The molecular docking software is Autodock vina ( https://autodock.scripps.edu ); The visual processing software uses the open-source version of Pymol2.1 (Archived: Python Extension Packages for Windows-Christoph Gohlke (uci.edu)); GROMACS ( https://www.gromacs.org/ ) is used for molecular dynamics simulation. TCMSP-Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (tcmsp-e.com); Protein Database (RCSB PDB: Homepage) and Discovery studio visualizer (Free Download: BIOVIA Discovery Studio Visualizer-Dassault Systemes (3ds.com)).

dPLA2 protein crystal structure (PDB ID: 1ijl) was retrieved from the PDB database and downloaded as a receptor; The six Chinese herbs selected for study are Hedyotis diffusa Willd, Lobelia chinensis Lour, Andrographis paniculata Nees, Houttuynia cordata Thunb, Sedum sarmentosum Bunge, Eclipta prostata. All the molecular components of these herbs were obtained from the TCMSP database, and screened according to bioavailability OB ≥ 30% and drug properties DL ≥ 0.18. The 3D structure of the selected small molecules was downloaded as ligands for docking with the PLA2 of Deinagkistrodon acutus.

The small molecule and dPLA2 protein were dehydrated and hydrogenated, respectively, and the small molecule twisting bond was detected before molecular docking; The processed small molecules and proteins were imported into Autodock software [20] . The docking site refers to the smallest spatial extent that encompasses the active center of the receptor. This encompasses the area defined by the docking box during the operation. Consequently, the size and the center coordinates of the box serve as a representation of this range throughout the procedure. The active site of dPLA2 comprises the region delineated by residues Y27, G29, G31, H47, D48, and D89. Consequently, we established the docking site with central coordinates at X = 8.391401, Y = 11.560492, Z = 47.501534, and size of 15 × 15 × 15. Then initiating molecular docking; The docking result files are saved in pdbqt and pdb formats; The former was imported into PYMOL [21] for visualized, the latter imported the Discovery studio visualizer to analyze the receptor-ligand interaction and derive the 2D diagram of the interaction [22] ; After docking, the results obtained were compared with the selected reference (phosphatidylcholine and gallic acid) in terms of binding site, binding energy, hydrogen bonding, etc.

GROMACS software was used for molecular dynamics simulation [23] . First, the ligand and receptor were separated from the docking result file, and the charge information was added separately to form two independent files. Secondly, topological files of ligand and receptor are generated, respectively. Then the atom information of the ligand topological file is integrated with the information of the receptor topological file to form the complex topological file, and the solvent is added to the complex topological file to generate the constraint box. After that, charge detection, balance charge, energy minimization, nvt balance and npt balance were carried out. Finally, dynamic simulation was carried out, and the simulation time was set to 100 ns. The changes of RMSD and RMSF values and the number of hydrogen bonds of complex and receptor were analyzed respectively.

Oral bioavailability (OB) and drug-like properties (DL) are two important basic parameters for virtual screening of small molecules with potential pharmacological effects. Using OB ≥ 30% and DL ≥ 0.18 as criteria, 287 molecules of 6 Chinese herbs were screened, and a total of 60 components were obtained. The results were shown in Table 1 . Molecular docking of these 60 compounds was performed. As shown in Table 1 , it is known that the lower the binding energy, the more stable the ligand-receptor binding is. The binding energy of all 60 compounds is less than 0, indicating that they can all bind to dPLA2. Among them, the binding energy of 53 small molecules was lower than that of gallic acid (−6.3 kJ/mol), and 59 small molecules was lower than phosphatidylcholine. Among the 10 small molecules with the lowest binding energy, 5 were from Lobelia chinensis, 2 were from Hedyotis diffusa, and there was 1 species of Houttuynia cordata, Sedum sarmentosum, Eclipta prostata respectively.

Since the lower the binding energy, the more stable the binding is, the 10 small molecules with the lowest binding energy among the 60 small molecules were

mainly selected for interaction analysis. The 10 small molecules were lobelanidine (−9.3 kJ/mol), Spinasterol (−9.2 kJ/mol), Lobeline (−9.1 kJ/mol), norlobelanine (−9.0 kJ/mol), Stigmasterol (−8.9 kJ/mol), Poriferasterol (−8.8 kJ/mol) and DFV (−8.8 kJ/mol), Pratensein (−8.8 kJ/mol), Cycloeucalenol (−8.8 kJ/mol), Lobelanine (−8.7 kJ/mol), and the docking interactions were shown in Table 2 . The binding energy of 10 small molecules was lower than that of gallic acid (−6.3 kJ/mol). Lobelanidine was the lowest and Lobelanine was the highest, and there was no significant difference between them.

The number of hydrogen bonds is an important factor in determining the stability of ligand binding. As shown in Table 2 , lobelanidine, norlobelanine and Pratensein form hydrogen bonds with three amino acid residues of dPLA2 respectively, lobeline forms hydrogen bonds with two amino acid residues of dPLA2, and the other molecules have only one hydrogen bond. Therefore, four small molecules were selected according to the number of hydrogen bonds, which were lobelanidine, norlobelanine, Pratensein and lobeline.

The docking results of four small molecules were visualized, as shown in Figure 1 . From the comparison of the overall spatial structure of the four small molecules and dPLA2 complex, it can be seen that They are all bound in the same surface pocket of dPLA2, which is formed by the calcium ion binding ring (TYR27, GLY29, GLY31, ASP48), calcium ion and its surrounding amino acids, and the surface pocket is the same as the gallic acid and phosphatidylcholine binding active pocket, with similar binding orientation of the four molecules, as shown in Figure 2 .

From the comparison of interactions, it can be seen that, on the one hand, the docking interaction results of the four small molecules are similar to gallic acid: they all interact with calcium ions; lobelanidine, norlobelanine and gallic acid form hydrogen bonds with GLY29 and ASP48; Lobeline and gallic acid form hydrogen bonds with GLY31; Lobelin, Pratensein and gallic acid all have hydrophobic effects with PHE5 and HIS47. On the other hand, phosphatidylcholine is a natural substrate for dPLA2 and binds to the active center of dPLA2. By analyzing the hydrogen bond and hydrophobic action of phosphatidylcholine, lobelanidine, norlobelanine, Pratensein and lobeline binding with dPLA2, it was found that GLY31 and LYS60 of dPLA2 were involved in the binding with phosphatidylcholine. And there were hydrogen bonds between Pratensein and these two amino acids; Norlobelanine and lobeline form hydrogen bonds with GLY31; Lobelanidine does not interact with either of them; And Lobeline, Pratensein and phosphatidylcholine all form hydrophobic interactions with CYS44 and HIS47.

Through the interaction analysis, Lobeline, norcloberanine, Pratensein and lobelanidine all had the possibility of inhibiting dPLA2.

Molecular dynamics is a set of molecular simulation methods widely used in physics, chemistry, biology, materials, medicine and other fields. This method mainly relies on Newtonian mechanics to simulate the motion of the molecular system, and extracts samples from the system composed of different states of the molecular system, so as to calculate the configuration integral of the system, and further calculate the thermodynamic quantity and other macroscopic properties of the system based on the results of the configuration integral. The stability of ligand-binding receptors can be evaluated more directly through molecular dynamics simulation of protein-ligand complexes.

Root mean square deviation (RMSD) and root mean square fluctuation (RMSF) are two important parameters in molecular dynamics simulation. They represent the mean difference between the complex conformation during the simulation time and the initial complex conformation, and the mean deviation between the atomic position of the complex during the simulation time and its initial position. Respectively reflecting the stability of the overall conformation of the complex and the flexibility of the overall structure. Based on these two parameters, the stability of dPLA2 binding to four small molecules was verified in this study.

Figure 3(a) shows the curve comparison of RMSD values of the four complexes

over the simulation time. It can be seen from the figure that the RMSD values of the four compounds tend to converge after 100 ns simulation, and their RMSD values are all lower than 4 ns. The lobelanidine curve is the most stable, and the RMSD value as a whole is lower than 0.2 ns, indicating that lobelanidine binding stability is the strongest. It was followed by Pratensein Lobeline, norcloberanine. Figure 3(b) shows the comparison of RMSD value of lobelanidine-dPLA2 complex and dPLA2. It can be seen from the figure that the RMSD value of lobelanidine-dPLA2 complex is lower than dPLA2 on the whole, indicating that the stability of dPLA2 combined with lobelanidine is enhanced.

Lobelanidine showed the best stability in both RMSD and RMSF analysis, so the hydrogen bond analysis of Lobelanidine was mainly performed. Figure 3(e) shows the variation in the number of hydrogen bonds between Lobelanidine and dPLA2 over the simulation time. It can be seen from the figure that the number of hydrogen bonds remained at 1 overall during the simulation time. In addition, When the distance between ligand and receptor is less than 0.35 nm, 3 pairs of hydrogen bonds are maintained, which is consistent with the molecular docking results.

dPLA2 is a calcium-dependent acid hydrolase, which mainly causes tissue hemolysis and edema [24] . The reason is that dPLA2 produces arachidonic acid by hydrolyzing phosphatidylcholine on red blood cell membranes, and the arachidonic acid is metabolized to produce inflammatory mediators such as prostaglandin, thrombin and leukotriene, which cause further hemolysis, vasodilation, increased vascular permeability, and finally tissue edema. On this basis, we identified molecules that potentially inhibit the hydrolysis of phosphatidylcholine by dPLA2 in order to block the hemolytic reaction and tissue edema.

The methods of inhibiting dPLA2 and small molecule binding sites are closely related. Studies have shown that dPLA2 binding sites are TYR27, GLY29, GLY31 and ASP48, and active sites are HIS47 and ASP89 [24] . Accordingly, dPLA2 interacts with phosphatidylcholine and gallic acid, respectively; the docking results are shown in Figure 1 and Table 1 . The comparison of the docking results showed that the surface parts of gallic acid and phosphatidylcholine binding dPLA2 were the same as those in the spatial structure analysis. The binding energies of gallic acid and phosphatidylcholine were −6.3 kJ/mol and −5.7 kJ/mol, respectively, and the lower the binding energies were, the stronger the binding affinity was. Therefore, the binding affinity of gallic acid and dPLA2 was stronger than that of phosphatidylcholine. From the interaction analysis, both gallic acid and phosphatidylcholine were bound to dPLA2 binding sites, but the binding amino acids were different. Gallic acid was bound to GLY29 and ASP48, and phosphatidylcholine was bound to GLY31, and both of them formed hydrophobic interaction with HIS47. In addition, phosphatidylcholine binds to the active site HIS47, but gallic acid does not bind to the active site, and gallic acid forms metallic bonds with calcium ions, but phosphatidylcholine does not bind to calcium ions. Therefore, we speculated that binding to the binding site is the basis of the effect of small molecules on dPLA2 activity, and binding to calcium ions may inhibit dPLA2.

Through molecular docking and molecular dynamics simulation, we obtained four potential dPLA2 inhibition molecules, lobelanidine, lobeline, norlobelanine and pratensein. The structural formulas of four small molecules are shown in Figure 4 . lobelanidine, lobeline, norlobelanine were all derived from Lobelia chinensis, and their structures were highly similar. The binding energy of the four small molecules were higher than that of gallic acid and phosphatidylcholine, and all of them were bound to dPLA2 binding sites and calcium ions but were not bound to HIS47 and ASP89, and their interaction characteristics were consistent with those of gallic acid. It is speculated that the molecular mechanism of its action is small molecules competitively inhibits the binding of dPLA2 with phosphatidylcholine, and is at an advantage in the competition, resulting in the overall inhibitory effect of dPLA2. Moreover, through molecular dynamics simulation, we found that the overall spatial conformation of four small molecules and dPLA2 complexes changed little and the overall flexibility decreased before and after simulation, indicating that the combination of small molecules and dPLA2 was stable.

As a respiratory stimulant, lobeline is widely used in clinical practice for neonatal asphyxia, asphyxia caused by carbon monoxide, intoxication of inhaled anesthetic and other central depressants (such as opioids and barbiturates), and respiratory failure caused by infectious diseases such as pneumonia and diphtheria, with mild side effects [25] [26] . In addition, studies have shown that lobeline can also affect learning and memory [27] , reduce depression-like behavior induced by nicotine withdrawal [28] [29] , reduce alcohol dependence [30] [31] , and reduce the function and metabolism of dopamine in the midbrain induced by alcohol in mice [32] [33] . Lobelanidine and norlobelanine are highly similar in structure to lobeline. For pratensein, studies have shown that it can reduce Aβ-induced oxidative damage and restore synaptic function [34] - [37] . Although these molecules are widely used in the treatment of diseases, there are no reports that they have inhibitory effects on dPLA2. To our knowledge, this is the first report of Lobelanidine, lobeline, norlobelanine and pratensein have an inhibitory effect on dPLA2, among which Lobelanidine had the strongest inhibitory effect. However, the actual inhibitory effect of these molecules on dPLA2 needs to be confirmed by further experiments.