Drug Repurposing for Influenza Virus Polymerase Acidic (PA) Endonuclease Inhibitor

Drug repurposing can quickly and effectively identify novel drug repurposing opportunities. The PA endonuclease catalytic site has recently become regarded as an attractive target for the screening of anti-influenza drugs. PA N-terminal (PAN) inhibitor can inhibit the entire PA endonuclease activity. In this study, we screened the effectivity of PAN inhibitors from the FDA database through in silico methods and in vitro experiments. PAN and mutant PAN-I38T were chosen as virtual screening targets for overcoming drug resistance. Gel-based PA endonuclease analysis determined that the drug lifitegrast can effectively inhibit PAN and PAN-I38T, when the IC50 is 32.82 ± 1.34 μM and 26.81 ± 1.2 μM, respectively. Molecular docking calculation showed that lifitegrast interacted with the residues around PA or PA-I38 T’s active site, occupying the catalytic site pocket. Both PAN/PAN-I38T and lifitegrast can acquire good equilibrium in 100 ns molecular dynamic simulation. Because of these properties, lifitegrast, which can effectively inhibit PA endonuclease activity, was screened through in silico and in vitro research. This new research will be of significance in developing more effective and selective drugs for anti-influenza therapy.


Introduction
Influenza is an infectious disease that causes 300,000 to 500,000 human deaths globally per year [1,2]. The Influenza virus genome is composed of eight negative RNA segments encoded by multiple viral proteins, including PB2, PB1, PA, HA, NP, NA, M1, and NS1 [3,4]. RNA-dependent RNA polymerase (RdRp) is critical for virus RNA transcription and replication and comprises PA, PB1, and PB2 ( Figure 1). The N-terminal of PA domain (PA N ) contains an endonuclease active pocket and plays a crucial role in influenza virus polymerase activity. Drugs that abolish PA N endonuclease activity or disturb the assembly of RdRp can effectively inhibit the replication of the influenza virus [5]. The PA N endonuclease domain is highly conserved among different influenza virus subtypes, indicating that PA N is a promising broad-spectrum anti-influenza therapeutic target because of its ability to inhibit virus proliferation during the initial mRNA synthesis stage [6].
The three classes of FDA-approved anti-influenza drugs are: neuraminidase inhibitor (oseltamivir and zanamivir), M2-ion channel inhibitor (adamantanes), and PA N endonuclease inhibitor (Baloxavir acid). Several classes of PA N endonuclease inhibitors researched included: 4-substituted 2,4-dioxobutanoic acids and 3,4-substituted 2,6-diketopiperazines [7], flutamide and derivatives [8][9][10], N-hydroxamic acid-scaffold compounds [11], catechins [12], etc. Baloxavir acid, the inhibitor targeted on the PA N endonuclease, was approved by the FDA as an inhibitor of influenza A and B after proving effective in clinical trials [13]. 2 of 10 drugs that overcome resistance, we screened inhibitors that target both PAN and mutant PAN -I38T. Experimental tests have verified that the drugs lifitegrast and saquinavir have an inhibitory effect on PAN, while the drugs lifitegrast and conivaptan have an inhibitory effect on PAN-I38T. In addition, molecular docking shows the interaction mechanism between lifitegrast and the active site of PAN or PAN-I38T. Research suggests that lifitegrast may be a potential anti-flu drug. Finally, the method employed in this work could be utilized as a fast and viable strategy for accelerating research in the treatment of influenza.  A mutation in the influenza viral genome led to the influenza's subsequent drug resistance. The hydrophobic interaction residues between PA N endonuclease and Baloxavir acid were disturbed after the Ile 38 mutated to Thr [14]. The I38T substitution in the PA N endonuclease domain is the primary mutant that leads to influenza's resistance to Baloxavir acid. This resistance reduces the effectiveness of Baloxavir acid as an antiinfluenza drug [15]. Using the I38T mutant as the drug screening target may help to develop and refine the next-generation endonuclease inhibitors.
Drug repurposing research can effectively identify new drug repurposing opportunities, quickly expand the drug market, and reveal new commercially valuable uses for existing drugs [16]. Some examples of drugs that have been successfully repurposed include thalidomide [17], sildenafil [18], bupropion [19], and fluoxetine [20]. These drugs are currently used for applications beyond their initially approved therapeutic indications. The combination of in vitro and in silico methods will increasingly be used in the discovery of novel medicine [21].
This research screened the effectivity of PA N endonuclease inhibitors from the FDAapproved database through in silico methods and in vitro experiments. PA N inhibitors can inhibit the entire PA N endonuclease activity. Since the purpose of this study is to identify drugs that overcome resistance, we screened inhibitors that target both PA N and mutant PA N -I38T. Experimental tests have verified that the drugs lifitegrast and saquinavir have an inhibitory effect on PA N , while the drugs lifitegrast and conivaptan have an inhibitory effect on PA N -I38T. In addition, molecular docking shows the interaction mechanism between lifitegrast and the active site of PA N or PA N -I38T. Research suggests that lifitegrast may be a potential anti-flu drug. Finally, the method employed in this work could be utilized as a fast and viable strategy for accelerating research in the treatment of influenza.

Virtual Screening and Compounds Selection
To develop the new PA N endonuclease inhibitor structure, we performed the virtual screening protocol based on the PA N and mutant PA N -I38T structure. The residue Ile38 was not essential in either metal-ion binding or in catalytic activity. The I38 substitution does not block the endonuclease reaction but offers resistance to the inhibitor because the I38 side chain interacts with the Baloxavir acid. After the virtual screening, the compounds interacting with PA N or PA N -I38T were sorted by affinity score. The top-ranked compounds listed in Tables S1 and S2 were selected. Nine compounds closely interacted with PA N and PA N -I38T as hit compounds. Finally, six compounds, saquinavir, conivaptan, lifitegrast, rifaximin, dutasteride, and lurasidone, were used for further basic gel endonuclease determination. In comparison, three compounds, Teniposide [22], Simeprevir [23], and Nilotinib [24], were filtered according to levels of cytotoxicity.

The Inhibitory Effect of PA N or PA N -I38T with Compounds
We identified the five compounds that could inhibit the endonuclease activity of PA N or PA N -I38T, respectively. Next, we incubated different concentrations of compounds with PA N or PA N -I38T and then detected the cleavage of a substrate (Figures 2 and 3).

Virtual Screening and Compounds Selection
To develop the new PAN endonuclease inhibitor structure, we performed the virtual screening protocol based on the PAN and mutant PAN-I38T structure. The residue Ile38 was not essential in either metal-ion binding or in catalytic activity. The I38 substitution does not block the endonuclease reaction but offers resistance to the inhibitor because the I38 side chain interacts with the Baloxavir acid. After the virtual screening, the compounds interacting with PAN or PAN-I38T were sorted by affinity score. The top-ranked compounds listed in Tables S1 and S2 were selected. Nine compounds closely interacted with PAN and PAN-I38T as hit compounds. Finally, six compounds, saquinavir, conivaptan, lifitegrast, rifaximin, dutasteride, and lurasidone, were used for further basic gel endonuclease determination. In comparison, three compounds, Teniposide [22], Simeprevir [23], and Nilotinib [24], were filtered according to levels of cytotoxicity.

The Inhibitory Effect of PAN or PAN-I38T with Compounds
We identified the five compounds that could inhibit the endonuclease activity of PAN or PAN-I38T, respectively. Next, we incubated different concentrations of compounds with PAN or PAN-I38T and then detected the cleavage of a substrate (Figures 2 and 3).
The results showed that the compound saquinavir could effectively inhibit the cleavage of PAN substrate ( Figure 2D), but not PAN-I38T substrate. Meanwhile, compound conivaptan can effectively inhibit the cleavage of PAN-I38T substrate ( Figure 3D), but not PAN substrate. However, compounds conivaptan, dutasteride, rifaximin, and lurasidone showed no inhibition of the endonuclease activity of PAN ( Figure S1); compounds dutasteride, rifaximin, saquinavir, and lurasidone showed no inhibition of the endonuclease activity of PAN-I38T ( Figure S2). Finally, the compound lifitegrast was shown to effectively inhibit the substrate cleavage of both PAN and PAN-I38T at 32.82 ± 1.34 μM and 26.81 ± 1.2 μM, respectively (Table 1).

Retained Stability of Conformations of PAN/PAN-I38T and Lifitegrast during MD Simulations
To gain further insight into the binding mode of the system after reaching equilibrium, 100 ns MD simulations were performed using an explicit solvent. The initial confirmation of lifitegrast was obtained from the optimal pose of molecular docking operations in virtual screening. The RMSD values are an essential parameter in assessing the stability of a protein-ligand complex. As shown in Figure 4A, the structure of both proteins and ligands acquired good equilibrium during 100.0 ns. Accordingly, we can draw the same conclusion from the radius of the gyration of protein analysis ( Figure 4B).
Furthermore, to estimate the structural flexibility, the mean RMSF values were calculated. Figure 4C indicates that residues that interact with lifitegrast at the active pocket of protein remain stable. Although the residue Asn55-Leu71 in the Loop region is highly flexible, the overall protein conformation is stable ( Figure 4D). The results showed that the compound saquinavir could effectively inhibit the cleavage of PA N substrate ( Figure 2D), but not PA N -I38T substrate. Meanwhile, compound conivaptan can effectively inhibit the cleavage of PA N -I38T substrate ( Figure 3D), but not PA N substrate. However, compounds conivaptan, dutasteride, rifaximin, and lurasidone showed no inhibition of the endonuclease activity of PA N ( Figure S1); compounds dutasteride, rifaximin, saquinavir, and lurasidone showed no inhibition of the endonuclease activity of PA N -I38T ( Figure S2). Finally, the compound lifitegrast was shown to effectively inhibit the substrate cleavage of both PA N and PA N -I38T at 32.82 ± 1.34 µM and 26.81 ± 1.2 µM, respectively (Table 1).

Retained Stability of Conformations of PA N /PA N -I38T and Lifitegrast during MD Simulations
To gain further insight into the binding mode of the system after reaching equilibrium, 100 ns MD simulations were performed using an explicit solvent. The initial confirmation of lifitegrast was obtained from the optimal pose of molecular docking operations in virtual screening. The RMSD values are an essential parameter in assessing the stability of a protein-ligand complex. As shown in Figure 4A, the structure of both proteins and ligands acquired good equilibrium during 100.0 ns. Accordingly, we can draw the same conclusion from the radius of the gyration of protein analysis ( Figure 4B).

The Interaction Mechanism between PAN/PAN-I38T and Lifitegrast
Next, the protein-ligand complex structures in the 100 ns molecular simulation trajectories with similar conformations were divided into the same clusters. The representative frame of the largest cluster was extracted for analysis. Figure 5A,B show that lifitegrast binds to PAN/PAN-I38T at the structure active site and forms hydrogen bond interactions with lifitegrast ( Figure 5C,D). Table 2 gives the detailed non-bond parameters of lifitegrast and PAN/PAN-I38T. Comparing the number of hydrogen bonds between protein and lifitegrast, it can be concluded that lifitegrast has three hydrogen-bond interactions with residues Arg124 of PAN, and four hydrogen-bond interactions with residues Trp88, Thr123, and Arg125 of PAN-I38T.  Furthermore, to estimate the structural flexibility, the mean RMSF values were calculated. Figure 4C indicates that residues that interact with lifitegrast at the active pocket of protein remain stable. Although the residue Asn55-Leu71 in the Loop region is highly flexible, the overall protein conformation is stable ( Figure 4D).

The Interaction Mechanism between PA N /PA N -I38T and Lifitegrast
Next, the protein-ligand complex structures in the 100 ns molecular simulation trajectories with similar conformations were divided into the same clusters. The representative frame of the largest cluster was extracted for analysis. Figure 5A,B show that lifitegrast binds to PA N /PA N -I38T at the structure active site and forms hydrogen bond interactions with lifitegrast ( Figure 5C,D). Table 2 gives the detailed non-bond parameters of lifitegrast and PA N /PA N -I38T. Comparing the number of hydrogen bonds between protein and lifitegrast, it can be concluded that lifitegrast has three hydrogen-bond interactions with residues Arg124 of PA N , and four hydrogen-bond interactions with residues Trp88, Thr123, and Arg125 of PA N -I38T.

Discussion
Researchers have persistently searched for a drug that could be used to treat all types of influenza viruses, including influenza A, influenza B, and influenza C, along with their mutants. In this study, we used PAN as the influenza drug virtual screening target because the PA endonuclease domain is highly conserved among different influenza virus subtypes. The I38T substitution in the PAN endonuclease domain is the primary mutant that leads to resistance to Baloxavir acid (PAN endonuclease inhibitor). Resistance develops when the hydrophobic interaction residues between PAN endonuclease and Baloxavir acid are disturbed after the Ile 38 mutates to Thr. Therefore, we used the mutant type of PAN-I38T as the virtual screening target to avoid drug resistance. We used the FDA database to effectively identify drug repurposing opportunities. Then, we discovered a potent inhibitor, lifitegrast, which differs from Baloxavir acid in that it has the ability to interact equally with both PAN and PAN-I38T.
Gel-based endonuclease inhibitory assay showed that lifitegrast effectively inhibits the substrate cleavage of both PAN and PAN-I38T. Molecular simulation displayed that lifitegrast interacted with the active site of PAN or PAN-I38T. Further molecular dynamic simulation analysis suggested that the bindings between PAN/PAN-I38T and lifitegrast were stable.
However, there are still numerous challenges. Subsequent viral experiments in vivo need to be implemented in order to verify the reliability of experiments in vitro. Meanwhile, lifitegrast has been used for the treatment of dry eye, but further toxic experimentation is required before lifitegrast can be used as a treatment for the influenza virus.
In this research, we screened the effectivity of the PAN endonuclease inhibitor from the FDA database through in silico and in vitro experiments. The findings suggest that lifitegrast may be a potential anti-flu treatment drug because it could solve the drug-resistant properties, which are superior to currently available flu treatments, such as Baloxavir acid. Finally, the method adopted in this work could be utilized as a fast and viable strategy for accelerating research on the treatment of influenza.

Discussion
Researchers have persistently searched for a drug that could be used to treat all types of influenza viruses, including influenza A, influenza B, and influenza C, along with their mutants. In this study, we used PA N as the influenza drug virtual screening target because the PA endonuclease domain is highly conserved among different influenza virus subtypes. The I38T substitution in the PA N endonuclease domain is the primary mutant that leads to resistance to Baloxavir acid (PA N endonuclease inhibitor). Resistance develops when the hydrophobic interaction residues between PA N endonuclease and Baloxavir acid are disturbed after the Ile 38 mutates to Thr. Therefore, we used the mutant type of PA N -I38T as the virtual screening target to avoid drug resistance. We used the FDA database to effectively identify drug repurposing opportunities. Then, we discovered a potent inhibitor, lifitegrast, which differs from Baloxavir acid in that it has the ability to interact equally with both PA N and PA N -I38T.
Gel-based endonuclease inhibitory assay showed that lifitegrast effectively inhibits the substrate cleavage of both PA N and PA N -I38T. Molecular simulation displayed that lifitegrast interacted with the active site of PA N or PA N -I38T. Further molecular dynamic simulation analysis suggested that the bindings between PA N /PA N -I38T and lifitegrast were stable.
However, there are still numerous challenges. Subsequent viral experiments in vivo need to be implemented in order to verify the reliability of experiments in vitro. Meanwhile, lifitegrast has been used for the treatment of dry eye, but further toxic experimentation is required before lifitegrast can be used as a treatment for the influenza virus.
In this research, we screened the effectivity of the PA N endonuclease inhibitor from the FDA database through in silico and in vitro experiments. The findings suggest that lifitegrast may be a potential anti-flu treatment drug because it could solve the drugresistant properties, which are superior to currently available flu treatments, such as Baloxavir acid. Finally, the method adopted in this work could be utilized as a fast and viable strategy for accelerating research on the treatment of influenza.

Virtual Screening
The PA N endonuclease protein crystal structure (PDB ID: 6FS6) and I38T mutant (PDB ID: 6FS7) in the PA N subunit were obtained from the RCSB Protein Data Bank database [25]. One thousand four hundred and ten compounds in mol2 format were downloaded from the FDA-approved drug database (updated in February 2018) for screening. The Raccoon program was used to convert the ligand from the mol2 format to the PDBQT format, which can be recognized by the virtual screening software [26]. Virtual screening was performed by the AutoDock Vina program [27], with the parameter values of x, y, z center set to 18.26, 95.54, 44.58, and the parameter values of the grid map in x, y, z-dimension set to 25 Å × 25 Å × 25 Å. Results of virtual screening were listed according to the binding energy score as shown in the supporting information.

Expression and Purified of PA N and PA N -I38T
The PA N endonuclease domain from influenza A type virus (H1N1) (PA N , residues 1-197) was cloned into the pET-28a vector, expressed in E. coli BL21 (DE3), and purified by IMAC. The positive clone was used for large-scale expression. Cells were expressed in LB medium at 37 • C for 4 h with 1 mM IPTG. Before the endonuclease activity assay, the protein sample was concentrated, and the buffer changed (10 mM Tris-HCl, 100 mM NaCl, and 2 mM MnCl 2 ) by ultrafiltration. The protein concentration was measured by the BCA method. The PA N -I38T was cloned from the PA N through site-directed mutagenesis. The mutant PA N -I38T was expressed and purified the same as the wild PA N .

Gel-Based Endonuclease Inhibitory Assay
The PA N endonuclease activity was detected through the digestion of a single-strand DNA substrate. In the digestion reaction, PA N or PA N -I38T was mixed with ssDNA in a digestion buffer that included 2 mM MnCl 2 . To determine the concentrate of PA N or PA N -I38T for endonuclease activity assay, 0-1.5 µM protein and 100 ng ssDNA were incubated at 37 • C for 60 min. After digestion, the samples were fractioned by 0.8% agarose gel and visible by a stain. Image J was used to determine the density of substrate ssDNA. Finally, 1.5 µM PA N or PA N -I38T were digested at 100 ng ssDNA in 10 µL volume for 60 min.
Compounds screened from the FDA library analysis determined the inhibition of endonuclease activity. In total, 1.5 µM PA N or PA N -I38T and 100 ng ssDNA mixed with different concentrations of compounds were used for the experiment test. Samples without inhibitors were used as the no-compound control. Samples without protein were used as the substrate control. After digestion, the samples were loaded on the agarose gel for separation and were visible by a stain. The amount of remaining ssDNA was determined by image J. The IC 50 value was calculated by GraphPad Prism 6.0 through non-Linear regression.

Molecular Dynamic (MD) Simulation
The MD simulation of PA N and PA N -I38T with or without compounds was performed using GROMACS version 2018 with the CHARMM36 all-atom force field (28 March 2019) to investigate the inhibition mechanism [28]. The initial confirmation of lifitegrast was obtained from the optimal pose of the virtual screening results. The CGenFF server was used for the topology generation of the compound [29,30]. The protein was centered in separate cubic boxes and solvated using the SPC216 water model [31]. Two additional NA ions were added to the system to automatically achieve electron neutrality because the protein had a total charge of −2.000 e. The structure was relaxed through energy minimization to ensure a reasonable starting structure in terms of geometry and solvent orientation. Convergence was achieved at a maximum force of less than 1000 kJ/mol·nm in any atom. Equilibration was conducted in two phases, i.e., the NVT (constant number of particles, volume, and temperature) and the NPT (constant number of particles, pressure, and temperature) ensembles, for 100 ps until the system was equilibrated. A total of 100 ns simulations were performed after the equilibration. The root mean square deviation (RMSD), root mean square fluctuations (RMSF), and gyration radius were calculated using the tools available in the GROMACS package. According to the GROMOS method [32], RMSD is used as a metric, and the cluster analysis of molecular simulation trajectories is divided into different groups by the GROMACS package. For each system, the cutoff value of RMSD was set to 2.5 Å. The protein trajectory was displayed, animated, and analyzed through the PyMOL visualization program [33][34][35].

Statistical Assay
The IC 50 was calculated by GraphPad Prism 6.0 using non-linear regression. The data were represented as mean ± SD.