Design, Synthesis, Molecular Docking Analysis and Biological Evaluations of 4-[(Quinolin-4-yl)amino]benzamide Derivatives as Novel Anti-Influenza Virus Agents

In this study, a series of 4-[(quinolin-4-yl)amino]benzamide derivatives as the novel anti-influenza agents were designed and synthesized. Cytotoxicity assay, cytopathic effect assay and plaque inhibition assay were performed to evaluate the anti-influenza virus A/WSN/33 (H1N1) activity of the target compounds. The target compound G07 demonstrated significant anti-influenza virus A/WSN/33 (H1N1) activity both in cytopathic effect assay (EC50 = 11.38 ± 1.89 µM) and plaque inhibition assay (IC50 = 0.23 ± 0.15 µM). G07 also exhibited significant anti-influenza virus activities against other three different influenza virus strains A/PR/8 (H1N1), A/HK/68 (H3N2) and influenza B virus. According to the result of ribonucleoprotein reconstitution assay, G07 could interact well with ribonucleoprotein with an inhibition rate of 80.65% at 100 µM. Furthermore, G07 exhibited significant activity target PA−PB1 subunit of RNA polymerase according to the PA−PB1 inhibitory activity prediction by the best pharmacophore Hypo1. In addition, G07 was well drug-likeness based on the results of Lipinski’s rule and ADMET prediction. All the results proved that 4-[(quinolin-4-yl)amino]benzamide derivatives could generate potential candidates in discovery of anti-influenza virus agents.


Introduction
Influenza virus remains to be a grand challenge to public health due to the high incidence and mortality. It belongs to pathogenic microorganisms causing acute respiratory diseases [1]. The course of the influenza mainly depends on the different flu strains in addition to the host immune response. Usually, influenza viruses are classified into four types (A, B, C and D) according to their nucleoprotein (NP) and matrix protein 1 (M1). Among the four types of influenza viruses, influenza A virus was further divided into many kinds of subtypes based on their surface glycoproteins, hemagglutinin (HA) and neuraminidase (NA). The influenza A virus infected a wide range of avian and mammalian hosts, while influenza B virus infected only humans [2]. These differences increased the severity of the disease [3,4].
Currently, influenza vaccination was the preferred method for prophylaxis of influenza. However, the initial supply of the vaccines might not be sufficient to meet the demand for them [5,6] due to the rapid mutations of the virus genes and the constraints

Biological Activities of All the Molecules
Each compound was performed to cytotoxicity assay to determine their CC50 values which was responsible for 50% reduction in cell viability (CC50). The results of the 26 target molecules in cytotoxicity study implied that no effect had emerged on Madin-Darby canine kidney (MDCK) cells until 100 µM except for molecules G16, G20 and G26 (Table  1). 26 target molecules were carried out CPE assay to evaluated their cytopathic inhibitory effect of MDCK cells infected by influenza A virus strain (A/WSN/33, H1N1). The results were expressed as the 50% effective concentration (EC50) which was defined as the compound concentration required to protect 50% of the MDCK cells against viral infection. Cytopathic cells were fewer in some target molecule-treated infected cells according to the CPE results. Subsequently, they were subjected to plaque inhibition screening assay to discover potential anti-influenza virus inhibitors (Tables 1 and 2). Furthermore, some compounds were fulfilled to RNP reconstitution assay to assess whether these compounds could interect with RNP of influenza virus. The structures of target compounds were confirmed by spectroscopic studies (MS, HRMS, 1 H NMR and 13 C NMR). The specfic spectrogram are available in the the Supplementary Information (Figures S9-S126).

Biological Activities of All the Molecules
Each compound was performed to cytotoxicity assay to determine their CC 50 values which was responsible for 50% reduction in cell viability (CC 50 ). The results of the 26 target molecules in cytotoxicity study implied that no effect had emerged on Madin-Darby canine kidney (MDCK) cells until 100 µM except for molecules G16, G20 and G26 (Table 1). 26 target molecules were carried out CPE assay to evaluated their cytopathic inhibitory effect of MDCK cells infected by influenza A virus strain (A/WSN/33, H1N1). The results were expressed as the 50% effective concentration (EC 50 ) which was defined as the compound concentration required to protect 50% of the MDCK cells against viral infection. Cytopathic cells were fewer in some target molecule-treated infected cells according to the CPE results. Subsequently, they were subjected to plaque inhibition screening assay to discover potential anti-influenza virus inhibitors (Tables 1 and 2). Furthermore, some compounds were fulfilled to RNP reconstitution assay to assess whether these compounds could interect with RNP of influenza virus.

Structure-Activity Relationships
The cytotoxicities of all target compounds (G01-G26) were tested. The cytopathic inhibitory effects of these target compounds were evaluated by CPE assay using influenza virus-infected MDCK cells (Tables 1 and S1-S4). For the target compounds (G01-G18), the CC 50 values of all these compounds were more than 100 µM as enlisted in Table 1.  (Table 1), respectively. In generally, G07 demonstrated significant anti-influenza virus activity both in CPE assay and plaque inhibition assay. Based on the above, retention of the substituent group (-NR 2 R 3 ) as 2-(2-methoxyphenoxy)ethan-1-amino and modification of the substituent group of quinoline ring resulted in target compounds N-[2-(2-methoxyphenoxy)ethyl]-4-(quinolin-4ylamino)benzamide derivatives (G19-G26). G19 and G23 exhibited significant anti-influenza virus activity (EC 50 values of G19 and G23 were less than 10 µM and inhibition rates of G19 and G23 at 100 µM were 82.30% and 77.88%, respectively), which implied that the positions of substituent groups in quinoline ring were essential for the anti-influenza virus activity. Unsubstituted group (G19) or the nitro group (G23) at 7-position in quinoline ring contributed a lot to the increase of anti-influenza virus activities. In addition, the relative contributions to anti-influenza virus activity of the substituent groups at 7-position in the scaffold were trifluoromethyl (95.91%) > nitro (77.88%) > methyl (11.50%). However, none of the target compounds (G19-G26) exhibited better anti-influenza virus activities than G07.
Residues 618−621, GLU623, VAL636, LEU640, LEU666, GLN670, ARG673, TRP706, and PHE710 of influenza A virus PA protein were several reported residues on the PA−PB1 interface [23,24]. Amino sequence alignment was used to analyze the full sequences of influenza A virus PA protein and influenza B virus PA protein. The specific report was shown in the Supplementary Information (Figure S1). Amino sequence alignment of influenza A virus PA protein (residues 614−716 of H1N1, H3N2 and H5N1 subtypes) and influenza B virus PA protein (residues 610−712) have the greatest similarity. As shown in Figure 3, amino sequence alignment of influenza A virus PA protein (residues 614−716 of H1N1, H3N2 and H5N1 subtypes) and influenza B virus PA protein (residues 610−712) are highly conserved. So, we hypothesized that G07 had antiviral potency against multiple types of influenza strains. Plaque inhibition assaies were conducted to test the antiviral activities of G07. The IC 50 value of G07 against influenza virus A/WSN/33 (H1N1) was 0.23 ± 0.15 µM (Table 2). G07 exhibited better anti-influenza virus activity than the control inhibitor amantadine (IC 50 > 100 µM) targeting influenza viral strain A/WSN/33 (H1N1) based on the plaque inhibition assay. In addition, G07 exhibited significanted anti-influenza virus activities against A/PR/8 (H1N1) and A/HK/68 (H3N1) with IC 50 values of 11.37 ± 2.38 and 7.51 ± 1.76 µM (Table 2), respectively. G07 also demonstrated significanted anti-influenza virus activities against influenza B virus (IC 50 = 10.99 ± 1.16 µM, Table 2). Obviously, G07 exhibited better anti-influenza virus activities than the control compound amantadine against these three influenza virus strains A/WSN/33 (H1N1), A/PR/8 (H1N1) and influenza B virus according to the plaque inhibition assay. The IC 50 value of amantadine against influenza virus A/HK/68 (H3N2) was 2.22 ± 0.57 µM (Table 2), which demonstrated stronger anti-influenza virus activity than G07 against influenza virus A/HK/68 (H3N2). According to the RNP reconstitution assay results, G07 could interect well with RNP of influenza virus with an inhibition rate of 80.65% at 100 µM.

PA−PB1 Inhibitory Activity Prediction by the Best Pharmacophore Hypo1
The generated pharmacophoric model Hypo1 was used to forecasted the activity of the 26 target molecules against PA−PB1 of RNA endonuclease. The results of the cost analysis, Fischer's randomization test and test set analysis confirm that the best pharmacophore model (Hypo1) could be selected as a significant pharmacophore model [2,[25][26][27][28][29][30][31][32][33][34][35] to further predicting the biological activities of the diverse structural entity (Figures S2-S8, Tables S5-S7). The estimated values (IC 50 ) of 26 target molecules were enlisted in Table 1. G06, G07 and G22 exhibited significant inhibitory activities against PA−PB1 subunit of the viral RNA with estimated values of 42.842, 35.4716 and 4.91662 µM, respectively. Four features (hydrogen bond acceptor (HBA), hydrophobic (HYD) and ring aromatic (RA)) were shown in the best pharmacophoric model Hypo1. However, the specific biological activities of these molecules targeting the PA−PB1 of RNA endonuclease need further experimental studies.
A typical compound G07 and a training set compound (H01) [25] with significant antiinfluenza virus activities were selected to analyze the pharmacophore mapping between compounds and Hypo1 ( Figure 4). The benzene ring of 2-(2-methoxyphenoxy)ethan-1-amino group in G07 mapped well with the RA feature. The methyl group in 2-(2methoxyphenoxy)ethan-1-amino group and the benzene ring of 4-aminobenzamide in G07 mapped well with the HYD features, respectively. Oxygen of the amide group in benzamide represented as the hydrogen acceptor and mapped well with the HBA feature. All these pharmacophore mapping results of G07 with Hypo1 were similar to that of H01 with Hypo1. H01 was by far a significant anti-influenza virus agent with the best anti-influenza virus activities against PA−PB1 subunit of RNA endonuclease. The estimated activity value of H01 was 0.982586 µM and experimental value was 1.1 µM. The estimated activity value of G07 was 35.4716 µM. Although G07 exhibited lower activity against PA−PB1 subunit of RNA endonuclease compared with H01, it still was of great advantage compared with other inhibitors targeting PA−PB1 subunit. As a corollary, G07 can strongly interact with PA−PB1 subunit of RNA endonuclease.

Lipinski's Rule and ADMET Prediction
In this study, the ADMET descriptors algorithm and toxicity prediction (extensible) module of Discovery Studio 3.0 were used to calculate the Lipinski's rule-of-five druglikeness properties for oral bioavailability and ADMET properties [36].
The results of Lipinski's rule calculation for G01-G26 included Alop, molecular weight (MW), number of hydrogen-bond acceptors, number of hydrogen-bond donors and number of rotatable bonds (Table S8). It was noticeable that all compounds were in line with the Lipinski's rule, except for G17 and G18 (the MW of G17 and G18 were more than 500) [37,38].
The ADMET prediction results (including ADME Solubility Level, ADME BBB Level, ADME Absorption Level, Hepatoxic, PPB Prediction and Toxicity Probability) were enlisted in Table S9. It was worth noting that G07 exhibited good absorption and low toxicity based on the results of ADME absorption level and toxicity probability [36].
In general, G07 was of well drug-likeness according to the results of Lipinski's rule and ADMET prediction.

Lipinski's Rule and ADMET Prediction
In this study, the ADMET descriptors algorithm and toxicity prediction (extensible) module of Discovery Studio 3.0 were used to calculate the Lipinski's rule-of-five druglikeness properties for oral bioavailability and ADMET properties [36].
The results of Lipinski's rule calculation for G01-G26 included Alop, molecular weight (MW), number of hydrogen-bond acceptors, number of hydrogen-bond donors and number of rotatable bonds (Table S8). It was noticeable that all compounds were in line with the Lipinski's rule, except for G17 and G18 (the MW of G17 and G18 were more than 500) [37,38].
The ADMET prediction results (including ADME Solubility Level, ADME BBB Level, ADME Absorption Level, Hepatoxic, PPB Prediction and Toxicity Probability) were enlisted in Table S9. It was worth noting that G07 exhibited good absorption and low toxicity based on the results of ADME absorption level and toxicity probability [36].
In general, G07 was of well drug-likeness according to the results of Lipinski's rule and ADMET prediction.

Molecular Dynamics Simulations
The 100 ns molecular dynamics (MD) simulations were performed with three selected complexes (G07−3CM8, G19−3CM8 and G23−3CM8) to examine their interactions between proteins and ligands. In addition, the contribution of key residues was elucidated during the binding process. The RMSD and RMSF values of three complexes during the whole MD simulations were enlisted in Figures 9 and 10. In Figure  9, the G07-3CM8 complex reached equilibrium after 30 ns and the protein RMSD value fluctuated around 5.5-6.0 Å. The G23−3CM8 complex reached equilibrium after 20 ns and

Molecular Dynamics Simulations
The 100 ns molecular dynamics (MD) simulations were performed with three selected complexes (G07−3CM8, G19−3CM8 and G23−3CM8) to examine their interactions between proteins and ligands. In addition, the contribution of key residues was elucidated during the binding process. The RMSD and RMSF values of three complexes during the whole MD simulations were enlisted in Figures 9 and 10. In Figure  9, the G07-3CM8 complex reached equilibrium after 30 ns and the protein RMSD value fluctuated around 5.5-6.0 Å. The G23−3CM8 complex reached equilibrium after 20 ns and Generally, 4-amino-N-[2-(2-methoxyphenoxy)ethyl]benzamide fragment contribu-ted a lot to the interaction between the active site and the target compounds. GLU623, LYS643 and TRP706 are three key amino acid residues interacted with ligands based on the docking results. We can speculate that these small molecules exhibited anti-influenza virus activities after blocking the PA−PB1 interface by competing with the PB1.

Molecular Dynamics Simulations
The 100 ns molecular dynamics (MD) simulations were performed with three selected complexes (G07-3CM8, G19-3CM8 and G23-3CM8) to examine their interactions between proteins and ligands. In addition, the contribution of key residues was elucidated during the binding process. The RMSD and RMSF values of three complexes during the whole MD simulations were enlisted in Figures 9 and 10. In Figure 9, the G07-3CM8 complex reached equilibrium after 30 ns and the protein RMSD value fluctuated around 5.5-6.0 Å. The G23-3CM8 complex reached equilibrium after 20 ns and the protein RMSD value fluctuated around 3.5-4.5 Å. However, G19-3CM8 showed obvious fluctuations and reached equilibrium after 80 ns, indicating that G19 had poor binding affinity with 3CM8 compared with G07 and G23. The RMSF plots were displayed in Figure 10. Majority fluctuations of residues were less than 2.5 Å. The states of protein secondary structures were monitored during the whole MD simulations (Figure 11). Obviously, the huge fluctuations in RMSF curves of G07-3CM8, G19-3CM8 and G23-3CM8 located in the non-alpha-helices and nonbeta-strands according to the protein secondary structure elements ( Figure 11). Among all of fluctuations in RMSF curves, the huge fluctuations of G07-3CM8 and G19-3CM8 exhibited from MET374 to PRO398, which was relatively small fluctuation of G23-3CM8 due to the generation of alpha-helices.
Int. J. Mol. Sci. 2022, 23, 6307 13 of 29 obvious fluctuations and reached equilibrium after 80 ns, indicating that G19 had poor binding affinity with 3CM8 compared with G07 and G23. The RMSF plots were displayed in Figure 10. Majority fluctuations of residues were less than 2.5 Å. The states of protein secondary structures were monitored during the whole MD simulations ( Figure 11). Obviously, the huge fluctuations in RMSF curves of G07−3CM8, G19−3CM8 and G23−3CM8 located in the non-alpha-helices and non-beta-strands according to the protein secondary structure elements ( Figure 11). Among all of fluctuations in RMSF curves, the huge fluctuations of G07−3CM8 and G19−3CM8 exhibited from MET374 to PRO398, which was relatively small fluctuation of G23−3CM8 due to the generation of alpha-helices.   obvious fluctuations and reached equilibrium after 80 ns, indicating that G19 had poor binding affinity with 3CM8 compared with G07 and G23. The RMSF plots were displayed in Figure 10. Majority fluctuations of residues were less than 2.5 Å. The states of protein secondary structures were monitored during the whole MD simulations ( Figure 11). Obviously, the huge fluctuations in RMSF curves of G07−3CM8, G19−3CM8 and G23−3CM8 located in the non-alpha-helices and non-beta-strands according to the protein secondary structure elements ( Figure 11). Among all of fluctuations in RMSF curves, the huge fluctuations of G07−3CM8 and G19−3CM8 exhibited from MET374 to PRO398, which was relatively small fluctuation of G23−3CM8 due to the generation of alpha-helices.     Figure 12 listed the type and ratio of interactions between proteins and ligands, including hydrogen bonds, hydrophobic contacts, ionic interactions, and water bridges. In Figure 12a, G07 formed hydrogen bonds with residues GLU623, LYS643, TRP706, SER709, which accounted for 31.8%, 8.5%, 10.3% and 22.2% of the simulation time, respectively. G07 generated water bridges with residues ASN412, GLU623 and LYS643. The hydrophobic interactions were generated between G07 and residues PRO625, LYS643, LEU666, TRP706, PHE707 and PHE710. In Figure 12b, G19 formed hydrogen bonds with residues ILE621, LYS643 and SER709, and generated water bridges with residues GLU623, SER624, VAL669, TRP706 and SER709. Among them, G19 formed strong hydrogen bonds with GLU623, which accounted for 18.7% of the simulation time. In addition, water bridges were also significant for the interactions, such as G19 formed water bridges with GLU623, SER624, VAL669, and TRP706, which accounted for 34.3%, 18.6%, 31.2% and 22.6% of the simulation time, respectively. The hydrophobic interactions were generated between G19 and residues PRO625, LYS643, LEU666, ARG673, TRP706, PHE707, PHE710 and HIS713. In Figure 12c, G23 formed hydrogen bonds with residues ASN412, GLU623, LYS643 and TRP706, which accounted for 6.6%, 34.7%, 7.2% and 10.6% of the simulation time, respectively. G23 generated water bridges with residues GLN408, GLU623 and LYS643. Among them, G23 formed strong water bridge interactions with GLU623 and LYS643, which accounted for 43.1% and 55.7% of the simulation time. The hydrophobic interactions were generated between G23 and residues PRO625, LYS643, LEU666, TRP706, PHE707, PHE710 and HIS713.

Alanine Scanning Mutagenesis Analysis
Alanine scanning mutagenesis (ASM) analysis was used to assess the role of the specific amino acid residue participating in protein-protein and protein-ligand interactions. It was usually based on the hypothesis that the main chain conformation did not modify and the side chains beyond the β-carbon for ligand-protein complexes were reduced after substituting an amino acid residue into alanine. As reported, ASM analysis has been regarded as an appealing alternative to in vitro experiments [45,46]. In this study, ASM analysis was used to validate the binding free energy decomposition analysis. The results of 100ns MD simulations indicated that G07-3CM8 and G23-3CM8 were more stable during MD simulations, which revealed that G07 and G23 might have stronger interactions with 3CM8. Generally, GLU623, LYS643, TRP706, PHE707 and PHE710 were key amino acid residues interacted with the ligands, which was similar with the previous reports [2,44].

Alanine Scanning Mutagenesis Analysis
Alanine scanning mutagenesis (ASM) analysis was used to assess the role of the specific amino acid residue participating in protein-protein and protein-ligand interactions. It was usually based on the hypothesis that the main chain conformation did not modify and the side chains beyond the β-carbon for ligand-protein complexes were reduced after substituting an amino acid residue into alanine. As reported, ASM analysis has been regarded as an appealing alternative to in vitro experiments [45,46]. In this study, ASM analysis was used to validate the binding free energy decomposition analysis.
Residues showing high interaction fraction from MD simulations were selected to mutate to alanine ( Figure 13). Positive ∆∆G values indicate that the wild-type residues could produce more favorable interactions with ligands than the mutated residues. Mutations like GLU623, PRO625, LYS643 and SER709 in complex G07-3CM8, mutations like GLU623 and LYS643 in complex G19-3CM8, and mutations like GLU623 and PRO625 in complex G23-3CM8 did not influence their binding behavior, indicating these residues interacted with ligands mainly through the scaffold atoms while the side chains did not contribute much to binding energy. The binding free energy changes caused by residue mutations in complex G07-3CM8 (residue mutations ASN412, TRP706, PHE707 and PHE710), complex G19-3CM8 (residue mutations SER624, LEU666, VAL669, ARG673, TRP706, PHE707, PHE710 and HIS713), as well as complex G23-3CM8(residue mutations TRP706, PHE707, PHE710 and HIS713) emphasized the importance of these unmutated residues. This reasonable result indicated that the mutations of the above residues into alanine shorten the length of residue side chains, and therefore decreased the interaction opportunity of residue with ligands. Interestingly, TRP706, PHE707 and PHE710 showed unexpected remarkable energy changes. It will be helpful to design more significant anti-influenza virus inhibitors through forming interactions with the side chain of TRP706, PHE707 and PHE710. Residues showing high interaction fraction from MD simulations were selected to mutate to alanine ( Figure 13). Positive ΔΔG values indicate that the wild-type residues could produce more favorable interactions with ligands than the mutated residues. Mutations like GLU623, PRO625, LYS643 and SER709 in complex G07−3CM8, mutations like GLU623 and LYS643 in complex G19−3CM8, and mutations like GLU623 and PRO625 in complex G23−3CM8 did not influence their binding behavior, indicating these residues interacted with ligands mainly through the scaffold atoms while the side chains did not contribute much to binding energy. The binding free energy changes caused by residue mutations in complex G07−3CM8 (residue mutations ASN412, TRP706, PHE707 and PHE710), complex G19−3CM8 (residue mutations SER624, LEU666, VAL669, ARG673, TRP706, PHE707, PHE710 and HIS713), as well as complex G23−3CM8(residue mutations TRP706, PHE707, PHE710 and HIS713) emphasized the importance of these unmutated residues. This reasonable result indicated that the mutations of the above residues into alanine shorten the length of residue side chains, and therefore decreased the interaction opportunity of residue with ligands. Interestingly, TRP706, PHE707 and PHE710 showed unexpected remarkable energy changes. It will be helpful to design more significant antiinfluenza virus inhibitors through forming interactions with the side chain of TRP706, PHE707 and PHE710.

Synthesis
Unless otherwise required, all solvents and reagents were purchased from commercial sources and were used without further purification. All chemical reactions were monitored through GF254 thin layer chromatography (TLC) plate and spots were visualized by UV light (254 nm). The structures of the target compounds were characterized by 1 H NMR spectra and 13 C NMR spectra on a Bruker 400 MHz or 101 MHz NMR spectrometer (Faellanden, Switzerland) with TMS as an internal standard and DMSO-d6 or CDCl3 as the solvent, chemical shifts (d values) and coupling constants (J values) are respectively given in ppm and Hz. The melting points were determined on a Buchi B-540 melting-point apparatus with a microscope, and were uncorrected. The IR spectra were recorded with KBr pellets on a Bruker IFS55 spectrometer (Faellanden, Switzerland). High-resolution mass (HRMS) spectral were performed on an Agilent Technologies 6530 Accurate-Mass Q-TOF Mass Spectrometer (Santa Clara, CA, USA). Compounds A (A01-A09), B (B01-B09), C (C01-C09), D (D01-D09) and E (E01-E09) [19,21,[47][48][49][50][51] were previously prepared. 4-Chloroquinoline (E01-E09, 0.01 mol), 4-aminobenzamide (0.011 mol), 20 mL ethanol and a catalytic amount (4 drops) of 37% hydrochloric acid were refluxed for 2 h. The reaction was allowed to cool to room temperature, and the precipitated was filtered off, washed with water (3 × 5 mL), and recrystallized by methanol [48]

Synthesis
Unless otherwise required, all solvents and reagents were purchased from commercial sources and were used without further purification. All chemical reactions were monitored through GF 254 thin layer chromatography (TLC) plate and spots were visualized by UV light (254 nm). The structures of the target compounds were characterized by 1 H NMR spectra and 13 C NMR spectra on a Bruker 400 MHz or 101 MHz NMR spectrometer (Faellanden, Switzerland) with TMS as an internal standard and DMSO-d 6 or CDCl 3 as the solvent, chemical shifts (d values) and coupling constants (J values) are respectively given in ppm and Hz. The melting points were determined on a Buchi B-540 melting-point apparatus with a microscope, and were uncorrected. The IR spectra were recorded with KBr pellets on a Bruker IFS55 spectrometer (Faellanden, Switzerland). High-resolution mass (HRMS) spectral were performed on an Agilent Technologies 6530 Accurate-Mass Q-TOF Mass Spectrometer (Santa Clara, CA, USA). Compounds A (A01-A09), B (B01-B09), C (C01-C09), D (D01-D09) and E (E01-E09) [19,21,[47][48][49][50][51] were previously prepared. 4-Chloroquinoline (E01-E09, 0.01 mol), 4-aminobenzamide (0.011 mol), 20 mL ethanol and a catalytic amount (4 drops) of 37% hydrochloric acid were refluxed for 2 h. The reaction was allowed to cool to room temperature, and the precipitated was filtered off, washed with water (3 × 5 mL), and recrystallized by methanol [48]  The MDCK cells and HEK293T cells were routinely cultured in minimum essential medium (MEM, Gibco) supplemented with 10% (v/v) fetal bovine serum (FBS, Gibco) at 37 • C in a humidified 5% CO 2 incubator. The 26 compounds which were dissolved in dimethyl sulfoxide (DMSO) during the cytotoxicity assay, CPE assay, anti-influenza virus assay and RNP reconstitution assay.

Cytotoxicity Assay
The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) method was performed to assess the cytotoxicity of target compounds in MDCK cells and HEK293T cells [27,52,53]. Firstly, MDCK cells or HEK293T cells were seeded into the 96-well plate. Then, the media containing test compounds replaced the growth media after 48 h. Incubate MDCK cells or HEK293T cells with test compounds for 72 h at 37 • C in a humidified 5% CO 2 incubator. Later, MTT solution (5 mg/mL in PBS) was added into each well and plates were incubated for 4 h at 37 • C. Then, add a solubilization solution to lyse cells. Finally, absorbance was read at 620 nm using an ELISA plate reader (Tecan Sunrise) after 3 h of further incubation at 37 • C. We set the values obtained from the wells treated with only DMSO as 100% of viable cells. The 50%-cytotoxic concentrations (CC 50 ) were gained by a non-linear least-squares fit in the software GraphPad Prism 7.

CPE Assay
The target compounds were assessed for their abilities in inhibiting influenza virus replication in MDCK cells by CPE reduction assay [54][55][56][57]. MDCK cells were seeded into a 96-well plate and were infected with virus at an MOI of 50 CCID 50 (50% cell culture infective dose) per well. The target compounds were added in serial dilutions. Then, replace the growth media by Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 0.2 µg/mL of TPCK-treated trypsin. Microscopy was performed to score virus-induced CPE after 72 h incubation at 35 • C. The concentration of test compound that protected half of the cells (EC 50 values) was calculated using a non-linear least-squares fit in the software GraphPad Prism 7. It is worth noting that cases where the culture was protected less than 50% before cytotoxicity became dominant as inhibitor concentrations were increased were designated "No effect".

Plaque Inhibition Assay
The anti-influenza virus activities of the selected compounds were evaluated by plaque inhibition assay. MDCK cells were seeded at 5000 cells/well on 96-well plates for a day before being infected with the influenza virus (A/WSN/33, H1N1). The infection medium was DMEM (High Glucose) containing 1% FBS and 0.2% trypsin (1 µg/mL) [58]. The selected compounds were added to the cell culture at 100 µM. Unless otherwise indicated, the MDCK cells were infected with influenza A virus at a multiplicity of infection (MOI) of 0.1. We added Promega CellTiter-Glo ® reagen to each well following the protocol provided by the supplier after 45 h of incubation. Molecular Device SpectraMax M2 plate reader was utilized to quantified the luminescence (RLU) emitted from each well. The concentration required to inhibit 50% (IC 50 values) of A/WSN/33 was calculated using the software GraphPad Prism 7.

PA−PB1 Inhibitory Activity Prediction by the Best Pharmacophore Hypo1
Firstly, 27 active and moderately active compounds [2,[25][26][27][28][29][30][31] were selected as training set compounds to generate a pharmacophore model [32][33][34][35] by Hypogen algorithm 3D-QSAR pharmacophore generation protocol. The best pharmacophore Hypo1 enlisted in Table S5 was characterized with lowest total cost value (94.2536), the highest cost difference (75.318), the lowest RMSD (0.973343), and the best correlation coefficient (0.937771). Then, cost analysis, Fischer's randomization test and test set analysis were used to validate the best pharmacophore model (Hypo1). The specific steps of pharmacophore model generation and validation were shown in the supporting information ( Figures S2-S8, Tables S5-S7). The target compounds were all screened by the best pharmacophore Hypo1 and the estimated values were enlisted in Table 1.

Lipinski's Rule and ADMET Prediction
A major filtration criterion for the drug design process was the prediction of adsorption, distribution, metabolism, excretion and toxicity (ADMET) properties. In this study, the ADMET modules in Discovery Studio 3.0 were used to calculate various mathematical predictive ADMET pharmacokinetic parameters, such as blood-brain-barrier penetration, human intestinal absorption, aqueous solubility, hepatotoxicity, plasma protein binding. Then, the 26 compounds were subjected to toxicity screening models using TOPKAT module of Discovery Studio 3.0.

Molecular Docking
Schrödinger's Glide docking protocol was performed to the virtual screening and study the interactions of the selected target compounds with the crystal structure (PDB ID: 3CM8) [40][41][42][43] which was retrieved from the protein data base (RCSB PDB, http: //www.rcsb.org, accessed on 12 August 2021). Then, Schrödinger's protein preparation wizards were used to prepare the protein (remove the cofactors and water molecules; add missing residues; add hydrogens; generate Het states, and optimize the selected protein) [42]. Subsequently, the prepared protein structure was further processed for generating grid [43]. The active sites were defined based on the key residues (Asn412, Gln408, Glu623, Trp706, Trp618, Ile621, Lys643, Arg673, and Gln670) of in-bound ligand according to the previous literature sources [2,39,44]. Finally, the selected target compounds were prepared to implement the molecular docking by Glide-XP (extra precision). The docking poses were visually analyzed.

Preparation for Molecular Dynamics Simulation
The 100 ns MD simulations were performed by Desmond v3.8 module in the Schrödinger suite (version 9.6, Schrödinger Inc., New York, NY, USA). Three complexes (G07-3CM8, G19-3CM8 and G22-3CM8) were carried out for 100 ns MD simulations. The system was solvated with SPC water and neutralized by adding an appropriate amount of counter ions in the orthorhombic box (10 Å × 10 Å × 10 Å) in order to generate a buffer area between the protein atoms and the side of the box. Then, OPLS_2005 force field was used to minimize the energy of the complex system. The maximum number of iterations was set as 5000 during the minimization process. The temperature was set to 300 K and the pressure was set to 1.01325 bar. Finally, the 100 ns MD simulations were carried out (record the time interval of each trajectory at every 100 PS) [60][61][62].
The simulation quality analysis tool was used to analysis the MD simulations. The quality of MD simulations was predicted by the simulation event analysis tool. The proteinligand interactions were identified through the simulation interaction diagram tool.

Prime/MM-GBSA Simulation
The molecular mechanics generalized born surface area (MM-GBSA) method was used to calculate the binding-free energy (∆G bind ) of each ligand according to the following equation [63]: ∆G bind = ∆E MM + ∆G solv + ∆G SA In the equation, ∆E MM was the difference in the minimized energies between the ligand-protein complexes and the sum of the energies of the protein and ligand in the unbound state. ∆G solv was the difference in the GBSA solvation energy of the ligandprotein complexes and the sum of the solvation energies for the protein and ligand in free form. ∆G SA was the difference in surface area energies for the ligand-protein complexes and the sum of the surface area energies for protein and ligand.

Alanine Scanning Mutagenesis
ASM analysis was usually used to investigate the role of a specific amino acid residue participating in protein-protein or protein-ligand interactions [45,46]. It was applied to further validate the binding free energy decomposition analysis.
∆∆G bind = ∆G bind,mutant − ∆G bind,wild type The difference between the binding free energies for the wild type (∆G bind,wild type ) and the mutant (∆G bind,mutant ) yield the changes of binding free energy (∆∆G bind ) arising from the substitution of a specific amino acid by an alanine. More positive ∆∆G bind value indicates that the single mutation caused more significant effects and this specific residue played an important role in ligand binding [45,46].

Conclusions
This study focused on the synthesis and antiviral activity studies of 4-(quinolin-4ylamino)benzamide derivatives. All the new compounds were evaluated for their cytotoxi-city in MDCK cells and the anti-influenza virus (A/WSN/33, H1N1) activities. The results indicated that 4-(quinolin-4-ylamino)benzamide derivatives exhibited anti-influenza virus activities. G07 demonstrated significant anti-influenza virus both in cytopathic effect assay (EC 50 = 11.38 ± 1.89 µM) and anti-influenza assay (IC 50 = 0.23 ± 0.15 µM) of influenza A virus strain (A/WSN/33, H1N1) in MDCK cells. In addition, G07 exhibited significant anti-influenza virus activities against other three different influenza virus strains A/PR/8 (H1N1), A/HK/68 (H3N2) and influenza B virus. According to the RNP reconstitution assay, G07 could interect well with RNP with an inhibition rate of 80.65% at 100 µM, and G07 exhibited significant activity against PA−PB1 subunit of RNA polymerase based on the PA−PB1 inhibitory activity prediction by the best pharmacophore Hypo1. Therefore, it can be concluded that G07 is a potential anti-influenza virus agent. The molecular docking and molecular dynamics simulation results indicated that G07, G19 and G23 could interact well with the PA−PB1 active site, and GLU623, LYS643, TRP706, PHE707 and PHE710 are key amino acid residues interacted with the ligands. It can be speculated that these small molecules exhibited anti-influenza virus activities after blocking the PA−PB1 interface by competing with the PB1. This study can enrich the diverse library of quinoline-based compounds and provides a novel series of molecules for developing potential anti-influenza agents against PA−PB1.