Abstract Purpose: This study aimed to identify the active components of Fuzheng Huayu (FZHY) formula and the mechanism by which they inhibit the viability of hepatic stellate cells (HSCs) by a combination of network pharmacology and transcriptomics. Methods: The active components of FZHY formula were screened out by text mining. Similarity match and molecular docking were used to predict the target proteins of these compounds. We then searched the STRING database to analyze the key enriched processes, pathways and related diseases of these target proteins. The relevant networks were constructed by Cytoscape. A network analysis method was established by integrating data from above network pharmacology with known transcriptomics analysis of quiescent HSCs-activated HSCs to identify the most possible targets of the active components in FZHY formula. A cell-based assay (LX-2 and T6 cells) and surface plasmon resonance (SPR) analysis were used to validate the most possible active component-target protein interactions (CTPIs). Results: 40 active ingredients in FZHY formula and their 79 potential target proteins were identified by network pharmacology approach. Further network analysis reduced the 79 potential target proteins to 31, which were considered more likely to be the target proteins of the active components in FZHY formula. In addition, further enrichment analysis of 31 target proteins indicated that the HIF-1, PI3K-Akt, FoxO, and chemokine signaling pathways may be the primary pathways regulated by FZHY formula in inhibiting the HSCs viability for the treatment of liver fibrosis. Of the 31 target proteins, peroxisome proliferator activator receptor gamma (PPARG) was selected for validation by experiments at the cellular and molecular level. The results demonstrated that schisandrin B, salvianolic acid A and kaempferol could directly bind to PPARG, decreasing the viability of HSCs (T6 cells and LX-2 cells) and exerting anti-fibrosis effects. Conclusion: The active ingredients of FZHY formula were successfully identified and the mechanisms by which they inhibit HSC viability determined, using network pharmacology and transcriptomics. This work is expected to benefit the clinical application of this formula. Keywords: Fuzheng Huayu formula, network pharmacology, transcriptomics, liver fibrosis, pharmacological mechanism Introduction Liver fibrosis is a protective mechanism of organ integrity with the accumulation of extracellular matrix and collagen proteins, which occurs as a result of large-scale apoptosis and necrosis ([35]Fagone et al., 2016). It is mainly associated with chronic viral hepatitis type B or viral hepatitis type C infection, biliary diseases, alcoholic steatohepatitis, and nonalcoholic steatohepatitis ([36]Friedman, 2008). In 2004, the leading cause of liver fibrosis was hepatitis type B in China, and viral hepatitis type C, alcoholic steatohepatitis and nonalcoholic steatohepatitis in United States and most European countries ([37]Afdhal, 2004; [38]Lavanchy, 2004; [39]McMahon, 2004). Without timely and effective treatment, liver fibrosis will gradually evolve into cirrhosis and even liver carcinoma, leading to death. A growing body of evidence has shown that liver fibrosis is a reversible process ([40]Hernandez-Gea and Friedman, 2011; [41]Wang et al., 2015; [42]Zhang C.Y. et al., 2016). Since the activation of hepatic stellate cells (HSCs) is the primary process of hepatic fibrosis ([43]Schuppan and Kim, 2013), inhibiting the activation of HSCs should be an effective treatment ([44]Liu et al., 2015). Currently, most researchers developed new drugs for liver fibrosis based on the “one drug for one target for one disease” assumption ([45]Mao and Fan, 2015; [46]Koyama et al., 2016; [47]Wang P. et al., 2016). For example, Sorafenib can inhibit the proliferation of HSCs through targeting tyrosine kinase ([48]Mejias et al., 2009); and Rimonabant, a cannabinoid receptor type 1 antagonist, can promote apoptosis of activated HSCs ([49]Dai et al., 2014). But the clinical utility of both drugs is limited by their off-target effects to cause severe adverse reactions. And network pharmacology studies concluded liver fibrosis to be of a complicated etiology, unamenable to effective treatment by intervention at a single node ([50]Liu et al., 2006). Accordingly, safer and more effective treatment strategies for hepatic fibrosis are urgently needed. Traditional Chinese Medicine (TCM) formulae have been widely used in clinical practice for thousands of years due to their efficacy and lack of serious side effects, and are an indispensable part of the current medical service system ([51]Zhang Y. et al., 2016). One characteristic of TCM formulae is that they generally comprise multiple ingredients, which operate in synergy and on multiple disease targets ([52]Li and Zhang, 2013). FZHY recipe, a classic TCM formula originally called Ganping capsule and 319 recipe, has been approved as an anti-fibrotic medicine by the State Food and Drug Administration. And it recently completed phase II clinical trials in the United States. This formula consists of six different herbs: Salviae miltiorrhizae Radix et Rhizoma (derived from Salvia miltiorrhiza Bge., Danshen; S. miltiorrhiza), Semen Persicae [derived from Prunus persica (L.) Batsch, Tao Ren; S. Persicae], Cordyceps sinensis [derived from Cordyceps sinensis (BerK.) Sacc., Dong Chong Xia Cao; C. sinensis], Gynostemma pentaphyllum [derived from Gynostemma pentaphyllum (Thunb.) Makino, Jiao Gu Lan; G. pentaphyllum], Schisandra chinensis Fructus [derived from Schisandra chinensis (Turcz.) Baill., Wu Wei Zi; S. chinensis] and Pollen Pini (derived from Pinus massoniana Lamb., Song Hua Fen; P. Pini). It has been shown to significantly improve liver function and clinical symptoms, decrease collagen synthesis, promote degradation of extracellular matrix, reverse hepatic fibrosis, and reduce mental stress in patients with chronic hepatitis B and liver cirrhosis ([53]Jia et al., 2012). Furthermore, the FZHY formula can reduce oxidative stress by down-regulating cytochrome P450 2E1 and tumor necrosis factor receptor type I ([54]Jia et al., 2012) expression. Also, several researchers reported the antifibrotic capability of the FZHY recipe in the treatment of liver fibrosis through inhibition of HSC activation ([55]Tao et al., 2014; [56]Pan et al., 2015; [57]Li et al., 2016). However, neither the active components of the FZHY formula nor their therapeutic target proteins in inhibiting HSCs viability are clear. In the emerging paradigm of network pharmacology, the concepts of drug discovery and drug development are moved from a “one target-one drug” mode to a “multi-target-multi-component” mode ([58]Li, 2011; [59]Li et al., 2011; [60]Xu et al., 2014). This latter mode in network pharmacology coincides well with the integrity and systemic nature of TCM formulae, which could solve above problems of the FZHY formula. Previous studies in our laboratory confirmed the utility of network pharmacology to clarify the synergistic molecular mechanisms of Sini decoction and Ku Shen ([61]Yang et al., 2013; [62]Chen et al., 2014; [63]Wang S.Q. et al., 2016). However, network pharmacology approaches routinely yield hundreds of potential target proteins for the active components in TCM formula ([64]Liu et al., 1999; [65]Jia et al., 2012; [66]Wang et al., 2012), and selection of the target proteins most suitable for validation is difficult. As a combination of approaches is thought most likely to bear fruit, we hypothesized that the combination of network pharmacology and transcriptomics could increase the accuracy of target identification of network pharmacology. In this study, network pharmacology and transcriptomics were used to identify the active ingredients in the FZHY formula, and the mechanisms by which they diminished HSCs viability. Subsequently, these predictions were verified in a series of experiments. A detailed flowchart is depicted in Figure [67]1. This is the first study to contemplate the mechanism of action of FZHY formula in the inhibition of HSCs viability for the treatment of liver fibrosis by the methods of network pharmacology and transcriptomics. FIGURE 1. FIGURE 1 [68]Open in a new tab The flowchart of network analysis approach. Materials and Methods Database Construction Information (structure, canonical name, and component identification number) pertaining to the compounds present in FZHY formula was obtained from the TCMs Integrated Database^[69]1 ([70]Xue et al., 2013), the TCM Database@Taiwan^[71]2 ([72]Chen, 2011), and the Chemistry Database^[73]3. The active compounds of all above herbs were found out by text mining from the National Center for Biotechnology Information PubMed database^[74]4. The search bar was comprised of a compound name in FZHY formula and liver fibrosis or hepatic fibrosis. Candidate target proteins relating to liver fibrosis were obtained by searches of the PubMed database. The most appropriate crystallographic structures of each target protein were then downloaded from the UniProt Database^[75]5 and RCSB Protein Data Bank^[76]6. Target Prediction Potential target proteins of the FZHY formula were identified using two methods. Firstly, since approved drugs of a similar structure may have common target proteins and similar curative effects, we used the drug similarity search tool in ChEMBL^[77]7 to identify compounds similar to the active components of the FZHY formula. Only compounds with a high similarity score (≥0.95) as compared with the structures of active components in FZHY formula were picked out, in order to obtain more accurate results. The therapeutic target proteins of these similar compounds were also collected in ChEMBL; those implicated in liver fibrosis were hypothesized to be targeted by the related active component of the FZHY formula. Secondly, molecular docking, which is very useful in rational drug design, can be used not only to predict the binding sites and pose(s) of drug candidates with their target proteins, but also to evaluate their binding affinities ([78]Kitchen et al., 2004). We used libdock in Discovery Studio 3.0^[79]8 to predict possible relationships between the target proteins of liver fibrosis and the active components of FZHY. The co-crystallized ligand binding with the target protein was regarded as a positive control. The dock scores of the positive control with corresponding proteins were regarded as cutoff point. If the dock score is higher than the cutoff point, the protein will be recognized as a potential target protein of the compounds. Based on these results, an active component-target protein interaction (CTPI) network can be constructed and displayed using Cytoscape 3.5.1 ([80]Smoot et al., 2011). A component and a related potential target protein are linked with an edge; and components or target proteins are represented by nodes. Gene Expression Profiles Gene expression microarray data ([81]GSE68001) of primary human quiescent HSCs and in vitro activated HSCs were obtained from the National Center for Biotechnology Information Gene Expression Omnibus^[82]9, a public functional genomics data repository. Network Construction and Analysis Based on the acquired identities of the potential target proteins and differential genes relating to the activation of HSCs, a protein-protein interaction (PPI) network was built by importing the gene names of above proteins and genes to the public database STRING (version 10.5^[83]10). The minimum required interaction score was set at 0.9, to improve the accuracy of the results. Cytoscape 3.5.1 ([84]Smoot et al., 2011) were used as a tool to visualize the PPI network. Subsequently, PPI combined with CTPI was used to build a new compound-target protein-differential genes (CTPG) network. This network analysis process can identify target proteins which can connect compounds in FZHY formula with differential genes. To facilitate scientific interpretation of identified potential targets, a STRING database was used to perform several analyses such as gene ontology (GO) enrichment analysis and pathway enrichment analysis. Quality Control of FZHY Formula The FZHY formula was purchased from Yifeng Pharmacy (Shanghai, China), and extracted by sonication in methanol for 60 min. UPLC-Q-TOF/MS was used to identify the active components of FZHY. Mass spectra was acquired in both negative and positive modes, and non-target compound identification was further conducted based on obtained fingerprints. Formulae were proposed based on the mass spectra and other rules, such as the general rule of the number of nitrogen atoms, double bond equivalent (DBE) index and ‘show isotopic’ function. And the compound was finally confirmed by the comparison with the authentic compound. Materials and Reagents Salvianolic acid B, dihydrotanshinone I, salvianolic acid A, tanshinone II-A, amygdalin, adenosine, cordycepin, schizandrin, schisandrin B, schisantherin A, gypenoside XLIX and kaempferol were purchased from EFEBIO (Shanghai, China^[85]11), and their structural information is shown in Supplementary Figure [86]S1. 15-deoxy-Δ12,14-prostaglandin J2 (15d-PGJ2) was used as a positive control ([87]Jin et al., 2016), and obtained from Abcam (Cambridge, United Kingdom). The structures of the above chemicals were unambiguously identified by ^1H NMR and MS spectra, and their purity was demonstrated to be 98% by HPLC-UV. Cell Counting Kit-8 (CCK8) detection kit was obtained from Beyotime (Shanghai, China). LX-2 cells and T6 cells were purchased from Cell Resource Center of Fudan IBS (Shanghai, China). These cells were incubated in Dulbecco minimal essential medium (Sigma, United States) with 10% fetal bovine serum (GIBCO, United States) and penicillin, streptomycin (GIBCO, United States) under a humidified atmosphere with 5% CO[2] at 37°C. The medium was renewed every 2 days. Cell Proliferation Assay A total of 5 × 10^3 cells were planted in 96 well plates and cultivated for 24 h. Then, cells were treated with various concentrations of compounds (6.25, 12.5, 25, 50, 100, 200 μM) and 15d-PGJ2 (100 μM) for 24 h. Cell viability was tested by CCK-8; the absorbance was directly detected using a Bio-Rad microplate reader (Synergy^TM 4, BioTek, United States) at 450 nm. All experiments were repeated three times. Surface Plasmon Resonance (SPR) Analysis Surface plasmon resonance (SPR) analyses were undertaken at 25°C on a BIA core T200 instrument (GE Healthcare, Little Chalfont, Buckinghamshire, United Kingdom), using a phosphate buffered solution with 5% dimethyl sulfoxide as running buffer, with a constant flow rate of 30 ml/min. Peroxisome proliferator-activated receptor gamma (PPARG, Proteintech) protein was immobilized on CM5 chips with levels of 2446.2 by applying 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide/N-hydroxy succinimide cross-linking reaction. The detection was performed according to the protocol provided by GE Healthcare. Gradient concentrations of components (0.5–256 μM) were dissolved in the running buffer and then injected into the channel for 60 s followed by disassociation for 120s. BIA evaluation 3.0 software (BIAcore) was used to analyze the data by a 1:1 binding model. Results and Discussion Active Components and Their Potential Target Proteins of FZHY Formula for the Treatment of Liver Fibrosis Forty active compounds of FZHY formula were retrieved from the PubMed database (Table [88]1). A molecular docking and similarity match were used to identify potential target proteins of active compounds in the FZHY formula. A total of 79 potential target proteins (Supplementary Table [89]S1) were obtained, of which 45 potential target proteins came from similarity match, and 39 from docking with targets of liver fibrosis (Supplementary Table [90]S2). Interestingly, the PPARG, STAT3, PGFRB, KS6B1, MK08 target proteins were found by both two approaches. Specifically, there were 14 active compounds in S. miltiorrhiza targeting 43 potential proteins; 12 active compounds in C. sinensis targeting 32 potential proteins; 6 active compounds in G. pentaphyllum targeting 31 potential proteins; one active compound in P. Pini targeting 35 potential proteins; five active compounds in S. Persicae targeting 31 potential proteins; and five active compounds in S. chinensis targeting 22 potential proteins. Table 1. Active components identified by in six herbs. Herbs Number Components S. miltiorrhiza 14 Salvianolic acid A, baicalin, dihydrotanshinone I, rosmarinic acid, ursolic acid, danshensu, ferulic acid, magnesium lithospermate B, protocatechuic aldehyde, tanshinol, rutin, tanshinone II-A, salvianolic acid B, β-sitosterol C. sinensis 12 Histidine, ergosterol, valine, adenosine, ascorbic acid, vitamin B12, vitamin A, nicotinic acid, glycine, cordycepin, linoleic acid, β-sitosterol S. chinensis 8 β-caryophyllene, β-elemene, schizandrin, vitamin k1, schisandrol B, schisantherin A, schisandrin B, β-sitosterol G. pentaphyllum 7 Ginsenoside-rb1, rutin, ginsenoside-rb2, gypenoside XLIX, gipsoside, gypenoside A, β-sitosterol S. Persicae 4 Chlorogenic acid, amygdalin, (+)-catechin, β-sitosterol P. Pini 1 Kaempferol [91]Open in a new tab Compound-Target Protein Network Construction and Analysis We used the above data to construct a CPTI (Figure [92]2), which contains 119 nodes (40 active compounds and 79 potential targets) and 172 edges. In this network, the rectangles and ellipses represent the active compounds and their target proteins, respectively. There are two red rectangles in the middle of the blue ellipse, which represent the common components of several herbs in FZHY formula. Specifically, β-sitosterol was present in S. miltiorrhiza, S. Persicae, C. sinensis, G. pentaphyllum, and S. chinensis; and rutin was an ingredient of both S. miltiorrhiza and S. chinensis β-sitosterol and rutin may be the key active compounds of FZHY formula. As shown in Figure [93]2, there are more active compounds in S. miltiorrhiza than in the other five herbs, and the active compounds in S. miltiorrhiza targeted the largest number of proteins compared with the active compounds in other five herbs. Thus, we infer that S. miltiorrhiza is the principal ingredient of FZHY, which is consistent with previous work ([94]Zhao C.Q. et al., 2006; [95]Yang et al., 2015). FIGURE 2. FIGURE 2 [96]Open in a new tab The Component-Target protein network. The rectangles represent the 45 candidate compounds in Fuzheng Huayu (FZHY), the cyan, red, green, yellow, purple, pink rectangles refers to active compounds in Pinus massoniana Lamb. (Song Hua Fen), Prunus persica (L.) Batsch (Tao Ren), Salvia miltiorrhiza Bunge (Dan Shen), Schisandra chinensis (Turcz.) Baill. (Wu Wei Zi), Cordyceps sinensis (BerK.) Sacc. (Dong Chong Xia Cao), Gynostemma pentaphyllum (Thunb.) Makino (Jiao Gu Lan), respectively. The red rectangles in the center of network represent the common components of several herbs. The blue circles represent the gene names of target proteins of the six herbs found by text mining and molecular docking. STRING (version 10.5, see foot note text 10) is a database which aims to collect and integrate all functional interactions between expressed proteins by consolidating known and predicted protein-protein association data for a large number of organisms. This platform can also be used to conduct functional enrichments of user inputs. In this study, we performed the gene ontology enrichment analysis and pathway enrichment analysis of potential target proteins with the functions in STRING. The meaningful pathways (Supplementary Table [97]S4), biological processes (Supplementary Table [98]S3), cellular components (Supplementary Table [99]S5) and molecular functions (Supplementary Table [100]S6) were selected with a p-value < 0.05. Figure [101]3A depicts the enriched molecular functions of the target protein, which are mainly associated with binding and steroid hormone receptor activity. The binding activities are primarily connected with protein binding and enzyme binding. As to the cellular components’ distribution, the target proteins were mainly distributed in nuclear part, organelle part and organelle lumen (Figure [102]3A). The main biological processes of the target proteins are also summarized in Figure [103]3A, which shows that the active ingredients in FZHY formula could exert an anti-liver fibrosis effect through cellular response to endogenous stimulus, positive regulation of macromolecule biosynthetic process, response to endogenous stimulus, regulation of cell death, positive regulation of nucleobase-containing compound metabolic process, transcription initiation from RNA polymerase II promoter, positive regulation of macromolecule metabolic process, cellular response to hormone stimulus, positive regulation of cellular biosynthetic process, and positive regulation of transcription. Among these processes, FZHY formula had been reported to cure liver fibrosis by positive regulation of macromolecule metabolic process ([104]Gao et al., 2016) and positive regulation of macromolecule biosynthetic process ([105]Liu et al., 2006; [106]Cheng et al., 2013). In addition, cellular responses to endogenous stimulus ([107]Huang et al., 2016) and regulation of cell death ([108]Zhong et al., 2017) have been reported to be closely related to liver fibrosis, further experiments are needed to validate the interaction between FZHY formula and these two biological processes. FIGURE 3. FIGURE 3 [109]Open in a new tab (A) The enrichment analysis in biological processes, cellular components and molecular functions of 79 identified target proteins by STRING database. (B) The Target protein-Pathway network. Blue nodes refer to target proteins. The pink nodes represent pathways, and the color is consistent with pathways related target protein numbers, the deeper the pink, the more number of proteins. In order to identify the significant pathways that the target proteins are involved in, we made a pathway enrichment analysis of target proteins by STRING database (Figure [110]3B). Logically, the pathway that contains more target proteins is more important than the pathway that contains fewer target proteins. Hence, a target protein-pathway network (Figure [111]3B) was constructed to indicate the most important pathways. The results showed that the FZHY formula exerted its protective effects against liver fibrosis primarily by regulating 15 pathways (Figure [112]3B), of which the TGF-beta signaling pathway has been reported ([113]Wang et al., 2012). In addition, although lots of references indicated that liver fibrosis was closely related to the