Investigation of the potential mechanism of the Shugan Xiaozhi decoction for the treatment of nonalcoholic fatty liver disease based on network pharmacology, molecular docking and molecular dynamics simulation

Abstract

Background

Nonalcoholic fatty liver disease (NAFLD) is a metabolic disease, the incidence of which increases annually. Shugan Xiaozhi (SGXZ) decoction, a composite traditional Chinese medicinal prescription, has been demonstrated to exert a therapeutic effect on NAFLD. In this study, the potential bioactive ingredients and mechanism of SGXZ decoction against NAFLD were explored via network pharmacology, molecular docking, and molecular dynamics simulation.

Methods

Compounds in SGXZ decoction were identified and collected from the literature, and the corresponding targets were predicted through the Similarity Ensemble Approach database. Potential targets related to NAFLD were searched on DisGeNET and GeneCards databases. The compound–target–disease and protein-protein interaction (PPI) networks were constructed to recognize key compounds and targets. Functional enrichment analysis of Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) was performed on the targets. Molecular docking was used to further screen the potent active compounds in SGXZ. Finally, molecular dynamics (MD) simulation was applied to verify and validate the binding between the most potent compound and targets.

Results

A total of 31 active compounds and 220 corresponding targets in SGXZ decoction were collected. Moreover, 1,544 targets of NAFLD were obtained, of which 78 targets intersected with the targets of SGXZ decoction. Key compounds and targets were recognized through the compound–target–disease and PPI network. Multiple biological pathways were annotated, including PI3K-Akt, MAPK, insulin resistance, HIF-1, and tryptophan metabolism. Molecular docking showed that gallic acid, chlorogenic acid and isochlorogenic acid A could combine with the key targets. Molecular dynamics simulations suggested that isochlorogenic acid A might potentially bind directly with RELA, IL-6, VEGFA, and MMP9 in the regulation of PI3K–Akt signaling pathway.

Conclusion

This study investigated the active substances and key targets of SGXZ decoction in the regulation of multiple-pathways based on network pharmacology and computational approaches, providing a theoretical basis for further pharmacological research into the potential mechanism of SGXZ in NAFLD.

Introduction

Nonalcoholic fatty liver disease (NAFLD) is a progressive condition ranging from simple steatosis (NAFL) to nonalcoholic steatohepatitis (NASH), hepatic fibrosis, cirrhosis, and even hepatocellular carcinoma (HCC) (Challa et al., 2019; Raza et al., 2021). NAFLD is a metabolic disease representing the hepatic manifestation of a systemic metabolic disorder (Tilg & Effenberger, 2020) and has been demonstrated to be associated with obesity-related disorders and diabetes (Liu et al., 2020c; Stefan, Häring & Cusi, 2019). With the epidemic of obesity and metabolic-related comorbidities, the global incidence of NAFLD is estimated to be 25% and continues to rise (Mundi et al., 2020). NAFLD is one of the leading courses of chronic hepatic disease, affecting approximately 1.7 billion individuals worldwide (Zhou et al., 2020a). Moreover, NAFLD is one of the most common indications for liver transplantation and NAFLD-related HCC, which can be induced by steatosis-related lipotoxicity, is now the fastest growing reason for liver transplantation in the United States (Younossi et al., 2019; Younossi et al., 2016; Ioannou, 2021). Lifestyle modification including a healthy diet and increased physical activity, is recommended as the first-line treatment in NAFLD management (Jeznach-Steinhagen et al., 2019). Nevertheless, its effectiveness is limited in NAFLD patients due to their low readiness to change lifestyle, especially in terms of increasing physical activity (Centis et al., 2013). Another challenge for the lifestyle intervention has been the occurrence of weight regain (Kantartzis et al., 2009; Wadden et al., 2014).

At present, there are no US Food and Drug Administration- or European Medicines Agency–approved pharmacological therapies for the treatment of NAFLD (Chapman & Lynch, 2020; Cusi et al., 2022; Konerman, Jones & Harrison, 2018). Insulin sensitizers (rosiglitazone, pioglitazone and metformin), antioxidants (vitamin E), anti-inflammatory, and lipid-lowering drugs (atorvastatin and simvastatin) have been used for the treatment of NAFLD (Ahsan et al., 2020; Takahashi et al., 2015). However, the usage of rosiglitazone could increase the risk of cardiovascular events (Nissen & Wolski, 2007), while the long-term use of vitamin E (Klein et al., 2011; Miller 3rd et al., 2005) has been reported to increase the risk of prostate cancer and might increase all-cause mortality. Moreover, adverse outcomes of pioglitazone (Liao et al., 2017; Lincoff et al., 2007), atorvastatin (Pal et al., 2015) and simvastatin (Abdoli, Azarmi & Eghbal, 2014) including weight gain, edema, heart failure, cytotoxicity, and hepatotoxicity, should not be neglected. In recent years, dual peroxisome proliferator-activated receptor (PPAR) α/ γ agonists have appealed to global attention as promising new treatment options for metabolic syndrome, including diabetic complications and NAFLD (Cheng et al., 2019). Saroglitazar is a novel dual PPAR α/ γ agonist acting as a regulator of lipids and glucose metabolism (Sarkar et al., 2021). India approved saroglitazar in 2020 for the treatment of type-2 diabetes mellitus and NASH (Sarkar et al., 2021). However, a phase 2 trial conducted in the United States indicated that the role of saroglitazar in a particular subset of patients with elevated alanine aminotransferase without histological endpoints cannot be recommended (Kumar, Kulkarni & Jagdish, 2021; Shuja, Eqbal & Rehman, 2021). Therefore, there is an urgent need to develop pharmacological strategies for the treatment of NAFLD.

The complex pathophysiological mechanism of NAFLD reduces the efficacy of the application of a single agent. Traditional Chinese medicine (TCM) has been commonly used to treat hepatic disease and metabolic disorders–related to obesity and type 2 diabetes mellitus (Fu et al., 2021; Li, Zhang & Li, 2020a; Zhang et al., 2021). The therapeutic effect of TCM on NAFLD mainly involves holistic regulations including lipid metabolism modulation, anti-oxidative stress, anti-inflammation, and gut microbiota modulation (Dai et al., 2021; Liang et al., 2019; Ma, Zhou & Li, 2017). Shugan Xiaozhi (SGXZ) decoction is a composite traditional Chinese prescription composed of 12 herbs, including pleuri Radix, Paeoniae radix alba (stir-baked), Aurantii Fructus Immaturus, Glycyrrhizae Radixet Rhizoma, Artemisiae Scopariae Herba, Gardeniae Fructus, Poria, Alismatis Rhizoma, Crataegi Fructus, Cassiae Semen, Nelumbinis Folium, and Pumex in a ratio of 2: 1: 3: 1: 6: 2: 4: 6: 6: 6: 6: 6. Previous clinical studies have demonstrated that SGXZ could partially protect and restore the liver functions in NASH patients, which was associated with a significant reduction in aminotransferases, total cholesterol and triacylglycerol, and reparation of the intestinal mucosal barrier (Tang et al., 2018; Zhai et al., 2014). Mechanically, in a rat model, SGXZ decoction exerts the hepatoprotective effect through regulating fatty acid β-oxidation, relieving intestinal microecological disorder, and repairing the intestinal mucosal barrier (Xing et al., 2016; Zhai et al., 2021). However, the therapeutic mechanisms of SGXZ decoction in the treatment of NAFLD have not been fully clarified, and the active ingredients and key targets of SGXZ remain to be further investigated.

Based on the integration of system biology, bioinformatics and pharmacology (Lai et al., 2020; Liu & Sun, 2012; Zhang et al., 2019), network pharmacology is currently used to explore the potential pharmacological effects and underlying mechanisms of a drug on a disease (Guo et al., 2019; Zhang et al., 2020). Network pharmacology concentrates on the elucidation of complex biological relationships among the drugs, targets, pathways and diseases from a systemic and holistic perspective (Wang et al., 2021). The wholeness, relevance and systematic nature of network pharmacology are in line with the concept and treatment theory of TCM, providing a novel method and powerful tool to decipher the therapeutic mechanisms of TCM (Zhou et al., 2020b). Molecular docking is a computational method based on the analysis of the binding pose and affinity between small molecule and macromolecular target, which is widely utilized to predict and identify potentially active compounds (Álvarez-Carretero et al., 2018; Jiao et al., 2021). In addition, molecular dynamics (MD) simulations have been used widely to provide a full consideration of the flexibility of protein targets and obtain accurate binding modes and stability of compounds through Newtonian mechanics (Liu et al., 2018). In our study, network pharmacology combined with molecular docking and MD simulations was applied to explore the potential pharmacological and molecular mechanisms involved in the treatment of NAFLD by SGXZ decoction. The overall flowchart of this study is described in Fig. 1.

Figure 1. Schematic diagram of the whole study design.

Material and Methods

Collection of compounds and the putative targets of SGXZ decoction

The active compounds in SGXZ decoction were identified and collected from the literature. Then, 2D or 3D conformations of the compounds were downloaded either from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) or through sketching in ChemDraw software. Moreover, their simplified molecular input line entry system (SMILS) numbers (Table S1) were also acquired from PubChem or were converted by ChemDraw. To predict the putative gene target of SGXZ decoction, the SMILES of each chemical compound was submitted to Similarity Ensemble Approach (SEA) database (https://sea.bkslab.org/), which relates proteins based on the set-wise chemical similarity among their ligands (Keiser et al., 2007). The predicted protein targets of SGXZ decoction are listed in Table S2.

Identification of potential targets of NAFLD

“Nonalcoholic Fatty Liver Disease” or “Non-alcoholic Fatty Liver Disease” were used as the keywords to predict the potential NAFLD-related gene targets using GeneCards (https://www.genecards.org/) and DisGeNET (https://www.disgenet.org/) disease database. DisGeNET is a discovery platform containing one of the largest publicly available collections of genes associated with human diseases (Piñero et al., 2020). Based on all annotated and predicted human genes, GeneCards provides comprehensive, user-friendly information via a searchable, integrative database (Stelzer et al., 2016). After removing the repetitive targets from the two databases, the NAFLD-related protein targets were retained for further study. The detailed information on the targets is provided in Table S3.

Construction of compound-target-disease network

SGXZ decoction–target network was first constructed based on the compounds and their predicted targets of SGXZ using Cytoscape software (Version 3.7.1). Venny diagram 2.1 version (https://bioinfogp.cnb.csic.es/tools/venny/) was used to intersect the protein targets between the predicted targets of SGXZ decoction and the potential targets of NAFLD. Then, the compound–target–disease network was established based on the intersected common targets and their corresponding compounds. The degree value of nodes in the compound–target–disease network was analyzed with the NetworkAnalyzer plug-in in Cytoscape software. The degree indicates the total number of nodes connected to this node in the network, and nodes whose degree value is higher play a more essential regulatory role in the network. The compounds and targets that contributed significantly to the construction of the compound–target–disease network were recognized and selected as key active compounds and targets according to their higher degree value compared with the average degree value.

Construction of protein–protein interaction network

PPI network can elucidate the interaction relationship between targets, which further helps to identify the key nodes and targets in the network. The intersected common targets were submitted to the STRING database (https://string-db.org/; version 11.5) (Szklarczyk et al., 2019). The species was limited to “Homo sapiens”, and the confidence of interaction score was set at ≥ 0.400 as default, which ensured a reasonable interaction between targets. Results with tabular text output (TSV) format of the PPI network in STRING were exported and then visualized through Cytoscape software. The nodes with the highest degree value were considered the key targets that mediated the interaction of the network and participated in the treatment of NAFLD by SGXZ decoction.

Enrichment analysis

Functional annotation of the common targets in the PPI network was analyzed through the “Database for Annotation, Visualization, and Integrated Discovery” (DAVID) database (https://david.ncifcrf.gov/) to identify enriched biological themes of Gene Ontology (GO) and visualize genes on Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway maps. DAVID can provide a comprehensive set of functional annotation tools for investigators to understand the biological meaning behind a large list of genes and is widely used in bioinformatics (Huang Da, Sherman & Lempicki, 2009a; Huang Da, Sherman & Lempicki, 2009b). The gene list of the common target proteins was uploaded to DAVID, and the species was limited to “Homo sapiens”. The functional annotation of biological processes (BPs), molecular functions (MCs), cellular components (CCs), and KEGG pathways with a P value <0.05 was then performed for further analysis. Accordingly, the target–pathway network was subsequently constructed using Cytoscape software.

Homology modeling and molecular docking

The 3D conformational SDF format of the selected key chemical compounds was downloaded from PubChem. The crystal structures of the key protein targets were retrieved in the RCSB Protein Data Bank database (PDB, https://www.rcsb.org/). For targets without released crystal structures, homology modeling was utilized to establish the structure based on the protein template. The construction was completed on the SWISS-MODEL website (https://swissmodel.expasy.org/), which is a fully automated protein structure homology–modeling server (Waterhouse et al., 2018). Then, the quality of the model structures was evaluated via the “Ramachandran plot” on the SAVES v6.0 website (https://saves.mbi.ucla.edu/). The target protein with neither crystal structure nor the template was excluded in molecular docking. Before docking, the compounds were charged with CHARMm force field and minimized for optimization in Discovery Studio 2020 (DS) software. The target proteins were also loaded in DS, and water and other redundant chains were first deleted. Then, excess structures of the protein were removed, incomplete amino acid residues were supplemented, and hydrogens were added to optimize the conformation. CDOCKER module in DS was applied to calculate the binding energy between the key active ingredients and key target proteins. CDOCKER is a grid-based molecular docking method based on CHARMm force field, which can produce high-precision docking results. For docking parameters of CDOCKER, the size of the docking box was set to 20 Å× 20 Å× 20 Å, “Top Hits” was set at 10; and “Pose Cluster Radius” was set at 0.5. Moreover, the docking site of each target protein was obtained from previous research or the active cavities detected in DS (Table S4). Only one top hit with the best docking pose of each complex was reported and saved for further analysis.

MD simulations

MD simulations using the Amber16 software package were performed on the complexes of the most potentially active compound and target proteins. Gaussian 16 and antechamber program were employed to prepare the force-filed parameters for the compound, and the ff14sb filed force parameters were applied for the proteins. The compound and protein were then loaded into the leap module, and the system was neutralized by automatically adding hydrogen atoms and antagonist ions. All the systems were immersed in the TIP3P water model with periodic boundary conditions and were first minimized through 2,000 steps of the steepest descent and 2,000 steps of conjugate gradients with the constraining of all atoms of protein and compound. Next, only the heavy atoms of the proteins were confined to minimize the water and the side chains by steepest descent minimization of 5,000 steps followed by a conjugate gradient minimization of 5,000 steps. Finally, the entire system was set with no constraints for the optimization with the same step as the last stage. The system was then heated slowly to 310 K and was balanced under nvt ensemble for 20 ps. Then, 100 ns simulations were performed, and the trajectory was saved every 500 ps. The CPPTRAJ module was exploited for the root mean square deviation (RMSD) analysis and the calculation of binding free energy of active compound and protein was carried out using molecular mechanics/generalized Born surface area (MM/GBSA) method by the MMPBSA.py program in AMBER 16.

Results

Active compounds of SGXZ decoction and the corresponding targets

A total of 31 active compounds of SGXZ decoction (SX01 to SX31) were identified and collected from the previous study (Xing et al., 2016) with 2D chemical structures displayed in Fig. 2. Totally, 220 targets of SGXZ decoction were screened after removing duplicates. The interaction of compounds in SGXZ decoction with their corresponding targets was constructed (Fig. 3). The interaction network comprised 252 nodes and 686 edges with 31 compounds and 220 corresponding targets, revealing the multitarget effect of SGXZ decoction.

Figure 2. 2D structures of the 31 identified compounds in SGXZ decoction.

Figure 3. SGXZ decoction–targets interaction network.

Red hexagon, pink triangle, and green circle represent SGXZ decoction, compounds, and targets, respectively.

Potential targets of NAFLD and the compound–target–disease interaction

Potential targets of NAFLD were independently screened in DisGeNET and GeneCards databases. After the overlapped targets were deleted, 1544 potential protein targets of NAFLD from the two databases were retained. The Venny diagram showed that there were 78 common targets between SGXZ and NAFLD, which represented the potential targets of SGXZ in the treatment of NAFLD (Fig. 4).

Figure 4. Overlapped targets between SGXZ decoction and NAFLD.

To further examine the direct regulatory effect of SGXZ decoction in treating NAFLD, the network of compounds with common intersection targets was constructed (Fig. 5A), which consisted of 109 nodes (31 compounds and 78 common targets) and 232 edges, suggesting multiple components involved with multiple targets of SGXZ decoction in the treatment of NAFLD. In addition, the top 10 compounds and common protein targets were obtained (Fig. 5B), and the detailed information is listed in Tables 1 and 2. SX01 (gallic acid) possessed the highest degree value (degree = 20), followed by SX27 (kaempferol-3-O-galactoside, degree = 19), SX31 (licoricone, degree = 17), SX13 (isochlorogenic acid A, degree = 16), SX07 (chlorogenic acid, degree = 12), SX11 (liquiritin, degree = 12), SX05 (gardenoside, degree = 11), SX02 (naringin, degree = 10), SX03 (neohesperidin, degree = 10), and SX04 (poncirin, degree = 9). These active compounds were recognized in SGXZ decoction to potentially exert the pharmacologic effect on NAFLD. Moreover, as listed in Table 2, the top 10 targets with the highest degree value included FGF1 (degree = 15), VEGFA (degree = 15), IL-2 (degree = 13), LGALS3 (degree = 11), SLC5A2 (degree = 11), F3 (degree = 10), ALDH1B1 (degree = 7), CES2 (degree = 7), IL-6 (degree = 7), and P4HB (degree = 7), which could be potentially acted on by the 10 key active compounds in SGXZ decoction.

Figure 5. Compound–target–disease interaction network.

(A) Interaction network of the compound–target–disease. Size of nodes corresponds to the value of the degree. SGXZ decoction chemical compounds and targets are colored in pink and blue, respectively; (B) Top 10 putative compounds and targets with the highest degree value in the compound–target–disease network.

Table 1. Key active compounds of SGXZ decoction in treatment of NAFLD.

Number	Ingredient	Molecular formula	Pubchem_CID	Degree
SX01	Gallic acid	C7H6O5	370	20
SX02	Naringin	C27H32O14	442428	10
SX03	Neohesperidin	C28H34O15	442439	10
SX04	Poncirin	C28H34O14	442456	9
SX05	Gardenoside	C17H24O11	24721095	11
SX07	Chlorogenic acid	C16H18O9	1794427	12
SX11	Liquiritin	C21H22O9	503737	12
SX13	Isochlorogenic acid A	C25H24O12	6474310	16
SX27	Kaempferol-3-O-galactoside	C21H20O11	5282149	19
SX31	Licoricone	C22H22O6	5319013	17

Table 2. Key targets of SGXZ decoction in the treatment of NAFLD.

UniprotKB	Target	Protein name	Degree
P05230	FGF1	Fibroblast growth factor 1	15
P15692	VEGFA	Vascular endothelial growth factor A	15
P60568	IL-2	Interleukin-2	13
P17931	LGALS3	Galectin-3	11
P31639	SLC5A2	Sodium/glucose cotransporter 2	11
P13726	F3	Tissue factor	10
P30837	ALDH1B1	Aldehyde dehydrogenase X, mitochondrial	7
O00748	CES2	Cocaine esterase	7
P05231	IL-6	Interleukin-6	7
P07237	P4HB	Protein disulfide-isomerase	7

Outcome of the PPI network

The PPI network was constructed from the common 78 targets (Fig. 6). Due to the lack of any interactions with other targets, chymotrypsin-C (CTRC) and alpha-ketoglutarate-dependent dioxygenase (FTO) were removed from the network. Therefore, the network contained 76 nodes and 376 edges, with an average degree value of 9.64. The target proteins with higher degree value than average degree value were IL-6, VEGFA, estrogen receptor (ESR1), hypoxia-inducible factor 1-alpha (HIF1A), matrix metalloproteinase-9 (MMP9), amyloid-beta precursor protein (APP), cAMP-responsive element binding protein 1 (CREB1), heat-shock protein family A member 5 (HSPA5), transcription factor p65 (RELA), and IL-2. The 10 targets were indispensable for constructing the PPI network. IL-6 is a proinflammatory factors and is associated with metabolic disorders (Giraldez et al., 2021). Increased IL-6 level promotes hepatic insulin resistance and impairs lipid metabolism (Akbari & Hassan-Zadeh, 2018; Singh et al., 2021). Similarly, upregulated IL-2 potentially leads to increased insulin resistance and several other metabolic inflammatory markers in the obese population (Kochumon et al., 2020). VEGFA is a major proangiogenic cytokine that regulates angiogenesis. Increased VEGFA accelerates angiogenesis and therefore drives hepatic inflammation and fibrosis in NAFLD (Lefere, Devisscher & Geerts, 2020; Surapaneni, Vishnu Priya & Mallika, 2015). Upregulated HIF1A expression in the liver leads to aggravated steatosis by suppression of fatty acid (FFA) β-oxidation and by induction of FFA uptake and inflammatory factors (Holzner & Murray, 2021; Wu et al., 2019). RELA is a pivotal transcription factor that regulates inflammatory molecules (Gasparini & Feldmann, 2012). FGF1 exerts a protective role in a series of metabolic disorders (Nies et al., 2015). IL-6, IL-2, FGF1 and VEGFA are located at the upstream of the PI3K–Akt pathway and can stimulate the pathway, and RELA can be modulated by the stimulation of PI3K–Akt. MMP9 has a vital role in modulating and degrading gelatins, collagens and other ECM compounds (D’Amico et al., 2010; Lachowski et al., 2019). A decreased level of MMP9 is associated with more advanced fibrosis and serum liver injury indexes in NAFLD patients (Goyale et al., 2021). GRP78 (HSPA5) is a chaperone heat shock protein playing the central role in maintaining ER proteostasis under excessive stress (Pobre, Poet & Hendershot, 2019). LGALS3 is an essential regulator of insulin resistance, fibrosis and inflammation cytokines including TNF- α, IL-6 and IL-1β, and has been shown to participate in glucose intolerance and lipid metabolism disorders (Li et al., 2020b; Yu et al., 2021). CREB1 has been found to promote insulin resistance (Yoon et al., 2021) and ESR1 is associated with adiposity and mitochondrial metabolism (Zhou et al., 2020c). These targets were considered the putative key targets of SGXZ decoction for the treatment of NAFLD.

Figure 6. Protein–protein interaction (PPI) network.

(A) PPI network of potential targets of SGXZ decoction for treating NAFLD. The size of the node with the shade of the color indicates the corresponding degree value; (B) The identified top 10 putative targets with degree value higher than the average according to the PPI network.

GO and KEGG enrichment analysis

The GO functionally annotated targets in the PPI network involved three main aspects, including molecular function (MF), cellular component (CC), and biological process (BP). In total, 185 GO entries, including 45 of MF, 25 of CC, and 115 of BP, were acquired based on the P value (P < 0.05). As for BP, the targets were mainly concentrated on the oxidation–reduction process, positive regulation of angiogenesis, negative regulation of the apoptotic process, regulation of insulin-like factor receptor signaling pathway, and so on. As for CC, the targets were mainly responsible for the extracellular space, extracellular exosome, cell surface, extracellular region, mitochondrion, extracellular matrix, and so on. As for MF, the intersection targets were distributed in enzyme binding, insulin-like growth factor binding, electron carrier activity, oxidoreductase activity, receptor binding, protein heterodimerization activity, iron ion binding, glycosphingolipid binding, and protein binding. The bubble plot of the most significant enriched GO terms is shown in Fig. 7.

Figure 7. Representative GO enrichment analysis based on the targets in the PPI network.

Bubble plot of the top 10 terms of BP, CC, and MF. The bigger the dot, the more genes are enriched in the term, while the redder color indicates the smaller P value.

Forty-three annotated pathways were also obtained based on the targets in the PPI network, and the top 20 pathways with the smallest significance value are shown in Fig. 8A and Table 3. The bubble plot of the pathways suggested a concentration of insulin resistance, amino acid metabolism, cancer-related pathways, and inflammation-related pathways including MAPK signaling pathway, HIF-1 signaling pathway, TNF signaling pathway, and PI3K–Akt pathway. Moreover, the common targets also concentrated on the endocrine biological processes such as the estrone signaling pathway and immunological pathway of T cell receptors. To reveal the network interaction of the pathways and the involved targets, the target–pathway network was constructed and analyzed, which consisted of 58 nodes and 130 edges (Fig. 8B), indicating a complex interaction among them. The largest node size was for the pathways in cancer, followed by PI3K–Akt, MAPK signaling pathway, proteoglycan in cancer, and HIF-1 pathway, suggesting their significant role in the treatment of NAFLD.

Figure 8. Results of KEGG pathway enrichment analysis based on the targets in the PPI network.

(A) The bubble plot of the top 20 pathway enrichment analysis; (B) target–pathway interaction network suggests the underlying mechanisms of SGXZ decoction for the treatment of NAFLD. Blue circle nodes indicate the targets and orange diamonds represent the pathways. The gray edges link the interaction between the targets and pathways, and the node size is proportional to the degree value.

Table 3. Annotation of KEGG pathways and the involved potential targets.

KEGG ID	Term	Count	P value	Targets
hsa04931	Insulin resistance	8	7.75E−05	RPS6KA3, PTPN1, IL-6, CREB1, PRKCE, PTPN11, RELA, NFKB1
hsa04066	HIF-1 signaling pathway	7	3.17E−04	IL-6, SERPINE1, PRKCA, HIF1A, RELA, NFKB1, VEGFA
hsa00380	Tryptophan metabolism	5	5.77E−04	MAOB, ALDH2, ALDH1B1, MAOA, CYP1A1
hsa04726	Serotonergic synapse	7	6.91E−04	APP, CYP2C8, MAOB, MAOA, ALOX5, PRKCA, ALOX12
hsa00340	Histidine metabolism	4	0.001191	MAOB, ALDH2, ALDH1B1, MAOA
hsa05030	Cocaine addiction	5	0.001253	CREB1, MAOB, MAOA, RELA, NFKB1
hsa05200	Pathways in cancer	12	0.001302	IL-6, MMP1, MMP2, RARA, PRKCA, IKBKG, FGF1, HIF1A, MMP9, RELA, NFKB1, VEGFA
hsa00330	Arginine and proline metabolism	5	0.001352	MAOB, P4HA1, ALDH2, ALDH1B1, MAOA
hsa05134	Legionellosis	5	0.001803	IL-6, RELA, NFKB1, HSPD1, HSPA1A
hsa05161	Hepatitis B	7	0.002742	IL-6, CREB1, PRKCA, IKBKG, MMP9, RELA, NFKB1
hsa04010	MAPK signaling pathway	9	0.002986	RPS6KA3, NR4A1, PRKCA, IKBKG, MAPT, FGF1, RELA, NFKB1, HSPA1A
hsa05205	Proteoglycans in cancer	8	0.003139	MMP2, PTPN6, PTPN11, PRKCA, HIF1A, ESR1, MMP9, VEGFA
hsa05142	Chagas disease (American trypanosomiasis)	6	0.003296	IL-6, SERPINE1, IKBKG, RELA, NFKB1, IL-2
hsa04668	TNF signaling pathway	6	0.003726	IL-6, CREB1, IKBKG, MMP9, RELA, NFKB1
hsa05202	Transcriptional misregulation in cancer	7	0.005495	IL-6, BCL2A1, IGFBP3, RARA, MMP9, RELA, NFKB1
hsa04151	PI3K-Akt signaling pathway	10	0.005885	NR4A1, IL-6, CREB1, PRKCA, IKBKG, FGF1, RELA, NFKB1, IL-2, VEGFA
hsa05219	Bladder cancer	4	0.007229	MMP1, MMP2, MMP9, VEGFA
hsa05215	Prostate cancer	5	0.010389	CREB1, SRD5A2, IKBKG, RELA, NFKB1
hsa04915	Estrogen signaling pathway	5	0.015491	CREB1, MMP2, ESR1, MMP9, HSPA1A
hsa04660	T cell receptor signaling pathway	5	0.016021	PTPN6, IKBKG, RELA, NFKB1, IL-2

Verification of the binding between the key compounds and targets

Molecular docking was employed to verify the binding between the key compounds and the key targets, revealing the potential therapeutic effect of SGXZ decoction acting on these targets. Crystal structures of the targets were downloaded from the PDB database, except for SLC5A2, CES2, and ALDH1B1 for their lack of 3D structures. Therefore, homology modeling was used to build the structures of SLC5A2, CES2, and ALDH1B1 (Song et al., 2021; Wang et al., 2019). The structural assessment of ALDH1B1 is provided in Fig. S1, which showed the good quality of the modeled structure. However, the crystal structure of CREB1 could not be obtained either from the PDB database or through the homology modeling since there were no templates of good quality. Besides, three duplicated key targets between the compound–target–disease network and the PPI network were removed. Thus, a total of 16 key targets were docked with 10 key compounds. After docking, 160 pairs of the compound–target complexes were retained according to their best binding affinity, and the heat map of their docking energy is shown in Fig. 9. Moreover, the binding energy of each pair of compound–target complex is provided in Table S5. Binding energy with a negative value indicated that the ligand molecule was able to combine with the receptor target proteins. The lower and more negative binding energy suggested a better binding affinity of active compounds with target proteins. Among the 10 compounds, SX01 (gallic acid), SX07 (chlorogenic acid), and SX13 (isochlorogenic acid A) could combine with all of the targets. The binding energy between the three compounds and targets and the residues that produce hydrogen bonds shared by all three active compounds are listed in Table 4. As shown, SX13 could bind with all of the key targets with the lowest binding energy, followed by SX07 and SX01. The binding location between the best poses of the three compounds and targets is displayed in Fig. 10. The results of molecular docking revealed that SX01, SX07, and especially SX13 might be the potentially active ingredients of SGXZ decoction for treating NAFLD through targeting these proteins.

Figure 9. Heat map of the docking energy (kcal/mol) between the key active ingredients and the key targets.

The bluer the color, the lower the binding energy and the stronger the binding ability.

Table 4. Binding energy between key active compounds and targets (kcal/mol).

Target proteins	Ingredients			Shared Hydrogen Bond Interactions (Amino acid)
	SX01	SX07	SX13
FGF1	−27.07	−22.06	−38.21	LYS112, LYS113, ASN18, LYS128, LYS118
VEGFA	−22.19	−16.77	−31.33	N/A
IL-2	−19.90	−15.25	−29.85	LYS35
LGALS3	−22.81	−17.71	−34.94	ASN174, GLU184
SLC5A2	−27.92	−31.68	−44.10	SER77, ASP294, GLN295
F3	−24.94	−5.43	−19.86	LYS65, LYS46
ALDH1B1	−29.62	−28.87	−24.23	CYS302, CYS303
CES2	−27.76	−23.35	−32.97	SER228
IL-6	−26.37	−17.42	−30.38	N/A
P4HB	−19.07	−12.82	−33.19	TRP396
ESR1	−35.57	−24.86	−3.86	GLU353, LEU346
MMP9	−23.77	−25.52	−48.88	HIS230, ALA191
HIF1A	−28.66	−28.22	−39.94	THR196, HIS199
APP	−25.31	−21.32	−32.41	LYS447, ASP444, HIS439
HSPA5	−33.25	−43.27	−53.91	GLY364, SER365, LYS296, GLU293, ARG297
RELA	−9.12	−15.06	−23.01	N/A
Average	−25.21	−21.85	−32.57

Notes.

^N/Anone hydrogen bond interaction residues shared by SX01, SX07 and SX13 with the protein.

Figure 10. Combination between SX01, SX07, SX13, and targets.

Target proteins are displayed as a cyan cartoon, while SX01 (gallic acid), SX07 (chlorogenic acid), and SX13 (isochlorogenic acid A) are displayed on sticks and colored in blue, yellow, and pink, respectively.

RMSD analysis and binding free energy

MD simulations were performed to validate the binding ability between the most potentially active compound and the key target proteins after molecular docking. SX13 was selected for MD simulations due to its most negative average docking energy compared to SX01 and SX07, which indicated a putative stronger binding ability of SX13. The proteins targets RELA, IL-6, MMP9, and VEGFA were chosen for their significant role both in the PPI network and the target–pathway interaction network. The equilibrium of all four systems was evaluated by RMSD, which represents the relatively stable combination of protein and compound. As shown in Fig. 11, the complex of VEGFA and SX13 maintained stability throughout the simulation, while the complexes IL-6/SX13 and MMP9/SX13 stabilized after 70 ns. As for the complex RELA/SX13, the conformation showed a slight larger fluctuation in the early stage and achieved the equilibrium at the last 20 ns simulations. As RELA is a macromolecular polymer consisting of more than 500 amino acids, it is reasonable that the complex of RELA and SX13 showed larger fluctuation than the other three complexes. Therefore, the MD trajectories between 80 and 90 ns simulations were selected for the calculations of binding free energy. The binding free energy and its components, including Van der Waals potential, electrostatic interaction, non-polar solvation free energy, and polar solvation free energy, which were calculated by MM/GBSA of the four systems, are listed in Table 5. The MM/GBSA results showed that the average binding free energy between SX13 and RELA, IL-6, MMP9, and VEGFA was −23.06 ± 1.69 kcal/mol, −21.93 ± 1.07 kcal/mol, −21.04 ± 1.31 kcal/mol, and −17.28 ± 1.37 kcal/mol, respectively. The second repeated simulation further strengthened the bind between SX13 and these target proteins (Fig. S2 and Table S6). The results demonstrated that SX13 could potentially combine directly with these key protein targets. The detailed binding of the protein with the compound is displayed in Fig. 12.

Figure 11. RMSD plot of the backbone atoms for isochlorogenic acid A with IL-6, RELA, MMP9, and VEGFA.

Table 5. Binding free energy and energy components calculated by MM/GBSA (kcal/mol).

	RELA	IL-6	VEGFA	MMP9
ΔE vdW	−34.79 ± 2.30	−27.05 ± 1.25	−30.54 ± 1.02	−33.50 ± 1.30
ΔE elec	−11.59 ± 6.18	−38.13 ± 1.85	−11.26 ± 3.36	−26.58 ± 4.06
ΔG GB	27.40 ± 5.81	47.48 ± 2.14	28.87 ± 3.42	43.96 ± 3.82
ΔG SA	−4.09 ± 0.22	−4.23 ± 0.19	−4.35 ± 0.15	−4.92 ± 0.18
ΔG bind	−23.06 ± 1.69	−21.93 ± 1.07	−17.28 ± 1.37	−21.04 ± 1.31

Notes.

ΔE_vdW

van der Waals energy

ΔE_elec

electrostatic energy

ΔG_GB

electrostatic contribution to the solvation

ΔG_SA

non-polar contribution to the solvation

ΔG_bind

binding free energy

Figure 12. Conformational diagrams of the RELA–SX13 (A), IL-6–SX13 (B), MMP9–SX13 (C), and VEGFA–SX13 interactions (D).

Discussion

In the present study, the compounds of SGXZ decoction with corresponding targets were first identified, and the overlapping targets between SGXZ decoction and NAFLD were considered the main targets these compounds acted on. Ten compounds with their important contributions to the compound–target–disease network were selected as the key bioactive ingredients for the treatment of NAFLD. Target proteins, including VEGFA, FGF1, IL-2, LGALS3, SLC5A2, and IL-6, were excessively expressed in the compound–target–disease network and were selected as key targets. In addition, 10 targets with the highest degree in the PPI network were also recognized as key targets acted on by SGXZ decoction. The binding between the key compounds and the key targets was verified by molecular docking, which showed that gallic acid (SX01), chlorogenic acid (SX07), and isochlorogenic acid A (SX13) could bind with all of the key targets. Among them, isochlorogenic acid A showed the lowest binding energy, indicating its potential pharmacological activity. A previous study has suggested that isochlorogenic acid A possesses hepatoprotective properties and anti-hepatitis B effects through suppressing oxidation (Hao et al., 2012). Moreover, it has been indicated that isochlorogenic acid A exerts a protective effect on liver fibrosis by inhibiting inflammation via the HMGB1/TLR4/NF-κB signaling pathway (Liu et al., 2020b). Gallic acid, a simple polyphenol, has been reported to reduce lipid accumulation related to β-oxidation and ketogenesis (Chao et al., 2020). Specifically, the hepatoprotective effect of gallic acid attributed to the repression of inflammatory signaling pathways including nuclear factor-κB (NF-κB)/TNF-α/IL-6 and ROS/NF-κB/TNF-α in NAFLD rats (Fanaei et al., 2021a; Fanaei et al., 2021b). Chlorogenic acid, a natural polyphenol extracted from Artemisiae Scopariae (Yin-Chen in Chinese) (Cai et al., 2020) and Lonicera japonica (Jin-Yin-Hua) (Liu et al., 2020a), could ameliorate HFD-induced hepatic steatosis and inflammation via inhibition of TNF-α and IL-6 in the liver, which was associated with the regulation of gut microbiota and an increase in glucagon-like peptide-1 secretion (Shi et al., 2021; Zamani-Garmsiri et al., 2021).

After that, GO functional enrichment analysis revealed that the targets in the PPI network mainly involved oxidation, positive regulation of angiogenesis, metabolic process, hypoxia, extracellular matrix and insulin-like factor binding, and other biological processes. In addition, KEGG enrichment analysis indicated that the action pathways mainly included insulin resistance, amino acid metabolism, cancer-related pathways, inflammation-related pathways, estrone signaling pathway, and T cell receptor pathway. The target–pathway interaction network suggested that SGXZ decoction might perform a therapeutic role via multiple targets and multiple pathways. Through literature retrieval, signaling pathways including PI3K–Akt, MAPK, insulin resistance, HIF-1, tryptophan metabolism, and estrogen signaling pathway were recognized as core pathways involved in the treatment of NAFLD. The PI3K–Akt signaling pathway, the second largest node size in the target–pathway, is associated with inflammation, which is the driving force for the development and evolution of NAFLD (Byun et al., 2021; Sutti & Albano, 2020). Previous research has provided evidence that the regulation of PI3K–Akt signaling mediators could improve hepatocyte damage, hepatic gluconeogenesis, and lipid disorder (Chen, Liu & Peng, 2019; Liao et al., 2018; Yang et al., 2019). Moreover, inhibiting apoptosis and promoting autophagy via the reactive oxygen species (ROS)/MAPK pathway significantly decreased total cholesterol and triglyceride levels of both plasma and liver in HFD-fed mice (Wu et al., 2021). In addition, increased ROS formation, triglyceride, and lipid accumulation in hepatocytes led to excessive oxidative stress and inflammatory responses, which further produced a hypoxic microenvironment (He et al., 2021; Lei et al., 2021). HIF-1 signaling pathway acts as the key response to hypoxia. Loss of HIF-1α (HIF1A) activates oxidative stress and promotes lipid deposition and secretion of proinflammatory factors IL-6 and TNF-α, suggesting its regulatory role in the progression of NAFLD (Arai, Tanaka & Goda, 2018; He et al., 2021). Indeed, increased FFAs and lipogenesis in hepatocytes, the release of proinflammatory cytokines from adipose tissue such as IL-6 and TNF-α, together with altered gut microbiota give rise to insulin resistance, which predisposes to the development of NAFLD and progression to NASH (Buzzetti, Pinzani & Tsochatzis, 2016; Khan et al., 2019; Lim et al., 2010). Tryptophan is an amino acid that can give rise to indoles by the bacterial enzyme trytophanase A in intestinal epithelial cells. Indoles promote intestinal barrier function and could translocate from the intestines to the liver to modulate hepatic lipid metabolism and inflammation to protect against NAFLD (Ding et al., 2019). In general, the abovementioned core pathways and their interaction correlated to NAFLD might be acted on by SGXZ decoction in the treatment of NAFLD.

Combining the PPI network with the target–pathway network revealed the significant roles of proteins RELA, IL-6, VEGFA, and MMP9. Besides, the four targets played an important regulatory role in the PI3K–Akt signaling pathway according to the KEGG reference pathway map. To further validate the binding between the four targets and the most latent compound isochlorogenic acid A after molecular docking, MD simulations were employed due to their more accurate calculation of binding ability between ligand molecule and protein receptor. The binding free energy of isochlorogenic acid A with RELA, IL-6, MMP9, and VEGFA was −23.06 ± 1.69 kcal/mol, −21.93 ± 1.07 kcal/mol, −21.04 ± 1.31 kcal/mol, and −17.28 ± 1.37 kcal/mol, respectively. This suggests that the proteins could be potentially targeted by isochlorogenic acid A to exert a pharmacological action. RELA, also known as p65, is one of the five members of the NF-κB family that is regulated by its upstream regulator PI3K/Akt. It is a pivotal transcription factor regulating inflammatory molecules (Gasparini & Feldmann, 2012). Inhibition of NF-κB signaling alleviates hepatic lipid accumulation and hepatic inflammation in NAFLD (Daniel et al., 2021; Yang et al., 2021). IL-6 has inflammatory properties and is associated with metabolic disorders (Giraldez et al., 2021). Numerous studies have highlighted that increased IL-6 level promotes hepatic insulin resistance (Akbari & Hassan-Zadeh, 2018) and impairs lipid metabolism (Singh et al., 2021). IL-6 deficiency ameliorates the hepatic inflammation and injury in NASH mice fed methionine and choline–deficient diet (Mas et al., 2009a). VEGFA is a major proangiogenic cytokine that regulates angiogenesis. Increased VEGFA has been reported in the liver of animal models and serum of NASH patients (Lefere, Devisscher & Geerts, 2020; Surapaneni, Vishnu Priya & Mallika, 2015), which accelerates angiogenesis and therefore drives hepatic inflammation and fibrosis in NAFLD (Lefere, Devisscher & Geerts, 2020). The onset of hepatic inflammation causes fibrogenesis in NAFLD, which is manifested by deposited ECM proteins, including collagens, elastin, and fibronectin (Mas et al., 2009b; Munsterman et al., 2018). MMP9 performs a vital role in modulating and degrading gelatins, collagens, and other ECM compounds (D’Amico et al., 2010; Lachowski et al., 2019). A decreased MMP9 level is associated with more advanced fibrosis and serum liver injury indexes (AST, GGT) in NAFLD patients, while increased MMP9 activity could precede the clearance of the fibrotic matrix (Goyale et al., 2021; Trojanek et al., 2020).

In light of the above, we assume that SGXZ decoction might comprehensively regulate inflammation, lipid deposition, insulin resistance, and fibrosis in NAFLD by targeting multiple pathways and targets.

Conclusions

In this study, network pharmacology combined with molecular docking and MD simulations was performed to investigate the putative active ingredients, novel targets, and pivotal signaling pathways of SGXZ decoction in the treatment of NAFLD. It was found that chlorogenic acid, gallic acid, and especially isochlorogenic acid A might exert the pharmacological effect through the PI3K–Akt signaling pathway; this regulation was associated with RELA, IL-6, VEGFA, MMP9, and other protein targets. We further found that the MAPK signaling pathway, insulin resistance, HIF-1 signaling pathway, tryptophan metabolism, estrogen signaling pathway, and other pathways play a role in the regulation of NAFLD. This work provided insight into the therapeutic strategies of NAFLD and evidence for future research. Nonetheless, additional in vitro and in vivo studies should be carried out to corroborate the identified potent active substances, key targets and pivotal pathways.

Funding Statement

This work is financially supported by the Shenzhen Science and Technology Planning Project (grant no. JCYJ20190812161605538), the Shenzhen Key Laboratory (ZDSYS20210623092000002), the Guangdong Natural Science Foundation (grant no. 2018A030313181), and Shenzhen Traditional Chinese Medicine Hospital “3030 Program” Chinese Medicine Clinical Project (G3030202115). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.