Dehydrovomifoliol Alleviates Nonalcoholic Fatty Liver Disease via the E2F1/AKT/mTOR Axis: Pharmacophore Modeling and Molecular Docking Study

Objective Herbal medicine discovery is a complex and time-consuming process, while pharmacophore modeling and molecular docking methods enable simple and economic studies. The pharmacophore model provides an abstract description of essential intermolecular interactions between chemical structures, and the molecular docking technology can identify novel compounds of therapeutic interests and predict the ligand-target interaction at the molecular level. This study was based on the two methods to elucidate the mechanism of dehydrovomifoliol, an active ingredient extracted from Artemisia frigida willd, in nonalcoholic fatty liver disease (NAFLD). Methods Bioinformatics analysis was performed to screen target genes of dehydrovomifoliol in NAFLD treatment, which were thus intersected with NAFLD-related differentially expressed genes (DEGs) and NAFLD-related genes. Venn diagram was used to identify candidate DEGs. A pharmacophore model was then generated, and molecular docking was performed. A protein-protein interaction (PPI) network was constructed to identify core genes, which were evaluated using GO and the KEGG enrichment analyses. Results Seven target genes of dehydrovomifoliol in NAFLD treatment were screened out, namely E2F1, MERTK, SOX17, MMP9, SULT2A1, VEGFA, and BLVRA. The pharmacophore model and molecular docking of candidate DEGs and dehydrovomifoliol were successfully constructed. E2F1 was identified as a core gene of dehydrovomifoliol in NAFLD treatment. Further enrichment analysis indicated the regulatory role of E2F1 in fat metabolism was associated with the regulation of the AKT/mTOR signaling pathway. Conclusion Overall, this study illustrates the anti-NAFLD mechanism of dehydrovomifoliol, which could be a useful compound for developing novel drugs in the treatment of NAFLD.


Introduction
Nonalcoholic fatty liver disease (NAFLD) is a metabolic disease closely associated with obesity and insulin resistance, leading to dysfunction of adipose tissues and liver [1]. NAFLD distresses one quarter of the worldwide population [2]. Metabolic abnormalities have been the hallmark of the development of NAFLD [3]. Hence, identifcation of metabolic drivers might reveal therapeutic targets to improve the treatment for NAFLD. Te pharmacophore model provides an abstract description of essential intermolecular interactions between chemical structures, which is usually derived both in a receptor-based and in a ligand-based manner [4]. An array of ligand-and structure-based approaches have been developed for improved pharmacophore model building, which have successful and extensive application in de novo design, virtual screening, and lead optimization [5]. Consistently, the molecular docking technology is an in silicostructure-based method that has been widely used in modern drug design, which can identify novel compounds of therapeutic interests and predict the ligand-target interaction at the molecular level [6]. It is interesting to note a widely applied in silico method molecular docking in drug discovery programs, which predicts the binding mode of a given molecule interacting with a specifc biological target [7]. Previous studies provided evidence for the feasibility of applying molecular docking approach in studying oxidative stress [8] and key signaling pathways involved in the mechanism of Qushi Huayu Decoction [9] in NAFLD, but these studies fail to single out core genes or illuminate the downstream mechanistic actions. Tus, further investigations are warranted for network pharmacology and molecular docking technology in NAFLD.
Dehydrovomifoliol is an active ingredient extracted from Artemisia frigida willd, a traditional Chinese medicine, and can be used as an early lead compound for developing novel drugs for the prevention of NAFLD due to its efects on hepatic lipogenesis and fatty acid oxidation [10]. Multiple kinds of traditional Chinese medicine have demonstrated the ability to regulate E2F transcription factor 1 (E2F1); for instance, Jieduan-Niwan formula can protect the liver of acuteon-chronic liver failure rats by inhibiting E2F transcription factor 1 (E2F1)-mediated apoptotic signaling pathway [11]. However, the possible regulation of dehydrovomifoliol on E2F1 has not been characterized and thus becomes the focus of this study. E2F1, a transcription factor, contributes to the development of liver pathology due to its regulatory role in lipid synthesis and glycolysis while its deletion has been shown to abrogate hepatic steatosis in a murine model of NAFLD [12]. Of interest, E2F1 has been reported to function as a critical antiapoptotic factor in human and rodent liver cancer via activation of the protein kinase B (AKT)/mammalian target of rapamycin (mTOR) signaling pathway [13]. In addition, inhibiting the AKT/mTOR signaling pathway reduces lipid accumulation and lipogenesis, while increasing fatty acids oxidation and very low density lipoprotein assembly and secretion in goose hepatocytes [14]. Terefore, the efect of dehydrovomifoliol may be associated with the E2F1/ AKT/mTOR axis. In lieu of the abovementioned fndings and evidence, the present study mainly centered on elaborating the mechanisms underlying the role of dehydrovomifoliol in NAFLD by combining the pharmacophore modeling and molecular docking approach.

Retrieval of Target Genes of Dehydrovomifoliol.
Te 3D chemical structure map of dehydrovomifoliol was downloaded from the PubChem database and saved in "SDF" format. Te SDF fle of dehydrovomifoliol was uploaded to the PharmMapper server, after which dehydrovomifoliol was simulated for molecular-target protein docking using the inverse docking tool PharmMapper server, with "Human Protein Targets Only" set as the target range and the remaining parameters were kept at their default values.

Acquisition of High-Troughput Transcriptome
Sequencing Dataset. NAFLD-related gene expression datasets GSE73500 and GSE89632 were downloaded from the GEO database. Te GSE73500 dataset includes three liver tissue samples from NAFLD rats and three control liver tissue samples. Te GSE89632 dataset includes nineteen liver tissue samples from NAFLD rats and 24 control liver tissue samples.

Quality Control and Reference Genome Alignment of
Datasets. Te quality of paired-end reads of raw sequencing data was assessed using the FastQC v0.11.8 software. Preprocessing of raw data was performed by using the Cutadapt v1.18 software: the illumina sequencing adaptors and poly (A) tail sequences were removed. Te reads with an N content of over 5% were removed using a Perl script. Te reads with a 70% base mass above 20 were extracted using FASTX Toolkit software (v0.0.13). Next, two-end sequences were repaired using BBMap software. Finally, the fltered high-quality reads fragment human reference genome was aligned using HISAT2 software (v0.7.12).

Diferential Expression Analysis.
With normal samples as the control, the R "limma" package (version 3.4.1) was applied for diferential analysis to identify the diferentially expressed mRNAs with |log2FC| > 1 and p value < 0.05 as the threshold.

NAFLD-Related Database
Retrieval. NAFLD-related target genes were retrieved from the GeneCards database with "nonalcoholic fatty liver disease" and "fat cell proliferation" as the key words. With Relevance Score ≥ 1 as the screening criteria, more relevant targets were obtained for subsequent analysis.
2.6. Venn Analysis. Te Draw Venn Diagram tool was used to perform Venn analysis on the dataset analysis results and the retrieval results of the GeneCards database to identify candidate genes.

Construction of Candidate Gene-Protein Interaction
Networks. Te Cytoscape (v3.8.2) plug-in Bisogenet and cytoNCA were used for intersection analysis. Bisogenet provides database information, while cytoNCA provides attribute information of nodes after network composition, and is utilized to calculate the degree centrality (DC) of nodes and observe the core genes intuitively. A proteinprotein interaction (PPI) network was constructed and topological analysis was performed. Core targets were screened according to the DC and the betweenness centrality (BC).
Trough the STRING database, the interaction network of E2F1-mediated genes related to fat metabolism, was obtained with the "Homo sapiens" as the species, which was then imported into Cytoscape software (v3.8.2) for analysis and ranking. Te Degree value and Combine score value were represented by color, and the candidate genes were ranked according to the Degree value. , and BLVRA (PDB ID : 2H63) were downloaded and screened from the Protein Data Bank. Te structure of dehydrovomifoliol was downloaded from the PubChem database, converted to a 3D structure using Chem3DUltra 14.0 software, and subjected to energy minimization using the MM2 algorithm. Te receptors of the seven target proteins were then subjected to dehydration and organic substance removal using PyMOL software. Target protein receptor molecules were hydrogenated, followed by charge calculation using AutoDockTools 1.5.6 software. Next, compounds and target protein receptors were converted into "PDBQT" fles with appropriate box center and box lattice parameters set. Finally, Vina 1.5.6 software was run to evaluate the molecular docking and calculate the docking energy value.

Functional Enrichment Analysis of Candidate Genes.
Te R "ClusterProfler" package was used for GO enrichment analysis, which consists of biological process (BP), molecular function (MF), and cellular component (CC), with p < 0.05 as the threshold. KEGG enrichment analysis was performed to analyze cellular functions and signaling pathways mainly afected by the potential targets and key targets with p < 0.05 as the threshold.

Statistical Analysis.
All statistical analyses in this study were performed using SPSS 21.0 software (IBM Corp. Armonk, NY). Measurement data were presented as mean ± standard deviation. Data obeying normal distribution and homogeneity of variance between two groups were compared using unpaired t test. A p value < 0.05 was considered signifcant.

Bioinformatics Analysis-Based Identifcation of Seven Target Genes of Dehydrovomifoliol in NAFLD Treatment.
Dehydrovomifoliol is an active ingredient extracted from the traditional Chinese medicine Artemisia frigida willd. Tis study explored the mechanism of dehydrovomifoliol in NAFLD based on pharmacophore model building and molecular docking methods. Te experimental schedule is shown in Figure S1. First, the chemical structure of dehydrovomifoliol was determined using the PubChem database (Figures 1(a) and 1(b)). Te PharmMapper server was then used to analyze and screen the target genes of dehydrovomifoliol, with 73 related genes obtained. Diferential analysis of the GSE73500 dataset yielded 635 diferentially expressed genes (DEGs), including 305 upregulated DEGs and 330 downregulated DEGs (Figures 2(a) and 2(b)). In addition, 1193 DEGs were obtained from the GSE89632 dataset, including 579 upregulated DEGs and 614 downregulated DEGs (Figures 2(c) and 2(d)). From the GeneCards database, 1505 NAFLD-related genes were identifed with Relevance Score ≥ 1 as the screening criteria.

Construction of Pharmacophore Models for Candidate
DEGs. When drug molecules interact with receptor targets, it is necessary to generate geometric and energy matched active conformations with targets. Pharmacophore can represent the characteristic elements of pharmacodynamics and their spatial arrangement in the active molecule that play an important role in the activity and can also refect the active site where the ligand and the receptor interact. Pharmacophore structure analysis of dehydrovomifoliol and the seven receptor targets involved in NAFLD treatment was performed using the PharmMapper database. Te results showed that four hydrophobic and one acceptor interact with E2F1; two hydrophobic, one donor, and one acceptor interact with MERTK; six hydrophobic, one donor, and one acceptor interact with SOX17; two hydrophobic, one positive region, two donor, and one acceptor interact with MMP9; three hydrophobic and one acceptor interact with SULT2A1; three hydrophobic, one negative region, and three acceptor interact with VEGFA; four hydrophobic and one acceptor interact with BLVRA (Figures 3(a)-3(h)). Accordingly, pharmacophore models for the interaction of candidate genes with dehydrovomifoliol were constructed for subsequent analysis.

Validation of the Molecular Docking between Candidate
Proteins and Dehydrovomifoliol. Molecular docking analysis of candidate proteins and dehydrovomifoliol was performed using AutoDockTools 1.5.6 and Vina 1.5.6 software. When the binding energy is < 0 kJ/mol, the molecular proteins are considered to bind spontaneously and interact with each other; the lower binding energy refects more stable molecular conformation. Te results of docking analysis of candidate proteins with dehydrovomifoliol are shown in Figures 4(a)-4(h). Te docking results of the seven candidate proteins and dehydrovomifoliol are displayed in 3D plots, which can visually demonstrate the binding mode of target protein receptors and compounds, as well as the interaction with the surrounding amino acid residues. Te results revealed that E2F1 had the highest binding free energy to dehydrovomifoliol compared to the other protein molecules, indicating that dehydrovomifoliol may participate in related mechanistic responses by regulating E2F1 protein in NAFLD treatment.    C10orf140  FMO1  TYMS  GINS2  WNT5A  JUB  PNRC1  BCL6  ELL  GADD45G  KLF11  CEBPD  GADD45B  MIR21  EPHA2  PPRC1  MAP3K8  FOSL2  FAM107A  ADRB2  IL4R  SERTAD1  PELI1  PIM1  BCL3  FILIP1L  FOSB  AXUD1  JUNB  PPP1R15A  CCL2  CYR61  ( Figure 6(b)); and MF of histone kinase activity, protein kinase C activity, tau protein binding and protein serine/ threonine phosphatase activity (Figure 6(c)). Additionally, the results of KEGG pathway enrichment analysis showed that the seven candidate DEGs were chiefy involved in the endocrine resistance, relaxin signaling pathway, fuid shear stress and atherosclerosis, and hepatitis B (Figure 6(d)). During the development of NAFLD, insulin resistance leads to increased liver lipogenesis and inhibits fat decomposition, resulting in fat storage in adipocytes in the form of triglycerides in the liver and the consequent formation of fatty liver [15]. Among the seven candidate DEGs, E2F1 can participate in the regulation of fat cell proliferation and fat storage, and plays an important regulatory role in the development of NAFLD. Moreover, diferential analysis of the GSE73500 dataset suggested that E2F1 was signifcantly highly expressed in rats with NAFLD ( Figure 6(e)).

E2F1 Regulates Fat Metabolism by Mediating the AKT/ mTOR Signaling Pathway.
Te accumulation of fat in liver tissues leads to an increase of free fatty acids and reactive oxygen species, which can cause oxidative stress and endoplasmic reticulum stress, further resulting in damage to hepatocytes and aggravating NAFLD [16]. To explore the regulation of fat metabolism by E2F1, related genes were retrieved from the GeneCards website with "fat cell proliferation" as the key word. Te top 50 fat cell proliferation-related genes and E2F1 were selected to construct a PPI network, and 693 edges were obtained, with the PPI enrichment p value < 1.0e − 16 (Figure 7(a)). Ten, 19 fat metabolism genes related to the core gene E2F1 were screened (Figure 7(b)). Further KEGG enrichment analysis suggested that the 19 genes were involved in the regulation of FoxO signaling pathway, EGFR tyrosine kinase inhibitor resistance, and ErbB signaling pathway (Figure 7(c)). Among them, the AKT/mTOR signaling pathway participates in various mechanisms related to fat metabolism, such as regulation of fatty acid β-oxidation, fatty acid catabolic processes, fat cell diferentiation, and fatty acid oxidation (Table S1). Evidence has shown that activation of the AKT/ mTOR signaling pathway is closely related to oxidative stress and fat metabolism in NAFLD [17,18]. Terefore, we speculated that E2F1 can regulate fat metabolism and participate in NAFLD development by mediating the AKT/ mTOR signaling pathway.

Discussion
Traditional Chinese medicine have gained widespread attention in recent years, but knowledge of the mechanisms by which they act is still very limited. In this work, we generated a qualitative pharmacophore model and applied the molecular docking technology to elucidate the mechanism underlying dehydrovomifoliol alleviating NAFLD. Te results demonstrated that dehydrovomifoliol may alter abnormalities in fat metabolism and consequently alleviate NAFLD by downregulating the E2F1/AKT/mTOR axis.
It is interesting to note that molecular docking, as a widely applied in silico method in drug discovery programs, can predict the binding mode of a given molecule interacting with a specifc biological target [7]. Subsequent pharmacophore model building and molecular docking method in this study revealed that E2F1 was a core gene of dehydrovomifoliol in NAFLD treatment. E2F1 is known as a key regulator of cell cycle, DNA-damage response, and apoptosis, and it plays a crucial role in the control of metabolism [25]. Additionally, E2F1 deletion has been shown to abrogate hepatic steatosis in diferent mouse models of NAFLD [12]. Te current results also indicated the involvement of E2F1 in the fat metabolism during NAFLD. In line with this, a recent study has identifed the increased expression of E2F1 in human NAFLD, and that increased E2F1 can function as a metabolic driver of NAFLD-related hepatocellular carcinoma [26]. Nonetheless, further indepth investigation is required owing to the insufcient evidence to support the correlation of dehydrovomifoliol with E2F1.
Further mechanistic investigations suggested that E2F1 may regulate fat metabolism and then participate in the development of NAFLD by mediating the AKT/mTOR signaling pathway. Additionally, the E2F1/DDX11 axis is capable of promoting malignant behaviors of hepatocellular carcinoma cells via activation of the PI3K/AKT/ mTOR signaling pathway [27]. Accumulating evidence suggests a close association of AKT/mTOR signaling pathway activation with oxidative stress and metabolism in NAFLD [17,18]. Recent data highlighted that salvianolic acid A, a traditional Chinese medicine extracted from Danshen, may curtail the process of liver fbrosis by regulating the PI3K/AKT/mTOR signaling pathway [28]. Tus, studying the E2F1/AKT/mTOR axis maybe benefcial to the mechanistic understanding of traditional Chinese medicine and the development of new therapeutic strategies for NAFLD.

Conclusion
In summary, the key fndings from the present study suggested that dehydrovomifoliol potentially alleviates NAFLD through disruption of the E2F1/AKT/mTOR axis ( Figure 8).
Tese results are very helpful to provide clues and useful exploration for drug research based on natural products, especially dehydrovomifoliol. Further studies based on clinical cohorts and in vitro and in vivo experiments are required to substantiate the mechanistic basis and therapeutic efects of dehydrovomifoliol in NAFLD.

Data Availability
Te datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethical Approval
Tis study did not involve human or animal material or data. Ethical approval was not required.