion assays demonstrated that the fasiglifam can positively cooperate with the FFAR1 ligand c-linolenic acid (c-LA). Virginie Binet et al. demonstrated that the reported γ-aminobutyric acid, type B (GABA_B) receptor positive allosteric regulator, CGP7930 (Fig. 6), can also activate GABA_B2 with agonistic activity.^{33, 34} Both [γ−³⁵S]GTP binding and IP production assays confirmed the allosteric activity of CGP7930, and the more sensitive IP assay revealed that CGP7930 can partially and directly activate the GABA_B receptor. Birgitte Holst et al. reported that the L-692429 (Fig. 6) can act as both an agonist and a positive allosteric modulator to the ghrelin receptor.³⁵ L-692429 displayed potency (25–60 nM) in signaling assays as an agonist and increased the potency of ghrelin for 4–10 folds as a positive allosteric modulator.

Figure 6.

Illustrated ago allosteric modulators.

2.3. Features and advantages of GPCRs allosteric modulation

There are successful stories on GPCR drug discovery campaigns, which are remarkable and exciting. Even though the traditional drug discovery efforts mainly focused on GPCR orthosteric binding pockets, developing allosteric modulators can possess advantages. Since the allosteric binding pockets are under less evolutionary pressure for conservation across a receptor family, the modulators can have a preferable subtype selectivity.³⁶ Besides, the allosteric modulators can have a better safety profile as their effects can only be exerted in cooperation with the corresponding orthosteric ligands.³⁷ Also, Yiran Wu et al. added that the distinct pathways and allosteric pockets may enable allosteric modulators cooperativities among different protein subtypes.³⁸

3. COVALENT BONDS AND COVALENT MODULATORS

A covalent bond is formed between two atoms which share an electron pair between molecular orbitals. The formation of a covalent bond results in a molecule with increased stability and therefore decreased energy when compared to the individual atoms.³⁹ Covalent modulation plays a major role in the regulation of physiological processes through biochemical reactions. Biochemical reactions such as phosphorylation, acetylation, glucuronidation, ubiquitination and their counter reactions all involve the formation or breaking of a covalent bond.⁴⁰ These reactions are known to regulate processes such as enzymatic activity, gene expression, drug metabolism and cellular waste recognition/removal. Covalent modification is also present throughout forms of pharmacological interventions. It has been actively discussed that this kind of drug has enhanced potency, selectivity, long-lasting inhibition, and enables both target and off-target identification (Table 1).⁴¹ During the last decade the FDA approved several drugs such as Osimertinib (2015),^{42, 43} Ibrutinib (2013),^{44, 45} and Afatanib (2013)⁴⁶ that were designed to behave as a hetero-Michael acceptor to react with a unique cysteine residue of specific protein (Fig. 7).

Table 1.

Strength and Weakness of covalent inhibitors

Strength	Reference
• Increase biochemical efficiency	47–51
• Less sensitive to pharmacokinetic parameters
• Longer duration of action (long residence time, less frequent dosing)
• Potential to avoid some resistance mechanisms
Weakness
• Higher risk if covalent inhibitor deficiencies specificity	52–57
• Not optimal for targets which need short residence time, transient or partial inhibition
• Protein adduct may cause an allergic response or hypersensitivity

Figure 7.

Illustrated FDA approved drugs as covalent inhibitors.

3.1. Designing and Optimizing a Covalent Modulator

Historically drugs have not been designed to be covalent modulators, instead after use it was discovered their mechanism of action involved a covalent bond formation. Covalent modulators can be highly selective molecules, but the difficulty in their design is that they can be very difficult to evaluate through a high-throughput screen or traditional assay format. When designing a covalent modulator, the compound must be potent enough to produce the desired response in the target protein and have a high enough specificity to prevent off target binding which leads to toxicities.

Although the IC₅₀ value is a hallmark of preclinical drug development, it may not be the most appropriate parameter to characterize covalent modulators. Due to the mechanism of covalent bond formation, it’s essential to consider the kinetics being the formation of covalent bonds.⁵⁸ A covalent bond is formed between a protein and drug molecule through two distinct steps. The first step is the formation of an initial binding complex through non-covalent bonding (electrostatics, hydrophobic interactions, etc.).⁵⁸ This step is described by K_I or the concentration of drug necessary for 50% potential for formation of a covalent bond. ⁵⁸ The second step is the interaction between electrophile and nucleophile to produce the covalent bond, a rate dependent step based on the concentration of drug.⁵⁸ For reversible inhibitors, once the protein is saturated with drug, the rate of inactivation is measured as k_inact. Overall, this process can be characterized by the ratio of K_I to K_inact. ⁵⁸ Including these parameters into bioassays during preclinical drug development can allow researchers to better characterize and select covalent inhibitors for further development. In addition to cellular and biochemical assays measuring the rate of covalent bond formation, utilizing mass spectrometry can confirm covalent binding by measuring a shift in mass based on the mass of protein + drug molecule^{58, 59}. Similar to metabolic studies later in a drug’s development, the adduct formed via covalent bond formation can be detected based on the mass shift. Zvonok et al utilized this methodology to characterize the binding site for human monoacylglycerol lipase (hMGL).⁵⁹ The binding site of hMGL contains three serine resides and the mechanism of covalent bond formation is a carbamylation between their ligand and one of the serine residues.⁵⁹ By performing tandem MS/MS on modified and unmodified hMGL, the researchers were able to identify Ser¹²⁹ as the site of covalent interaction.⁵⁹ Use of mass spectrometry was also used to confirm the kinetic results by monitoring mass shifts based on the mass of the inhibitor.⁵⁹ This study demonstrates the power that MS has to enhance the understanding and validity of screening results for a covalent modulator. A covalent interaction could be assessed by changes on IC₅₀ value when compound was pre-incubated for a definite time with the protein and then activity was tested, while the IC₅₀ will be higher when the assay was performed immediately after the contact between modulator and protein (time dependency). The washout experiment and the residence time experiment are typical assessments for a covalent interaction. It is critical to have the evaluation on the time-dependent effects of drug treatment. The washout experiment evaluates the phenotypic result of target engagement while the drug is removing from the system, which exhibits target vulnerability and the time-dependent drug activity.⁶⁰ The drug-target residence time reflects an experimental measure of the lifetime of a drug-target complex. The strategies of testing the residence time can range from direct / indirect kinetic radioligand binding assays to competition association assays and dual-point competition association assays.⁶¹ A long residence time is expected for a covalent interaction.

An alternative method for determining the kinetics/reactivity of a covalent modulator through the use of a Nuclear Magnetic Resonance (NMR) based screening has been developed.⁶² This assay is completed in solution with the selected nucleophilic amino acid from the protein target and the covalent modulator/warhead. The carboxy and amino groups of the amino acid are modified with protecting groups and in solution at a concentration 10 times the concentration of the covalent modulator to ensure the amino acid is present in excess and the reaction is only occurring between the R-group of the residue rather than the terminal groups. The assay begins with a set concentration of the modulator which gradually decreases with time. The researchers monitored the progress of the reaction between the two molecules by focusing on the NMR peaks which had the greatest change in intensity when the amino acid was introduced to the warhead. As the warhead concentration decreases the integral of the peaks was taken; this value was then plotted as a natural log to produce a linear representation of the change in intensity over time. It utilized for multiple types of covalent warheads including those which target cysteine, serine, threonine, lysine and arginine. Although this screening method does not necessarily demonstrate where/how binding is occurring at the active site of the protein, it does provide another avenue to compare the reactivities of selected ligands/compounds across a large diversity of potential covalent targets.⁶²

An additional approach for identifying and optimizing a covalent modulator is through computational screening. A new Docking-Based Virtual Screening (DBVS) approach for identifying novel covalent inhibitors was developed. It evaluates the reactivity of covalent inhibitors for Cathepsin K; a member of the lysosomal protease family. Even though covalent inhibitors based on low-throughput DBVS screening was successfully developed, Schröder and his colleagues developed a new method with high throughput capabilities to determine novel scaffolds/structures for their target of interest.

3.2. Electrophilic Warheads

Electrophilic warheads are molecules which have an electrophilic moiety which can be attacked by a nucleophilic residue within the active site of a protein. The residues commonly targeted by electrophilic moieties include cysteine and serine.⁶² Residues with lower activities; histidine, threonine, lysine and tyrosine residues are also targeted by electrophilic warheads. Figure 8 shows examples of interactions between covalent inhibitors and their target residues. Disulfide moiety, the electrophilic building block, designed to connect the recognition element of neurotransmitters via flexible linker, based on the compound FAUC50.⁶⁹ Noradrenaline, dopamine, serotonin, and histamine are constructed as nucleophilic neurotransmitters. Compounds containing sulfonyl halides are excellent electrophilic warheads which can be used in a broad range of applications.⁶³ In terms of stability, sulfonyl fluorides are less susceptible to reduction and may have better applicability in drug design. Active sites which contain cysteine are commonly targeted with molecules containing acrylamide moieties. Even though histidine and cysteine react with sulfonyl fluoride, these adducts are unstable relative to the adducts of tyrosine, lysine, serine, and threonine. Therefore, utilizing a sulfonyl fluoride allows for targeting of active sites which do not contain cysteine.^{70, 71} Unfortunately a major drawback of sulfonyl fluorides is a decreased solubility in aqueous solution when compared to acrylamide which may limit application to drug design. Nitrogen mustard compounds have a valuable class of cross-linking functionality.⁷² In aqueous solution, the mustard group (or 2-haloalkylamines) cyclizes to the corresponding aziridinium ion, which explains the nucleophilic alkylation of amino acid side chain. κ-Opioid receptor (KOR) agonist, Salvinorin A, has been reported.⁷³ Nucleophilic substitution addition of cysteine residue of protein to the 22-thiocyanato-derivative of salvinorin A results a formation of disulfide bond with C315^7.38. Clopidogrel is activated in the liver by oxidative ring opening and the nucleophilic free thiol of the active metabolite reacts with cysteine C97^3.25 in the P2Y₁₂ receptor resulting disulfide bond.^74–76 β-Funaltrexamine (β-FNA) is an opioid receptor antagonist used experimentally and it has α,β-unsaturated carbonyl derived function as a Michael acceptor.⁷⁷ This electrophilic moiety predominantly reacts with cysteines, but also react with nucleophilic lysin or histidine residues.⁷⁸ Table 2 contains representative examples of how nucleophilic residues can be targeted with electrophilic warheads.⁷⁹

Figure 8.

Covalent inhibitors/probes for GPCR. Electrophiles and nucleophiles are highlighted.

Table 2.

Representative examples of electrophilic protein modification

Type of nucleophile	Nucleophilic residue	Electrophile	Covalent bond product	Reference
Tyrosine				63
Cysteine and Lysine				64–66
Cysteine				67
Cysteine				68

Acrylamides or α,β-unsaturated amides have been used similarly to sulfonyl halides, but their electrophilic reactivity can be relatively low. Jackson et al. stressed the importance/reliance of structural substitutions surrounding the acrylamide moiety.⁸⁰ Depending on what moieties are substituted on opposite sides of the acrylamide, there can be a drastic difference in reactivity. This observation has major implications in tailoring a drug to have a specific action with limited kinome/proteome-wide effects.^{80, 81} To better avoid the off-target effects of covalent inhibitors, Michael acceptors equipped with α-cyanoacrylamides (or α-cyanoacrylates) have been shown to undergo rapid reversible thiol-carbon bond formation as reversible modifiers of cysteines. Taunton et al. also have demonstrated that the reversibility of the hetero-Michael addition of acrylonitriles can be modulated with heteroaromatic groups.⁸² The facile thiol or nucleophilic addition to the α-cyanoacrylamides (or α-cyanoacrylates) is on account of the two electron withdrawing groups at α-carbon. Contrarily, the increased α-proton acidity of the cyanoacrylate adduct crusades the reverse the reaction through a β-elimination. When targeting a cysteine residue, it is important to recognize whether the residue is catalytic or non-catalytic. For example, protein kinases are implicated in numerous forms of cancer and their function often relies on a catalytic cysteine.⁸¹ This residue can be irreversibly inhibited via a covalent bond with acrylamide to abrogate its function. In 2013 there were four covalent drugs in clinical trials which were protein kinase inhibitors, and each drug’s action relied on a covalent bond between acrylamide and a thiol of the kinase.⁸¹ There is a large diversity of chemical structures which can act as electrophilic warheads; many of these chemicals act as Michael Acceptors and undergo a Michael Addition. In order to keep this review brief, please refer to the review from Jackson et al. for a more in-depth overview of the diverse range of electrophilic warheads.⁸⁰

4. CASE STUDIES OF DEVELOPING GPCRs COVALENT ALLOSTERIC MODULATORS

Structural biology studies in recent years exhibit that covalent molecules are good probes to stabilize the GPCRs for X-ray crystallization^83–86 which significantly enhances drug discovery efforts in this area of research. Given the merits of having covalent allosteric modulators, there is an increasing number of stories for developing GPCRs allosteric modulators. In this section, typical examples will be discussed for class A, B, and C GPCRs.

4.1. Human cannabinoid 1 receptor

Research on design and synthesis of novel electrophilic and photoaffinity covalent allosteric modulator for the cannabinoid 1 receptor (CB1) has been published.⁸⁷ Cannabinoid receptors are class A rhodopsin-like GPCRs and are critical target for the endocannabinoid system.^{88, 89} CB1 and CB2 are two major subtypes of the CB family. CB1 is predominately expressed in the central nervous system (CNS).^{89, 90} Targeting CB1 selectively can show therapeutic potential for multiple diseases and disorders including drug addictions, motor controls, neurodegenerative diseases, and anxiety responses.⁹¹ As describe above, CB1 has a reported co-crystal structure with the orthosteric agonist CP55940 and the negative allosteric modulator ORG27569 (Fig. 3d). The modulator occupies the cholesterol binding site at the extracellular surface of TM2 and TM4. In their research, they added a chemical reactive electrophilic isothiocyanate to the classical CB1 allosteric modulators 1 (ORG27569) (Fig. 9).⁹² These modifications result in an irreversible covalent allosteric modulator 2 (GAT100). The GAT100 functions as a PAM that increases the binding of CB1 orthosteric agonist CP55940⁹³ by 2.25 folds. The isothiocyanate was incorporated to replace the chloride on 1 (ORG27569) to create the capability of covalently interacting with distinct amino acids. The molecular docking studies afterward indicated that the residue cysteine C382 is the most likely amino acid to interact with the isothiocyanate moiety.⁹⁴ Their findings also illustrate that the amine nucleophiles and sulfhydryl groups of cysteine can have better reactivity with isothiocyanate than other nucleophiles like alcohol or water. In the following cellular cAMP, G-protein-independent β-arrestin, and G protein-dependent GTPγS functional assays, 2 (GAT100) turned out to possess the higher potency than its parent compound ORG27569. 2 (GAT100) was an effective inhibitor of β-arrestin recruitment (EC50 = 2 nM) and illustrated good activity in the cAMP assay (EC50 = 174 nM).⁸⁷ In particular, without showing inverse agonistic activities, 2 (GAT100) did not affect the constitutive activity of the receptor in the GTPγS assay, indicating a lower risk of adverse events.⁹⁴ Mariano Stornaiuolo et al. published a paper in 2015 comparing both the endogenous and exogenous allosteric modulators for CB1.⁹⁵ Their group designed and synthesized the CB1 covalent ligand, 4 (ORG27569ALK3), by adding an alkyne moiety to the nitrogen atom of the indole ring. The sulfhydryl and hydroxyl group of an amino acid can act as a nucleophile and attack the alkyne to form the covalent bond between the small molecule and the amino acid residues. The LC/MS analysis revealed that the 4 (ORG27569ALK3) formed the covalent bonds between a serine residue on TM2 (S158) or TM3 (S206).

Figure 9.

Chemical structures of ORG27569, GAT100, and ORG27569ALK3. Key moieties involved in forming covalent bonds are marked in red.

4.2. Human Glucagon-like peptide-1 receptor

Small molecule 5 (BETP) that behaves as a PAM at the glucagon-like peptide-1 receptor (GLP-1R) has been reported (Fig. 10).⁹⁶ 5 (BETP) covalently modifies cysteines C347 at TM6 and C438 at C-terminal in GLP-1R. The MS-based proteomics and mutagenesis studies with two GLP-1R mutants (C347A and C438A) were used for the binding site identification. GLP-1R is a member of the class B GPCRs and it is known to be associated with type 2 diabetes.⁹⁷ GLP-1 is a peptide hormone which is secreted from enteroendocrine L cells. GLP-1 can bind to GLP-1R on pancreatic β-cells to stimulate the secretion of insulin causing a decrease in plasma glucose levels. Currently there are a limited number of discovered orthosteric ligands for GLP-1R, and none are capable of fully reproducing the pharmacological activities of GLP-1.^{96, 98} In their research, developing covalent allosteric modulators targeting GLP-1R was a proposed solution. 5 (BETP) is observed to have the potential for intracellular calcium mobilization in response to GLP-1(9–36)NH₂⁹⁹ in recombinant cell lines.⁹⁶ They conducted cAMP washout experiments to confirm the covalent modulations. The cAMP washout experiment is a typical way to determine the covalent activity. 5 (BETP) was used to treat both CHO-K1 and INS-1 cells first and then washed followed by GLP-1(9–36)NH₂. CHO-K1 cells stably overexpress recombinant human GLP-1R and in insulin-secreting INS-1 cells endogenously express rat GLP-1R. The mechanism of covalent modification of GLP-1R was supported as 5 (BETP) maintained the capability to enhance the cAMP activity of GLP-1(9–36)NH₂ in the two cell lines after washout. The 5 (BETP) analog 7 (PETP) was also investigated and observed to show positive allosteric activities towards GLP-1R in CHO-K1 cells. 7 (PETP) potentiated the GLP-1(9–36)NH₂-dependent cAMP production to the similar level of potency and efficacy as 5 (BETP). Compound 6 is another GLP-1R PAM with a distinct scaffold.⁹⁶ The covalent binding of 6 to the GLP-1R was suggested by the cAMP accumulation-washout experiments similar to 5 (BETP).

Figure 10.

Chemical structures of BETP, PETP, and compound 6.

4.3. Metabotropic glutamate receptors

Study on design, synthesis, and pharmacological characterization of the covalent allosteric probe for the metabotropic glutamate receptor 2 (mGlu₂) has been disclosed (Fig. 11).¹⁰⁰ The mGlu₂ is a class C GPCR which is activated by glutamate.¹⁰¹ The allosteric modulation of the mGlu₂ has been touted to have the potential for treatment of neurological disorders including schizophrenia and anxiety.¹⁰² To date, the structural information is available for the extracellular venus flytrap domain (endogenous glutamate binding domain), while the understanding towards the transmembrane domain (allosteric binding domain) is more based on the structural information of mGlu₁ and mGlu₅.^{27, 103, 104} To further the understanding of mGlu₂ allosteric modulation, compound 11 was prepared as a covalent binding PAM based on the known compound 10 (JNJ-46281222). The fluorosulfonylphenyl moiety was selected as the warhead for covalent binding as it possesses the capacity to interact with nucleophilic amino acids. The positive allosteric activity of compound 11 was confirmed via a [³⁵S]GTPγS binding assay. The washout assay was combined with the [³⁵S]GTPγS binding assay to find that compound 11 maintained the ability to induce [³⁵S]GTPγS binding after the washing cycles. The mutagenesis studies as well as in silico simulations further identified the threonine T791^7.29×30 as the amino acid for the covalent interaction.

Figure 11.

Modification of Chemical structures of JNJ-46281222 and covalent interaction to nucleophilic residue.

Study of the photoswitchable negative allosteric modulator, 13 (OptoGluNAM4.1) for the metabotropic glutamate receptor 4 (mGlu₄) was performed (Fig. 12).¹⁰⁵ The mGlu₄ is another member of class C GPCR. Targeting mGlu₄ is an attractive strategy for treating chronic pain, Parkinson’s disease, etc.^{106, 107} 13, OptoGluNAM4.1 possesses a reactive moiety for covalent interactions with the surrounding amino acid residues and a blue-light-activated azobenzene moiety that permits the reversible photocontrol both in vitro and in vivo. The illumination with blue light will cause the trans to cis configuration change of the OptoGluNAM4.1 photoisomerization. The cis configuration (14) can no longer possess the mGlu₄ activity that the trans configuration (13) exhibited. Mutational studies on leucine L756 on TM5 to serine and valine V826 on TM7 to methionine were performed to confirm the boundary of the allosteric binding pocket and supported binding location of the compound. The in vitro assay showed that 13 (OptoGluNAM4.1) decreased the orthosteric agonist induced activities in a dose-dependent manner. The inclusion of an azobenzene group provides a small molecule with photochromic activity. From the single-cell calcium imaging experiments, they found the 13 (OptoGluNAM4.1) prevented the receptor activation by the agonist L-AP4 while the L-AP4 induced intracellular calcium responses can be restored under 430 nm illumination. The in vivo assays were followed to investigate the compound in animal behaviors. The mouse model of inflammatory chronic pain was first applied. 13 (OptoGluNAM4.1) was capable of blocking the analgesic effect caused by the injection of LSP4–2022, an mGlu₄ orthosteric agonist. Transparent zebrafish larvae were used for behavioral assays. The zebrafish locomotion was monitored and showed that animals treated with 13 (OptoGluNAM4.1) increased the free-swimming distance.

Figure 12.

Photo active compound 13, OptoGluNAM4.1 and its photo isomer 14.

The discovery of clickable photoaffinity NAM for the metabotropic glutamate receptor 5 (mGlu₅) has been reported (Fig. 13).¹⁰⁸ The mGlu₅ is a well-established drug target for drug discovery campaigns on central nervous system disorders including depression, autism, and schizophrenia.^109–111 With the photoaffinity labeling, the receptor-ligand contacts can be detected.^{112, 113} To advance the structural biological knowledge of mGlu₅ allosteric modulation, photoprobes were designed and synthesized. For a photoprobe, a photoactive functional group and another bio-orthogonal/click chemistry functional group are introduced. The photoactive functional group will form the covalent bond with the receptor to function as the photoaffinity label. On the other hand, the bio-orthogonal/click chemistry functional group can have bio-orthogonal conjugation reaction to attach a chosen tag like biotin and fluorophore.¹¹⁴ Based on the structures of acetylenic mGlu₅ NAM lead compounds 15-18, photoreactive and click chemistry functional groups were introduced to result in compounds 19-21. The intracellular Ca²⁺ mobilization assay and the [³H]mPEPy inhibition binding assay were conducted to confirm that benzophenone-alkyne (19) and diazido (20) conjugates of lead compounds are NAMs. The compact azido-alkyne derivative (21) of compound 17 was tested to be a NAM as well.¹⁰⁸

Figure 13.

Chemical structures of compounds 15–21.

5. FUTURE PERSPECTIVES AND SUMMARY

Besides the successful stories described above, challenges and opportunities can remain on the following aspects, (1) identification of allosteric binding pockets and locations, (2) issues associated with irreversible binding, and (3) development of the bitopic ligands. The identification of the binding pocket is key for compound design. Class C GPCRs have the conserved rigid transmembrane domain as the site for allosteric interaction while the Class A and B GPCRs are using flexible loops and the external surface of helixes as sites for allosteric modulation. The uncertainty among locations of the binding pockets and the residues flexibility among these pockets intensified the difficulty of rational compound discovery. Irreversible binding can cause undesired drug-induced toxicity including hepatotoxicity, mutagenicity or carcinogenicity. The off-target covalent binding may further attenuate the usage of hyperactive warheads. Instead, soft electrophilic moieties like acrylamide may be referred to as safer functional groups.¹¹⁵ Bitopic ligands, which are also called as dualsteric ligands, are a novel type of ligands that simultaneously target both the orthosteric and allosteric binding pockets.¹¹⁶ Introducing a linker to combine the orthosteric and the allosteric ligands is a common strategy for bitopic ligand design. The developed bitopic ligand is supposed to maintain the high affinity of the orthosteric ligand and the high selectivity of the allosteric ligand towards the protein target, which suggests a new opportunity for future drug discovery.¹¹⁷ As a fast-growing and promising area of research, more studies on GPCRs covalent allosteric modulation are expected. We are optimistic to see how drug discovery campaigns on GPCRs can be benefited by adapting this compound design strategy.

Highlights.

• Designing covalent allosteric modulators brings new opportunities to the field of drug discovery towards GPCRs.

• Reported allosteric binding pockets are summarized for class A, B, and C GPCRs.

• The latest advances on allosteric drug discovery with covalent warheads are investigated in detail.

• Successful case studies on developing GPCRs covalent allosteric modulators are reviewed and discussed.