In Silico and In Vitro Identification of Pan-Coronaviral Main Protease Inhibitors from a Large Natural Product Library

The main protease (Mpro or 3CLpro) in coronaviruses represents a promising specific drug target as it is essential for the cleavage of the virus polypeptide and has a unique cleavage site that does not exist in human host proteases. In this study, we explored potential natural pan-coronavirus drugs using in vitro and in silico approaches and three coronavirus main proteases as treatment targets. The PyRx program was used to screen 39,442 natural-product-like compounds from the ZINC database and 121 preselected phytochemicals from medicinal plants with known antiviral activity. After assessment with Lipinski’s rule of five, molecular docking was performed for the top 33 compounds of both libraries. Enzymatic assays were applied for the top candidates from both in silico approaches to test their ability to inhibit SARS-CoV-2 Mpro. The four compounds (hypericin, rosmarinic acid, isorhamnetin, and luteolin) that most efficiently inhibited SARS-CoV-2 Mpro in vitro were further tested for their efficacy in inhibiting Mpro of SARS-CoV-1 and MERS-CoV. Microscale thermophoresis was performed to determine dissociation constant (Kd) values to validate the binding of these active compounds to recombinant Mpro proteins of SARS-CoV-2, SARS-CoV-1, and MERS-CoV. The cytotoxicity of hypericin, rosmarinic acid, isorhamnetin, and luteolin was assessed in human diploid MRC-5 lung fibroblasts using the resazurin cell viability assay to determine their therapeutic indices. Sequence alignment of Mpro of SARS-CoV-2 demonstrated 96.08%, 50.83%, 49.17%, 48.51%, 44.04%, and 41.06% similarity to Mpro of other human-pathogenic coronaviruses (SARS-CoV-1, MERS-CoV, HCoV-NL63, HCoV-OC43, HCoV-HKU1, and HCoV-229E, respectively). Molecular docking showed that 12 out of 121 compounds were bound to SARS-CoV-2 Mpro at the same binding site as the control inhibitor, GC376. Enzyme inhibition assays revealed that hypericin, rosmarinic acid, isorhamnetin, and luteolin inhibited Mpro of SARS-CoV-2, while hypericin and isorhamnetin inhibited Mpro of SARS-CoV-1; hypericin showed inhibitory effects toward Mpro of MERS-CoV. Microscale thermophoresis confirmed the binding of these compounds to Mpro with high affinity. Resazurin assays showed that rosmarinic acid and luteolin were not cytotoxic toward MRC-5 cells, whereas hypericin and isorhamnetin were slightly cytotoxic. We demonstrated that hypericin represents a potential novel pan-anti-coronaviral agent by binding to and inhibiting Mpro of several human-pathogenic coronaviruses. Moreover, isorhamnetin showed inhibitory effects toward SARS-CoV-2 and SARS-CoV-1 Mpro, indicating that this compound may have some pan-coronaviral potential. Luteolin had inhibitory effects against SARS-CoV-2 Mpro.


In Silico Studies
By using a ZINC library of 39,442 natural-product-like compounds and a second natural product library of 121 compounds that were preselected from medicinal plants with known antiviral activity, we performed virtual screening with PyRx. A total of 89 hits from the ZINC natural-product-like library and 32 hits from the antiviral natural product library were selected on the basis of their lowest PyRx-based binding energies to SARS-CoV-2 M pro . These compounds were assessed using the Lipinski rules, and compounds with a molecular weight ≤ 500 and an octanol-water partition coefficient of log p ≤ 5 were considered for further investigation, i.e., 21 candidate compounds from the ZINC database and 12 candidates from the antiviral natural product library. These 33 compounds were subjected to molecular docking with AutoDock 4.2.6. The lowest binding energy values (LBEs) and predicted inhibition constants (pKi's) of the 12 candidates from the antiviral natural product library that were used for subsequent experiments are shown in Table 1. Figure 1 depicts the structures of these 12 compounds and the known M pro inhibitor, GC376, which was used as a positive control. To identify similarities between the main protease of seven human coronaviruses, we performed protein alignments. The results revealed 96.08%, 50.83%, 49.17%, 48.51%, 44.04%, and 41.06% identity of SARS-CoV-2 M pro to SARS-CoV-1, MERS-CoV, HCoV-NL63, HcoV-OC43, HcoV-HKU1, and HcoV-229E, respectively ( Table 2). Figure 2 shows the highly conserved amino acid residues between the seven human coronaviruses.

Inhibition of M pro Enzyme Activity
We performed in vitro enzymatic assays to validate whether the compounds selected from the in silico studies inhibit the activity of M pro of SARS-CoV-2. As expected, hypericin, rosmarinic acid, isorhamnetin, and luteolin were the most active natural products among the 12 compounds preselected by our bioinformatical approach. These compounds inhibited enzymatic activity by more than 50% ( Figure 3). Therefore, these four compounds were subjected to subsequent dose-response experiments to calculate the concentration of each compound required to inhibit M pro activity by half (IC 50 ). The percentage of activity versus the log concentration of the inhibitors was used to calculate the IC 50 values ( Figure 4A-C). The IC 50 values for hypericin, rosmarinic acid, isorhamnetin, and luteolin for SARS-CoV-2 CL pro were 23. 30, 9.43, 8.42, and 11.81 µM, respectively. The IC 50 values for the inhibition of M pro of SARS-CoV-1 by hypericin and isorhamnetin were 19.43 and 13.13 µM, respectively. Rosmarinic acid and luteolin inhibited the enzymatic activity of M pro of SARS-CoV-1 only at the highest concentration of 100 µM by 31% and 44%, respectively. The IC 50 value for the inhibition of MERS-CoV M pro by hypericin was 49.65 µM. At a concentration of 100 µM, rosmarinic acid, luteolin, and isorhamnetin inhibited MERS-CoV M pro activity to 14.9%, 21.3%, and 26.3%, respectively (Table 3).

Microscale Thermophoresis
Microscale thermophoresis is a sensitive technique used to determine the binding between unlabeled molecules and labeled macromolecules (i.e., proteins). The labeled recombinant M pro of SARS-CoV-2, SARS-CoV-1, and MERS-CoV were titrated against different concentrations of the selected compounds ( Figure 5A

Binding of the Top Candidates
Molecular docking in silico revealed high binding affinities of the best candidate compounds and the control inhibitor, GC376, to M pro of seven human coronaviruses ( Table 5). Figure 6A-C shows the molecular interactions of potential inhibitors and GC376 with SARS-CoV-2, SARS-CoV-1, and MERS-CoV M pro . The best candidate compounds shared the same binding site at M pro as the control inhibitor, GC376. Hypericin, rosmarinic acid, isorhamnetin, luteolin, and the control inhibitor, GC376, formed hydrogen bonds or hydrophobic interactions with at least one of the catalytic residues (Cys 145 , His 41

Cell Viability Assay
The inhibitory effects on the cell viability of the potential M pro inhibitors toward human MRC-5 fibroblasts were assessed using the resazurin assay. As shown in Figure 8, luteolin and rosmarinic acid did not show significant inhibitory effects within the tested concentration range. Hypericin and isorhamnetin showed a slight inhibition of viability of MRC-5 cells with CC 50 values of 55.46 ± 2.2 µM and 36.80 ± 3.4 µM, respectively ( Table 6). The therapeutic indices of hypericin for SARS-CoV-2, SARS-CoV-1, and MERS-CoV M pro were 2.38, 2.85, and 1.11, respectively. For isorhamnetin, the therapeutic index was 4.37 and 2.8 for SARS-CoV-2 and SARS-CoV-1, respectively (Table 7).

Cell Viability Assay
The inhibitory effects on the cell viability of the potential M pro inhibitors toward human MRC-5 fibroblasts were assessed using the resazurin assay. As shown in Figure 8, luteolin and rosmarinic acid did not show significant inhibitory effects within the tested concentration range. Hypericin and isorhamnetin showed a slight inhibition of viability of MRC-5 cells with CC50 values of 55.46 ± 2.2 μM and 36.80 ± 3.4 μM, respectively (Table

Discussion
During the past two decades, highly infectious pathogens rapidly emerged, such as SARS-CoV-1 in 2003, MERS-CoV in 2012, and SARS-CoV-2 at the end of 2019. Therefore, there is an urgent need to investigate new broad-spectrum anti-CoVs drugs. The main protease was proposed as a promising target for the development of pan-coronaviral drugs as it significantly differs from human proteases and is highly conserved between coronavirus family members [16].
In this study, we first performed a computer-based approach to screen 39,442 naturalproduct-like compounds from the ZINC database and 121 preselected natural products from medicinal plants with known antiviral activity to find candidate compounds with a high binding affinity to SARS-CoV-2 M pro . As a result of PyRx-based virtual drug screening, assessment using the Lipinski rule of five, and molecular docking using AutoDock 4.2.6, 33 compounds were selected and subjected to AutoDock to validate the virtual screening results. Twelve compounds were selected for further in vitro experiments. To analyze whether these 12 compounds affect SARS-CoV-2 M pro activity in vitro, we performed M pro enzyme activity inhibition assays. Hypericin, rosmarinic acid, isorhamnetin, and luteolin inhibited SARS-CoV-2 M pro . The binding site of SARS-CoV-2 includes a catalytic dyad (His41 and Cys145) and several subsites (S1-S5). The S1 subunit consists of His163, Glu166, Cys145, Gly143, His172, and Phe140. The S2 subunit comprises Cys145, His41, and Thr25. S3-S5 consists of Met165, Met49, His41, Glu166, and Gln189. These subunits play a key role in substrate binding [32,33]. Molecular docking revealed that hypericin, rosmarinic acid, isorhamnetin, and luteolin not only bound to M pro through these subunits but also interacted (hydrophobic or hydrogen binding) with at least one of the catalytic center residues (His41 and Cys145) ( Figure 7A). Moreover, microscale thermophoresis confirmed the binding of these four natural compounds to SARS-CoV-2 M pro . Although the K d values were different, all compounds showed a high binding affinity to SARS-CoV-2 M pro . Consequently, we concluded that the in silico data reflected the in vitro situation as there was a good correlation between the computationally predicted lowest binding energies (−12.44, −9.98, −9.06, and −9.01 kcal/mol) and the experimentally measured percentages of activity (4.95%, 8.28%, 8.56%, and 10.10%) of SARS-CoV-2 M pro in the presence of hypericin, rosmarinic acid, isorhamnetin, and luteolin, respectively. To develop a potential pan-HCoV inhibitor, we also performed M pro enzyme activity inhibition assays for SARS-CoV-1 and MERS-CoV. Hypericin and isorhamnetin inhibited SARS-CoV-1 M pro , while only hypericin inhibited MERS-CoV M pro . Microscale thermophoresis confirmed that these two compounds were bound to SARS-CoV-1 and MERS-CoV M pro with high affinities.
Cell viability assays showed that luteolin and rosmarinic acid did not inhibit human fetal MRC-5 lung fibroblasts in the highest concentration tested (100 µM) while hypericin and isorhamnetin showed slight toxicity.
Hypericin is a natural polyquinone from Hypericum perforatum (St. John's wort) and is traditionally used as an anti-depressive and wound-healing drug [34]. Hypericin has antitumor, antivirus, and anti-depression activity. It exhibits in vitro activity against infectious bronchitis virus (IBV) by the inhibition of apoptosis in host cells and the production of reactive oxygen species [35]. Additionally, hypericin inhibits hepatitis C virus (HCV) replication via downregulation of heme oxygenase-1 expression and deacetylation in vitro [36]. However, hypericin caused phytotoxicity without detectable anti-HCV activity in patients with chronic HCV infection who were provided oral doses of 0.05 and 0.10 mg/kg/d [37]. Although hypericin inhibits human immunodeficiency virus (HIV) in vitro and in vivo [38,39], a clinical trial revealed phytotoxicity of orally administered hypericin (0.5 mg/kg daily) without antiretroviral activity in a limited number of patients [40]. Moreover, hypericin inhibits the replication of α-coronaviruses (PEDV and TGEV) through the inhibition of M pro [41]. Hypericin inhibits M pro of SARS-CoV-2 with a CC 50 value of 63.6 µM [42]. Our in silico and in vitro results indicated that hypericin both binds and inhibits M pro of β-coronaviruses. Previous in silico and in vitro studies showed that hypericin has anti-inflammatory activity [43][44][45] and is a potential treatment for rheumatoid arthritis [44]. Thus, hypericin is a promising pan-CoV inhibitor that, due to its anti-inflammatory effects, may be used in coronavirus-infected patients suffering from autoimmune reactions ("long COVID") who are prohibited from obtaining anti-coronavirus vaccinations. This is an advantageous feature that distinguishes this compound from other approved drugs.
Rosmarinic acid, an ester of caffeic acid and 3,4-dihydroxy phenyl lactic acid, is present in most Lamiaceae species [46]. It has a broad inhibitory effect on a variety of viruses, e.g., rosmarinic acid inhibits HBV replication in HBV-infected cells by specifically targeting DNA polymerase ε binding [47]. It also inhibits influenza viruses and enterovirus 71 [48,49]. Although previous studies suggested that rosmarinic acid inhibits SARS-CoV-2 replication with an IC 50 value of 25.47 ng/µL, the mechanism of action is still unknown [50]. Our results indicated that rosmarinic acid binds to SARS-CoV-2 M pro with an IC 50 value of 9.43 µM. Additionally, several animal studies revealed that rosmarinic acid has anti-inflammatory activity through the inhibition of NF-κB and STAT3 signaling pathways [51,52] and may be applied against arthritis, inflammatory bowel disease, and asthma [53]. Hence, rosmarinic acid is a potential therapeutic against COVID-19, especially for fighting immunological overreactions (i.e., the cytokine storm) during severe courses of the disease.
Isorhamnetin is a flavonoid from Hippophae rhamnoides L. [54], Artemisia absinthium L. [55], and other plants. Isorhamnetin has a wide range of pharmacological effects on cardiovascular diseases, a variety of tumors, and neurodegenerative diseases [54]. Isorhamnetin exerts antiinfluenza effects in vitro and in vivo by inhibition of hemagglutinin and neuraminidase [56]. Isorhamnetin also inhibits SARS-CoV-2 entry through inhibition of the Spike protein [57]. Our results showed that isorhamnetin binds and inhibits M pro of SARS-CoV-2 and SARS-CoV-1 with IC 50 values of 8.42 and 13.13 µM, respectively. Isorhamnetin has anti-inflammatory effects against different diseases, such as inflammatory bowel disease [58], osteoarthritis, and periodontitis, by suppressing the production of inflammatory mediators, cytokines, and reactive oxygen species [54]. Hence, targeting the Spike protein and M pro makes isorhamnetin a promising drug candidate for the inhibition of coronavirus entry and replication.
Luteolin is a natural flavonoid that is extensively present in many plant species [59]. It has multiple biological effects such as anti-inflammation, antiallergy, and anticancer activities [60]. In vivo and in vitro studies demonstrated that luteolin inhibits HBV replication through ERK-mediated downregulation of HNF4α expression [61]. It also exhibits antiviral activity against influenza A virus, HIV-1, and JEV [62][63][64]. Our results demonstrated that luteolin binds to SARS-CoV-2 M pro and inhibits its activity with an IC 50 value of 11.81 µM. Several in vivo and in vitro studies revealed that luteolin has an anti-inflammatory effect by blocking the NF-κB and AP-1 activation pathways [65][66][67][68]. Luteolin was also suggested as a potential therapeutic strategy for various autoimmune diseases, such as Hashimoto's thyroiditis and multiple sclerosis [69,70]. Therefore, luteolin may be beneficial for COVID-19 patients with overshooting autoimmune reactions.
In conclusion, we demonstrated that it is possible to identify natural products that exert activity against several coronaviruses and may be useful for developing pan-coronaviral drugs. There was some selectivity between the inhibition of coronaviral M pro and cytotoxic activity toward human lung cells. Though the cytotoxicity was very low, the inhibitory rates toward the tested coronaviral main proteases were in the micromolar but not nanomolar range. Hence, animal experimentation should clarify whether this in vitro activity is reflected in vivo. Furthermore, the chemical scaffolds of the identified natural products may serve as lead structures when generating (semi)synthetic derivatives with improved activity. The concept of developing pan-coronaviral drugs is attractive for being prepared for future outbreaks of epidemic or pandemics by known or novel coronaviruses.

Compounds
The chemical structures of natural products were downloaded from ZINC and Pub-Chem databases in three-dimensional SDF format. Based on in silico studies, 12 selected compounds were provided by Fischer Analytics/Fischer Organics GmbH (Weiler, Germany). The compounds had a purity of >95%.

Virtual Screening
In this study, the PyRx software was used for the virtual screening of 39,442 naturalproduct-like compounds from the ZINC database and 121 natural compounds used in herbal medicines against viral diseases. As a target, the dimeric form of SARS-CoV-2 M pro (PDB ID: 6XMk) was chosen to identify compounds with high binding affinity and low binding energy (kcal/mol). AutoDock version 1.5.6 was used to convert the Protein Data Bank files of target proteins (PDB) to PDBQT files. The energy of the compounds was minimized and converted from SDF format to PDBQT format using the PyRx software.

Inhibition of M pro Enzyme Activity
Enzymatic assays were performed using the SensoLyte SARS-CoV-2 3CL Protease Activity Assay Kit (AnaSpec, San Francisco, CA, USA), SARS-CoV-1 Assay Kit, and 3CL Protease MERS-CoV Assay Kit (BPS Bioscience, San Diego, CA, USA). Twelve selected compounds were diluted in assay buffer to a final concentration of 1 mM. Compound aliquots of 10 µL were added to black 96-well plates (Greiner, Frickenhausen, Germany), and 40 µL of 0.1 mg/mL M pro was added to each well of the plates and incubated with the compounds at 37 • C for 30 min. The enzymatic reactions were initiated by adding a fluorescent substrate. The final concentration of the compound was 100 µM. Fluorescence was measured using an Infinite M2000 Pro plate reader (Tecan, Crailsheim, Germany). All values were subtracted from blank values. Then, compounds exhibiting more than 50% inhibitory activity at a fixed concentration of 100 µM were selected for dose-response studies in a concentration range from 0 to 100 µM for SARS-CoV-2, SARS-CoV-1, and MERS-CoV to calculate 50% inhibition concentrations (IC 50 ). The activity percentage of M pro was calculated using the following equation: Activity % = 100 − [(RFU Vehicle control − RFU tested sample )/RFU Vehicle control × 100].

Cell Viability Assay
Cell viability was measured using the resazurin assay as previously described in [72]. Human diploid MRC-5 lung fibroblasts was kindly provided by Dr. rer. nat. Sebastian Zahnreich (Department of Radiation Oncology and Radiation Therapy, University Medical Center of the Johannes Gutenberg University, Mainz, Germany) were seeded (5 × 10 5 cells per well) into 96-well culture plates and incubated overnight before treatment. On the second day, the cells were treated with 10 concentrations of the four compounds in a range of 0.3-100 µM. After 72 h incubation, 20 µL 0.01% resazurin (Promega, Mannheim, Germany) was added to each well. Fluorescence was detected after 4 h incubation using an Infinite M2000 Pro plate reader (Tecan) at Ex/Em = 550 nm/590 nm wavelength. Cell viability was calculated in comparison to DMSO control. The DMSO final concentration was 0.5%. The 50% cytotoxicity concentration (CC 50 ) values were calculated in comparison to the DMSO-treated control. Each experiment was independently repeated three times with six wells for each concentration. Therapeutic indices were calculated using the following equation: Therapeutic index = TD 50 /ED 50 .

Conclusions
Overall, our in silico and in vitro results demonstrated that hypericin is a potential novel pan-anti-coronaviral agent as it binds to and inhibits M pro of human-pathogenic coronaviruses. Moreover, isorhamnetin showed inhibitory effects toward SARS-CoV-2 and SARS-CoV-1 M pro , while luteolin revealed inhibitory effects against SARS-CoV-2 M pro ( Figure 9). Our results need to be further validated in animal models and clinical trials. A typical feature of natural products is that they are frequently multi-specific, i.e., they exert specific activities against several targets [73]. Therefore, the anti-coronaviral activity of the compounds we investigated is complemented with known anti-inflammatory effects reported in the literature. This may qualify these compounds not only to inhibit coronavirus replication but also to improve inflammatory conditions in severe courses of COVID-19 and other coronavirus infections. Natural products are generally considered to be of low toxicity. This is another property that speaks in favor of the four compounds we investigated.