Gut metabolites and inflammation factors in non-alcoholic fatty liver disease: A systematic review and meta-analysis

The interaction of gut microbiota, related metabolites and inflammation factors with nonalcoholic fatty liver disease (NAFLD) remains unclearly defined. The aim of this systematic review and meta-analysis was to synthesize previous study findings to better understand this interaction. Relevant research articles published not later than September, 2019 were searched in the following databases: Web of Science, PubMed, Embase, and Cochrane Library. The search strategy and inclusion criteria for this study yielded a total of 47 studies, of which only 11 were eligible for meta-analysis. The narrative analysis of these articles found that there is interplay between the key gut microbiota, related metabolites and inflammation factors, which modulate the development and progression of NAFLD. In addition, the results of meta-analysis showed that probiotic supplementation significantly decreased tumor necrosis factor-α (TNF-α) in NAFLD patients (standardized mean difference (SMD) = −0.52, confidence interval (CI): −0.86 to −0.18, and p = 0.003) and C-reactive protein (CRP) (SMD = −0.62, CI: −0.80 to −0.43, and p < 0.001). However, whether therapies can target TNF-α and CRP in order treat NAFLD still needs further investigation. Therefore, these results suggest that the interaction of the key gut microbiota, related metabolites and inflammation factors with NAFLD may provide a novel therapeutic target for the clinical and pharmacological treatment of NAFLD.

Study characteristics. The characteristics of the 47 eligible studies are summarized in Appendices 2-4, and in Table 1. Specifically, 10 studies were included to demonstrate gut microbiota in NAFLD and 12 studies to demonstrate gut microbial metabolites in NAFLD (Appendices 2, 3). In addition, gut microbiota in NAFLD has been described in terms of its phylum, family, and genus. Meanwhile, the different roles played by different gut microbial metabolites in NAFLD have been described as well as the main results of the corresponding studies. Furthermore, the relationship between gut metabolites and inflammatory factors in NAFLD, and the main results including different gut metabolites that would induce different inflammatory factors in NAFLD, have been described in Table 2. Moreover, in relation to the NAFLD and its inflammatory factors after probiotics treatment, Figure 1. Flowchart of study selection. Showing the process by which relevant studies were retrieved from the databases, assessed, and selected, or excluded. Preferred reporting items for systematic reviews and metaanalyses (PRISMA) diagram for study search. Gut metabolites and inflammatory factors in NAFLD. The gut bioactive metabolites include the bacterial DNAs, endotoxins, lipopolysaccharides, peptidoglycan and extracellular vesicles, which are derived from the gut microbiota as well as the metabolites such as short-chain fatty acids (SCFAs), branched-chain amino acids, bile acids, cholic acid, indole, butyrate, propionate, endogenous ethanol and trimethylamine-N-oxide (TMAO). Appendix shows the details. Table 2 summarizes the association between gut metabolites and inflammatory factors in modulating the pathological process of NAFLD. Among them, endotoxin and lipopolysaccharide increase the release of a range of inflammatory factors including IL-6, IL-12, IL-1β and TNF-α in the development of NAFLD. Additionally, bile acids and cholic acids not only had direct antibacterial properties, but also they mitigated inflammatory response through the inhibition of nuclear factor kappa-B (NF-κB) signaling and cytokines generation in macrophages. Meanwhile, peptidoglycan from gram-positive bacteria promoted inflammatory response. Bacterial DNA played an important role in activating immune cells such as NK cells, dendritic cells, macrophages, B cells and macrophages, followed by the secretion of IL-12 and TNF-α. Studies have also shown that indole-3-acetic acid dose-dependently reduced lipopolysaccharide-induced proinflammatory cytokines, including MCP-1, IL-1β, and TNF-α.
The effects of probiotic therapies on inflammatory factors in NAFLD. Table 1 presents the NAFLD and its association with inflammatory factors after probiotics treatment, while Fig. 3 presents forest plots of the results of the Hereditary obesity, insulin resistance and NAFLD may be prevented or reversed by glycine-β-muricholic acid, an intestinal FXR antagonist, which possesses the ability to change the intestinal bacterial composition.  inflammatory factors and their association with NAFLD between the probiotics group and placebo group. It was observed that probiotics treatment (without additional intervention) decreased CRP (SMD = −0.62, CI − 0.80 to −0.43, p < 0.001), with no heterogeneity (I² = 0); and levels of TNF-α (SMD = −0.52, CI − 0.86 to −0.18, p = 0.003) but with considerable heterogeneity (I² = 79%). However, the two groups did not differ significantly with respect to IL-6 concentrations. Sensitivity analysis showed that there was a slight change in the SMD and its corresponding 95% CI when each study was removed in turn, indicating that the current meta-analysis data were relatively stable. Furthermore, only studies that included the TNF-α group were eligible for publication bias assessment, which was done by Egger funnel plot and Egger's test (Appendix). Accordingly, there was no significant publication bias detected (t = −0.61, df =10, p = 0.5539). Moreover, the results of risk of bias were, A, B, and C, implying that the risk of bias was negligible (Appendix).

Discussion
This study has revealed that there may be an interaction of gut microbiota, related metabolites and inflammation factors with NAFLD. In addition, this study assessed the efficacy of probiotic therapies in modifying inflammation factors in NAFLD. www.nature.com/scientificreports www.nature.com/scientificreports/ As regard to gut microbiota and gut metabolites, although gut microbiota of humans was highly variable, this review has shown that patients with NAFLD reported more harmful bacteria (such as Gammaproteobacteria) and less beneficial bacteria (such as Bifidobacteriaceae) than the control group 17 . Considering several contradictory findings of related studies with similar designs (such as Lachnospiraceae and Prevotellaceae), additional large-scale studies on the gut bioactive metabolites and metagenomics characteristics of gut microbiota in NAFLD are needed to confirm these results 16 .
It has long been known that high-fat diet could induce gut dysbiosis, and prolonged gut dysbiosis could produce different gut metabolites, including trimethylamine, SCFAs, glucagon-like peptide 1, lipopolysaccharide, free fatty acids, free cholesterol, and ethanol 10 . In addition, these gut metabolites may promote systemic inflammatory response, such as lipopolysaccharide, which can activate Kupffer cells to produce TNF-α and cause insulin resistance 13 . Thus, insulin resistance increases the accumulation of hepatic triglycerides by promoting triglyceride synthesis, hepatic uptake of free fatty acids and peripheral lipolysis in NAFLD 7 . Therefore, patients affected by NAFLD may have a gut dysbiosis, accompanied by small intestine bacterial overgrowth and elevation of gut metabolites such as endotoxin. Noteworthy, increased intestinal permeability may lead to increased bacterial migration and, under this condition, gut metabolites and some bacteria may enter the portal circulation by permeating the intestinal barrier and reach the liver 12,18 . Clinical studies have also shown that almost all patients with NAFLD had abnormal levels of inflammatory cytokines, which triggered the inflammatory response pathway of the gut microflora in NAFLD. Furthermore, experimental studies have shown that cytokine-mediated oxidative stress, mitochondria and endoplasmic reticulum dysfunction may promote the development of steatohepatitis from simple steatosis via the toll-like receptor signaling pathway 12,19 .
Accumulated evidence suggested that oxidative stress played an important role in the multi-hit hypothesis model of progression of NAFLD 18 . At the organelles level, mitochondria and endoplasmic reticulum are the main sites of ROS formation, producing pro-apoptotic peroxidized lipids (4-hydroxy-2-nonenal) and Bax, and promoting the synthesis of respiratory proteins through PGC1α as well as the transfer of Ca 2+ and ONOO 20,21 . Consequently, these may promote activation of stellate cells, macrophages, and Kupffer cells, which further promote endoplasmic reticulum and mitochondrial dysfunction, hence inducing inflammation, fibrosis and apoptosis in NAFLD 12 . Therefore, these results supported the hypothesis that interactions among the liver, the immune . Forest plot of inflammation factors between probiotics group and placebo group. Study effect sizes of IL-6, TNF-α and CRP concentration differences between NAFLD and controls. Each data marker represents a study, and the size of the data marker is proportional to the total number of individuals in that study. The summary effect size for each IL-6, TNF-α and CRP concentration is denoted by a diamond. NAFLD, nonalcoholic fatty liver disease; SMD, standardized mean difference. A, IL-6; B, TNF-α; C, CRP. (2020) 10:8848 | https://doi.org/10.1038/s41598-020-65051-8 www.nature.com/scientificreports www.nature.com/scientificreports/ system and the gut microflora related metabolites have a key role in the development of NASH from simple steatosis. An illustration of this hypothesis is shown in Fig. 4 (by KA).
Considering the gut metabolites and inflammatory factors in NAFLD, accumulating evidence has found that gut microbiota communities generate a variety of metabolites, which are bioactive substances that interact with the host liver cells in NAFLD 13 . Among them, endotoxin and lipopolysaccharide are active metabolites of gram-negative bacterial envelope components, and significantly increase the (mainly through activation of TLR4 signaling pathway) release of a range of inflammatory factors, including IL-6, IL-12, IL-1β and TNF-α in the development of NAFLD 22 . Additionally, exposure of the liver to endotoxin and lipopolysaccharide causes pathological and metabolic changes that induce an acute inflammatory response and early accumulation of inflammatory cells, such as monocytes, neutrophils and lymphocytes. Moreover, inflammatory cells release proteases, other enzymes and reactive oxygen metabolites, leading to pathological progression of liver injury in NAFLD. www.nature.com/scientificreports www.nature.com/scientificreports/ Furthermore, bile acids are regulators of lipid, glucose metabolism, and modulate inflammation in NAFLD. Therefore, the gut microbiota converts the primary bile acids (including chenodeoxycholic acid and cholic acid at the colon and distal small intestine) to secondary bile acids, such as ursodeoxycholic acidursodeoxycholic acid, lithocholic acid, and deoxycholic acid 23 . Also, bile acids can regulate the composition of gut microbiota, which in turn can regulate the capacity and composition of bile acids 24 . Thus, the destruction of bile acid-flora interaction can promote the development of NAFLD inflammation 25 . Meanwhile, studies have shown that farnesoid X receptor (FXR) plays an important role in bile acid-flora interactions. Therefore, bile acids are the ligands for the FXR, a member of a family of nuclear hormone receptors, widely found in the ileum and liver that regulate a variety of metabolic pathways 26 . The regulation of FXR signal has become a potential target for the prevention and treatment of NAFLD and related metabolic disorders 26 . Furthermore, not only do bile acids and cholic acids have direct antibacterial properties, but also mitigate inflammatory response through the inhibition of NF-κB signaling and cytokines generation in macrophages, mainly by activating Takeda-G-protein-receptor-5 27 . Meanwhile, peptidoglycan from gram-positive bacteria promotes inflammatory response, primarily through the activation of mitogen-activated protein kinase (MAPK)/NF-κB of the mediated proinflammatory cytokine generation. Subsequently, it also plays an important role in activating immune cells such as NK cells, dendritic cells, macrophages, B cells and macrophages. In addition, bacterial DNA activates MAPK/NF-κB (mainly through activation of TLR9 signaling pathway) 28 .
It is noteworthy that the hepatic steatosis is considered to be the benign beginning of the NAFLD, which is reversible without serious liver injury at this stage. Therefore, the preceding discussion in relation to the results of this study clarifies the roles of gut microbiota, related metabolites and inflammation factors in the pathophysiological process of pure steatosis of liver to NASH. It is hoped that this evidence may provide new insights for future researches and development of drugs targeted at preventing the hepatic steatosis progression to NASH, steatohepatitis-related associated cirrhosis and hepatocellular carcinoma. Interestingly, a recent meta-analysis also showed that probiotics reduced liver transaminase, total cholesterol, TNF-α, and insulin resistance in NAFLD patients 29 . However, that meta-analysis analyzed studies which considered only one inflammatory factor, TNF-α. In this regard, that meta-analysis did not investigate other inflammatory factor levels with respect to the association between probiotics and NAFLD, which has been addressed in this study. Therefore, this study has shown that there is an indirect evidence of the effects of the probiotic (beneficial gut microbiota) interventions on inflammation and liver damage, which indicated significant reductions in inflammatory factors (CRP and TNF-α) in NAFLD after probiotics were used. Nevertheless, there was no statistically significant difference in the concentrations of IL-6 between the NAFLD patients and the controls, which may be attributed to the small sample sizes involving IL-6. Therefore, large-scale clinical studies on this phenomenon are needed in the future.
Additionally, a systematic review suggested that pentoxifylline, an anti-TNF-α agent, directly targets the inflammatory process. In this regard, reduced AST and ALT levels may improve liver histological scores in patients with NALFD/NASH 30 . However, anti-TNF-α therapy for NAFLD may have significant side effects. Therefore, large, prospective and well-designed randomized controlled studies are needed to explore medications with better efficacy for NAFLD. Besides, new therapeutic targets for inflammatory signaling pathways are also worth exploring, as well as the treatment of NAFLD with CRP as the target.
Also, in agreement with the findings of this study, some studies have shown that activated carbon could reduce metabolites in NAFLD. For example, in addition to the probiotic supplementation, Yaq-001 (Yaqrit Ltd.) carefully designed and developed a novel synthetic (both macroporous and microporous) activated carbon for treating gut microbiota and related metabolites modulation in NAFLD 31 . In this regard, it has been shown that Yaq-001 (when administered orally) could reduce the transintestinal migration of gut microbiota and related metabolites, such as dimethylarginine, ammonia, bacteria-derived products, acetaldehyde, hydrophobic bile acids, and inflammation factors, including TNF-α and IL-6 31 . A safety and efficacy study on Yaq-001 is being conducted through the European Commission Horizon program. A clinical trial is also under way (NCT03962608). Future studies should focus on developing more specific novel therapeutics for NAFLD based on bioengineering technology with high efficacy and specificity. For example, utilizing targeted engineered individual microbiota instead of fecal transplants, or utilizing engineered microbiota capable of producing anti-inflammatory or antioxidants molecules to replace the antibiotics, or utilizing engineered compounds that modulate or adsorb the gut microbiota-derived metabolites of interest.
There are some limitations of this study, which should be acknowledged. Firstly, the heterogeneity was very high in the meta-analysis of TNF-α. Given that the subjects in this study came from different countries and ethnic groups, this heterogeneity may be due to the large individual differences in TNF-α concentrations, especially considering the significant side effects of TNF-α in the pathogenesis of NAFLD, hence the findings on TNF-α in this study should be interpreted with caution. Secondly, this study only analyzed the overall level of each study and did not quantify the important clinical indicators of individual patient level (also due to the lack of information in the original study), such as race, smoking, alcohol drinking and blood pressure, which could also be a reason for the large heterogeneity. Thirdly, the probiotic intervention was a well-designed experiment, and whether it is possible to replicate the results in the general population is unknown. Therefore, large population cohort studies are needed in the future to provide a causal relationship between intestinal flora and its metabolites and NAFLD.
Although there is currently no efficient treatment available for NAFLD, the findings of this systematic review/ meta-analysis suggest that the gut microbiota, related metabolites and inflammation factors (especially CRP and TNF-α) network presently appear to be innovative and potentially promising treatment targets. Future studies should focus on developing more specific novel therapeutics for NAFLD based on bioengineering technology with high efficacy and specificity.

Data extraction and preparation.
A standardized data extraction form was developed to record data of each eligible study in connection with the aims of this systematic review and meta-analysis. Therefore, data related to the following variables were extracted from each eligible study: (1) first author's name and year of publication; (2) subject's characteristics such as NAFLD or NASH; (3) characteristics of gut microbiota such as phylum, family, and genus; (4) characteristics of gut microbiota and its metabolites; (5) characteristics of gut metabolites and inflammatory factors; and (6) characteristics of inflammatory factors after probiotics treatment. Two researchers [KA and SW] independently extracted data from each study by using EpiData 3.0 and Microsoft Excel 2010. Discrepancies in this activity were resolved by [AL]. Data for the risk of bias assessment for the randomized studies was extracted using the Cochrane Collaboration's tool by two researchers [KA and SW] who also used Egger's test to assess the significance of the bias 33 .
Statistical analysis. Random effects model was used to assess the differences in concentrations of the inflammatory factors in NAFLD between the probiotic group and the placebo group. This model was chosen because it is suitable when synthesizing findings of studies with varying populations and criteria used to define outcomes, which was the case with the eligible studies for this meta-analysis. The R software (version R 3.5.2) was used to perform meta-analysis in the R packages, 'meta' and 'metafor' . Specifically, the group differences in the concentrations of the inflammatory factors in NAFLD were measured in terms of standardized mean difference (SMD), calculated as Cohen's d, and corresponding 95% confidence intervals (CI) 34 . A high effect size was represented by SMD > 0.8, whereas 0.5 ≤ SMD ≤ 0.8 and SMD < 0.5 represented moderate and low effect sizes, respectively. In addition, heterogeneity between-study was assessed using the Cochrane Q statistic and the level of heterogeneity was measured by the I² statistic 35,36 . Thus, low, moderate and high heterogeneity was represented by the I² statistic of less than 25%, 25~75% and greater than 75%, respectively 37 . Moreover, sensitivity analysis was repeated every time each study was omitted in turn 38 . All the statistical tests were two-sided and performed at the 5% significance level. In addition to meta-analysis, a narrative description was performed for: gut microbiota's phylum, family, and genus in NAFLD; the different roles played by different gut microbial metabolites in NAFLD; the relationship between gut metabolites and inflammatory factors in NAFLD; and different gut metabolites that would induce different inflammatory factors in NAFLD.