The active compounds of ZGBSF were retrieved from the Traditional Chinese Medicine Systems Pharmacology (TCMSP) Database and Analysis Platform (http://tcmspw.com/tcmsp.php) (23) by using ‘Yin Yang Huo’, ‘Xu Duan’, ‘Bai Shao’, ‘Chuan Xiong’, ‘Huang Qi’, ‘Shu Di Huang’, ‘Du Zhong’, ‘Niu Xi’, ‘Hong Hua’ and ‘Dang Gui’ as keywords to screen targets correlated with ZGBSF. All target names were standardized through the UniProt database (https://sparql.uniprot.org/) (24).

The potential targets related to OP were screened from DrugBank (https://go.drugbank.com/) (25), Genecards (https://www.genecards.org/) (26), Therapeutic Target Database (TTD, http://bidd.nus.edu.sg/group/ttd/ttd.asp) (27), PharmGkb (https://www.pharmgkb.org/) (28) and Online Mendelian Inheritance in Man (OMIM, https://www.genecards.org/) (29). After elimination of repetitive targets, the potential targets correlated with OP were obtained.

A drug-target-disease interaction network was constructed by inputting active compounds in ZGBSF and common targets of these active compounds and OP into Cytoscape 3.8 software. The degree was used to evaluate the importance of compounds and targets, which represented the total number of routes associated with the node and other nodes. And the higher the degree, the more important it was (30).

To clarify the functional interactions among the screened potential proteins, they were all imported to the String database (https://string-db.org/) to construct Protein-Protein Interaction (PPI) network (31). Then the PPI network was inputted into Cytoscape 3.8 software, and key proteins were identified by CytoNCA software according to betweenness centrality (BC), closeness centrality (CC), degree centrality (DC), Eigenvector centrality (EC), local average connectivity-based method (LAC), and network centrality (NC), which were greater than or equal to the median value (32).

The GO and KEGG functional enrichment analyses were performed through R software to obtain the core biological processes and signaling pathways. The top 10 items of GO analysis and 30 items of KEGG analysis were mapped as bubble plots. Finally, R software was utilized to generate the key signaling pathways related to OP (33).

Primary antibodies against total (t)-Akt1, Bcl-2, Bax, Parp were purchased from Ruiying Biological Co. (Jiangsu, China). Primary antibodies against Caspase-3, p-Akt1(S473), PI3K were obtained from Hua’an Biological Co. (Hangzhou, China). Primary antibodies against alkaline phosphatase (Alp), Osterix (Osx) were from CUSABIO Biological Co. (Wuhan, China). And primary antibody against Runx2 was supplied by Abcam Company Ltd. (Cambridge, MA, USA). Alexa Fluor^® 488 rabbit anti-goat IgG (H+L) secondary antibody (green) was obtained from Zhongshan Jinqiao Biotech Co. (Beijing, China). TUNEL Bright Green Apoptosis Detection Kit was supplied by Vazyme Biotech Co. (Nanjing, China). The remaining chemicals were from Sigma-Aldrich (St. Louis, Mo, USA) unless otherwise stated.

Ten raw herbs of ZGBSF were provided from the pharmacy department of the first affiliated hospital of Zhejiang Chinese Medical University (Hangzhou, China). A voucher of the specimen was stored at the first affiliated hospital of Zhejiang Chinese Medical University (Table 1). Epimedium, Dipsaci Radix, Paeoniae Radix Alba, Chuanxiong Rhizoma, Hedysarum Multijugum Maxim., Rehmanniae Radix Praeparata, Eucommiae Cortex, Achyranthis Bidentatae Radix, Carthami Flos, Angelicae Sinensis Radix were mixed as a proportion of 2:1.5:1.5:1:1.5:2:1.5:1.5:1:1.5 (w/w), and then soaked in 12 volumes of distilled water (v/m) for 1 h, decocted for 2 times, and 1.5 h each time. Finally, the water-based decoction was concentrated to 1 g/mL and stored at -20°C prior to use.

Twenty-four C57BL/6J female mice (8-week-old, 20 ± 1 g) were provided by the animal center of Zhejiang Chinese Medical University (Hangzhou, China) and housed at the Animal Care Facility of Zhejiang Chinese Medical University according to the institutional guidelines for laboratory animals. All proposals concerning animals were approved by the Ethics Committee of Zhejiang Chinese Medical University (20211220–02), and all laboratory procedures were performed following the Regulations for the Administration of Affairs Concerning Experimental Animals approved by the State Council of the People’s Republic of China.

All mice were randomly divided into 3 groups: Sham group, Ovariectomy (OVX) group, ZGBSF group (n = 8 in each group). After anesthesia with 1.5% isoflurane (R510-22-16, RWD Life Science, Shenzhen, China) in 100% oxygen flowing at 1.0 L/min, two ovaries of OVX and ZGBSF mice were removed, while the ovaries of Sham mice were preserved, and only part of the fat tissue around the ovaries was removed. Two days after OVX surgery, mice in the ZGBSF group were orally administered with ZGBSF (0.2 mL/10 g body weight), once a day for 8 consecutive weeks. Mice in the other two groups were given the same dosage of normal saline. All mice were euthanized 8 weeks after intragastric administration, and the femurs were collected for further analysis.

After fixation with paraformaldehyde, the femurs were scanned using micro-CT (μCT) equipment (Skyscan 1176, Bruker μCT N.V., Kontich, Belgium) at a voltage of 50 kV with a current of 500 μA and a resolution of 9 μm per pixel. The NRecon v1.6 and CTAn v1.15 software were utilized to reconstruct the three-dimensional (3D) structure of the distal femoral metaphysis as we previously described (34).

After μCT analysis, the femurs were decalcified with 14% EDTA (pH 7.4) for 3 weeks, then dehydrated, embedded in paraffin, and processed into 5-μm-thick coronal-oriented sections. Then sections were stained with hematoxylin-eosin (H&E) and tartrate-resistant acid phosphatase (TRAP) as we previously described (35, 36). For immunohistochemistry (IHC) or immunofluorescence (IF) analysis, sections were incubated with primary antibodies of Runx2 (diluted 1:300), Alp (diluted 1:200), Osx (diluted 1:300), t-Akt1 (diluted 1:300), p-Akt1(S473) (diluted 1:50), PI3K (diluted 1:50), Bcl-2 (diluted 1:300), Bax (diluted 1:300), Caspase-3 (diluted 1:300), Parp (diluted 1:300) at 4°C overnight. Secondary biotinylated goat anti-rabbit antibody (diluted 1:1000) (Invitrogen) was added for 30 min the following day, and a diaminobenzidine solution (Invitrogen) was used to detect IHC staining. For IF analysis, secondary goat anti-rabbit antibody conjugated with fluorescence (diluted 1:1000) was added for 30 min the next day. The quantification of positive cells was estimated in a blinded manner with Image-Pro Plus 6.0 (Media Cybernetics, Silver Spring, MD, USA).

To further evaluate apoptosis of osteoblasts in femoral tissues, the DNA damage, another hallmark of aopotosis, was determined by the TUNEL Bright Green Apoptosis Detection Kit (Vazyme Biotech, Nanjing, China). All operations were performed according to the manufacturer’s instructions. The Image-Pro Plus 6.0 (Media Cybernetics, Silver Spring, MD, USA) was used to calculate the number of positive cells by using three sections from each sample in three randomly selected fields of view with blind method. And the total number of cells was calculated by DAPI staining.

All numerical data were expressed as mean ± standard error of mean (SEM). One-way analysis of variance (ANOVA) with a Student-Newman-Keuls (SNK) post hoc test was performed using GraphPad Prism 6.0 (GraphPad Software Inc., La Jolla, CA). When P < 0.05, the differences were considered to be statistically significant.

A total of 145 active compounds of ZGBSF were screened, including 23 from Yin Yang Huo, 8 from Xu Duan, 13 from Bai Shao, 7 from Chuan Xiong, 20 from Huang Qi, 2 from Shu Di Huang, 28 from Du Zhong, 20 from Niu Xi, 22 from Hong Hua, and 2 from Dang Gui. And 2729 related targets of ZGBSF were obtained from the TCMSP database, containing 511 types in Yin Yang Huo, 63 types in Xu Duan, 123 types in Bai Shao, 42 types in Chuan Xiong, 462 types in Huang Qi, and 34 types in Shu Di Huang, 532 types in Du Zhong, 444 types in Niu Xi, 449 types in Hong Hua, and 69 types in Dang Gui. After excluding duplicates, a total of 86 compounds and 233 targets were selected for the following research.

To identify potential targets of OP, we searched 5 databases for OP-related research reports. A total of 4725 targets of OP were screen out, of which 384 were from the DrugBank database, 4273 from the GeneCards database, 31 from the TTD database, 3 from the PharmGkb database, 34 from the OMIM database. After expurgating duplicate targets, 4392 targets related to OP were screened out (Figure 2A).

To obtain the intersection of ZGBSF target genes and OP target genes, the Venn diagram analysis was performed. The results showed that a total of 164 target genes were identified (Figure 2B), which matched with the related targets of 86 active compounds. Then, to clarify the relationship among active compounds, targets, and OP, a drug-target-disease network, including 250 nodes and 847 edges, was constructed by using Cytoscape 3.8 software (Figure 2C). The top 10 core compounds of ZGBSF against OP were listed in Table 2.

To obtain the pivotal proteins of ZGBSF in treating OP, after removing 3 isolates, a PPI network with 161 nodes and 3021 edges was constructed by using a String database (Figure 2D). Then, the thresholds of BC, CC, DC, EC, LAC, and NC were set in two steps according to the topological parameters calculated by CytoNCA. And a total of 22 pivotal proteins were filtered out, including Caspase-3, BCL2L1, TP53, IL-1β, Akt1, ESR1, FN1, CCND1, VEGFA, HIF1A, EGFR, CXCL8, CCL2, CTNNB1, JUN, PPARG, FOS, NFKBIA, PTGS2, MYC, MMP9, EGF, which were strongly linked to OP (Figures 2E, F).

To explore the crucial biological processes of ZGBSF in the treatment of OP, the GO functional enrichment analyses were performed. The results of GO analysis revealed that the 164 genes were enriched in 2709 GO entries, consisting of 2401 biological progress (BP), 92 cellular components (CC), and 216 molecular functions (MF) (P < 0.05), and the top 10 entries of BP, CC and MF were shown in Figure 3A. BP analysis revealed that related targets were mainly centered on cellular response to chemical stress, drug, oxidative stress, etc. CC analysis indicated that related targets were mainly centered on membrane microdomain, membrane raft, etc. Moreover, MF analysis showed that potential targets were primarily focused on nuclear receptor activity, transcription factor activity, RNA polymerase II−specific DNA−binding transcription factor binding, etc.

To investigate the representative signaling pathways associated with the key targets, the KEGG enrichment analysis was performed. And the results showed that 160 significantly enriched signaling pathways were retrieved (P < 0.05). The top 30 significantly enriched signaling pathways closely correlated with OP were shown in Figure 3B, including apoptosis, PI3K-Akt, and TNF signaling pathways, etc. Furthermore, we found that apoptosis and PI3K-Akt signaling pathways were most closely related to OP in these enriched pathways, and the predicted targets corresponding to these two signaling pathways were shown in Figures 3C, D.

To verify the core pharmacological mechanism of ZGBSF in treating OP predicted by the above-mentioned network pharmacology analysis, OVX model mice were established and treated intragastrically with ZGBSF for 8 weeks. The femurs of OVX mice were harvested and radiographic alterations were assessed by μCT analysis. The results of μCT analysis showed that compared with sham mice, the bone mass of OVX mice significantly reduced, whereas ZGBSF intervention significantly restored the bone mass (Figure 4A). Moreover, the results of bone morphological parameter analysis showed that ZGBSF enhanced the decrease of bone mineral density (BMD), bone volume fraction (BV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N) in OVX mice. However, it had no significant recovery effect for the widened trabecular separation (Tb.Sp) (Figures 4B–F). In parallel, the histological changes of the femur were examined using H&E staining. The results showed that the femoral tissues of OVX mice exhibited less trabecular bone, and a large number of lipid droplets were present instead. ZGBSF administration significantly reversed these alterations (Figures 4G–I). All these findings showed that ZGBSF effectively attenuated bone loss and the progression of OP.

Bone homeostasis and remodelling is regulated by a balance between osteogenesis of osteoblast and osteoclastogenesis of osteoclast (37). To examine the effects of ZGBSF on osteogenesis, protein levels of Runx2, Alp, and Osx were determined by IHC analysis. The results of IHC analysis revealed that Runx2, Alp, and Osx were significantly decreased in OVX mice. The treatment of ZGBSF significantly increased the expression of Runx2, Alp, and Osx (Figures 5A–D). Subsequently, to explore the effect of ZGBSF on osteoclast formation, the number of osteoclasts was evaluated by TRAP staining. Compared with sham mice, the number of osteoclasts in the distal femurs of OVX mice was markedly increased. Unexpectedly, ZGBSF intervention had no significant effect on the increase of osteoclast number in OVX mice (Figures 5E, F).

The prediction results from our network pharmacology analysis revealed that the apoptosis and PI3K-Akt signaling pathways were pivotal biological events in ZGBSF against OP. Therefore, we first determined the expression of Bcl-2, Bax, Caspase-3, and Parp, indicators of apoptosis, by IF analysis. The IF results showed that Bcl-2 was markedly decreased in the distal femur of OVX mice, while Bax, Caspase-3 and Parp were significantly increased. ZGBSF significantly reversed these alterations (Figures 6A–E). In parallel, the anti-apoptosis effect of ZGBSF was further confirmed by TUNEL staining. The results of TUNEL staining revealed that ZGBSF significantly inhibited the increased apoptosis ratio of osteoblasts in femoral tissues of OVX mice (Figures 6F, G). All these findings showed that ZGBSF suppressed apoptosis of osteoblasts and contributed to osteogenesis, thus ameliorating bone loss of OVX mice.

Next, to verify the effect of ZGBSF on PI3K−Akt signaling pathway predicted by KEGG enrichment analysis, the phosphorylation status of Akt1, and expression of PI3K protein were examined by IHC analysis of t-Akt1, p-Akt1, and PI3K. The IHC results showed that the levels of p-Akt1 and PI3K protein in distal femoral tissue of OVX mice were lower than those in sham mice. ZGBSF treatment significantly upregulated the expression of p-Akt1 and PI3K protein in ZGBSF mice (Figures 7A–C).