The active components of GEB were obtained by searching for ‘TIAN MA’ or ‘Gastrodia elata BI’ in the Encyclopedia of Traditional Chinese Medicine (ETCM; http://www.tcmip.cn/ETCM/index.php/Home/Index/), Bioinformatics Analysis Tool for Molecular mechANisms of Traditional Chinese Medicine (BATMAN-TCM; http://bionet.ncpsb.org.cn/batman-tcm/) and literature from China National Knowledge Infrastucture (CNKI). Subsequently, using the PubChem database (http://pubchem.ncbi.nlm.nih.gov/), the obtained active components were converted to Canonical SMILES format and imported into the SwissTargetPrediction database (http://swisstargetprediction.ch/) (21) with species set to ‘Homo sapiens’ to acquire the targets of the active components of GEB.

GeneCards (https://www.genecards.org/), OMIM (https://omim.org/), CTD (http://ctdbase.org/) and DisGeNET (http://www.disgenet.org/) databases (22,23) were used to screen potential targets associated with AD by searching for the keyword ‘Alzheimer disease’. The results from the four databases were integrated and deduplicated for the following analyses.

Limma (version 3.44.3) (24) R package was used to screen for differential genes (DEGs) between 10 AD and 13 control samples in GSE5281(18) with |log2FC|>1 and adjusted P-value <0.05. The potential therapeutic targets for the anti-AD effects of GEB were obtained by overlapping targets of active components, AD-related targets and DEGs. The drug-active component-target-AD network was then constructed and visualized using Cytoscape software (version 3.2.1) (25).

Next, potential therapeutic targets were input into STRING database (https://string-db.org). In order to build a PPI network with a confidence level >0.4, Cytoscape software (version 3.2.1) was used to visualize the PPI network and calculate the degree of each node in the PPI network to obtain the core target. Core targets were defined as the top five nodes with the highest degrees. Heat maps were created using the Pheatmap (version 0.7.7) (26) package in R for potential targets.

ClusterProfiler (version 3.18.0) (27) R package was used for GO and KEGG enrichment analyses of GEB-AD-related core targets. The GO system contains three main categories including biological process (BP), cellular component (CC) and molecular function (MF). An adjusted P-value of <0.05 was considered to indicate significant enrichment.

Core targets acting as large molecular receptors, and the corresponding active compounds, acting as small molecule ligands, were selected for molecular docking analysis. The 3D structures of the core targets were downloaded from the PDB database (https://www.rcsb.org/). The 2D structures of the corresponding active ingredients were downloaded from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/). The raw files of core targets and the corresponding active components were processed with the AutoDock (version 4.2) (28) tool and converted to PDBQT format for molecular docking. Binding activity was expressed as binding energy; the lower the bonding energy, the more stable the docking module (29). In general, a docking energy <-5 kcal/mol indicates good docking (30).

4,4'-Dihydroxydiphenyl methane (DM) and protocatechuic aldehyde (PA) were purchased from Chengdu Alfa Biotechnology Co., Ltd. (purity ≥98%). Other reagents included RNAsimple Total RNA Kit (cat. no. DP419; Tiangen Biotech Co., Ltd.), GoScript™ Reverse Transcription System (cat. no. A5000; Promega Corporation), GoTaq^® qPCR Master Mix (cat. no. A6001; Promega Corporation). Transgenic C. elegans strains CL4176 [dvIs27 (myo-3p::Aβ1-42::let-851 3'UTR) + rol-6 (su1006)] and CL2006 [dvIs2 (pCL12 (unc-54/human Aβ1-42 minigene) + pRF4)], and Escherichia coli OP50 were obtained from the Caenorhabditis Genetics Center (CGC) of the University of Minnesota (Minneapolis, USA). The network pharmacology workflow is presented in Fig. 1.

The strain CL2006 was propagated at 20˚C, and the strain CL4176 was propagated at 16˚C. All C. elegans strains were routinely propagated on solid nematode growth medium (NGM; consisting of 300 ml deionized water including 0.82 g peptone, Oxoid Limited; Thermo Fisher Scientific; 1.2 g NaCl, Tianjin Fengchuan Chemical Reagent Co., Ltd.; 5.15 g agar, Beijing Solarbio Science & Technology Co., Ltd.; autoclaved and supplemented 1 mM MgSO4, 1 mM CaCl2, Tianjin Fengchuan Chemical Reagent Co., Ltd.; 12.9 mM cholesterol solution, MACKLIN) with E. coli OP50 as a food source. C. elegans was treated with sodium hypochlorite, and synchronized larvae (L1 stage) were obtained after eggs were incubated overnight in M9 buffer (containing 22 mM KH2PO4, 42 mM Na2HPO4, autoclaved and supplemented with 1 mM MgSO4; all the reagents were from Tianjin Fengchuan Chemical Reagent Co., Ltd.) for experiment. Transgenic C. elegans strains CL2006 and CL4176 were used as AD animal models. C. elegans were randomly divided into control and drug pretreatment groups. The preparation was dissolved in 0.1% DMSO and diluted with ultrapure water to the appropriate concentration. Based on a previous study (31), various concentrations of DM (0.125, 0.25, 0.5, 1, 2 mM) and PA (0.125, 0.25, 0.5, 1 mM) were used to assess safety and efficacy in C. elegans.

CL4176, a temperature-sensitive transgenic model, has been demonstrated to inactivate the smg-1 system when the temperature increases from 16˚C to 25˚C, which leads to overexpression of Aβ in muscles and a paralysis phenotype. CL4176 nematodes synchronized to the L1 stage were transferred to 35 mm culture plates with or without OP50 and drugs, cultured at 16˚C for 36 h and then at 25˚C for transgene induction. After 24 h of culture at 25˚C, the paralyzed worms were observed every 2 h. When they could not move or respond to platinum wire stimulation, they were considered paralyzed (32).

Transgenic CL2006 worms synchronized to the L4 stage (early adult stage) were inoculated on NGM plates with or without drugs, and N2 was used as a negative control for Aβ deposition (32). Following incubation at 20˚C for 48 h, the worms were collected with M9 buffer and fixed in 4% paraformaldehyde/PBS (pH 7.4) (cat. no. BL539A; Biosharp Life Sciences) at 4˚C for 24 h. The worms were then treated with 5% β-mercaptoethanol (cat. no. M828395; Macklin, Inc.), 1% Triton X-100 (cat. no. MB2486; Dalian Meilun Biology Technology Co., Ltd.) and 125 mM Tris (pH 7.4) (cat. no. T8060; Beijing Solarbio Science & Technology Co., Ltd.) at 37˚C for 24 h. The samples were stained with 0.125% thioflavin S (cat. no. S19293; Shanghai Yuanye Bio-Technology Co., Ltd.) in 50% ethanol at room temperature for 2 min, and then with 50% ethanol. The samples were thereafter rinsed and transferred to a glass slide for observation under a laser scanning confocal microscope (LSM900; Carl Zeiss AG). The number of thioflavin S-reactive deposits in the area anterior to the pharyngeal bulb was counted.

Synchronized L1 stage CL4176 nematodes were transferred to NGM medium with or without drugs (~500 nematodes in each plate) and cultured at 15˚C for 36 h, and then at 25˚C for a further 36 h. The worms were collected in an EP tube with M9 buffer, rinsed, and the supernatant was discarded (33). Total RNA was extracted using RNAsimple Total RNA Kit, and the concentration and purity were measured with an ultra-micro spectrophotometer (SMA6000; Merinton Instrument, Inc.). cDNA was obtained by reverse transcription using GoScript™ Reverse Transcription System and a thermocycler (Veriti Applied Biosystems; Thermo Fisher Scientific, Inc.), and target mRNA was quantified using GoTaq^® qPCR Master Mix and real-time PCR (C1000 Touch PCR; Bio-Rad Laboratories, Inc.). Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) was performed under the following operating conditions: 95˚C for 10 min, followed by 40 cycles of 95˚C for 15 sec, 55˚C for 30 sec and 72˚C for 30 sec. The reaction was then cooled and maintained at 4˚C. β-actin was used as a housekeeping gene. Relative gene expression was calculated using the 2^-ΔΔCq method (34), and each reaction was repeated three times. The primers used in this assay are listed in Table SI.

GraphPad Prism software 8.0 (GraphPad Software, Inc.) was used for statistical analyses. All data are presented as the means ± SD. All experiments were repeated three times. Curves were plotted using Kaplan-Meier survival curves and statistical analyses of differences between groups in the paralysis assays were performed using the log-rank test. In addition, unpaired Student's t-test was used to compare differences between two groups. P<0.05 was considered to indicate a statistically significant difference.

A total of 21 and 23 active components in GEB were obtained from the ETCM and BATMAN-TCM databases, respectively. After integrating and de-duplicating the data, 34 active components were obtained. After removing three active components (bis(4-hydroxybenzyl)ether mono-β'-d-glucopyranoside, tris-[4-(β'-d-glucopyranosyloxy)benzyl]citrate and vanillyl alcohol) with no canonical SMILES format, the remaining 31 active components (Table I) were imported into the SwissTargetPrediction database. A total of 869 targets of the active components of GEB were obtained. By mining the GeneCards, OMIM, CTD and DisGeNET databases, and after removing duplicate data, a total of 26,255 AD-related targets were obtained.

To obtain more robust AD-related targets, a total of 1,684 DEGs between AD and control samples were identified, among which 975 were upregulated and 709 were downregulated. The red dots represent upregulated genes, the green dots represent downregulated genes, and the gray dots represent genes with no significant difference (Fig. S1). Thereafter, by overlapping 869 targets of the active components of GEB (26,255 AD-related targets and 1,684 DEGs), a total of 59 potential targets of the anti-AD therapeutic effects of GEB were obtained (Fig. 2A and Table SII). The heat map shows the expression of 59 potential therapeutic targets in AD and control samples (Fig. S2). Each square represents a gene. The greater the expression, the darker the color (red, upregulated; blue, downregulated). Each row represents the expression of each gene in different samples; the column represents expression of all genes in each sample.

Based on these results, a drug-active component-target-AD network with 92 nodes (AD, TIANMA, 31 active compounds, and 59 target genes) and 269 edges (Fig. 2B) was constructed. Furthermore, to investigate the interactions of the 59 therapeutic targets of GEB, a PPI network was constructed using the STRING database. The PPI network was composed of 47 nodes and 119 edges, revealing that therapeutic targets had a strong and complex association with each other. The color of the nodes, ranging from blue to purple, indicates the degree, ranging from small to large (Fig. 2C). GAPDH, EP300, HSP90AB1, KDM6B and CREBBP had the highest degrees, and where therefore considered to be core targets and were used for molecular docking analysis.

Next, the biological functions of the 59 potential therapeutic targets were explored, which were significantly enriched in the 163 BP, 9 CC, 38 MF and 21 KEGG pathways. The top 10 BP, MF and CC pathways were involved in gene transcription, such as ‘histone demethylation’, ‘histone modification’, ‘covalent chromatin modification’, ‘DNA-binding transcription factor binding’, ‘transcription coactivator activity’ and ‘NAD binding’ (Fig. 3A). The top 10 enriched KEGG pathways were closely associated with neurodegeneration and neuroendocrinology, including the ‘cAMP signaling pathway’, ‘HIF-1 signaling pathway’ and ‘Glucagon signaling pathway’ (Fig. 3B and Table SIII).

As mentioned above, GAPDH, EP300, HSP90AB1, KDM6B and CREBBP may be the core potential targets of GEB in the treatment of AD. GAPDH is an internal reference, and was therefore not considered further. DM and PA, two components which are at the forefront in the network diagram were investigated. These two components, with high moderate values in the network diagram, were matched with four core protein targets, and their docking potential was studied. The binding energies were between -8.12 and -5.04 kcal/mol (Table II), exhibiting good docking activity (the more negative the value, the stronger the molecular docking). Both compounds interact with the four target proteins mainly through hydrogen bonds.

The results revealed that DM forms a hydrogen bond with MET-1124 on EP300 (Fig. 4A). DM forms five hydrogen bonds with ASN-1163 and MET-1133 on CREBBP (Fig. 4B). DM forms three hydrogen bonds, with GLN-1489, SER-1184 and PRO-1181, on KDM6B (Fig. 4C). DM forms two hydrogen bonds, with GLN-18, on HSP90AB1 (Fig. 4D). PA forms four hydrogen bonds, with GLN-1077, ASP-1080 and PRO-1078, on EP300 (Fig. 4E). PA forms five hydrogen bonds, with ASN-1163 and MET-1133, on CREBBP (Fig. 4F). PA forms five hydrogen bonds, with LEU-1438, ASP-1251, TRP-1435 and PRO-1436, on KDM6B (Fig. 4G). PA forms five hydrogen bonds, with GLY-103, ASP-97 and TRP-157, on HSP90AB1 (Fig. 4H).

DM (0, 0.125, 0.25, 0.5, 1, 2 mM) and PA (0, 0.125, 0.25, 0.5 mM) had no obvious toxicity. When the concentration of DM exceeded 0.5 mM, the effect of delaying C. elegans paralysis exhibited a downward trend, and 0.5 mM DM exerted the most marked effect (P<0.01) (Fig. 5A). When the concentration of PA exceeded 0.5 mM, it displayed a toxic effect; a PA concentration of 0.25 mM exhibited the most marked therapeutic effect (P<0.01) (Fig. 5B). Therefore, DM and PA have potential anti-AD activity.

In order to study the effects of DM and PA on the aggregation of Aβ, the formation of amyloid fibrils was observed using the thioflavin-S fluorescence method. The transgenic strain CL2006 used in this experiment expressed the Aβ protein fragment related to the development of AD. These worms showed expression and aggregation of Aβ in muscle, leading to progressive paralysis. CL2006 was stained with triterpenes for Aβ at the end of DM and PA treatment (green fluorescent labeling). The fluorescence image of the head area of CL2006 nematodes showed that compared with untreated worms (negative control), Aβ deposition in those treated with 0.5 mM DM or 0.25 mM PA significantly decreased. The wild N2 strain exhibited no Aβ deposition in any part of the animal (Fig. 6A). These results indicated that both DM and PA inhibit the aggregation and deposition of Aβ in muscle cells of transgenic worms (Fig. 6B).

Based on the results of network pharmacology and molecular docking, the regulatory effects of two active components of DM and PA with the core targets EP300, HSP90AB1, KDM6B and CREBBP were analyzed and verified using RT-qPCR. In C. elegans, the gene homologue of EP300 and CREBBP is CBP-1, that for KDM6B is JMJD3.1, and that for HSP90AB1 is HSP90 (35-37). Therefore, CBP-1, JMJD3.1 and HSP90 were examined. Compared with the control group, the expression levels of HSP90 in the DM and PA groups were significantly increased (P<0.01) (Fig. 7). There was no significant effect on the expression levels of CBP-1 in the DM or PA group. JMJD3.1 was upregulated in the DM group (P<0.01).