To date, a number of old drugs have been clinically tested for potential efficacy against SARS-CoV-2, such as lopinavir/ritonavir, hydroxychloroquine, darunavir and cobicistat20,21,24,25. Although some of these molecules are effective in cellular models, clinical trials showed no significant improvement in symptoms and length of hospitalization21,26; thus, the efficacy of these drugs remains controversial. Since the targets of these drugs on SARS-CoV-2 are unclear, the mechanism is ambiguous, which suggests that it is essential and reliable to start from a defined target to identify candidate drugs. Although coronaviruses undergo extensive mutagenesis, some key proteins, especially replication-related enzymes, such as the main protease, RNA-dependent RNA polymerase and helicase, are highly conserved among different strains5,8,27. In the current study, the crystal structures of SARS-CoV-2 Mpro were selected as the target for molecular screening. Then, an in silico high-throughput screening strategy and automatic pipeline were established by using classic docking software and our in-house program, which greatly accelerated the screening process.
Virtual drug screening in silico allows us to rapidly identify possible candidates on a large scale, making drug exploration more effective and economic than traditional strategies28. However, the accuracy of prediction obtained with current software could not reach 100%, even when different software programs were combined29. Here, we found the flexibility of the protein structure, especially the active pocket, has a strong impact on the performance of the docking process. Fortunately, several crystal structures of Mpro of SARS-CoV-2 have been solved and are currently available17,30, so we used these structures as targets for multiple molecular docking. The average score is a much better indicator of the real diversity of molecular interactions than molecular dynamics simulations for flexible structures. The KDs determined by SPR were consistent with the prediction made by using Autodock, which demonstrated that the protocol used for screening candidate drugs is reliable.
Based on the analysis of docking diversity and effectiveness, this study indicates that the search for molecules targeting Mpro is constrained by similarity searching based on molecular fingerprints. Moreover, until many positive molecular samples are obtained, the effective use of artificial intelligence for drug exploration is difficult. Therefore, we suggest that structural optimization of the above high-affinity drugs as lead compounds is an important method for exploration of new Mpro inhibitors.
This study identified four drugs, namely, delafloxacin, nelfinavir, saquinavir and dolutegravir, which exhibited affinities higher than e-5 M with Mpro from SARS-CoV-2, indicating that these drugs may serve as potential antiviral drugs against SARS-CoV-2. In addition, three other drugs, namely, indinavir, baloxavir marboxil and tadalafil, showed moderate affinities at the e-5 M level, so a higher dose might be required for clinical administration, which would be a challenge for relevant studies.
Delafloxacin is a fluoroquinolone antibiotic used to treat skin infections and pneumonia caused by bacteria in adults31,32. In this study, this drug exhibited the highest affinity with Mpro as determined by SPR; thus, it may serve as a novel candidate drug for the treatment of SARS-CoV-2 infection. Delafloxacin is widely used in clinical practice, given the activity and safety of this drug, it is feasible to conduct exploratory clinical treatment in special cases, such as during the SARS-CoV-2 pandemic. Interestingly, three of the drugs mentioned above have already been evaluated in cellular experiments. Among them, nelfinavir is an HIV-1 protease inhibitor that was reported to inhibit the replication of SARS-CoV-2 efficiently, with effective concentrations for 50% and 90% inhibition (EC50 and EC90) at 1.13 µM and 1.76 µM, respectively33. In the current study, the KD between nelfinavir and Mpro was determined to be 2.36e-6 M. In addition, the concentrations of saquinavir and indinavir required to inhibit 50% of virus replication were reported to be 8.63 µM and 59.14 µM, respectively33. Our results also showed that these drugs exhibited high affinity with Mpro, and the KDs were comparable with the IC50s. The high KDs determined in this study provided solid evidence that Mpro is the specific target for the drug-mediated replication inhibition of SARS-CoV-2.
Moxifloxacin has ever been widely used to control subsequent infection during COVID-19 treatment34. This study indicated a significant affinity for Mpro at the e-4 M level, which should be slightly helpful in suppressing the virus, but a retrospective investigation is needed because it might be difficult to achieve an obvious effect in the body at the conventional dose. Moreover, darunavir, an HIV protease inhibitor35, has also been confirmed to suppress SARS-CoV-2 in cellular experiments33, but our results showed that its affinity is lower than that of dihydroergotamine and moxifloxacin.
Hydroxychloroquine is an anti-malarial drug36 that has been previously reported to exhibit an anti-SARS-CoV-2 effect in cellular assays, with an IC50 of 1.31 µM37, which is nearly the same as that of nelfinavir33. Clinical studies were designed and implemented to evaluate the efficacy of hydroxychloroquine in patients with COVID-19, but the outcomes were contradictory25,38. In this study, SeeSAR predicted that hydroxychloroquine may interact with Mpro, but it was confirmed in the subsequent molecular experiment that the KD of hydroxychloroquine is only e-3 M; thus, the affinity is very poor and close to that of streptomycin. These results demonstrated that hydroxychloroquine may exploit another mechanism to suppress viruses ex vivo, but not by targeting Mpro.
Taken together, based on virtual drug screening combined with experimental confirmation, we have identified seven drugs that specifically target SARS-CoV-2 Mpro, and the high affinity with Mpro indicates that these drugs may serve as promising candidate drugs for the treatment of SARS-CoV-2 infection. Furthermore, our results also provided a possible target and rational explanation for the efficacy of five candidate drugs that have already been evaluated in cell models, including nelfinavir, saquinavir, indinavir, darunavir, and hydroxychloroquine.