We first obtained active compounds and corresponding targets for GXV from our previous study (17). In another previous study (6), we identified the effective components in GXV by ultra-performance liquid chromatography-quadrupole time-of-flight high-resolution mass spectrometry (UPLC-Q-TOF/HRMS E). In this study, we supplemented the targets of these effective components in SwissTargetPrediction by the structure of each effective component (18), as described previously (19). The targets of VR were obtained from our previous study (17). Taking “oxidative stress” as a keyword, we obtained oxidative stress-related targets from GeneCards, a searchable and integrative database that provides comprehensive information on all annotated and predicted human genes (20).

All targets were standardized in Universal Protein (UniProt) (21), and then taking the intersection of the targets in GXV, VR, and oxidative stress, these shared targets were considered as key therapeutic targets related to GXV against oxidative stress of VR, as described previously (22), and were visualized by Venn diagram. We also constructed a GXV-component-target-VR (G-C-T-V) network to visualize the relationships of targets between GXV and VR. Moreover, the cytoHubba plugin (23) in Cytoscape (24) was applied to evaluate the multiple centralities. A protein–protein interaction (PPI) network of all shared targets was then constructed with all genes as the background (25). The network diagram was completed in the string database (26) with the organism option was set to Homo sapiens (Human) and medium confidence was more than 0.400. Then, we used the MCODE plugin (27) in Cytoscape (24) to cluster the PPI network based on the topology to find densely connected regions.

DAVID Bioinformatics Resources (28) was utilized to conduct functional enrichment analysis {including Gene Ontology [GO] terms (29) and Kyoto Encyclopedia of Genes and Genomes [KEGG] pathways (30)} of all shared targets. We also used Homo sapiens (Human) as the background. To avoid over counting duplicated genes, based on corresponding DAVID gene IDs, we calculated the Fisher Exact statistics to remove all redundancies in original IDs. The threshold of EASE score, a modified Fisher Exact P value, was applied for the evaluation of gene-enrichment analysis. Fisher Exact P value range from 0 to 1, ≤0.05 was considered strongly enriched, and equal to 0 were considered perfect enrichment. All results had to pass the thresholds (EASE ≤ 0.05 and the number of genes annotated with a GO term or KEGG pathway was more than or equal to 5) to ensure only significant enrichment terms were displayed. The top 20 results were visualized by the ggplot2 package in R (31). KEGG Mapper is a collection of tools for KEGG mapping, which could realize the visualization of the situation of the certain gene set in the corresponding signaling pathways or other high-level enrichment features (32). We used this tool to mapper specific signaling pathways.

In addition, the ClueGO plugin (33) in Cytoscape (24) was used to create and visualize a functionally grouped network of terms in cluster networks from the PPI network.

Refer to our previous research results (17), we selected two key targets from established “two GXV pathways” (TGF-β1 and Caspase-3) and their corresponding compounds from the established G-C-T-V network (MOL000358 [Beta-Sitosterol], MOL000006 [Luteolin], MOL007154 [Tanshinone IIA], MOL002714 [Baicalein]) to validate computationally by molecular docking. Detailly, we used receptor-ligand molecular docking to assess these interactions. The structures of protein crystal and compound were obtained from Protein Data Bank (34) and PubChem (35), respectively. The binding energy was calculated by AutoDockTools (36) and the docking was visualized by Discovery Studio.

H9c2(2-1) cells were incubated with H₂O₂ (300 μM) for 24 h to establish the oxidative stress model and then treated with GXV (1 g/L) for another 24 h (17). Supernatants and cells were harvested and the antioxidant activities were investigated via malondialdehyde (MDA, NJJCBIO, Nanjing, China), superoxide dismutase (SOD, mlbio, Shanghai Enzyme-linked Biotechnology Co., Ltd., Shanghai, China), lactate dehydrogenase (LDH, NJJCBIO, Nanjing, China), catalase (CAT, Jiangsu Meimian Industrial Co., Ltd., Yancheng, China), total antioxidant capacity (T-AOC, Beyotime, Shanghai, China), and glutathione peroxidase (GSH-PX, NJJCBIO, Nanjing, China) according to the instructions of the manufacturers.

We obtained 119 active compounds and corresponding 196 targets for GXV from our previous study (17). Moreover, we supplemented 77 targets for 15 effective components in GXV. After integrating all the targets, we obtained 251 GXV-related targets. We got 11,425 known therapeutic targets for VR from our previous study (17). A total of 9,727 targets were identified as stress-related oxidative.

The Venn diagram showed that there were 206 shared targets among 251 GXV-related targets, 11,425 VR-related targets, and 9,727 oxidative stress-related targets (Supplementary Figure S1). These 206 shared targets were considered as key therapeutic targets related to GXV against oxidative stress of VR. Subsequently, we built the G-C-T-V network to illustrate the potential mechanism of GXV acting on VR. To simplify the network, we only used the shared targets and their related compounds to construct the G-C-T-V network, which is composed of 314 nodes and 1,503 edges (Figure 2). Moreover, we revealed the most important nodes in this network. Multiple centralities demonstrated that MOL000006 (Luteolin), MOL003896 (7-Methoxy-2-Methyl Isoflavone), MOL007154 (Tanshinone IIA), 5705531 (Dihydrocatalpol), MOL007100 (Dihydrotanshinlactone), MOL000358 (Beta-Sitosterol), and MOL002714 (Baicalein) were the most important compounds in our G-C-T-V network (Table 1); while PTGS2, NCOA1, SCN5A, ADRB2, NCOA2, PTGS1, CHRM1, CHRNA7, F2, RXRA, ACHE, and CA2 were the most important targets in our G-C-T-V network (Table 2). We have reasons to believe that these compounds and targets played key roles in GXV for treating VR.

Moreover, in this study, we also constructed a PPI network of the 206 shared targets, which consisted of 206 nodes and 6,116 edges (Supplementary Figure S2). This means that the proteins have more interactions among themselves than a random set of proteins of similar size drawn from the genome. Such an enrichment indicates that the proteins are at least partially biologically connected as a group and the network highlights the complexity of interactions among proteins. We later applied the PPI network with known interactions supported by the experimentally determined evidence hiding disconnected nodes in the PPI network (Figure 3). Through the MCODE plugin, we obtained 4 cluster networks with the highest clustering scores (Table 3, Figure 4).

To clarify the oxidative stress mechanism of GXV in treating VR, we conducted GO and KEGG functional enrichment analysis for 206 shared targets. GO enrichment items were classified into three functional groups: molecular function (Figure 5A), biological process (Figure 5B), and cellular component (Figure 5C). The results indicated that numerous molecular functions were involved in enzyme binding, protein homodimerization activity, drug binding, protein binding, protein heterodimerization activity, identical protein binding, transcription factor binding, ubiquitin-protein ligase binding, type I transforming growth factor beta receptor binding, and transforming growth factor beta receptor, pathway-specific cytoplasmic mediator activity (Figure 5A). The results indicated that numerous biological processes were related to the treatment of VR, including response to the drug, response to hypoxia, aging, negative regulation of the apoptotic process, positive regulation of cell proliferation, positive regulation of transcription from RNA polymerase II promoter, positive regulation of transcription, DNA-templated, positive regulation of smooth muscle cell proliferation, transforming growth factor beta receptor signaling pathway, and cell proliferation (Figure 5B). The results indicated that numerous cellular components were involved in the plasma membrane, cytosol, membrane raft, extracellular space, an integral component of the plasma membrane, cell surface, transcription factor complex, SMAD protein complex, receptor complex, and nucleoplasm (Figure 5C). The top 20 enriched KEGG pathways for the 206 shared targets are shown in Figure 5D. Among these pathways, the HIF-1 signaling pathway, FoxO signaling pathway, neuroactive ligand-receptor interaction, calcium-signaling pathway, PI3K-Akt-signaling pathway, cell cycle, and TGF-β signaling pathway were involved in the development and pathogenesis of VR. In a word, these enrichment findings support the potential pharmacological mechanism of GXV in the treatment of VR. Importantly, the “two GXV pathways” (TGF-β signaling pathway and apoptosis pathway) we obtained before (17) have been verified here. We reconstructed the PPI network diagram of the targets related to the “two GXV pathways” and visualized the pathways (Supplementary Figure S3).

Through GO biological process enrichment of 4 cluster networks from the PPI network, we found that cluster 1 enriched in negative regulation of G1/S transition of the mitotic cell cycle, positive regulation of fibroblast proliferation, and histone phosphorylation; cluster 2 enriched in Notch receptor processing and peptidase activator activity; cluster 3 enriched in SMAD protein signal transduction, mesenchyme morphogenesis, striated muscle cell proliferation, positive regulation of pri-miRNA transcription by RNA polymerase II, and production of miRNAs involved in gene silencing by miRNA; and cluster 4 enriched in activation of cysteine-type endopeptidase activity involved in the apoptotic signaling pathway, positive regulation of pri-miRNA transcription by RNA polymerase II, nuclear receptor activity, and regulation of insulin receptor signaling pathway (Figure 6, Supplementary Table S1).

The PDB entry codes for TGF-β1 and Caspase-3 are 6P7J (37) and 5I9B (38), respectively, and the PubChem IDs of Beta-Sitosterol, Luteolin, Tanshinone IIA, and Baicalein are 222284, 5280445, 164676, and 5281605, respectively. The compounds from GXV likely interacted strongly with the identified key targets (Figure 7A). The three-dimensional structural diagrams of molecular docking are shown in Figures 7B–F.

H9c2(2-1) cells incubated with 300 μM H₂O₂ had increased LDH and MDA and decreased SOD, GSH-Px, CAT, and T-AOC (Figure 8) indicating that the level of oxidative stress was elevated. Administration of GXV could reverse the elevated oxidative stress (Figure 8). Together, we concluded that GXV could consider as an antioxidant.