Inhibition activity of a traditional Chinese herbal formula Huang-Lian-Jie-Du-Tang and its major components found in its plasma profile on neuraminidase-1

Huang-Lian-Jie-Du-Tang (HLJDT), a traditional formula with four TCM herbs, has been used for hundred years for different diseases. The current study aimed to assess the inhibitory activity of HLJDT against H1N1 neuraminidase (NA-1), and identify potent NA-1 inhibitors from its plasma profile. The in vitro NA-1 study has shown that the water extract of HLJDT potently inhibited NA-1 (IC50 = 112.6 μg/ml; Ki = 55.6 μg/ml) in a competitive mode. The IC50 values of the water extracts of its four herbs were as follows: Coptidis Rhizoma, 96.1 μg/ml; Phellodendri Chinensis Cortex, 108.6 μg/ml; Scutellariae Radix, 303.5 μg/ml; Gardeniae Fructus, 285.0 μg/ml. Thirteen compounds found in the plasma profile of HLJDT were also identified as potent NA-1 inhibitors, which included jatrorrhizine, palmatine, epiberberine, geniposide, oroxylin A, berberine, coptisine, baicalein, wogonoside, phellodendrine, wogonin, oroxylin A-7-O-glucuronide and baicalin (sorted in ascending order by their IC50 values). Their inhibitory activities were consistent with molecular docking analysis when considering crystallographic water molecules in the ligand-binding pocket of NA-1. Our current findings suggested that HLJDT can be used as a complementary medicine for H1N1 infection and its potent active compounds can be developed as NA-1 inhibitors.

Highly infectious influenza A virus is pandemics and recurrent annual epidemics, and causes severe respiratory illness and death, especially in the elderly, children, and weakness. Neuraminidase (NA), a surface glycoprotein antigen, is one of biomarkers for subtype classification of influenza A virus. NA facilitates the release of influenza A virus via hydrolyzing glycosidic linkages of terminal sialic acid residues, which is critical to infection progression in the host. Current treatment strategy for influenza virus infection is to inhibit NA function 1 . Several crystal structures of NA are obtained, and these structures facilitate structure-based drug discovery of NA inhibitors 1 . Two commercial drugs zanamivir (Relenza) and oseltamivir (Tamiflu), as derivatives of sialic acid, have been developed through this process. However, the supply of these drugs is limited. It is not possible to prescribe these drugs in the countryside of China when patients get influenza virus infection without serious symptoms. Therefore, it is necessary to discover new drug candidates for treating H1N1 infection. Currently, natural products (e.g. chlorogenic acid 2 , quercetin-7-O-glucoside 3 and catechins 4 ) are considerable resources for the discovery of NA inhibitors.
Although HLJDT is not traditionally used for the treatment of influenza A virus infection, some of its active components, such as baicalein 10 , berberine 11 and coptisine 12 , have been identified as effective inhibitors of various NA subtypes. Other major components, detectable in the plasma profile of HLJDT, are supposed to be a potential resource for discovering NA inhibitors due to their similar structures. The aim of our current study was to evaluate the inhibitory activity of the water extracts of HLJDT and its four herbs on NA-1, and identify potent NA-1 inhibitors from its plasma profile (see chemical structures in Fig. 1) by in vitro inhibition study. Further, the inhibition of active compounds against NA-1 was also evaluated by in silico molecular simulation, which shows a better understanding for the binding mechanisms of the active compounds in ligand-binding pocket of NA-1. The results would provide information for further investigation on HLJDT as a complementary medicine in clinics for treating H1N1 infection, and its potent NA-1 inhibitors can also be a chemical resource for new drug development.
For graphical inspection on the inhibition type of HLJDT, Primary Lineweaver-Burk plot (obtained by reciprocal of reaction velocities versus reciprocal of MUNANA's concentrations) and Dixon plot (obtained by reciprocal of reaction velocities versus HLJDT's concentrations) were firstly applied. As shown in Fig. 2B,C, the straight lines did not intersect on the x-axis or first quadrant in the Primary Lineweaver-Burk plot, but intersected on the x-axis in the Dixon plot. However, possibly due to experimental deviation, the inhibition type can not be confirmed by Primary Lineweaver-Burk plot and Dixon plot. For confirmation of the inhibition type, Secondary Dixon plot (obtained by the slopes of the regression lines in the Dixon plot versus reciprocal of MUNANA's concentrations) was further drawn. As shown in Fig. 2D, the straight line goes through the origin, showing a competitive inhibition of HLJDT on NA activity 13 . Secondary Lineweaver-Burk plot for K i (obtained by the slopes of the regression lines in the Primary Lineweaver-Burk plot versus HLJDT's concentrations) in Fig. 2E showed that the K i value of HLJDT on NA-1 activity was 55.6 μg/ml. This is the first time to report NA-1 inhibition activity of HLJDT. Since HLJDT is a clinically used formula, it is very safe for TCM practitioners to prescribe it for patients with influenza-like symptom before prescribing western drugs like Tamiflu.
In vitro inhibition study of each herb. In Fig. 3, when using MUNANA concentration at 20 μM, the IC 50 and IC 10 values of each herb on NA activity were as follows: Phellodendri Chinensis Cortex (HB), IC 50 = 108.6 ± 8.6 μg/ml and IC 10 = 8.4 ± 3.3 μg/ml; Coptidis Rhizoma (HL), IC 50 = 96.1 ± 7.6 μg/ml and IC 10 = 9.9 ± 1.1 μg/ml; Scutellariae Radix (HQ), IC 50 = 303.5 ± 21.9 μg/ml and IC 10 = 28.0 ± 8.7 μg/ml; Gardeniae Fructus (ZZ), IC 50 = 285.0 ± 16.6 μg/ml and IC 10 = 37.4 ± 2.7 μg/ml. According to One-way ANOVA with Tukey's multiple comparisons post-test, there were no significant difference among the IC 50 values of HB, HL and HLJDT (P > 0.05). Meanwhile, the IC 50 value of HLJDT was significantly lower than the one of HQ and ZZ (P < 0.01). These suggested that HB and HL were the major ingredients of HLJDT responsible for NA-1 inhibition, and HQ and ZZ were also involved in NA-1 inhibition. Since the HLJDT formula was designed according to the TCM theory, its safety has been observed in its clinical application for hundreds of years. Its composition should not be changed in its application.

Molecular docking analysis of potent NA-1 inhibitors.
For the validation of docking procedure, zanamivir was firstly re-docked to its co-crystalized NA-1 structure (PDB ID 3B7E). As shown in the Fig. 4A, the pose of re-docked zanamivir (in purple sticks) was very close to the original crystallographic one (in yellow sticks) with a RMSD (Root-Mean-Square Deviation) of 0.285 Å which is much smaller than the cutoff of 2 Å. This indicated that the parameters used in the current procedures are accurate for molecular docking.
Current approach using Autodock Vina provided maximum 9 possible binding conformations of each compound in the ligand-binding pocket with different binding free energies, but this method ignored the existence of water molecules by removing water molecules at the beginning of docking procedures. The favorable conformations of zanamivir, jatrorrhizine and phellodendrine were observed with respective lowest binding free energies of −8.1, −6.8 and −7.4 kal/mol, and did not overlap the poses of crystallographic water molecules in the ligand-binding pocket of NA-1 (Table 2). However, the poses of other compounds with lowest binding free energies overlapped those of crystallographic water molecules in the ligand-binding pocket. Since the water molecule has a spherical volume with a radius of 1 Å, the binding conformations of each compound with lower binding free energies were re-selected with distance (more than 1 Å) to corresponding crystallographic waters 20 . As shown in Table 2, the binding free energy of the re-selected pose of sialic acid was −6.5 kal/mol. Except baicalin and MUNANA, all other tested compounds bound to the ligand-binding pocket with lower binding free energies and appropriate distance to the corresponding crystallographic waters when compared to the one of sialic acid. As reported before, the 150 cavity, as part of the ligand-binding pocket, is closed when the inhibitor bound to the ligand-binding pocket of NA-1 1 . As such, the volume of NA-1 co-complexed with zanamivir (PDB ID 3B7E) used in the current study was smaller than the one of its native structure (PDB ID 3BEQ). This is may be the reason that there is no proper conformations of MUNANA and baicalin with bigger volumes in the ligand-binding pocket of NA-1 (PDB ID 3B7E).
Molecular simulation of NA-1 inhibitors. As mentioned above, the binding conformations of active compounds with lower binding free energy and proper distances (more than 1 Å) to the water molecules in the ligand-binding cavities of NA-1 were selected, and simulated by both LigPlot + and PyMOL. Among these residues, eight charged and polar residues (e.g. Arg118, Asp151, Arg152, Arg224, Glu276, Arg292, Arg371 and Tyr406) directly interacted with the substrate in the ligand-binding cavities (the catalytic site) of NA-1 1 . As illustrated in Fig. 4A and summarized in Table 3, the binding conformation of zanamivir showed that as reported 1 , it directly interacted via Hydrogen-bonding with amino acid residues including Arg118, Asp151, Arg152, Trp178, Glu227, Glu276, Arg292, and Arg371, and through hydrophobic contacts with Glu119, Arg224, Ser246, Glu277, Asn294, and Tyr406, which contributed to the inhibition of zanamivir in the catalytic site. Similarly, inhibitors from the plasma profile of HLJDT bound to the catalytic site by interacting with these amino acid residues via H-bonding and/or hydrophobic contacts. These interactions took responsibility for the inhibition of these compounds on NA-1-mediated 4-MU formation. Among these compounds, jatrorrhizine and palmatine interacted with amino acid residues in the catalytic site of NA-1 only through hydrophobic contacts. Besides hydrophobic contacts, other inhibitors tested in this study strongly interacted amino acid residues through hydrogen bonds. Hydroxyl groups at C-5, C-6 and C-7 of baicalein interacted with Glu227, Glu277 and Tyr406 via H-bonding. Hydroxyl group at C-7 of oroxylin A generated H-bonding with Arg118 and Arg371, while oroxylin A 7-O-β-D-glucuronide generated H-bondings via hydroxyl group at C-5 with Asn294, as well as its glucuronide with Asn221, Gly244, and Ser246. Wogonin formed H-bondings through ketone group at C-3 with Arg292 and Tyr406, hydroxyl group at C-5 with Arg292, and hydroxyl group at C-7 with Ser246. Likewise, wogonoside developed H-bondings through ketone group at C-3 with Asn294, and glucuronide with Asp151, Arg156, Trp178, Glu227, and Glu277. Methoxyl group at C-9 and 1,3-dioxolane at C-2, 3 of berberine interacted via H-bondings with Asn294 and Ser179, respectively, which were different from those of bacterial NA 12 . Ser246 formed H-bondings with 1,3-dioxolane at C-2, 3 of coptisine and epiberberine, hydroxyl group at C-2 of phellodendrine, respectively. Hydrogen bonds of geniposide were formed through hydroxymethyl group at C-7 with Arg292, Arg371 and Tyr406, 7-(hydroxymethyl)-methyl ester with Ser246, 4-carboxylic acid with Asn294, and glucosyl group with Glu227, Trp178 and Asp151.
Structure-activity relationship. According to the inhibition activities (IC 50 values) of tested compounds on NA-1, structure-activity relationships of flavones and isoquinoline alkaloids tested in the current study were explored and summarized as follows.
In silico prediction for oral toxicity in rodents. The LD 50 values for the compounds tested in this study are predicted by PROTOX and shown in Table 4. Jatrorrhizine and Palmatine showed more toxic with oral LD 50

Conclusion
This is the first time to report the NA-1 inhibition activity of HLJDT as a ready-to-use potent agent for anti-H1N1 infection, and it is suggested that it is valuable for TCM practitioners to use HLJDT and evaluate its efficacy in patients with influenza-like symptoms when they can not be diagnosed at the beginning. The in vitro inhibition activities of 13 compounds found in the plasma profile of HLJDT should be responsible for the possible therapeutic effect of HLJDT due to the correlation of their Cmax and NA-1 inhibition activity. Besides, it is the first time to report the inhibition effects of epiberberine, oroxylin A, oroxylin A 7-O-β-D-glucuronide, and phellodendrine on NA-1 in vitro and in silico.

Materials and Methods
Materials. All  Raw herbs for the formula (containing Coptidis Rhizoma, 9 g; Phellodendri Chinensis Cortex, 9 g; Scutellariae Radix, 6 g; and Gardeniae Fructus, 6 g) were extracted by boiling with water (300 mL) for 60 min. After filtered, the residue was extracted again in the same way for another hour. The supernatant was mixed and subjected to freeze-dry. The dried extract was kept in the desiccator before use. The dried yield of HLJDT extract was ~31%. Each herb with respective amount in the formula was individually prepared by the same extraction method with 300 mL of water, and their extraction yields were as follows: Scutellariae Radix, ~55%; Coptidis Rhizoma, ~22.6%; Phellodendri Chinensis Cortex, ~24.6%; Gardeniae Fructus, ~35.4%.

Ligands
H-bonding to amino acid residues (distance, Å) Hydrophobic contacts to amino acid residues   The binding modes of active compounds with lower binding free energy and proper distances (more than 1 Å) to water molecules in the ligand-binding cavity of NA were chosen for further analysis of docking conformation. The 2D and 3D simulation results were illustrated by LigPlot + v.1.4.5 (http://www.ebi.ac.uk/thornton-srv/software/LIGPLOT/) 22  In silico prediction for oral toxicity in rodents. PROTOX, a webserver for predicting oral toxicities of small molecules in rodents, was used for evaluating the oral toxicity of the compounds tested in this study 23 . The structures of the tested compounds were uploaded in the website, and results were generated by the proteinligand-based pharmacophore models with the in-house toxicity database. Data analysis. Data were expressed as mean ± standard error of Mean (SEM). IC 50 value was calculated by non-linear regression analysis with Prism version 5.0 (GraphPad Software, CA, USA). The inhibition constants (K i ) and modes of different inhibitors to neuraminidase-1 were measured by graphical inspection from different plots (e.g. Primary Lineweaver-Burk Plot, Dixon plot, and Secondary Lineweaver-Burk plot for Ki) according to the previous report 13,24 . Data were analyzed by One-way ANOVA with Tukey's multiple comparisons post-test. A p value less than 0.05 was considered statistically significant.