The combination of computational and experimental strategies is of great value for the identification and development of desirable compounds. Structure-based virtual screening (SBVS) has been used for decades in early drug discovery to search chemical compound libraries for novel bioactive molecules against a particular drug target (Kontoyianni, 2017). The computational technique underlying virtual screening, molecular docking, is widely used in modern drug design to explore ligand conformations at the binding site of macromolecular targets (Ferreira et al., 2015). The ligand-receptor binding free energy, estimated by assessing the ligand binding mode during intermolecular recognition and the corresponding intermolecular interactions, provides a scoring ranking of docked compounds based on the binding affinity of the ligand-receptor complex (Huang and Zou, 2010; Lopez-Vallejo et al., 2011). The technique is fast enough to allow virtual screening of ligand libraries containing tens of thousands of compounds in a short period of time (Forli et al., 2016). In this study, a virtual screening procedure was established as shown in Figure 1. Initially, 73355 compounds were filtered based on Lipinski’s rule of five and ADMET property prediction, and 35414 compounds remained. Next, molecular docking was performed using the Libdock module to perform a rough screening of the compounds and the top 4% compounds with the highest docking scores were retained. Then, an exact docking was performed using the Glide XP module and retained the top 20% compounds with the highest scores. Finally, hierarchical clustering was performed using the Canvas module to yield 20 groups of compounds, from where 19 compounds were selected by analyzing the receptor-ligand binding mode, ligand spatial site resistance and backbone characteristics of the compounds. We successfully purchased 11 of them (No 1-11 as listed in Table 1) for subsequent bioassays.

We choseDrosophila as a screening platform, which is less expensive to raise, breed and screen than traditional mammalian models and has a shorter testing cycle (Pandey and Nichols, 2011). Studies have shown that the herbicide paraquat has neurotoxic effects (Chaudhuri et al., 2007) and Drosophila models exposed to PQ are often used as the models of neurodegenerative diseases to find new neuroprotective compounds (Jaiswal et al., 2012; Fernández-Hernández et al., 2016; Hajji et al., 2019). Here, Drosophila were treated with 6 mg/ml Paraquat dichloride (methyl viologen; PQ) in a sucrose solution, which provides basic energy for survival, for 24 h (Reiszadeh Jahromi et al., 2021). The data showed that 6 mg/ml PQ exposure led to lethal effects in Drosophila, and the survival rate was 50%. The survival rates of all groups with single addition of compounds were 100%, demonstrating that all screened compounds had no effect on the normal growth of Drosophila at this concentration. Compared with the experimental group (5% sucrose + 6 mg/ml PQ), the survival rate of Drosophila in the experimental group (compound + 6 mg/ml PQ) was improved. Five compounds showed significant protective effects against PQ-induced toxicity after pretreatment (Figure 2). Overall, these five compounds appear to be a potential neuroprotective candidate against PQ toxicity in PQ-induced toxicity model in Drosophila.

Experiments were designed to test the effect of compounds on the PGK-catalyzed reaction in vitro based on PGK kinase activity. As the PGK self-catalyzed reaction is rather rapid and not easily detectable, an indirect approach was used by coupling the PGK-catalyzed reaction with the pyruvate dehydrogenase (GAPDH) catalyzing the production of 1,3-BPG from glyceraldehyde phosphate. The substrate for the PGK forward reaction is 1,3-BPG, and if the PGK-catalyzed reaction is positively enhanced, which can cause a significant decrease in the concentration of 1,3-BPG in the reaction system, promoting the GAPDH forward reaction. The product of the GAPDH forward reaction, NADH, has strong absorption at 339 nm, and thus the strength of the PGK reaction can be monitored by the changes in NADH. To test the effect of compounds on the activity of PGK1, an in vitro assay was carried out. Purified hPGK1 was provided with substrates (ADP and 1,3-BPG) and the positive response accumulating NADH was then measured. The results showed that within a certain range, activation of PGK1 by 7979989 (10 μM-1 nM), Z112553128 (10 μM-1 nM), and AK-693/21087020 (1 μM-1 nM) caused a positive PGK response, thus promoting a positive GAPDH response to produce NADH. However, AK-693/21087020 inhibited PGK1 activity at 10 μM, and this difference may be due to the difference in drug-PGK1 interaction forces. The binding modes of these three compounds are shown in Figure 3. Theoretically, 7979989 and Z112553128 have the same molecular backbone, so they produce similar forces when they bind to the surrounding amino acid residues, and thus they are similar in terms of activity. While the structure of AK-693/21087020 is quite different from others resulting in different interactions with protein and different activities.

The biosafe concentrations of the compounds were determined by MTT assays. Cells were exposed to concentrations of candidate compounds ranging from 25 to 1.5625 μM for 24 h (Figure 4B). MTT assays showed that almost 100% of PC12 cells treated with 12.5 μM or lower concentrations survived compared to the controls, indicating that the different concentrations of compounds ranging from 12.5 to 1.5625 μM did not cause any significant decrease in cell decrease in viability. However, the higher concentrations reduced the cell viability (AK: 80.74 ± 3.66% of control value, p < 0.005 vs. control; 797: 89.61 ± 1.91% of control value, p < 0.01 vs. control; Z11: 70.30 ± 1.61% of control value, p < 0.0001 vs. control), suggesting that concentrations above 25 μM are detrimental to neuronal cells.

To explore the protective effect of candidate compounds against neuronal OGD/R injury, we treated OGD/R-injured PC12 cells with these compounds (12.5–1.5625 µM). The obtained data showed a significant decrease in cell viability after OGD/R injury (AK: 54.36 ± 5.76% of control value; 797: 53.74 ± 5.76% of control value; Z11: 52.39 ± 5.76% of control value, p < 0.0001 vs. control). Significant improvements in cell viability were observed in the compound groups at concentrations of 12.5, 6.25, 3.125, and 1.5625 μM compared to the OGD/R group (AK: 65.53 ± 3.15%, 71.53 ± 1.49%, 80.37 ± 1.19%, and 71.69 ± 4.32% of control value, respectively; 797: 72.37 ± 5.60%, 77.19 ± 1.02%, 78.45 ± 2.55%, and 80.86 ± 1.60% of control value, respectively; Z11: 65.65 ± 3.07%, 74.24 ± 0.38%, 74.78 ± 5.13%, and 76.57 ± 1.53% of control value, respectively) (Figure 4A). The graph does not show dose-dependence probably because hypoxic conditions increase the expression of hypoxia-inducible factor-1 (HIF-1). Under hypoxic conditions, most eukaryotic cells can convert mitochondrial respiration to increased glycolysis to maintain ATP levels. The presence of HIF-1 leads to increased expression of glycolytic enzymes such as PGK1, which accelerates the glycolytic process and correspondingly increases the level of lactate, a glycolytic product that leads to a decrease in intracellular pH, resulting in cellular acidosis and ultimately apoptosis. Although the compound activates PGK1 with neuroprotective effects, the elevation of PGK1 levels under hypoxic conditions causes cellular acidosis and eventually apoptosis.

LDH release was significantly increased in OGD/R-treated PC12 cells (228 ± 1.26%) compared to the control group (Figure 4C). A decrease in LDH release could be recorded following compounds treatment. OGD/R + AK (3.125 μM), OGD/R +797 (1.5625 μM), and OGD/R + Z11 (1.5625 μM) treatment reduced the values to 191 ± 0.45%, 184 ± 1.39%, and 162 ± 6.15%, respectively.

The effect of compounds on OGD/R-induced apoptosis in PC12 cell was shown by flow cytometry using Annexin V-FITC/PI double staining. As shown in Figure 5, the apoptosis rate after OGD/R injury was much higher than the control group, while in the OGD/R + AK (3.125 μM), OGD/R +797 (1.5625 μM), and OGD/R + Z11 (1.5625 μM) groups, the apoptosis rate after OGD/R injury was significantly reduced, indicating that the compounds were effective in preventing OGD/R-induced apoptosis.

The MTT experiment was performed to ascertain whether 7979989 and Z112553128 have cytotoxicity on RAW264.7 cells. The results were shown in Figure 6, from where we can see that compound 7979989, from 0.00256 to 200 μM for 24 h, did not decrease the cell viability compared with the control group. As for Z112553128, the results appeared a decreased trend from the concentration 40–200 μM. Hence, in the following steps, we chose a concentration range of 0.0128–8 μM to verify preliminarily the anti-inflammatory properties by using the Griess method.

Lipopolysaccharide (LPS) is a major component of the outer cell membrane of Gram-negative bacteria and is intrinsically important for inflammation caused by pathogens. When bound with LPS, the receptor TLR4 is initiated, thereby promoting downstream events. NF-κB drives a classical signaling pathway and gene regulation that is associated with various of diseases including inflammatory diseases. Activation of NF-κB initiates transcription and expression of pro-inflammatory genes, which then leads to excessive accumulation of various mediators such as tumor necrosis factor (TNF)-α, interleukin (IL)-β and IL-6. In this experiment, we finally determined the LPS (1 μg/ml) according to the conditions, and dissolved it in PBS. But for stimulation, the environment has no special requirements, just add LPS directly to the cell culture medium. Finally, by verifying the NO release of the LPS group and the non-adding group to determine whether the inflammation model is successful.

The RAW264.7 cells induced with LPS (1 μg/ml) were exposed to various concentrations of 7979989 and Z112553128 ranging from 0.00256 to 8 μM. The obtained results showed that NO concentration released by RAW264.7 cells increased significantly when added with LPS compared with control group without LPS, and the treatment with 7979989 and Z112553128 decreased the LPS-induced NO production. As shown in Figure 7, the two compounds have similar inhibitory trend on NO, and both illustrated significant inhibitory effects at the con-centration of 8 μmol/L. Then the suppression effect decreased with the decrease of compound concentrations, and disappeared at around 0.0128 μM. At the concentration of 8, 1.6 and 0.32 μM, they showed better inhibitory effect than other concentrations. Therefore, we chose 8, 1.6 and 0.32 Μm for subsequent experiments.

To investigate the inhibitory effect of 7979989 and Z112553128 on IL-1β and TNF-α production, we added the compounds to the RAW264.7 cells pre-treated with LPS for 1.5 h, and collected the cell supernatant after 24 h for centrifugation to conduct ELISA detection of different factors. As shown in Figure 8, compared with the unstimulated cells of control group, the level of IL-1β and TNF-α were increased significantly in LPS-induced cells of the model group. After adding 7979989 and Z112553128, the release of inflammatory factors produced by macrophages were inhibited to varying degrees. Shown in Figure 8A, the compound 7979989 displays an effective inhibitory effect on IL-1β production while it only slightly hampered the production of TNF-α (Figure 8C). In addition, Z112553128 has a specific characteristic of inhibiting the release of IL-1β and TNF-α as shown in Figures 8B,D, which is consistent with the previous NO results.

To evaluate the effect of 7979989 and Z112553128 on pro-inflammatory mediators, we determined the expressions of iNOS and COX-2. Figure 9A shows that compound 7979989 could reduce the expressions of iNOS and COX-2 proteins (p < 0.05) compared with the LPS alone-treated model group and the appearance of iNOS and COX-2 were slightly weaker than LPS alone-treated model group. Figures 9B,C summarize the protein expression level and trend, where the levels of iNOS and COX-2 were significantly reduced. Compound Z112553128 also reduces the expressions of these two proteins in Figures 9D–F. Thus, both 7979989 and Z112553128 play essential roles in inhibiting the expression of inflammatory mediators such as iNOS and COX-2.

To evaluate the effects of compounds 7979989 and Z112553128 on oxidative stress induced by LPS, fluorescence staining and flow cytometry were used to examine intracellular ROS production. Figure 10 shows that after LPS stimulation alone, the release of ROS was strongly stimulated as shown by the green normal distribution curve. After adding 7979989, the normal distribution curve had a significant shift in cytofluorogram to the left and almost reached a coincidence with the control group. Compared with the model group, the distribution of the Z112553128 also had an obvious shift to the left, indicating that it had an inhibitory effect on the production of reactive oxygen species in the cells. The inhibitory effect gradually weakened as the concentrations decreased, and a significant decline in the average fluorescence intensity was detected. Compared with the model group, the image also indicated a sharp decrease after treatment with 7979989 and Z112553128, while the average fluorescence intensity was significantly changed in the concentration range of 8 to 0.32 µM (p < 0.05). Therefore, compounds 7979989 and Z112553128 can reduce LPS-induced production of intracellular ROS.

Results of experiment show the average value of the results from three independent repeated experiments. Statistical analysis was performed using the GraphPad Prism 5.0 statistical software. Data are expressed as means ± standard deviation (SD). Statistical analysis was carried out using one-way ANOVA followed by Tukey’s multiple comparison test, with p < 0.05 considered statistically significant.