The seamless integration of dietary plant-derived natural flavonoids and gut microbiota may ameliorate non-alcoholic fatty liver disease: a network pharmacology analysis

Abstract We comprised metabolites of gut microbiota (GM; endogenous species) and dietary plant-derived natural flavonoids (DPDNFs; exogenous species) were known as potent effectors against non-alcoholic fatty liver disease (NAFLD) via network pharmacology (NP). The crucial targets against NAFLD were identified via GM and DPDNFs. The protein interaction (PPI), bubble chart and networks of GM or natural products- metabolites-targets-key signalling (GNMTK) pathway were described via R Package. Furthermore, the molecular docking test (MDT) to verify the affinity was performed between metabolite(s) and target(s) on a key signalling pathway. On the networks of GNMTK, Enterococcus sp. 45, Escherichia sp.12, Escherichia sp.33 and Bacterium MRG-PMF-1 as key microbiota; flavonoid-rich products as key natural resources; luteolin and myricetin as key metabolites (or dietary flavonoids); AKT Serine/Threonine Kinase 1 (AKT1), CF Transmembrane conductance Regulator (CFTR) and PhosphoInositide-3-Kinase, Regulatory subunit 1 (PIK3R1) as key targets are promising components to treat NAFLD, by suppressing cyclic Adenosine MonoPhosphate (cAMP) signalling pathway. This study shows that components (microbiota, metabolites, targets and a key signalling pathway) and DPDNFs can exert combinatorial pharmacological effects against NAFLD. Overall, the integrated pharmacological approach sheds light on the relationships between GM and DPDNFs.


Introduction
Non-alcoholic fatty liver disease (NAFLD) is an initial stage to develop malignant cascade, which might lead to hepatocellular death as irreversible condition: cirrhosis [1] and hepatocellular carcinoma (HCC) [2].These diseases represent chronic inflammation, including liver damage to be aggravated by NAFLD [3,4].Moreover, alterations of gut microbiota (GM) have notable characteristics such as intestinal dysbacteriosis; enhancement of intestinal penetrability and disparity of endogenous compounds: bile acids, choline and free fatty acids [5].
In the context of metabolites from GM, they play critical roles to modulate inflammatory level via gut-liver crosstalk, suggesting that metabolites from GM can traverse liver through portal vein [6,7].Most recently, it has been suggested that equol converted from daidzein in the gut has potent antioxidative, antiandrogenic and oestrogenic effects in host system [8,9].Interestingly, individual variation to produce equol depends on the diversity of GM, which is responsible for converting from precursor [10].It implies that GM might play important roles to produce valuable flavonoids against diseases.Furthermore, dietary plant-derived natural flavonoids (DPDNFs) and GM have interconnected intricately to biological benefits for human health [11].The DPDNFs in myriad plants, fruits and vegetables own the therapeutic efficacy such as antioxidant, anti-inflammatory, anticancer and antiviral qualities [12].
An animal experiment indicated that myricetin in flavonoid species has alleviative effects by modulating hepatic lipid mechanism on NAFLD in obese mice [13].The spinach, peas, turnip, carrot, cauliflower and cranberry are in plentiful supply of myricetin, its key mechanism diminished the amount of triglycerides and cholesterol [14].Additionally, another animal test suggested that luteolin reduces inflammatory factors and alleviates dysbiosis by GM, which leads to the restoration in aggravated intestinal mucosal barrier [15].The peppers, carrots, broccoli, celery, cabbages and parsley are abundant providers of luteolin, which have been taken as alternatives for alleviation in hypertension, inflammatory diseases and cancer [16].
Collectively, we investigated promising components to ameliorate NAFLD by comprising GM and DPDNFs.In parallel, we explored its interactive metabolites, targets and a key signalling pathway against NAFLD, via incorporating GM and DPDNFs.
Specifically, network pharmacology (NP) is an important methodology to elucidate the integrated biological information, including complex microbiome analysis [17,18].To pioneer the pharmacological pathways and relevant mechanisms is significant processing to interpret the combinatorial effects via NP concept [19].Thus, we utilized NP-based analysis to expound the relationships between GM-metabolites-targets and a key signalling pathway with DPDNFs against NAFLD.From this holistic viewpoint, we needed to make a new methodology and concept to reveal the combinatorial effects on GM and DPDNFs.At present, four crucial components (GM, targets, metabolites and a key signalling pathway) and key DPDNFs have yet to be documented.
The aim of this study is to uncover the combinatorial value in microbiome components and DPDNFs against NAFLD.Firstly, we retrieved GM targets and key DPDNFs from public database.Then, we browsed flavonoids-related and NAFLD-related targets with bioinformatics database.Also, the identified crucial targets were utilized to analyse PPI and a bubble chart.Moreover, we utilized molecular docking test (MDT) to obtain affinity between the targets and molecules on a key signalling pathway against NAFLD.More importantly, we integrated relationships with GM or DPDNFs metabolitestargetsa key signalling pathway.In the merged networks, it has been shown that GM or DPDNFs might exert combinatorial effects.Collectively, we interpret the medicinal value of GM as endogenous species, DPDNFs as exogenous species in integrated pharmacological viewpoints.The workflow of this study is displayed in Figure 1.

Database platforms and data analysis associated with NP
Biologically significant databases that give a large amount of data linked to the relationship between ligands and targets, which is enables the researcher to utilize NP as a powerful tool for drug development.The utilized databases in this study were profiled in Supplementary Table S1.All these web-based databases are accessible freely to users who retrieve valuable information, moreover, which are useful resources to employ microbiome and NP concept at the same time.In this study, we incorporated the connectivity of the GM and DPDNFs, based on public databases.

The selection of metabolites from GM and its targets
The metabolites from GM were identified by gutMGene as a collective database of GM's targets and the metabolites (http://bio-annotation.cn/gutmgene) (accessed on 3 July 2022) [20].The targets were collected by the Similarity Ensemble Approach (SEA) (https://sea.bkslab.org/)(accessed on 3 July 2022) [21] and SwissTargetPrediction (STP) (http:// www.swisstargetprediction.ch/) (accessed on 4 July 2022) [22] as cheminformatics databases to identify potential targets.Specially, we selected only overlapping targets from the two databases to achieve the rigorousness.The Venny 2.1 tool was utilized to select the overlapping targets between SEA and STP, which was visualized by Venn diagram plotter.

The collection of DPDNFs and identification of its targets
The DPDNFs were collected by literature and the drug-likeness qualities of them were evaluated by SwissADME platform based on Lipinski's rule of five [23].With exactness and rigour, the targets responded to DPDNFs were retrieved by SEA and STP.

The retrieval of intersecting targets and NAFLD-related targets
The targets identified by metabolites from GM and DPDNFsrelated targets were analysed by Venny 2.1.The targets linked directly to NAFLD were found in both DisGENet as a platform of target-disease correlation (https://www.disgenet.org/) (accessed on 5 July 2022) and Online Mendelian Inheritance in Man (OMIM) as a database of target-etiology association (accessed on 5 July 2022) (https://omim.org/).

The identification of crucial targets and its PPI networks
The overlapping targets between the intersecting targets and NAFLD-related targets were considered as crucial targets deeply associated with NAFLD, based on the GM metabolites' targets and DPDNFs targets.The crucial targets were input into STRING database version 11.5 to construct PPI network (accessed on 7 July 2022) [24], the result of which was obtained via Excel file.The Excel file was input into R Package to describe the connectivity between each target.The PPI network depicts the networks of protein interacted by biochemical responses and/or distinctive biological events in host system [25].Noticeably, we found two significant targets on PPI network analysis.

The construction of bubble chart to identify a key signalling pathway
We obtained crucial targets with combinatorial properties on the GM metabolites' targets and DPDNFs targets.On the bubble chart, the rich factor (the number of targets in the pathway/number of all targets in the total target set) was considered as a key selective factor [26].Then, a signalling pathway with the lowest rich factor was chosen as the uppermost mechanism, indicating that the mechanism might function as a suppressor to ameliorate NAFLD.

The preparation of metabolites and targets for MDT
The metabolites (or flavonoids) associated with the key signalling pathway were changed .sdfformat from PubChem into .pdbformat using Pymol, and they were changed into .pdbqtformat through Autodock tool.The three targets were linked to a key signalling pathway, ie AKT Serine/Threonine Kinase 1 (AKT1) (PDB ID: 3O96), PhosphoInositide-3-Kinase, Regulatory subunit 1 (PIK3R1) (PDB ID: 4JPS) and CF Transmembrane conductance Regulator (CFTR) (PDB ID: 6O1V) were extracted by STRING on RCSB PDB (https://www.rcsb.org/),(accessed 8 July 2022).Additionally, the PDB formats of proteins were converted into .pdbqtvia Autodock tool to perform the MDT.
Also, each metabolite formed the most stable conformer on the key three targets (AKT1, PIK3R1 and CFTR) was considered as a crucial effector.In other words, a molecule with the lowest binding energy on each target is defined as the most stable effector [30].After MDT, the metabolites with the lowest binding energy were chosen to describe conformers on targets in Pymol.

Toxicological verification of two key metabolites
During drug development, the level of toxicity is an important factor to be successful therapeutic agents.The toxicity was verified by the following assays: hERG blockers [31], Human Hepato-Toxicity (H-HT) [32], Rat Oral Acute Toxicity [33], Carcinogenicity [34], Eye Corrosion [35], Respiratory Toxicity [36].The toxicity parameters are assessed by using ADMETlab platform [37].

The networks of GM or natural resources -metabolitestargets-key signalling pathway (GNMTK)
To describe a holistic viewpoint, we constructed the network of GM or natural resources-metabolites -targets-key signalling pathway (GNMTK) via R Package.The GNMTK was depicted as nodes in the network, and their correlations were indicated as edges [38].The GNMTK network was depicted on a size plot with the number of edges connected to each node.In the network diagram, yellow circle (node) represented GM; purple circle (node) stood for metabolites; orange circle (node) indicated targets and green circle (node) described a key signalling pathway.The merged network was visualized utilizing R Package.

The identification of potential targets from GM
In the gutMGene, we identified a total of 208 metabolites from GM (Supplementary Table S2) and the metabolites were searched for its targets through SEA (1256 targets) and STP (947 targets) cheminformatics platforms (Supplementary Table S2).The number of 668 overlapping targets was identified between 1256 targets in SEA and 947 targets in STP (Figure 2(A); Supplementary Table S2).The targets were considered as respondents modulated by metabolites from GM.

The DPDNFs' mining and drug-likeness selection
The sixteen DPDNFs were selected through a literature study, and their chemical information was profiled in Table 1.The selected compounds' physicochemical properties or druglikeness qualities were assessed utilizing Lipinski's rule of five, the parameters of which were confirmed by SwissADME webbased tool.Based on Lipinski's rule of five, 14 out of 16 DPDNFs were accepted as drug-likeness molecules without violating of the criteria: (i) molecular weight 500; (ii) Hbond donors 5; (iii) H-bond acceptors 10; (iv) MLogP 4.15 1 (v) and Topological Polar Surface Area (TPSA) <140 (Å 2 ) [39,40].The violated compounds were Epigallocatechingallate and Procyanidins, which were excluded in the analysis.

The identification of crucial targets against NAFLD
The targets related to the fourteen DPDNFs were retrieved via SEA and STP databases, indicating 173 targets in SEA and 200 targets in STP (Figure 2(B); Supplementary Table S3).The 112 intersecting targets were identified by the two databases (Supplementary Table S3).The 668 overlapping targets were associated with metabolites from GM, suggesting that the 112 intersecting targets were a subset of the 668 metabolites' targets (Figure 2(C)).Coordinately, NAFLD-related targets were retrieved by DisGeNET, and OMIM online database, resulting in 1836 targets (Supplementary Table S3).Collectively, the number of 47 overlapping targets (Supplementary Table S3) was identified between NAFLDrelated targets (1836) and the 112 intersecting targets (Figure 2(D)).The obtained 47 targets were considered as significant targets in determining integrated application of GM and DPDNFs against NAFLD.

PPI network construction
The STRING database was incorporated to construct PPI network on the crucial 47 targets, constructing via R Package.The 44 out of 47 targets were interacted closely with each other, however, the three targets (CA3, ST6GAL1 and ERN1) were not correlated with one another.The 44 targets were represented with 44 nodes and 241 edges (Figure 3).In the PPI networks, the two targets with the highest degree value were AKT1 (31) and VEGFA (31) (Table 2).The two targets were considered as the most significant protein-coding gene in the entire PPI network.

Bubble chart on signalling pathways against NAFLD
The results of Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway enrichment analysis indicated that the 47 targets were associated with 30 signalling pathways (False Discovery Rate < 0.05).The 30 signalling pathways were involved in occurrence and development of NAFLD, suggesting that the implicated pathways might be significant mechanisms against NAFLD.The description of the 30 signalling pathways was profiled in Table 3.Additionally, the bubble chart represented that cAMP (cyclic Adenosine MonoPhosphate) signalling pathway with the lowest rich factor might be a key mechanism against NAFLD (Figure 4).

Toxicological evaluation of two key metabolites in silico
The two metabolites (luteolin and myricetin) was assessed through ADMETlab web-based tool.The result has been shown that the two compounds had no noticeable toxic properties (Table 5).

The networks of microbiota or natural resourcesmetabolites-targets-key signalling pathway
The GM or natural resources-metabolites-targets-key signalling pathway (GNMTK) suggested that they have strong relationships to exert the therapeutic effects.Hence, GNMTK network represented that the orchestration of DPDNFs and favourable GM might have more beneficial effects against NAFLD.All in all, both Luteolin and Myricetin might bind stably to AKT1, PIK3R1 and CFTR to dampen cAMP signalling pathway, suggesting that the effectors are bona fide agents in DPDNFs and Enterococcus sp.45, Escherichia sp.12, Escherichia sp.33 and Bacterium MRG-PMF-1 are critical GM in the treatment of NAFLD (Figure 6).

Discussion
Our study suggested that combinatorial application with natural resources and beneficial GM might favourable pharmacological efficacy on NAFLD.Most recently, a report demonstrated that NP is an efficient methodology to elucidate active molecules in Silybum marianum against NAFLD [41].Similarly, another report has been shown that Xiaochaihu decoction plays an important role to ameliorate NAFLD via NP-based analysis [42].Certainly, we assembled the GM database on natural products database to unravel the relationships between GM and DPDNFs, in the pursuit of scaled-up approach.This methodology is needed to optimize key GM and metabolites against some diseases and would provide new insight into NAFLD therapeutics.Accordingly, we employed the systemic analysis by integrating key components to be identified by data mining, thereby, luteolin (referred to as postbiotics or metabolites), myricetin (referred to as postbiotics or metabolites), AKT1 (target), CFTR (target), PIK3R1 (target), Enterococcus sp.45 (probiotics), Escherichia sp.12 (probiotics), Escherichia sp.33 (probiotics) and Bacterium MRG-PMF-1 (probiotics) were associated with cAMP signalling pathway.In parallel, we surveyed luteolin-rich and myricetin-rich natural food resources.As a result, the pepper, carrot, broccoli, celery, cabbage and parsley are luteolin-rich natural resources, which are related deeply to AKT1 and CFTR.Likewise, Enterococcus sp.45 (probiotics) can convert orientin (prebiotics) (flavone species) and cymaroside (prebiotics) (flavone species) into luteolin (postbiotics) [43,44].
An animal test demonstrated that cAMP level in NAFLD rats was elevated to as fivefold as control groups, conversely,   Lys672,Glu76,Pro3  the cAMP level had negative correlation with high density lipoprotein (HDL) cholesterol [46].Furthermore, cAMP triggers gluconeogenesis in NAFLD as well as causes insulin resistance [47][48][49].It has been suggested that inhibitors of cAMP signalling pathway might be important therapeutic agents in the treatment of NAFLD.Both luteolin and myricetin are secondary metabolites in natural resources or postbiotics in specific GM.The luteolin improves sensitivity level of insulin as an antidiabetic agent, including antioxidant and anti-inflammatory properties [50].Likewise, myricetin has not only potent stimulatory effect of insulin, but accelerates lipogenesis in rat adipocytes [51].The MK-2206 as an AKT inhibitor reached at a phase II clinical trial suppressed the progression of NAFLD [52].Uncommonly, there was no relationship between CFTR and hepatic steatosis in the development of NAFLD [53].Wortmannin as PIK3R1 (PI3K) inhibitor diminished the expression level of interleukin 1b (IL-1b) and interleukin 18 (IL-18), which leads to protecting mice from NAFLD [54].The results suggested that targets (AKT1, PIK3R1) and effectors (luteolin, myricetin) might be critical elements to alleviate NAFLD on a pharmacological viewpoint.Thus, our innovative methodology gives significant clues to clarify the nuanced relationships between GM and DPDNFs.
The top 10 signalling pathways related to NAFLD were briefly discussed as follows, the selective criteria of which was based on false discovery rate (FDR< 0.05).cAMP signalling pathway: Metformin with AMPK (AMP-activated protein Kinase)-dependent mechanisms dampens mitochondrial respiratory system, activating AMPK and lowering cAMP and thereby enhancing liver enzymes and fat content in NAFLD [55].Interestingly, an animal test demonstrated that metformin as cAMP suppressor is associated with diversity of GM and reduced the production of bacterial endotoxin in NAFLD [56].It elicits that cAMP blocker might have synergistic effects via positive pharmacological mechanism and favourable GM composition on NAFLD.
Cyclic Guanosine MonoPhosphate Protein Kinase G (cGMP-PKG) signalling pathway: The Gene Expression Omnibus (GEO) database analysis shows that cyclic Guanosine MonoPhosphate Protein Kinase G (cGMP-PKG) signalling pathway is related directly to insulin resistance and gives rise to steatosis via aggravating cytotoxic cascades of NAFLD [57].JAnus Kinase-Signal Transducer and Activator of Transcription (JAK-STAT) signalling pathway: An animal experiment demonstrated that the deletion of JAK2 induces fatty liver, thereby elevating liver triglyceride as twenty times as normal group [58].Neurotrophin signalling pathway: A logistic regression study demonstrated that brain-derived neurotrophic factor (BDNF) has a negative independent factor with 0.041 (Odds ratio) in NAFLD patients, by comparison with control groups [59].T cell receptor signalling pathway: The recognition of antigen on T cell is an initial step to maintain the homeostasis of the GM, affecting liver immune system in gut-liver axis [60].NF-kappa B (NFKB) signalling pathway: O. indicum seed extract (OISE) with anti-inflammatory effects inhibits the activity of NFKB, leading to the development non-alcoholic steatohepatitis (NASH) [61].Chemokine signalling pathway: The concentrations of the C-C Motif Chemokine Ligand 2 (CCL2) and CXC Motif Ligand 9 (CXCL9) in NAFLD group were significantly greater than the control group [62].Oxytocin signalling pathway: Oxytocin is an accelerator to promote liver regeneration via autophagy, liver metabolic homeostasis [63,64].Insulin signalling pathway: Insulin resistance a key risk factor to aggravate NAFLD, insulin sensitizers (thiazolidinediones) taken to ameliorate NAFLD might be  critical therapeutic agents [65,66].Thyroid hormone signalling pathway: The thyroid hormone can reduce the concentration of blood lipid and attenuate the progression of NAFLD [67].
It implies that thyroid hormone can alleviate the hyperlipidaemia and NAFLD.
To sum up, we believe that the integrated analysis of prebiotics, probiotics, postbiotics and targets is an essential strategy to decode the complex microbiome system.In parallel, DPDNFs and GM play important role against NAFLD.Collectively, we indicate relationships of the key GM or natural resources, metabolites, targets and a key signalling pathway in the treatment of NAFLD.From a holistic perspective, we suggest the GM and DPDNFs components might exert the combinatorial effects on NAFLD.The key findings of this study were displayed in Figure 7.

The pros and cons of this study
With the aid of NP, the connectivity of GM and DPDNFs is a strategy to decrypt novel uses for existing, validated, or known therapeutic ingredients that are outside extent of its real clinical features in microbiome research.This study is based on the scientific evidence that metabolites and DPDNFs from GM affect some key targets against NAFLD.At present, thanks to the development of bioinformatics, cheminformatics and microbiome databases, a great deal of biodata has been achieved.This enables to analyse the promising natural resources, GM, mechanisms, targets and metabolites against human diseases, including NAFLD.The aim of this applicability is to represent the utilization of these integrated pharmacological method to predict the relationships and to discover for new effectors in certain diseases.At this stage, the devised platform allows new data to input in the stylized format.Nevertheless, there are some limitations that our suggests are theoretical points and dependent on data-driven study.It is a challenge that we should overcome in further study.

Conclusion
Overall, this study indicated key probiotics (Enterococcus sp.45, Escherichia sp.12, Escherichia sp.33 and Bacterium MRG-PMF-1) or natural resources, postbiotics (luteolin and myricetin) and targets (AKT1, PIK3R1 and CFTR) might block cAMP signalling pathway against NAFLD.In the scenario, we shed light on the significance of both luteolin and myricetin as metabolites from GM (Enterococcus sp.45, Escherichia sp.12, Escherichia sp.33 and Bacterium MRG-PMF-1) and DPDNFs in the treatment of NAFLD.To conclude, we suggest that combinatorial application of favourable GM and DPDNFs might make positive efficacy in the alleviation of NAFLD.

Figure 1 .
Figure 1.The stepwise workflow of this study.

Figure 2 .
Figure 2. (A) The number of 668 overlapping targets between SEA (1256 targets) and STP (947 targets).(B) The number of 112 overlapping targets related to 14 flavonoids between SEA (173 targets) and STP (200 targets).(C) The number of 112 overlapping targets between (A) and (B).(D) The number of 47 crucial targets against NAFLD.

Figure 4 .
Figure 4. Bubble chart of 30 signalling pathways associated with progression of NAFLD.

Table 1 .
The physicochemical properties of the number of 16 dietary plant derived flavonoids.Bold font: Epigallocatechin-gallate and Procyanidins are excluded due to violation of Lipinski's rule.

Table 2 .
The degree of value in PPI network.

Table 3 .
The description of 30 signalling pathways against NAFLD.

Table 4 .
The binding energy and interactions of key metabolites and targets.

Table 5 .
The verification of toxic parameters on two key metabolites.