New insight into gut microbiota-derived metabolites to enhance liver regeneration via network pharmacology study

Abstract

We intended to identify favourable metabolite(s) and pharmacological mechanism(s) of gut microbiota (GM) for liver regeneration (LR) through network pharmacology. We utilized the gutMGene database to obtain metabolites of GM, and targets associated with metabolites as well as LR-related targets were identified using public databases. Furthermore, we performed a molecular docking assay on the active metabolite(s) and target(s) to verify the network pharmacological concept. We mined a total of 208 metabolites in the gutMGene database and selected 668 targets from the SEA (1,256 targets) and STP (947 targets) databases. Finally, 13 targets were identified between 61 targets and the gutMGene database (243 targets). Protein–protein interaction network analysis showed that AKT1 is a hub target correlated with 12 additional targets. In this study, we describe the potential microbe from the microbiota (E. coli), chemokine signalling pathway, AKT1 and myricetin that accelerate LR, providing scientific evidence for further clinical trials.

Graphical Abstract

Keywords:

• liver regeneration
• network pharmacology
• molecular docking assay
• chemokine signaling pathway
• AKT1
• myricetin

Introduction

Liver regeneration (LR) is a critical process to reorganize tissue and is the most significant process associated with relapse after liver injury [1]. The risk attributed to loss of hepatocytes results in liver fibrosis, liver cirrhosis, liver failure and even liver cancer [2,3]. These stages are causative factors that increase the mortality rate, which leads to the diverse pathological spectrum ranging from atherosclerosis and diabetes mellitus [4]. This finding suggests that regeneration in liver injury is the uppermost curative process to dampen the development of other unexpected complications. Additionally, a report showed that the main markers of LR include hepatocyte growth factor (HGF), interleukin-6 (IL-6) and vascular endothelial growth factor (VEGF), which are crucial components influencing on hepatocyte proliferation [5].

These factors are representative endogenous components in liver injury as well as main elements in the recovery of function, suggesting that the endogenous targets might represent key regulators that accelerate LR.

In probiotics, a report showed that the diversity of gut microbiota (GM) produces dynamic effects on its recovery rate during LR after surgical operation of liver as well as GM regulation under antibiotics or probiotics [6]. In particular, optimal colonization of the GM is essential to maintain healthy liver conditions, whereas gut dysbiosis leads to chronic liver diseases [7]. In detail, gut microbial dysbiosis has negative effects on hepatic functionality; therefore, metabolites detected in serum have been altered due to biased gut microbial reserves [8]. Dysbacteriosis results in a lack of natural killer (NK) T cells, which are vulnerable to acute liver injury [9]. An animal study demonstrated that a reduced percentage of Firmicutes to Bacteroidetes was correlated with improved cell proliferation, and significant changes in observational results were noted 12 h after partial hepatectomy (PH) [10].

In prebiotics, the supply of growth supplements for GM to drive different composites or metabolic changes can be an alleviative strategy. For instance, fructose well-known as prebiotics protect hepatocytes from apoptosis induced by tumour necrosis factor (TNF) via activating JNK pathway [11]. Lactulose accelerates H2 production intrinsically and promotes LR after 70% partial PH in rats [12]. Presumably, it elicits that key prebiotics digested by GM converted into key metabolites which can exert biological activity.

In metabolites, indoles and its derivatives as the representative metabolites produced by GM intaking tryptophan in the gut play a significant role in enhancing LR through stabilizing the gut mucosal epithelial barrier [13]. More recently, endogenous microbiota-derived components, such as proteins and compounds produced from living microbiota communities, have exerted favourable effects on hosts [14]. Metabolites derived from GM can reach the liver via the portal vein, which can ameliorate liver diseases related to immunity [15].

The metabolites produced by GM and exogenous substrate (ES) affect via gut-liver axis, which might manage the expression of cytokines in intestinal epithelial cells and hepatocytes [13]. Accordingly, they play important roles to promote LR [16,17]. Furthermore, many liver diseases have been implicated in inhibiting LR [18]. The metabolites (postbiotics) and immune system are significant factors influencing on LR via GM. The diversity of GM is a critical element to control above the two factors (metabolites and prebiotics) and thus acts on LR.

Hence, the multiple components as in GM (probiotics), ES (prebiotics) and metabolites (postbiotics) can be utilized to accelerate the rate of LR. However, the definite mechanisms of these inter-networking remain unclear. In the context of incomplete project, we employed the network pharmacology to unravel the conundrum in the complex microbiome for LR.

Network pharmacology is a systemic perspective to elucidate the effector mechanism of targets and ligands (metabolites) in complex biological pathways [19]. Recently, a report showed that network pharmacology was utilized to uncover the effects of the infectious GM against ulcerative colitis [20]. A study suggested that expounding the promising pharmacological mechanisms connected with microbiota diversity affects the host pathological response directly or indirectly [21]. Therefore, the use of an integrated concept to decode the complex scenario in the microbiome, as such relationships between GM, signalling pathways, targets and metabolites, is imperative.

In this study, network pharmacology was utilized to reveal key components for improving LR and to determine how the microbiota-signalling pathways-targets-metabolites (MSTM) relationship induces positive effects on LR. First, we utilized the gutMGene database to retrieve targets and metabolites related to the microbiota. Then, metabolite-associated targets and LR-related targets were identified by two public databases. Second, the overlapping targets between the two public databases were obtained via a Venn diagram, in which the intersecting targets were determined between LR-related targets. The final overlapping targets were identified by comparison with the intersecting targets and reported targets in the gutMGene database. Third, the final overlapping targets were chosen to construct a protein–protein interaction (PPI) network and a bubble plot to identify the key signalling pathways in LR. After identifying the key signalling pathways, we performed a molecular docking assay (MDA) to verify the potential metabolites on key targets. Finally, we integrated the relationships of MSTM networks to reveal how the four components interact with one another in LR. Therefore, this analysis might provide critical hints for further experimental studies and the utilization of metabolites for LR. The workflow of this study is displayed in Figure 1(A). Additionally, pioneering clinical intervention of various liver diseases directing to facilitate LR associated with metabolism by adjusting to GM is a promising therapeutic strategy in the future research area.

Figure 1. (A) A workflow of this study. (B) The number of 668 overlapping targets between SEA (1256) and STP (947). (C) The number of 61 targets between 668 targets and liver regeneration-related targets (365). (D) The identification of 13 core targets between 61 targets and Gut human targets (223 targets) from gutMGene database. SEA: Similarity Ensemble Approach; STP: SwissTargetPrediction; PPI: protein–protein interaction; GO: gene ontology; MSTM: microbiota-signalling pathways-targets-metabolite.

Materials and methods

Database platforms and data analysis associated with network pharmacology

Biologically significant databases that give a great amount of data associated with the relationship between compounds and proteins enable the researcher to employ network pharmacology as a powerful way for drug discovery (Supplementary Table S1). All these web-based databases are accessible freely to users who pursue to retrieve valuable information, moreover, which is functional methodology to be applicable to microbiome and network pharmacology principle at once. In this study, we tried new approach to explore the complex microbiome and its network with phytochemicals, via public databases.

Table 1. The degree value of core targets.

No.	Target	Degree of value
1	AKT1	12
2	EGFR	11
3	IL6	11
4	PPARG	11
5	CASP3	10
6	MAPK8	10
7	HDAC1	9
8	MAPK14	9
9	NFE2L2	9
10	TLR4	9
11	MAPK1	7
12	HDAC3	4
13	NR1H4	4

The mining of metabolites and targets from the gut microbiota

The gutMGene v1.0 (http://bioannotation.cn/gutmgene/) (accessed on 15 June 2022) was employed to retrieve the metabolites and targets for further analysis. The metabolites converted by GM were identified by a sub-folder of downloads section (http://bio-annotation.cn/gutmgene/public/res/gutMGene-human.xlsx) in gutMGene v1.0, which was considered as key metabolites reported to date. The Simplified Molecular Input Line Entry System (SMILES) formats of metabolites were adopted by PubChem (https://pubchem.ncbi.nlm.nih.gov/) (accessed on 15 June 2022).

The acquisition of crucial targets for regeneration

Using the SMILES format, the related targets on metabolites were browsed by both the Similarity Ensemble Approach (SEA) as a cheminformatics database to identify potential targets from repository of compounds (http://sea.bkslab.org/) (Accessed on 16 June 2022) and SwissTargetPrediction (STP) as a target-selecting database platform from reservoir of ligands (http://www.swisstargetprediction.ch/) (Accessed on 16 June 2022) on ‘Homo Sapiens’ mode. Specifically, Dr Shoichet’s group assembled the ‘Similarity Ensemble Approach (SEA)’ to identify the ligand-target affinity and eventually uncover its interaction [22]. In addition, STP is a web-based searching tool developed by SIB (Swiss Institute of Bioinformatics), accessing approved 376,342 compounds and 3068 targets since 2014 [23].

We obtained the intersecting targets between the overlapping targets selected by the SEA and STP databases and LR-related targets obtained by DisGeNET as a platform of target-disease correlation (https://www.disgenet.org/) (accessed on 16 June 2022) and OMIM as a database of target-etiology association (https://www.omim.org/) (Accessed on 16 June 2022) on a Venn diagram. Crucial targets between the intersecting targets and targets reported in the gutMGene database were obtained. Thus, we considered these targets to be the most significant targets for LR.

The protein–protein interaction network

The PPI network was identified by STRING (https://string-db.org/) analysis (accessed on 20 June 2022), the network of which was imported into R Package software. The middlemost target with the greatest degree value in the PPI network was considered the uppermost target.

Gene ontology and Kyoto Encyclopaedia of Genes and Genomes analysis

Gene ontology (GO) molecular function (MF) analysis was mainly utilized to represent the functions of targets, including biological processes (BPs) and cellular components (CCs). This information was displayed on a bubble plot utilizing the R Package. The signalling pathways on the KEGG (Kyoto Encyclopaedia of Genes and Genomes) database were analysed based on the crucial targets, which allows the identification of the signalling pathways related to metabolites from the GM.

Metabolite-target preparation for molecular docking assay

The metabolites linked to the key targets were translated into .sdf files from PubChem to the .pdb format utilizing PyMOL. Then, files in the .pdbqt format were obtained through AutoDock software and used to perform MDT. The main targets were retrieved from the STRING database via RCSB (https://www.rcsb.org/) (accessed on 18 June 2022). Files in the .pdb format extracted from RCSB were converted into the .pdbqt format for incorporation into AutoDock −1.5.6 for MDT.

Molecular docking assay

AutoDockTools-1.5.6 was adopted to generate the target centre and the grid box dimensions of the active site. With the default 10 different poses of the metabolites, AutoDock4 software was set up with 4 energy ranges and 8 exhaustiveness to identify the greatest score. For docking, a cubic box was created with solid dimensions based on the active site, which was set to 40 Å × 40 Å × 40 Å. The tangible centres to bind the metabolites on the two main targets were X: 6.313, Y: −7.926, Z: 17.198 and X: −20.844, Y: 40.173, Z: 21.018. The hydrophobic and hydrophilic residues of the complex were obtained by LigPlot+ 2.2 (https://www.ebi.ac.uk/thornton-srv/software/LigPlus/download.html) (Accessed on 19 June 2022)) [24]. The cut-off value of MDA was −6.0 kcal/mol [25], and the key ligand with the lowest docking score on a specific target was considered the uppermost metabolite for LR.

Drug-likeness properties and toxicological evaluation

The drug-likeness properties of the two metabolites were evaluated using the SwissADME database [26]. Conventionally, the metabolite is prone to exhibit hydrophilic properties for easy excretion from the body, indicating that these metabolites have low bioavailability. Thus, it is necessary to profile their physicochemical characteristics using an in silico assay.

In addition, the level of toxicity is the primary factor that fails during the development of new agents. Toxicity is assessed using the following agents and assays: hERG blockers [27], Human Hepato-Toxicity (H-HT) [28], Ames mutagenicity [29], Skin sensitization [30], Lethal Dose 50 (LD50) of acute toxicity [31] and Drug Induced Liver Injury (DILI) [32]. The toxicity findings are confirmed utilizing the ADMETlab platform [33].

Construction of the microbiota-signalling pathways-targets-metabolite network

The crucial targets were input into the STRING database to examine the proteins under the Homo sapiens setting. We subsequently identified the main signalling pathways for LR. Then, microbiota and metabolites related directly to targets were selected via the gutMgene database. MSTM networks were visualized using R Package. The microbiota, signalling pathways, targets and metabolites are described as nodes in the network, and the relationships between the four components are depicted as edges [34]. The MSTM network was constructed on a size plot based on the number of edges for each node. In the network plot, purple circles (nodes) represent the GM; red circles (nodes) represent the signalling pathways; orange circles (nodes) describe the targets; and pink circles (nodes) indicate the metabolites. The size of the purple circles describes the number of connectivity to signalling pathways, metabolites and targets; the size of the red circles represents the number of interactions with GM; the size of orange circles represents the number of associations with signalling pathways; and the size of pink circles depicts the number of correlations with targets. The integrated network was incorporated using R Package.

Results

The promising targets and metabolites from gut microbiota

A total of 208 metabolites obtained by a sub-folder in gutMGene v1.0 converted by the GM were identified through the gutMGene database (Supplementary Table S2).

Table 2. The number of 36 signalling pathways and targets related to LR.

KEGG ID & description	Targets	False discovery rate
hsa04933: AGE-RAGE signalling pathway in diabetic complications	AKT1, CASP3, MAPK1, MAPK8, MAPK14, IL6	0.00000000197
hsa04664: Fc epsilon RI signalling pathway	AKT1, MAPK1, MAPK8, MAPK14	0.00000110000
hsa04620: Toll-like receptor signalling pathway	AKT1, IL6, TLR4, MAPK1, MAPK8, MAPK14	0.00000000197
hsa04917: Prolactin signalling pathway	AKT1, MAPK1, MAPK8, MAPK14	0.00000122000
hsa04657: IL-17 signalling pathway	CASP3, IL6, MAPK1, MAPK8, MAPK14	0.00000005640
hsa04668: TNF signalling pathway	AKT1, CASP3, IL6, MAPK1, MAPK8, MAPK14	0.00000000235
hsa04370: VEGF signalling pathway	AKT1, MAPK1, MAPK14	0.00004220000
hsa04625: C-type lectin receptor signalling pathway	AKT1, IL6, MAPK1, MAPK8, MAPK14	0.00000007720
hsa04012: ErbB signalling pathway	AKT1, MAPK1, MAPK8, EGFR	0.00000229000
hsa04068: FoxO signalling pathway	AKT1, MAPK1, MAPK8, MAPK14, IL6, EGFR	0.00000000405
hsa04066: HIF-1 signalling pathway	AKT1, MAPK1, IL6, EGFR, TLR4	0.00000008920
hsa04912: GnRH signalling pathway	MAPK1, MAPK8, MAPK14, EGFR	0.00000292000
hsa04660: T cell receptor signalling pathway	AKT1, MAPK1, MAPK8, MAPK14	0.00000441000
hsa04926: Relaxin signalling pathway	AKT1, MAPK1, MAPK8, MAPK14, EGFR	0.00000020600
hsa04722: Neurotrophin signalling pathway	AKT1, MAPK1, MAPK8, MAPK14	0.00000656000
hsa04071: Sphingolipid signalling pathway	AKT1, MAPK1, MAPK8, MAPK14	0.00000686000
hsa04919: Thyroid hormone signalling pathway	AKT1, MAPK1, HDAC1, HDAC3	0.00000726000
hsa04920: Adipocytokine signalling pathway	AKT1, MAPK8	0.00320000000
hsa04621: NOD-like receptor signalling pathway	MAPK1, MAPK8, MAPK14, IL6, TLR4	0.00000079400
hsa04622: RIG-I-like receptor signalling pathway	MAPK8, MAPK14	0.00320000000
hsa04662: B cell receptor signalling pathway	AKT1, MAPK1	0.00390000000
hsa04915: Oestrogen signalling pathway	AKT1, MAPK1, EGFR	0.00037000000
hsa04910: Insulin signalling pathway	AKT1, MAPK1, MAPK8	0.00037000000
hsa04010: MAPK signalling pathway	AKT1, MAPK1, MAPK8, MAPK14, CASP3, EGFR	0.00000020600
hsa04072: Phospholipase D signalling pathway	AKT1, MAPK1, EGFR	0.00045000000
hsa04015: Rap1 signalling pathway	AKT1, MAPK1, MAPK14, EGFR	0.00004220000
hsa04630: JAK-STAT signalling pathway	AKT1, IL6, EGFR	0.00057000000
hsa04014: Ras signalling pathway	AKT1, MAPK1, MAPK8, EGFR	0.00006310000
hsa04152: AMPK signalling pathway	AKT1, PPARG	0.00820000000
hsa04371: Apelin signalling pathway	AKT1, MAPK1	0.00950000000
hsa04024: cAMP signalling pathway	AKT1, MAPK1, MAPK8	0.00120000000
hsa04151: PI3K-Akt signalling pathway	AKT1, MAPK1, IL6, EGFR, TLR4	0.00001430000
hsa04921: Oxytocin signalling pathway	MAPK1, EGFR	0.01190000000
hsa04150: mTOR signalling pathway	AKT1, MAPK1	0.01200000000
hsa04022: cGMP-PKG signalling pathway	AKT1, MAPK1	0.01350000000
hsa04062: Chemokine signalling pathway	AKT1, MAPK1	0.01740000000

The targets related to the collected metabolites were identified using the SEA (1256 targets) and STP (947 targets) databases (Figure 1(B)) (Supplementary Table S3), and the two databases revealed 668 overlapping targets (Figure 1(C)) (Supplementary Table S3). In total, 61 targets overlapped between 668 targets and 365 LR-related targets (Supplementary Table S3). Finally, 13 core targets overlapping between 61 targets and 223 targets from the gutMGene database were selected to analyse the PPI network (Figure 1(D)) (Supplementary Table S3).

Table 3. The evaluation of drug-likeness properties on two key metabolites.

No.	Compounds	Lipinski Rules				Lipinski’s violations	Bioavailability score	TPSA (Å^2)
		MW	HBA	HBD	MLog P
		< 500	< 10	≤ 5	≤ 4.15	≤1	> 0.1	<140
1	Myricetin	318.24	8	6	−1.08	1	0.55	151.59
2	Compound K	653.80	8	4	1.72	1	0.55	131.17

Note: MW: molecular weight; HBA: hydrogen bond acceptor; HBD: hydrogen bond donor; MLog P: Moriguchi octanol-water partition coefficient; TPSA: topological polar surface area.

Protein–protein interaction network

The 13 core targets were incorporated into PPI network analysis, highlighting the most essential proteins in the network. The PPI network consisted of 13 nodes and 58 edges. AKT1 had the highest degree of value (12), showing that AKT1 might be a significant target to accelerate LR (Figure 2(A)) (Table 1). Up to date, many of the research indicated that AKT1 might be a hub target to relieve LR. Accordingly, our research lines up with the implication demonstrated by the survey. The targets with the next highest degrees of value included EGFR (11), IL-6 (11), PPARG (11), CASP3 (10) and MAPK8 (10).

Figure 2. (A) The protein–protein interaction (PPI) networks (13 nodes and 58 edges). (B) Bubble plot of 36 signalling pathways related to liver regeneration (LR). (C) Gene ontology (GO) enrichment analysis.

Gene ontology and signalling pathway enrichment analysis

To further facilitate the therapeutic application of the metabolites from the GM in the pharmacological concept of LR, the 13 core targets were assessed by signalling pathway enrichment (Figure 2(B)) and GO enrichment analysis (Figure 2(C)) using the STRING bioinformatics database. The 13 core targets were directly associated with 36 signalling pathways in LR (Table 2), and the circle size represents the number of targets connected to the pathway. As an observational result, the chemokine signalling pathway with the lowest rich factor (gene ratio) represents a potent inhibitive mechanism for LR. In KEGG pathway, the lowest rich factor is, the more inhibitive signalling pathway is [35]. Additionally, the GO enrichment analysis comprised three modules: MF, BP and CC.

Molecular docking assay analysis

The binding affinities of a total of 20 metabolites and two targets (AKT1 and MAPK1) related to the chemokine signalling pathway at the cocrystalized complex were assessed. Regarding AKT1 (PDB ID: 3O96), myricetin exhibited the highest binding energy of −8.6 kcal/mol followed by kampferol (−8.5 kcal), apigenin (−8.4 kcal/mol) and diosmetin (−8.2 kcal/mol). Compared with three positive controls (miransertib: −7.7 kcal/mol; uprosertib: −7.7 kcal/mol; and afuresertib: −7.6 kcal/mol) (Supplementary Table S4), myricetin could bind stably to the active sites of AKT1. The hydrogen bond interactions of Asn199, Trp80 and Asn53 and the hydrophobic interactions of Gln59, Leu78, Ala58, Ser56, Gln79 and Val201 facilitated myricetin-AKT1 (PDB ID: 3O96) complex formation (Figure 3(A)).

Figure 3. (A) The docking result of myricetin-AKT1. (B) The docking result of Compound K-MAPK1. (C) Microbiota-signalling pathways-targets-metabolite (MSTM) networks (144 nodes, 499 edges).

Table 4. The evaluation of toxicity on two key metabolites.

Parameter	Compound K	Myricetin
1) hERG blockers	Non-blockers	Non-blockers
2) H-HT	Negative	Negative
3) Ames Mutagenicity	Negative	Negative
4) SkinSen (skin sensitization, (r)LLNA)	Negative	Negative
5) LD50	3.403 mg/kg	2.691 mg/kg
6) DILI	Negative	Negative

Notes: 1) Human ether-a-go-go-related gene (hERG): encoding potassium ion (K+) channel; 2) Human Hepatotoxicity (H-HT): Chemical-driven liver damage; 3) Ames Mutagenicity: Salmonella mutagenicity assay; 4) Local Lymph Node Assay (LLNA or (r) LLNA): A murine model to verify the skin sensitization risk of compounds; 5) Lethal Dose for 50% kill (LD50): The amount of a toxic chemical to kill 50% population; 6) Drug Induced Liver Injury (DILI): A liver injury induced from various drugs.

Notably, with a strong binding affinity of −14.0 kcal/mol, colibactin harboured a significant binding pocket for MAPK1 (PDB ID: 5IZ5). However, previous studies have demonstrated that colibactin is a potent genotoxin that is secreted by some bacteria, including the human gut microbiome [36,37]. Thus, we did not consider it as a potential agent in this study. Instead, Compound K, with the second-best binding affinity of −8.4 kcal/mol, was regarded as a promising effector for LR. Moreover, the binding stability of Compound K-MAPK1 (PDB ID: 5IZ5) had a higher affinity than that of the positive controls DH-6-OH-rotenone (−7.8 kcal/mol) and Cu-CPT22 (−6.0 kcal/mol) (Supplementary Table S5). The hydrogen bond interactions of Gln306, Ser51, His299, Thr295 and Glu303 and the hydrophobic interactions of Ile39, Ser47, Glu305, Ala44, Glu37, Lys300, Lys292 and Phe48 facilitated Compound K-MAPK1 (PDB ID: 5IZ5) complex formation (Figure 3(B)).

Drug-likeness and toxicity evaluation

The two metabolites (myricetin and Compound K) were verified via in silico methodologies, including assessment of drug-likeness properties, such as Lipinski’s rule and Topological Polar Surface Area (TPSA) (Cut-off value: <140 Ų) (Table 3). Hence, it is suggested that the two metabolites from the GM can be administered orally to facilitate LR. The toxicity evaluation of Compound K and myricetin was performed using the ADMETlab online tool (Table 4). The result shows that the two metabolites did not exhibit obvious hazardous properties that would serve as hurdles if these metabolites were utilized as drugs.

Analysis of MSTM networks

The MSTM network represented the relationships between microbiota, signalling pathways, targets and metabolites, consisting of 144 nodes (25 microbiota, 36 signalling pathways, 11 targets and 72 metabolites) and 499 edges (Figure 3(C)). The purple circles represent the GM, the red circles indicate the signalling pathways, the orange circles represent the targets, and the pink circles describe the metabolites. The size of each circle indicates the number of relationships between the two. The analysis was performed using R Package. We found that Escherichia coli is the uppermost microbiota with 271 degrees of value, and the AGE-RAGE signalling pathway in diabetic complications, Toll-like receptor signalling pathway, TNF signalling pathway, HIF-1 signalling pathway, and PI3K-Akt signalling pathway were identified as significant mechanisms involved in LR. We also observed that the most significant targets are MAPK1 (32) and AKT1 (31), which exhibit higher degrees of value than other targets. In parallel, phenylacetylglutamine is the most notable metabolite due to the highest degree of value for LR. The suggested 4 elements represent underlying hallmarks of LR, indicating that these factors might orchestrate favourable effects on LR.

Discussion

Our study provides significant perspectives into systemic relationships between the microbiome and LR, which is an important and robust methodology to clarify the function of potential bioactive metabolites at the integrated level. Additionally, our study demonstrated that the 12 metabolites related to a key target (AKT1) were classified into 5 categories: flavonoids (6 compounds), coumarins (2 compounds), indoles (2 compounds), steroids (1 compound) and carboxylic acids (1 compound).

Flavonoids represented 50% of the 12 metabolites, suggesting that flavonoids are significant compounds that exert more pharmacological effects in LR compared with other metabolites. According to the physicochemical properties of drug-likeness, the 90th percentile of agents accepted by Lipinski’s rule are entered into phase II [38]. Additionally, myricetin bound stably to AKT1 and is a flavonoid compound that inhibits the PI3K/AKT signalling pathway [39]. Notably, myricitrin can be converted into myricetin via Escherichia coli sp. 12 and Escherichia coli sp. 33 [40]. The result is consistent with our observational result, indicating that Escherichia sp. could positively affect LR after liver injury. In parallel, the myricetin dampens the activation of hepatic stellate cell (HSC) and alleviates liver fibrosis induced by carbon tetrachloride (CCl₄) as well as diminishes TNF-α expression [41]. Besides, proinflammatory cytokines as in TNF-α and IL-6 expression level in hepatocytes were dramatically reduced (p value < .001) by myricetin [42]. The hepatectomy is a mainstay for treatment of the hepatocellular carcinoma (HCC); however, some subjects are exposed to critical post-hepatectomy complications such as subphrenic infection and urinary tract infection related to inflammatory responses [43,44]. An animal test demonstrated the pharmacological efficacy of flavonoid silymarin, suggesting that silymarin promoted LR by activating cell cycle in PH liver [45]. It implies that flavonoids might be potential agents to accelerate LR after PH.

The Lycium chinense combined to myricetin has favourable efficacy on antioxidant activity and the regenerative ratio of residue liver tissue after 70% PH in rats [46]. A report showed that human hepatocytes grow gradually its size for 3, 6 and 10 weeks after transplanting of myricetin-treated hepatocytes [47]. Collectively, myricetin might be an effective agent to relieve inflammation and to accelerate cell proliferation during LR.

The metabolites from the GM had no connectivity to MAPK1. Compound K is a potential metabolite derived from the microbiota. Recently, a report demonstrated that Compound K has potent protective effect on high fat diet (HFD)-induced hepatocyte injury [48]. Still, therapeutic value of Compound K is under consideration including LR.

Reportedly, both myricetin and Compound K enhance wound healing at the cellular level, demonstrating that they have potent anti-inflammatory efficacy [39,49,50]. Other reports have shown that AKT1 and MAPK inhibitors play a crucial role in ameliorating specific injury correlated with inflammasomes in cardiomyocytes [51,52]. In particular, the relative low levels of CXC chemokine are linked to LR and repair, whereas its overexpression levels were related to hepatotoxicity [53–56]. It can thus be hypothesized that the two metabolites (myricetin and Compound K) from GM can dampen inflammation with synergistic effects on each target by inactivating the chemokine signalling pathway.

GO enrichment analysis indicated that metabolites from GM mainly play a role in MAP kinase activity and NAD-dependent histone deacetylase activity (H3-K14 specific) in the MF category; host cell nucleus and spindle in the CC category; and response to ultraviolet-A (UV-A) and cellular response to oxidized low-density lipoprotein particle stimulus in the BP category.

The associations of the top 10 pathways with LR were concisely described as follows, and these findings are based on rich factors. Activation of the AGE-RAGE signalling pathway can enhance intestinal permeability. Thus, lipopolysaccharides (LPS) can cross the intestinal epithelial barrier [57]. This finding implies that the AGE-RAGE signalling pathway induces systemic inflammation and exacerbates liver injury. The Fc epsilon RI signalling pathway is activated by immunoglobulin E (IgE), which has negative effects on LR [57,58]. The Toll-like receptor signalling pathway has two functions. On one hand, the pathway results in favourable effects that improve LR. On the other hand, the pathway leads to delayed LR [59]. From this perspective, studies on TLR in LR might reveal an important mechanism for developing new therapeutic strategies. Similarly, the prolactin signalling pathway is also a double-edged sword with respect to cellular inflammation under some conditions [60].

Treatment with IL-17 promoted proliferation in a liver progenitor cell (LPC) line through the IL-17 signalling pathway [61]. It has been suggested that IL-17 might represent a potential agent for LR via further studies. TNF is a considered trigger for the development of LR, which simultaneously induces an inflammatory cascade via the TNF signalling pathway [62,63]. In addition, an animal test demonstrated that TNF does not promote LR [64]. Still, the exact function of TNF was not exposed clearly to LR. An animal test showed that VEGF plays a significant role in healing injured tissue and enhancing regenerative capacity after PH [65].

A previous report indicated that the deletion of the C-type lectin receptor has detrimental effects on LR after hepatectomy [66]. It can be postulated that the activation of the C-type lectin receptor might be attributed to LR. The ErbB signalling pathway exhibits different expression patterns between foetal and adult hepatocytes but remains poorly understood [50]. FoxO3 suppresses LR by upregulating Nox4 and downregulating Nr4a1 in the Forkhead Box Transcription Factors (FoxO) signalling pathway [67]. This finding indicated that FoxO3 regulation might represent a crucial target against liver damage and for LR. The systemic network is a method used to predict the biological relationships and potential components to reveal the metabolites from complex microbiome interactions [68].

Collectively, we identify key microbiota, mechanisms and targets core metabolites, and our results suggest that these components play important roles in LR. Based on mechanistic insight into network pharmacology, this study provides an integrated methodology to elucidate the relationships of MSTM networks for LR. Despite limited data on the complex microbiome, the therapeutic mechanisms of metabolites from GM were identified by utilizing multiple data. In subsequent analyses, the network models for microbiome analysis provide scientific-based evidence for further clinical trials.

Conclusion

In summary, this study demonstrated promising effectors (AKT1 and MAPK1) in the treatment of LR using a network pharmacology approach. We revealed that Escherichia sp., AKT1 and myricetin might play significant roles in LR by dampening the chemokine signalling pathway. In addition, our MDA also showed that Compound K can bind stably to MAPK1, positively affecting LR. These results suggest a significant point for further investigation. However, this study has some limitations, and clinical trials should be analysed to further verify our findings.

Author contributions

Ki-Tae Suk, Dong Joon Kim and Ki-Kwang Oh had substantial contributions to the conception or design of the work and interpretation of data for the work. Ki-Tae Suk, Ki-Kwang Oh and Ickwon Choi were involved in the acquisition, analysis or interpretation of data for the work. Ki-Tae Suk, Ki-Kwang Oh, Haripriya Gupta, Raja Ganesan, Satya Priya Sharma, Sung-Min Won and Jin-Ju Jeong drafted the work and revised it critically for important intellectual content and finally approved the version to be published. Su-Been Lee, Min-Gi Cha, Goo-Hyun Kwon, Min-Kyo Jeong, Byeong-Hyun Min, Ji-Ye Hyun, Jung-A Eom, Hee-Jin Park, Sang-Jun Yoon and Mi-Ran Choi agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Acknowledgments

This study has been implemented with the support of the Hallym University Research Fund, the Basic Science Research Program through the National Research Foundation of Korea, funded by the Ministry of Education, Science and Technology.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

All data generated or analysed during this study are included in this published article (and its Supplementary Information files).

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This research was supported by the Hallym University Research Fund, the Basic Science Research Program through the National Research Foundation of Korea, funded by the Ministry of Education, Science and Technology (NRF-2020R1I1A3073530 and NRF-2020R1A6A1A03043026).