Network Pharmacology Study to Interpret Signaling Pathways of Ilex cornuta Leaves against Obesity

: Ilex cornuta Leaves (ICLs) are a representative and traditional prescription for controlling obesity. Nevertheless, the corresponding therapeutic compounds and related pharmacological mechanisms of such medication remain undocumented. The compounds from ICLs were identiﬁed by gas chromatography-mass spectrum (GC-MS), and SwissADME conﬁrmed their physicochemical properties. Next, the target proteins related to compounds or obesity-associated proteins were re-trieved from public databases. RPackage constructed the protein–protein interaction (PPI) network, a bubble chart, and signaling pathways–target proteins–compounds (STC) network. Lastly, a molecular docking test (MDT) was performed to evaluate the afﬁnity between target proteins and ligands from ICLs. GC-MS detected a total of 51 compounds from ICLs. The public databases identiﬁed 219 target proteins associated with selective compounds, 3028 obesity-related target proteins, and 118 overlapping target proteins. Moreover, the STC network revealed 42 target proteins, 22 signaling pathways, and 39 compounds, which were viewed to be remedially signiﬁcant. The NOD-like receptor (NLR) signaling pathway was considered a key signaling pathway from the bubble chart. In parallel, the MDT identiﬁed three target proteins (IL6, MAPK1, and CASP1) on the NLR signaling pathway and four compounds against obesity. Overall, four compounds from ICLs might show anti-obesity synergistic efﬁcacy by inactivating the NLR signaling pathway.


Introduction
Obesity has now sharply hit epidemic levels and has become a significant cause of global death [1]. A recent report indicates that obesity is closely linked to metabolic disorders that distinctly develop psychological stress and often exacerbate obesity-related complications [2]. Obesity can present in all ages; in 2016, almost 13% of the world population were overweight [3]. Moreover, obesity is deeply associated with other metabolic diseases such as diabetes, hypertension, atherosclerosis, and heart failure [4]. The main driving factors of metabolic disorders are cytokines, which are mainly implicated with a high-fat diet [5]. Currently, available anti-obesity medications include sibutramine, rimonabant, and orlistat, which may lead to side effects like diarrhea, fecal incontinence, flatulence, and dyspepsia [6,7]. In contrast, herbal solutions based on natural organic compounds have been used for thousands of years with high efficacy and safety [8]. To date, many natural herbal plants are used in a diet regime or as an alleviator for anti-obesity [9], for example, Ilex cornuta Leaves (ICLs) are potentially used to treat obesity. A topical patent on ICLs summarized that the extract could be effective for various metabolic diseases [10]. Moreover, some reports demonstrated that ICLs extract has potent anti-inflammatory effects associated with obesity in adipocytes [11,12]. Another study stated that the Ilex species, including ICLs, are known for regulating lipid metabolism and weight-loss activity [13]. Until now, research of ICLs has been focused on a broad range of metabolic disorders without defining the exact action mechanism for particular diseases. Therefore,

Chemical Compounds from ICLs
A total of 52 chemical compounds in ICLs were identified by the GC-MS analysis (Figure 2), and the name of compounds, PubChem ID, retention time (min), peak area (%), and pharmacological activities were enlisted in Table 1. Lipinski's rules accepted the number of 51 out of 52 chemical compounds (molecular weight ≤ 500g/mol; Moriguchi octanol-water partition coefficient ≤4.15; number of nitrogen or oxygen ≤10; number of NH or OH ≤5), and the selected 51 chemical compounds (excluding lactose) corresponded with the standard of 'Abbott Bioavailability Score (>0.1)' through SwissADME. Additionally, lactose was excluded due to the number of nitrogen or oxygen and the number of NH or OH. The TPSA (topological polar surface area) value of the selected 51 chemical compounds (excluding lactose) was also accepted ( Table 2).

Overlapping Target Proteins between SEA and STP
A total of 525 target proteins from SEA and 576 target proteins from STP connected to 51 chemical compounds were identified (Supplementary Table S1). The Venn diagram showed that 219 target proteins were overlapped between the two compound databases (Supplementary Table S1) ( Figure 3A).

Overlapping Target Proteins between Obesity-Related Target Proteins and 219 Target Proteins
A total of 3028 target proteins associated with obesity were selected by retrieval from DisGeNET and OMIM databases (Supplementary Table S2). The Venn diagram result revealed 118 overlapping target proteins between obesity-associated 3028 target proteins and the 219 overlapping target proteins ( Figure 3B) (Supplementary Table S3).

Analysis of Signaling Pathways against Obesity
The results of the KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway enrichment analysis revealed that 22 signaling pathways were related to 42 target proteins (false discovery rate < 0.05). The 22 signaling pathways were directly connected to obesity, suggesting that these 22 signaling pathways might be the significant pathways of ICLs against obesity. The description of 22 signaling pathways is provided in Table 4. A bubble chart showed that both NOD-like receptor signaling pathway and MAPK signaling pathway have the same rich factor of 0.024 ( Figure 5). Additionally, NOD-like receptor signaling pathway was directly related to IL6 (the highest degree of value) but MAPK signaling pathway was unconnected to IL6 (listed in Table 4).

STC Networks Analysis of ICLs against Obesity
STC networks of ICLs against obesity are exhibited in Figure 6. There were 22 pathways, 42 target proteins, and 39 compounds (103 nodes, 333 edges). The nodes stand for a total number of three components: signaling pathways-target proteins-compounds (STC). The edges stand for the association of a total number of three components. The STC networks suggested that the network was associated with the therapeutic efficacy against obesity. Particularly, MAPK1 target protein (the highest degree value) was related to 20 out of 22 signaling pathways in STC networks, connected to NOD-like receptor signaling pathway. The main three target proteins of ICLs against obesity are indicated in the KEGG pathway diagram (Figure 7).
On MDT, the most potent compounds on IL6 were ethyl-α-d-glucopyranoside and 3,5dihydroxy-6-(hydroxymethyl)oxan-2-one having an aliphatic heteromonocyclic structure. It was reported that a hetero-aliphatic ring is associated with a better appeal to develop a drug due to its high solubility, low lipophilicity (logP < 5), low albumin-binding and cytochrome P450 inhibition [34,35]. Similarly, Orlistat is a representative anti-obesity drug with an aliphatic heteromonocyclic structure [36] which implies that compounds with an aliphatic heteromonocyclic structure might be potential candidates for anti-obesity drug development.
On the MAPK1 target protein, the most potent compound was 4-dehydroxy-N-(4,5methylenedioxy-2-nitrobenzylidene)tyramine with an aromatic heteropolycyclic structure. Likewise, Cetilistat is a typical drug for anti-obesity with an aromatic heteropolycyclic structure [36]. On the CASP1 target protein, the most potent compound was pentanoic acid, 3-[(adamantan-1-ylmethyl) carbamoyl] -4-phenyl-) with an aromatic homopolycyclic structure. Moreover, Oleoyl-estrone is a standard drug to reduce the body fat, having an aromatic homopolycyclic structure [36]. Based on these similarities, our study suggests that the four compounds in ICLs have a high chance of offering synergistic effects against obesity.
A bubble chart displayed that ICLs compounds on obesity were involved in 42 target proteins. Furthermore, the outputs of the KEGG pathway enrichment analysis of 42 target proteins indicated that 22 signaling pathways were connected to the progression of obesity, suggesting that these signaling pathways might be the molecular mechanisms of ICLs against obesity. The associations of the 22 signaling pathways with obesity were discussed as follows.
PPAR (peroxisome proliferator-activated receptor) signaling pathway: PPARs are ligand-regulated receptors, and many standard anti-obesity drugs are related to this signaling pathway [37]. VEGF (vascular endothelial growth factor) signaling pathway: VEGF-A (vascular endothelial growth factor A) has anti-inflammatory effects against diet-induced obesity [38]. HIF-1 (hypoxia-inducible factor 1) signaling pathway: inactivation of HIF-1 in adipose tissue alleviates obesity, suggesting that HIF-1 is a new target to develop antiobesity agents [39]. Fc epsilon RI signaling pathway: an animal experiment demonstrated that mice with obesity increased the expression level of Fc epsilon RI more than eight times as compared to lean mice [40]. It implies that the inactivation of Fc epsilon RI can inhibit obesity. Prolactin signaling pathway: a report shows that prolactin accelerates fat accumulation in diverse animal models; particularly, an increased level of PRL was recorded for obese women in accordance with the visceral fat amount [41]. Interleukin 17 (IL17) signaling pathway: IL17 expression level is enhanced in obese individuals, a mediator to induce pro-inflammatory reactions [42,43]. AGE-RAGE signaling pathway in diabetic complications: AGE-RAGE is deeply interconnected to obesity-involved renal damage; both AGE and RAGE induce a pro-inflammatory reaction and are associated with obesity [44,45]. Tumor necrosis factor (TNF) signaling pathway: TNF inhibits lipoprotein lipase to break triglyceride, known as a primary factor of obesity [46]. Calcium signaling pathway: the activation of calcium signaling promotes energy consumption, which facili-tates metabolism and differentiation of adipocytes, thus preventing obesity [47]. Thyroid signaling pathway: obesity is evidently related to Hashimoto's thyroiditis, suggesting that prevention of obesity is significant for recovering thyroid function [48]. B cell receptor signaling pathway: B cells aggravate obesity-related metabolic disorders and secrete cytokines to stimulate inflammation [49]. T cell receptor signaling pathway: T cell damage is accelerated by obesity; accordingly, T cell dysfunction is detrimental to maintain the immune system [50,51]. Relaxin signaling pathway: the activation of Relaxin-2 attenuates obesity in high-fat-diet mice; furthermore, Relaxin-3 associated with hypercholesterolemia is a potential target protein against obesity [52,53]. Insulin signaling pathway: insulin resistance is a crucial factor in aggravating obesity; especially, adipose tissues in obese individuals produce pro-inflammatory agents that stimulate progressive insulin resistance [54]. Rap-1 signaling pathway: an animal experiment demonstrated that the lack of Rap-1 induces weight gain due to abdominal fat accumulation [55]. Gonadotropin-releasing hormone (GnRH) signaling pathway: GnRH agonist treatment induces fat accumulation; mainly, inhibition of GnRH is a preventive method to treat obesity [56]. Renin-angiotensin system (RAS) signaling pathway: obesity is linked to RAS activation; in contrast, blockers of RAS diminished the type 2 diabetes by 22% in severe populations [57]. Oxytocin signaling pathway: the insufficiency of oxytocin and/or its receptor expression leads to obesity, which is implicated in metabolic disorders [58]. Phosphoinositide 3-Kinase-Protein kinase B (PI3K-AKT) signaling pathway: the dysfunction of the PI3K-AKT signaling pathway causes obesity, in other words, inhibition of the PI3K-AKT signaling pathway exacerbates metabolic processes [59]. Janus kinase/signal transducer and activator of transcription proteins (JAK-STAT) signaling pathway: the JAK-STAT signaling pathway is involved in preventing metabolic diseases including obesity, suggesting that the JAK-STAT pathway is a potential therapeutic mechanism for the treatment of obesity [60]. MAPK signaling pathway: a study indicated that obese mice have shown activated MAPK (known as ERK) expression, while the blocking of MAPK diminishes lipolysis in both mice and human adipose tissue [61]. NOD-like receptor (NLR) signaling pathway: the NLR pathway is overexpressed in the adipocytes from the obesity which increases inflammasome activity [62]. On NOD-like signaling pathway in KEGG pathway enrichment, each CASP1, MAPK1, and IL6 target protein is associated with proinflammatory reactions. In detail, CASP1, MAPK1, and IL6 are deeply involved in metabolic diseases, and their activation leads to obesity-related diseases [63].
According to the degree value of each target protein in the PPI network, IL6 was regarded as a key target of ICLs against obesity, which was directly connected to 52 out of 118 target proteins. In addition, based on the degree value of each target protein in the STC network, MAPK1 was considered as an uppermost target of ICLs against obesity, which was enriched in 20 out of 22 signaling pathways. Specifically, both the MAPK signaling pathway and the NLR signaling pathway had the same rich factor of 0.024. Between the two signaling pathways, a signaling pathway associated with IL6 and MAPK1 was the NLR signaling pathway. Thus, the NLR signaling pathway might be the key signaling pathway of ICLs against obesity. The four target proteins associated with the NLR signaling pathway were IL6, MAPK1, P2RX7, and MAPK1. The four target proteins were used to perform MDT with ligands connected to each target protein; also, MDT was conducted for positive controls to compare each affinity with ligands from ICLs. From the MDT, P2RX7 was excluded due to invalid binding energy (< −6.0 kcal/mol). Each ligand from ICLs bound to three other target proteins (IL6, MAPK1, and CASP1) and exposed higher affinity than the positive controls. Thus, these results suggest that inhibition of the three targets on the NLR signaling pathway might develop a synergistic effect to alleviate the obesity. Our research shows that four compounds, including ethylα-d-glucopyranoside (1); 3,5-dihydroxy-6-(hydroxymethyl)oxan-2-one (2); 4-dehydroxy-N-(4,5-methylenedioxy-2-nitrobenzylidenetyramine (3); and pentanoic acid, 3-[(adamantan-1-ylmethyl) carbamoyl] -4-phenyl-(4) from ICLs were noted as promising ligands on the three targets (IL6, MAPK1,and CASP1) (Figure 9).

Plant Material Collection and Classification
The Ilex cornuta leaves (ICLs) were collected from Mihogil of Bomunmyeon (Latitude:  35

Plant Preparation and Extraction
The ICLs were dried in a shady area at room temperature (20-22 • C) for 7 days, and dried leaves were powdered using an electric blender. Approximately 30 g of C. maackii flower powder was soaked in 500 mL of 100% methanol (Daejung, Korea) for 7 days and repeated 3 times for the highest extraction. The solvent extract was collected, filtered, and evaporated using a vacuum evaporator (IKA-RV8, Japan). The evaporated sample was dried under a boiling water bath (IKA-HB10, Japan) at 40 • C to obtain the yield.

GC-MS Analysis Condition
Agilent 7890A was used to carry out the GC-MS analysis. The GC was equipped with a DB-5 (30 m × 0.25 mm × 0.25 µm) capillary column. Initially, the instrument was maintained at a temperature of 100 • C for 2.1 min. The temperature was increased to 300 • C at the rate of 25 • C/min and maintained for 20 min. The injection port temperature and helium flow rate were confirmed as 250 • C and 1.5 mL/min, respectively. The ionization voltage was 70 eV. The samples were injected in split mode at 10:1. The MS scan range was set at 35-900 (m/z). The fragmentation patterns of mass spectra were compared with those stored in the W8N05ST Library MS database. The percentage of each compound was calculated from the relative peak area of each compound in the chromatogram. The concept of integration used was the ChemStation integrated algorithms.

Chemical Compounds Database Construction and Drug-Likeness Identification
The chemical compounds from the ICLs leaves were identified through the GC-MS analysis. Then, the GC-MS detected chemical compounds that were filtered by Lipinski's rule and TPSA value in SwissADME (http://www.swissadme.ch/) (accessed on 23 April 2021) [64] to confirm the 'drug-likeness' physicochemical properties. The PubChem repository (https://pubchem.ncbi.nlm.nih.gov/) (accessed on 23 April 2021) was utilized to select the SMILES (simplified molecular input line entry system) format.

PPI Networks and Signaling Pathways on a Bubble Chart
On the final overlapping target proteins, STRING (https://string-db.org/) (accessed on 11 May 2021) [67] was utilized to analyze the PPI networks. Thereby, RPackage was used to identify the degree of value, which is defined as the numbers of connectivity to a target protein (node). Then, signaling pathways directly related to obesity were visualized on a bubble chart via RPackage. Thus, the signaling pathways provide important clues for the therapeutic effect of ICLs against obesity.

A Signaling Pathways-Target Proteins-Chemical Compounds Network
The signaling pathways-target proteins-chemical compounds (STC) network was utilized to construct a size map, based on the degree of values. In the network, green circles (nodes) represented signaling pathways; pink rectangles (nodes) represented target proteins, and orange triangles (nodes) represented chemical compounds. The size of the pink rectangles stood for the number of connectivity with signaling pathways; the size of the orange triangles stood for the number of connectivity with target proteins. The merged networks were constructed using RPackage.

Preparation of Target Proteins for MDT
The target proteins of two key signaling pathways (MAPK signaling pathway, NODlike receptor signaling pathway), i.e., FGF1 (PDB ID:

Preparation of Ligands for MDT
The ligand molecules were converted to .sdf from PubChem into .pdb format using Pymol, and the ligand molecules were converted into .pdbqt format through Autodock.

Preparation of Ligand Molecules for MDT
The ligand molecules were converted to .sdf from PubChem into .pdb format using Pymol, and the ligand molecules were converted into .pdbqt format through Autodock.

Ligand-Protein Docking
The ligand molecules were docked with target proteins utilizing autodock4 by settingup 4 energy range and 8 exhaustiveness as default to obtain 10 different poses of ligand molecules [68]. The active site's grid box size was x = 40 Å, y = 40Å, and z = 40Å. The 2D binding interactions were used with LigPlot+ v.2.2 (https://www.ebi.ac.uk/thornton-srv/ software/LigPlus/) (accessed on 14 May 2021) [69]. After docking, ligands of the lowest binding energy (highest affinity) were selected to visualize the ligand-protein interaction in Pymol.

Toxicological Properties Prediction by admetSAR
Toxicological properties of key ligands from ICLs were demonstrated utilizing the admetSAR web-service tool (http://lmmd.ecust.edu.cn/admetsar1/predict/) (accessed on 14 May 2021) [70] because toxicity is a critical factor in developing new drugs. Hence, Ames toxicity, carcinogenic properties, acute oral toxicity, and rat acute toxicity were predicted by admetSAR.