Lactobacillus lactis and Pediococcus pentosaceus-driven reprogramming of gut microbiome and metabolome ameliorates the progression of non-alcoholic fatty liver disease

Abstract

Background: Although microbioa-based therapies have shown putative effects on the treatment of non-alcoholic fatty liver disease (NAFLD), it is not clear how microbiota-derived metabolites contribute to the prevention of NAFLD. We explored the metabolomic signature of Lactobacillus lactis and Pediococcus pentosaceus in NAFLD mice and its association in NAFLD patients.

Methods: We used Western diet-induced NAFLD mice, and L. lactis and P. pentosaceus were administered to animals in the drinking water at a concentration of 10⁹ CFU/g for 8 weeks. NAFLD severity was determined based on liver/body weight, pathology and biochemistry markers. Caecal samples were collected for the metagenomics by 16S rRNA sequencing. Metabolite profiles were obtained from caecum, liver and serum. Human stool samples (healthy control [n = 22] and NAFLD patients [n = 23]) were collected to investigate clinical reproducibility for microbiota-derived metabolites signature and metabolomics biomarker.

Results: L. lactis and P. pentosaceus supplementation effectively normalized weight ratio, NAFLD activity score, biochemical markers, cytokines and gut-tight junction. While faecal microbiota varied according to the different treatments, key metabolic features including short chain fatty acids (SCFAs), bile acids (BAs) and tryptophan metabolites were analogously restored by both probiotic supplementations. The protective effects of indole compounds were validated with in vitro and in vivo models, including anti-inflammatory effects. The metabolomic signatures were replicated in NAFLD patients, accompanied by the comparable levels of Firmicutes/Bacteroidetes ratio, which was significantly higher (4.3) compared with control (0.6). Besides, the consequent biomarker panel with six stool metabolites (indole, BAs, and SCFAs) showed 0.922 (area under the curve) in the diagnosis of NAFLD.

Conclusions: NAFLD progression was robustly associated with metabolic dys-regulations in the SCFAs, bile acid and indole compounds, and NAFLD can be accurately diagnosed using the metabolites. L. lactis and P. pentosaceus ameliorate NAFLD progression by modulating gut metagenomic and metabolic environment, particularly tryptophan pathway, of the gut-liver axis.

Keywords: gut-liver axis; indole; metabolites; microbiome; non-alcoholic fatty liver disease.

FIGURE 1

Effect of Lactobacillus and Pediococcus on the Western diet‐induced liver disease. (A) Experiment design of WD model. (B) Gross specimen of mice liver. (C) L/B ratio. (D) Pathological effects of strains on the liver. Hematoxylin and eosin staining of liver sections were analyzed (x 20) and analyze to NAFLD activity score (NAS). (E) Liver function test and cholesterol level of mice. *p < 0.05 as compared with WD. L/B, Liver weight/body weight; LL, Lactobacillus lactis; NC, normal control; PP, Pediococcus pentosaceus; WD, Western diet

FIGURE 2

Effects of L. lactis and P. pentosaceus on gut‐liver axis. (A) The expression of tight junction markers, occludin and Zo‐1, in the mice intestine (n = 5/group). (B) Trans‐epithelial electrical resistance assay using Caco2 cell. (C) The levels of endotoxin in mice serum (n = 5/group) using LAL assay kit. (D) Representative microphotographs and measured areas of CD68 immunohistochemistry. (E) The levels of inflammatory cytokines tumor necrosis factor (TNF)‐α, interleukin (IL)‐1β and interleukin (IL)‐6 in mice liver (n = 5/group). (F) Effects of strains on WD‐induced activation of MAPKs and NF‐kB in mice liver. (G) The analysis of adipokines such as leptin and adiponectin in mice liver (n = 5/group). (H) The levels of bile acid (BA) regulation‐related genes, Cyp7A1 and FXR, in mice liver (n = 5/group). *p < 0.05 as compared with WD. LL, Lactobacillus lactis; MAPKs, Mitogen‐activated protein kinases; NC, normal control; NF‐kB Nuclear factor kappa‐light‐chain‐enhancer of activated B cells; PP, Pediococcus pentosaceus; WD, Western diet

FIGURE 3

The alteration in the microbial taxonomic composition of the Western diet‐mice caecal samples. (A) The alpha diversity based on species richness. (B) The beta diversity by principal coordinate analysis (Bray‐Curtis distance), and (C) Firmicutes to Bacteroidetes (F/B) ratio as compared among four different groups (NC, WD, LL, and PP). (D) Relative abundances of caecal microbiome at genus level in different experimental groups. (E) Heatmap analysis for significantly different species. (F) Comparative analysis of the estimated functional profiles based on KEGG orthology in different experimental group. KEGG, Kyoto encyclopedia of genes and genomes LL, Lactobacillus lactis; NC, normal control; PP, Pediococcus pentosaceus; WD, Western diet

FIGURE 4

The caecal metabolomic dysregulation by Western diet and the normalization by the supplementation with L. lactis and P. pentosaceus . (A) Chemical classification of identified metabolites in mouse caecum provided by HMDB (http://www.hmdb.ca). A total of 256 compounds (91%) are categorized into nine super classes. (B) The score scatter plot of 282 caecal metabolites by principal component analysis (PCA). Most variation was imposed by PC1 (30.8%) and PC2 (12.3%). *p < 0.05 as compared with WD. (C) Overview of the metabolic features. Pie charts present the number of metabolites that were significantly different in other groups, respectively compared to WD (Student's t‐test, p < 0.05). Red and blue colours present significantly higher or lower abundance in other groups, respectively compared to WD (p < 0.05). The network is constructed based on chemical structural similarity (Tanimoto score) and KEGG reaction pair (substrate‐product relation), which results in distinctive metabolic modules indicated by box. Red and blue colours present significantly higher or lower abundant in NC, LL, and PP groups, respectively compared to WD (Student's t test, p < 0.05). Node sizes are determined by the ratios. HMDB, human metabolome database; KEGG, Kyoto encyclopedia of genes and genomes; LL, Lactobacillus lactis; NC, normal control; PC1, Climatic index 1; PC2, Climatic index 2; PP, Pediococcus pentosaceus; WD, Western diet

FIGURE 5

The characteristic alteration in the major caecal metabolites according to different diet types and strains supplementation. (A) The metabolites that show common abundance pattern in other groups compared to WD group. A total of 36 metabolites are significantly different in all three groups. (B) The score scatter plot of the 36 metabolites by principal component analysis. (C) Heatmap of auto‐scaled abundances (mean‐centered and divided by the standard deviation of each variable) of the common metabolites. (D) Volcano plot of the common metabolites (36 metabolites) that are significantly different in all three groups, respectively compared to WD group. The x‐axis is log2‐fold change, and the y‐axis is log10‐p value. Indole derivatives (indole‐3‐propionic acid and methyl indole‐3‐acetic acid) show the highest fold‐increases in all three groups. Bile acids (taurocholic acid and taurochenodeoxycholic acid) present the highest fold‐decreases in all three groups. The name of metabolites in common among three groups is only visualized. The levels of caecal short chain fatty acids (SCFAs) (E) and bile acids (BAs) (F) in different diet groups (n = 4–7). The log10‐transformed abundances (ion intensities) are shown as violin plot. *p < 0.05 as compared with WD group by Mann–Whitney U test. **p < 0.05 as compared with WD group by nonparametric Kruskal–Wallis test and Dunn's test adjusted by Benjamini‐Hochberg correction. LL, Lactobacillus lactis; NC, normal control; PP, Pediococcus pentosaceus; WD, Western diet

FIGURE 6

Ameliorative effects of gut microbe‐derived indoles on non‐alcoholic fatty liver disease (NAFLD) progression. The levels of indoles in caecal samples (A) and livers (B) in different diet groups (n = 4‐7). The log10‐transformed abundances (ion intensities) are shown as violin plot. *p < 0.05 as compared with WD group by Mann–Whitney U test. **p < 0.05 as compared with WD group by nonparametric Kruskal–Wallis test and Dunn's test adjusted by Benjamini–Hochberg correction. (C) Anti‐inflammatory effects of indoles on Raw 264.7 cell exposed to LPS (100 ng/ml, n = 5). Following the subsequent indole‐treatment (indole acrylic acid, 100 μM; indole‐3‐acetic acid, 500 μM; indole‐propionic acid, 100 μM), inflammatory cytokine gene expression is analyzed based on qRT‐PCR. *p < 0.05. (D) The expression levels of aryl hydrocarbon receptor. (E) Experiment design of WD indole mice model and L/B ratio (F) level of liver enzymes, cholesterol, and cytokine. IPA, indole‐propionic acid; L/B, Liver weight/body weight; LPS, lipopolysaccharides; LL, Lactobacillus lactis; NC, normal control; PP, Pediococcus pentosaceus; qRT‐PCR, quantitative reverse transcription polymerase chain reaction; WD, Western diet

FIGURE 7

Human faecal metabolomic features in the non‐alcoholic fatty liver disease (NAFLD) patients. The levels of human faecal short chain fatty acids (SCFAs) (A), bile acids (BAs) (B), and indoles (C) in two different groups, healthy control (HC) (n = 22) and the patients with NAFLD‐elevated liver enzyme (ELE) (n = 25). The log10‐transformed abundances (ion intensities) are shown as violin plot. *p < 0.05 as compared with HC group based on Student's t‐test. **p < 0.05 as compared with HC after multiple comparison adjustment by false discovery rate. (D) Receiver operating characteristic (ROC) curve analysis of biomarker panel included by multiple faecal metabolites for discriminating the HC (n = 22) and non‐alcoholic steatohepatitis (NASH) (n = 25). Each biomarker cluster includes Marker_1 (indole‐3‐propionic acid, glycocholic acid, propionic acid, butyric acid, valeric acid), Marker_2 (indole‐3‐propionic acid, indole‐3‐acetic acid, glycocholic acid, propionic acid, butyric acid, valeric acid). The area under curve value is 0.918 (95% confidence interval: 0.801–0.978) and 0.922 (95% confidence interval: 0.805–0.980), respectively. Optimal cutoff is determined using the closest to top‐left corner, and the 95% confidence interval is calculated using 1000‐bootstrapping. The area under ROC curve and predicative accuracy are calculated through 1000 times permutation test

FIGURE 8

Microbial community profiles of human in response to different diet type and microbial preventive therapy. (A) Pie charts of phylum distribution in human stool. The distribution of phylum was compared between healthy control (HC) (n = 22) control and non‐alcoholic fatty liver disease‐elevated liver enzyme (NAFLD‐ELE) (n = 25). *p < 0.05 (B) Firmicutes to Bacteroidetes ratio. (C) Box and whisker plot of Alpha‐diversity indices. *p < 0.05 and **p < 0.01 as compared between the NC and the NAFLD‐ELE groups based on Mann–Whitney U test. (D) Relative abundance of faecal microbiome at the genus level in HC and NAFLD‐ELE. (E) Volcano plot of stool microorganisms at genus level. The x‐axis presents log10 transformed fold‐change, and y‐axis presents ‐log10 transformed p‐value calculated by Student's t‐test. Red and blue colours present higher or lower abundance in NAFLD‐ELE compared to HC. (F) Co‐inertia analysis of metabolomic and taxonomic profiles. The x‐ and y‐axis present the first two vectors that most explain the variance composed with integrative metabolomic and taxonomic profiles. Circles and squares represent the faecal metabolome and microbiome, respectively. Green and blue‐sky colours indicate NC and NAFLD groups, respectively. Lines connect the faecal metabolome and microbiome from the same individual. The shorter the length of line is, the stronger the level of association is. (G) Co‐occurrence matrix of individual metabolite and microbial composition. The correlation structure consists of gut microbial feature (genus level) and the metabolic features that showed the significant alteration in the NAFLD patients. The correlation coefficient was calculated based on Spearman rank analysis. *p < 0.05, **p < 0.01, and ***p < 0.001