Deciphering the Underlying Mechanism of Eucommiae Cortex against Osteoporotic Fracture by Network Pharmacology

Background Du Zhong (DZ), or Eucommiae Cortex, a traditional Chinese herbal medicine, has been used to treat osteoporosis. Although it has been reported that DZ can improve bone mass in ovariectomized rats, its pharmacological mechanisms in treating osteoporotic fractures (OPF) remain unclear. Methods In this study, we used a network pharmacological manner to explore its potential complicated mechanism in treating OPF. We obtained DZ compounds from TCMSP and BATMAN-TCM databases and collected potential targets of these compounds through target fishing based on TCMSP and BATMAN-TCM databases. Next, we collected the OPF targets by using CTD, GeneCards, OMIM, HPO, and GenCLiP 3 databases. And then the overlapping genes between DZ potential targets and OPF targets were used to build up the protein-protein interaction (PPI) network and to analyze their interactions and find out the big hub genes in this network. Subsequently, clusterProfiler package in R language was utilized to conduct the enrichment of Gene Ontology biological process and KEGG pathways. Results There were totally 93 active compounds and 916 related targets in DZ. After the enrichment analysis, we collected top 25 cellular biological processes and top 25 pathways based on the adjusted P value and found that the DZ anti-OPF effect was mainly associated with the regulation of ROS and inflammatory response. Furthermore, 64 hub genes in PPI network, such as MAPK1 (degree = 41), SRC (degree = 39), PIK3R1 (degree = 36), VEGFA (degree = 31), TP53 (degree = 29), EGFR (degree = 29), JUN (degree = 29), AGT (degree = 29), MAPK1, SRC, PIK3R1, VEGFA, and TP53, were considered as potential therapeutic targets, implying the underlying mechanisms of DZ acting on OPF. Conclusion We investigated the possible therapeutic mechanisms of DZ from a systemic perspective. These key targets and pathways provided promising directions for the future research to reveal the exact regulating mechanisms of DZ in treating OPF.


Introduction
Osteoporosis is a systemic bone disorder characterized by low bone density, poor bone quality, reduced bone strength, and an accompanying increased incidence of fractures [1].
e latest study noted that over 10 million people suffer from osteoporosis in the United States. Meanwhile, the treatment burden of osteoporosis, $22 billion in 2008, is expected to rise due to the consistently increasing aging population. Related osteoporosis treatments should be used to alleviate or reduce symptoms such as fractures, which could result in considerable cost savings [2]. Osteoporotic fractures (OPF) have serious consequences, such as declining functions and independence as well as increased morbidity and mortality [3,4]. Hip fractures are considered to be the most expensive and debilitating of fractures because approximately 10 to 20% of patients will end up with a disability in the year following a hip fracture and half of patients will lose their independence. Osteoporotic fractures are also associated with motor function decline when occurring at other skeletal sites [5]. To curb bone loss, maintain bone mass, and decrease the risk of OPF, antiresorptive agents, anabolic agents, and bone mineral drugs have been widely used in clinical treatment [6]. However, it is noteworthy that almost all these antiosteoporosis treatments have limitations and side effects, such as increasing the risk of breast cancer, jaw necrosis, or atypical femur fracture [7,8]. erefore, alternative and safe intervention strategies may be promising.
Recently, traditional Chinese medicine (TCM) has attracted worldwide attention and has served as a main alternative treatment in East Asia, North America, and Europe due to its satisfactory curative effect, relatively low toxicity, and low cost [9][10][11]. TCM has been used in China for thousands of years, and it exerts a beneficial therapeutic effect and reduces side effects through multiple herb combinations to prevent and treat various diseases [12]. Meanwhile, TCM has increasingly been found to be effective in the treatment of osteoporosis [13,14].
Du Zhong (DZ), or Eucommiae Cortex, is one of the most commonly used TCM herbs in the treatment of osteoporosis. DZ has been found to facilitate osteogenesis through activating osteoblasts and to inhibit osteolysis by suppressing osteoclast activity [15]. An animal experiment confirmed that Du Zhong Wan, consisting of Eucommiae Cortex and Radix Dipsaci, is capable of improving trabecular bone mineral density and bone biomechanical properties and thus plays a key role in the treatment of osteoporosis [16]. Although DZ exerted therapeutic effects against osteoporosis, the therapeutic mechanism of its exact is still unclear. Similar to TCM, DZ exerts its therapeutic efficacy by regulating multiple molecules in the human body. erefore, it is still a major challenge to prove the mechanism of DZ through scientific trials that are used in Western medicine. Fortunately, the development of systems pharmacology offers researchers an alternative opportunity and an option to investigate the pharmacological mechanisms of TCM [17]. Recently, Wang et al. utilized a network pharmacology method to clarify the synergistic mechanism of Er-Xian decoction in treating osteoporosis [18]. Similarly, Gan et al. employed this systems pharmacology method to dissect the mechanisms of the therapeutic effect of Rhizoma Drynariae on osteoporosis [19].
In this network pharmacology work, we aim to comprehensively dissect the mechanisms of DZ in treating OPF. We collected related compounds of DZ from multiple databases and obtained the compounds' potential targets via target fishing. en, we matched these targets with OPFrelated targets that were collected from a multisource database. Next, using overlapping targets obtained from the previous process, we built a protein-protein network to analyze their internal interactions and then screened out the hub genes. Furthermore, we used the clusterProfiler package in R to conduct biological process and KEGG pathway enrichment analyses. e protocol of our experimental procedures is shown in Figure 1. [20], which is a frequently used platform for systems pharmacology. e BATMAN-TCM platform (http://bionet.ncpsb.org/batman-tcm/) [21], one of the largest comprehensive TCM platforms, uses the similarity of known TCM drug-target interactions to predict potential compound-target interactions. In total, ninetythree chemical compounds were collected in this part.

Compound Screening and Target Prediction
2.2.1. OB Evaluation. Oral bioavailability (OB) refers to the percentage of an orally administered drug that reaches systemic circulation, and it is one of the most important pharmacokinetic profiles for drug screening. e TCMSP platform has adopted the OBioavail1.1 system, which integrates P450, 3A4, and P-glycoprotein information to obtain the OB value [22]. To determine the active ingredients used for further steps, we set the OB threshold at 30%.

DL Prediction.
Drug-likeness (DL) is a qualitative index that represents the degree to which the target compound is "drug-like" and is used to remove chemically unsuitable compounds. TCMSP uses the Tanimoto similarity method to calculate the DL index by comparing the target compound to all 6511 molecules in the DrugBank database [23]. In this process, the compounds that did not meet the condition that DL ≥ 0.18 were removed.

BATMAN-TCM.
To collect candidate compounds, we set the potential TAR score cutoff ≥20 and P-value cutoff <0.05.

Target Prediction of the Active Ingredients.
e compounds have effects on the targets that induce them to exert their biological functions. us, we used the TCMSP and BATMAN-TCM platforms to predict the targets of active compounds.

Disease Target Identification by Multiple Databases.
We used multiple databases to collect OPF-related targets, and the terms "Osteoporotic" and "Fracture" were used as the key words for the search. e databases in this step included the Comparative Toxicogenomics Database (CTD, http://ctdbase.org/) [24], GeneCards (https://www. genecards.org/) [25], OMIM (https://www.omim.org/) [26], HPO (https://hpo.jax.org/app/) [27], and GenCLiP3 (http://ci.smu.edu.cn/genclip3/analysis.php) [28]. In total, three thousand four hundred-thirteen OPF targets were found through these databases. en, we constructed a Venn diagram to determine the overlap between OPF and the active compound targets because these overlapping targets may play a critical role in the treatment of OPF. ese targets were analyzed by String (https://string-db.org/) [29], in which the minimum required interaction score was set at the highest confidence level (0.900), and then, the proteinprotein interaction (PPI) data were exported.

Construction Method.
Two main networks were built in this process, including the compound-hub gene network and the hub gene-pathway network. e target information was obtained from the KEGG pathway enrichment results. Cytoscape 3.6.2 (http://www.cytoscape.org/) was used for all the network visualization work. Cytoscape 3.6.2 is one of most powerful open-source software programs for constructing different networks visually [30]. However, there are several limitations faced by network pharmacology research. For example, questions including whether compounds activate or inhibit targets, how they exact their effects on targets, whether they use binding or catalysis, and so on remain unclear.

Topological Feature Definition.
We used three parameters to describe and quantify the importance of nodes in these networks because nodes that bridge many edges with their neighbors are more likely to exert crucial mediating functions. (1) "Degree" means the number of edges shared with other nodes. Examining the node's degree is the most straightforward method of quantifying its centrality [31]. (2) "Betweenness centrality" is used to measure node centrality based on the shortest paths. A high-value node would play a more important role in the network because more regulation information will pass through it [32]. (3) "Closeness centrality" represents the mean distance between the node and all the other nodes in the network and is the reciprocal of the sum of the length of the shortest paths between a node and all the other nodes in the network. erefore, the central node would be more likely to be close with all other nodes [33].

Biological Process and Pathway Enrichment Analysis.
We used the clusterProfiler package in R (R 3.6.2 for Windows) to perform Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis to identify biological processes and systemic involvement of pathways.

Active Compounds of DZ.
A total of 147 related compounds were found in DZ collected from the TCMSP database, and of these, 27 candidate compounds remained after the screening of ADME thresholds (OB ≥ 30%, DL ≥ 0.18). Using the BATMAN-TCM database, we obtained 73 active compounds that matched the filter criteria. In total, we collected 93 unique compounds.

Target Prediction and Analysis.
For these unique compounds, we obtained 961 unique related targets, 104 from TCMSP and 857 from BATMAN-TCM. We integrated the Evidence-Based Complementary and Alternative Medicine OPF genes that were obtained from multisource databases, including the CTD, GeneCards, OMIM, HPO, GenCLiP 3 databases, and a total of 3,834 related genes were collected. After construction of the Venn diagram, three hundred twenty-five overlapping targets between the related targets of DZ and OPF were selected as the key targets on which DZ exerts its anti-OPF effects ( Figure 2). e data obtained from the String database were used to establish the PPI network for the 325 overlapping targets. In this network, there were 249 nodes and 1006 edges in total.
en, we further built the big hub nodes-compound network ( Figure 4) based on these 64 key targets. is network included 40 compound nodes and 64 large hub target nodes. Subsequently, we reordered these compound nodes in descending order of degree and found that Civetone was relevant to 25 large hub genes, Epiquinidine to 19 genes, quercetin to 13 genes, Gamma-Sitosterol to 7 genes, and Pinosylvin to 7 genes.

GO Biological Process Enrichment Analysis.
After sorting the 343 biological process (BP) terms in ascending order of adjusted P value, we found that the top 25 biological processes ( Figure 5) mainly concentrated on aspects of the reactive oxygen species metabolic process and its response to nutrient levels and epithelial cell proliferation. e following processes showed an enriched gene ratio: gland development (GO:0048732, 18.6%), response to nutrient levels (GO: 0031667, 18 .3%). Based on these BP enrichment analyses, we found that the anti-OPF effect of DZ may result from its regulatory effect on the reactive oxygen species metabolic process, response to nutrient levels, and epithelial cell proliferation.

Discussion
Osteoporosis is becoming a serious health burden due to the aging population. erefore, prevention of osteoporosis and of osteoporotic fracture has considerable social and economic significance [34]. Drug therapy, such as bisphosphonates and denosumab, is a practical way to treat osteoporosis by inhibiting osteoclast development, formation, and survival [35,36]. Although these antiresorptive drugs are effective in treating osteoporosis, the long-term use or high dosage of these drugs can cause serious side effects, such as atypical long bone fractures and jaw osteonecrosis [37,38]. In terms of fracture healing, the use of bisphosphonates remains controversial because callus remodeling is impaired in testing on ovariectomized rats [39]. Furthermore, these antiresorptive drugs cannot rebuild bone that is lost in the progression of osteoporosis, so there is a need to find a method capable of promoting bone building [40].
DZ has been found to alleviate improved osteoblast activity and reverse bone loss in ovariectomized (OVX) mice, which may result from the interaction between its active compounds and related targets. us, we investigated the anti-OPF mechanism of DZ through a network pharmacology method. Based on the big hub genes-compounds network (Figure 4), we found that five key compounds, including Civetone, Epiquinidine, quercetin, Gamma-Sitosterol, and Pinosylvin, play dominant roles in this network.
erefore, these active compounds may lay the foundation for the promising anti-OPF effect of DZ.
In the present study, 325 common gene targets of DZ and OPF were identified and 64 hub genes were identified that may play critical roles in this treatment. Some anti-OPF effects of these genes have been confirmed by clinical trials or animal experiments. Chow et al. found that low-magnitude high-frequency vibration stimulation is able to promote fracture healing by positively regulating the impaired innate immune response (lower expression of TNF-α and IL-6) and promoting macrophage polarization in delayed osteoporotic fracture healing [41]. Wang et al. suggested that organic gallium can improve osteoporotic fracture healing by positively modulating the OPG/RANKL ratio and inhibiting the expression of TNF-α and IL-6 [42]. Zhang et al. found that 17β-estradiol, which is used in the clinic to prevent and treat postmenopausal women with osteoporotic fracture, can significantly downregulate ROS production and the expression of TNF-α and IL-1β, as well as the phosphorylation of ERK2 (MAPK1) [43]. Wang et al. suggested that ESR1 is a susceptibility gene for osteoporotic fracture in postmenopausal Chinese women [44]. De-Ugarte et al. found that miR-320a, which is known to target CTNNB1 of osteoblasts, is overexpressed in osteoporotic fractures [45]. Swanberg et al. found that elderly women who were carriers of the IFNG variant alleles had lower BMD and a lower risk of incident fracture [46]. Based on a large population-based cohort study, Rivadeneira et al. found that ESR2 variants were related to an increased risk of fragility fracture in postmenopausal women [47]. ese gene targets play key roles in osteoblast differentiation and osteoclastogenesis during fracture repair. Moreover, the results presented in Figure 4 show that one active compound can interact with multiple genes, one gene with dicompound can interact with multiple genes, and one gene can interact with different compounds, which is in accordance with the modern drug theory of "multi-ingredient, multitarget" [48].
We conducted GO biological process and KEGG enrichment analyses of the overlapping genes to identify their functions. We found 44 (13.6%) genes involved in reactive oxygen species metabolic processes. Reactive oxygen species play key roles in regulating cell proliferation, apoptosis, migration, and differentiation [49], as well as in the regulation of osteoclasts and osteoblasts [50,51]. EGFR is the overlapping gene between reactive oxygen species (ROS) and the PI3K-Akt signaling pathway (27 genes enriched, 9.6%) based on the results of the GOBP and KEGG pathway enrichment analyses. e PI3K-Akt signaling pathway was first found in tumor cells and proved to regulate numerous cell functions, and it has been proven to be related to the promotion of osteogenesis by upregulating the proliferation and differentiation of bone marrow mesenchymal stem cells [52][53][54]. Furthermore, activation of the EGFR-Akt-mTOR pathway can effectively protect osteoblasts against dexamethasone, which increases the level of ROS production in osteoblastic cells [55]. MAPK1 is a common gene in reactive oxygen species, the PI3K-Akt signaling pathway, the MAPK signaling pathway (31 genes enriched, 11.1%), and the TNF signaling pathway. MAPKs play important roles in cell proliferation, differentiation, and apoptosis, as well as in regulating inflammation [56]. ROS can lead to serious periodontal tissue destruction by promoting osteoclastic bone resorption, which is closely related to the protein expression of MAPKs and NF-κB [57]. Srinivasa et al. found that the facilitation of ROS on osteoclast differentiation and resorption can be reversed by inhibiting the NF-κB and calcineurin-NFAT pathways [58]. e activation of MAPKs also plays a key role in osteoclastogenesis induced by RANKL. However, these ROS-directed upregulations of MAPK and NF-κB signaling can be attenuated by strengthening the nuclear factor-erythroid 2-related factor 2 [59]. erefore, the anti-OPF effect of DZ may depend on the regulatory effect on ROS through these key genes and pathways.
Forty-five (13.9%) genes were involved in the regulation of the inflammatory response. In vitro and in vivo studies have shown that the proinflammatory cytokines IL-6 and TNF-α are involved in the pathogenesis of osteoporosis and influence bone metabolism [60,61]. Several studies have also 8 Evidence-Based Complementary and Alternative Medicine found that high expression of inflammatory markers is closely related to increased bone loss [62][63][64] and that these markers are also risk factors for incidents of nontraumatic fractures. Women with increased expression of inflammatory markers have a 3-fold risk of hip fractures [65]. IL-6 and TNF-α are the overlapping genes between the regulation of inflammatory response (17 genes enriched, 6.1%) and the TNF signaling pathway (45 genes enriched, 13.9%) based on the results of the BP and KEGG pathway enrichment analyses. Ali Aydin et al. found higher expression levels of TNF-α and IL-6 in OVX model rats, and these cytokines reached their apex when a fracture occurred [66]. Moreover, these cytokines also increase the production of oxygen radicals, which can enhance osteoclastic activity while decreasing osteoblastic activity [67,68]. erefore, the anti-OPF effect of DZ may depend on the regulation of the inflammatory response through these key genes and pathways.

Conclusion
By utilizing network pharmacology, we investigated the potential targets of DZ and the underlying mechanism of its anti-OPF effects, which may be based on the regulation of ROS and the inflammatory response. According to the KEGG pathway enrichment results, we found that the PI3K-Akt signaling pathway, MAPK signaling pathway, and TNF signaling pathway may be the main pathways in treating OPF. us, we believe that the anti-OPF effect of DZ is mainly based on its direct or indirect regulation of the abovementioned potential targets and pathways and that DZ provides promising directions for future research, which is essential to reveal its exact regulatory mechanisms.

Conflicts of Interest
e authors declare that they have no conflicts of interest regarding the publication of this paper.