Systematic Elucidation of the Potential Mechanism of Erzhi Pill against Drug-Induced Liver Injury via Network Pharmacology Approach

Objective The purpose of this work was to investigate the bioactive compounds, core genes, and pharmacological mechanisms and to provide a further research orientation of Erzhi pill (EZP) on drug-induced liver injury (DILI). Methods At first, we collected information of bioactive compounds of EZP from Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP) and previous studies. And then, the targets related to bioactive compounds and DILI were obtained from 4 public databases. At last, Cytoscape was used to establish a visual network. Moreover, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses and network analysis were performed to investigate potential mechanism of EZP against DILI. Results A total of 23 bioactive compounds and 89 major proteins of EZP were screened out as potential players against DILI. Association for bioactive compounds, core targets, and related pathways was analyzed, implying that core targets related to these pathways are ALB, AKT1, MAPK1, EGFR, SRC, MAPK8, IGF1, CASP3, HSP90AA1, and MMP9, and potential mechanisms of EZP acting on DILI are closely related to negative regulation of apoptosis process, improvement of lipid metabolism, and positive regulation of liver regeneration process. Conclusion This study demonstrated the multicompound, multitarget, and multichannel characteristics of EZP, which provided a novel approach for further research the mechanism of EZP in the treatment of DILI.


Introduction
Drug-induced liver injury (DILI), which is defined as a liver injury due to medications, xenobiotics, and herbs that leads to liver dysfunction or abnormal liver serology, has been an important health concern around the world [1]. Crude annual incidence rate of DILI was 19.1 cases per 100000 inhabitants [2]. DILI can cause serious consequences. However, there are few drugs that have liver-protective effect. erefore, safe and effective drugs against DILI are urgently needed. For treating DILI, traditional Chinese medicine (TCM) has unique advantages and has been proven to have liver-protective effects [3].
Erzhi pill (EZP), which is composed of Ligustri lucidi fructus (LLF) and Ecliptae herba (EH) on the ratio of 1 : 1, is a classic TCM formula and widely used to treat hepatic disease with a long history in China. e history of EZP treating hepatic disease can be traced back to Wu Minji's series "Fu shou Jing Fang" in the Ming Dynasty. In TCM theory, EZP can tonify liver and kidney, nourish Yin, and stop bleeding, [4] which is applied for collapse of liver and kidney Yin deficiency. In our previous study, we have found that the bioactive compound of EZP shows liver-protective effect via enhancing the antioxidative defense system, suppressing the inflammatory response and cell apoptosis of liver [5]. However, this study still focused on "single target and single pathway." A holistic "multiple compounds, multiple targets, and multiple pathways" study is necessary to clarify how EZP produces liver-protective effect on DILI.
Network pharmacology, first proposed in 2007 [6], has become an effective tool to systematically analyze the mechanism of TCM formula with multiple compounds. Applications of network pharmacology to investigate mechanism of TCM have become an indispensable method for development of TCM [7]. In many previous studies, network pharmacology has successfully predicted potential targets and pathways of TCM [8][9][10]. erefore, network pharmacology has been proved to be an effective method to explore potential targets and pathways of TCM by analyzing network of biological systems.
However, studies about the liver-protective effect of EZP on DILI are absenct. For the first time, this study explored the protective effect of EZP on DILI through network pharmacology and bioinformatic analysis. Workflow of this work is shown in detail in Figure 1.

Collection of Bioactive Compounds of EZP. Information of compounds of EZP was collected from Traditional Chinese
Medicine Systems Pharmacology Database (TCMSP, http:// lsp.nwu.edu.cn/tcmsp.php, Version: 2.3), a website that can provide information of herbal ingredients and structures. In addition, TCMSP also provides absorption, distribution, metabolism, and excretion (ADME)-related parameters of herbal ingredients, such as oral bioavailability (OB), druglikeness (DL), and half-life [11]. OB and DL were used to filter bioactive compounds of EZP after data were collected from TCMSP. OB, a major pharmacokinetic parameter of orally administered drugs, is used to measure the speed and extent of drug absorption into blood circulatory system [12]. DL is a qualitative principle to predict possibility of a chemical compound to become a drug, which can be applied to help optimize pharmacokinetics and pharmaceutical features in drug development [9]. Only compounds with OB ≥ 30% and DL ≥ 0.18 were identified for further study. As per this consideration, some compounds were removed by ADME screening, but these ingredients were identified as the main constituents of EZP in previous studies. So, we also identified these compounds as bioactive molecules.

Establishment of Bioactive Compounds Potential Targets
Database. All the targets related to bioactive compounds of EZP were collected from PharmMapper (http://lilab-ecust. cn/pharmmapper/, Version 2017) and Swiss TargetPrediction (http://www.Swiss.Target.Prediction.ch/, 2019 version) by uploading the structure of bioactive compounds, which was acquired from e PubChem Compound Database (https://www.ncbi.nlm.nih.gov/pccompound) or drawn by Chem3D 16.0. PharmMapper and Swiss Target-Prediction are web servers for potential drug target prediction by reversed pharmacophore matching query compound against an in-house pharmacophore model database [13]. Only targets with a norm fit score (in PharmMapper) or Probability (in Swiss TargetPrediction) higher than 0.60 would be selected; the purpose of doing this is to ensure the reliability of prediction.

Construction of Target Database of DILI.
e targets related to DILI were acquired form DrugBank (https://www. drugbank.ca/, version 5.1.4) and GeneCards (https://www. genecards.org/). ese two databases illuminate relationship between targets and disease from different perspectives. DrugBank is a comprehensive, freely available database, from which the user can obtain information on detailed drug, drug target, drug action, and drug interaction of FDAapproved drugs or experimental drugs going through the FDA approval process [14]. GeneCards is also a comprehensive, freely available database, which provides information about targets related to disease, mutations and polymorphisms, gene expression, gene function, pathways, protein-protein interactions, and so on [15]. By searching the key word "drug-induced liver injury," the targets related to DILI were collected. On the website of DrugBank, targets related to DILI were filtered by approved drug by the FDA. For keeping the reliability of the target prediction, we only chose the FDA-approved drugs in DrugBank and the targets with a norm fit score higher than 20 in GeneCards.

Network Establishment and Pathway
Analyses. In order to investigate the possible mechanisms of EZP on DILI, common targets that related to DILI and putative targets of bioactive compounds were selected as EZP's targets against DILI. e nodes of network are bioactive compounds of LLF and EH networked with relevant disease targets [9]. All the targets were transferred to "ENTRY" by UniProt (https:// www.uniprot.org/) before the establishment of network. e networks were established by Cytoscape 3.7.1, an opensource software project that is used for integrating biomolecular interaction networks with high-throughput expression data and other molecular states into a unified conceptual framework [16]. e pathways of EZP related to DILI were analyzed based on e Database for Annotation, Visualization and Integrated Discovery (DAVID, https:// david.ncifcrf.gov/home.jsp, Vision 6.8), and KEGG (https:// www.kegg.jp/, Release 91.0). e results of GO and KEGG pathway enrichment were considered to have statistically significant and necessary functional mechanisms of DILI, when P < 0.05.

Protein-Protein Interaction (PPI) Data Collection.
Search Tool for the Retrieval of Interacting Genes (STRING, https://string-db.org/) was used to collect possible proteinprotein interactions (PPI) by uploading 89 common targets that related to DILI and putative targets of active compounds. Species was limited to "Homo sapiens" with a confidence score >0.4.

Bioactive Compounds' Screen of EZP.
A total of 166 compounds were collected from TCMSP: 119 in LLF and 47 in EH; among them, 5 compounds were duplicated and removed. erefore, 161 compounds were identified from EZP. After ADME screening by OB ≥ 30% and DL ≥ 0.18, 20 compounds, 13 compounds from LLF and 9 compounds from EH with two repeated compounds (luteolin and quercetin), were identified as bioactive compounds of EZP (Figures 2(a) and 2(b)). Furthermore, some compounds (oleanolic acid, salidroside, and specnuezhenide) were removed by ADME screening, but these ingredients were identified as the main constituents of EZP in previous studies [17,18]. At last, 23 compounds were identified as potential bioactive molecules for further study. e results of selected 23 compounds from LLF and EH are presented in Table 1.        (Figure 3(a)). Active compounds associated with selected overlapping targets are listed in Supplementary Table S3. A visual EZP common-target network with 112 nodes (including 23 bioactive compounds and 89 targets) and 883 edges was established by Cytoscape (Figure 3(b)).

PPI Network of Common
Targets. PPI network was obtained from STRING database by uploading 89 common targets. A combined score >0.4 and "Homo sapiens" was selected. And then, we established PPI network, which had 84 nodes and 811 edges by Cytoscape. In this network, the protein with greater degree was described by larger node and darker color, and the edge with greater combined score was described by thicker and darker line ( Figure 4). 10 targets with highest degree score were select as core targets for DILI. e core targets, which may play an essential role against DILI, were serum albumin (ALB), RAC-alpha serine/threonine-protein kinase (AKT1), mitogen-activated protein kinase 1 (MAPK1), epidermal growth factor receptor (EGFR), insulin-like growth factor I (IGF1), caspase-3 (CASP3), proto-oncogene tyrosine-protein kinase Src (SRC), mitogen-activated protein kinase 8 (MAPK8), heat shock protein HSP 90-alpha (HSP90AA1), and matrix metalloproteinase-9 (MMP9).

GO and KEGG Pathway Enrichment Analysis.
In order to explore possible mechanism of EZP against DILI, we analyzed GO term and KEGG pathway enrichment results executed by DAVID. GO term enrichment results were divided into biological process (BP), cell compound (CC), and molecular function (MF). Top 10 enriched conditions in BP and top 5 in CC and MF were shown in Figure 5(b). As shown in Supplementary Table S4, 223 BPs, 25 CCs, and 65 MFs enriched for these targets have a P value less than 0.05. In GO term enrichment, the biological process of EZP against DILI may relate to negative regulation of apoptotic  Figure 5(a) and  regulation of actin cytoskeleton, progesterone-mediated oocyte maturation, and so on. Interestingly, the results of pathways enrichment analysis can be divided into two function modules, including cell cycle regulation (such as MAPK, PI3K-Akt, and VEGF signaling pathways) and metabolic pathway (such as insulin signaling pathway, and central carbon metabolism in cancer).

Discussion
DILI, which carries a high mortality rate, [23] has been a major public concern impacting patients, doctors, drug researchers, and drug regulators [24]. TCM has its unique advantages to treat complex disease for "holistic treatment concept." However, multicompound and multitarget characteristics of TCM also brought a lot of difficulties for Chinese medicine research and restrained the development of TCM. Fortunately, network pharmacology, which is especially suitable for multicompound and multitarget research, provides a prospective method to solve this problem.
In this study, we predict and analyze the potential mechanisms of EZP from the perspective of systematic network pharmacology method. e results of GO and KEGG enrichment analysis indicated that mechanisms of EZP against DILI may be closely associated with negative regulation of apoptosis process, improvement of lipid metabolism, and positive regulation of liver regeneration process.
According to GO term enrichment results, negative regulation of apoptotic process was the biological process with most targets (22 targets) involved with P < 0.05. Necrosis and apoptosis of hepatocytes, cholangiocytes, and endothelial cells are typical features of DILI. IGF1, SRC, ALB, MMP9, CASP3, EGFR, AKT1, and MAKP8, which are involved in this biological process, are included in the top ten targets of the PPI network with highest degree. As we know, CASP3 is a key enzyme in the execution of apoptosis. Evidence has shown that there is a significant upregulation of CASP3 in DILI [25]. In addition, there are two MAPK proteins involved in core targets. MAPK1 (extensively known as extracellular signal-regulated kinase 2, ERK2) takes party in multiple cellular processes such as cell proliferation, differentiation, adhesion, migration, and survival [26]. MAPK8 (known as c-Jun N-terminal kinase 1, JNK1) has diverse functions in cell cycle, such as cell death, regeneration, and differentiation [27]. ese 3 genes (CASP3, MAPK1, and MAPK8) are all involved in MAPK signaling pathway ( Table 2). ese results indicated that negative regulation of apoptotic process and these proteins may play an essential role in EZP against DILI. In addition, KEGG enrichment analysis also showed that the mechanisms of EZP against DILI are closely related to PI3K-Akt, FoxO, MAPK, and VEGF signaling pathways. It is interesting to note that those pathways are all associated with cell cycle. In a previous study, it has been proved that the effect of EZP e protein with greater degree is described by larger node and darker color, and the edge with greater combined score is described by thicker and darker line. 10 Evidence  Evidence-Based Complementary and Alternative Medicine inhibition of hepatocyte apoptosis was closely associated with PI3K-Akt signaling pathway [28]. MAPK signaling pathway comprises the classic MAP kinase pathway, JNK and P38 MAP kinase pathway, and ERK5 pathway. Among them, JNK and P38 MAP kinase pathway is closely related to DILI. Drugs can be metabolized by P450s to reactive metabolites, which can activate JNK pathway to induce apoptosis through the recruitment of Bax [27]. By analyzing the results of KEGG pathway enrichment, we also found that insulin signaling pathway and insulin resistance have a significant result in KEGG pathway enrichment analysis as shown in Figure 5(a). ese pathways are emerged as key players in glucose and lipid metabolism. Drug-induced steatohepatitis (DIS), which pathological feature is intracellular accumulation of lipids in hepatocytes, is another form of DILI [29]. e mechanisms of DIS can be aligned with the four aspects: increased fatty acid synthesis; decreased lipoprotein export; decreased fatty acid β-oxidation; and increased mobilization and uptake of fatty acids [29][30][31]. ese pathways show that EZP may have potential for improving lipid metabolism function, which is beneficial to ameliorate DILI, by mediating the inhibitory action of insulin or insulin-like growth factor. In addition, these pathways also take part in cell metabolism, differentiation, oxidative stress, autophagy, and aging [32].
PPI network analysis, as well as GO and KEGG pathway analysis indicated that there were 1 core target and 2 pathways closely associated with liver regeneration, namely VEGF, VEGF signaling pathway, and PI3K-Akt signaling pathway. e liver is an organ with strong ability to regenerate. ere are three phases, priming stage, proliferative phase, and termination phase, involved in the overall process of liver regeneration [33]. e VEGF, a core target of PPI network, belongs to the angiogenic factors that potently involves in endothelial cell proliferation and survival in liver regeneration following damage [34]. VEGF promotes proliferation of hepatocytes through reconstruction of liver sinusoids by proliferation of sinusoidal endothelial cells [35]. As described above, PI3K-Akt signaling pathway emerged as a key player in negative regulation of apoptosis. However, the regulation of liver regeneration is dual through PI3K-Akt signaling pathway.
On the one hand, PI3K-Akt signaling pathway plays an essential role in liver regeneration, which has been testified [36]. On the other hand, PI3K inhibition can diminish the expression of IL-6 and TNF-α, which ultimately leads to Evidence-Based Complementary and Alternative Medicine 13 attenuated regeneration [34]. Hence, EZP's effect on liver regeneration via PI3K-Akt signaling pathway is complex and needs further research.
In this work, we investigate the potential mechanism of EZP against DILI; however, network pharmacology is just a prediction. Whether EZP acts against DILI by regulating these pathways and proteins needs further experimental verification.

Conclusion
In summary, this study explored the protective effect of EZP on DILI through network pharmacology and bioinformatic analysis for the first time. 23 bioactive compounds of EZP and 89 targets associated with DILI were identified, and 10 core targets were identified by analyzing PPI network analysis. GO and KEGG pathway enrichment analysis indicates that the mechanisms of EZP against DILI may be related to negative regulation of apoptosis process, improvement of lipid metabolism, and positive regulation of liver regeneration process through PI3K-Akt, MAPK, Foxo, VEGF, and insulin signaling pathways, as well as insulin resistance.

Data Availability
e data used to support the findings of this study are included within the article and the supplementary information file(s).

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
Shao-jie Huang, Fei Mu, Fei Li, and Wen-jun Wang contributed equally to this work. Table S1: putative targets for chemicals contained in EZP. Chemname was obtained from TCMSP. Entry, Entry name, and Protein name were obtained from PharmMapper. Entry was converted to Gene name by Uniprot. (https://www. uniprot.org/) Table S2: targets related to DILI. Detailed information of targets related to DILI, there were 267 targets from DrugBank and 357 targets from GeneCards with 42 targets duplicated. Table S3: common targets related to bioactive compounds. Detailed information of 89 common targets related to 23 bioactive compounds.