Gut-Microbial and -Metabolomic Signatures in the Prevention of Non-Alcoholic Fatty Liver Disease by Lactobacillus Lactis and Pediococcus Pentosaceus


 Background: Despite a recent preventive evidence of Lactobacillus and Pediococcus on non-alcoholic fatty liver disease (NAFLD) progression, the underlying mechanistic is less understood. We explored the causality of L. lactis and P. pentosaceus on the gut-metabolomic modulation in the prevention of NAFLD progression in mouse model and subsequently discovered metabolic biomarkers based on NAFLD patients.Results: Six-week-old male C57BL/6J mice were divided into 4 groups (control, Western diet [WD], and 2 WD with strains [L. lactis and P. pentosaceus]). Given completely reproduced data (liver/body ratio, pathology, and metagenomic profiles), comorbid etiologies including inflammation and diabetes were significantly amended by 2 strains. The comprehensive metabolomic profiles of the mouse cecum revealed unique compositional characteristic according to the groups. L. lactis and P. pentosaceus supplementation restored the dysregulation in short chain fatty acids (SCFAs), bile acids, and tryptophan metabolites. Indole derivatives (indole-3-acetic acid, indole-3-propionic acid, and indole-3-acrylic acid) showed anti-inflammatory activities by suppressing pro-inflammation cytokines. Human data (healthy control [n=30] and NAFLD patients [n=74]) were analyzed for clinical association and biomarker. The Fermicutes/Bacteroidetes ratio of NAFLD (4.3) was significantly higher compared with control (0.6)(p<0.05), accompanied by the dysregulation in the key metabolic signatures identified in mouse model. Metabolic panel with 5 stool metabolites (indole, bile acids, and SCFAs) revealed 0.868 (area under the curve; 95% CI 0.773-0.933) in the diagnosis of NAFLD.Conclusion: L. lactis and P. pentosaceus ameliorate NAFLD by modulating gut metabolic environment, particularly the indole pathway, of the gut-liver axis. NAFLD progression was associated with metabolic deterioration in the SCFAs, bile acid, and indole pathways. ClinicalTrials.gov: NCT04339725

tryptophan metabolites have been shown to affect the development of NAFLD [6]. Indoles are major products of tryptophan-derived metabolites, and some indole family metabolites are reported to reduce the production of pro-in ammatory cytokines by downregulating macrophages, scavenging free radicals, and reducing oxidative stress [5,7].
In our previous report, Lactobacillus and Pediococcus supplementation have been demonstrated to improve NAFLD by the modulating the gut microbiota and in ammation [8]. In addition, our preliminary study with strains licensed for human use and reproducibility experiments with selected strains have both shown promising results in Western diet (WD) induced NAFLD (Supplementary table 1 and  supplementary Fig. 1). In this study, we comparably examined gut micro-environments of WD-induced NAFLD mice and of patients with NAFLD using metagenomics/metabolic pro ling. We deepened the mechanistic understanding of the preventive effects of L. lactis and P. pentosaceus based on comprehensive examination on biochemical and pathology parameters.

Strain preparation
Lactobacillus and Bi dobacterium used for the preliminary study were isolated from various source such as sour milk, cheese, healthy Korean adult feces and new-born's feces. L. lactis is lactic acid bacterium that was isolated from sour milk. P. pentosaceus (KCTC 18308P) is a strain of lactic acid bacterium that was isolated from nger millet (Eleusine coracana) gruel. L.lactis and P. pentosaceus were inoculated into a ask containing de Man, Rogosa, and Sharpe media (BD/Difco). The strains were incubated under anaerobic conditions at 37°C for 24 h. Stocks of each strain were prepared by mixing the culture broth with an equivalent 20% skim milk solution and then storing the mixture at -80°C.

Pathology
Specimens were xed with 10% formalin and routinely embedded in para n, and the tissue sections were processed with hematoxylin and eosin, Masson's trichrome, and CD 68 staining. NAFLD activity score (NAS), an objective index for classifying the grade of fatty liver, was assessed [9]. All biopsy specimens were analyzed by a hepato-pathologist (S. H. H.).

Page 5/27
A prospective cohort study was carried out between April 2017 and March 2020. A total of 104 patients comprising of healthy control (HC, n=30), NAFLD patients with normal liver enzymes (NAFLD-NLE, n=37), and NAFLD patients with elevated liver enzyme (NAFLD-ELE, n=37) were prospectively enrolled (ClinicalTrials.gov: NCT04339725). This study involved patients with liver disease who were followed-up at the hepatology department of Hallym University. The diseases of the patients were treated regardless of the study. This study was conducted in accordance with the ethical guidelines from the 1975 Helsinki Declaration, as re ected by a prior approval by the institutional review board for human research in hospitals participating in the trial (2016-134). Informed consent for enrollment was obtained from each participant. Patients with autoimmune hepatitis, malignancy, drug induced liver injury, and virus induced hepatitis were excluded.
Stool analysis for the metagenomics Metagenomic DNA was extracted with a QIAamp stool kit (cat. no. 51504) and ampli cation of the V3 -V4 region of the bacterial 16S rRNA gene was conducted using barcoded universal primers. PCR was performed according to following steps: an initial denaturation at 95 °C for 5 min, 20 cycles of 95 °C for 30 seconds, 55 °C for 30 seconds, and 72 °C for 30 seconds, followed by a nal extension at 72 °C for 10 min. Puri cation of the amplicons was conducted with an Agencourt AMPure XP system (Beckman, USA) and quanti cation of the puri ed amplicons was conducted using PicoGreen and quantitative PCR. After pooling of the barcoded amplicons, sequencing was carried out using a MiSeq sequencer on the Illumina platform (ChunLab Inc., Republic of Korea) according to the manufacturer's speci cation.
Microbiota pro ling was conducted with the 16S-based Microbial Taxonomic Pro ling platform of EzBioCloud Apps (ChunLab Inc., Republic of Korea). After taxonomic pro ling of each sample, comparative analyzer of EzBioCloud Apps was used for the comparative analysis of the samples. Taxonomic assignment of the reads was conducted with ChunLab's 16S rRNA database (DB ver. PKSSU4.0) [10]. OTU picking was conducted with UCLUST and CDHIT with 97% of similarity cutoff [11].
Subsequently, Good's coverage, rarefaction, and alpha-diversity indices including ACE, Chao1, Jackknife, Shannon, Simpson, and NPShannon were calculated. Beta-diversity including PCoA and UPGMA clustering was shown in the comparative MTP analyzer. All 16S rRNA sequences were deposited in the ChunLab's EzBioCloud Microbiota database and sequencing reads of the 16S rRNA gene from this study were deposited in the NCBI Short Read Archive under the bioproject number PRJNA532302.

Metabolic pro ling of mouse cecal samples and human stool samples
The metabolomic pro les of mouse cecum were acquired using a combination of GC-MS and two LC-MS methods. Cecal samples were thawed at 4 °C and mixed with 1.1 ml of cold extraction solvent (acetonitrile/water 1:1, v/v). The mixtures were vortexed for 1 minutes and sonicated for 5 minutes under ice and centrifuged at 13,200 rpm for 5 min at 4°C. Each supernatant (500 µl) was transferred into a new 2 ml tube for short chain fatty acids (SCFAs) analysis (Method 1, see the supplementary method for details) [12]. The rest of supernatant was mixed with 600 µl of cold extraction solvent (acetonitrile/methanol, 1:3, v/v). For the second extraction step, the mixtures were vortexed for 1 min and centrifuged at 13,200 rpm for 5 minutes at 4 °C. The supernatants (500 µl) were aliquoted and transferred to new 1.5-ml tubes for gas-chromatography time-of-ight mass spectrometry (Method 2) and liquidchromatography Orbitrap mass spectrometry (Method 3). The aliquots were concentrated to complete dryness using a speed vacuum concentrator (SCANVAC, Korea) [13]. Similarly, human stool metabolomic pro les were obtained based on Methods 1, 2 and 3. The cecal data acquired by GC-TOF MS has been retrieved from our previous study [8].
Real-time Reverse Transcription-Polymerase Chain Reaction (RT-PCR) Analysis Isolation of total RNA from tissue was performed using a Trizol reagent kit (Invitrogen, Gaithersburg, MD, USA) according to the manufacturer's instructions. Aliquots of total RNA (2 μg) were converted to cDNA using cDNA reverse transcription kit (Applied Biosystems, Foster City, CA). The cDNA was ampli ed for quantitative PCR using the Luna® Universal Probe qPCR Master Mix (New England Biolabs Beverly, MA, USA) and each target speci c probe-primer (Applied Biosystems, Foster City, CA).

ELISA
Tissue homogenates were incubated with PRO-PREP™ Protein Extraction Solution (iNtRON Biotechnology, Korea) at 4 °C for 30 min, vortexing 3 min after added stainless bead and then centrifuged at 10,000 g for 10 min. Resulting supernatants were harvested and analyzed for the levels of pro-in ammatory cytokines including TNF-α, IL-1β, and IL-6 were analyzed by enzyme-linked immunosorbent assay (R & D Systems, Minneapolis, MN), according to the manufacturer's instructions.

Western blots
Tissue homogenates were incubated with PRO-PREP™ Protein Extraction Solution (iNtRON Biotechnology, Korea) on ice for 30 min, vortex mixer 3 min after added stainless bead and then centrifuged at 10,000 g for 10 min. Thirty micrograms of proteins were resolved by 10% SDS-PAGE and transferred to nitrocellulose membranes. The blots were probed with the indicated primary antibodies RBP4 (1:1000, Abcam, Cambridge, MA, USA), phospho-NF-κB p65, MAPKs, phosphor-MAPKs, and GAPDH (1: 1000, Cell Signaling Technology, Beverly, MA, USA) followed by incubation with the corresponding horseradish peroxidase-conjugated secondary antibodies (1:10000 dillution). The membrane was reacted with the enhanced chemiluminescence (ECL) substrate solution and analyzed by Amersham Imager 680 (GE Healthcare UK Ltd, Buckinghamshire, UK).

Statistical analysis
One-way ANOVA, the Kruskal-Wallis test, and independent sample T-test were performed for the body weight, liver function test, histology analysis, RT-PCR, ELISA, and LAL test. A p value<0.05 was considered to indicate statistical signi cance. All statistical analyses were performed using GraphPad prism software ver. 8 (GraphPad Software Inc., San Diego, CA, USA). The statistical analyses were conducted on all continuous variables acquired from GC-MS and LC-MS. That analysis described to the supplemental information.

Results
L. lactis and P. pentosaceus suppress the progression of non-alcoholic fatty liver disease The L. lactis (4.76±0.31) and P. pentosaceus (5.16±0.23) supplementations were associated with signi cant improvement in the liver/body weight ratio (%) when compared with that in the WD group (6.51±0.28) (Fig. 1B and 1C)(p<0.001). The 9-week WD induced steatosis and in ammation in the liver pathology whereas the two strain groups showed improvement in the pathologic ndings. WD induced steatosis (2.67±0.52) was reduced in the strain groups (LL 1.33±0.52 and PP 1.83±0.41). Regarding the in ammation grade, the WD group (2.67±0.52) exhibited elevated scores in comparison to the strain groups which had a score of 1. The WD-induced increase of NAFLD activity score (NAS)(5.83±0.75) was signi cantly decreased in the LL (2.5±0.84) and PP (3.83±0.41) groups (Fig. 1D).
Western diet-induced dysbiosis is ameliorated by L. lactis and P. pentosaceus supplementation In analysis of stool samples, the compositions of Proteobacteria, Verrucomicrobia, Deferribacteres, Actinobacteria, Bacteroidetes, Firmicutes, and ETC were different according to the diet groups ( Fig. 2A). At the genus level, each group revealed different compositions (Fig. 2B). The Firmicutes-to-Bacteroidetes ratio (F/B ratio) has been broadly studied in human and mouse gut microbiotas [14]. The F/B ratio in the WD group (60.1) was decreased in the LL (17.8) and PP (13.6) groups (Fig. 2C). In the analytics for beta diversity for microbiota taxonomic pro ling, each group showed a different location (Fig. 2D). WDinduced decrease in species richness and diversity index is not changed by strains supplementation (Fig.  2E). In a heatmap for the comparison of species, each groups was associated with different patterns compared with that of the WD group, and WD-induced changes in a heatmap for the functional biomarker expression is recovered by L. lactis and P. pentosaceus supplementation (Fig. 2F). Lactobacillus abundance was increased in the PP group. Also, signi cantly recovered Bacteroides acidifaciens abundance by LL group. That reported to may have potential for treatment of metabolic diseases such as diabetes and obesity [15](Supplementary gure 3A). The top 42 statistically signi cant markers not shown in Figure 2 are presented in supplementary table 2.
The composition of the gut-microbiota differs according to the progression of liver disease In the analysis of human stool, the proportions of phyla were different according to the progression of liver diseases (Fig. 3A). In the taxonomic composition at the phylum level, Firmicutes abundance increased in the NAFLD-ELE group when compared with HC group. At the genus level, each group revealed different compositions (Fig. 3B). Bacteroidetes abundances and the F/B ratio were signi cantly different between the HC, NAFLD-NLE, and NAFLD-ELE groups ( Fig. 3C)(p<0.001). In the comparison for beta diversity, each group occupied a different area (Fig. 3D). The species richness and diversity indexes were signi cantly decreased according to disease progression (Fig. 3E). When looking at the heatmap for the comparison of species, each group demonstrated different patterns, and functional biomarkers were also expressed with different patterns (Fig. 3F). The top 38 statistically signi cant markers not shown in Figure 3 are presented in supplementary table 3. Prevotella was decreased as the disease progressed (Supplementary gure 3B).
Short chain fatty acids are characteristically altered by diet type and strain supplementation Overall, the WD group was associated with signi cant decreases in the short chain fatty acids (SCFAs) except iso-valeric acid when compared with those in the NC, LL, and PP groups (p<0.05) (Fig. 4A). The LL group showed marginal differences after multiple comparison adjustment whereas the PP group presented signi cant differences in all SCFAs compared to the WD group. The differences were not signi cant between the two strains-supplemented groups (Supplementary table 4).
Subsequently, we analyzed SCFAs from NC (n=35) and NAFLD-ELE (n=26). Similarly, the levels of the main SCFAs were observed signi cantly lower levels in the patients with NAFLD-ELE than in the HCs (p<0.05)( Fig. 4B and supplementary table 4).
Comprehensive pro ling of the mouse cecal metabolome shows unique different patterns among groups To comprehensively pro le cecal metabolic context, we performed untargeted and targeted metabolic pro ling of cecal samples based on GC-and LC-MS analysis. The metabolomic pro les were analyzed for the NC, WD, LL, and PP groups. Metabolic features were assigned to 282 unique compounds based on reference comparisons, spectra library searching, and retention time indexes. The chemical ontology analysis classi ed the compounds, and approximately 50% of metabolites were categorized as organic acids and lipid molecules (Fig. 4C). The sub-categories of the major classes were as follows: carboxylic acids, fatty acyls, and steroids accounted for 60, 32, and 14 compounds, respectively (Supplementary  table 5). Subsequent pathway analysis demonstrated a wide range of coverage, including amino acid metabolism, carbohydrate metabolism, and fatty acid metabolism (Supplementary gure 4A).
First, the metabolic phenotypes of the four groups were characterized by unsupervised multivariate statistics. Principal component analysis showed the distinctive clusters between the NC group and the others (Fig. 4D). We, then, evaluated cecal metabolomic traits that were altered by the WD as a baseline. A signi cant difference was found for a total of 135 compounds out of 282 (Supplementary table 6). 11 percent of metabolites were signi cantly enriched in the WD group, whereas 37 percent were depleted when compared to those in the NC group (Fig. 4E). The largest depletion was determined to be 5hydroxyindole-3-acetic acid content in the WD group (Supplementary table 6). Other indole derivatives were concomitantly depleted including indole-3-propionic acid, methyl indole-3-acetic acid, indole-3-acetic acid, and indole-3-acrylic acid (Supplementary table 6). In addition, pathway overrepresentation analysis implied repression of carbohydrate metabolism (Supplementary gure 4B). Taurine conjugated bile acids were the compound with highest increases in WD group compared to NC group. Taurocholic acid and taurochenodeoxycholic acid were associated 50-/43-fold increases in WD group compared to NC group, respectively. Other increases associated with WD group included enriched metabolites were glutamic acid, cholesterol, 2'-deoxycytidine, and glycocholic acid (>10-fold changes). Overall, the most signi cantly upregulated compounds were determined to be associated with amino acid metabolism and primary bile acid biosynthesis based on pathway analysis (Supplementary gure 4B, right panel).
Exploration of gut microbiota-derived remedial therapeutics: common metabolic signatures among the 3 groups relative to the Western diet group We explored common metabolic features among the NC, LL, and PP groups in comparison to the WD group, which may determine which molecular phenomic signatures were microbiota-derived and potentially had roles in remediating NAFLD. The metabolic re-programming by the two strains were similar overall (Fig. 4E). To effectively identify the metabolic features, we constructed an integrated metabolic network that was tiered by chemical structural similarity and enzymatic reaction connectivity. To effectively identify the metabolic features, we constructed an integrated metabolic network (MetaMapp) that was tiered by chemical structural similarity (Tanimoto score) and enzymatic reaction connectivity (KEGG reaction pair). The network provided a general overview at the level of the metabolic module and comprehensive details at the individual metabolite level [13]. Overall, the LL and PP groups showed compatible patterns in major metabolic modules, which coincided with the similar levels of the preventive effect of WD-induced liver damage.
We further interrogated the common metabolites that were similarly regulated in the NC and strain-fed groups. A total of 33 metabolites showed similar patterns among the three groups when compared to those in the WD group ( Fig. 5A and supplementary gure 5). The metabolic pro les of the 33 metabolites were associated with the highest discrimination among the three groups, the NC, strain-fed groups, and the WD group (Fig. 5B).
Compared to those in the WD group, gut-bacterial metabolites (indole-3-propionic acid and methyl indole-3-acetic acid) showed the highest fold-change in the NC and strain-fed groups ( Fig. 5C and D). On contrary, indole-3-lactic acid and indole-3-pyruvic acid were exclusively different in the LL group and the PP group, respectively. Whereas indole-3-acrylic acid was highly abundant only in the NC group ( Fig. 5E  and supplementary table 7). Subsequently, we evaluated the indoles from human fecal samples and compared the levels between HC and patients with NAFLD-ELE. Indole-3-propionic acid was at signi cantly-reduced level in the patient group (Student's t-test, p<0.01). Indole-3-acrylic acid and indole-3acetic acid were marginally reduced (p=0.225,0.473) whereas indole-3-lactic acid was moderately increased in the NAFLD patients compared to HC (Student's t-test, p=0.100) (Fig. 5F).
Bile acid homeostasis in the farnesoid X receptor pathway is altered by diet and is regulated by the tight junction of the intestine.
Primary BAs conjugated with taurine were most dramatically upregulated in the WD group, compared to those in the other groups (Fig. 6A). Glycocholic acid was signi cantly enriched in the WD group relative to that in the NC and LL groups. In contrast, secondary BAs following deconjugation/dehydroxylation of primary BAs presented a decreasing tendency in the WD group except taurodeoxycholic acid (supplementary table 6). Overall, the fecal BAs showed elevated levels in the patients diagnosed with NAFLD-ELE (Fig. 6B), which were not identical to the pro les of the cecal BAs in the WD group. Nonetheless, comparable dysregulation was identi ed for fecal taurocholic acid (p=0.05, FDR=0.1) and glycocholic acid (p=0.104, FDR=0.156) in the patients diagnosed with NAFLD-ELE (Fig. 6B). The deconjugated BAs (cholic acid and chenodeoxycholic acid) were the most signi cantly enriched in the patients, consistent with a previous report [16].
BA homeostasis is critically regulated by the farnesoid X receptor (FXR), which is activated by BAs [17]. The BA synthesis-, BA transport-, and hepatic acid regulation-related genes NCTP, Cyp7A1, SHP, and FXR were downregulated in the WD group. However, the LL and PP groups were associated with partial recovery of aforementioned downregulation (Fig. 6C).
We analyzed the expression of the tight junction occludin and ZO-1 genes. The LL and PP groups were shown with increased occludin and ZO-1 gene expression compared with that in the other groups (Fig.  6D). Caco-2 cells were cultured to complete con uence and co-incubated with a bacterial suspension in MEM for eight hours. The treatment of Caco-2 cells with LL and PP increased the trans-epithelial electrical resistance values by 2.3-and 1.9-fold, respectively, compared with those of non-treatment controls. And endotoxin analysis in serum was performed. The elevated levels of endotoxin in the WD group were reduced in the LL group. However, those in the PP group were not signi cant (Fig. 6E). These results suggest that LL and PP supplementation strengthened intestinal-barrier function and reduced bloodstream endotoxin in ltration from the intestine.
Carbohydrate metabolism was correspondingly altered according to the diet type. All monosaccharides were signi cantly more enriched in the NC group than in the WD group. Among the intestinal monosaccharides, glucose and xylose were signi cantly depleted in the WD group, respectively compared to those in the other groups (Fig. 6F). The strain-fed groups showed glucose levels equivalent to those in the NC group. Fructose was not differentially regulated among the WD groups, whereas galactose and mannose levels in the PP group were at comparable levels to those in the NC group (Supplementary table  8  L. lactis and P. pentosaceus attenuate in ammation and insulin resistance in the liver Altered metabolite composition due to gut microbiota dysbiosis can damage the liver and induce in ammation through the gut-liver axis. Elevated in ammatory cytokines, such as tumor necrosis factor (TNF)-α, interleukin (IL)-1β, and IL-6 in the WD group were signi cantly decreased via the LL and PP groups (Fig. 7A). Immunohistochemical analyses for CD68, a marker for macrophages, were performed in representative cases. The mean value of the positive-stained area measured in random areas of the liver was determined. The strain groups showed signi cant reduction in stained area compared to the area for the WD group (Fig. 7B). Various MAPKs, such as p38, c-Jun NH2-terminal kinase and extracellular signalregulated kinase participate in the expression of pro-in ammatory mediators during in ammatory responses [18]. Elevated activation of the MAPK-NF-kB pathway in the WD group was decreased in the LL and PP groups (Fig. 7C). In ammation in adipose tissues is a mechanism to induce insulin resistance and is mediated by the activation of cellular stress-induced in ammatory signaling pathways. The proin ammatory adipokines retinol-binding protein 4 and leptin were elevated the WD group and signi cantly decreased the LL and PP groups. In addition, the anti-in ammatory adipokine adiponectin decreased in the WD group and was signi cantly elevated in the LL group compared with that in the WD group (Fig. 6D  and E).

Biomarkers for non-alcoholic steatohepatitis
We performed binary logistic regression analysis to examine whether a single molecule or composed set of metabolites can predict NAFLD-ELE (n=45) over HCs (n=35). Based on the metabolites (SCFAs, indoles, and BAs), we applied receiver operating characteristic (ROC) curve analysis for accuracy, speci city, and sensitivity. The area under the curve (AUC) for indole-3-propionic acid was 0.782 (95% CI, 0.676-0.867) (Fig. 8A)

Discussion
Our current investigation is the rst report on the accomplishment of the comprehensive stool metabolomics-based elucidation of the preventive mechanism of strain-supplementation therapy for NAFLD. Besides, we link the metagenomic and metabolomic features of the mouse model to human data, which leads to the discovery of potential biomarker, precisely diagnosing NAFLD progression.
Supplementation with L. lactis and P. pentosaceus strains are known to be associated with bene cial effects in various diseases such as enteritis, in ammation, and hypertension [21,22]. Pediococcus inactivated hepatic stellate cells among with a decreasing hepatic in ammatory response by via suppression of macrophages and in ammatory cytokines [23]. In this study, L. lactis and P. pentosaceus were found to attenuate in ammation and insulin resistance in the liver. Overall, L. lactis and P. pentosaceus supplementation should be considered a potentially effective therapeutic options in the treatment of NAFLD.
WD induces dysbiosis and elevation of endotoxin levels in the gut [8]. These bacterial endotoxins activate macrophages by engaging Toll-like receptor 4. Macrophages produce in ammatory cytokines or chemokines which lead to liver damage through the MAPK-NF-kB signaling pathway. Our study also demonstrated that the WD is related with endotoxemia, alteration of microbiota composition, macrophage related elevation of cytokine levels, and MAPK-NF-kB signaling. In addition, P. pentosaceus decreased serum endotoxin level via restoring gut-permeability and curved release of pro-in ammatory cytokines by inactivating MAPK-NF-kB signaling.
In our study, L. lactis and P. pentosaceus supplementation signi cantly decreased the L/B ratio, NAS score, and serum chemistry markers as compared to the WD group. Previous studies suggested that the gut-liver axis plays an important role in the pathophysiology of NAFLD and that probiotics can modulate the elements of the gut-liver axis, especially the gut microbiota [8,24]. Altogether, our results showed that L. lactis and P. pentosaceus supplementation might be effective for bringing about weight reduction and histologic improvement in patients with NAFLD.
Regarding metagenomics, it is well-known that WD in uences the composition of the intestinal micro ora in animals, and this difference is also found in patients with NAFLD [8]. In our research, administration of different strains resulted in differential composition of the microbiota. Whether the alteration of intestinal microbiota is a result of disease progression or compensatory reaction to disease is not clearly understood; more research is needed to understand the mechanism or ecosystem of these strains. The metagenomic pro les did not reveal a correspondence in the proportions of the gut microbiota between the normal control and probiotic-fed groups. This inequivalence implies that the common molecular mechanism drives the protective effects of probiotic supplementation on WD-induced NAFLD.
Indeed, a growing evidence implicates gut microbiota-derived metabolites as active modulators, encompassing diet, microorganisms, and the host, attributed to causal functionality, including SCFAs and tryptophan metabolites [19,25]. Tryptophan metabolites in particular have been shown to affect the development of NAFLD by altering the composition of the gut microbiota [5,26]. Indole compounds are major products of tryptophan-derived metabolites, and include indole-3-acrylic acid, indole-3-acetic acid, and indole-3-propionic acid [5,7,20]. In a previous study, indole-3-acrylic acid was reported to promote intestinal epithelial barrier function and mitigate in ammatory responses via NRF2 activation [19]. Indole-3-acetate is associated with attenuation of indicators of in ammation in macrophages and cytokinemediated lipogenesis in hepatocytes [20]. Additionally, indole-3-propionic acid has been found to improve HFD-induced intestinal epithelial barrier damage and inhibits endotoxin-induced production of proin ammatory cytokines via inactivation of NF-κB signaling [27]. Indeed, our metabolomic analysis revealed that the indole compounds were differentially regulated by the L. lactis and P. pentosaceus and the levels of indole compounds in the probiotic-feeding groups were the comparable levels to those of normal control. Particularly, indole-3-propionic acid was associated with the highest fold-change when normal control and two probiotic-feeding groups were compared with the western diet group. The differential regulation coincided with results for the NAFLD-ELE patients. Accordingly, we examined the functionality of three indole compounds, which were linked to suppression of cytokines such as TNF-α and IL-1β, in LPS-treated macrophages. These results imply that a diet with L. lactis and P. pentosaceus at least partially protects against NAFLD progression via the production of anti-in ammatory metabolites such as indole compounds.
BAs have direct or indirect antimicrobial effects and modulate the composition of the microbiota, which in turn has a role in regulating the size and composition of the BA pool [28]. BA homeostasis is critically regulated by FXR, which is activated by BAs. FXR is known to exert tissue-speci c effects in regulating BA synthesis and transport [17]. Previous studies showed that mice with FXR de ciency were a icted with hepatic steatosis as well as glucose and insulin intolerance, the main hallmarks of NAFLD in humans [29,30]. There have been incoherent results regarding BAs, implicating the complexity in which different spatial and temporal examinations have been performed under non-identical experimental settings. For instance, there have been a relatively few number of investigations on intestinal BAs. Nonetheless, the WD-induced NAFLD in our study was best characterized by dysregulation of unconjugated BAs, which is consistent with other recent studies [31]. Besides, our results clearly revealed that the WD-induced dysregulation of BAs was largely restored by the strains administration. The fecal BAs in the patients with NAFLD-ELE did not share identical pattern compared to that of WD-induced NAFLD mouse model; however, the general trend was similar in regards to alteration of unconjugated BAs by WD. In particular, the diet-induced accumulation of taurocholic acid has been reported to perturb gut microbial symbiosis in a mouse model [32,33] and to stimulate hepatic in ammation and brosis [34].
It is plausible that a WD induces severe reduction of microbial diversity, which leads to the depleted SCFAs content [35] triggering accelerated glucose consumption by host enterocytes and colonocytes to compensate for depletion of the main energy sources. Accordingly, we observed a decreased glucose level in the WD group, which was ameliorated by strains supplementation, resulting in glucose levels equivalent to those in the normal control group. It is also worth noting that a signi cantly higher content of mannose was identi ed in the normal control and strain-fed groups. The bene cial effect of dietary supplementation of mannose is associated with the modulation of gut microbiome, leading to prevention of diet-induced obesity and amended host metabolism [36].
In our study, the diagnostic model utilizing biomarkers were shown to have promising results in AUC (0.782-0.868). Although there are few studies related to biomarkers through fecal metabolomics analysis of nonalcoholic fatty liver disease with elevated liver enzymes, a recent study using microbiota data demonstrated diagnostic accuracy (AUC 0.87) for detecting brosis in NASH [37]. If the method using feces overcomes cost and time challenges, it will be used in various elds of liver disease in the future.

Conclusion
In summary, we investigated the intestinal microbial normalization and anti-in ammatory effects of L. lactis and P. pentosaceus in the NAFLD mouse model. Diet with the L. lactis and P. pentosaceus inhibited NAFLD progression via normalization of BA composition, which in turn was achieved by modulation of gut microbiota composition and production of anti-in ammatory metabolites such as indole compounds (Fig. 8B).