Gut Microbiome and Metabolites in Patients with NAFLD and after Bariatric Surgery: A Comprehensive Review

The prevalence of non-alcoholic fatty liver disease (NAFLD) is increasing, as are other manifestations of metabolic syndrome such as obesity and type 2 diabetes. NAFLD is currently the number one cause of chronic liver disease worldwide. The pathophysiology of NAFLD and disease progression is poorly understood. A potential contributing role for gut microbiome and metabolites in NAFLD is proposed. Currently, bariatric surgery is an effective therapy to prevent the progression of NAFLD and other manifestations of metabolic syndrome such as obesity and type 2 diabetes. This review provides an overview of gut microbiome composition and related metabolites in individuals with NAFLD and after bariatric surgery. Causality remains to be proven. Furthermore, the clinical effects of bariatric surgery on NAFLD are illustrated. Whether the gut microbiome and metabolites contribute to the metabolic improvement and improvement of NAFLD seen after bariatric surgery has not yet been proven. Future microbiome and metabolome research is necessary for elucidating the pathophysiology and underlying metabolic pathways and phenotypes and providing better methods for diagnostics, prognostics and surveillance to optimize clinical care.


Introduction
Non-alcoholic fatty liver disease (NAFLD) is considered the hepatic manifestation of metabolic syndrome, next to obesity, insulin resistance and hyperlipidemia. The prevalence of obesity is increasing worldwide, as is the prevalence of NAFLD [1]. The term NAFLD comprises a spectrum of liver diseases, ranging from hepatic steatosis to non-alcoholic steatohepatitis (NASH) with or without fibrosis [2]. Progression of NASH can ultimately lead to liver cirrhosis or hepatocellular carcinoma. Moreover, NAFLD is considered to become the number one cause of chronic liver disease worldwide, with end-stage fatty liver disease as the second indication for liver transplantation [3][4][5].
Consequently, health care costs of liver disease are increasing, including costs for regular follow-up necessary to detect disease progression. The early stages of fatty liver disease can be reversible with a healthy lifestyle and weight loss. Currently, bariatric surgery is a long-term effective therapy to prevent the progression of NAFLD and other manifestations of metabolic syndrome, such as obesity and type 2 diabetes [4,6].
However, the exact pathophysiology of fatty liver disease progression remains to be clarified. The gut microbiome has been proposed as one potential contributing factor in the pathogenesis of NAFLD. The gut microbiome consists of microorganisms present in the gastrointestinal tract. The bacteria produce metabolites which are transported to the liver, where they can act as signaling molecules and have systemic effects. In addition, several studies have shown an influence of the gut microbiome on metabolism and suggest a causal Actinobacteria [14]. The composition of the microbiome can be affected by diet, antibiotics, and other medication such as metformin and proton pump inhibitors [15].
One of the functions of the gut microbiome is to contribute to host metabolism and homeostasis. Therefore, a distinct balance between bacterial species should be maintained. Disturbance of this balance is recognized as dysbiosis and has been associated with gastrointestinal complaints and metabolic alterations. For example, in a comparison of compositions of the gut microbiome between those with normal glucose, impaired glucose, and a diabetic glucose control, differences were observed [16,17]. Methods to determine microbiome composition include 16S ribosomal RNA (16S RNA) sequencing of ubiquitous genes and whole genome shotgun sequencing of the entire gene component. Additional functional analysis of the microbiome can provide insight into the metabolic microbial capacity.
The gut microbiota are metabolically active. Ingested food is metabolized by specific microorganisms and gut microbiota are also able to excrete metabolites. These gut microbiome-derived metabolites are absorbed by enterocytes and enter the circulation where they are able to exert systemic effects ( Figure 1). Some of these metabolites are thought to be the key components in influencing host metabolism [18]. Dysbiosis, the disbalance of the gut microbiome, alters (diet-derived) metabolite production and thus crosstalk between the host and gut microbiome. Research on gut microbiome-derived metabolites helps to distinguish which ones are key players in regulating metabolism at cell level [18,19].
A major class of intestinal bacteria produced metabolites are short-chain fatty acids (SCFAs). They are produced during the fermentative activity of microbiota when digesting dietary fiber, and SCFA production is suggested to modulate intestinal pH and intestinal barrier integrity [20][21][22]. The main SCFAs are propionate, butyrate and acetate, which also function as signaling molecules in immune response regulation, intestinal homeostasis and energy metabolism [23]. Other classes of metabolites include amino acids such as glycine and its precursor serine, aromatase amino acids (AAAs) including phenylalanine and tyrosine, and branched chain amino acids (BCAAs) such as leucine or valine. BCAAs promote glucose uptake and skeletal protein synthesis [24,25]. However, increased serum levels of BCAAs are seen in individuals with insulin resistance, and two bacterial species were identified as the main species driving the association between the biosynthesis of BCAAs and insulin resistance [26]. The relationship between the gut microbiome and glucose metabolism has been reviewed elsewhere [27,28].
Gut microbiota conjugate primary bile acids, produced by hepatocytes, into secondary bile acids [31,32]. Bile salt hydrolase (BSH) is one of the microbial enzymes involved in driving the deconjugation of bile acids, with metabolites such as glycine and taurine as end products [33]. Besides direct involvement in fat digestion, secondary bile acids also function as signaling molecules in lipid, glucose and energy metabolism [34,35]. Most bile acids are actively reabsorbed from the intestinal lumen to the liver. This gut-to-liver axis plays a role in the enterohepatic circulation of bile acids, influencing host metabolism. The Farnesoid X receptor (FXR) is involved in bile acid synthesis control and enterohepatic circulation [36,37]. Bile acid-induced TGR5 activation decreases inflammation, and alterations of bile acid metabolism are associated with gut microbiota dysbiosis and obesity [33,38,39]. Furthermore, imbalance in bile acids is associated with gut barrier dysfunction [40].
Lipopolysaccharides (LPS) are endotoxins, components of the Gram-negative bacterial cell membrane, and play a role in immune response [41]. They can also be transported beyond the liver (that filters most of these) into the systemic circulation, where they dysregulate the inflammatory tone and contribute to metabolic disease [42].
The development of advanced sequencing techniques and the use of machine learning has increased the possibilities to gain insight in the human microbiome and derived metabolites and differences between specific populations, forming the foundation of a whole new scale of possible therapies [43][44][45]. The influence of the gut microbiome in diseases such as diabetes mellitus or inflammatory bowel disease has been intensely investigated [46,47]. Intervention studies with prebiotics, probiotics, antibiotics and fecal microbiota transplantations (FMT) focus on elucidating a potential causal relationship between gut microbiome and development as therapy for various human diseases as well.

Fatty Liver Disease
As briefly mentioned before, NAFLD includes various forms of fatty liver disease and can progress to severe and irreversible liver disease ( Figure 1). Fatty liver disease encompasses hepatic steatosis and non-alcoholic steatohepatitis (NASH). NAFLD is defined as lipid droplet accumulation in more than 5% of hepatocytes, and is sometimes referred to "simple" steatosis [2]. NASH, characterized by inflammation, includes steatohepatitis with and without fibrosis and requires a liver biopsy for diagnosis. NASH is histologically characterized by inflammation and hepatocyte ballooning in combination with steatosis. Histological evaluation of liver biopsies is performed via scoring systems to classify liver disease, such as the NAFLD Activity Score (NAS) and the Steatosis Activity and Fibrosis Score (SAF). Another frequently used score is the NASH Clinical Research Network (NASH CRN) scoring system: F0 = no fibrosis; F1 = perisinusoidal or portal/periportal fibrosis; F2 = perisinusoidal and portal/periportal fibrosis; F3 = bridging fibrosis; and F4 = cirrhosis. European guidelines recommend the SAF score because of its higher accuracy in distinguishing the intermediate category from mild and severe fatty liver disease [48].
It is important to realize that although NAFLD is seen as the hepatic manifestation of metabolic syndrome, not all patients with NAFLD are obese, and several patients with obesity do not have NAFLD either [49].

Gut Microbiome in Non-Alcoholic Fatty Liver Diesease
Differences in composition of the gut microbiome of patients with fatty liver disease have been observed (Table 1) [50,51]. Compared to individuals without NAFLD, fatty liver disease is associated with an increased abundance of Gram-negative microbiota: the abundance of the phylum Proteobacteria is increased [52,53]. The genera Escherichia, Enterobacteriaceae, Dorea and Peptinophilus are enriched, and abundances of Anaerosporobacter, Coprococcus, Eubacterium, and Faecalibacterium and Prevotella, Rikenellaceae and Ruminococcaceae are decreased compared to healthy controls [52,[54][55][56]. Functional analyses of these microbiome differences reported increased microbial capacity for the metabolism of BCAAs and AAAs [52,[54][55][56]. Furthermore, increased intestinal permeability, which is associated with dysbiosis, is seen in NAFLD [57]. Increased intestinal permeability and inflammation was observed in NAFLD patients in combination with dysbiosis of the gut microbiota: higher abundance of Escherichia was observed in fatty liver disease [58]. Development of fatty liver disease to NASH is associated with increased innate immune activation and inflammation [59,60]. The specific inflammatory factors associated with NAFLD are reviewed elsewhere [61]. Furthermore, differences between the gut microbiome of the different manifestations of fatty liver disease are also reported. One study comparing the gut microbiome found higher abundance of the phylum Firmicutes in patients with mild NAFLD compared to advanced fibrosis. The species Eubacterium rectale and Bacteroides vulgatus were the most abundant species in the mild NAFLD group. The advanced fibrosis group had higher abundance of the phylum Proteobacteria and, the most abundant species were Bacteroides vulgatus and Escherichia coli. Ruminococcus obeum CAG: 39, R. obeum, and E. rectale were significantly lower in advanced fibrosis than in mild/moderate NAFLD. A study including 203 individuals compared patients with biopsy-proven NAFLD-cirrhosis to individuals with fatty liver without advanced fibrosis, and to controls without fatty liver disease, as determined via imaging. An enriched abundance of Streptococcus was observed in both NAFLD groups. Megasphaera was only enriched in NAFLD-cirrhosis, and the highest abundance in this group was family Enterobacteriaceae and genera Streptococci and Gallibacterium. Patients with fatty liver disease without advanced fibrosis and controls had similar abundances of Bacillus and Lactococcus, which was enriched compared to NAFLD-cirrhosis. Controls had a higher abundance of Faecalibacterium prausnitzii species compared to both NAFLD groups [62]. A recent prospective study, where fatty liver status was assessed via transient elastography, found decreased relative abundance of Clostridium (sensu stricto) in patients with liver fibrosis, and an enrichment of Enterobacteriaceae, Escherichia and Shigella compared to individuals with severe steatosis without fibrosis [63].
A study comparing liver biopsies of 57 patients with NAFLD reported that increased abundance of the family Bacteroidaceae and decreased abundance of the families Prevotellacea and Erysipelotrichaceae is associated with increased disease severity. Compared to patients without NASH, NASH was associated with decreased Prevotella and increased Bacteroides abundance. Fatty liver disease with significant fibrosis (F3/F4) is associated with the increased abundance of Ruminococcus and Bacteroides, and increased abundance of Prevotella compared to mild fibrosis [64]. In children, high abundance of Prevotella copri is associated with more severe fibrosis [65]. Another study found that compared to controls, the genus Collinsella was most strongly associated with patients with NASH, and Ruminococcoaceae, a SCFA-producing genus, was decreased [66].
Certain strains, such as the Proteobacteria species Klebsiella pneumonia, are associated with endogenous alcohol production and are found to be increased in fecal microbiome of individuals with NAFLD and NASH, and abundance decreased with clinical improvement of the fatty liver disease [9]. In children, NASH was associated with an increased abundance of Proteobacteria and increased peripheral blood ethanol levels. Between obese children and children with NASH, the microbiome abundance was only statistically different for the abundance of Proteobacteria, Enterobacteriaceae and Escherichia [52].
As already mentioned, NAFLD can also develop in non-obese patients. Additionally, it appears that microbiome and metabolome signatures of non-obese patients with NAFLD differ from those of obese patients [67]; however, this requires further study.

Metabolites in Fatty Liver Disease
The liver plays a central role in metabolism and homeostasis, and the majority of blood to it is supplied via the portal vein. The portal vein drains a large part of the gastrointestinal tract, and thus the liver is the first organ to process absorbed components, such as gut microbiome-derived metabolites.
A cohort study involving obese, non-diabetic women, associated hepatic steatosis with microbiome-derived metabolite phenylacetic acid [29]. Furthermore, low microbiome gene richness was correlated with plasma BCAAs leucine, valine and isoleucine and hepatic steatosis, and the microbial capacity for metabolism of BCAA and AAAs such as phenylalanine, tyrosine and tryptophan was increased in patients with steatosis [29]. Plasma choline and phosphocholine were not found to be negatively correlated with steatosis, but urinary choline excretion was increased [29]. However, TMA (the microbial processed form of dietary choline and carnitine) and hepatic (FMO3-processed TMA) TMAO were found to be inversely correlated with steatosis [29]. In a study with 86 patients with biopsy-proven NAFLD plasma metabolites, hypoxanthine and inosine were found to be enriched in individuals with mild NAFLD. In contrast, the plasma metabolites succinate, malate, alfa-ketoglutarate, glutamine, serine and fumarate (associated with carbon metabolism) were enriched in individuals with advanced NAFLD with fibrosis [53].
Analysis of the plasma metabolome from patients with NAFLD, NASH and cirrhosis identified several metabolites as biomarkers relevant for determining disease stage. The metabolites isocitric acid and isoleucine were decreased in controls and increased with disease progression. In contrast, xanthine, glutathione and glycolic acid were found to be higher in controls and decreased with disease progression. Valine (BCAA), asparagine, propanoic acid (SCFA), butanoic acid (SCFA), phenylalanine (AAA), palmitic acid (FA), stearic acid (FA) and taurocholic acid (BA) were also identified as relevant for the metabolic signature of NAFLD; these metabolites are linked to pathways involved in bile acid, lipid, and amino acid metabolism. With increases in disease severity, taurocholic acid, phenylalanine and BCAAs increase, while glycolic acid (SCFA) and glutathione decrease [68]. Interestingly, a deficiency of serine, a precursor of glycine which is a precursor of glutathione, is associated with fatty liver disease [69,70]. In rodents, treatment with glycine improves NAFLD via glutathione synthesis [71].
SCFAs are associated with anti-inflammatory properties, in accordance with the observed decrease in NAFLD disease severity [72]. In contrast, BCAAs are associated with inflammation and insulin resistance [25,73]. Increased serum levels of BCAAs are seen in individuals with insulin resistance, and Prevotella copri and Bacteroidetes vulgatus have been identified as the main species for the association between the biosynthesis of BCAAs and insulin resistance [26]. Insulin resistance is associated with increased hepatic de novo lipogenesis, contributing to fatty liver disease [74]. Although Masarone and colleagues did not describe gut microbiome data [68], the reported metabolites such as BCAAs and AAAs, SCFAs and bile acids are in line with other studies on the gut microbiome, metabolites and NAFLD, as described in this review ( Table 1).
As mentioned before, bile acids are metabolized and reabsorbed in enterohepatic circulation. Only approximately 5% of bile acids are excreted in the feces, and thus can be measured as stool metabolites. Higher serum and fecal bile acids levels are associated with advanced NAFLD and fibrosis in obese patients [75,76]. Increased serum bile acids were primary conjugated bile acids such as glycocholic acid (GCA) and secondary conjugated bile acids [75]. Increased fecal bile acids were mainly secondary unconjugated bile acids such as deoxycholic acid (DCA). Serum CGA levels and fecal DCA levels correlated with the abundance of Bacteroidaceae and Lachnospiraceae [75]. In patients with NASH, total primary bile acids were elevated and secondary bile acids were decreased [77]. In patients with fibrosis, primary bile acids were also elevated, mainly, glycine-conjugated bile acids, but secondary bile acids were the same compared to patients without fibrosis [76]. Glycogenodeoxycholic acid (GCDCA) and glycocholic acid (CGA) had strongest association with advancing fibrosis grade, as well as secondary bile acids 7-keto-deoxycholic acid (7-Keto-DCA) and glycoursodeoxycholic acid (GUDCA) [76]. Patients with borderline NASH (NAS = 3-4) had higher levels of total bile acids, total primary bile acids, and primary glycine-conjugated bile acids compared to simple steatosis (NAS = 1-2); no differences were observed between simple steatosis and definite NASH (NAS 5-8) [76]. Secondary bile acids 7-Keto-DCA and 7-Keto-lithocholic acid (7-Keto-LCA), both formed by microbial transformation, were increased in patients with definite NASH [76]. Primary bile acids, including total glycine-and taurine-conjugated bile acids, were increased in lobular inflammation and were also more increased with higher inflammation grade [76]. In steatosis, serum total cholate/chenodeoxycholate ratio is increased. Compared to steatosis or controls, in patients with NASH, total conjugated primary bile acids, conjugated/unconjugated chenodeoxycholate, cholate, and total primary bile acids are increased [77].

Causality
Causality can be established with intervention studies. One of the first indicators for causality regarding gut microbiome and metabolic syndrome was when the transfer of intestinal microbiota of lean donors to patients with metabolic syndrome increased the insulin sensitivity [8,78,79]. Interestingly, metformin alters the gut microbiome of individuals with treatment-naive type 2 diabetes, probably contributing to the therapeutic effects of the drug [80]. However, most studies are cross-sectional and therefore cannot imply causality. A causal role for the gut microbiome and metabolites in the pathogenesis of NAFLD was indicated when donor feces of patients with hepatic steatosis induced steatosis after transplantation into germ-free mice. In addition, chronic treatment with phenylacetic acid also triggered steatosis [29]. FMT of human NAFLD gut microbiota into mice promoted the progression of NASH, by increasing the accumulation of intrahepatic B-cells, thus suggesting gut microbial-driven factors contributing to hepatic inflammation and fibrosis [81].
A recent publication studied the effect of FMT from vegan donors on histologic improvements of liver biopsies from patients with NAFLD. Although no significant histological changes were observed, positive changes in liver gene expression level were detected, as well as changes in metabolites and gut microbiome [82]. Another recently published study showed that donor FMT can improve liver stiffness in obese subjects [83]. Although promising, larger studies are needed to support these data.

Bariatric Surgery and Other Weight Loss Interventions
Bariatric surgery is a successful therapy for obesity and related comorbidities, resulting in permanent weight loss and improvements of metabolic and inflammatory status, such as insulin resistance and fatty liver disease [86][87][88][89]. The most performed types of surgery are laparoscopic Roux-en Y gastric bypass (RYGB) and laparoscopic sleeve gastrectomy (SG). Other surgical interventions include omega-loop gastric bypass (OAGB), biliopancreatic diversion, adjustable gastric banding (AdGB) and vertical banded gastroplasty (VGB).

Gut Microbiome in Obese Population and after Bariatric Surgery
The gut microbiome of individuals with obesity differs from that of lean individuals [90]. The gut microbiome of obese patients has increased the capacity for energy harvest [91]. Bacterial diversity and gene richness are usually decreased in patients with obesity, and microbial gene richness is inversely correlated with the severity of obesity [7]. Moreover, a small but significant negative association between microbial species alpha diversity (the species variation at individual sample level) and BMI has been reported [90].
After bariatric surgery, both anatomy and physiology rapidly adapt to the new physiological state [93,94]. The specific mechanisms of how bariatric surgery affects the composition of the gut microbiome, however, remains to be clarified. However, changes in diet habits, gastrointestinal anatomy, nutrients and gastrointestinal transit time, bile acid metabolism and gastrointestinal pH are probably all contributing factors in altering the gut microbiome [94]. Studies show that bariatric surgery increases microbial gene richness, although it still remains low and does not increase to the average richness seen in lean controls (Table 2) [7,95].
Several studies investigated the influence of RYGB on gut microbiome. In a study of 16 patients who underwent RYGB, the gut microbiome before and 3 months after surgery was analyzed. Before surgery, the phyla Firmicutes and Actinobacteria were more abundant and Verrucomicrobia was less abundant compared to lean subjects. After RYGB, the abundance of these phyla was similar to the healthy controls. Only Proteobacteria abundance was enriched after RYGB and lower in controls [96]. Thus, compared to before surgery, the phyla Firmicutes, Verrucomicrobia and Proteobacteria were increased, and the phylum Actinobacteria was decreased [96], as was Bacteroidetes [95]. At the genus level, Blautia, Roseburia, Faecalibacterium (Firmicutes) and Bifidobacterium (Actinobacteria) were decreased; however, these genera were still more abundant when compared to lean controls [96]. Furthermore, at the species level, Streptococcus spp., Akkermansia muciniphila (Verrucomicrobia), Roseburia feces, Roseburia hominis and Enterococcus faecalis were increased [96,97], and Faecalibacterium prausnitzii was decreased after RYGB [97,98]. F. prausnitzii, a butyrate producer, was associated with beneficial effects on host metabolism and negatively correlated with inflammation markers [99][100][101][102].
Studies specifically analyzing gut microbiome composition after SG found enriched abundance of the species C. comes, D. longicatena, Clostridiales bacterium, Anaerotruncus colihominis, Akkermansia muciniphila and B. thetaiotaomicron [92,103]. Relative abundance of Firmicutes, Prevotella and Bacteroides fragilis was decreased, and the abundance of species Akkermansia muciniphila, Roseburia spp., Bacteroidetes, and Bifidobacterium was increased. One year after SG, the abundance of phylum Actinobacteria was increased compared to baseline and three months post-operation [103].
After both RYGB and SG, the species Escherichia coli (Proteobacteria), Klebsiella pneumoniae (Proteobacteria) and Haemophilus parainfluenzae (Proteobacteria) were increased [97]. Abundance of aero-tolerant bacteria, such as Escherichia coli and buccal species, such as Streptococcus and Veillonella spp., was relatively higher after RYGB. In contrast, anaerobes such as Clostridium were more abundant after SG. Akkermansia muciniphila was enriched in both surgeries [97]. Functional analysis found that pathways involved in cholesterol transporters, nitrate respiration and propionate production via kinase increased in RYGB, and glutamate degradation module was more abundant in SG [97].
Long-term effects of RYGB on the gut microbial composition in seven women nine years after surgery showed enriched Gammaproteobacteria and Proteobacteria, and at the genus level, increased relative abundance of Escherichia, Klebsiella and Pseudomonas. Three species of Firmicutes, Clostridium difficile, Clostridium hiranonis and Gemella sanguinis, were lower after RYGB [104]. Functional analysis found enriched microbial genes associated with TMAO in both RYGB and ABG compared to controls, but only after RYGB increased was plasma TMAO observed.
Some patients have considerable weight regain after RYGB, and the gut microbiome may be involved. A study comparing patients at least 5 years after RYGB with and without weight regain found differences in the composition of the gut microbiome [105] In the non-regain group, higher Akkermansia genus abundance was found, compared both to the control group and weight-regain group. Compared to the control group, the non-regain group also had a higher abundance of Phascolarctobacterium genus and lower abundance of the SMB53 genus. Bacteroidetes still had a lower abundance after RYGB, both with and without weight regain compared to the control group, and in the control group genera Bacteroides and SMB53 were increased [105].
Multiple studies report changes in microbial composition after bariatric surgery. Different interventions showed different microbial profiles and only partial restoration towards lean microbiome composition. However, variations in results were observed, possibly due to different methods, small sample sizes, and/or comorbidities such as type 2 diabetes [106,107].

Plasma Metabolites after Bariatric Surgery
Plasma metabolites such as tyrosine, phenylalanine, leucine, isoleucine, valine and glutamate are all increased in patients with obesity [92]. Plasma glycine and glutamine levels are lower in obese patients [73]. Lower branched chain fatty acids (BCFAs) and higher BCAA levels are found in obese patients compared to controls [109]. Alterations of SCFA production are thought to play a contributing role in the development of obesity [20,110]. Colonic infusion with SCFA mixture increased fat oxidation and energy expenditure and decreased lipolysis in obese men [111]. Studies correlate metabolites with findings in the gut microbiome. For example, Liu et al. reported a decreased abundance of Bacteroides thetaiotaomicron, a glutamate-fermenting commensal, in Chinese patients with obesity, and this decreased abundance was inversely correlated with serum glutamate concentration [92].
As described previously, the obesity-associated gut microbiome after bariatric surgery is partially restored towards that of lean patients [92]. This phenomenon can also be observed in metabolite levels ( Table 2). After both SG and RYGB, the production of AAAs and BCAAs is reduced and normalizes towards lean control levels [92]. Glycine levels relatively increase and normalize towards lean control levels [92]. In addition to the BCAAs leucine, isoleucine, and valine, the aromatic AAs phenylalanine and tyrosine, as well as ornithine, citrulline, and histidine, all decreased after RYGB. In a study comparing diet intervention with RYGB, this effect was not seen after diet intervention [108]. Furthermore, during 50 weeks of dietary weight loss, plasma SCFA levels did not change; only acetate concentration decreased with overall weight loss [112].
Correlation analysis of metabolites and clinical parameters after RYGB indicated that a cluster of metabolites, including glycine, acetylglycine, and methylmalonate, were increased and were negatively correlated with decreases in body corpulence and adipocyte diameter [7]. After RYGB and AGB, acetylglycine and glycine were negatively associated with improved weight and body composition [7]. Corresponding pathways involved in carbohydrate fermentation, the citrate cycle, glycosaminoglycan degradation, and LPS synthesis were normalized towards levels of lean controls. In patients with type 2 diabetes who underwent RYGB, plasma LPS, the endotoxin associated with insulin resistance and increased inflammatory tone, was reduced by 20% during follow-up [113]. Additionally, inflammatory stress markers decreased, and gut permeability was reduced after RYGB. Although an increase in Gram-negative, LPS-producing Proteobacteria was seen after RYGB, plasma LPS decreased, possibly due to the decreased gut permeability in combination with the increase in SCFAs, which also decreased gut permeability.
When correlated with gut microbiome composition, plasma metabolites positively associated with microbial gene richness are 3-methoxyphenylacetic acid, phloretate, hippurate, 3-hydroxphenylacetate, L-histidin, and three unidentified metabolites. Specifically, plasma glutamate is negatively correlated with microbial gene richness [7]. Changes in methylmalonate and glycine are significantly correlated with the change of Bacteroides finegoldii and Coprobacillus spp., which were also associated with improvements in body composition. Moreover, stool metabolite SCFAs acetate, propionate and butyrate were decreased after RYGB and VBG. After RYGB, the SCFA/BCFA ratio was decreased without changes in total fiber intake and with decreased protein intake, indicating that the findings were not due to dietary consumption [104].
Finally, plasma bile acids are also altered after SG and RYGB [84,114]. Especially, after RYGB, the bilio-enteric flow is altered with changes in plasma bile acid concentrations and decreases in FGF-19 [114]. Changes in bile acid profile after laparoscopic sleeve gastrectomy are associated with improvements in metabolic profile and fatty liver disease (as measured via serum cytokeratin 18) [84]. A study analyzing the gut microbiome and serum bile acids of seven patients after RYGB with controls provided evidence for an increased abundance of microbial genes involved in converting primary to secondary bile acids [104].
Both SG and RYGB lead to histological improvements of NAFLD, as scored by NAS [117]. A prospective analysis on the effect of bariatric surgery on hepatic inflammation and fibrosis, as assessed in liver biopsies from 32 patients, found significant improvements in steatosis, lobular inflammation, ballooning, and fibrosis, and NASH was resolved in three out of four patients [118]. A recent large, retrospective, cohort analysis reported decreased progression of NAFLD to cirrhosis after bariatric surgery: 1.7% of the nonsurgical population and 0.5% of the bariatric surgery cohort had progression to cirrhosis [119]. In a retrospective analysis comparing RYGB with diet intervention on the potential effects on the severity and course of NASH, a 93.3% regression of NASH was shown after RYGB, as well as a 27.3% regression after diet [120]. In addition, bariatric surgery can be performed in selected patients with liver cirrhosis [121,122]. However, with more profound fibrosis, the long-term net clinical benefit compared to conservative weight loss therapy needs to be determined.
Bariatric surgery has positive effects on the clinical parameters of NAFLD; therefore, the pathophysiological mechanism has been studied intensively. However, studies focusing on evaluating the effect of microbiota of individuals before and after bariatric surgery on NAFLD are scarce. Whether gut microbiomes and metabolites contribute to the metabolic improvement and improvements of NAFLD seen after bariatric surgery has not been proven. For example, the plasma metabolite glycine is relatively decreased in individuals with obesity or NAFLD, and after bariatric surgery, it increases. Glycine levels are inversely correlated with BCAA levels. In rodent studies, glycine suppletion improved fatty liver disease [71], but suppletion did not lower elevated BCAA levels [123,124]. The glycine increase after bariatric surgery possibly contributes to the beneficial effect on NAFLD. Furthermore, a low-isoleucine diet in rodents reprogrammed liver and adipose metabolism, increasing hepatic insulin sensitivity and increasing energy expenditure [125]. However, further research in humans is necessary to validate hypotheses regarding the underlying metabolic processes of how bariatric surgery affects the pathophysiology of fatty liver disease. As mentioned before, overlap of differences in microbial composition exists between different metabolic diseases, including NAFLD and diabetes. To determine the exact isolated microbial effect of bariatric surgery on NAFLD, including its bacterial metabolites, more research is necessary in human studies with large sample sizes to find the subtle differences between the metabolic diseases. Additionally, it is expected that adding machine learning bioinformatics will further increase insights into genes and pathways involved in the improvement of NASH after RYGB [126].

Discussion and Future Perspectives
Gut microbiome and plasma metabolite research is rapidly evolving, and heterogeneous new methods have been introduced. This certainly sometimes contributes to contradictory findings. Study population, chosen time points after intervention, differences in the handling of samples and storage also have an impact on study results. Validation and reproduction of results is necessary before findings can be declared as facts. Correct sample sizes with sufficient power can contribute to that. Furthermore, conflicting results and overlap of altered microbiota between metabolic diseases such as obesity, NAFLD, and type 2 diabetes are complicating factors in interpretation and in allocating findings to a distinct signature per disease: a recent review focused on microbial signatures of patients with NAFLD and the overlap with other metabolic disorders [54,67]. Thus, further research is necessary for validation and the continued comparing and interpreting of results.
Gut microbiome and metabolome research continues to evolve in two directions. First, it is evolving towards better understanding of the pathophysiology and underlying metabolic pathways of various diseases and phenotypes, using an integration of metabolome, microbiome, and other omics data [127]. For example, besides the gut microbiome, the intrahepatic microbiome and its relationship with fatty liver disease has been investigated to better understand the gut-liver axis [128,129].
Secondly, the research is evolving towards improving clinical care via determining biomarkers (serum and/or stool) for providing better methods for diagnostics, prognostics and surveillance. Furthermore, the possibilities of clinical therapies with next-generation (bacterial strain based) probiotics, albeit via fecal microbiota transplantations or oral probiotics, are explored [130]. One study showed that administration of the probiotic Akkermansia muciniphila reduces insulin resistance in overweight individuals [131]. A randomized controlled trial of patients with NAFLD showed that 1 year of administration of a synbiotic combination (probiotic and prebiotic) with Bifidobacterium animalis subspecies lactis BB-12 as a component, altered the fecal microbiome but did not reduce liver fat content or markers of liver fibrosis [132]. An RCT including patients with NAFLD scheduled for SG did not find improvements of hepatic, inflammatory or clinical outcomes when comparing the effect of generic probiotics and placebo [133]. Limitations of intervention therapies with microorganisms include the inability to control the development of the particular gut microbiome, specific abundances, or optimal ratios. Furthermore, antibiotic therapies have taught us that influencing the gut microbiome comes with adverse effects. Nowadays, the only accepted microbiome intervention to restore dysbiosis is the treatment of Clostridium difficile [134,135]. Developing therapies focused on specific receptors identified in pathophysiological studies could enable more specific interventions, with fewer uncontrolled influencing variables and thus less variance in preferred outcomes.
Although outside the scope of this review, animal studies can be used to further develop and evaluate hypotheses [136][137][138][139]. For example, a recent study in rats concluded that RYGB led to greater liver fat loss compared to a low-calorie diet, possibly due to increased fasting bile acid levels and the increased expression of modulators of liver fat oxidation, FXR and PPARα (peroxisome proliferator activator receptor alpha) [140]. Intervention with resveratrol altered the gut microbiota and improved hepatic steatosis and insulin resistance in mice [141]. Whether microbiota and their metabolites are a causal factor interacting with pathophysiological processes remains to be demonstrated. Intervention studies including FMT studies continue to explore the causality of the gut microbiome and derived metabolites in fatty liver disease [82,142].
Thus, the positive effects of bariatric surgery on NAFLD and its underlying mechanisms become more elucidated every year. Bioinformatics are used to exert insights into genes and pathways involved in the improvement of NASH after RYGB [46]. However, most patients with NAFLD will not undergo bariatric surgery. Therefore, expanding knowledge and the translation of gained insights is necessary to optimize general care. Although research can focus and reveal processes that cannot be seen by the human eye, we cannot escape the bigger picture of the impact of metabolic syndrome and NAFLD: there are limits to what health care can comprehend and support. The boundaries of resilience have become painfully clear during the ongoing COVID-19 pandemic: simple steatosis is not as benign as assumed [143].  Acknowledgments: Figure 1 was created with BioRender.com.

Conflicts of Interest:
M.N. is on the Scientific Advisory Board of Caelus Health, the Netherlands, and Kaleido Biosciences, USA. None of these possible conflicts of interest bear direct relation to the content of this review. All other authors declare that they have no competing interests.