Validation and invalidation of SARS-CoV-2 main protease inhibitors using the Flip-GFP and Protease-Glo luciferase assays

Abstract

SARS-CoV-2 main protease (M^pro) is one of the most extensively exploited drug targets for COVID-19. Structurally disparate compounds have been reported as M^pro inhibitors, raising the question of their target specificity. To elucidate the target specificity and the cellular target engagement of the claimed M^pro inhibitors, we systematically characterize their mechanism of action using the cell-free FRET assay, the thermal shift-binding assay, the cell lysate Protease-Glo luciferase assay, and the cell-based FlipGFP assay. Collectively, our results have shown that majority of the M^pro inhibitors identified from drug repurposing including ebselen, carmofur, disulfiram, and shikonin are promiscuous cysteine inhibitors that are not specific to M^pro, while chloroquine, oxytetracycline, montelukast, candesartan, and dipyridamole do not inhibit M^pro in any of the assays tested. Overall, our study highlights the need of stringent hit validation at the early stage of drug discovery.

Graphical abstract

FlipGFP and protease-Glo luciferase assays, coupled with the FRET and thermal shift binding assays, were applied to validate/invalidate the reported SARS-CoV-2 M^pro inhibitors.

1. Introduction

SARS-CoV-2 is the causative agent for COVID-19, which infected 229 million people and led to 4.71 million deaths as of September 23, 2021. SARS-CoV-2 is the third coronavirus that causes epidemics and pandemics in human. SARS-CoV-2, along with SARS-CoV and MERS-CoV, belong to the β genera of the coronaviridae family¹. SARS-CoV-2 encodes two viral cysteine proteases, the main protease (M^pro) and the papain-like protease (PL^pro), that mediate the cleavage of viral polyproteins pp1a and pp1ab during viral replication2, 3, 4. M^pro cleaves at more than 11 sites at the viral polyproteins and has a high substrate preference for glutamine at the P1 site⁵. In addition, the M^pro is highly conserved among coronaviruses that infect human including SARS-CoV-2, SARS-CoV, MERS-CoV, HCoV-OC43, HCoV-NL63, HCoV-229E, and HCoV-HKU1. For these reasons, M^pro becomes a high-profile drug target for the development of broad-spectrum antivirals. Structurally disparate compounds including FDA-approved drugs and bioactive compounds have been reported as M^pro inhibitors6, 7, 8, 9, 10, several of which also have in vivo antiviral efficacy against SARS-CoV-211, 12, 13, 14, 15.

FRET assay is the gold standard assay for protease and is typically used as a primary assay for the screening of M^pro inhibitors. However, the FRET assay conditions used by different groups vary significantly in terms of the protein and substrate concentrations, pH, reducing reagent, and detergent. Reducing reagent is typically added in the assay buffer to prevent the non-specific oxidation or alkylation of the catalytic C145 in M^pro. Nonetheless, many studies do not include reducing reagents in the FRET assay buffer, leading to debatable results¹⁶. Regardless of the assay condition, FRET assay is a cell free biochemical assay, which does not mimic the cellular environment; therefore, the results cannot be used to accurately predict the cellular activity of the M^pro inhibitor or the antiviral activity. Moreover, one limiting factor for M^pro inhibitor development is that the cellular activity has to be tested against infectious SARS-CoV-2 in BSL-3 facility, which is inaccessible to many researchers. For these reasons, there is a pressing need of secondary M^pro target-specific assays that can closely mimic the cellular environment and be used to rule out false positives.

In this study, we report our findings of validating or invalidating the literature reported M^pro inhibitors using the cell lysate Protease-Glo luciferase assay and the cell-based FlipGFP assay, in conjunction to the cell-free FRET assay and thermal shift-binding assay. The purpose is to elucidate their target specificity and cellular target engagement. The Protease-Glo luciferase assay was developed in this study, and the FlipGFP assay was recently developed by us and others17, 18, 19, 20. Our results have collectively shown that majority of the M^pro inhibitors identified from drug repurposing screening including ebselen, carmofur, disulfiram, and shikonin are promiscuous cysteine inhibitors that are not specific to M^pro, while chloroquine, oxytetracycline, montelukast, candesartan, and dipyridamole do not inhibit M^pro in any of the assays tested. The results presented herein highlight the pressing need of stringent hit validation at the early stage of drug discovery to minimize the catastrophic failure in the following translational development.

2. Methods and materials

2.1. Protein expression and purification

The tag-free SARS CoV-2 M^pro protein with native N- and C- termini was expressed in pSUMO construct as described previously³.

2.2. Enzymatic assays

The FRET-based protease was performed as described previously². Briefly, 100 nmol/L of M^pro protein in the reaction buffer containing 20 mmol/L HEPES, pH 6.5, 120 mmol/L NaCl, 0.4 mmol/L EDTA, 4 mmol/L DTT, and 20% glycerol was incubated with serial concentrations of the testing compounds at 30 °C for 30 min. The proteolytic reactions were initiated by adding 10 μmol/L of FRET-peptide substrate [Dabcyl-KTSAVLQ/SGFRKME (Edans)] and recorded in Cytation 5 imaging reader (Thermo Fisher Scientific) with 360/460 filter cube for 1 h. The proteolytic reaction initial velocity in the presence or absence of testing compounds was determined by linear regression using the data points from the first 15 min of the kinetic progress curves. IC₅₀ values were calculated by a 4-parameter dose−response function in Prism 8.

2.3. Thermal shift assay (TSA)

Direct binding of testing compounds to SARS CoV-2 M^pro protein was evaluated by differential scanning fluorimetry (DSF) using a Thermal Fisher QuantStudio 5 Real-Time PCR System as previously described². Briefly, SARS CoV-2 M^pro protein was diluted into reaction buffer to a final concentration of 3 μmol/L and incubated with 40 μmol/L of testing compounds at 30 °C for 30 min. DMSO was included as a reference. SYPRO orange (1 ×, Thermal Fisher, catalog No. S6650) was added, and the fluorescence signal was recorded under a temperature gradient ranging from 20 to 95 °C with incremental step of 0.05 °C/s. The melting temperature (T_m) was calculated as the mid log of the transition phase from the native to the denatured protein using a Boltzmann model in Protein Thermal Shift Software v1.3. ΔT_m was the difference between T_m in the presence of testing compounds and T_m in the presence of DMSO.

2.4. FlipGFP M^pro assay

The construction of FlipGFP-M^pro plasmid was described previously¹⁷. The assay was carried out as follows: 293T cells were seeded in 96-well black, clear bottomed Greiner plate (catalog No. 655090) and incubated overnight to reach 70%–90% confluency. 50 ng of FlipGFP-M^pro plasmid and 50 ng SARS CoV-2 M^pro expression plasmid pcDNA3.1 SARSCoV-2 M^pro were transfected into each well with transfection reagent TransIT-293 (Mirus catalog No. MIR 2700) according to the manufacturer's protocol. Three hours after transfection, 1 μL of testing compound was directly added to each well without medium change. Two days after transfection, images were taken with Cytation 5 imaging reader (Biotek) using GFP and mCherry channels via 10 × objective lens and were analyzed with Gen5 3.10 software (Biotek). The mCherry signal alone in the presence of testing compounds was utilized to evaluate the compound cytotoxicity.

2.5. Protease-Glo luciferase assay

pGlosensor-30F DEVD vector was obtained from Promega (Catlog No. CS182101). pGloSensor-30F M^pro plasmid was generated by replacing the original caspase cutting sequence (DEVDG) was with SARS CoV-2 M^pro cutting sequence (AVLQ/SGFR) from BamHI/HindIII sites. The DNA duplex containing M^pro cutting sequence was generated by annealing two 5′-phosphoriated primers: forward: GATCCGCCGTGCGCAGAGCGGCTTTCAGA; and reverse: AGCTTCTGAAGCCGCTCTGCAGCACGGCG. Protease-Glo luciferase assay was carried out as follows: 293T cells in 10 cm culture dish were transfected with pGlosensor-30F M^pro plasmid in the presence of transfection reagent TransIT-293 (Mirus catalog No. MIR 2700) according to the manufacturer's protocol. 24 h after transfection, cells were washed with PBS once, then each dish of cells was lysed with 5 mL of PBS+ 1% Trition-X100; cell debris was removed by centrifuge at 2000 × g for 10 min. Cell lysates was freshly frozen to −80 °C until ready to use. During the assay, 20 μL cell lysate was added to each well in 96-well flat bottom white plate (Fisherbrand Catalog No. 12566619), then 1 μL of testing compound or DMSO was added to each well and mixed at room temperature for 5 min. 5 μL of 200 nmol/L Escherichia coli expressed SARS CoV-2 M^pro protein was added to each well to initiate the proteolytic reaction (the final M^pro protein concentration is around 40 nmol/L). The reaction mix was further incubated at 30 °C for 30 min. The firefly and Renilla luciferase activity were determined with Dual-Glo Luciferase Assay according to manufacturer's protocol (Promega Catalog No. E2920). The efficacy of testing compounds against M^pro was evaluated by plotting the ratio of firefly luminescence signal over the Renilla luminescence signal versus the testing compound concentrations with a 4-parameter dose−response function in Prism 8.

2.6. Antiviral assay in Calu-3 cells

The antiviral assay was performed as previously described²¹. Calu-3 cells (ATCC, HTB-55) were plated in 384 well plates and grown in minimal Eagle's medium supplemented with 1% non-essential amino acids, 1% penicillin/streptomycin, and 10% FBS. The next day, 50 nL of compound in DMSO was added as an 8-pt dose response with three-fold dilutions between testing concentrations in triplicate, starting at 40 μmol/L final concentration. The negative control (DMSO, n = 32) and positive control (10 μmol/L remdesivir, n = 32) were included on each assay plate. Calu-3 cells were pretreated with controls and testing compounds (in triplicate) for 2 h prior to infection. In BSL-3 containment, SARS-CoV-2 (isolate USA-WA1/2020) diluted in serum free growth medium was added to plates to achieve an MOI of 0.5. Cells were incubated with compounds and SARS-CoV-2 virus for 48 h. Cells were fixed and then immunostained with anti-dsRNA (J2) and nuclei were counterstained with Hoechst 33342 for automated microscopy. Automated image analysis quantifies the number of cells per well (toxicity) and the percentage of infected cells (dsRNA + cells/cell number) per well. SARS-CoV-2 infection at each drug concentration was normalized to aggregated DMSO plate control wells and expressed as percentage-of-control:

POC = % Infection_sample/Avg% Infection_{DMSO cont} (1)

A non-linear regression curve fit analysis (GraphPad Prism 8) of POC Infection and cell viability versus the log₁₀ transformed concentration values to calculate EC₅₀ values for Infection and CC₅₀ values for cell viability. Selectivity index (SI) was calculated as a ratio of drug's CC₅₀ and EC₅₀ values:

SI = CC₅₀/IC₅₀ (2)

3. Results and discussion

3.1. Assay validation using GC-376 and rupintrivir as positive and negative controls

The advantages and disadvantages of the cell lysate Protease-Glo luciferase assay and the cell-based FlipGFP assay compared to the cell free FRET assay are listed in Table 1. To minimize the bias from a particular assay, we apply all these three functional assays together with the thermal shift-binding assay for the hit validation.

Table 1.

Advantages and disadvantages of the three functional assays used in this study.

Assay	Advantage	Disadvantage
FRET assay	•High throughput	•Compounds that quench the fluorophore will show up as false positives •Assay interference from fluorescent compounds, detergents, and aggregators. •Cannot be used to predict the cellular antiviral activity •No standard condition among scientific community
FlipGFP assay	•Can rule out compounds that are cytotoxic, membrane impermeable, or substrates of drug efflux pump •A close mimetic of virus-infected cell •Can be used to predict the cellular antiviral activity •Reveals cellular target engagement	•The assay takes 48 h, thus it cannot be used for cytotoxic compounds •Interference from fluorescent compounds •Not high throughput
Protease-Glo luciferase assay	•High throughput •Reveals cellular target engagement •Can be used to test cytotoxic compounds	•Cannot be used to predict the cellular antiviral activity

In the cell-based FlipGFP assay, the cells were transfected with two plasmids, one expresses the SARS-CoV-2 M^pro, and another expresses the GFP reporter²². The GFP reporter plasmid expresses three proteins including the GFP β10–β11 fragment flanked by the K5/E5 coiled coil, the GFP β1–9 template, and the mCherry (Fig. 1A). mCherry serves as an internal control for the normalization of the expression level or the quantification of compound toxicity. In the assay design, β10 and β11 were conformationally constrained in the parallel position by the heterodimerizing K5/E5 coiled coil with a M^pro cleavage sequence (AVLQ↓SGFR). Upon cleavage of the linker by M^pro, β10 and β11 become antiparallel and can associate with the β1–9 template, resulting in the restoration of the GFP signal. In principle, the ratio of GFP/mCherry fluorescence is proportional to the enzymatic activity of M^pro. The FlipGFP M^pro assay has been used by several groups to characterize the cellular activity of M^pro inhibitors^17,19,20.

Figure 1.

Principles for the FlipGFP and Protease-Glo luciferase assays and assay validation with control compounds. (A) Assay principle for the FlipGFP assay. Diagram of the FlipGFP M^pro reporter plasmid is shown. (B) Assay principle for the Protease-Glo luciferase assay. Diagram of pGlo-M^pro luciferase reporter in the pGloSensor-30F vector is shown. (C) Representative images from the FlipGFP-M^pro assay. Dose-dependent decrease of GFP signal was observed with the increasing concentration of GC-376 (positive control); almost no GFP signal change was observed with the increasing concentration of rupintrivir (negative control). The scale bar is 300 μm. (D) and (E) Dose−response curve of the ratio of GFP/mCherry fluorescence with GC-376 and rupintrivir; mCherry signal alone was used to normalize protein expression level or calculate compound cytotoxicity. (F)–(G) Protease-Glo luciferase assay results of GC-376 and rupintrivir. Left column showed Firefly and Renilla luminescence signals in the presences of increasing concentrations of GC-376 and rupintrivir; Right column showed dose−response curve plots of the ratio of FFluc/Rluc luminescence. The results are mean ± standard deviation from three repeats. Compound concentration, μmol/L.

In the cell lysate Protease-Glo luciferase assay, the cells were transfected with pGloSensor-30F luciferase reporter (Fig. 1B)²³. The pGloSensor-30F luciferase reporter plasmid expresses two proteins, the inactive, circularly permuted firefly luciferase (FFluc) and the active Renilla luciferase (Rluc). Renilla luciferase was included as an internal control to normalize the protein expression level. The firefly luciferase was split into two fragments, the FF 4–354 and FF 358–544. The SARS-CoV-2 M^pro substrate cleavage sequence (AVLQ/SGFR) was inserted in between the two fragments. Before protease cleavage, the pGloSensor-30F reporter comprises an inactive circularly permuted firefly luciferase. The cells were lysed at 24 h post transfection, and M^pro and the luciferase substrates were added to initiate the reaction. Upon protease cleavage, a conformational change in firefly luciferase leads to drastically increase in luminescence. In principle, the ratio of FFluc/Rluc luminescence is proportional to the enzymatic activity of M^pro.

To calibrate the FlipGFP and Protease-Glo luciferase assays, we chose GC-376 and rupintrivir as positive and negative controls, respectively. The IC₅₀ values for GC-376 in the FlipGFP and Protease-Glo luciferase assays were 2.75 and 0.023 μmol/L, respectively (Fig. 1C, D, and F). The IC₅₀ value in the FlipGFP assay is similar to its antiviral activity (Table 2), suggesting the FlipGFP can be used to predict the cellular antiviral activity. In contrast, rupintrivir showed no activity in either the FlipGFP (IC₅₀ > 50 μmol/L) (Fig. 1C second row and 1E) or the Protease-Glo luciferase assay (IC₅₀ > 100 μmol/L) (Fig. 1G), which agrees with the lack of inhibition from the FRET assay (IC₅₀ > 20 μmol/L). Nonetheless, rupintrivir was reported to inhibit SARS-CoV-2 replication with an EC₅₀ of 1.87 μmol/L using the nanoluciferase SARS-CoV-2 reporter virus (SARS-CoV-2-Nluc) in A549-hACE2 cells²⁴ (Table 2). The discrepancy indicates that the mechanism of action of rupintrivir might be independent of M^pro inhibition. Overall, the FlipGFP and Protease-Glo luciferase assays are validated as target-specific assays for SARS-CoV-2 M^pro.

Table 2.

Summary of results.

Compd.	FRET IC50 (μmol/L)	TSA ΔTm (°C)	FlipGFP IC50 (μmol/L)		pGlo-Mpro luciferase (μmol/L)	Anti-viral (μmol/L) Vero CPE	PDB code	Comment
Control compounds
	0.030 ± 0.008 0.15 ± 0.0325 0.052 ± 0.00726	18.302		2.75 ± 1.06	0.023 ± 0.002	3.37 ± 1.682 0.7025 10 ± 4.226	6WTT2 6WTJ27 7C6U25	Positive control
	>202 >1008	0.01		>50	>240	(Nluc)1.87 ± 0.4724	N.A.	Negative control
HCV protease inhibitors
	4.13 ± 0.612 2.9 ± 0.628 8.0 ± 1.525 3.129 3.7 ± 1.730	6.672	18.33 ± 3.54		4.49 ± 1.42	1.31 ± 0.582 19.628 15.5725 >5030 5.4 (293T)28	6XQU29 7C6S25 7COM31	Validated Mpro inhibitor
	24.2 ± 6.1 18.7 ± 6.428 1829 17.9 ± 4.530	1.03	19.9 ± 3.0		41.91 ± 6.82	>5028 20.5 (293T)28	6XQS29 7C7P31 7LB732	Validated Mpro inhibitor
	5.73 ± 0.672 2.2 ± 0.428 5.129	5.182	23.8 ± 6.5		10.99 ± 1.96	7.72 15 (293T)28	6XQT29 7D1O33	Validated Mpro inhibitor
HIV protease inhibitors
	>602 234 ± 9830	−0.60	>20		>240	(Nluc)9.00 ± 0.4224 19 ± 834 2535	N.A.	Not a Mpro inhibitor
	>202 >100030	−0.65	>20		>240	> 10035	N.A.	Not a Mpro inhibitor
	>6036 7.5 ± 0.337	0.19	>60		>240	2.0 ± 0.1238	N.A.	Not a Mpro inhibitor
	>202 118 ± 1830	−0.60	>10		>240	3.330 (Nluc)0.77 ± 0.3224 3.1 ± 0.0634	N.A.	Not a Mpro inhibitor
	>20 6.7 ± 0.639	−0.65	>20		>240	(Nluc)2.74 ± 0.2024	N.A.	Not a Mpro inhibitor
Calcium channel blocker
	64.2 ± 9.8 4.81 ± 1.8740	0.45	>10		>240	N.A.	N.A.	Not a Mpro inhibitor
Hits from drug repurposing
	0.97 ± 0.272 8.98 ± 2.026	6.652	>60		0.60 ± 0.11	2.07 ± 0.762 27 ± 1.426	6XA43	Validated Mpro Inhibitor Cell-type dependent
	0.45 ± 0.062 6.48 ± 3.426	7.862	38.71 ± 5.66		0.79 ± 0.10	0.49 ± 0.182 1.3 ± 0.5726	6XBH3	Validated Mpro Inhibitor Cell-type dependent
	>6041 0.67 ± 0.0916,42 >10026	0.1441	>60		>60	4.67 ± 0.8016,42 >10026	7BAK42	Not a Mpro inhibitor
	>6041 9.35 ± 0.1816	0.2141	>60		>240	not active16	N.A.	Not a Mpro inhibitor
	28.2 ± 9.541 1.82 ± 0.0616 >10026	0.3541	>60		>240	>10026	7BUY43	Not a Mpro inhibitor
	>6041 21.39 ± 7.0616	−0.1441	>60		>240	not active16	N.A.	Not a Mpro inhibitor
	>6041 1.55 ± 0.3016	−0.2141	>60		>240	not active16	N.A.	Not a Mpro inhibitor
	>6041 15.75 ± 8.2216 15.0 ± 3.026	0.4041	>20		>240	>10026	7CA844	Not a Mpro inhibitor
	>60 0.39 ± 0.1145 0.94 ± 0.2046	0.21	>60		>240	2.92 ± 0.0645 2.94 ± 1.1946	N.A.	Not a Mpro inhibitor
	>20036 3.9 ± 0.237	0.0936	>200		>800	1.1347	N.A.	Not a Mpro inhibitor
	>20036 2.9 ± 0.337	0.1636	>200		>800	2.71 to 7.3648	N.A.	Not a Mpro inhibitor
	>6036 15.2 ± 0.937	0.1636	>60		>240	N.A.	N.A.	Not a Mpro inhibitor
	13.5 ± 1.036 7.3 ± 0.537	−0.6836	>60		>240	N.A.	N.A.	Not a Mpro inhibitor
	>6036 Ki = 0.62 ± 0.0537	0.2336	>60		>240	N.A.	N.A.	Not a Mpro inhibitor
	29.4 ± 3.236 0.60 ± 0.0137	−0.1936	>60		>240	N.A.	N.A.	Not a Mpro inhibitor

N.A., not available.

3.2. HCV protease inhibitors

The HCV protease inhibitors have been proven a rich source of SARS-CoV-2 M^pro inhibitors^2,28,29. From screening a focused protease library using the FRET assay, we discovered simeprevir, boceprevir, and narlaprevir as SARS-CoV-2 M^pro inhibitors with IC₅₀ values of 13.74, 4.13, and 5.73 μmol/L, respectively, while telaprevir was less active (31% inhibition at 20 μmol/L)². The binding of boceprevir to M^pro was characterized by thermal shift assay and native mass spectrometry. Boceprevir inhibited SARS-CoV-2 viral replication in Vero E6 cells with EC₅₀ values of 1.31 and 1.95 μmol/L in the primary CPE and secondary viral yield reduction assays, respectively (Table 2). In parallel, Fu et al.²⁵ also reported boceprevir as a SARS-CoV-2 M^pro inhibitor with an enzymatic inhibition IC₅₀ of 8.0 μmol/L and an antiviral EC₅₀ of 15.57 μmol/L. The X-ray crystal structure of M^pro with boceprevir was solved, revealing a covalent modification of the C145 thiol by the ketoamide (PDBs: 6XQU²⁹, 7C6S²⁵, 7COM³¹).

In the current study, we found that boceprevir showed moderate inhibition in the cellular FlipGFP M^pro assay with an IC₅₀ of 18.33 μmol/L (Fig. 2A and B), a more than 4-fold increase compared to the IC₅₀ in the FRET assay (4.13 μmol/L). The IC₅₀ value of boceprevir in the cell lysate Protease-Glo luciferase assay was 4.49 μmol/L (Fig. 2E). In comparison, telaprevir and narlaprevir showed weaker inhibition than boceprevir in both the FlipGFP and Protease-Glo luciferase assays (Fig. 2A, C, D, F, and G), which is consistent with their weaker potency in the FRET assay (Table 2). Overall, boceprevir, telaprevir, and narlaprevir have been validated as SARS-CoV-2 M^pro inhibitors in both the cellular FlipGFP assay and the cell lysate Protease-Glo luciferase assay. Therefore, the antiviral activity of these three compounds against SARS-CoV-2 are likely due to M^pro inhibition. Although the inhibition of M^pro by boceprevir is relatively weak compared to GC-376, several highly potent M^pro inhibitors were subsequently designed as hybrids of boceprevir and GC-376^11,17,31 including the Pfizer oral drug candidate PF-07321332, which contain the dimethylcyclopropylproline at the P2 substitution.

Figure 2.

Validation/invalidation of hepatitis C virus NS3/4A protease inhibitors boceprevir, telaprevir, and narlaprevir as SARS CoV-2 M^pro inhibitors using the FlipGFP assay and Protease-Glo luciferase assay. (A) Representative images from the FlipGFP-M^pro assay. Dose-dependent decrease of GFP signal was observed with the increasing concentration of boceprevir, telaprevir or narlaprevir. The scale bar is 300 μm. (B)–(D) Dose−response curve of the GFP and mCherry fluorescent signals for boceprevir (B), telaprevir (C) or narlaprevir (D); mCherry signal alone was used to normalize protein expression level or calculate compound toxicity. (E)–(G) Protease-Glo luciferase assay results of boceprevir (E), telaprevir (F) or narlaprevir (G). Left column shows Firefly and Renilla luminescence signals in the presences of increasing concentrations of boceprevir, telaprevir or narlaprevir; Right column shows dose−response curve plots of the ratio of FFluc/Rlu luminescence. Renilla luminescence signal alone was used to normalize protein expression level. The results are mean ± standard deviation from three repeats. Compound concentration, μmol/L.

3.3. HIV protease inhibitors

HIV protease inhibitors, especially Kaletra, have been hotly pursued as potential COVID-19 treatment at the beginning of the pandemic. Kaletra was first tested in clinical trial during the SARS-CoV outbreak in 2003 and showed somewhat promising results based on the limited data⁴⁹. However, a double-blinded, randomized trial concluded that Kaletra was not effective in treating severe COVID-19^50,51. In SARS-CoV-2 infection ferret models, Kaletra showed marginal effect in reducing clinical symptoms, while had no effect on virus titers⁵².

Keletra is a combination of lopinavir and ritonavir. Lopinavir is a HIV protease inhibitor, and ritonavir is used as a booster. Ritonavir does not inhibit the HIV protease and it is a cytochrome P450-3A4 inhibitor⁵³. When used in combination, ritonavir can enhance other protease inhibitors by preventing or slowing down the metabolism. In cell culture, lopinavir was reported to inhibit the nanoluciferase SARS-CoV-2 reporter virus with an EC₅₀ of 9 μmol/L²⁴. In two other studies, lopinavir showed moderate antiviral activity against SARS-CoV-2 activity with EC₅₀ values of 19 ± 8 μmol/L³⁴ and 25 μmol/L³⁵. As such, it was assumed that lopinavir inhibited SARS-CoV-2 through inhibiting the M^pro. However, lopinavir showed no activity against SARS-CoV-2 M^pro in the FRET assay from our previous study (IC₅₀ > 60 μmol/L)². Wu et al.⁵⁴ also showed that lopinavir was a weak inhibitor against SARS-CoV M^pro with an IC₅₀ of 50 μmol/L. In the current study, we further confirmed the lack of binding of lopinavir to SARS-CoV-2 M^pro in the thermal shift assay (ΔT_m = −0.60 °C, Table 2). The result from the FlipGFP assay was not conclusive as lopinavir was cytotoxic. Lopinavir was not active in the Protease-Glo luciferase assay. Taken together, lopinavir is not a M^pro inhibitor.

Ritonavir was not active in both the FlipGFP M^pro and the Protease-Glo luciferase assays, which is consistent with the lack of activity in the FRET assay and the thermal shift binding assay (Table 2). Therefore, ritonavir is not a M^pro inhibitor.

We also tested additional HIV antivirals including atazanavir, nelfinavir, and cobicistat. Atazanavir and nelfinavir were reported as a potent SARS-CoV-2 antiviral with EC₅₀ values of 2.0 ± 0.12³⁸ and 0.77 μmol/L²⁴ using the infectious SARS-CoV-2 and the nanoluciferase reporter virus (SARS-CoV-2-Nluc), respectively. A drug repurposing screening similar identified nelfinavir as a SARS-CoV-2 antiviral with an IC₅₀ of 3.3 μmol/L³⁰. Gupta et al.³⁹ showed that cobicistat inhibited M^pro with an IC₅₀ of 6.7 μmol/L in the FRET assay. Cobicistat was also reported to have antiviral activity against SARS-CoV-2 with an EC₅₀ of 2.74 ± 0.20 μmol/L using the SARS-CoV-2-Nluc reporter virus²⁴. However, our FRET assay showed that nelfinavir and cobicistat did not inhibit M^pro in the FRET assay (IC₅₀ > 20 μmol/L), which was further confirmed by the lack of binding to M^pro in the thermal shift assay (Table 2). The results of atazanavir, nelfinavir, and cobicistat from the FlipGFP assay were not conclusive due to compound cytotoxicity (Fig. 3A–F). None of the compounds showed inhibition in the Protease-Glo luciferase assay (Fig. 3G–K).

Figure 3.

Validation/invalidation of HIV protease inhibitors lopinavir, ritonavir, atazanavir, nelfinavir, and cobicistat as SARS CoV-2 M^pro inhibitors using the FlipGFP assay and Protease-Glo luciferase assay. (A) Representative images from the FlipGFP-M^pro assay. The scale bar is 300 μm. (B)–(F) Dose−response curve of the GFP and mCherry fluorescent signals for lopinavir (B), ritonavir (C), atazanavir (D), nelfinavir (E), and cobicistat (F); mCherry signal alone was used to normalize protein expression level or calculate compound cytotoxicity. (G)–(K) Protease-Glo luciferase assay results of lopinavir (G), ritonavir (H), atazanavir (I), nelfinavir (J), and cobicistat (K). Left column shows Firefly and Renilla luminescence signals in the presences of increasing concentrations of lopinavir, ritonavir, atazanavir, nelfinavir, and cobicistat; Right column shows dose−response curve plots of ratio of FFluc/Rluc luminescence. Renilla luminescence signal alone was used to normalize protein expression level. None of the compounds shows significant inhibition in the presence of up to 240 μmol/L compounds. The results are mean ± standard deviation from three repeats. Compound concentration, μmol/L.

Collectively, our results have shown that the HIV protease inhibitors including lopinavir, ritonavir, atazanavir, nelfinavir, and cobicistat are not M^pro inhibitors. Nonetheless, given the potent antiviral activity of lopinavir, atazanavir, nelfinavir, and cobicistat against SARS-CoV-2, it might worth to conduct resistance selection to elucidate their drug target(s).

3.4. Bioactive compounds from drug repurposing

Several bioactive compounds have been identified as SARS-CoV-2 M^pro inhibitors through either virtual screening or FRET-based HTS. We are interested in validating these hits using the FlipGFP and the Protease-Glo luciferase assays.

Manidipine was identified as a SARS-CoV-2 M^pro inhibitor from a virtual screening and was subsequently shown to inhibit M^pro with an IC₅₀ of 4.81 μmol/L in the FRET assay⁴⁰. No antiviral data was provided. When we repeated the FRET assay, the IC₅₀ was 64.2 μmol/L (Table 2). Manidipine also did not show binding to M^pro in the thermal shift assay. Furthermore, manidipine showed no activity in either the FlipGFP assay or the Protease-Glo luciferase assay (Fig. 4A, B, and F). Therefore, our results invalidated manidipine as a SARS-CoV-2 M^pro inhibitor. A recent study independently confirmed our results and suggested that manidipine might form colloidal aggregates⁵⁵, leading to inactivation of M^pro. In the presence of 0.05% Tween-20, manidipine was not active against M^pro in the FRET assay (IC₅₀ > 100 μmol/L).

Figure 4.

Validation/invalidation of manidipine, calpain inhibitors II and XII, and ebselen as SARS CoV-2 M^pro inhibitors using the FlipGFP assay and Protease-Glo luciferase assay. (A) Representative images from the FlipGFP-M^pro assay. The scale bar is 300 μm. (B)–(E) Dose−response curve of the GFP and mCherry fluorescent signals for manidipine (B), calpain inhibitor II (C), calpain inhibitor XII (D), and ebselen (E); mCherry signal alone was used to normalize protein expression level or calculate compound cytotoxicity. (F)–(I) Protease-Glo luciferase assay results of manidipine (F), calpain inhibitor II (G), calpain inhibitor XII (H), and ebselen (I). Left column shows Firefly and Renilla luminescence signals in the presences of increasing concentrations of lopinavir, ritonavir, atazanavir, nelfinavir, and cobicistat; Right column shows dose−response curve plots of the ratio of FFluc/Rluc luminescence. Renilla luminescence signal alone was used to normalize protein expression level. (J)–(K) Antiviral activity of remdesivir (J), calpain inhibitor II (K), and calpain inhibitor XII (L) against SARS-CoV-2 in Calu-3 cells. The results are mean ± standard deviation from three repeats. Compound concentration, μmol/L.

In the same screening which we identified boceprevir as a SARS-CoV-2 M^pro inhibitor, calpain inhibitors II and XII were also found to have potent inhibition against M^pro with IC₅₀ values of 0.97 and 0.45 μmol/L in the FRET assay². Both compounds showed binding to M^pro in the thermal shift and native mass spectrometry binding assays. The Protease-Glo luciferase assay similarly confirmed the potent inhibition of calpain inhibitors II and XII against M^pro with IC₅₀ values of 0.60 and 0.79 μmol/L, respectively (Fig. 4G and H). However, calpain inhibitor II had no effect on the cellular M^pro activity as shown by the lack of inhibition in the FlipGFP assay (IC₅₀ > 60 μmol/L, Fig. 4A and C), while calpain inhibitor XII showed weak activity (IC₅₀ = 38.71 μmol/L, Fig. 4A and D). A recent study by Cao et al.⁴⁵ using a M^pro trigged cytotoxicity assay similarly found the lack of cellular M^pro inhibition by calpain inhibitors II and XII. These results contradict to the potent antiviral activity of both compounds in Vero E6 cells². It is noted that calpain inhibitors II and XII are also potent inhibitors of cathepsin L with IC₅₀ values of 0.41 and 1.62 nmol/L, respectively³. One possible explanation is that the antiviral activity of calpain inhibitors II and XII against SARS-CoV-2 might be cell type dependent, and the observed inhibition in Vero E6 cells might be due to cathepsin L inhibition instead of M^pro inhibition. Vero E6 cells are TMPRSS2 negative, and SARS-CoV-2 enters cell mainly through endocytosis and is susceptible to cathepsin L inhibitors⁵⁶. To further evaluate the antiviral activity of calpain inhibitors II and XII against SARS-CoV-2, we tested them in Calu-3 cells using the immunofluorescence assay (Fig. 4J, K, and L). Calu-3 is TMPRSS2 positive and it is a close mimetic of the human primary epithelial cell⁵⁷. As expected, calpain inhibitors II and XII displayed much weaker antiviral activity against SARS-CoV-2 in Calu-3 cells than in Vero E6 cells with EC₅₀ values of 30.34 and 14.78 μmol/L, respectively (Fig. 4K and L). These results suggest that the FlipGFP assay can be used to faithfully predict the antiviral activity of M^pro inhibitors. The lower activity of calpain inhibitors II and XII in the FlipGFP assay and the Calu-3 antiviral assay might due to the competition with host proteases, resulting in the lack of cellular target engagement with M^pro.

In conclusion, calpain inhibitors II and XII are validated as M^pro inhibitors but their antiviral activity against SARS-CoV-2 is cell type dependent. Accordingly, TMPRSS2 positive cell lines such as Calu-3 should be used to test the antiviral activity of calpain inhibitors II and XII analogs.

Ebselen is among one of the most frequently reported promiscuous M^pro inhibitors. It was first reported by Jin et al.¹⁶ that ebselen inhibits SARS-CoV-2 M^pro with an IC₅₀ of 0.67 μmol/L and the SARS-CoV-2 replication with an EC₅₀ of 4.67 μmol/L. However, it was noted that no reducing reagent was added in the FRET assay, and we reasoned that the observed inhibition might be due to non-specific modification of the catalytic cysteine 145 by ebselen. To test this hypothesis, we repeated the FRET assay with and without reducing reagent DTT or GSH, and found that ebselen completely lost the M^pro inhibition in the presence of DTT or GSH⁴¹. Similarly, ebselen also non-specifically inhibited several other viral cysteine proteases in the absence of DTT including SARS-CoV-2 PL^pro, EV-D68 2A^pro and 3C^pro, and EV-A71 2A^pro and 3C^pro41. The inhibition was abolished with the addition of DTT. Ebselen also had no antiviral activity against EV-A71 and EV-D68, suggesting that the FRET assay results obtained without reducing reagent cannot be used to predict the antiviral activity. In this study, we found that ebselen showed no inhibition in either the FlipGFP assay or the Protease-Glo luciferase assay (Fig. 4A, E, and I), providing further evidence for the promiscuous mechanism of action of ebselen. Another independent study by Gurard-Levin et al.²⁶ using mass spectrometry assay reached similar conclusion that the inhibition of M^pro by ebselen is non-specific and inhibition was abolished with the addition of reducing reagent DTT or glutathione. In contrary to the potent antiviral activity reported by Jin et al.¹⁶, the study from Gurard-Levin et al.²⁶ found that ebselen was inactive against SARS-CoV-2 in Vero E6 cells (EC₅₀ > 100 μmol/L). Chen et al.⁵⁸ reported that ebselen and disulfiram had synergistic antiviral effect with remdesivir against SARS-CoV-2 in Vero E6 cells. It was also proposed that ebselen and disulfiram act as zinc ejectors and inhibited not only the PL^pro59, but also the nsp13 ATPase and nsp14 exoribonuclease activities⁵⁸, further casting doubt on the detailed mechanism of action of ebselen.

Despite the accumulating evidence to support the promiscuous mechanism of action of ebselen, several studies continue to explore ebselen and its analogs as SARS-CoV-2 M^pro and PL^pro inhibitors^42,60,61. A number of ebselen analogs were designed and found to have comparable enzymatic inhibition and antiviral activity as ebselen. MR6-31-2 had slightly weaker enzymatic inhibition against SARS-CoV-2 M^pro compared to ebselen (IC₅₀ = 0.824 vs. 0.67 μmol/L); however, MR6-31-2 had more potent antiviral activity than ebselen (EC₅₀ = 1.78 vs. 4.67 μmol/L) against SARS-CoV-2 M^pro in Vero E6 cells. X-ray crystallization of SARS-CoV-2 M^pro with MR6-31-2 (PDB: 7BAL) and ebselen (PDB: 7BAK) revealed nearly identical complex structures. It was found that selenium coordinates directly to Cys145 and forms a S–Se bond⁴². Accordingly, a mechanism involving hydrolysis of the organoselenium compounds was proposed. Similar to their previous study, the M^pro enzymatic reaction buffer (50 mmol/L Tris pH 7.3, 1 mmol/L EDTA) did not include the reducing reagent DTT. Therefore, the M^pro inhibition by these ebselen analogs might be non-specific and the antiviral activity might arise from other mechanisms⁴².

Overall, it can be concluded that ebselen is not a specific M^pro inhibitor, and its antiviral activity against SARS-CoV-2 might involve other drug targets such as nsp13 or nsp14.

Disulfiram is an FDA-approved drug for alcohol aversion therapy. Disulfiram has a polypharmacology and was reported to inhibit multiple enzymes including urease⁶², methyltransferase⁶³, and kinase⁶² through reacting with cysteine residues. Disulfiram was also reported as an allosteric inhibitor of MERS-CoV PL^pro64. Yang et al. reported disulfiram as a M^pro inhibitor with an IC₅₀ of 9.35 μmol/L. Follow up studies by us and others showed that disulfiram did not inhibit M^pro in the presence of DTT or GSH^26,41. In this study, disulfiram had no inhibition against M^pro in either the FlipGFP assay or the Protease-Glo luciferase assay (Fig. 5A, B and N).

Figure 5.

Validation/invalidation of disulfiram, carmofur, PX-12, tideglusib, shikonin, baicalein, chloroquine, hydroxychloroquine, oxytetracycline, montelukast, candesartan, and dipyridamole as SARS CoV-2 M^pro inhibitors using the FlipGFP assay and Protease-Glo luciferase assay. (A) Representative images from the FlipGFP-M^pro assay. The fluorescent background for dipyridamole at high drug concentrations (last row) was caused by the fluorescence emission from the compound itself. The scale bar is 300 μm. (B)–(M) Dose−response curve of the ratio of GFP/mCherry fluorescent signal for disulfiram (B), carmofur (C), PX-12 (D), tideglusib (E), shikonin (F), baicalein (G), chloroquine (H), hydroxychloroquine (I), oxytetracycline (J), montelukast (K), candesartan (L), and dipyridamole (M); mCherry signal alone was used to normalize protein expression level or calculate compound cytotoxicity. (N)–(Y) Protease-Glo luciferase assay results of disulfiram (N), carmofur (O), PX-12 (P), tideglusib (Q), shikonin (R), baicalein (S), chloroquine (T), hydroxychloroquine (U), oxytetracycline (V), montelukast (W), candesartan (X), and dipyridamole (Y). Left column shows Firefly and Renilla luminescence signals in the presences of increasing concentrations of disulfiram, carmofur, PX-12, tideglusib, shikonin, baicalein, chloroquine, hydroxychloroquine, oxytetracycline, montelukast, candesartan, and dipyridamole; Right column shows dose−response curve plots of the ratio of FFluc/Rluc luminescence. Renilla luminescence signal alone was used to normalize protein expression level. The results are mean ± standard deviation from three repeats. Compound concentration, μmol/L.

Similar to disulfiram, carmofur, PX-12 and tideglusib, which were previously claimed by Jin et al.¹⁶ as M^pro inhibitors, showed no inhibitory activity in either the FlipGFP or Protease-Glo luciferase assay (Fig. 5A, C, D, E, O–Q), which is consistent with their lack of inhibition in the FRET assay in the presence of DTT.

Shikonin and baicalein are polyphenol natural products with known polypharmacology. Both compounds showed no inhibition in either the FlipGFP or the Protease-Glo luciferase assay (Fig. 5A, F, G, R, and S), suggesting they are not M^pro inhibitors. These two compounds were previously reported to inhibit SARS-CoV-2 M^pro in the FRET assay16, 45, 46 and had antiviral activity against SARS-CoV-2 in Vero E6 cells. However, our recent study showed that shikonin had no inhibition against SARS-CoV-2 M^pro in the FRET assay in the presence of DTT⁴¹. Studies from Gurard-Levin et al.²⁶ using FRET assay and mass spectrometry assay reached the same conclusion. X-ray crystal structure of SARS-CoV-2 M^pro in complex with shikonin showed that shikonin binds to the active site in a non-covalent manner⁴⁴.

In addition to the proposed mechanism of action of M^pro inhibition, Zandi et al.⁶⁵ showed that baicalein and baicalin inhibit the SARS-CoV-2 RNA-dependent RNA polymerase. Overall, shikonin and baicalein are not M^pro inhibitors and the antiviral activity of baicalein against SARS-CoV-2 might involve other mechanisms.

A recent study from Li et al.³⁷ identified several known drugs as SARS-CoV-2 M^pro inhibitors from a virtual screening. The identified compounds include chloroquine (IC₅₀ = 3.9 ± 0.2 μmol/L; K_i = 0.56 ± 0.12 μmol/L), hydroxychloroquine (IC₅₀ = 2.9 ± 0.3 μmol/L; K_i = 0.36 ± 0.21 μmol/L), oxytetracycline (IC₅₀ = 15.2 ± 0.9 μmol/L; K_i = 0.99 ± 0.06 μmol/L), montelukast (IC₅₀ = 7.3 ± 0.5 μmol/L), K_i = 0.48 ± 0.04 μmol/L), candesartan (IC₅₀ = 2.8 ± 0.3 μmol/L; K_i = 0.18 ± 0.02 μmol/L), and dipyridamole (K_i = 0.04 ± 0.001 μmol/L). The discovery of chloroquine and hydroxychloroquine as M^pro inhibitor was particularly intriguing. Several high-throughput screenings have been conducted for M^pro30,66, and chloroquine and hydroxychloroquine were not among the list of active hits. In our follow up study, we found that none of the identified hits reported by Luo et al. inhibited M^pro either with or without DTT in the FRET assay⁵. In corroborate with our previous finding, the FlipGFP and Protease-Glo luciferase assays similarly confirmed the lack of inhibition of these compounds against M^pro (Fig. 5A, H–M, T–Y). Therefore, it can be concluded that chloroquine, hydroxychloroquine, oxytetracycline, montelukast, candesartan, and dipyridamole are not SARS-CoV-2 M^pro inhibitors. Other than the claims made by Luo et al., no other studies have independently confirmed these compounds as M^pro inhibitors.

4. Conclusions

The M^pro is perhaps the most extensive exploited drug target for SARS-CoV-2. A variety of drug discovery techniques have been applied to search for M^pro inhibitors. Researchers around the world are racing to share their findings with the scientific community to expedite the drug discovery process. However, the quality of science should not be compromised by the speed. The mechanism of action of drug candidates should be thoroughly characterized in biochemical, binding, and cellular assays. Pharmacological characterization should address both target specificity and cellular target engagement. For target specificity, the drug candidates can be counter screened against unrelated cysteine proteases such as the viral EV-A71 2A^pro, EV-D68 2A^pro, the host cathepsins B, L, and K, caspases, calpains I, II, and III, etc. Compounds inhibit multiple cysteine proteases non-discriminately are most likely promiscuous compounds that act through redox cycling, inducing protein aggregation, or alkylating catalytic cysteine residue C145. For cellular target engagement, the FlipGFP and Protease-Glo luciferase assays can be applied. Both assays are performed in the presence of competing host proteins at the cellular environment. Collectively, our study reaches the following conclusions: 1) for validated M^pro inhibitors, the IC₅₀ values with and without reducing reagent should be about the same in the FRET assay; 2) validated M^pro inhibitors should show consistent results in the FRET assay, thermal shift binding assay, and the Protease-Glo luciferase assay. For compounds that are not cytotoxic, they should also be active in the FlipGFP assay; 3) compounds that have antiviral activity but lack consistent results from the FRET, thermal shift, FlipGFP, and Protease-Glo luciferase assays should not be classified as M^pro inhibitors; 4) compounds that non-specifically inhibit multiple unrelated viral or host cysteine proteases are most likely promiscuous inhibitors that should be triaged. 5) X-ray crystal structures cannot be used to justify the target specificity or cellular target engagement. Promiscuous compounds have been frequently co-crystallized with M^pro including ebselen, carmofur, and shikonin (Table 2). Overall, we hope our studies will promote the awareness of the promiscuous SARS-CoV-2 M^pro inhibitors among the scientific community and call for more stringent hit validation.

Author contributions

Chunlong Ma performed the FlipGFP assay, Protease-Glo luciferase assay, and thermal shift assay with the assistance from Haozhou Tan. Juliana Choza and Yuyin Wang expressed the M^pro and performed the FRET assay. Jun Wang wrote the draft manuscript with the input from others; Jun Wang submitted this manuscript on behalf of other authors.

Conflicts of interest

The authors have no conflicts of interest to declare.