Abstract Network pharmacology is considered as the next paradigm in drug discovery. In an era when obesity has become global epidemic, network pharmacology becomes an ideal tool to discover novel herbal-based therapeutics with effective anti-obesity effects. Zanthoxylum bungeanum Maxim (ZBM) is a medicinal herb. The mature pericarp of ZBM is used for disease treatments and as spice for cooking. Here, we used the network pharmacology approach to investigate whether ZBM possesses anti-obesity effects and reveal the underlying mechanism of action. We first built up drug–ingredient–gene symbol–disease network and protein–protein interaction network of the ZBM-related obesity targets, followed by Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analyses. The results highlight apoptosis as a promising signaling pathway that mediates the anti-obesity effects of ZBM. Molecular docking also reveals quercetin, a compound in ZBM has the highest degree of connections in the compound-target network and has direct bindings with the apoptotic markers. Furthermore, the apoptotic effects of ZBM are further validated in 3T3-L1 adipocytes and in the high-fat diet–induced obesity mouse model. These findings not only suggest ZBM can be developed as potential anti-obesity therapeutics but also demonstrate the application of network pharmacology for the discovery of herbal-based therapeutics for disease treatments. Keywords: Zanthoxylum bungeanum Maxim, network pharmacology, high-fat diet, obesity, apoptosis Introduction Obesity has reached epidemic proportions globally. Based on the data reported by the Centers for Disease Control and Prevention, the prevalence of obesity in the United States has increased from 30.5 to 42.4%, and the prevalence of severe obesity has increased from 4.7 to 9.2% in the last decade ([40]Agha and Agha, 2017). Asian countries such as China have 46% of adults being obese or overweight (Wang et al., 2019b). Moreover, the prevalence of obesity or overweight in youngsters and childhood is also increasing worldwide. The WHO reports that the number of obese children and adolescents had a tenfold increase, which had already reached 124 million in 2016, and 216 million children in the world were overweight. Obesity is associated with many comorbid conditions and is the main risk factor for many noncommunicable diseases. Overweight and obesity are closely associated with polycystic ovary syndrome, which is an endocrine condition that causes enlarged ovaries, prevents proper ovulation, and reduces fertility ([41]Barthelmess and Naz, 2014). Obesity is also the risk factor for type 2 diabetes, high blood pressure, heart disease and strokes, sleep apnea, osteoarthritis, fatty liver disease, kidney disease, and certain types of cancers such breast cancer, colorectal cancer, and kidney cancer ([42]Pi-Sunyer, 2009). Every year, at least 2.8 million people are dying because of being overweight or obese. Obesity and its associated conditions cast a heavy burden on the health sector. Recently, network pharmacology has been used to explore the therapeutic effects and therapeutic targets of Chinese medicinal herbs and bioactive compounds. The “network target, multi-components” concept of network pharmacology is the most suitable tool to explore the therapeutic effects of herbal medicine at the molecular level ([43]Li and Zhang, 2013; [44]Zhang et al., 2013). The network pharmacology approach is a new research paradigm that facilitates the development of evidence-based medicine and novel herbal-based drug discovery. Zanthoxylum bungeanum Maxim (ZBM) is a Chinese medicinal herb; it belongs to the family Rutaceae and is widely distributed in south central and southwest China. ZBM has been recorded in “Shen Nong’s Herbal Classic” and “Compendium of Materia Medica” and is described as “hot, nontoxic.” In Chinese medicine practice, the mature pericarp of ZBM is used to treat colds, stomach and abdomen pains, vomiting, and diarrhea, while the seeds are used to treat edema, tumescence, and dyspnea due to phlegm and retained fluid. In Chinese Pharmacopoeia, ZBM is described as a medicinal herb for somebody who lost appetite. In our daily life, the mature pericarp of ZBM is also used in cooking as spice. It is both medicine and food. Pharmacological studies have demonstrated that ZBM has anti-obesity properties ([45]Gwon et al., 2012), and some constituents occurring in ZBM, for example, quercetin ([46]Kim et al., 2015), rutin ([47]Yuan et al., 2017), and sanshool (Wang et al., 2019a) have been reported to exert anti-obesity effects. However, the anti-obesity mode and mechanism of action of ZBM are not fully understood. In this study, we aimed to employ network pharmacology to explore whether the mature pericarp of ZBM possess anti-obesity effects and delineate the underlying mechanism of action. Furthermore, both in vitro and in vivo studies have been done to validate its anti-obesity effects. A schematic diagram is shown in [48]Figure 1. FIGURE 1. [49]FIGURE 1 [50]Open in a new tab Flowchart showing the network pharmacological and experimental studies for the investigation of the anti-obesity effect of Zanthoxylum bungeanum Maxim. Materials and Methods Screening of Bioactive Ingredients from the Mature Pericarp of ZBM All of the ingredients containing in ZBM were obtained from Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP, [51]http://tcmspw.com/tcmsp.php) ([52]Ru et al., 2014), Chinese Academy of Sciences Chemistry Database ([53]www.organchem.csdb.cn), TCM Database@taiwan ([54]http://tcm.cmu.edu.tw), and Traditional Chinese Medicine Integrated Database ([55]https://omictools.com/tcmid-too) ([56]Yang et al., 2019). The effective components from ZBM are mainly filtered according to their oral bioavailability (OB) and drug-likeness (DL) indices. Absorption, distribution, metabolism, excretion, and toxicity modeling as a tool for rational drug design has significant effects in new drug discovery ([57]Wang et al., 2015). OB is one of the most important pharmacokinetic parameters in the absorption, distribution, metabolism, excretion and toxicity characteristics of drugs, indicating the ratio of the oral drug to the oral dosage of the blood circulatory system ([58]Xu et al., 2012; [59]Liu et al., 2013). High OB values is often an important consideration for the development of bioactive molecules as therapeutic agents ([60]Alam et al., 2015). In order to filter out compounds which are not likely to be drugs, the OB was calculated using in-house software OBioavail 1.1 ([61]Liu et al., 2018). This software is based on a dataset of 805 structurally diverse drug and drug-like molecules that have been critically evaluated for their OB (%F) in humans. DL evaluation is used in drug design to evaluate whether a compound is chemically suitable for use as a drug and how a drug-like molecule is with respect to parameters that affect its pharmacodynamic and pharmacokinetic profiles, which ultimately impact its ADME properties ([62]Walters and Murcko, 2002). The Tanimoto coefficient was used to evaluate the DL index of the molecules in ZBM using the following formula: [MATH: T(α,β)=α×βα2+β< /mi>2−α×β, :MATH] where a is the molecular property of the ZBM ingredient on the basis of Dragon software ([63]www.talete.mi.it/products/dragon_description.htm) and β denotes the average molecular property for all drugs in the DrugBank database ([64]www.drugbank.ca/) ([65]Mauri et al., 2006). Hence, we further selected the major ingredients based on the literature to identify the potential therapeutic effects. Although some ingredients, such as volatile oil have lower DL values, they were selected because the effect had been experimental verification ([66]Zheng et al., 2020). Meanwhile, the chemical information of these ingredients (structure, specification name, and CID number) for computational analysis was also collected according to the PubChem ([67]https://pubchem.ncbi.nlm.nih.gov/) and DrugBank ([68]https://www.drugbank.ca/drugs). Identification of ZBM-Associated Molecular Targets The potential molecular targets of ZBM were predicted using the TCMSP ([69]Ru et al., 2014), SwissTargetPrediction ([70]Gfeller et al., 2014), and the Search Tool for Interacting Chemicals ([71]Szklarczyk et al., 2016). Identification of Obesity-Associated Molecular Targets The obesity-associated targets were comprehensively collected from four databases including Therapeutic Target Database, Kyoto Encyclopedia of Genes and Genomes (KEGG), the Comparative Toxicogenomics Database, and GeneCards v4.9.0 ([72]www.genecards.org/). Ingredient–Target Network Construction The obtained drug-related targets and the disease-related targets were intersected, and a Venn diagram of the intersected gene symbols was obtained. Then, a complex information network was constructed based on the interaction of drug (ZBM), components, gene symbol, and disease (obesity). Cytoscape 3.7.1 software ([73]Shannon et al., 2003) was used for visual analyses of the drug–component–target–disease network. Protein–Protein Interaction Network Construction STRING online database ([74]https://string-db.org/) ([75]Hsia et al., 2015) was applied to obtain the PPI data of the molecular targets of ZBM, where the parameter organism was set to Homo sapiens, and other basic settings were the default value. Cytoscape software was employed to establish the PPI relationship network and perform topological analysis. Enrichment of Gene Ontology and Kyoto Encyclopedia of Genes and Genomes Pathways The GO analysis and KEGG pathway enrichment were employed by using Bioconductor (R) v3.8 bioinformatics software ([76]http://bioconductor.org/). Terms with expression analysis systematic explorer scores of ≤0.05 were collected for functional annotation clustering. The pathway enrichment analysis was performed using the KEGG database to verify the functional categories of statistically significant genes (p < 0.05). Terms with thresholds of count of ≥2 and Expression Analysis Systemic Explorer scores of ≤ 0.05 were screened for functional annotation clustering. Network Construction and Analysis The compound–target network was generated by linking bioactive constituents and putative targets. Based on the predicted targets and the predicted obesity-related signaling pathways, a target–pathway network was established. The compound–pathway network was constructed based on all the compounds and the signaling pathways. In our network, the nodes represent the candidate compounds, potential targets, or signaling pathways, while the edges represent the compound–target or target–pathway interactions. Cytoscape software was used to construct the network. Computational Validation of Ingredient–Target Interactions To further evaluate the results obtained in systemic pharmacologic analyses, quercetin, a compound in ZBM with the highest degree of connection among all the obesity-related targets, was selected for the test of its apoptotic effects in adipocytes. The three-dimensional (3D) structures of caspase 3 (PDB ID: 2XYH), Bcl-2 (PDB ID: 2W3L), and Bax (PDB ID: 5W63) were downloaded from RCSB Protein Data Bank ([77]http://www.rcsb.org/pdb). In addition, we also performed the docking assay with three inhibitors of caspase 3, Bax, and Bcl-2 as positive controls to each target to verify whether our model is robust and reliable. The 3D structure of quercetin was drawn by ChemBioDraw Ultra 14.0 and ChemBio3D Ultra 14.0 software. Structures of the compounds were sketched using MarvinSketch ([78]www.chemaxon.com), and ligand molecules were converted from 2D to 3D using ChemBioDraw Ultra 14.0 and ChemBio3D Ultra 14.0 software. The docking study was performed using AutoDock Vina ([79]Trott and Olson, 2011), and input files necessary for AutoDock program were prepared using AutoDockTools ([80]Morris et al., 2009). The size of the grid box in AutoDock Vina was kept as 40 × 40 × 40 for X, Y, and Z, and the default setting was kept for energy range. The automated program yielded nine possible conformations with distinguished binding energy for each ligand output. The final model was selected based on the binding affinity and molecular contacts. The molecular contacts were calculated using the program CONTACT available in CCP4 suite ([81]Winn et al., 2011). Docked complexes were analyzed, and figures were rendered using PyMOL ([82]www.pymol.org). Cell Culture and Reagents 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) and DMSO were purchased from Sigma Chemicals Ltd. (St. Louis, MO, USA). Antibodies against cleaved PARP (No. 5625 s), cleaved caspase 3 (No. 9661 s), and cleaved caspase 7 (No. 8438 s) were purchased from Cell Signaling Technology (Beverly, MA, USA); cleaved caspase 6 (No. ab108335), cleaved caspase 8 (No. ab108333), cleaved caspase 12 (No. ab62484), Bax (No. ab182733), and Bcl-2 (No. ab182858) were purchased from Abcam (Cambridge, MA, USA); Mcl-1 (No. sc-819) and β-actin (No. sc81178) were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Their corresponding secondary antibodies and protein marker were supplied by Bio-Rad (Hercules, CA, USA). All materials for cell culture were obtained from Life Technologies Inc (GIBCO, USA). Extraction of Mature Pericarp of ZBM ZBM, originated from Sichuan province, China, was purchased from the Chinese Medicine Clinic of the Hong Kong Baptist University and authenticated in accordance with the corresponding monograph in the Chinese Pharmacopoeia by the corresponding author. Voucher specimen of ZBM (No. 170901) was deposited at the School of Chinese Medicine, Hong Kong Baptist University. The mature pericarp of ZBM (100 g) was reflux-extracted twice with 50% ethanol (1:10, w/v) for 2 h each. The combined extracts were filtered after cooling and then concentrated under reduced pressure to remove ethanol. The powdered extracts (yield: 11.87%, we name the extract ZBM hereafter) were obtained by lyophilizing the concentrated samples with a Virtis Freeze Dryer (The Virtis Company, New York, USA). UPLC/Q-TOF-MS Analysis Liquid chromatography was performed on an Agilent 1200 system coupled with an ACQUITY UPLC T3 C18 column (2.1 mm × 50 mm I.D., 1.8 μm) maintained at 32°C. Elution was performed with a mobile phase of A (0.1% FA in water) and B (0.1% FA in ACN) under a gradient elution of 10–20% B at 0–5 min, 20–50% B at 5–10 min, 50–70% B at 10–25 min, and 70–100% B at 25–32 min was employed. The flow rate was 0.4 mL/min, and the injection volume was 5 μL. Mass spectrometric detection was carried out on an Agilent 6540 Q-TOF mass spectrometer (Hewlett Packard, Agilent, USA) with an electrospray ionization interface. The positive ion mode was used with the mass range setting at m/z 100–1,700. Optimized ionization conditions were as follows: gas temperature, 300°C; drying gas (N[2]) flow rate, 8 L/min; nebulizer, 40 psi; sheath gas temperature, 350°C; sheath gas flow, 8 L/min; capillary voltage, 4,500 V; fragmentor, 175 V; skimmer voltage, 65 V; and Octopole RF peak, 600 V. Data were collected with LC-MS-QTOF MassHunter data acquisition software ver. A.01.00 (Agilent Technologies) and analyzed with Agilent MassHunter qualitative analysis software B.06.00, respectively. The peaks were tentatively identified by matching with empirical molecular formulae and mass fragments. 3T3-L1 Preadipocyte Differentiation 3T3-L1 preadipocytes (ATCC) were induced to differentiate into mature white adipocytes with differentiation-inducing medium containing 1 mM dexamethasone, 0.5 mM isobutylmethylxanthine, and 1.67 mM insulin in Dulbecco’s modified Eagle’s medium with 10% FBS for 4 days before switching to Dulbecco’s modified Eagle’s medium with only 10% FBS and 10 μg/mL insulin for an additional 3 days ([83]Su et al., 2020). Cell Viability Assay Cell viability was determined by the MTT assay. Briefly, 3T3-L1 cells were seeded into a 96-well plate and treated with ZBM at the indicated concentrations for 24 or 48 h. Vehicle served as control. After the treatment, MTT solution (5 mg/mL) was added and incubated with the cells for another 4 h at 37°C in dark. Dimethylsulfoxide (Sigma-Aldrich) was then used to dissolve the formazan precipitate. The absorbance of each well was measured at wavelength 570 nm in a microplate reader (Bio-Rad Laboratories). Cell viability was calculated according to the absorbance of each well with the following formula: cell viability (%) = [(A570 sample−A570 blank)/(A570 control−A570 blank)] ×100%, where A570 sample, A570 blank, and A570 control stand for the absorbance of treatment group, blank group (no cells), and control group (vehicle), respectively. Western Blot Assay After treatments, 3T3-L1 cells, subcutaneous adipose tissues, or visceral adipose tissues were collected and lysed for 30 min on ice with lysis buffer. Samples were centrifuged at 15,000 rpm for 10 min at 4°C, and total protein concentrations were measured by using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific) and then denatured. Aliquots of 20–40 µg of cell or tissue lysates were separated in a 6–12% sodium dodecyl sulfate polyacrylamide gel along with PageRuler^™ Prestained Protein Ladder (Thermo Scientific) and transferred onto a polyvinylidene difluoride membrane preactivated by methanol. The membrane was blocked for 2 h at room temperature with 5% nonfat milk powder dissolved in TBST (0.1% Tween-20 in TBS), and then incubated with primary antibodies against cleaved PARP, cleaved caspase 3, 6, 7, 8, and 12, Bax; Bcl-2; Mcl-1; or GAPDH for 12 h in 4°C. The secondary antibodies were diluted in TBST containing 5% milk and incubated for 1 h at room temperature. The immune-reactive targets were detected by an ECL Western Blotting Substrate Kit (Thermo Fisher Scientific). Band density was analyzed by ImageJ software and normalized with internal control. Quantification of Apoptosis Apoptotic cells were assessed using annexin V-fluorescein isothiocyanate apoptosis detection kit I (BD, Bioscience) following the manufacturer’s instruction. Samples of 100,000 stained cells were analyzed using flow cytometry (Leica TCS SP8). Animal Handling All the animal studies were approved and performed according to the guidelines of the Department of Health HKSAR, and animal research ethics panel in the Hong Kong Baptist University. C57 male mice of 4–5 weeks old were purchased from the Chinese University of Hong Kong. Mice were randomly selected to have either control diet (D12450J Research Diets) or high-fat diet ([84]D12762 Research Diets) which was used to induce obesity. Both diet and water were supplied ad libitum. Body weight of each mouse was recorded every week. After 10 week of dietary intervention, the body weight of high-fat diet (HFD)–fed mice and comparable control diet (CD)–fed mice were significantly different, indicating the diet-induced obesity (DIO) mouse models were established. The DIO mice were then given either ZBM (4 g/kg) or vehicle control by oral administration. Behavioral changes, body weight, and food intake of these mice were recorded every day. Results Identification of Bioactive Components From ZBM According to the UPLC/Q-TOF-MS analyses, 13 bioactive compounds in ZBM were tentatively identified by matching with the empirical molecular formulae and mass fragments ([85]Supplementary Figure S1). Details were shown in [86]Supplementary Table S1. In addition, according to the high-performance liquid chromatography analysis, two characteristic constituents in ZBM, such as hydroxy-α-sashool and hydroxy-β-sanshool, were also identified with reference standards ([87]Supplementary Figure S2). After combining with the identified compounds and the compounds collected in three databases, a total of eighty-four candidate compounds were identified ([88]Supplementary Table S2). To identify the active ingredients of ZBM, two classical ADME parameters, OB and DL, were used for screening. OB ≥15% and DL ≥0.1 were considered to have relatively better pharmacological properties. Although some ingredients do not meet the screening criteria, they have clinical therapeutic effects; we also kept these ingredients in our research for a comprehensive analysis. For example, rutin had a low OB, while it is a major and active constituent of ZBM ([89]Zhang et al., 2017). Increasing studies have demonstrated that rutin reduces obesity by activating brown fat ([90]Yuan et al., 2017). Besides, in HFD-induced obese rats, rutin increases muscle mitochondrial biogenesis coupled with AMP-activated protein kinase activation and reduces body weight by increasing brown adipose tissue mitochondrial biogenesis ([91]Seo et al., 2015). Another compound, hyperoside, has a low DL, but it has a beneficial effect on the controlling body weight because it inhibits adipogenesis ([92]Berkoz, 2019). It is important to note that although the pharmacokinetic parameters of these components are relatively low, they are bioactive and therefore are considered as the candidate ingredients. Hence, we expanded the screening standard procedure beyond the ADME principles; we assumed that if the candidate components in ZBM intersected with the obesity targets, they were considered as the active components. Therefore, as shown in [93]Table 1, a total of 20 ingredients were selected as active ingredients in ZBM. TABLE 1. A total of twenty active ingredients were obtained from ZBM. No CAS Name References