key: cord-0293796-ust4u61c authors: Xia, Xin; Zhang, Yuwei; Li, Songling; Lin, Hengwei; Yan, Zhiqiang title: Structure-based screening of drug candidates targeting the SARS-CoV-2 envelope protein date: 2021-08-25 journal: bioRxiv DOI: 10.1101/2021.08.25.457645 sha: 7293bc0d4f293b94458d091bfc2a907ba6d05db0 doc_id: 293796 cord_uid: ust4u61c The COVID-19 (coronavirus disease 2019) pandemic is caused by SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2). SARS-CoV-2 produces a small hydrophobic envelope (E) protein which shares high homology with SARS-CoV E protein. By patch-clamp recording, the E protein is demonstrated to be a cation-selective ion channel. Furthermore, the SARS-CoV-2 E protein can be blocked by a SARS-CoV E protein inhibitor hexamethylene amiloride. Using structural model and virtual screening, another E protein inhibitor AZD5153 is discovered. AZD5153 is a bromodomain protein 4 inhibitor against hematologic malignancies in clinical trial. The E protein amino acids Phe23 and Val29 are key determinants for AZD5153 sensitivity. This study provides two promising lead compounds and a functional assay of SARS-CoV-2 E protein for the future drug candidate discovery. To test if E protein of SARS-CoV-2 is indeed an ion channel, the full sequence was synthesized and expressed in CHO cells utilizing pcDNA3.1 vector with an IRES-GFP. We then performed whole-cell patch-clamp recording of CHO cells expressing E protein. Compared with the untransfected CHO cells, the cells transfected with SARS-CoV-2 E gene showed moderate inward and relatively larger outward currents under voltage stimulation ( Figure. 1, B and C). Using NaMES, K-gluconate and CsMES solution, we further revealed that E protein channels are permeable to monovalent cations Na + , K + and Cs + ( Figure. We characterized the E protein with reverse potentials of -5.36±1.36 mV (n=15) in Na + / K + solutions and -5.77±1.43 mV (n=15) in Na + / Cs + solutions ( Figure. 2 E). The relative permeabilities were calculated as P Na /P K ≈ 0.83 and P Na /P Cs ≈ 0.82 ( Figure. 2 F). The numbers did not exhibit significant differences ( Figure. 2, E and F). Previous studies demonstrated that the drug hexamethylene amiloride (HMA) blocked the channel activity of SARS-CoV E protein and inhibited virus replication (4, 5) . Here, we showed that HMA could also inhibit the conductance of the SARS-CoV-2 E protein. With 10 μM HMA application in the bath solution, the channel currents of SARS-CoV-2 E protein were largely reduced to the level of un-transfected CHO cells ( Figure. 3). We also demonstrated that the antiinfluenza A virus drug amantadine had no effect on SARS-CoV-2 E protein activity with a concentration of 26.6 μM (Figure. S1). Using homology modeling and the structure of SARS-CoV E protein (6), we built a structural showed that E protein channel currents were abolished in the mutants p.Glu8Ala, p.Glu8Lys, p.Asn15Ala, p.Asn45Ala, p.Ser50Ala and p.Tyr57Ala ( Figure. S2), suggesting that these amino acids were critical for the E protein channel activity; consistent with previous studies which showed that p.Asn15Ala mutant abolished SARS-CoV E protein channel activity (2, 7) . The cells transfected with mutants p.Leu28Ala, p.Leu51Ala produced robust channel currents compared to the un-transfected CHO cells. While wild-type E protein channel was inhibited by HMA ( Figure. 3), the currents of mutants p.Leu28Ala, p.Leu51Ala were not suppressed by HMA ( Figure. 4) . These results showed that Leu28 and Leu51 are critical amino acid mediating HMA inhibition. Drug repurposing uses de-risked compounds, thus potentially has lower overall costs and shorter timelines for development (8) . Thus, we further used the structural model of SARS-CoV-2 E protein and performed a virtual screening with listed drug library, clinical phase drug library and natural product library with total 5000 compounds (Supplementary table). We obtained 35 candidate compounds (Table 1 ) and tested their effects on SARS-CoV-2 E protein. We showed that AZD5153, a bromodomain protein 4 inhibitor against hematologic malignancies in clinical trial (9) , suppressed the currents of SARS-CoV-2 E protein at 40 μM concentration ( Figure. Our research demonstrated the ion channel activity of SARS-CoV-2 E protein and identified HMA and AZD5153 as its inhibitors. Our point mutation and recording experiments further identified that Leu28, Leu51 are key determinants of HMA inhibition and Phe23, Val29 are required for AZD5153 sensitivity. In Figure. 4, we showed that HMA did not inhibit the channel activity of mutants p.Leu28Ala, p.Leu51Ala. The two residues that are critical for E protein and HMA are dominated by non-polar effects. The results strongly supported that HMA binds to the upper pocket consisted of Leu28 in one subunit and Asn45, Ser50, Leu51, Tyr57 in adjacent subunit ( Figure.4 ). In Figure. S2, we showed that all three mutants of another predictive HMA binding pocket (p.Glu8Ala, p.Glu8Lys, p.Asn15Ala) induced E protein channel function loss. In previous studies, the mutation p.Asn15Ala was demonstrated to knock down the channel conductivity of SARS-CoV E protein (10) and further attenuate SARS-CoV virulence (2) . In the SARS-CoV E protein, the residues Asn15 and Glu8 were predicted to face the lumen of the channel pore (7, 10) . In our hypothetical model, HMA might interact with Asn15 and Glu8 ( Figure. 4) . Given the high homology between the SARS-CoV E protein and SARS-CoV-2 E protein, we speculated that HMA probably binds to a pocket that locates in the pore and thus blocks the channel. A previous study suggested two binding sites of HMA on the SARS-CoV E protein, one in the C-terminus near Arg38 and another in the N-terminus near Asn15. The authors presumed that HMA may bind to the channel by hydrogen bonding to the side-chain carbonyl of Asn15 and the guanidinium moiety of Arg38 (5) . The binding sites model near Asn15 in this report was similar to the binding pocket of HMA in SARS-CoV-2 E protein defined by Glu8, Glu8, Asn15 in adjacent subunits in our study. In Figure. 5, we showed that the residues Phe23 and Val29 were essential for AZD5153 sensitivity. In our predictive binding model, Phe23 and Val29 were two nonpolar amino acids located in two adjacent subunits, which might form non-polar interactions with AZD5153. The SARS-CoV-2 E protein pentamer is composed of five identical monomers. So this binding 6 pattern could be between the A and B subunits, or between the B and C subunits, or between the other two subunits. And it's possible that multiple compounds could simultaneously bind to the E protein pentamer. It was reported that the IC50 of AZD5153 for BRD4 is <10 nM (9) and therefore this compound is much less potent against the E protein and is unlikely to be an actual therapeutic candidate. It might serve as an interesting starting point for developing more potent inhibitors. A recent research showed that SARS-CoV-2 E protein could cause acute respiratory distress syndrome (ARDS) damages in lungs of mice and inhibitors of SARS-CoV-2 E protein could reduce the viral load in lungs of SARS-CoV-2-infected mice, which suggested that SARS-CoV-2 E protein could be a drug target (11) . In summary, first, our results suggest HMA and AZD5153 as lead compounds to develop other compounds to inhibit SARS-CoV-2 E protein. Second, the established functional essay of SARS-CoV-2 E protein by patch-clamp recording in cultured cells offers an opportunity for drug repurposing such as screening FDA approved drugs as potential drug candidates to treat COVID-19. With this functional essay, we discovered an active molecule against hematologic malignancies as the inhibitor of SARS-CoV-2 E protein. The SARS-CoV-2 envelope (E) protein gene (NCBI Reference Sequence: NC_045512.2) was synthesized and subcloned to a pcDNA3.1 vector with an IRES-GFP by GENEWIZ. The E SARS-CoV-2 protein gene variants were generated from the wild-type SARS-CoV-2 E protein gene by homologous recombination method and verified by automated sequencing. Chinese hamster ovary (CHO) cell line was cultured in F-12/DMEM media with 10% fetal bovine serum (FBS) and 1% antibiotic-antimycotic mixture (Invitrogen) at 37 °C with 5% CO2. Plasmid was transiently transfected into cells, at a total amount of 2000 ng per dish in 35 mm culture dishes, using Lipofectamine 3000 (Invitrogen) according to the manufacturer's 7 instructions. Whole-cell voltage-clamp recordings were performed using an Axopatch 700B amplifier Series resistance compensation (＞80%) was performed in the experiments. The whole-cell current traces of different groups were generated by voltage steps ranging from -100 to 100 mV with 10 mV increments from a holding potential of 0 mV. The currents at different voltage potentials were measured at peak levels to generate the current-voltage relationship. Analyses of data were performed with GraphPad Prism7. Pooled data are shown as means±SEM. Data were analyzed by unpaired t test or ANOVA. For determination of the relative permeability ratio of PX/PY for monovalent cations, we used the simplified Goldman-Hodgkin-Katz (GHK) equation: At 23ºC, RT/zF has the value of 25.5. Erev represents the reversal potential of the current. Reversal potential for each cell in different bi-ionic condition was calculated from interpolation of the relative current-voltage data. Reversal potentials and permeability ratios were presented as mean ± SEM. Modeling of SARS-CoV-2 E protein was carried out on SWISS-MODEL server. The protein sequence of SARS-CoV-2 E protein (YP_009724392.1) was obtained from NCBI database. PDB structure of SARS-CoV E protein (PDB ID: 5X29), which shares homology of 94.7% with SARS-CoV-2 E protein, was used as a template. Furthermore, a QMEAN score of -9.66 confirmed the reliability of the homology structural model of SARS-CoV-2 E protein. The SiteMap protocol was used to predict the possible binding pocket of SARS-CoV-2 E protein in Schrodinger. As a result, three possible binding pocket was identified. The pocket located within the possible central ion permeation path was chosen to perform the virtual screening. Structural optimization was conducted in Schrodinger with the default protocol of the Protein Preparation Wizard. Energy minimization was carried out using the OPLS-2005 force-field. Since the five monomers of SARS-CoV-2 E are identical in the pentameric model, amino acids Lipinski's Rule of Five was used to filter out the unqualified small molecules. All small compounds were prepared using the MMFFs force-field in the LigPrep module. High throughput virtual screening (HTVS) was carried out by Schrodinger. We have screened the L6000 Targetmol compound library against the SARS-CoV-2 E protein structure. Compounds which were screened successfully from HTVS were further subjected to SP (standard-precision) docking to get more accurate results. Furthermore, XP (extra precision) docking was used to remove the false-positive results. The top-ranked 283 molecules obtained from this virtual screening were then grouped into hierarchical clusters based on their chemical similarity calculated using Tanimoto similarity scores of binary fingerprints in the Canvas. From each cluster, the small compounds with the highest docking score were selected to retain, and finally 35 small compounds with different chemical structures were obtained. All data are contained within the manuscript. This article contains supporting information. Thanks Woo-Ping Ge, Jun Li and Weizhi Sun for discussion. Viroporins: structure and biological functions Severe acute respiratory syndrome coronavirus envelope protein ion channel activity promotes virus fitness and pathogenesis SARS coronavirus E protein forms cationselective ion channels Hexamethylene amiloride blocks E protein ion channels and inhibits coronavirus replication Structure and inhibition of the SARS coronavirus envelope protein ion channel Structural model of the SARS coronavirus E channel in LMPG micelles Conductance and amantadine binding of a pore formed by a lysine-flanked transmembrane domain of SARS coronavirus envelope protein Drug repurposing: progress, challenges and recommendations AZD5153: A Novel Bivalent BET Coronavirus E protein forms ion channels with functionally and structurallyinvolved membrane lipids Drosophila NOMPC is a mechanotransduction channel subunit for gentle-touch sensation The authors declare that they have no conflicts of interest with the contents of this article.