Xinqiang SONG, Yu ZHANG, Erqin DAI, Lei WANG，Zhiguo FENG

1. Department of Biological Sciences, Xinyang Normal University, Xinyang 464000, China; 2. School of Science, Qiongtai Normal University, Haikou 571127, China

Abstract [Objectives] To elucidate potential targets and mechanisms of action of Epimedium brevicornu in treating ovarian cancer. [Methods] The Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform was used to screen active components of E. brevicornu for disease control and prevention, and potential targets were collected from the DisGeNET database. These sets of bioactive and targets were analyzed using Ingenuity Pathway Analysis (IPA) to predict molecular networks affected by E. brevicornu in ovarian cancer. Venny 2.1.0 software was used to screen for proteins affected by interactions between disease and active components, which were input into the STRING 11.0 platform to construct a protein-protein interaction network. Then IPA and STRING were used to analyze common targets which were obtained from the two data analysis platform. [Results] A total of 23 major active components of E. brevicornu and 200 potential human targets were screened. IPA analysis identified 363 pathways and 24 networks shared between the set of predicted Yinyanghuo targets and ovarian cancer-associated proteins. These pathways are involved mainly in molecular mechanisms of cancer, glucocorticoid receptor signaling pathways, pancreatic adenocarcinoma signaling pathways, aryl hydrocarbon receptor signaling pathways, and macrophage function. The 24 networks have been implicated mainly in cancer, endocrine system disorders, body damage and abnormality, cell growth and proliferation, connective tissue development and function, tissue development, and other biological functions. IPA and STRING combined analysis suggested that AKT1, CASP3, JUN, FOS and CCND1 are the most likely targets of Yinyanghuo in treating ovarian cancer. [Conclusions] Our network pharmacology analysis identified several pathways that Yinyanghuo may influence to reduce ovarian cancer risk; in particular, it identified specific protein targets, including AKT1, CASP3, JUN, FOS and CCND1.

Key words Epimedium folium, Ovarian cancer, TCMSP, IPA, Network pharmacology

1 Introduction

Ovarian cancer is the fifth most common cause of cancer-related deaths in women and the most fatal gynecologic cancer. There are more than 200 000 new cases of ovarian cancer and more than 150 000 deaths annually worldwide[1-2]. Initial treatment options for epithelial ovarian cancer (EOC) include cytoreductive surgery followed by platinum and taxane chemotherapy[3-5], or neoadjuvant chemotherapy followed by interval debulking surgery[6-9]. Unfortunately, up to 75% of patients are diagnosed when their ovarian cancer is in an advanced stage, because the presenting symptoms are ill-defined and the disease can rapidly spread peritoneally, which makes screening difficult[10]. Thus, there is a great need to improve therapies available for ovarian cancer patients.

Traditional Chinese medicine, widely used in China and other countries, can be quite effective[11-13], but how it works is difficult to understand.Epimediumbrevicornuor "horny goat weed" (Yinyanghuo in Chinese) has been used to treat diseases for approximately 2 000 years. The Chinese name comes from the observation that goats’ sexual drive increases after they graze on these herbs. In traditional Chinese medicine theory, Yinyanghuo reinforces kidney yang, strengthens tendons and bones, and relieves rheumatic conditions[14-17]. Modern studies indicate thatEpimediumspecies exhibit many antiviral, antinociceptive, anti-aging, anti-oxidant, and neuroprotective effects; they also enhance immunity and promote estrogen biosynthesis. Several studies have documented the anti-cancer effects of Yinyanghuo[18-26], but the mechanism of action is unclear.

We decided to analyze the potential anti-cancer targets and mechanisms of action of Yinyanghuo using network pharmacology, which is based on the principles of network theory and systems biology[27]. Network pharmacology is built on the belief that drugs targeting multiple nodes in interconnected molecular systems, rather than individual molecules, may show better efficacy and fewer adverse effects. This fits well with the traditional Chinese medicine theory of multiple compounds and multiple targets[28-30]. Indeed, network pharmacology has been used effectively in several studies to predict main active ingredients and potential targets of traditional Chinese medicines[13,31-37].

Here we investigated the potential targets and mechanisms ofE.brevicornu(Yinyanghuo) using a network pharmacology approach. We first collected the chemical components and known targets ofE.brevicornuand ovarian cancer-associated target proteins from several databases. Afterwards, pathway enrichment analysis and network analysis were conducted using STRING and Ingenuity Pathway Analysis (IPA).

2 Materials and methods

The network-based analysis of Yinyanghuo active ingredients and their potential targets is shown schematically in Fig.1, and described in detail below.

Fig.1 Flow diagram for IPA of possible molecular mechanisms of Yinyanghuo reducing risk of ovarian cancer

2.1 Active ingredient screeningAll chemical ingredients inYinyanghuowere obtained from the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP,http://lsp.nwu.edu.cn/tcmsp.php)[38]. A combination of oral bioav-ailability (OB) screening and drug-likeness property evaluation was applied to explore the active substance ofYinyanghuo, Then, the potential compounds with OB≥30% and drug-likeness index≥0.18 were collected from the herbal constituents of Yinyanghuo.

2.2 Construction of Yinyanghuo target pathways and networksKnown and predicted targets of Yinyanghuo were obtained from the TCMSP and the Swiss Target Prediction database (www.swisstargetprediction.ch/). Using this website, it is able to estimate the most probable macromolecular targets of a small molecule, assumed as bioactive. The prediction is founded on a combination of 2D and 3D similarity with a library of 370 000 known actives on more than 3 000 proteins from three different species. Targets were uploaded to the IPA website (https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis/) accor-ding to symbols. Yinyanghuo targets pathways and networks were constructed using IPA and STRING platform.

2.3 Construction of ovarian cancer-associated protein pathways and networksOvarian cancer-associated proteins were obtained from DisGeNET (www.disgenet.org/), protein symbols were uploaded to the IPA website, and ovarian cancer-associated proteins pathways and networks were constructed using IPA and STRING platform.

2.4 Prediction of overlapping pathways and interaction networks affected by YinyanghuoOverlapping pathways and interaction networks affected by Yinyanghuo were constructed using IPA. Proteins related to ovarian cancer and proteins potentially targeted by Yinyanghuo were denominated "key molecules", which were represented as nodes with different shapes depending on their function. Lines were drawn between nodes that were previously associated with one another based on the literature or other canonical information in the Ingenuity Knowledge Base.

2.5 Construction of common networks and key molecules screeningProteins that were both previously associated with ovarian cancer and predicted to be targets ofYinyanghuowere collected using Venny 2.1.0 software (http://bioinfogp.cnb.csic.es/tools/venny/), and potential protein-protein interactions were analyzed in STRING 11.0 (https://string-db.org/) after filtering with "Homosapiens". STRING is a database of known and predicted protein-protein interactions. The interactions include direct (physical) and indirect (functional) associations; they stem from computational prediction, from knowledge transfer between organisms, and from interactions aggregated from other (primary) databases. Target proteins whose values for the topological attributes of node degree distribution and betweenness centrality were above the means were defined as "key molecules". The degree of a node is the number of nodes to which it is linked, while betweenness reflects the extent to which nodes lie between one another[39]. Finally, core targets were analyzed using IPA.

3 Results

3.1 Main active ingredients in Yinyanghuo and their potential targetsA total of 23 main chemical ingredients were collected from TCMSP (Table 1), each of which may target many proteins. Filtering by "Homosapiens", 200 targets were identified (Table 1).

Table 1 Main active components of Epimedium folium based on the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform

3.2 Pathways and networks containing Yinyanghuo targets

Yinyanghuo target symbols were imported into IPA, which generated 376 canonical pathways and 23 protein-protein interaction networks (Fig.2, only showed the top 10 canonical pathways and the top 2 networks). Proteins potentially targeted by Yinyanghuo participate primarily in cellular growth and proliferation; connective tissue development and function; cell death and survival; organismal injury and abnormalities; DNA replication, recombination, and repair; organismal injury and abnormalities; cardiovascular system development and function; cellular development, and cellular function and maintenance.

Note: Only the top 10 pathways are shown in panel (A).

3.3 Pathways and networks involving ovarian cancer-related proteinsA total of 2 202 ovarian cancer-associated proteins were obtained from DisGeNET and uploaded to IPA, which revealed 379 pathways and 25 networks. These pathways involve primarily molecular mechanisms of cancer, glucocorticoid receptor signaling, and IL-8 signaling. The networks involve mainly cancer, endocrine system disorders, organismal injury and abnormalities, gene expression, cellular development, embryonic development, cell cycle, neurological disease, organismal injury and abnormalities, and reproductive system disease (Fig.3).

Note: Only the top 10 pathways are shown in panel (A).

3.4 Overlapping pathways and interaction networksThe "Canonical Pathway" and "Networks" modules of IPA identified 363 pathways and 24 networks shared between the set of predicted Yinyanghuo targets and ovarian cancer-associated proteins. These overlapping pathways involve primarily molecular mechanisms of cancer, glucocorticoid receptor signaling, pancreatic adenocarcinoma signaling, aryl hydrocarbon receptor signaling, role of macrophages, fibroblasts and endothelial cells in rheumatoid arthritis. The networks involve primarily cancer, endocrine system disorders, organismal injury and abnormalities, gene expression, cellular development, embryonic development, cellular growth and proliferation, tissue development, cell death and survival, as well as connective tissue development and function (Fig.4).

Note: A: Signaling pathways; B-C: Networks involving predicted Yinyanghuo targets (purple lines) and ovarian cancer-associated proteins (black lines).

3.5 Overlapping targets and their interaction networksIntersections between the set of potential Yinyanghuo targets and ovarian cancer-related proteins were analyzed using Venny software, which identified 109 shared proteins (Fig.5A). STRING suggested that the 109 proteins can interact with one another via 2 207 interactions (edges) (Fig.5B). For protein-protein interactions, the mean degree distribution was 40.5, and the mean betweenness was 6.10E-3. A total of 31 potential protein targets showed higher values for degree and betweenness than these mean values, suggesting that they occupy particularly important positions in the protein-protein interaction networks. These may be particularly good targets for treating ovarian cancer.

Note: A: Venn diagram showing overlap; B: Protein-protein interactions among the overlapping proteins; circles represent proteins; lines between circles, protein-protein interaction; different colors, different interactions.

Table 2 Topological parameters of key targets

3.6 IPA prediction of ovarian cancer-related proteins targeted by YinyanghuoProteins linked to ovarian cancer and potentially targeted by Yinyanghuo participate in several canonical pathways subserving a range of biological activities. To demonstrate the ability of our integrative bioinformatics approach to propose specific protein targets for further mechanistic studies, we selected a top pathway in the IPA category "molecular mechanisms of cancer" that was linked to ovarian cancer and targeted by Yinyanghuo. Several nodes in this pathway emerged as potential direct targets of Yinyanghuo in ovarian cancer: RAS, JNK, AKT, CASP3, JUN, FOS and CCND1 (Fig.6). Combining these results with STRING analysis (shown in Section3.5) identified AKT1, CASP3, JUN, FOS, and CCND1 as particularly likely targets of Yinyanghuo in ovarian cancer.

Note: Proteins likely to be targeted by Yinyanghuo are marked with purple boxes. Triangles mean proteins with enzymatic activity and circles mean proteins with no enzymatic activity.

4 Discussion

In this study, we applied a network pharmacology approach to predict protein targets and mechanisms by which Yinyanghuo may treat ovarian cancer. Our analysis drew on large and publicly available databases of proteins previously linked to ovarian cancer and proteins known or predicted to be affected by Yinyanghuo. As a result of the large size of both databases, numerous signaling pathways and networks were identified to be potentially linked to ovarian cancer and potentially regulated by Yinyanghuo. Then we were able to predict several specific proteins likely to be affected by Yinyanghuo in ovarian cancer. These are strong leads for detailed mechanistic studies.

Our network analysis implicated several pathways by which Yinyanghuo may reduce ovarian cancer risk: organismal injury and abnormalities, gene expression, cellular development, embryonic development, cellular growth and proliferation, tissue development, cell death and survival, as well as connective tissue development and function. These results are consistent with several studiesinvitroandinvivosuggesting that Yinyanghuo exerts anticarcinogenic activity by inducing ovarian cancer cell apoptosis through activation of p53 and inhibition of Akt/mTOR pathway, by stimulating macrophages to secrete abundant anti-cancer cytokines, and by promoting the maturation of dendritic cells and their antigen presentation[40].

Our analysis of canonical pathways in IPA suggests that Yinyanghuo may reduce ovarian cancer risk by altering pathways involved in molecular mechanisms of cancer, glucocorticoid receptor signaling, pancreatic adenocarcinoma signaling, aryl hydrocarbon receptor signaling. In particular, we identified AKT1, CASP3, JUN, FOS and CCND1 as particularly promising potential targets of Yinyanghuo in ovarian cancer.

In agreement with our results, Yinyanghuo has been shown to inhibit the Akt/NF-κB pathway and thereby repress TNF-α-induced migration and endothelial-mesenchymal transition of cancer cellsinvitroandinvivo[20,26,41]. Similarly, for example, the Yinyanghuo component icariside II has been shown to increase the ratio of Bax/Bcl2, causing cytochrome c release and activation of caspases 3 and 9 activation in lung cancer cells and acute myeloid leukemia cells. We are unaware of studies linking JUN, FOS or CCND1 with ovarian cancer. Future work on these candidate therapeutic targets is justified.

5 Conclusions

Network pharmacology involves the application of systems biology approaches, combining with the pharmacokinetics and pharmacodynamics evaluations, to study their targets and effects of drugs. Network pharmacology analysis generally counts on a large number of genes or proteins to construct networks for evaluating the drug action and understanding the therapeutic mechanisms. As a major tool, the network analysis based on widely existed databases permits us to form an initial understanding of the action mechanisms within the context of systems-level interactions. This approach may be suitable for analyzing the mechanism of action of other bioactive compounds. Our network analysis allowed us to identify several pathways by which Yinyanghuo may reduce ovarian cancer risk. These pathways are involved mainly in organismal injury and abnormalities, gene expression, cellular development, embryonic development, cellular growth and proliferation, tissue development, cell death and survival, as well as connective tissue development and function. Our network analysis also allowed us to identify several specific proteins that Yinyanghuo may help regulate in ovarian cancer, including AKT1, CASP3, JUN, FOS and CCND1.