We obtained TMGT lyophilized powder from the Research Center of Traditional Chinese Medicine Affiliated Hospital of Changchun University of Chinese Medicine (Changchun, China). Each medicinal material used for experiments was tested by the Chinese Medicine Appraisal Laboratory of Changchun University of Chinese Medicine. The final yield of the aqueous extract of TMGT was 43.19%. The aqueous extract of TMGT was stored at −80°C before use. TMGT powder (prepared as previously described) was dissolved in 50% methanol made with ultrapure water to generate a solution of concentration 70 mg/ml for use in HPLC analysis (Deng et al., 2020). Eight different batches of TMGT powder were subjected to chromatographic separation using a ZORBAX 300 Exten-C18 column (4.6 × 250 mm, 5 μm, Agilent, CA, United States) then were analyzed for chemical composition using an HPLC system (LC-2030c, Shimadzu Corporation, Kyoto, Japan) coupled with an ultraviolet-visible (UV) detector (Lu et al., 2020). Acetonitrile (Mobile phase A) and 0.1% phosphoric acid in water (Mobile phase B) served as HPLC mobile phases (Supplementary Table S1). The flow rate was 1.0 ml/min at 30°C, and the detection wavelength was set to 254 nm. Gastrodin, parishin A, parishin B, parishin C, chlorogenic acid, ferulic acid, quercetin, and p-hydroxybenzyl alcohol served as known reference standards for use in compositional analysis of TMGT based on retention time (Figure 1A, Shanghai Yuanye Biotechnology, Shanghai, China). Chemical constituents corresponding to eight major HPLC peaks obtained for TMGT extract were identified based on comparisons of retention times to those of known standards (Figure 1B). Relative proportions of gastrodin, p-hydroxybenzyl alcohol, chlorogenic acid, parishin B, parishin C, ferulic acid, parishin A, and quercetin were 9.929, 0.382, 1.806, 0.078, 0.035, 0.045, 0.060, and 0.005%, respectively. Quality assessments of TMGT preparations, which were conducted using Chinese Medicine Chromatographic Fingerprint Similarity Evaluation System (2012 Edition) guidelines, demonstrated a high degree of overall TMGT batch-to-batch reproducibility, as reflected by the results of HPLC fingerprint analysis (Figure 1C). In fact, for all eight TMGT batches, the overall similarity exceeded 98%, thus demonstrating batch-to-batch consistency and reproducibility of TMGT preparations generated in this study.

Keywords of “GouTeng,” representing GT associated active ingredients, and “TianMa,” representing TM-associated active ingredients, were used to search TCMSP (http://tcmspw.com/) and BATMAN-TCM (http://bionet.ncpsb.org.cn/batman-tcm/) databases, respectively (An et al., 2020). Next, we screened the GeneCards database (https://auth.lifemapsc.com/) to obtain CI-related proteins, and transcriptomic information (Wang Y. et al., 2020; Chen et al., 2021). Based on the abovementioned screening results, we drew a Venn diagram to display the interaction networks of TMGT-associated components and CI-related targets using Venny (2.1.0) (http://bioinformatics.psb.ugent.be/webtools/venn/) (Wang et al., 2021). Finally, we constructed the interaction network of chemical composition-targets of action, and the potential targets of TMGT against CI using Cytoscape (3.6.1) (Xu et al., 2021).

GO enrichment analysis was performed to predict possible functions of target genes using the DAVID 6.8 database (https://david.ncifcrf.gov/). KEGG (Kyoto Encyclopedia of Genes and Genes) pathway enrichment was carried out to predict pathways associated with proteins corresponding to genes predicted using OmicShare tools (https://www.omicshare.com/tools/) (Xu et al., 2020). Homo sapiens was selected as the species item, and p < 0.05 was chosen as the screening condition.

Interactions between bioactive ingredients of TMGT and CI-associated target genes were verified through the construction of molecular docking models using the Maestro 11.9 software (Li et al., 2021). In brief, the crystal structures of the target proteins corresponding to predicted target genes were first downloaded from the RCSB Protein Data Bank database (https://www.rcsb.org/) then were imported into Maestro. Next, water molecules and irrelevant ligands in the crystal structure were systematically removed by adding hydrogen atoms using OPLS3e force field, then the adjusted crystal structures were optimized by determining their energy-minimized conformations. As described previously, structures of active ingredients of TGMT were obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) and were saved in SDF format (Yuan et al., 2021). Using the LigPrep module of Maestro, the energy-minimized conformation was selected using force-field OPLS3e and Epik set to pH 7.0 ± 2.0 to produce the candidate ligand for docking. The setting “Dock ligands with length” of the Glide module was defined as 15 Å for use in creating a binding pocket. Finally, the XP docking method in the ligand docking module was selected for docking analysis, and the RMSD (root mean square deviation) value was chosen to evaluate the displacement of the two state coordinates before and after docking. The Glide Score was selected to evaluate the binding of the ligand and the receptor, and the PyMOL software was chosen to visualize the top-ranked compounds.

The BV2 and PC12 cell lines were purchased from The American Type Culture Collection (ATCC, Rockefeller, Maryland, United States). BV2 cell lines were cultured in MEM with 10% fetal bovine serum (FBS, CLARK Bioscience, Claymont, DE, United States), while PC12 cell lines were cultured in DMEM (Gibco, New York, NY, United States) supplemented with 7.5% FBS (CLARK Bioscience) and 2.5% horse serum (Gibco). Both cell lines were cultured in a medium supplemented with penicillin-streptomycin (each at 100 Units/ml, MedChemExpress, Shanghai, China) at 37°C in an incubator with humidified atmosphere and 5% CO₂. Before they were used in the experiments, PC12 cells were incubated with 50 ng/ml nerve growth factor (NGF, Gibco) for 24 h to induce neurite formation in order to establish the neuronal cell model (Hu et al., 2018; Huang et al., 2018). To establish OGD-induced cell injury models, differentiated PC12 cells were maintained in a serum-free, glucose-free medium and incubated under oxygen-free conditions (95% N₂, 5% CO₂) in a BioSpa automated incubator (BioTek, United States) for 0–8 h. Cells of the control group were cultured in a normal medium, while cells in OGD or LPS model groups were subjected to OGD or LPS exposure, respectively. For treatment groups, TMGT or quercetin (40 μM) extracts were dissolved in DMSO to create high-concentration stock solutions that were further diluted to generate solutions of various concentrations according to the requirements of individual experiments. Next, cells were pretreated with different concentrations of TMGT or quercetin for 24 h followed by exposure to OGD conditions or LPS (Le et al., 2020).

PC12 or BV2 cells were seeded into wells of a 96-well plate at a density of 3 × 10³ cells/well, then a 100-μl DMEM/MEM medium was added to each well. PC12 cells were treated with TMGT or quercetin (PHR1488, Sigma-Aldrich) after the cells grew to 80% confluence in each well followed by incubation for 24 h. Untreated cells cultured under OGD conditions (for the same amounts of time as for treated cells) served as controls. Next, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, 5 mg/ml, Sigma-Aldrich) was added to cells to determine cell viability according to the instructions provided with the kit. A total concentration of 10% MTT (v/v) was added to cells in different groups, then DMSO (150 μl) was added to each well to dissolve the cells (Babaei et al., 2021). The OD value of each well was detected using a microplate reader (TECAN Infinite M200pro, ZH, Switzerland) at a wavelength of 490 nm.

PC12 cells were seeded into wells of 96-well plates at a density of 3 × 10³ cells/well and exposed to OGD conditions for 2 h followed by no treatment or treatment with TMGT or quercetin for 24 h. LDH release was measured using an assay kit to detect levels of LDH in supernatants of PC12 cells (Jin et al., 2019). Based on absorbance values at 450 nm measured according to the manufacturer’s instructions.

Supernatants of OGD-induced PC12 cells and LPS-induced BV2 cells treated with different concentrations of TMGT or quercetin were collected. ELISAs were performed to evaluate the levels of the inflammatory factors (IL-6, IL-1β, and TNF-α) in PC12 and BV2 cells according to the manufacturer's instructions (He et al., 2017). ELISA kits to detect IL-6 (# BMS603-2, # ERA31RBX5), IL-1β (# BMS6002TEN, # BMS630TEN), and TNF-α (# 88-7324-86, # KRC3011C) were obtained from Thermo Fisher Scientific (Waltham, MA, United States).

PC12/BV2 cells were treated with RIPA lysate buffer containing a protease inhibitor (Roche, Basel, Switzerland) and a phosphatase inhibitor (Roche). Concentrations of proteins within supernatants were determined using a BCA kit (Roche). Proteins (35 μg per lane) were loaded into wells of SDS-PAGE gels followed by electrophoresis to separate proteins by molecular weight (8–10% SDS-PAGE separation gel and a 5% SDS-PAGE concentrated gel). Next, proteins were transferred to polyvinylidene fluoride membranes (PVDF, Roche) using a wet transfer method. PVDF membranes bound to proteins were blocked by immersion in 5% (w/v) bovine serum albumin (BSA), then membranes were probed with antibodies anti-IκBα, ab76429; anti-HIF-1 alpha, ab179483, anti-IKB alpha (phospho S36), ab133462; Abcam, Cambridge, MA, United States; anti-NF-κB p65, #8242; anti-phospho-NF-κB p65 (Ser536), #3033; anti-β-Actin, #4970 (Cell Signaling Technology, Beverly, MA, United States); and anti-PPAR gamma, bs-0530R (Bioss, Beijing, China) as dilutions of 1:1,000 followed by incubation at 4°C overnight. Next, membranes were incubated with secondary antibody (HRP) at room temperature for 1 h. An enhanced ECL reagent was added to membranes to visualize target proteins attached to membranes (Qi et al., 2019). Antibody-binding to target protein as a measure of the relative protein expression level was calculated by measuring areas and intensities of bands using AlphaView SA.

GraphPad Prism 6.0 software (San Diego, California, United States) was used for all data analysis. Experiments in this study were repeated three times. The one-way ANOVA test (Turkey’s post hoc) was chosen to calculate the significance of multiple comparisons. The level of significance was set to p < 0.05.

Screening of published literature for information regarding TMGT active ingredients using the keywords “TianMa” and “GouTeng” led to the identification of a total of 46 active ingredients (Supplementary Table S2). These ingredients were subsequently verified through searches of TCMSP and BATMAN-TCM databases based on the criteria that included oral bioavailability OB ≥30% and drug-like DL ≥0.18. Meanwhile, a total of 3,268 genes were searched for CI-related targets after entering “Cerebral Infarction” in GeneCards, then the results were subjected to VENNY analysis, which led to the identification of 36 common targets of CI and TMGT, as shown in Figure 2A. Ultimately, the selected 46 compounds and 36 targets were imported into Cytoscape (3.6.1) to generate an interaction network (Figure 2B) containing 69 nodes and 169 sidebars.

Target genes identified from analysis of interaction networks (mentioned in Figure 3) were input into STRING (https://string-db.org/) to generate the protein–protein interaction (PPI) network. Next, indirect connections were made based on reported functions of selected target proteins or protein characteristics obtained from related published studies (Figure 3A), resulting in the creation of a PPI network containing 70 nodes, with neighboring nodes >20 in the PPI network listed (Figures 3B,C). Among them, JUN protein had the highest score due to its connections with 60 adjacent nodes. Putative close relationships among adjacent nodes were predicted using a polar coordinate histogram. We speculate that targets identified through this analysis may play important roles in mechanisms underlying beneficial TMGT effects on CI-induced CNS cell damage.

For GO enrichment analysis, a total of 3,848 GO entries were obtained for the 46 common targets. Results showed that 83.63% of targets were functionally associated with “Biological Process (BP),” 7.02% with “Cellular Component” (CC), and 9.36% with “Molecular Function” (MF). However, the proportions of the three abovementioned categories were listed as 54.42, 3.38, and 5.46% for results with p values <0.05. The top 20 molecules were selected for use in drawing a bubble chart (Figure 4). The results suggested that the top 20 molecules were mainly related to neuron projection terminus, neurotransmitter binding, and positive regulation of multicellular organismal process. Meanwhile, KEGG pathways were shown to consist of aggregates of pathway maps based on current knowledge of molecular interaction networks (Su et al., 2018). Followed by the KEGG enrichment pathway, 169 genes functionally associated with key pathways were obtained, among which 46.75% exhibited differential expression characters. The top 20 selected pathways are shown in Supplementary Table S3. These results revealed that key pathways associated with common target proteins were mainly related to functional terms neuroactive ligand–receptor interaction, the HIF-1 signaling pathway, and the NF-kappa B signaling pathway (Figures 4E,F).

Molecular docking technology is a common method for exploring functions/effects of drugs (Li et al., 2021). Here we chose core proteins for docking studies, including PPARG, HIF1A, and NFKBIA, which were identified from the PPI network as research objects (Figure 5). TMGT active ingredients that interacted with these three core proteins were molecularly docked. Figure 5 shows the 2D interaction molecular docking diagram. From Figures 5A–C, it can be seen that quercetin could form hydrogen bonds with several amino acid residues of HIF1A (Figure 5A), NFKBIA (Figure 5B), and PPARG (Figure 5C) to form stable spatial structures (with binding energies <-5 kcal/mol, mentioned in Supplementary Table S4). Thus, quercetin may be a key bioactive core ingredient of TMGT. Meanwhile, PPARG also bound stably to vincoside lactam_qt (Figure 5D), yohimbine (Figure 5E), and tetrahydroalstonine (Figure 5F). PPARG bound most strongly with the indole alkaloid vincoside lactam_qt (with binding energies = 11.54 kcal/mol), a component commonly incorporated in anti-inflammatory and antivasoconstriction drugs, thus suggesting that TMGT may exert beneficial blood vessel dilation and anti-inflammatory effects. Notably, the root-mean-square deviation (RMSD) values of all of the abovementioned molecular docking models were <2Å, thus demonstrating the reliability of these models.

As shown in Figures 6A,B, we first conducted MTT assays to confirm that TMGT treatment did not decrease the viability of PC12 and BV2 cells. Results showed that 24-h TMGT treatment concentrations ranging from 3.125 to 400 μg/ml had no significant effect on the survival of PC12 and BV2 cells. To establish OGD models, PC12 cells were incubated under anaerobic (95% N₂, 5% CO₂) and low-glucose conditions for different amounts of time. As shown in Supplementary Figure S1A, cells cultured under OGD conditions exhibited decreased cell viability in a time-dependent manner based on OGD exposure duration. Moreover, 2-h cell culture under OGD conditions obviously decreased cell viability (to 53.77%) as compared to control group viability. As shown in Figure 6C, OGD-induced decline in cell viability was reversed by cell treatment with various concentrations of TMGT. Specifically, cell viability could be greatly increased by as much as 1.31-fold by treatment with a TMGT concentration of 50 μg/ml, suggesting protective effects of TMGT against OCD-induced PC12 cell damage. Also, OGD exposure of PC12 cells led to greatly increased LDH release by the cells as compared to that of control cells. By contrast, dramatically decreased LDH release by PC12 cells (by 0.67- to 0.89-fold) was observed after TMGT treatment (Figure 6C), suggesting that TMGT treatment protected PC12 cells from OCD-induced damage.

In general, the occurrence of diseases is accompanied by inflammatory biological processes. Based on the aforementioned MTT results and the results of experiments performed using PC12 cells, we found that 2-h culture of cells under OGD conditions significantly increased IL-6 and TNF-α release by PC12 cells (Supplementary Figures S2A,B). We also conducted ELISAs to assess levels of inflammatory factors in PC12 cells under TMGT treatment. Figures 7A–C show that quantities of IL-6, IL-1β, and TNF-α secreted by PC12 cells cultured under OGD conditions were all significantly increased (relative to quantities secreted by control cells without OGD exposure); however, markedly lower quantities of all three factors were secreted by cells that had been pretreated with TMGT prior to culture under OGD conditions.

To establish a microglial inflammation model, BV2 cells were treated with different concentrations of LPS for various amounts of time, then IL-6 and TNF-α release from the cells was measured via ELISAs. Results showed significantly increased release of IL-6 and TNF-α after cells were exposed to 100 ng/ml of LPS for 6 h (Supplementary Figures S2C,D), with decreased release of IL-6, IL-1β, and TNF-α observed in cells pretreated with TGMT prior to LPS exposure in a TMGT-dose-dependent manner (Figures 7D–F). Taken together, these findings suggested that TMGT treatment may inhibit CI-induced neuroinflammation.

Based on abovementioned network pharmacology results, the HIF-1α/PPARγ/NF-κB signaling pathway may be the key pathway targeted by TMGT to protecting effects against CNS damage induced by CI. Therefore, we focused on this signaling pathway in our subsequent investigation of signaling-based mechanisms underlying beneficial TMGT effects on OGD-exposed PC12 cells and LPS-exposed BV2 cells. Previous evidence obtained in other studies had pointed out that modulation of PPARγ/NF-κB signaling pathway activity prevented neuronal damage after craniocerebral injury by suppressing intracellular inflammation (Song et al., 2018; Feng et al., 2020). In this work, we found, as shown in Figure 8, that levels of PPARγ were significantly decreased, while p-IκBα levels were markedly increased either in OGD-induced PC12 cells or in LPS-induced BV2 cells. After TMGT treatment, levels of activated p-P65, HIF-1α, and IKBα were all significantly altered in a TMGT concentration-dependent manner relative to model group levels, in accordance with predicted pharmacology network analysis results. Thus, these data implied that TMGT intervention may alleviate CI-triggered neuronal and microglial cell damage and neuroinflammatory processes by regulating the HIF-1α/PPARγ/NF-κB pathway.

According to our molecular docking-based predictions, quercetin may be a key bioactive component involved in TMGT prevention of CI-induced CNS nerve damage, aligning with a previous study. That study reported that quercetin could reduce neurocytotoxicity under OGD conditions through its effects on the ERK/AKT pathway, while also exerting anti-inflammatory and anti-apoptotic effects by inhibiting the activation of MCAO-induced apoptotic pathway signal molecules caspase-3 and PARP, thereby preventing functional impairment of nerve cells (Park et al., 2018; Wang YY. et al., 2020). In a similar vein, evaluation of TMGT effects on OGD-induced PC12 cells (89.95%) revealed an increase in cell viability to 82.40% after quercetin treatment (Figure 9A), and no significant difference was found between effects observed for TMGT and quercetin on cell viability. Similar effects of quercetin in LDH levels were discovered in OGD-induced PC12 cells (Figure 9B). To further explore the relationship between quercetin and nerve inflammation, the expression levels of inflammatory factors were evaluated using two cell lines based on models of CI. As shown in Figures 9C,D, it can be seen that quercetin has an inhibitory effect on the inflammatory factors secreted by OGD-induced PC12 cells and LPS-induced BV2 cells, but it is not as effective as TMGT. Using this strategy, the expression levels of HIF-1α, P65, and PPARγ were tested in cells treated with or without quercetin to uncover the putative regulatory pathway acted on by quercetin to prevent neuronal damage and neuroinflammation (Figures 9F,G). These results suggested that quercetin treatment could prevent neuronal damage and CI-induced inflammatory processes by regulating HIF-1α/PPARγ/NF-κB pathway activity. The abovementioned results thus verify our previous hypothesis that quercetin is a key bioactive component of TMGT that may exert beneficial effects for alleviating cellular dysfunction and neuroinflammation induced by CI.