GEB was used as a keyword to retrieve its active ingredient using the encyclopedia of TCM (ETCM; tcmip.cn/ETCM/index.php/Home/Index/) and systematically review the chemical structure of active ingredients in the TCM system pharmacology database and analysis platform (TCMSP; https://tcmsp-e.com/). The detailed screening process is shown in Fig. S1.

Target genes for the major active compounds in GEB were mined using the ETCM and SwissTargetPrediction (swisstargetprediction.ch/) databases. A merge-and-de-duplicate analysis produced a list of the final compound target genes. SwissTargetPrediction database predicted the target based on two- and three-dimensional (3D) similarity analysis of known ligands (21).

Disease-associated genes were accessed through five databases: GeneCards (genecards.org), Gene-Disease Networks (DisGeNet) (disgenet.org), DrugBank (drugbank.ca), Online Mendelian Inheritance in Man (OMIM) (ncbi.nlm.nih.gov/omim) and Therapeutic Target Database (TTD) (systemsdock.unit.osit.jp/iddp/home/index). The disease-associated genes were collected using IS as a keyword and merge-de-duplication analysis.

GSE16561 dataset (ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE16561) containing peripheral whole-blood RNA sequencing data from 39 patients with IS and 24 healthy control subjects was downloaded from the GEO database. Subsequently, differential expression analysis was performed using R package limma (version 3.44.3) (22) with IS vs. healthy subjects to identify DEGs associated with IS. |Log₂fold-change (FC)|>0.5 and P<0.05 were used to determine differential expression.

Functional enrichment analysis focused on key genes. The overlap of compound target and potential disease-associated genes with IS-associated DEGs were defined as key genes for treatment of IS using GEB. R package clusterProfiler (version 3.18.0) (23) used for the enrichment analysis of key genes was applied to GO and KEGG. GO contains three main categories, biological processes (BP), cellular components (CC) and molecular functions (MF). Adjusted P<0.05 indicated significantly enriched entries.

The potentially active compounds and key genes of GEB and IS were entered into Cytoscape software (version 3.2.1) (24) to develop the herb-compound-targets-disease network, where nodes represented potentially active compounds of GEB and the key targets of IS. The edges (linear segments) showed association between these factors.

STRING (version 11.0) (25) database was used to analyze interactions between the key genes previously identified for GEB treatment of IS (confidence score, 0.4) (26). Cytoscape (version 3.2.1) (24) was used to generate a PPI network map. In addition, the CytoNCA plugin (version 2.1.6) (27) in Cytoscape was used to analyze the connectivity of proteins in the PPI network (27). The results were sorted by degree and the top five genes were identified as core candidates for subsequent analysis.

Molecular docking was performed using AutoDock Vina (version 1.1.2) (28) and PyMOL (version 2.4.1) (29) to confirm the binding activity of the core components of GEB to potential targets. The binding activity is expressed as binding energy. The lower the binding energy, the more stable the docking module (30,31). In general, docking energy <-4.25 kcal/mol has docking activity, docking energy <-5 kcal/mol has good docking activity, while docking energy <-7 kcal/mol has very strong docking activity (32). Briefly, protein crystal structures of key genes were first obtained from the Protein Data Bank (rcsb.org/), where water and other small molecules were removed using PyMOL and hydrogen atoms and charge manipulations were added using AutoDock. Moreover, the 3D structures of core compounds were downloaded from the PubChem database (pubchem.ncbi.nlm.nih.gov/) and imported into AutoDock to add charge and display the rotatable bonds. The core compounds were used as ligands and proteins corresponding to the core genes were used as receptors for molecular docking. Subsequently, the binding affinity between these core compounds and the core proteins was used as the evaluation criteria. The structure with the lowest docking energy (the highest binding affinity) in the output results was selected. PyMOL was used to visualize the combination of the best docking scores.

All statistical analysis was performed using R software (version 4.0.3). Volcano maps were plotted using R package ggplot2 (version 3.3.2) (33) to visualize the distribution of DEGs. The expression heat map of key genes in the GSE16561 dataset was generated using the R package Pheatmap (version 0.7.7) (34). The bubble and chord plots showing the functional enrichment results were visualized using the R package GOplot (version 1.0.2) (35). The thresholds representing the statistical significance were |log₂fold-change (FC)|>0.5 and P<0.05.

Materials. Para-hydroxybenzaldehyde, gastrodin and alexandrin were purchased from Chengdu Alfa Biotechnology Co., Ltd. (purity, ≥98%). Sodium dithionite (Na₂S₂O₄) was purchased from Shanghai Macklin Biochemical Co., Ltd. HT22 cells were obtained from Shanghai Biological Technology Co., Ltd. Primary antibodies against phosphorylated (p)-STAT3 (cat. no. AF3293), STAT3 (cat. no. AF6294) and matrix metalloproteinase (MMP)9 (cat. no. AF5228) were bought from Affinity Biosciences and primary antibodies against myeloperoxidase (MPO) (cat. no. 22225-1-AP) and β-actin (cat. no. 66009-1-Ig) were purchased from ProteinTech Group, Inc. Secondary anti-rabbit and anti-Mouse IgG (whole molecule)-peroxidase antibody were procured from Sigma-Aldrich Trading Co., Ltd.

Cell culture and drug treatment. The cells were cultured in high-glucose DMEM (cat. no. 2024059) supplemented with 10% FBS (cat. no. 04-001-1A) and 1% Penicillin-Streptomycin solution (cat. no. 2114092) (these were all purchased from Biological Industries) for 24 h in a constant temperature incubator at 37˚C with 5% CO₂, then treated at 37˚C with 10 mM Na₂S₂O₄ in the glucose-free medium (cat. no. 2029548; Biological Industries) for 2 h for oxygen and glucose deprivation. Subsequently, the medium was replaced with high-glucose DMEM and reoxygenated at 37˚C for 2 h, which was defined as the oxygen-glucose deprivation/reperfusion model (OGD/R). The cells were randomly divided into Control, OGD/R and drug pretreatment groups. The drug formulation was solubilized in 0.1% DMSO and diluted with DMEM. First, cells were incubated with alexandrine (2.5, 5.0, 10.0, 20.0, 40.0 and 80 µM), para-hydroxybenzaldehyde (2.5, 5.0, 10.0, 20.0, 40.0 and 80.0 µM) and gastrodin (1, 5, 25, 50, 100 and 200 µM) at 37˚C for 24 h. Subsequently, according to the experimental results and previously published protocols (36-38), the pretreatment group was incubated at the same time with the complete DMEM containing 2.5, 5.0 and 10.0 µM para-hydroxybenzaldehyde, 2.5, 5.0 and 10.0 µM alexandrin or 1, 5 and 25 µM gastrodin for subsequent experiments.

Cell viability analysis. Cell viability following treatment was measured by Cell Counting Kit (CCK)-8 assay (Beijing Solarbio Science & Technology Co., Ltd.). Following 2 h reoxygenation, 10 µl CCK-8 solution was added to all groups in a 96-well plate and incubated for 4 h. The optical density (OD) was measured at 450 nm on an enzyme-labeling instrument (Thermo Fisher Scientific, Ltd.). The cell viability rate was calculated as follows: Cell viability rate=[(experimental group OD₄₅₀-blank OD₄₅₀)/(control group OD₄₅₀-blank OD₄₅₀)] x100.

Western blot analysis. Cells were treated with Lysis buffer (PSMF:100 mM RIPA, 1:100) (cat. no. 042121210730; 051021210825 Beyotime Biotech Inc) for lysis on ice for 20 min. The lysate was collected by centrifugation at 14,300 x g and 4˚C for 10 min and protein concentration was determined in the supernatant by the BCA method. Total protein (80 µg/lane) was separated by SDS-PAGE on a 6% gel and transferred onto a PVDF membrane. Subsequently, the membrane was blocked with 5% bovine serum albumin at room temperature for 1 h. Membranes were incubated with primary antibody against p-STAT3 (1:1,000), STAT3 (1:1,000), MMP9 (1:1,000), MPO (1:1,000) and β-actin (1:25,000) at 4˚C overnight. Following primary antibody incubation, the membranes were incubated with anti-Rabbit or Anti-Mouse secondary antibody (1:5,000) for 2 h at room temperature. The immunoreactive bands were developed using enhanced chemiluminescence reagent (cat. no. A38555; Thermo Fisher Scientific) and images were captured in the Bio-Rad ChemiDoc™ XRS gel imaging system (Bio-Rad Laboratories, Ltd.). ImageJ Lab™ V4.0 software (Bio-Rad Laboratories, Ltd.) was used to analyzed.

Experimental statistical analysis. GraphPad Prism 9.0.0 software (GraphPad Software, Inc.) was used for statistical analysis. Normal distribution was used Kolmogorov-Smirnov test, and homogeneity of variance was used Brown-Forsythe test. All data conformed to normal distribution and homogeneity of variance. Therefore, multiple comparisons were performed using one-way ANOVA followed by Bonferroni's post hoc test. P<0.05 was considered to indicated a statistically significant difference. Data are presented as the mean ± standard error of the mean. Cell viability experiments were repeated 6 times, and western blot experimental was repeated 3 times.

A total of 21 active ingredients of GEB were retrieved from the ETCM database. Subsequently, the 2D structure, molecular formula, and relative molecular weight of each ingredient was obtained from the TCMSP database. The drug-like properties of active ingredients of GEB, including LogD, AlogP, oral bioavailability (OB), drug-likeness (DL), Caco-2 permeability, and BBB transmission, were also obtained (Table I).

The targets of the aforementioned 21 active ingredients were predicted using the ETCM and SwissTargetPrediction databases. Notably, only 13 [sitosterol, vanillin, palmitic acid, sucrose, succinic acid, alexandrin, para-hydroxybenzaldehyde, citric acid, protocatechuic aldehyde, bis(4-hydroxybenzyl)ether mono-β-D-glucopyranoside, bis(4-hydroxybenzyl)ether, 4,4'-dihydroxydiphenyl methane and gastrodin] active ingredients were able to predict targets in the aforementioned databases. Following de-duplication, 245 potential targets of GEB were retrieved from the ETCM database (Table SI). Moreover, 571 potential targets were obtained from SwissTargetPrediction database (Table SII). Finally, the potential targets were obtained by merging the two databases and de-duplicated to obtain the unique results, resulting in 752 targets for the active ingredient of GEB (Table SIII).

Using IS as the keyword, 3,855, 1,159, 10, 106 and 15 IS targets were retrieved from GeneCards (Table SIV), DisGeNET (Table SV), DrugBank (Table SVI), OMIM (Table SVII) and TTD (Table SVIII) databases, respectively. A total of 4,158 potentially disease-associated genes were obtained by combined analysis (de-duplication to retain unique results; Table SIX).

In addition, 476 IS-associated DEGs between 39 IS samples and 24 healthy control subjects were identified based on GSE16561 dataset. Of these, 364 genes were upregulated in IS and 112 were downregulated in IS samples (Fig. S2; Table SX).

Venn analysis of 752 targets of GEB active ingredients, 4,158 potential disease-associated genes and 476 IS-associated DEGs yielded 32 key genes for GEB in treating IS (Fig. 2A; Table SXI). Heat map demonstrated the expression patterns of the 32 key genes between patients with IS and healthy control subjects in the GSE16561 dataset (Fig. 2B).

To explore the biological functions of key genes, GO analysis was performed (Table SXII). In the BP category, ‘response to oxidative stress’, ‘neutrophil degranulation’ and ‘neutrophil activation involved in immune response’ were the three most enriched terms (Fig. 3A) and also involved most key genes (all counts, 10). These key genes were also associated with ‘regulation of blood circulation’ (count, 6). A total of nine entries in the CC category were enriched, with ‘secretory granule membrane’ the most significant and the term with the most key genes involved (count, 6; Fig. 3B). Among the 30 MF category terms that were significantly enriched, ‘carbohydrate binding’ was significant (count, 8; Fig. 3C). KEGG analysis suggested that eleven pathways were associated with key genes (Fig. 3D). ‘Toxoplasmosis’, ‘arachidonic acid metabolism’, ‘lipid and atherosclerosis’, ‘galactose metabolism’ and ‘PD-L1 expression and PD-1 checkpoint pathway in cancer’ were the top five KEGG pathways (Table SXIII).

Based on the aforementioned results, the active ingredients of GEB and 32 key genes were matched. A herb-compound-target-disease network was constructed, which consisted of 48 nodes (IS, one herb, 14 active compounds and 32 target genes) and 234 edges (Fig. 4A).

Subsequently, all 32 key genes were uploaded into STRING to construct the PPI network. At a confidence level of 0.4, discrete proteins were hidden to obtain a PPI network consisting of 26 genes with 52 edges (Fig. 4B; Table SXIV). The degree value of each gene was calculated using the CytoNCA plugin; toll-like receptor 4 (TLR4), STAT3, MPO, PTGS2 and MMP9 were the top five genes in terms of degree value (Fig. 4C; Table SXV) and were selected as core genes for downstream analysis.

TLR4, STAT3, MPO, PTGS2 and MMP9 may be core potential targets of GEB for treating IS. Therefore, the active components and five core gene targets were matched and the docking potential was examined. The binding energy for all compounds and core genes was in the range of -8.2~-4.4 kcal/mol; the larger the absolute value of binding energy, the stronger the molecular docking effect (30,39). Only three core genes, STAT3, MPO and MMP9 had binding energy >-5 kcal/mol with active components of GEB and showed good docking activity (Table II). The present results suggested that the compounds in GEB interacted with core genes to counteract IS.

Molecular docking illustrated the interaction between core genes and compounds; all five compounds interacted with the corresponding targets, primarily via hydrogen bonds. Specifically, palmitic acid formed three hydrogen bonds, primarily with residues LEU-119, PHE-122 and PRO-145 on the TLR4 protein (Fig. 5A). SER-69 residue on STAT3 protein was bound to alexandrin through one hydrogen bond (Fig. 5B). Para-hydroxybenzaldehyde formed two hydrogen bonds with two residues (GLN-257 and PHE-498) on the MPO protein (Fig. 5C). Palmitic acid formed two hydrogen bonds with residue HIS-3375 on the PTGS2 protein (Fig. 5D). Finally, gastrodin was bound to the MMP9 protein by nine hydrogen bonds (Fig. 5E).

The present results indicated that gastrodin (1, 5 and 25 µM) did not cause notable cytotoxicity. At concentrations of gastrodin ≥50 µM, cell viability significantly decreased (Fig. 6A). In addition, para-hydroxybenzaldehyde and alexandrin did not affect viability of HT22 cells (Fig. 6B and C). OGD/R significantly decreased viability of HT22 cells. Compared with the OGD/R group, the gastrodin 1 µM group improved cell viability, although not significantly. The para-hydroxybenzaldehyde 2.5 µM group and gastrodin 5.0 µM group increased cell viability increased. Conversely, the gastrodin 25 µM and para-hydroxybenzaldehyde 5 and 10 µM groups increased cell viability significantly (Fig. 6D and E). The cell viability was significantly increased in the 2.5, 5.0 and 10.0 µM groups of alexandrin (Fig. 6F). Thus, it may be speculated that para-hydroxybenzaldehyde, gastrodin and alexandrin promoted viability of HT22 cells.

Based on network pharmacology and molecular docking results, western blotting analysis was performed to confirm the regulatory effects of the three active components of GEB, namely alexandrin, para-hydroxybenzaldehyde and gastrodin, on expression of STAT3, MPO and MMP9, which are the core targets, in HT22 cells. The present results showed that compared with the control, the OGD/R group did not show any significant change in the expression of p-STAT3 protein (Fig. 7A) but upregulated protein expression of MPO and MMP9 (Fig. 7B and C). However, expression of p-STAT3 in the alexandrin group was significantly upregulated compared with that in the OGD/R group (Fig. 7A), while total protein expression of STAT3 remained unchanged. MPO expression in the para-hydroxybenzaldehyde group was significantly lower than that in the OGD/R group (Fig. 7B). Moreover, the expression of MMP9 was significantly lower in the gastrodin group than in the OGD/R group (Fig. 7C).