The TCM@Taiwan (http://tcm.cmu.edu.tw/zh-tw/) Database (Chen, 2011), Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP http://lsp.nwu.edu.cn/tcmsp.php) (Ru et al., 2014), BATMAN-TCM (L et al., 2016) (a bioinformatics analytical tool for the molecular mechanisms of TCM: http://bionet.ncpsb.org.cn/), Traditional Chinese Medicine Information Database (Kang et al., 2013) (TCM-ID http://bidd.group), and relevant literature were utilized to acquire all identified AB–DA compounds. ADME (absorption, distribution, metabolism, and excretion) properties were applied to screen bioactive ingredients, and the screening criteria were set as “oral bioavailability (OB)≥ 30, drug-likeness (DL)≥ 0.18” (Gao et al., 2022) to screen the compounds from the TCMSP database. Similarly, compounds from different sources were screened in the SwissADME (Daina et al., 2017) database using their pharmacokinetic properties (http://www.swissadme.ch).

To predict potential targets based on their spatial configuration, the compounds generated in the previous step were imported into the SwissTargetPrediction (Daina et al., 2019) (http://www.swisstargetprediction.ch/) and PharmMapper databases (Liu et al., 2010) (http://www.lilab-ecust.cn/pharmmapper/). The UniProt ID of the target was converted into a standardized gene name using the UniProt database (Holzhüter and Geertsma, 2022) (https://www.uniprot.org). The keyword “osteoporosis” was searched in the GeneCards (Barshir et al., 2021) (https://www.genecards.org/), DisGeNET (Piñero et al., 2020) (https://www.disgenet.org), and Online Mendelian Inheritance in Man (Li et al., 2012) (OMIM, https://omim.org) databases to obtain relevant targets. Then, the overlapping targets identified by Venn diagram were considered as targets of AB–DA for the treatment of osteoporosis after merging and removing duplicates.

To identify potential hub genes, a protein–protein interaction (PPI) network was established in the STRING (Szklarczyk et al., 2021) (http://string-db.org, Version 11.5) database, with a focus on the anti-osteoporosis efficacy of AB–DA in Homo sapiens and an interaction score threshold of 0.4. Topological analysis was performed, and the core targets of AB–DA for treating osteoporosis were accurately selected by using the Cytoscape plug-ins MCODE (molecular complex detection) and CytoHubba (Ye et al., 2022).

Analysis of Gene Ontology (GO) functions, including cellular component (CC), molecular function (MF), and biological process (BP), and Kyoto Encyclopedia of Genome and Genome (KEGG) pathway enrichment analysis, was utilized to clarify the key anti-osteoporosis mechanism of the AB–DA herb pair. When entering the targets into the Metascape database (http://www.metascape.org/), the cut-off p-value, minimum overlap value, and concentration value were set to 0.01, 3, and 1.5, respectively (Zhou et al., 2019). False-positive rate (FPR) analysis was eliminated using the Benjamini–Hochberg method with a q-value of 0.05 or lower (Zou et al., 2016). The enriched findings were displayed as bar and bubble plots on the bioinformatics website using the R package (http://www.bioinformatics.com.cn/). Comprehensive information on the most significantly enriched pathway was then extracted and colored (Kanehisa and Sato, 2020). Finally, a herb–compound–target–pathway–disease network was created using Cytoscape software (v.3.9.1, https://cytoscape.org/) to present the complicated network of the AB–DA herb pair in the treatment of osteoporosis.

The SwissDock platform (Grosdidier et al., 2011) (http://www.swissdock.ch/) is an online molecular docking (MD) tool to determine the binding affinity from each binding site between small molecule ligands and receptor proteins. The X-ray diffraction of the protein crystal structure of key targets were downloaded from the Protein Data Bank (PDB) database (www.rcsb.org) (Nakamura et al., 2022). The binding sites were ranked based on their binding affinity scores, with the site having the smallest score considered the best binding site. Discovery Studio 2019 software (https://www.3ds.com) was used to visualize the binding details (Sultana et al., 2022).

To investigate the stability of the complexes between small-molecule ligands and proteins, molecular dynamics simulations (MDS) were performed using the Standard Dynamics Cascade subunit of the Discovery Studio 2019 software. The ligand–protein complex with the lowest binding affinity score according to molecular docking analysis was selected (Hu et al., 2022). In this simulation system, water molecules are used to fill the solvent chamber, and Cl and Na+ ions are used to maintain an electrically neutral state. The simulation time was set as 300 ps, and the heating, balancing, and manufacturing phases were carried out after the system was balanced by an NPT ensemble, which fixed the pressure, temperature, and particle number. The analysis of the locus was performed using root mean square fluctuation (RMSF), root mean square deviation (RMSD), and hydrogen bond properties to produce the results.

Sitogluside, identified as one of the most promising bioactive compounds in the AB–DA herb pair, was further investigated for its anti-osteoporotic effects and associated mechanisms. Sitogluside was procured from the Dalian Meilunbio company and solubilized in dimethyl sulfoxide (DMSO). Human fetal osteoblast (hFOB) cells were obtained from the American Type Culture Collection (ATCC) and cultured in six-well plates using Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS, Gibco, United States), 1% penicillin–streptomycin (P/S), and 0.3 mg/mL Geneticin (G418). The JNK-specific inhibitor (SP600125) and p38-specific inhibitor (SB203580) were acquired from Absin (Shanghai). The cells were maintained in a humidified sterile atmosphere at 34 °C. They were subsequently treated with sitogluside when they reached a confluence of 60%–80% per well.

For alkaline phosphatase (ALP) activity analysis, the BCIP/NBT Alkaline Phosphatase color development kit (Beyotime Institute of Biotechnology, Shanghai, China) was used according to the manufacturer’s procedure. Briefly, osteoblast precursor cells were seeded at 3 × 10⁴ cells/well in 24-well plates and grown for 14 days in osteogenic media (DMEM +10% FBS +1% P/S +100 M ascorbic acid +2 mM 2-glycerophosphate +10 nM dexamethasone). The stained culture plates were photographed using a microscope (Leica image analysis system, Q500MC) and quantified using ImageJ software (National Institutes of Health, Bethesda, MD, United States). In addition, a 2% ARS reagent (Beyotime Institute of Biotechnology) was used to detect matrix mineralization, with the same protocol as ALP assay except the dye. In addition, we also investigated the effect of sitogluside on the osteoclast to fully illustrate the anti-OP effects. The RAW264.7 (osteoclast precursor) cells were treated with a different concentration of sitogluside and supplemented with 50 ng/mL receptor activator of NF-κB ligand (RANKL) as an osteoclast formation stimulator. Cells were then fixed in paraformaldehyde and stained with tartrate-resistant acid phosphatase (TRAP) after the intervention.

This study received ethical approval from the Animal Care Committee of the Second Xiangya Hospital of Central South University. A total of 30 10-week-old female C57BL/6 mice were procured from SLAC Laboratory Animal Co. Ltd. (SLACCAS, Shanghai, China). They were acclimatized in specific pathogen-free (SPF) cages for 1 week, during which measures were taken to minimize animal suffering through the use of anesthesia and sterile techniques during the surgical procedures. Bilateral ovariectomy (OVX) or sham surgery (retroperitoneal incision without ovariectomy) was performed based on group assignment, as described in the following paragraph.

Following the surgical procedures, the mice were randomly assigned to one of three groups: sham group (non-OVX mice, n = 10), vehicle group (OVX mice, n = 10), and sitogluside group (OVX mice intraperitoneally injected with sitogluside at a dose of 10 mg/kg/day, n = 10).

After 12 weeks, all mice were euthanized by cervical dislocation, and the right femurs were dissected and fixed in 4% paraformaldehyde (PFA) for 48 h. High-resolution micro-computed tomography (μCT40, Scanco, Zurich, Switzerland) was employed for bone analysis at the following parameters: scanning voltage = 80 kV, electric current = 80 μA, and resolution = 10 μm. The relevant trabecular bone volume fractions (BV/TV), trabecular number (Tb. N), trabecular thickness (Tb. Th), and trabecular separation (Tb. Sp) were subsequently calculated to assess the protective efficacy of sitogluside in OVX mice. In addition, to investigate the biosafety of sitogluside in the OVX model, the main organs were also obtained and detected using hematoxylin-eosin (H&E) staining. The heart, liver, spleen, lung, and kidney were hence fixed with formalin and embedded in paraffin cut to a 4 μm section. They were dewaxed in xylene, rehydrated with concentration gradient ethanol, and then stained with H&E for histological examinations and morphometric analysis. The serum of the mice was also collected to examine the biomarkers of alanine aminotransferase (ALT), creatine kinase (CK), and blood urea nitrogen (BUN).

Human osteoblast (hFOB) cells were treated with 40 μM sitogluside or combined with the specific inhibitor of JNK and p38, depending on the groups, with ascorbic acid added in the osteogenic media. The cells were then harvested using the RNeasy Mini kit (QIAGEN, CA, United States) to extract total RNA following the manufacturer’s protocol. Subsequently, cDNA synthesis was performed using the reverse transcriptase kit (Takara Biotechnology, Japan). Real-time PCR analysis was performed using the SYBR Premix Ex Taq kit (Takara Biotechnology, Japan). The PCR parameters were set as follows: 40 cycles (denaturation at 95 °C for 10 s and amplification at 60 °C for 30 s). The resultant data were recorded as cycle threshold (Ct) values, and the 2^−ΔΔCT method was employed for further analysis of RNA expression. In addition, to determine the modulated effect of sitogluside with JNK and osteogenic genes, the knockdown and activation of JNK expression were applied. ShRNA (shGnai3) was thus used to downregulate the expression level of JNK, and ASM was used as an activator to increase the p-JNK level (Meng et al., 2021). The alterations of osteogenic biomarkers were detected via q-PCR.

The hFOB cells were harvested by trypsin and then lysed in RIPA for 30 min on ice. Cell lysates were centrifuged at 12,000 g for 15 min at 4 C; the supernatant was collected, and protein content was quantified via the BSA protein assay kit following the manufacturer’s instruction. Proteins were separated by electrophoresis on 10%–12% SDS-PAGE at 100 V for 1.5 h and transferred onto a 0.45 μm polyvinylidene difluoride (PVDF) membrane at 250 mA for 1 h. The PVDF membrane was blocked with 5% non-fat milk in TBST buffer for 1 h at room temperature and incubated with primary antibody at 4 °C overnight. They were then incubated with secondary antibody for 1 h at room temperature and detected using the Chemiluminescence Kit.

Statistical analyses were conducted using GraphPad Prism 8.0.2 software (San Diego, United States). The data are presented as means ± standard deviation (SD). Data comparisons were performed using one-way analysis of variance (ANOVA), and statistical significance was determined by a p-value of less than 0.05.

According to TCM databases, 185 components of AB and 82 compounds of DA were obtained in this work (Figures 3A, B). After filtration by screening criteria, 19 potential bioactive compounds of AB were obtained: arjunolic acid, baicalein, baicalin, berberine, chondrillasterol, coptisine, delta-7-stigmastenol, epiberberine, inophyllum E, kaempferol, oleanol, palmatine, quercetin, sitogluside, spinasterol, spinoside A, stigmasterol, wogonin, and β-ecdysterone. The 12 compounds from DA were 2,6-dihydroxycinnamic acid, caffeate, cauloside A, gentisin, isochlorogenic acid A, japonine, loganetin, loganic acid, loganin, sweroside, and sylvestroside III. Sitosterol is a common compound of both AB and DA (details shown in Supplementary Table S1; structures shown in Figure 2).

After merging and duplicating results, 413 potential targets associated with 31 bioactive compounds were obtained, and 827 relevant targets of osteoporosis were acquired. Some 83 overlapped genes on Venn between AB–DA ingredients and osteoporosis were regarded as potential therapeutic targets (Figure 3C; details shown in Supplementary Table S2).

A total of 83 targets of AB–DA against osteoporosis were imported to the STRING database; after deleting the disconnect targets, a PPI network with 78 nodes and 702 edges was constructed. Cytoscape software was utilized for further visualization, and plug-ins of MCODE and CytoHubba based on the topological parameters were applied to screen the hub genes, including AKT1, IGF1, CASP3, MMP9, EGFR, PPARG, ESR1, MAPK1, MAPK8, and MAPK14 (Figures 3D–F).

GO and KEGG pathway enrichment analyses were performed in the Metascape database, and 83 targets with 875 GO items were enriched. It contained 772 biological processes (BP), 29 cellular components (CC), and 74 molecular functions (MF) items (p < 0.01, adjusted q < 0.05) (details shown in Figure 4A).

A total of 135 KEGG pathways were significantly enriched, which mainly included the MAPK signaling pathway, osteoclast differentiation, PI3K-Akt signaling pathway, and endocrine resistance (Figures 4B, C). We further mapped and colored the regulation details of the MAPK cascade in the KEGG mapper database; in that map, red objects represent targets of AB–DA against osteoporosis, blue objects show the targets of AB–DA without the therapeutic effects of osteoporosis, and the other untargeted targets of osteoporosis are colored yellow (Figure 4E).

A drug–compound–target network was constructed using Cytoscape to illustrate core compounds and targets (Figure 4D); the core bioactive compounds included sitogluside, arjunolic acid, chondrillasterol, stigmasterol, spinasterol, spinoside A, cauloside A, sylvestroside III, β-ecdysterone, sitosterol, oleanol, and baicalin (ranked by degree value) (Figure 4G). A drug–compound–target–pathway–disease (D–C–T–P–D) network was then constructed to exhibit the complex molecular mechanisms of AB–DA anti-osteoporosis with multi-compound, multi-target, and multi-pathway characteristics. The dark green diamond represents the drugs (AB and DA), the light green hexagon represents the compounds of AB–DA, the circles colored orange to red represent targets with low to high degrees, the blue V icon represents the core enriched pathways, and the yellow rectangle indicates the disease (osteoporosis) (Figure 4F).

The binding affinity between the core compounds and core targets are shown by heatmap (Figure 5A). According to relevant theories of molecular docking, the results of binding affinity < −5.0 kcal/mol suggest that there is a good spontaneous binding activity between molecule ligands and protein receptors, and results < −7.0 kcal/mol are stronger. Our research results showed that all the compounds had good binding activity with core targets, with binding affinities ranging from −5.64 kcal/mol to −9.43 kcal/mol. Sitogluside has the highest binding activity with IGF1. As shown in Figures 5B–D, the molecular interaction forces between IGF1 and sitogluside include π–donor hydrogen bond, π–alkyl bond, conventional hydrogen bond, carbon–hydrogen bond, and alkyl bond. The distances between the sitogluside atoms and amino acid residues of IGF1 range from 1.72 Å (number 1133, histidine residue) to 5.43 Å (number 1154, phenylalanine residue).

The previous analysis of molecular docking showed the strong binding affinity between AB–DA compounds and core targets, and molecular dynamics simulation was utilized to identify the stability of the ligand–protein complex after docking by conformation alternation with potential energy under Newton’s law of motion. After the 300 ps simulation, the energy and temperature alternate tendency of the ensemble, hydrogen bond, RMSD, and RMSF changes of ligand–receptor interaction were calculated for stability analysis. RMSD was used to analyze the conformational alternation of receptors made by the ligand, and the results showed that curves and fluctuations only occurred at the beginning of the 80 ps simulation, and then tended to be stable (Figure 6A). The RMSF curve was utilized to monitor the conformational alternation of amino acid residues, and results show that the whole process is stable with only some small random fluctuations—which also reflects the whole ensemble’s stability to some extent (Figure 6B). The same tendency is also observed in the hydrogen bond heatmap (Figure 6C). The energies and temperature alternation of the whole ensemble were also stable and controlled. Therefore, the overall results exhibited good stability between the small-molecule ligand–protein receptor complex.

To further investigate the impact of sitogluside on alkaline phosphatase (ALP) activity and mineralization in osteoblasts, we conducted ALP activity and ARS experiments. As illustrated in Supplementary Figure S3D, the sitogluside group exhibited increased ALP activity and mineralization compared to the control group. However, when shGnai3 intervention was introduced, the stimulatory effect of sitogluside on ALP activity and mineralization in osteoblasts was partially diminished. Subsequently, with the addition of the p-JNK activator ASM, the promotion of ALP activity and mineralization was restored. Additionally, our results indicated that sitogluside might not have a significant effect on osteoclast formation (Supplementary Figure S2).

The OVX model was employed to further investigate the anti-osteoporotic effects of sitogluside in vivo. As depicted in Figure 7A, the μCT scan results demonstrated a significant loss of bone in the vehicle group (OVX mice) compared to the sham group, confirming the successful establishment of the OVX model. Additionally, the sitogluside-treated group exhibited a mitigating effect on bone loss compared to the vehicle group. Specifically, the sitogluside-treated OVX mice demonstrated increased bone mineral density (BMD), trabecular thickness (Tb. Th), bone volume fraction (BV/TV), and trabecular number (Tb. N), while exhibiting a decreased bone-surface-to-bone-volume ratio (BS/BV) and trabecular separation (Tb. Sp). These findings collectively indicate the anti-osteoporotic efficacy of sitogluside in the OVX mouse model (Figure 7B). Furthermore, histological examination of major organs in the drug-treated group, including the heart, liver, spleen, lungs, and kidneys, was conducted using (H&E) staining. Additionally, mouse serum was analyzed to assess cardiac, hepatic, and renal functions. As depicted in Supplementary Figure S1, the experimental results indicate that sitogluside did not exert significant toxic effects on mouse organs.

To gain further insights into the underlying mechanism of sitogluside’s anti-osteoporotic effects, we examined the expression of osteogenic-related genes using q-PCR and Western blot. As shown in Figure 7, treatment with sitogluside for 48 h resulted in a significant increase in the expression of osteogenic markers, such as Osterix (OSX) and osteocalcin (OCN). These findings suggest that sitogluside has the potential to promote osteogenic activity. Additionally, we investigated the core target and pathway associated with sitogluside in the treatment of osteoporosis via blocking the JNK and p38 cascade (Figure 7D). Our results revealed that targeting JNK could reduce the osteogenic efficacy of sitogluside while blocking the p38 pathway without significant expression changes of OSX. To further investigate whether sitogluside exerts its osteogenic effects through the JNK pathway, we employed shRNA to silence the JNK pathway. As shown in Supplementary Figure S3C, compared to the sitogluside group, silencing the JNK pathway with shRNA (shGnai3) resulted in a partial reduction in the expression levels of osteogenic markers, including Runx2, OSX, OCN, and ALP. However, when treated with the p-JNK activator anisomycin (ASM), these osteogenic markers increase. Moreover, as shown in Supplementary Figures S3A, B, the phosphorylation level of p-JNK increases with sitogluside intervention. These further suggest that sitogluside may promote osteoblast differentiation through the JNK signaling pathway. We also investigated the effect of sitogluside on the Smad 1/5/8 protein in the BMP signaling pathway, which also showed increasing expression (Supplementary Figures S3A, B).