SARS-CoV-2 M^pro Inhibitors: Achieved Diversity, Developing Resistance and Future Strategies

Abstract

While the COVID-19 pandemic seems to be on its decline, the unclear impacts of long-COVID cases, breakthrough infections in immunocompromised individuals, vaccine hesitancy, and inhomogeneous health-care accessibility constitute a not to be underestimated threat. These cases, along with pandemic preparedness, ask for an alert identification of new drugs and the optimization of existing drugs as therapeutic treatment options for this and potential future diseases. M^pro inhibitors were identified early on as potent drug candidates against coronaviruses, since they target viable processing machinery within the virus, i.e., the main protease that cleaves the polyproteins encoded by the viral RNA into functional proteins. Different strategies, including reversible and irreversible inhibition as well as allosteric inhibitors, mostly from drug repurposing endeavors, have been explored in the design of potent SARS-CoV-2 M^pro antivirals. Ambitious screening efforts have uttered an outstanding chemical and structural diversity, which has led to half a dozen lead compounds being currently in clinical trials and the emergency FDA approval of ritonavir-boosted nirmatrelvir as a COVID-19 therapeutic. This comprehensive analysis of the achieved inhibitor diversity sorted into irreversible, reversible, and allosteric M^pro binders, along with a discussion of emerging resistance reports and possible evasion strategies, is aimed at stimulating continuing M^pro drug design efforts.

1. Introduction

With the current COVID-19 pandemic slowly transitioning to its endemic state, we are confronted with a devastating reality: over 6.6 million deaths worldwide between January 2020 and December 2022, persistent long-COVID cases, and ~3 million confirmed weekly infections (as of December 2022) continue to dramatically impact the socioeconomical motor of society [1]. The vaccination campaign, along with persistent viral exposure, is thought to result in herd immunity; however, the highly adaptive nature of the vaccine targeted spike protein [2,3], campaign fatigue, and globally inhomogeneous vaccine access constitute viral escape routes that fuel a continuous endemic. Mutational analysis of the current SARS-CoV-2 and genomic comparison with previous coronaviruses suggest alternate targets with significantly lower mutation rates and higher structural conservation [4,5]. These include functional proteins that can be targeted with small-molecule inhibitors, which, compared to vaccines, have generally accepted advantages of easier formulation and production, storage, and administration. Among these, two viable main proteases, namely, PL^pro and M^pro, show a high sequence identity of 86% [6] and 96% [7] between SARS-CoV and SARS-CoV-2, thus reflecting druggable molecular targets. Of the two main proteases, M^pro seems the more beneficial target for small-molecule inhibitor development for the following reasons:

• The smaller, spatially more defined active site of M^pro, allowing the synthesis of smaller and stiffer inhibitors, whereas efficient binding to PL^pro usually relies on tight distal inhibitors [8].
• Dimerization of M^pro for activity and close proximity of active site and dimerization interface, suggesting inhibitor design that additionally prevents dimerization [9].
• The lower mutation rates of M^pro (nsp5) compared to PL^pro (nps3), mitigating the risk of mutation-mediated drug resistance [4].
• Less off-target effects expected, since M^pro’s glutamine (Gln) cleavage site recognition is unique and has not been not observed for any human protease. PL^pro inhibitors might interfere with human ubiquitin binding motives [10].

Coronavirus replication relies on the expression of two overlapping polyproteins (pp1a and pp1b) as well as four structural proteins encoded by the viral RNA. Both polyproteins liberate vital proteins needed for replication by the action of the main proteases. M^pro, a 3-chymotrypsin-like protease (3CL), performs 11 cleavages, whereas PL^pro performs only three cleavages, and M^pro uniquely recognizes a glutamine in P1 and small, nonpolar residues in P1′ and P2.

This review aims to highlight the current landscape of SARS-CoV-2 M^pro inhibitors with emphasis on the mechanistic differences in inhibition and achieved efficacies. Therefore, Pubmed and SciFinder databases were cross-searched using the search term string “SARS-CoV-2 M^pro inhibitor” OR “SARS-CoV-2 3CL inhibitor”, excluding in silico studies using the NOT “in-silico” term. The ~900 research articles found dating January 2020–November 2022 were sorted manually into three groups: irreversible competitive, reversible competitive, and noncompetitive/allosteric inhibition strategies, and reflect on structure optimization efforts as well as pitfalls. Only chemically synthesized or repurposed drugs that have been biochemically tested and ideally verified by co-crystallization are considered here. Their efficacy, potency, and toxicological data along with available pharmacokinetic profiles are compared. As there is growing evidence for potential viral resistance of some of these inhibitors, including the FDA-approved paxlovid M^pro inhibitor nirmatrelvir, we summarize recent reports and point to potential extended therapy solutions.

2. Competitive Inhibitor Design Strategies

M^pro is a homodimeric cysteine protease, and effective inhibition of the active site relies on the covalent interaction with the catalytic dyad consisting of cysteine residue 145 (Cys145) located on domain II and histidine 41 (His41) located on domain I (Zhang 2020, [7,11] (Figure 1). Nucleophilic attack of the inhibitor by the cysteine thiolate is preceded by His41-moderated thiol deprotonation, increasing sulfur nucleophilicity [12]. This reflects a generally accepted two-step mechanism that is different from the mechanism of serine proteases such as chymotrypsin that have a catalytic triad that supports a concerted mechanism. Tight binding of the inhibitor causes a slight conformational change that brings both catalytic residues into proximity, thus initiating covalent interaction with the inhibitor warhead. Promiscuity in M^pro seems to follow the general trend of being somewhat higher than that in serine proteases [13], as indicated by the longer distance between the dyad residues, giving rise to various peptide and nonpeptide inhibitor geometries and different warhead chemistries. The PDB currently hosts 294 structures of covalently bound competitive M^pro inhibitors, which serve as a basis for the structural and mechanistic analysis in this chapter. Chemoinformatic analysis indicates that the vast majority of crystallized inhibitors follow Lipinski design rules, including a molecular weight below 500 Da (M_w,aver = 449.5 ± 144.9 g/mol). Aside from covalently bound inhibitors, which are covered in the first part of this chapter, an increasing number of inhibitors that are noncovalently bound in the active site can be found in the recent literature, and these are described in a third subsection. These generally tend to be smaller in size (M_w,aver = 346.8 ± 122 g/mol), as is obvious from the comparative structural analysis of their PDB entries (123 PDB structures in total).

2.1. Predominantly Irreversible Warheads

The search for SARS-CoV-2 M^pro inhibitors was initiated by a structure-based high-throughput screening of a >10,000 compound database of approved drug molecules, advanced clinical trial drug candidates, and other bioactive compounds. The pseudo-tetrapeptide N3, a Michael acceptor, previously reported as a SARS-CoV and MERS-CoV M^pro inhibitor, was found to bind SARS-CoV-2 M^pro time-dependently with k_obs/[N3] of 11,300 M⁻¹s⁻¹ [14] (Figure 2; Table 1). A plaque reduction assay (VeroE6 cells) provided an EC₅₀ value of 16.77 µM. The crystal structure of the SARS-CoV-2 M^pro complex (PDB 6lu7) confirms the β-vinyl carbon being covalently bound to Cys145’s sulfur atom (1.8 Å C-S distance; Figure 2). The P1 lactam ring is held in position by various contacts, including His163, Glu166, and backbone atoms of Phe140, Asn142, and His172 and two water molecules. The side chain of Leu in P2 fits into the hydrophobic enzyme S2 pocket, maintaining several van-der-Waals interactions. The isopropyl group of Val in P3 is solvent-exposed, while Ala in P4 is surrounded by Met165, Leu167, Phe185, Gln192, and Gln189. The same screening study also revealed the 5-hydroxytryptamine receptor antagonist cinanserin as an M^pro binder (IC₅₀ = 125 µM; EC₅₀ (VeroE6) = 20.61 µM), reflecting another example of a repurposed drug that is known as a potent inhibitor of other coronaviruses [15].

Cinanserin also resembles an α,β-unsaturated carbonyl drug, which is supposed to covalently bind to Cys145 similarly to N3; however, crystal structure confirmation of this interaction is yet not available. Pharmacological profiles for both drugs have not been accessed; however, toxicological data point to an acceptable safety level (CC₅₀ >130 µM (N3) and >200 µM (cinanserin)) [14]. In vitro screening of successful SARS-CoV M^pro inhibitors uncovered the N3 analog 1 as a nanomolar inhibitor of SARS-CoV-2 M^pro (IC₅₀= 0.15 µM), able to inhibit replication in VeroE6 cells (EC₅₀ = 2.88 µM) [16]. In the crystal structure, inhibitor 1 forms a covalent bond between its β-vinyl carbon and the Cys145 sulfur atom of the main protease (PDB 7jt7 and 7jw8). The ester portion expands into subsite S1′, the γ-lactam ring in P1 is stabilized in S1 by hydrogen bonding involving Phe140, His164, and His166, and the hydrophobic leucine and O-tert-butyl threonine side chains occupy S2 and the surface S3 subsites, respectively. Virtual screening efforts to develop peptidomimetic N3 analogs resulted in a series of α,β-unsaturated carbonyl compounds with variable hydrophobic P1–P3 substituents. Biological evaluation of the promising lead compound 2 from this set found only moderate M^pro inhibition (IC₅₀ = 47 µM, [17]). In a search to expand the chemical space of anti-SARS-CoV-2 M^pro α,β-unsaturated carbonyl drugs, recently, also nonpeptidomimetic compounds have been designed [18]. They feature a fused benzo[b]-[1,4]oxazinone-imidazo [2,1-b]thiazole system, which beneficially orients substituents on the aromatic portion of these molecules towards the main protease subsites, as suggested by molecular docking. Biological testing revealed compound SIMR-2418 (Figure 3) as a micromolar SARS-CoV-2 M^pro inhibitor (IC₅₀ = 21 µM) with acceptable ADME properties (t_0.5 > 1 h; mouse LMS).

Another virtual screening study identified nonpeptidomimetic acrylamides as cysteine reactive M^pro inhibitors [19]. The most potent compound from this series is only active in its S-configuration (IC₅₀ =2.9 µM, K_i = 38 µM; Table 1). A co-crystal structure of LON-WEI-adc59df6-47 (PDB 7nw2) confirms covalent bond formation between the β-vinyl carbon and the active site sulfur atom. From the natural pool of Michael acceptors, the flavonoid myricetin was tried for its anti-SARS-CoV-2 properties (Figure 3) [20]. In vitro screening confirmed sub-micromolar inhibition (IC₅₀ = 0.2–0.6 µM), and in the solid-state structure, the covalent linkage to C6 of the pyrogallol unit is obvious (PDB 7b3e). Alternatively, a noncompetitive inhibition mechanism is proposed in the literature [21]. Myricetin’s carbonyl-rich structure enables stabilization by hydrogen-bonding, which is moderated through active site water molecules [22]. Covalent bond formation via pyrogallol-C6 can be explained by a preceding oxidation of pyrogallol to o-quinone, enabling attack of the nucleophilic Cys145 thiolate. Myricetin was further shown to block viral infection in vitro (EC₅₀ = 8 μM (VeroE6, [22]); 0.9 μM (Calu-3), [21]) and reduce lung inflammation in a mouse model of bleomycin-induced pulmonary inflammation, making it potentially useful for the symptomatic treatment of COVID-19 [23]. To increase potency, myricetin analogs were designed that included compound 3 with a diphenyl phosphate to the 7-OH of myricetin, resulting in a stronger antiviral effect (Figure 3; EC₅₀ = 3.15 μM) [22].

Figure 3. Chemical structure, M^pro binding affinity, and antiviral activity of different designed Michael acceptor inhibitors 1 [16], 2 [17], SIMR-2418 [18], LON-WEI-adc59df6-47 [19], myricetin [22], and 3 [22]. The (β-vinyl)-carbon reacting with Cys145 is highlighted in each structure. In the case of myricetin and its analog 3, the reactive Michael acceptor structure becomes accessible after oxidation of the pyrogallol unit.

Figure 3. Chemical structure, M^pro binding affinity, and antiviral activity of different designed Michael acceptor inhibitors 1 [16], 2 [17], SIMR-2418 [18], LON-WEI-adc59df6-47 [19], myricetin [22], and 3 [22]. The (β-vinyl)-carbon reacting with Cys145 is highlighted in each structure. In the case of myricetin and its analog 3, the reactive Michael acceptor structure becomes accessible after oxidation of the pyrogallol unit.

An early drug-repurposing study identified the anticancer drug Carmofur, featuring a carbamoyl warhead, as another type of irreversible M^pro inhibitor [24] that also inhibits other SARS-CoV-2 target proteases [25]. Upon acylation of Cys145, releasing 5-fluoro-uracil as side product, the 1-hexylcarbamoyl chain maintains several interactions with Gly143 and hydrophobic residues in S2, as obvious from its co-crystal structure (PDB 7buy). Flexibility and small overall size of the active inhibitor result in an IC₅₀ of 1.82 µM and a half-maximal effective concentration (VeroE6) of ~25 µM. Cell toxicity was determined to be around 150 µM (Table 1).

Recently, there has been renewed interest in developing optimized peptide and peptide-mimetic inhibitors based on the initial success of Michael acceptor warheads. The exchange of the α,β-unsaturated carbonyl for hydroxymethyl and alkoxymethyl ketones gave rise to another class of potent M^pro inhibitors [26,27]. Reaction of these highly electrophilic ketones with the active site thiolate resulted in a hemithioketal that, unlike Michael-type inhibitors, possesses a coordination-relevant OH group supporting additional stabilizing interactions with the active site residues. Among these, the hydroxymethyl ketone PF-00835231 (Figure 4) has been identified as a nanomolar M^pro binder (IC₅₀ = 0.007 µM) and effective antiviral (EC₅₀ = 35.9µM, Vero6, USA/WA1/2020), the potency of which can be increased almost 80-fold by the application of an efflux inhibitor (EC₅₀ = 0.46µM, 2µM CP100356; Table 1). Thermal shift assays as well as an X-ray crystal structure (PDB 6xhm) confirm tight binding to the active site, including a dense array of stabilizing hydrogen bonds in S1′ (Figure 4) [28]. The compound is safe in cell assays and displays pharmacokinetic profiles in animals (rat, dog, monkey) that were beneficial for preclinical studies (Table 1). Solubility issues and low oral bioavailability (>2% in rat, dog, monkey) can be overcome by transformation of PF-00835231 into its phosphate prodrug PF-07304814, which is suitable for intravenous injection formulation, releasing about 75% active drug by human alkaline phosphatase cleavage [26]. Extension of the C-terminal alcohol by an alkyl or aryl group results in inhibitors with an α-acyloxymethyl ketone warhead. Unlike the hydroxymethyl ketone PF-00835231, acyloxymethyl ketones possess an electrophilic α-carbon that can be attacked by the active site thiolate, resulting in an irreversible covalent bond, as confirmed by the cocrystal structure of 4 (PDB 7mbi; [27]). By experimental variation, benzoyloxymethyl ketones were identified as the most potent M^pro inhibitors, with 2,6-disubstituted aryl rings additionally lowering the pKa of the corresponding acid, thus reducing IC₅₀ values down to single-digit nanomolar (Figure 5). EC₅₀ values of 4 and 5 are 0.3 and 0.16 µM, respectively, with no observed cell toxicity up to the highest tested concentration (200 µM) [27]. With a mouse plasma half-life > 4 h, logD of 4.4, and basolateral-to-apical permeability > 20 × 10⁻⁶ cm/s, 4 promises beneficial ADME properties and nondiscriminant thiol reactivity. The six-membered lactam ring seems only to have a marginal effect on the potency of these inhibitors. While highly selective for cysteine proteases over other cellular proteases, these inhibitors act promiscuous towards human cathepsin B and S with IC_50′s for 4 of 0.2 µM for CatB and 0.05 µM for CatS.

Similar to α-acyloxymethyl ketones, α-halomethyl ketones are known as alkylating agents and potent inhibitors for cysteine proteases such as human calpain and cathepsin B [29]. Promiscuity and metabolic stability issues [30] need to be addressed to develop highly selective agents against SARS-CoV-2 M^pro. Although peptidic α-halomethyl ketones occur as synthetic intermediates towards α-acyloxymethyl ketones, their inhibitory action towards SARS-CoV-2 M^pro has not been tested in the recent literature, probably due to the mentioned lack of selectivity. Structural design therefore seems to concentrate on more specific peptide mimetics such as α-chloroacetamide 6, which has been identified as a potent SARS-CoV-2 M^pro inhibitor (IC₅₀ = 0.4 µM) in a comprehensive in vitro study [31].

The crystal structure of 6 confirms displacement of the chloride by the active site thiolate and covalent bond formation (PDB 7mlf). The pyridine sits in the S1 pocket, fixed by a hydrogen bond with the nearby His163, the tert-butylphenyl substituent spreads out into S2, maintaining hydrophobic interactions, and the cycloalkyl group expands towards S4 (Figure 6). Structural optimization of the “side chain” residues shows that a smaller cyclopentyl ring (R3) gives slightly better (IC₅₀ = 0.38 µM) and 4-pyridyl or 5-methylpyrazin-2-yl moieties in R1 slightly worse (IC₅₀ = 0.8–1 µM) inhibition. Regarding the warhead, a chloromethyl group is more effective than fluoromethyl is, as a consequence of the higher electrophilicity of C-Cl. This electrophilicity can even be further increased by attachment of a second or third halogen, as demonstrated in a related study by Ma et al. using a slightly altered backbone structure [32]. The α,α-dichloroacetamide Jun9-62-2R, accessible by a simple one-pot Ugi four-component reaction, shows beneficial inhibitory data (IC₅₀ = 0.43 µM; EC₅₀ = 0.9 µM) and low cell cytotoxicity (CC₅₀ > 100 µM, Vero6; Table 1). The tribromoacetamide Jun9-88-2R is even more potent (IC₅₀ = 0.08 µM; EC₅₀ = 0.58 µM) but has a significant lower therapeutic index (CC₅₀ = 5.5 µM, Vero6), likely due to interference with other vital proteases that seems predictable for this substitution pattern [33]. Within their comprehensive SAR study, the authors found a tighter inhibition for the R-configurated isomers, as observed from the cocrystal structure of Jun9-62-2R (PDB 7rn1, Figure 6). A noteworthy, optimized framework uses an α-chloro-α-fluoroacetamide moiety as warhead, allowing for a strong and selective inhibition of SARS-CoV-2’s main protease [34]. A lead derivative from this series, 7, with a p-pentafluorosulfanyl phenyl group in P2 and a 5-pyrimidinyl group in P1, has an IC₅₀ of 0.056 µM (K_i = 1.34 µM) and acts selectively only in its (R,R)-cis configuration. More recently, novel pyrazoline skeleton derivatives bearing a chloroacetamide warhead also showed promising SARS-CoV-2 M^pro inhibition [35]. A lead compound from this series, 8, has an IC₅₀ of 0.53 µM in its R-configuration, while the S-enantiomer is five times less active. The crystal structure of this compound interestingly shows no occupants in the coordination-rich S1 and S2 subsite; therefore, the 3-phenyl substituent points in the S4 direction and the 5-quinoxalin-6-yl group towards S1 (Figure 6). Instalment of a benzene on C4 and optimization of the C3 substituent gave an even tighter M^pro binder, as confirmed by a fluorescent substrate assay. Nonpeptidomimetic chloroacetamide-based inhibitors have been developed in silico; however, likely due to more flexible backbone geometries, none of the tested lead compounds have sub-micromolar inhibition capacity [36]. None of the herein presented inhibitors with chloromethyl ketone warhead have been tested for their pharmacokinetic parameters yet.

Table 1. Different warhead chemistries, antiviral, and pharmacokinetic data of irreversible SARS-CoV-2 M^pro inhibitors.

Warhead	Drug Example a	Efficacy	Potency	Toxicology	Pharmacology			Lit.
		Ki/IC50 (µM)	EC50 (µM)	CC50 (µM)	t0.5 (h)	Cmax (ng/mL)	Clear. (mL/min/kg)
Michael acceptor	N3, Cinanserin, 1	- 1: 0.15 (IC50)	N3: 16.77 1: 2.88	N3: >130 1: >200	-	-	-	[14] [16]
Acrylamide	LON-WEI-adc59df6-47	2.9 (IC50) 38.4 (Ki)	-	-	-	-	-	[19]
Carbamoyl	Carmofur	1.82	24	133	-	-	-	[24]
Hydroxymethyl ketone	PF-00835231	0.0069 (IC50)	0.46 b	>50 (VeroE6)	0.72 (rat)	1250 (rat)	27 (rat)	[26]
Acyloxymethyl ketone	4, 5	0.001	0.16	>200	>4 (mouseplasma)	-	-	[27]
Chloroacetamides	6, Jun9-62-2R, Jun9-88-2R	6a: 0.4 (IC50) Jun9-62-2R: 0.43 (IC50) Jun9-88-2R: 0.08 (IC50)	6a: -Jun9-62-2R: 0.9 Jun9-88-2R: 0.58	6a: -Jun9-62-2R: >100 Jun9-88-2R: 5.48	-	-	-	[31]
Chlorofluoro- acetamides	7	0.056 (IC50) 1.34	-	-	-	-	-	[34]
Vinylsulfonamide	11	2.3 (Ki) 0.17 (IC50)	-	-	-	-	-	[31]

^a Bold entries resemble inhibitors corresponding to reported activity, safety, and pharmacology data; ^b using an efflux inhibitor.

Table 1. Different warhead chemistries, antiviral, and pharmacokinetic data of irreversible SARS-CoV-2 M^pro inhibitors.

Warhead	Drug Example a	Efficacy	Potency	Toxicology	Pharmacology			Lit.
		Ki/IC50 (µM)	EC50 (µM)	CC50 (µM)	t0.5 (h)	Cmax (ng/mL)	Clear. (mL/min/kg)
Michael acceptor	N3, Cinanserin, 1	- 1: 0.15 (IC50)	N3: 16.77 1: 2.88	N3: >130 1: >200	-	-	-	[14] [16]
Acrylamide	LON-WEI-adc59df6-47	2.9 (IC50) 38.4 (Ki)	-	-	-	-	-	[19]
Carbamoyl	Carmofur	1.82	24	133	-	-	-	[24]
Hydroxymethyl ketone	PF-00835231	0.0069 (IC50)	0.46 b	>50 (VeroE6)	0.72 (rat)	1250 (rat)	27 (rat)	[26]
Acyloxymethyl ketone	4, 5	0.001	0.16	>200	>4 (mouseplasma)	-	-	[27]
Chloroacetamides	6, Jun9-62-2R, Jun9-88-2R	6a: 0.4 (IC50) Jun9-62-2R: 0.43 (IC50) Jun9-88-2R: 0.08 (IC50)	6a: -Jun9-62-2R: 0.9 Jun9-88-2R: 0.58	6a: -Jun9-62-2R: >100 Jun9-88-2R: 5.48	-	-	-	[31]
Chlorofluoro- acetamides	7	0.056 (IC50) 1.34	-	-	-	-	-	[34]
Vinylsulfonamide	11	2.3 (Ki) 0.17 (IC50)	-	-	-	-	-	[31]

^a Bold entries resemble inhibitors corresponding to reported activity, safety, and pharmacology data; ^b using an efflux inhibitor.

Vinyl sulfones are another known warhead for irreversible cysteine protease inhibition [37]. Upon reaction with the active site thiolate, they establish a covalent bond with the inhibitor β-vinyl carbon atom. Decorating the discussed pyrazoline skeleton with such a functionality results in the nanomolar SARS-CoV-2 M^pro binder 9 (Figure 7, IC₅₀ = 0.035 µM), which shows promiscuity to other coronaviruses, including SARS-CoV, HCoV-NL63, IBV, and HCoV-229E [35]. Other nonpeptidic inhibitors with vinyl sulfone warheads have been developed based on the successful α-chloroacetamide 6. Swapping out the warhead gives molecule 10 with a IC₅₀ of 0.42 µM, which was further optimized by tuning the P3 residue to a 3-chlorophenyleth-2-yl group (11; IC₅₀ = 0.17 µM) [31]. Isothermal titration calorimetry as well as a crystal structure of 10 (PDB 7mlg) confirm covalent bond formation. The orientation of the molecule in the active site is due to the unchanged P1 and P2 residues almost identical with 6.

Inhibitors with an ester warhead are another category of irreversible M^pro agents that has been considered in drug development during this pandemic. Nucleophilic attack of the thiolate results here in the departure of a leaving group and irreversible enzyme acylation. Interestingly, usage of this warhead has so far been limited to small-molecule inhibitors involving rigid pyridine and/or indole systems, exemplified by the molecule GRL-0820 (Figure 8). Among a screen of structural analogs, GRL-0820 stands out with an M^pro IC₅₀ of 0.073 µM and tolerable anti-SARS-CoV-2 activity (EC₅₀ = 15 µM, VeroE6, [38]). Alternate connection or aromatic substitution of the indole unit leads to even more potent M^pro binders [15,39]; however, the viral activity of these ester compounds is either unknown [39] or negligible [15]. Recently, a similarly rigid 5-chlopyridinyl ester of salicylic acid was proposed as a drug candidate [40], showing micromolar potency and anti-SARS-CoV-2-viral activity (12, IC₅₀ = 4.9 µM, EC₅₀ = 25 µM). Pharmacokinetic evaluation of ester warhead inhibitors is still pending.

2.2. Reversible Warheads

Reversible inhibitors possess a reactive carbonyl function that, upon reaction with the nucleophilic cysteine thiolate, forms a temporary addition product. The resulting covalent bond is less strong than that for the previously described irreversible warheads, thus supporting time-dependent reversal of the binding event and drug clearance. Functional groups that establish a reversible covalent bond include foremost aldehydes, α-ketoamides, and some α-substituted ketones as well as nitriles. Aldehydes and ketones form hemithioacetals, whereas nitriles form a reversible thioimidate (Figure 9). It seems an acceptable consensus that reversible inhibition is preferable as, in general, it allows for more selective target binding and fewer off-target effects, and promotes bioavailability and efficacy of respective inhibitors and their ultimate clearance from the infected host. It therefore does not surprise that among the 500+ covalent anti-SARS-CoV-2 M^pro drug candidates, predominantly those with reversible warhead chemistries were forwarded to clinical testing (see next chapter).

Aldehydes represent the biggest investigational group among reversible SARS-CoV-2 M^pro inhibitors. Chronologically, the two peptides 13 and 14 were the first tested aldehydes, rationally modeled after successful SARS-CoV M^pro inhibitors [41]. They feature a γ-lactam group in P1, a six-membered aliphatic or aromatic side chain in P2, and an indole moiety in P3. Both drug compounds show remarkable in vitro activity (IC₅₀ = 0.053 and 0.04 µM, 13 and 14), strong antiviral effect (EC₅₀ = 0.53 and 0.72 µM, 13 and 14), low cytotoxicity, and preferable pharmacokinetic parameters. Beneficial complementarity of the used side chains is highlighted in the obtained X-ray structures (PDB 6lze (13) and 6m0k (14)): an additional H-bond between the thioacetal –OH group and Cys145 backbone enhances stabilization of the covalent bond. The γ-lactam group maintains H-bonds with Glu166 and His163 and an additional contact with Phe140 in S1, the cycloaliphatic (13) or aromatic (14) P2 residue penetrates the S2 pocket, supported by several hydrophobic interactions, and the indole moiety is located at the solvent interface (S4).

A structural analog of 14 with a benzyl group in P2 (15) shows improved binding (IC₅₀ = 0.034 µM) and even better antiviral activity (EC₅₀ = 0.29 µM) [42]. Vuong et al. explored the known peptide aldehyde GC373 and its bisulfite adduct GC376 as potent reversible anti-SARS-CoV-2 M^pro agents [43]. Both drugs were initially investigated as inhibitors of a feline coronavirus (FCoV) that causes a fatal inflammatory disease in cats (feline infectious peritonitis). Under physiological conditions, the bisulfite adduct GC376 reverses to the active aldehyde drug GC373 (Figure 10). Since FCoV and SARS-CoV-2 M^pro are highly homologous cysteine proteases and share close resemblance of their active site geometries, a repurposing attempt seemed reasonable. Indeed, both drugs proved effective as new anti-COVID-19 agents with half-maximal inhibitory concentrations, M^pro-IC_50, of 0.40 µM (GC373) and 0.19 µM (GC376) [43]. Several hydrogen bonds with the oxyanion hole residues Gly143, Ser144, and Cys145 stabilize the hemithioacetal of GC373, as is obvious from its cocrystal structure [43,44]. Inhibitor interactions in S1 and S2 are identical to those of the aldehydes 13 and 14, the benzyloxycarbonyl group in P3 coordinates with Glu166. In the cell infectivity assay, both drugs showed micromolar effectivity (GC373: EC₅₀ = 1.5 µM; GC376: EC₅₀ = 0.9 µM) and low cell toxicology (CC₅₀ for both >100 µM). Efficacy of GC376 against SARS-CoV-2 infection was shown in a transgenic mouse model expressing the human angiotensin-converting enzyme type 2 (K18 hACE2) [45]. Pharma-cokinetic profiling performed by Shi et al. indicated a quick and durable uptake of GC376 after intramuscular injection in mice and rats (t₀.₅ = 1.1 h (mice) and 3.9 h (rat)), a high peak plasma concentration, and good clearance [46] (Table 2). Reversibility of GC376 binding was demonstrated with a ¹³CHO-labeled GC376 derivative in a time-dependent NMR experiment [47]. To find even more potent structural analogs of GC376, the Vederas group explored various synthetic derivatives with different P2 and P3 substitutions for their anti-SARS-CoV-2 behavior [48]. The two lead compounds 16 and 17, with a cyclopropyl group in P2 and an m-halophenyl substituent in P3, were identified in this series, displaying improved IC₅₀ values of 0.07 and 0.08 µM, respectively. Both compounds showed slightly better antiviral susceptibility compared to GC376 (EC₅₀ = 0.57 and 0.7 µM, respectively) and no cytotoxicity [48]. A crystal structure of 16 with SARS-CoV-2 M^pro highlights an enhanced interaction of the P2/P3 groups with respective enzyme subsites (PDB 7lco). Further structural optimization of the GC376 core led to the two potent antiviral drug candidates UAWJ247 (IC₅₀ = 0.045 µM) [49] and NK01-63 (coronastat, IC₅₀ = 0.016 µM) [50]. The latter compound has a pronounced antiviral effect (EC₅₀ = 0.006 µM; Huh-7ACE2-infected cells) and high selectivity over other human proteases. Aside from the common interaction patterns, the NK01-63 cocrystal structure features two additional H-bonds of the P3 trifluoromethylphenyl group to nearby Asn142, reiterating the prominent stabilization effect of halogens (PDB 7tiz). In vitro mice tests confirmed good microsomal stability, and in vivo pharamacokinetic analysis underlined high uptake via both intraperitoneal and oral routes, with an extended half-life in critical tissues such as lung (I.P., t_0.5 = 2.2 h; P.O. t₀.₅ = 3.4 h). Another example of activity-guided optimization of GC376′s structure led to analog 18, which has a constrained bicyclic ring system in P3 [51]. M^pro FRET substrate assay of 18 indicated an IC₅₀ of 0.18 µM and nanomolar efficacy in the plaque reduction assay (EC₅₀ = 0.035 µM; VeroE6).

Inspired by the structurally related FDA-approved antivirals boceprevir and telaprevir, Xia et al. also synthesized and tested boceprevir/telaprevir-GC376 chimera [52]. Some of these analogs turned out to be nanomolar SARS-CoV-2 M^pro inhibitors (19: IC₅₀ = 0.054 µM; 20: IC₅₀= 0.051 µM) and (sub)micromolar antivirals (VeroE6, 19: EC₅₀ = 0.37 µM; 20: EC₅₀ = 2.6 µM). X-ray data for both compounds (PDB 7lyi (19) and 7lyh (20)) confirmed successful insertion of the bicyclic proline moieties into the S2 subpocket. Additional modification of the P3 substituent, as tried in an extensive SAR study by Qiao et al., led to the superior compound MI-23 with an IC₅₀ of 0.008 µM [53]. In the cocrystal structure, the extended 1-ethyl-3,5-difluorobenzene substituent in P3 interacts via hydrophobic contacts with nearby Leu167 and Pro168, additionally contributing to a tight fit of this inhibitor. Presumably due to a limited cell membrane permeability, MI-23 and other inhibitors of this series had reduced activity in the cell protection assay (EC₅₀ = 5.6 µM), incentivizing the authors to select further structurally modified P3 analogs for pharmacological testing. The analog MI-30 (21) with a (2,4-dichlorophenoxy)acetyl group in P3 had a persistent low IC₅₀ of 0.017 µM, good in vitro efficacy (EC₅₀ = 0.54 µM), no common cell toxicology, and protected transgenic mice that expressed hACE2 against SARS-CoV-2 infection. In an SD rat model, MI-30 showed good in vivo stability (t₀.₅ = 2.35 h), high plasma concentrations (C_max = 23582 ng/mL), and sufficient clearance (CL = 17.1 mL/min/kg).

To improve the surface interaction of inhibitory peptides, Yang et al. shifted their focus to optimizing tripeptides as SARS-CoV-2 M^pro inhibitors [54]. A potent aldehyde from their series (MPI8) has the common γ-lactam ring in P1, a cyclohexyl group in P2, and a benzyloxycarbonyl-protected tert-butylserine in P3, and displayed an IC₅₀ of 0.11 µM and good in vitro antiviral efficacy (EC₅₀ = 2.5 µM). Indeed, the cocrystal structure of MPI8 (PDB 7jq5) reveals a bifurcated orientation of the P3 residue covering an extended area of the solvent-exposed S3/S4 site, highlighting enhanced interaction with several hydrophobic M^pro residues. In an extended study by the same group, the authors classify MPI8 as dual inhibitor of M^pro and cathepsin L, showing nanomolar half-inhibition concentrations for both cysteine proteases [55]. Cathepsin L, among other membrane proteases, is suspected to facilitate coronaviral entry, giving dual inhibitor aldehydes such as MPI8 superior viral-neutralization qualities. This concept was further explored in various drug-repurposing studies focusing on well-known proteasome inhibitors. As such, MG-132 was shown to effectively bind to and inhibit both SARS-CoV-2 and cathepsin L [56,57]. Another set of studies proved calpain inhibitors II and XII as potent SARS-CoV-2 M^pro binders (IC₅₀ = 0.97 and 0.45 µM, respectively) with a micromolar antiviral effect [39,58,59].

Figure 10. Chemical structures of aldehyde-based M^pro inhibitors 13–15 [41,42], GC376 and GC373 [43], 16 and 17 [48], UAWJ247 [49], NK01-63 [50], 18 [51], 19 and 20 [52], MI-23 and MI-30 (21) [53], MPI8 [54], as well as MG-132 [56]. Inhibitory (IC₅₀) and antiviral data (EC₅₀) are given below each structure.

Figure 10. Chemical structures of aldehyde-based M^pro inhibitors 13–15 [41,42], GC376 and GC373 [43], 16 and 17 [48], UAWJ247 [49], NK01-63 [50], 18 [51], 19 and 20 [52], MI-23 and MI-30 (21) [53], MPI8 [54], as well as MG-132 [56]. Inhibitory (IC₅₀) and antiviral data (EC₅₀) are given below each structure.

A second important reversible warhead functional group that has been extensively implemented in M^pro inhibitor design comprises α-ketoamides. This function is especially valuable since, in addition to the hemithioacetal generated by the reaction of the ketone with the cysteine sulfur, the neighboring ketoamide can also engage in stabilizing hydrogen bonds. A recurring bonding motif for this function involves a hydrogen bond from the hemithioacetal –OH with His41 and hydrogen bonds from the amide group with backbone amides Gly143, Ser144, and Cys145. Some of the first proposed peptide-based inhibitors during the pandemic were compounds 22 and 23 (Figure 11), prominent SARS-CoV inhibitors. Their sub-micromolar performance against SARS-CoV-2 was confirmed by FRET M^pro assay, and optimized compound 23 blocked SARS-CoV-2 viral replication with EC_50′s > 5 µM [7,60]. The solid-state structure of 23 with M^pro confirms the expected hydrogen bond network around the active site Cys145, orienting the C-terminal benzyl group towards the S1′ site. S1 is accommodated by the five-membered ring glutamine surrogate, S2 hosts the cyclopropyl side chain, and the pyridyl moiety in P3 seems beneficial in supporting additional hydrogen bonds with Glu166 that line up the N-terminal Boc group towards the surface S3 pocket. Beneficial ADME and pharmacokinetic data of 23 (Table 2) suggest pyridone-containing inhibitors as a viable framework towards anticoronaviral drugs. Based on the GC376 core structure, α-ketoamides UAWJ246 and UAWJ248 were developed (Figure 11) [49]. In vitro testing indicated a negligible effect of exchanging the aldehyde for a ketoamide warhead, with UAWJ246 scoring an IC₅₀ of 0.045 µM and an EC₅₀ of 4.61 µM. Other repurposing studies identified the hepatitis C antivirals boceprevir and telaprevir as promising model compounds to combat SARS-CoV-2 (Figure 11) [61,62]. Cocrystal structures with both drugs confirm hemithioactal formation and a surprisingly snug fit of both drugs into the protease subsites (boceprevir: PDB 6zru; telaprevir: PDB 6zrt). While the IC₅₀ of telaprevir was determined as 18 µM [63] and cell studies indicated cytotoxicity above 60 µM [64], boceprevir seemd more active (IC₅₀ = 4.13 µM), inhibited SARS-CoV-2 replication with an EC₅₀ value of 1.95 µM, and was not toxic in Caco-2 cells (CC₅₀ > 100 µM) [61]. A different study found the proteasome inhibitor A, which inhibits proliferation of various cell cancer lines in nanomolar concentrations, a suitable COVID-19 antiviral [65]. The VeroE6 cell assay determined an even better antiviral effect (EC₅₀ = 1.28 µM) of A compared to the related tripeptides with an aldehyde warhead.

A third category of reversible M^pro inhibitors uses a nitrile warhead to establish a temporary covalent bond with the enzyme. A nucleophilic Cys145 thiolate attack of the electrophilic nitrile carbon results here in the formation of a reversible thioimidate. Although nitrile-based peptidomimetic SARS-CoV-2 inhibitors developed as past (corona)viral defense agents showed only moderate inhibition success [66], the nitrile function is a chemically valuable way to reduce the number of hydrogen bond donors in the molecule [67]. This feature is helpful for the development of orally available drug candidates, as a reduced number of hydrogen bond donor functions increases drug permeability in the gut. Using this chemical masking strategy, Pfizer chemists developed, in a comprehensive drug development program, nirmatrelvir, influenced by structural optimization of the α-hydroxymethyl ketone PF-00835231 and the benzothiazol-2-yl ketone candidate 24 (Figure 12) [67]. The concept proved successful, with nirmatrelvir showing nanomolar SARS-CoV-2 M^pro inhibition (IC₅₀ = 0.019 µM), a strong antiviral effect (EC₅₀ = 0.075 µM), and no cell toxicity [68]. The X-ray structure shows thioimidate formation of the inhibitor with Cys145 of the main protease, supported by stabilizing hydrogen bonds of the imine with nearby Gly143 (PDB 7vh8 and 7vlq) [69]. The choice of an N-terminal trifluoroacetyl group in the drug design is supported by stabilizing contacts with Gln192 in the S4 pocket. In vivo animal tests confirmed desired intestinal permeability features, excellent oral bioavailability, and proper drug clearance (Table 2). A mouse-adapted SARS-CoV-2 (SARS-CoV-2 MA10) model responded with significantly reduced virus levels in the lungs after 4 days of a twice-daily treatment regime with nirmatrelvir (300 mg/kg) compared to the placebo group. Excellent in vitro and pharmacokinetic performance as well as favorable off-target selectivity qualified the drug candidate for clinical trials, consequentially leading to emergency FDA approval of nirmatrevir in December 2021. During the preclinical trials, it was found that coadministration of the anti-HIV drug ritonavir increased drug performance due to ritonavir’s inhibitory action on CYP-mediated metabolism. This ultimately resulted in coformulation of nirmatrelvir with ritonavir in the final drug paxlovid. In light of the mutational adaptivity and widespread transmission of different SARS-CoV-2 variants of concern, it can be seen encouraging that several studies confirmed nirmatrelvir’s ability to efficiently neutralize infection of beta, delta, and omicron variants of the virus in animal models [70,71,72,73,74]. Regarding effectivity in humans, a manufacturer-conducted study indicated persistent pharmacokinetic performance, viral clearance, and drug safety in healthy participants [75]. Related studies with renally impaired subjects showed increased peak plasma concentrations of the drug and delayed clearance correlating with severity of renal disease, thus recommending a dose reduction for this impairment group [76]. Driven by the success of nirmatrelvir, several analogs with adaptations of the P2 and P3 sites have been developed. A related patent by Pfizer identified compound 25 as a strong main protease binder (K_i = 0.004 µM) and effective SARS-CoV-2 antiviral (EC₅₀ = 0.019 µM) [77]. A different group considered elements of GC376 and its derivatives for their inhibitor optimization screening. Among their synthesized compounds, nitrile 26 was the most active in the inhibition assay (IC₅₀ = 0.009 µM) and potent in reducing viral replication (EC₅₀ = 2.2 µM) [78].

Table 2. Reversible inhibitors of SARS-CoV-2 M^pro with inhibitory and pharmacological data.

Warhead	Drug Example	Efficacy	Potency	Toxicology	Pharmacology			Lit.
		IC50 (µM)	EC50 (µM)	CC50 (µM)	t0.5 (h)	Cmax (ng/mL)	Clear. (mL/min/kg)
Aldehydes	13–21, GC376, MPI8, MI-23	GC376: 0.19	GC376: 0.92	GC376: >100 (Vero6)	GC376: 1.1 (mouse), 3.9 (rat)	GC376: 46700 (mouse), 12560 (rat)	GC376: 33.1 (mouse), 20.1 (rat)	[43] [46]
Ketoamides	23, UAWJ246	23: 0.67 UAWJ246: 0.045	23: 4–5 UAWJ246: 4.61	UAWJ246: >250 (Vero6)	23: 1.0 (mouse)	23: 334 (mouse)	23: 566 (mouse)	[7] [49]
Nitriles	nirmatrelvir	0.019	0.075	>100 (Vero6)	1.5 (rat, P.O.)	1290 (mouse, P.O.)	-	[68]

Table 2. Reversible inhibitors of SARS-CoV-2 M^pro with inhibitory and pharmacological data.

Warhead	Drug Example	Efficacy	Potency	Toxicology	Pharmacology			Lit.
		IC50 (µM)	EC50 (µM)	CC50 (µM)	t0.5 (h)	Cmax (ng/mL)	Clear. (mL/min/kg)
Aldehydes	13–21, GC376, MPI8, MI-23	GC376: 0.19	GC376: 0.92	GC376: >100 (Vero6)	GC376: 1.1 (mouse), 3.9 (rat)	GC376: 46700 (mouse), 12560 (rat)	GC376: 33.1 (mouse), 20.1 (rat)	[43] [46]
Ketoamides	23, UAWJ246	23: 0.67 UAWJ246: 0.045	23: 4–5 UAWJ246: 4.61	UAWJ246: >250 (Vero6)	23: 1.0 (mouse)	23: 334 (mouse)	23: 566 (mouse)	[7] [49]
Nitriles	nirmatrelvir	0.019	0.075	>100 (Vero6)	1.5 (rat, P.O.)	1290 (mouse, P.O.)	-	[68]

An odd case of covalent inhibition of SARS-CoV-2 M^pro concerns the repurposing studies of the anti-inflammatory agent ebselen, which has been in the inhibitor race since early 2020. Enzymatic testing indicated strong inhibition and antiviral activity of ebselen (IC₅₀ = 0.67 µM, EC₅₀ = 4.67 µM, VeroE6) [14,79]. Two inhibition mechanisms are discussed in the literature: covalent attachment to and ultimate selenylation of Cys145 [79] and allosteric inhibition at a dimerization site (discussed later). Covalent inhibition proceeds through S_N-type opening of the benzisoselenazol ring and subsequent hydrolysis, releasing a phenolic byproduct and the selenylated cysteine sulfur, as suggested by DFT and mass spectrometric analysis (Figure 13) [80,81]. The final sulfur selenium adduct was confirmed by X-ray crystallography (PDB 7bak) [79]. Considering the possibility for plural inhibition within M^pro, ebselen analogs were designed and tested. One design avenue considered ebselen derivatives with a modified nitrogen substituent [79], where again it was found that halide substitution beneficially influences inhibitor properties. The most active compound from this series, 27, showed similar potency to that of the parent ebselen and had about three times improved antiviral activity (EC₅₀ = 1.78 µM; VeroE6). Substitution of the selenium atom for sulfur gives the isosteric compound ebsulfur, which also acts as a potent M^pro inhibitor. Among a series of analogs, compound 28 bound with an IC₅₀ of 0.11 µM to the target; however, antiviral and pharmacokinetic data have not been reported yet.

2.3. Noncovalent Active Site Inhibitors

A virtual screening study performed by Yang et al. identified Z1759961256 (Figure 14) as one of six hits with micromolar M^pro potency and viral activity against SARS-CoV-2 [82]. Similarly, in the optimization of the molecular library probe ML300 in a comprehensive study design, CCF981 evolved as lead compound with a strong M^pro inhibition capability (IC₅₀ = 0.068 µM) and a pronounced antiviral effect (EC₅₀ = 0.56 µM; VeroE6) [83]. The drug binds well into M^pro’s active site cavity with the imidazole system pointing towards S1′, the benzotriazole unit in S1, and the m-chlorobenzene occupying S2, as abstracted from the binding modes of structurally similar analog crystal structures. Following the dual inhibitor approach to maximize the antiviral effect, Elseginy et al. synthesized inhibitor 29 (Figure 14) after mapping a >50,000 compound library for structural hits [84]. In vitro tests identified 29 not only as a potent M^pro binder (IC₅₀ = 0.11 µM) but similarly as a good inhibitor of PL^pro and human furin protease. Cell assays indicated cytotoxicity in mammalian cells that were overcome by synthesis of analogs of 29. Inspired by the weak anti-SARS-CoV-2 M^pro activity of the antiepileptic perampanel, Zhang et al. synthesized and tested 27 structural analogs in their search for better inhibitory activity [85]. While most synthesized compounds showed sub-micromolar M^pro inhibition, VeroE6 cell assays indicated considerate toxicity. One compound (30), however, stood out, having a low M^pro IC₅₀ of 0.17 µM, potent antiviral effect (EC₅₀ = 0.98 µM), and displaying no cell toxicity within the tested concentration range (CC₅₀ > 100 µM). The crystal structure of a related analog shows the cloverleaf-like molecular structure occupying three subpockets of the protease (PDB 7l11). Using an in-house structural library, Unoh et al. recently discovered another cloverleaf-like structure that shows beneficial anti-SARS-CoV-2 properties and binds noncovalently in the active site of M^pro [86]. Virtual screening with predefined filters that include two hydrogen bond acceptors in S1 and a hydrophobic element in S2 narrowed down the selection of hit structures suitable for biological screening. Among them, the compound S-217622 displayed very strong binding to M^pro (IC₅₀ = 0.013 µM) and pronounced antiviral effect in the plaque reduction assay (EC₅₀= 0.37 µM). Further biochemical assays confirmed that S217622’s antiviral activity extends to common SARS-CoV-2 variants, including the omicron strain (EC₅₀= 0.29–0.5 µM), while other proteases, including cathepsins and thrombin, remain uninhibited. In the X-ray structure, the 1,3,5-triazine-2,4(1H,3H)-dione core sits in the middle of the catalytical site, with the three wing-like substituents reaching in S1′, S1, and S2 pockets (PDB 7vu6). A tight fit is achieved by a dense H-bond network involving Gly146, Ser147, His165, and Glu168 of the enzyme as well as π–π interactions between His43 and the aromatic P2 inhibitor moiety (Figure 14). The superior in vitro performance of S-217622 stimulated in vivo testing in mice infected with SARS-CoV-2 Gamma strain, where viral titers in lung homogenates were determined 12 h after oral dose of S-217622 and 24 h after infection. The drug demonstrated a peak plasma concentration that at all tested doses (2–32 mg/kg single dose, P.O.) was above the protein-adjusted EC₅₀, thereby reducing the viral titer at the highest two doses to the detection limit (1.8 log₁₀[TCID₅₀]/mL). Full pharmacokinetic profiling in hamster revealed a stable concentration level (C_max = 62 µM at 30 mg/kg single dose P.O., t_0.5 = 1.2 h) and reduced lung damage from acute SARS-CoV-2 infection in hamster recipients [87].

3. Allosteric Inhibitors

Aside from the Cys145-hosting main active site, several surface-exposed pockets of M^pro support catalytic activity assisting crucial enzyme dimerization or substrate processing. Experimentally, to date, six allosteric sites have been suggested, three of which are located at the dimerization interface and three more at distal surface regions (Figure 15). Since dimerization is integral for the activity of the main protease, allosteric inhibitors targeting dimerization close sites seem especially valuable. While there are numerous studies reporting on in silico-generated noncompetitive inhibitors (and the interested reader is referred to those reviews for details, for example [88]), we focus solely on the few experimentally confirmed allosteric M^pro inhibitors that have been biologically validated and ideally crystallographically characterized. There are currently 35 PDB entries concerning noncompetitive inhibitors, possessing vastly different structural features and lipophilicity. The average size of cocrystallized allosteric inhibitors is about 200 Da less than that of covalently bound competitive inhibitors (M_w,aver = 236 ± 75 g/mol).

A comprehensive drug screening effort of 5000 FDA- or near-FDA-approved drugs identified 37 M^pro inhibitors [89]. In cell-based assays, seven compounds had antiviral activity without toxicity. From their X-ray structures with M^pro, one drug showed noncovalent binding within the active site, while four more drugs were located at distal sites. The S. aureus antibacterial MUT056399 (Figure 16) coordinated near the active site and moderately inhibited SARS-CoV-2 viral activity (EC₅₀ = 38.2 µM, CC₅₀ > 100 µM; Table 3) in the Vero E6 cell assay. From the cocrystal structure, it is obvious that the diphenyl ether blocks access to the catalytic dyad, with other parts of the molecule binding to S1 and S2 pocket residues (PDB 7ap6). The flattened structure of the anticancer agent pelitinib covers a hydrophobic pocked within M^pro’s C-terminal dimerization domain, thereby interfering with activity-determining enzyme dimerization. It displays a potent antiviral effect (EC₅₀ = 1.25 µM) and unfortunately, moderate VeroE6 cell toxicity (CC₅₀ =14 µM). While the 3-cyanoquinoline moiety interacts with the C-terminal helix around Ser301, the ethyl ether forms bridging interactions with a loop from the opposing protomer (PDB 7axm). Similarly, the vasodilator ifenprodil and the investigational chemokine RS-102895 (Figure 16) also bind to this allosteric site, resulting in moderate antiviral activity (EC₅₀ ≥ 20 µM) for both drugs (Table 3). Ifenprodil is undergoing clinical phase III studies to test its efficiency in acute intervention for hospitalized COVID-19 patients [90]. The same study identified a deep groove within the dimerization domain as a second allosteric site, which can be occupied by the investigational anticancer agent AT7519 (Figure 16) [89]. Binding to the grove is supported by van-der-Waals interactions of the chlorinated benzene ring with a hydrophobic residue wall as well as hydrogen-bonding with Gln110 and Arg298 (PDB 7aga). Mutational analysis demonstrated that Arg298 is crucial for dimerization, explaining AT7519’s moderate antiviral effect (EC₅₀ = 25.2 µM; CC₅₀ > 100 µM). The anthelminthic drug niclosamide was identified as a wonder antiviral for various viruses, including SARS, MERS, ZIKV, HCV, and human adenovirus [91]. In a repurposing attempt, Li and coworkers tested a library of 300 niclosamide derivatives for their anti-SARS-CoV-2 properties [92]. While several analogs were identified with low micromolar IC₅₀ values and good antiviral effect but unfortunately also considerable cell toxicity, one compound, JMX0286 (Figure 16), stands out due to an increased selectivity index (IC₅₀ = 4.8 µM, EC₅₀ = 2.3 µM, CC₅₀ = 53 µM, A549-hACE2 cells). Kinetic experiments showed a lowered maximum velocity but similar Km for inhibitor-treated samples, suggesting a noncompetitive inhibition mechanism. Molecular modeling points to a similar binding mode and allosteric site as that observed for AT7519. A recent clinical phase II trial showed no significant effect of niclosamide on decreasing the contagious period of SARS-CoV-2 infection [93]. Considering the structural similarity of M^pro’s binding site and its direct vicinity with that of thrombin (FactorXa), Chaves et al. were curious about the inhibitory effect of anticoagulants apixaban (Figure 16) and rivaroxaban [94]. In the in vitro assay, apixaban (IC₅₀ = 0.01 µM) was superior to rivaroxaban (IC₅₀ = 0.14 µM) and 21-fold more active than GC-376. Kinetic analysis suggests noncompetitive binding that seems to occur at the same allosteric site as that binding pelitinib (molecular docking). Apixaban inhibited SARS-CoV-2 replication in infected Calu-3 cells, with an EC₅₀ of 1.84 µM, 60-times less potent than remdesivir, and showed in this assay a CC₅₀ of 491 µM. Despite a low free plasma concentration (C_max = 0.55 mM) due to high binding affinity for albumin, apixabans free fraction was ~10-times higher than its SARS-CoV-2 M^pro affinity, serving as a good basis for further anti-coronaviral pharmacological studies. A structure-based virtual screening of 3.8 million small-molecule ligands recently identified three hits that showed promising complexation with the M^pro enzyme [95]. Among them, GR20 (Figure 16) showed with IC₅₀ = 91.8 µM the best M^pro inhibition. Although competitive inhibition cannot be ruled out by the authors, kinetic analysis suggested a mixed inhibition model. Molecular docking gives the best fitting of GR20 into allosteric groove 2 (Figure 15), similar to that of the anticancer agent AT7519. From the natural pool of antivirals, the broad-spectrum drugs chebulagic acid and punicalagin have been demonstrated to show potent activity against M^pro [96]. Both drugs displayed IC₅₀ and EC₅₀ values in the lower micromolar range, with enzyme kinetic data indicating noncompetitive binding. From initial docking, it is suggested that these structurally extensive plant-based antivirals fit into a groove between domains II and III distal from the active site.

As an alternative anti-COVID-19 drug design, Xia and co-workers originally envisioned thiophilic metal complexation of the active site Cys145 using colloidal bismuth subcitrate (CBS) to inhibit M^pro’s activity [97]. Indeed, incubation of SARS-CoV-2 M^pro with CBS resulted in excellent inhibition (IC₅₀ = 0.93 µM), but unexpectedly, k_cat/Km remained unchanged with both Km and v_max decreasing, indicating noncompetitive binding. Inductively coupled plasma mass spectrometry revealed an inhibitor:enzyme monomer ratio of 2. By mutational analysis, one binding site was identified as the C-terminal Cys300 residue, which is essential to stabilize the dimeric structure of M^pro [98]. The VeroE6 antiviral assay yielded a half maximal effective concentration of CBS against SARS-CoV-2 replication of 177.3 µM, sparking excitement for further drug optimization endeavors.

Based on the extended conformation of the native M^pro monomer, Geng and co-workers recently designed inhibitory nanobodies using the Nanoluc Binary Technology (nanoBiT) [99]. Application of a phage display library obtained 11 high-affinity nano-bodies from M^pro-immunized camel. An in vitro M^pro FRET assay indicated two antibodies as sub-micromolar M^pro binders (NB1A2 IC₅₀ = 0.19 µM, NB2B4 IC₅₀ = 0.12 µM; Table 3). Size-exclusion chromatography showed that binding to M^pro leads to dissociation to catalytically inactive monomers. Moreover, X-ray crystallographic analysis showed different epitopes for the two antibodies. In the cocrystal structure with NB2B4, the antibody is located at the α-helical domain of M^pro, distal from the catalytic site (PDB 7vfb). The other antibody binds instead to an epitope that overlaps with the substrate binding site, thus suggesting NB1A2 as not only an allosteric but also a competitive inhibitor (PDB 7vfa).

Table 3. In Vitro inhibition data of allosteric M^pro inhibitors and their PDB codes.

Inhibitor	Non-Competitive Binding in: a	IC50 (µM)	EC50 (µM) b	CC50 (µM) b	PDB Code	Ref.
MUT056399	active site	-	38.2	>100	7ap6	[89]
pelitinib	allosteric site 1	-	1.25	14	7axm	[89]
ifenprodil	allosteric site 1	-	46.9	>100	7aqi	[89]
RS-102895	allosteric site 1	-	19.8	55	7abu	[89]
AT7519	allosteric site 2	-	25.2	>100	7aga	[89]
JMX0286	allosteric site 2 *	4.8	2.3 (A549-hACE2)	53(A549-hACE2)	-	[92]
apixaban	allosteric site 1 *	0.01	1.84 (Calu-3)	491(Calu-3)	-	[94]
GR20	allosteric site 2 *	91.8	-	-	-	[95]
chebulagic acid	alternative site	9.76	9.09	~100	-	[96]
punicalagin	alternative site	7.2	4.62	~100		[96]
colloidal bismuth-subcitrate (CBS)	alternative site	0.93	177.3	-	-	[97]
NB1A2	alternative site	0.19	-	-	7vfa	[99]
NB2B4	alternative site	0.12	-	-	7vfb	[99]

^a An asterisk indicates predicted binding modes according to molecular docking. ^b Determined in VeroE6 cell assay unless stated otherwise.

Table 3. In Vitro inhibition data of allosteric M^pro inhibitors and their PDB codes.

Inhibitor	Non-Competitive Binding in: a	IC50 (µM)	EC50 (µM) b	CC50 (µM) b	PDB Code	Ref.
MUT056399	active site	-	38.2	>100	7ap6	[89]
pelitinib	allosteric site 1	-	1.25	14	7axm	[89]
ifenprodil	allosteric site 1	-	46.9	>100	7aqi	[89]
RS-102895	allosteric site 1	-	19.8	55	7abu	[89]
AT7519	allosteric site 2	-	25.2	>100	7aga	[89]
JMX0286	allosteric site 2 *	4.8	2.3 (A549-hACE2)	53(A549-hACE2)	-	[92]
apixaban	allosteric site 1 *	0.01	1.84 (Calu-3)	491(Calu-3)	-	[94]
GR20	allosteric site 2 *	91.8	-	-	-	[95]
chebulagic acid	alternative site	9.76	9.09	~100	-	[96]
punicalagin	alternative site	7.2	4.62	~100		[96]
colloidal bismuth-subcitrate (CBS)	alternative site	0.93	177.3	-	-	[97]
NB1A2	alternative site	0.19	-	-	7vfa	[99]
NB2B4	alternative site	0.12	-	-	7vfb	[99]

^a An asterisk indicates predicted binding modes according to molecular docking. ^b Determined in VeroE6 cell assay unless stated otherwise.

4. Inhibitors in Action—Drug Performance in the Clinic

Several herein described SARS-CoV-2 antivirals that target its main protease, M^pro, have entered clinical studies and/or trials with varying success [100,101]. As mentioned before, reversible covalent inhibitors provide more beneficial pharmacokinetic data to qualify for clinical testing (Figure 17). Among the trial candidates, nirmatrelvir established the golden ticket to become the first covalent SARS-CoV-2 M^pro inhibitor, approved in combination with ritonavir for the treatment of mild-to-moderate COVID-19 in adults and pediatric patients 12 years or older. The US has committed to purchasing 20 million treatment courses, with close to 10 million treatment courses currently prescribed [102]. The first meta-analysis studies indicate effectivity of the marketed drug paxlovid in reducing mortality and hospitalization rates in patients with COVID-19, with benefits especially for vulnerable cohorts including older patients, immunosuppressed patients, and patients with underlying neurological or cardiovascular disease [103,104]. Nonetheless, one must be aware of the limitations of a single target treatment regime, most importantly, the occurrence of viral resistance by mutational escape (“variants of concern”), as has been suggested in recent literature [105,106]. Recently, a higher reinfection rate after completing an antiviral treatment course compared to a control group has been observed in a prospective cohort study (preprint) [107]. Interestingly, this seems not to be attributable to virus mutation or impaired immunity but rather to a suboptimal drug exposure for certain patients [108], leaving room for further drug design and formulation development. Potential viral resistance has not only been recognized for paxlovid’s M^pro-active ingredient nirmatrelvir but also for other late-stage M^pro antivirals such as GC376 and boceprevir. In fact, a study by Peng and co-workers investigated the development of resistance in a feline coronavirus (FIPV) under pressure from GC376 and found that a mutation in nsp12 of its main protease rendered the enzyme up to three times less susceptible for the inhibitor [109]. Encouraging data from cell culture studies hint that GC376 performs efficiently against M^pro’s from SARS-CoV-2 and various seasonal coronaviruses (NL63, 229E, and OC43), suggesting beneficial inhibition promiscuity of this drug that can be used as a basis for future structure development and antiviral testing [110]. This is in contrast to the molecular docking analysis that suggested comparable affinity of boceprevir and GC376 for the different analyzed coronaviral M^pro enzymes, leading the authors to caution against overinterpretation of in silico data during antiviral therapy development.

5. Future Strategies/Outlook

The threat of a potentially uncontrollable pandemic has mobilized unprecedented scientific efforts to identify druggable targets of SARS-CoV-2. Influenced by successful past (corona)viral research, the main protease of the current coronavirus, M^pro, was identified as one such target due to its critical role in virus replication and the expected low mutational susceptibility. Over the past three years, various research groups and companies around the globe entered the quest for potent SARS-CoV-2 M^pro antivirals, revealing at least two dozen lead structures with predominantly reversible warheads. About half a dozen peptidomimetic and small-molecule inhibitors have since entered clinical trials [101], with nirmatrelvir (Pfizer) receiving FDA emergency approval after successful pass of clinical trials and Shionogi’s nonpeptidic S-217622 advancing to late-stage Phase III evaluation [86,111]. Central to the rapid discovery of vastly different inhibitor scaffolds was the accessibility of research data, catalyzed partly by extended open access policies of scientific journals and global molecular design initiatives such as COVID moonshot [112]. However, considering emerging reports of breakthrough infections after antiviral treatment and the constant threat of viral adaptability, and learning from the persistence of other viral epidemics such as HIV, the scientific community is challenged to be on the lookout for extended and alternative treatment options. One such strategy that has been proven effective, for example, in epidemiologic HIV management, is combination therapy, to explore synergistic drug effects and limit the potential mutational escape of mono-target antiviral therapies. Alternative viral targets with successfully developed inhibitors that can be co-targeted in a combination therapy include the RNA-dependent RNA polymerase (RdRP), the nucleocapsid protein (N), and the spike protein (S), among others [113]. An example of a synergistic combination therapy with high potential to further condemn viral replication is pairing the paxlovid drug nirmatrelvir with the RNA polymerase-attacking molnupiravir [114,115].

Another approach of enhanced antiviral combat considers alternative warhead chemistries to increase target specificity as well as drug bioavailability and clearance after successful therapy. An example that explores the chemical space around the reactive electrophilic warhead shows promising potential to finetune inhibitor properties to desired parameters solely by exchange of the reactive head group [116]. Aside from the already employed diversity of functional groups, novel chemistries that have the potential to improve target interaction can be derived from wider drug repurposing screens, including, for example, boronic acids, such as in ixazomib [117]. The example of colloidal bismuth subcitrate as a strong M^pro inhibitor highlights the potential of thiophilic metal complexes in anti-M^pro pharmacophore design [97]. Especially in the light of minimizing off-target effects, the excellent in vitro and in vivo performance of selected noncovalent leads such as S-217622 should be considered [86]. Lastly, available new drug delivery technologies can help to reduce dosage and unwanted side effects, potentially allowing the consideration of drug candidates with a smaller selectivity index that have previously been excluded [118]. Target overlap in vulnerable COVID-19 patients that are treated for pre-existing conditions, including cardiovascular and viral diseases, might lead to side-effects upon COVID-19 treatment. This can be desired if such interactions are synergistic, but drug exposure should be limited if contraindications have been identified. For such cases, new formulation technologies are anticipated to reduce the amount of active drug needed, minimizing off-target effects and avoiding drug accumulation. With these tools in development, new or improved SARS-CoV-2 M^pro inhibitors remain an exciting and beneficial avenue for the successful treatment and eradication of COVID-19 and its complications.