Microbiota, Liver Diseases, and Alcohol

ABSTRACT

Being overweight and obesity are the leading causes of liver disease in Western countries. Liver damage induced by being overweight can range from steatosis, harmless in its simple form, to steatohepatitis, fibrosis, cirrhosis, and hepatocellular carcinoma. Alcohol consumption is an additional major cause of liver disease. Not all individuals who are overweight or excessively consume alcohol develop nonalcoholic fatty liver diseases (NAFLD) or alcoholic liver disease (ALD) and advanced liver disease. The role of the intestinal microbiota (IM) in the susceptibility to liver disease in this context has been the subject of recent studies. ALD and NAFLD appear to be influenced by the composition of the IM, and dysbiosis is associated with ALD and NAFLD in rodent models and human patient cohorts. Several microbial metabolites, such as short-chain fatty acids and bile acids, are specifically associated with dysbiosis. Recent studies have highlighted the causal role of the IM in the development of liver diseases, and the use of probiotics or prebiotics improves some parameters associated with liver disease. Several studies have made progress in deciphering the mechanisms associated with the modulation of the IM. These data have demonstrated the intimate relationship between the IM and metabolic liver disease, suggesting that targeting the gut microbiota could be a new preventive or therapeutic strategy for these diseases.

METABOLIC LIVER DISEASES: FROM DIAGNOSIS TO CURRENT TREATMENTS

Pathogenesis of NAFLD and ALD

Excessive alcohol consumption and being obese/overweight are the leading causes of chronic liver disease in Western countries. Nonalcoholic fatty liver disease (NAFLD) encompasses all liver lesions that can be observed in overweight/obese patients, ranging from pure steatosis to steatohepatitis (nonalcoholic steatohepatitis [NASH]), fibrosis, cirrhosis, and even hepatocellular carcinoma (HCC). Alcoholic liver disease (ALD) defines liver lesions observed in patients with alcohol abuse and includes steatosis, hepatitis, fibrosis, cirrhosis, and HCC. It usually occurs when alcohol consumption is higher than 70 g/day in men and 60 g/day in women. However, only some chronic alcohol consumers develop liver injury, suggesting that factors other than excessive alcohol consumption play a role in ALD. Moreover, some patients with other liver diseases, such as chronic hepatitis C or NAFLD, also consume alcohol. Alcohol may lead to liver injury at a lower level of consumption in these patients with another underlying liver disease.

NAFLD usually occurs in patients with a metabolic syndrome. Such patients are overweight and may have hypertension, dyslipidemia, high blood glucose, or diabetes. However, as with chronic alcohol consumers, overweight or obese patients do not systematically develop NAFLD.

NAFLD and ALD encompass a wide spectrum of pathological lesions. Liver lesions in both diseases include a combination of steatosis, inflammation, hepatocyte necrosis, and fibrosis. Thus, NASH is merely a stage of NAFLD. Although the lesions are similar, the mechanisms of the two diseases share similarities, but there are also differences (1, 2).

Steatosis is an accumulation of triglycerides in the cytoplasm of hepatocytes. Steatosis is reversible after withdrawal of its cause (alcohol intake, being overweight), but if its cause remains, it may progress towards inflammation (Fig. 1). The presence of an inflammatory process in alcoholic patients with simple steatosis suggests that they will progress towards alcoholic hepatitis in the absence of alcohol withdrawal (3), whereas the transition from simple steatosis to steatohepatitis in overweight patients is less well understood (4). Another difference between the two diseases is the degree of the liver inflammation. The inflammatory infiltrate in both alcoholic and overweight patients may contain lymphocytes and polymorphonuclear neutrophils along with ballooning and necrosis of hepatocytes. Nevertheless, the intensity of this process is generally much higher in alcoholic than overweight patients. Megamitochondria is often observed in both diseases. The distinction of simple steatosis from true NASH may be difficult. The lesion is considered to be NASH when lobular inflammation and liver cell clarification/ballooning are present (5). The inflammatory process progresses towards fibrosis and cirrhosis in approximately 20% of alcoholic patients. The natural incidence of cirrhosis in overweight patients is less clear, possibly due to death from other diseases associated with being overweight, such as cardiovascular diseases. The incidence of HCC in cirrhotic livers is approximately 3 to 5% each year (6). HCC may occur without cirrhosis in overweight patients, whereas cirrhosis is an obligatory step in alcoholic patients (7). Severe alcoholic hepatitis is a particular form of ALD, characterized by acute hepatocellular insufficiency, severe histological liver injury, and a high mortality rate, as the 1-month survival of such patients is between 50 and 65%. Among patients who survive beyond 1 month, 50% die within the following year (8).

FIGURE 1.

Histology of the liver. Paraffin sections (4 μm thick) were stained with hematoxylin and eosin. Images were obtained using a Hamamatsu scanning module (Hamamatsu LX2000) and appropriate software (magnification, ×100). (A) Healthy tissue; (B) steatosis; (C) steatosis with inflammation; (D) fibrosis.

NAFLD and ALD: Clinical Aspects and Diagnosis

A diagnosis of NAFLD is clinically suspected in asymptomatic patients with elevated transaminases or radiological evidence of steatosis. Diagnosis of NAFLD may be associated with other components of metabolic syndrome, such as being overweight/obese, type 2 diabetes mellitus, hypertension, or dyslipidemia. When symptoms occur, they are nonspecific, consisting of fatigue and/or vague right-upper-quadrant abdominal pain. ALD should be suspected in all patients with excessive alcohol intake. The WHO safety threshold for alcohol consumption is approximately 20 g/day for women and 30 g/day for men, with 1 day of abstinence from alcohol each week. Higher consumption may lead to complications, but individual susceptibility to alcohol varies widely. For example, we have shown that overweight patients are more sensitive to alcohol-dependent liver toxicity, due to alcohol-induced inflammation in adipose tissue (9).

Elevated transaminases and gamma-glutamyltransferase are often the only abnormalities detected in overweight and alcoholic patients. A ratio of ALT to aspartate aminotransferase (AST) of <1 is suggestive of ALD in most cases, whereas an ALT/AST ratio of >1 suggests the presence of NAFLD. An ALT/AST ratio of <1 may also be observed when NAFLD progresses towards advanced fibrosis/cirrhosis. Transaminase values do not reflect the extent of liver injury, and a normal transaminase level does not guarantee the absence of underlying steatohepatitis or fibrosis. Conversely, increased transaminase levels are often associated with inflammation. The level of transaminases in ASH and NASH is rarely higher than twice the upper limit of the normal value (Table 1), and prothrombin time and bilirubin and albumin levels remain normal in the absence of hepatocellular insufficiency. Abdominal imaging aids in diagnosing steatosis. Abdominal ultrasound is noninvasive and can detect a steatosis involving more than 30% of hepatocytes (10). Computed tomography scan lacks sensitivity and introduces a radiation hazard. Magnetic resonance imaging (MRI) is expensive but provides the highest precision for quantifying steatosis. Nevertheless, liver imaging cannot distinguish between simple steatosis and NASH in the absence of liver dysmorphy due to cirrhosis (reviewed in reference 1).

TABLE 1.

Main differences between ALD and NAFLD

	ALD	NAFLD
Clinical characteristic	Alcohol intake	Overweight
		Metabolic syndrome
		Hypertension
Biology	ALT/AST < 1	ALT/AST > 1 in 66% of cases
		ALT/AST < 1 usually in patients with advanced fibrosis
		Dyslipidemia
		Increased blood glucose
Liver biopsy	Steatosis	Steatosis
	In patients with alcoholic hepatitis, high inflammatory process	In patients with NASH, usually low inflammatory process

Liver biopsy remains the gold standard to assess liver injury because of the lack of good correlations between simple blood tests and histological liver lesions. Noninvasive biological and physical tests, such as transient elastography, are often used for initial screening to limit the use of liver biopsy.

The mortality rate is higher in patients with NAFLD than in the general population (11). NAFLD is associated with an increased risk of cardiovascular disease and liver-related mortality.

Nevertheless, most patients with NAFLD display only simple steatosis, without inflammation. Their risk of progression from steatosis to steatohepatitis (i.e., NASH) is very low. These patients do not generally show a progression of liver injury and increased mortality risk after 20 years of follow-up. Conversely, patients with NASH may progress after several years towards fibrosis, cirrhosis, and HCC, with increased mortality. The three major causes of NASH-related mortality are cardiovascular diseases, all-cause malignancy, and liver-related death. NASH is associated with a >10-fold-increased risk of liver-related death (2.8% versus 0.2%) and a 2-fold-increased risk of death from cardiovascular disease (15.5% versus 7.5%) (12, 13). The presence and severity of fibrosis are the strongest determinants of long-term prognosis (14). The progression of fibrosis is generally slow, taking approximately 8 years to progress from no fibrosis to low-stage fibrosis. However, some patients progress more rapidly (15), similar to what is seen with those with chronic hepatitis C. Factors leading to more-rapid progression are a body mass index (BMI) of >30 kg/m², persistently abnormal liver enzyme levels with an AST/ALT ratio of >1, type 2 diabetes mellitus, hypertension, high triglycerides, low high-density lipoprotein (HDL) cholesterol levels, and a family history of diabetes mellitus (13, 16).

Of note, 30 to 60% of patients with NASH who develop HCC harbor advanced fibrosis without cirrhotic features (7, 17).

NAFLD: Current Treatments

Metabolic syndrome and its components are the primary cause of NAFLD. The management of NAFLD includes treatment of the risk factors that are commonly associated with metabolic syndrome through lifestyle modifications. No treatment is needed for simple steatosis, as it is not associated with increased morbidity. Thus, treatment should be considered in patients with NASH.

Dietary modifications leading to a weight loss of 7% or more may improve liver histology. A hypocaloric diet leading to a weight loss of 0.5 to 1.0 kg/week is generally recommended (18). Nevertheless, most patients do not achieve this weight loss goal.

Exercise, independent of weight loss, may also improve liver histology, decreasing liver fat content and increasing muscle mass (19). Lasting weight loss can be achieved in patients with NAFLD by combining diet and exercise for longer than 12 months (20). No specific treatment is recommended for NASH, despite several clinical trials. The best-studied treatments are insulin sensitizers, such as metformin, thiazolidinediones, and glucagon-like peptide 1 agonists (liraglutide); antioxidants, such as vitamin E; and cytoprotective agents, such as ursodeoxycholic acid (UDCA) and obeticholic acid (OCA) (21). Thiazolidinediones are peroxisome-proliferator-activated receptor gamma agonists. In addition to increasing insulin sensitivity, they promote peripheral fatty acid uptake and thus divert them from the liver towards adipose tissue. They improve steatosis, inflammation, and ballooning degeneration. No significant effect on liver fibrosis has been found. These drugs are no longer used in Europe due to their side effects: weight gain, increased risk of heart failure, and possible increased risk of bladder cancer (22). Metformin led to encouraging results in animal models, but no improvement of NASH has been observed in human clinical trials. However, metformin may decrease the risk of liver cancer and modulate the intestinal microbiota (IM), which may contribute to its beneficial effects. Metformin increases the abundance of the mucin-degrading bacterium Akkermansia, which may result in an improved metabolic profile in patients with type 2 diabetes (23). Vitamin E has shown a benefit in steatosis and inflammation, but not fibrosis, in nondiabetic patients with NAFLD (24). However, vitamin E does not improve insulin resistance in these nondiabetic patients and may increase all-cause mortality (24). UDCA may prevent hepatocyte apoptosis and downregulate inflammatory pathways. Improvement of liver transaminases following treatment with UDCA was shown in one clinical trial but not another, and there was only mild improvement of liver lesions, if any. OCA activates the farnesoid X receptor (FXR), whereas UDCA is a weak agonist of FXR. In a clinical trial, OCA improved liver transaminase levels and pathological lesions, fibrosis, hepatocellular ballooning, steatosis, and lobular inflammation. The safety of OCA must be tested, as it also increases alkaline phosphatase and low-density lipoprotein cholesterol (25). Bariatric surgery may be recommended in obese patients with a BMI of >40 kg/m² or >35 kg/m² with a comorbidity (26). Bariatric surgery improves liver histology and reduces mortality from both NASH-related complications and all-cause mortality in the 7 to 10 years that follow the bariatric procedure.

Overall, lifestyle modifications can provide a strong benefit to patients but are difficult to follow in the long term. Some treatments may be beneficial for patients, but the magnitude of the effects is small, in particular on liver fibrosis.

ALD: Current Treatments

Treatment of ALD relies principally on alcohol withdrawal. Nevertheless, patients with severe alcoholic hepatitis, defined by a Maddrey discriminant function of >32, must also be treated with corticosteroids, which is the reference treatment. Corticosteroids decrease liver inflammation and improve 1-month and possibly 6-month survival. In one of the largest trials to date, mortality after 28 days was 17% (45 of 269 patients) in the placebo group versus 14% (38 of 266 patients) in the prednisolone group (27). However, survival curves converged after 28 days, such that prednisolone therapy provided no benefit to patients after 90 days or 1 year. Pentoxifylline, a phosphodiesterase inhibitor, may reduce tumor necrosis factor alpha production. Its efficacy, alone or in association with corticosteroids, is controversial (28). It may increase survival relative to placebo but does not do so relative to prednisolone (29). N-Acetylcysteine is a glutathione precursor. In one trial, its addition to prednisolone increased survival of patients, but its efficacy needs to be confirmed (30). Enteral or parenteral nutritional supplementation may improve the prognosis, as ALD is often associated with protein, vitamin, and mineral deficiencies, but its impact on survival remains unclear (31). Nevertheless, nutritional support should be considered in patients with severe alcoholic hepatitis. Liver transplantation is generally proposed to patients with severe liver insufficiency after at least 6 months of abstinence. However, the mortality rate of patients with severe alcoholic hepatitis is very high (more than 50% during the first 3 months). Thus, it has been proposed to proceed rapidly to liver transplantation for patients with severe alcoholic hepatitis who do not respond to corticosteroids. In a clinical trial of 26 patients undergoing liver transplantation, the 6-month survival rate was 77%, similar to that of patients who respond to corticosteroids. After 2 years, 71% were still alive and only 3 patients resumed the consumption of alcohol (32).

Therapeutic resources for metabolic liver diseases are scarce, and the recent discovery of the involvement of the IM in NAFLD and ALD opens new therapeutic perspectives.

NAFLD AND THE GUT MICROBIOTA

NASH and Dysbiosis

Dysbiosis is defined as a microbial imbalance associated with a disease. Dysbiosis in NAFLD was first described in animal studies. For example, the severity of steatosis was associated with an increase in the Lactobacillus population in mice fed a high-fat diet (HFD) for 10 weeks (33). Since then, an increasing number of studies have investigated the gut microbiota in relation to human NAFLD, using culture-independent techniques, such as quantitative PCR (qPCR) or sequencing (Table 2). Children and adolescents (with or without NASH) were found to have a higher abundance of Bacteroidetes and a lower abundance of Firmicutes than healthy controls. Proteobacteria, Enterobacteriaceae, and Escherichia were the only phylum, family, and genus, respectively, exhibiting a significant difference between microbiomes from obese and NASH individuals and were found to be more abundant in patients with NASH (34). In adults, the comparison of 30 obese patients with clinically defined NAFLD and 30 nonobese controls revealed an overrepresentation of Lactobacillus species and selected members of the phylum Firmicutes (Lachnospiraceae, Dorea, Robinsoniella, and Roseburia) in NAFLD patients (35). Another study comprising 53 NAFLD patients and 32 healthy subjects found that Escherichia, Anaerobacter, Lactobacillus, and Clostridium cluster XI were more abundant in the gut microbiotas of NAFLD patients, whereas Alistipes, Odoribacter, Oscillibacter, and Flavonifractor were less abundant (36).

TABLE 2.

Microbiotas associated with different stages of NAFLD^a

Condition	Phylum	Family	Genus	Reference
NAFLD versus healthy controls	[Bacteroidetes]	Porphyromonadaceae ⇩		Healthy (n = 30, BMI = 22)NAFLD (n = 30, BMI = 33)Raman et al., 2013 (35)
	[Firmicutes]	Veillonellaceae ⇧
		Lactobacillaceae ⇧	Lactobacillus ⇧
		Ruminococcaceae ⇩	Oscillibacter ⇩
		Lachnospiraceae ⇧	Robinsoniella ⇧
			Roseburia ⇧
			Dorea ⇧
	[Proteobacteria]	Kiloniellaceae ⇧
		Pasteurellaceae ⇧
	[Bacteroidetes]	[Rikenellaceae]	Alistipes ⇩	Healthy (n = 32, BMI = 22)NAFLD (n = 53, BMI = 26)Jiang et al., 2015 (36)
		[Porphyromonadaceae]	Odoribacter ⇩
	[Firmicutes]	[Lactobacillaceae]	Lactobacillus ⇧
		[Oscillospiraceae]	Oscillibacter ⇩
		[Clostridiaceae]	Anaerobacter ⇧
		[Lachnospiraceae]	Clostridium XI ⇧
		[Streptococcaceae]	Streptococcus ⇧
			Flavonifractor ⇩
	[Proteobacteria]	[Enterobacteriaceae]	Escherichia ⇧
	Lentisphaerae ⇩
NASH versus healthy controls	Bacteroidetes ⇩			Healthy (n = 17, BMI = 26)
				NASH (n = 22, BMI = 32)
				Mouzaki et al., 2013 (37)
	Bacteroidetes ⇧	Prevotellaceae ⇧	Prevotella ⇧	Healthy (n = 16, BMI = 20)NASH (n = 22, BMI = 34)ChildrenZhu et al., 2013 (34)
		Rikenellaceae ⇩	Allistipes ⇩
	Firmicutes ⇩	Ruminococcaceae ⇩	Oscillospira ⇩
			Ruminococcus ⇩
		Lachnospiraceae ⇩	Blautia ⇩
			Coprococcus ⇩
			Eubacterium ⇩
			Roseburia ⇩
	Proteobacteria ⇧	Enterobacteriaceae ⇧	Escherichia coli ⇧	Healthy (n = 22, BMI = 22)NASH (n = 16, BMI = 29)Wong et al., 2013 (38)
	Actinobacteria	Bifidobacteriaceae ⇩	Bifidobacterium ⇩
	Firmicutes ⇩	Unclassified Clostridiales ⇩	Faecalibacterium ⇩
		[Ruminococcaceae]	Anaerosporobacter ⇩
		[Lachnospiraceae]	Allisonella ⇧
		[Veillonellaceae]
	[Bacteroidetes]	Porphyromonadaceae ⇧	Parabacteroides ⇧
	[Proteobacteria]	Succinivibrionaceae ⇧
NASH versus obese	Bacteroidetes ⇩	[Lachnospiraceae]	Clostridium coccoides ⇧	Obese (n = 11, BMI = 28)
				NASH (n = 22, BMI = 32)
	[Firmicutes]			Mouzaki et al., 2013 (37)
	Proteobacteria ⇧	Enterobacteriaceae ⇧	Escherichia coli ⇧	Obese (n = 25, BMI = 33)
				NASH (n = 22, BMI = 34)
				Zhu et al., 2013 (34)
NAFLD patients NASH versus no NASH	[Bacteroidetes]	Bacteroidaceae ⇧	Bacteroides ⇧	No NASH (n = 22, BMI = 30)
				NASH (n = 35, BMI = 32)
				Boursier et al., 2016 (122)

^a Microbiota were analyzed using 16S deep sequencing, except in the study by Mouzaki et al., in which qPCR was used. Additional taxonomic information for discriminating taxa is in brackets. Taxa in boldface are those identified in more than one study.

In a prospective, cross-sectional study, the gut microbiota composition (analyzed using qPCR) was compared between adults with biopsy-proven NAFLD (simple steatosis or NASH) and healthy controls. Patients with NASH had a lower proportion of Bacteroidetes than individuals with simple steatosis and healthy controls, even after adjusting for body mass index (37). Wong et al. investigated the microbiome composition using 16S rRNA gene sequencing in 16 biopsy-proven NASH patients and 22 healthy controls. The order Aeromonadales, the families Succinivibrionaceae and Porphyromonadaceae, and the genera Parabacteroides and Allisonella were found to be more abundant in NASH patients. Conversely, the class Clostridia, the order Clostridiales, and the genera Anaerosporobacter and Faecalibacterium were less abundant in NASH patients (38). Differences in fecal microbiota between nonobese adults with and without NAFLD have also been found, characterized by a higher abundance of Bacteroidetes and a lower abundance of Firmicutes in nonobese patients with NAFLD (39).

The effect of a choline-deficient diet on the composition of the human gut microbiome and the development of fatty liver has been studied by Spencer et al. They found that a baseline microbial composition correlated with the development of fatty liver in response to choline deficiency. The abundance of Gammaproteobacteria negatively correlated with the development of steatosis, whereas the abundance of Erysipelotrichia was positively associated, suggesting that baseline levels of these populations may predict susceptibility to fatty liver due to choline deficiency (40).

Altogether, these human studies revealed taxonomic differences between the gut microbiotas of healthy controls and patients with NASH. However, the results were heterogeneous, even contradictory, despite the identification of dysbiosis at the phylum or genus levels. Factors that may explain these differences include differences in the study design and methods of gut microbiota analysis, the variety of clinical endpoints, or changes of the gut microbiota due to diet or medications. The specific phyla or genera that are the causative agents in NASH development remain unclear.

A Causative Role for the Gut Microbiota in NAFLD

Patients suffering from NAFLD harbor a gut microbiota different from that of healthy controls, but the previously described association studies do not indicate whether this dysbiosis is a cause or a consequence of liver disease. Strategies have been developed in animal models to prove the causality of the gut microbiota in disease development. These strategies aim to transmit a phenotype from one mouse to another by a microbiota transfer that can be performed via cohousing of conventional mice or fecal transplant to germ-free or antibiotic-treated mice. These microbiota transfer strategies have already proven the causative role of the gut bacteria in several diseases including obesity and inflammatory bowel disease (41, 42). It was first demonstrated that cohousing germ-free mice with their conventional counterparts leads to increased glycogenesis in the liver of ex-germ-free mice prior to triggering increases in hepatic triglyceride synthesis. Colonization was also associated with altered hepatic Cyp8b1 expression and changes in bile acid (BA) composition. The hepatic triglyceride, glucose, and glycogen levels were strongly associated with the Coriobacteriaceae family (phylum Actinobacteria) (43). Henao-Mejia et al. further established the causative role of the gut microbiota in the development of liver disease using mouse models deficient in the proinflammatory multiprotein complex called the inflammasome (44). These inflammasome-deficient mice exhibited exacerbated NAFLD phenotypes when fed a methionine choline-deficient diet or HFD. This was associated with changes in the composition of the gut microbiota, especially an increase in the proportion of bacteria belonging to the Porphyromonadaceae family. Strikingly, cohousing of wild-type mice with these inflammasome-deficient mice resulted in increased glucose intolerance and obesity, as well as hepatic steatosis and liver inflammation, in the wild-type mice, indicating that dysbiosis itself can induce NAFLD progression. Another study conducted in mouse models demonstrated that the susceptibility to develop NAFLD can be transmitted by fecal transplant to recipient germ-free mice (45). They first observed that conventionally raised mice displayed various levels of weight gain, glucose intolerance, and steatosis when given an HFD for 16 weeks. They selected two donor mice for microbiota transplant based on their opposite responses to the HFD. Although the mice were the same weight, one displayed low fasting glycemia and slight steatosis (“nonresponder”), and the other displayed insulin resistance and marked steatosis (“responder”). Two groups of germ-free mice were transplanted with the gut microbiota of the two selected mice by oral gavage of donor cecal content. Only the mice transplanted with the responder microbiota developed fasting hyperglycemia, hyperinsulinemia, and liver steatosis after being fed an HFD and showed increased expression of genes involved in lipogenesis. This suggests that differences in susceptibility to develop NAFLD may be controlled primarily by the gut microbiota. Lachnospiraceae and Barnesiella were found to be overrepresented in the gut microbiotas of mice with glucose intolerance and steatosis, whereas the group of mice that did not develop features of NAFLD had a higher population of Bacteroides vulgatus (45).

Mechanisms Linking the Gut Microbiota to NAFLD

Several mechanisms have been proposed to link the gut microbiota to NAFLD, including small intestinal bowel overgrowth (SIBO), gut leakiness and resulting endotoxemia and inflammation, endogenous ethanol production, or BA and choline metabolism.

SIBO is a condition in which colonic bacteria colonize the small bowel due to altered intestinal motility. The diagnosis of bacterial overgrowth is performed using various techniques including breath testing and the culture of an aspirate from the jejunum, whereas culture-independent techniques, such as qPCR and sequencing, have been rarely applied. Most controlled trials using breath testing revealed a higher prevalence of SIBO in NAFLD patients than in healthy subjects (reviewed in reference 46). Moreover, SIBO correlated with the severity of steatosis but not NASH (47). However, no difference was found in bacterial numbers between patients suffering from NAFLD or NASH and healthy controls in a study using qPCR (37), suggesting that the use of molecular techniques may revise the current view on the role of SIBO in NAFLD.

Numerous studies have shown that the IM tunes the homeostasis of the gut barrier. A disruption of gut barrier integrity, characterized by the disruption of tight junctions and increased permeability, has been found in biopsy-proven NAFLD patients (47). Similarly, increased gut permeability has been observed in children with NAFLD and was found to correlate with the severity of the disease (48). This loss of barrier function leads to increased amounts of microbial products reaching the liver via portal circulation (49). Among these microbial products, lipopolysaccharide (LPS), a component of the bacterial cell wall of Gram-negative bacteria, binds to Toll-like receptor 4 (TLR4), triggering not only inflammation but also production of the extracellular matrix by hepatic stellate cells, leading to fibrosis (Fig. 2) (50, 51). In addition to TLR4, TLR2 and TLR9 may also be involved in NAFLD development, as TLR2- and TLR9-deficient mice display less steatosis and inflammation than their wild-type counterparts. In addition to increased gut permeability, HFD results in increased intestinal LPS absorption through high chylomicron production by intestinal epithelial cells, leading to low-grade endotoxemia (52). Animal studies suggest that endotoxemia alone may induce NASH, and higher endotoxin levels have been observed in NASH patients than in subjects with steatosis (53, 54). However, endotoxemia is not always higher in NASH (55), suggesting that LPS is not the only component that triggers the cascade of increased proinflammatory cytokines, insulin resistance, and fibrosis.

FIGURE 2.

Intestinal microbiota in liver disease. Diet and alcohol influence the composition of the gut bacteria. Dysbiosis is associated with changes in bacterial metabolites such as SCFA and BAs. The gut barrier is also altered, leading to increased endotoxemia (LPS). Acetaldehyde is specifically produced by the gut bacteria in ALD. Modifications of BAs and activation of their receptors, FXR and TGR5, participate in the development of liver lesions.

Endogenous ethanol production has also been proposed as a mechanism linking the gut microbiota to NAFLD (56). Indeed, gut bacteria, including the Enterobacteriaceae family, carry out mixed-acid fermentation, a major product of which is ethanol. This microbially produced ethanol reaches the liver via the portal vein and is converted to acetate and acetaldehyde. Acetate is a fatty acid substrate, whereas acetaldehyde triggers the production of reactive oxygen species. Moreover, ethanol alters gut permeability, leading to increased endotoxemia. Thus, endogenous ethanol production may be another component that triggers the proinflammatory cascade. Higher ethanol levels have been detected in patients with histology-proven NAFLD than in healthy controls (57) and in non-alcohol-consuming adolescents with NASH than in obese or healthy controls (34). This was associated with an increased abundance of Escherichia bacteria, ethanol-producing members of the Enterobacteriaceae family.

Choline, a phospholipid component of the cell membrane, is an essential nutrient required for lipid metabolism and neurotransmitter synthesis. Diets deficient in choline lead to triglyceride accumulation in the liver and reduced hepatic secretion of very-low-density lipoproteins, resulting in liver steatosis, which is reversed after choline supplementation (58). The gut microbiota can reduce choline bioavailability, thus mimicking a choline-deficient diet. Different gut bacteria can convert choline to methylamines, which may further induce inflammation when absorbed by the liver. Indeed, the high capacity of the gut microbiota to metabolize choline into methylamines explains the high susceptibility to develop NAFLD in response to an HFD in mouse strain 129S6 (59). However, it is not currently known whether bacterial choline metabolism contributes to NAFLD susceptibility in humans.

Primary BAs (in humans, cholic acid [CA] and chenodeoxycholic acid [CDCA]) are synthesized from cholesterol in the liver and conjugated to either taurine or glycine (60). They are then excreted into the small intestine, where their deconjugation and dehydroxylation are dependent on the gut bacterial species and are therefore modified when dysbiosis occurs (61). More than 95% of the BAs secreted in bile are reabsorbed in the distal ileum and return to the liver, taking part in the enterohepatic cycle. The main function of BAs is to assist the absorption of dietary lipids and lipid-soluble nutrients. However, they are also signaling molecules through the activation of receptors, such as the FXR or Takeda G protein-coupled receptor 5 (TGR5). Thus, they may modulate the inflammatory response and lipid, glucose, energy, and drug metabolisms, as well as their own biosynthesis (62). FXR-deficient mice develop features of NAFLD, including increased steatosis and inflammation. Beneficial effects of BAs on glucose and lipid metabolism, as well as on NAFLD, have been recently reported. OCA, a synthetic FXR ligand, improves liver injuries in biopsy-proven NASH patients (25). However, less than 50% of the patients receiving OCA showed histological improvement, whereas transaminases and glucose intolerance returned to baseline levels when the treatment ceased. Dyslipidemia was a side effect of OCA administration, preventing its use as an adjuvant treatment for NAFLD. The fraction of BAs that escape enterohepatic circulation, approximately 5% (i.e., 200 to 800 mg daily in humans), pass into the colon, where they undergo bacterial metabolism leading to the production of over 20 different secondary BAs in adult human feces (63). The capacity to activate the FXR and TGR5 depends on the type of BA. The primary BA, CDCA, is the most potent, whereas secondary BAs activate the FXR and TGR5 to a much lesser extent. The metabolism of the gut microbiota may thus modulate FXR and TGR5 activation, and subsequently NAFLD susceptibility, by altering the composition of the BA pool.

Specific BAs produced by the gut microbiota can also directly affect liver health. For example, the dehydroxylation of CDCA leads to formation of lithocholic acid, which is toxic to liver cells (64), and high levels of deoxycholic acid (DCA) are associated with an increased risk of liver cancer (65). Some studies also showed that DCA can impair gut barrier function (66). Conversely, UDCA, produced by few intestinal bacteria through the epimerization of the 7α-hydroxyl group of CDCA (67), is thought to be chemopreventive and is used to treat cholesterol gallstones and primary biliary cirrhosis. UDCA is also able to reverse the effect of DCA on the gut barrier (66). Finally, the gut microbiota may also modulate host lipid metabolism and NAFLD development through the metabolism of BAs by changing their emulsification and absorption properties, which may affect fatty acid storage in the liver.

Probiotics, Prebiotics, and Symbiotics as Therapeutic Approaches

Prebiotics and probiotics are known modulators of the gut microbiota and have demonstrated beneficial effects in both animal models of NAFLD and humans. Several probiotic strains have been found to be effective in different experimental rodent models of NAFLD. For example, supplementation with Lactobacillus rhamnosus GG reduced steatosis in fructose-induced NAFLD (68). This was associated with restored gut barrier function and decreased expression of proinflammatory cytokines in the liver. Similarly, Bacteroides uniformis CECT 7771 and Bifidobacterium pseudocatenulatum CECT 7765 improved steatosis and immune defense mechanisms in HFD-fed mice (69, 70), whereas Lactobacillus casei strain Shirota was shown to protect against NASH induced by a methionine- and choline-deficient diet (71). Also, Lactobacillus paracasei F19 reduced steatosis and inflammation in a rat model of ischemia-reperfusion (72). Finally, administration of VSL#3, a preparation composed of eight bacterial strains, improved liver steatosis and insulin resistance in HFD-fed mice (73) and reduced liver inflammation and serum ALT in leptin-deficient (ob/ob) mice (74). Few randomized, prospective clinical trials have been performed in humans to assess the effect of probiotic administration on NAFLD. However, a meta-analysis of four trials involving 134 NAFLD/NASH patients provided evidence that probiotic therapies may be effective in improving NAFLD markers, despite the use of different bacterial strains and administration protocols (75). In particular, the use of probiotics was associated with lower plasma aminotransferase and total cholesterol levels, lower systemic inflammation, and improved insulin resistance.

Prebiotics, which can promote the growth of beneficial bacteria, may also improve markers of NAFLD. For example, fructo-oligosaccharide (FOS) administration reduced endotoxemia and steatosis in HFD-fed mice, probably via the restoration of Bifidobacteria and Akkermansia muciniphila populations (76, 77). Similarly, an HFD supplemented with fungal chitin-glucan was found to decrease steatosis as well as fat mass development and hyperglycemia in mice. These beneficial effects correlated with the abundance of Roseburia spp., which was decreased by the HFD and restored by the prebiotics. These changes in gut microbiota composition, due to prebiotic treatment, are thought to improve gut barrier integrity through mechanisms including the production of endogenous glucagon-like peptide 2 and endocannabinoids.

Finally, symbiotics, which consist of a mix of probiotics and prebiotics, have been recently used in patients with NAFLD. A 24-week treatment with Bifidobacterium longum and FOS (78), together with lifestyle modifications, reduced endotoxemia and serum transaminase and tumor necrosis factor alpha levels, as well as steatosis and liver inflammation (79). However, these parameters also improved in the placebo group. Similarly, a 28-week treatment with a mixture of seven bacterial strains associated with FOS had beneficial effects on hepatic inflammation and overall liver function (80).

These encouraging results strongly suggest that the use of pro-, pre-, and symbiotics may be an effective addition in the treatment of NAFLD.

ALCOHOL LIVER DISEASE

Microbiota, a Key Player in ALD

In 1995, Adachi et al. reported that antibiotics protected rats against alcohol-induced liver injury by lowering Kupffer cell activation, without linking antibiotic use to gut bacteria (81). ALD is also associated with elevated plasma endotoxin levels in alcoholic patients and rodent models of alcohol consumption (82–84), which is further explained by increased intestinal permeability (85). The disruption of the gut barrier is now largely understood (86–88) and is explained by the lower expression of several proteins of the tight junctions (89). Moreover, ethanol consumption modulates the glycosylation of mucin, which modifies the protective mucus layer and, potentially, the adherent bacterial species (90). Using germ-free mice, it was shown that the alcohol-associated increase of intestinal permeability is not induced by alcohol, per se, but at least in part by acetaldehyde produced by bacterial ethanol metabolism (91). These data do not exclude that other bacterial metabolites may be involved in the disruption of tight junctions (92). Thus, the IM participates in the disruption of the gut barrier, which may impair liver homeostasis by increasing the level of bacterial products or bacterial translocation into the blood and lymph nodes. This translocation may be particularly deleterious in the case of dysbiosis. Indeed, dysbiosis has already been described in alcohol-fed rodents (93–95). Alcohol intake in rats was associated with a decrease in the abundance of lactic acid bacteria, especially those of the genera Lactobacillus, Pediococcus, Leuconostoc, and Lactococcus (95). An increase in the abundance of Proteobacteria and Actinobacteria and decreases in Bacteroidetes and Firmicutes were described in alcohol-fed mice. Moreover, these modifications were accompanied by a decrease in the bacterial diversity (93). In alcoholic patients, dysbiosis of the colon IM was associated with a decrease in the abundance of Bacteroidetes and an increase in Enterobacteriaceae and Proteobacteria (96). These studies show that dysbiosis is associated with alcohol-induced liver lesions, but the causal role of the IM has only recently been demonstrated.

Patients with severe alcoholic hepatitis (sAH) harbor an IM different from that of alcoholic patients without liver lesions (noAH) (97). The propensity of alcohol-induced inflammation was shown to be transmissible from patients to mice via the transplantation of the IM. Germ-free mice that received the IM of an alcoholic patient with sAH developed more-severe liver lesions than mice that received the IM of an alcoholic patient without AH. They also developed more-pronounced disruption of the intestinal barrier, associated with visceral inflammation. In conventional mice, efficient IM transplantation of a human stool sample from an alcoholic patient with noAH reverted the development of liver lesions in alcohol-fed mice that initially received the IM of an alcoholic patient with sAH (97). Altogether, these findings show that individual susceptibility to ALD is dependent on the IM and provide strong evidence for a causal role of the IM in ALD.

The connections between the microbiota and their host in the development of ALD have been established, but the mechanisms that are involved in the deleterious role mediated by the IM remain unclear (98). However, metabolites produced by bacteria, such as short-chain fatty acids (SCFA) and, more importantly, volatile organic compounds (VOC) or BAs seem to be key players in ALD.

Alcohol consumption is associated with a specific mixture of VOC in stools (99). Among these VOC, SCFA propionate and isobutyrate, involved in intestinal epithelial cell homeostasis and gut barrier integrity, are less abundant in the stools of alcoholics than in those of nonalcoholic healthy individuals. BA profiles are also altered by alcohol intake in rats (100). In humans, alcohol intake induces an increase in total stool BA levels, specifically lithocholic acid and DCA, and more generally an increase in the secondary/primary BA ratio (101). However, these studies compared alcoholic and nonalcoholic individuals, and alcohol as a specific carbohydrate substrate could itself induce marked changes.

Analysis of the fecal metabolome from alcohol-fed mice has highlighted the involvement of BAs in ALD (97). Mice transplanted with the IM of a patient with severe alcoholic hepatitis developed more-severe liver lesions than those transplanted with the IM of an alcoholic patient without liver disease. Comparison of the fecal metabolites allowed the identification of 13 biomarkers related to alcohol-induced liver toxicity. The most discriminating molecules were BA derivatives and hydroxygenated/oxygenated fatty acids. Higher amounts of the primary BA, CDCA, were found in fecal samples of mice without liver lesions than in those of mice with alcohol-induced liver lesions. The secondary BA, UDCA, was more abundant in the fecal metabolome of mice that received the MI of the alcoholic patient without liver disease than of those that received the MI of the patient with ALD.

The synthesis of BAs is dependent on a negative-feedback loop following FXR activation (102, 103). This cycle also involves several transporters that take part in the secretion of BAs from the liver to the gallbladder and BA reabsorption from the ileum. Specific BAs play a role in gut barrier disruption. UDCA and tauroursodeoxycholic acid attenuate alcohol-induced liver lesions by preventing liver damage (104). UDCA has hepatoprotective properties and was a discriminant metabolite between patients with and without alcohol-induced liver lesions (97). DCA impairs the gut barrier function in the jejunum and colon, and UDCA has a protective effect on the colon against DCA-induced barrier dysfunction (66). Moreover, mice deficient for the BA receptor FXR showed an increase of liver lesions, suggesting that the disruption of BA homeostasis was involved in alcohol-induced liver lesions (105). Thus, the dysregulation of the enterohepatic cycle by IM dysbiosis may be involved in alcohol-induced liver lesions, but further studies are needed to better understand this relationship (Fig. 2).

Disruption of the gut barrier is a prerequisite for ALD, resulting in bacterial translocation (106). The mucus layer and antimicrobial peptides are essential for protecting the intestinal epithelium against bacteria, and their alteration is associated with ALD. The production of mucin by goblet cells specifically decreases in mice with alcohol-induced liver lesions depending on the IM (97). Surprisingly, mice deficient for mucin production were protected against alcohol-induced liver lesions. This deficiency was associated with a large increase in defensin production, as well as lectins Reg3β and Reg3γ, increasing the killing of commensal bacteria and preventing bacterial overgrowth (107). Moreover, defensin, Reg3β, and Reg3γ levels decreased in mice fed alcohol in a mouse model of ALD (95), and mice deficient for Reg3β and Reg3γ showed increased adherent bacteria and a high level of bacterial translocation, promoting alcohol-induced liver lesions (108). Conversely, mice that overexpressed Reg3γ were protected from alcohol-induced hepatic toxicity, associated with decreased adherent bacteria and a low-level bacterial translocation. Moreover, mice harboring a specific IM, that were protected from the hepatic toxicity of alcohol, also expressed high levels of defensins (109). These data suggest that the control of bacterial overgrowth and adherent bacterial content is essential to the protective mechanisms against alcohol-induced liver toxicity.

Recent data have emphasized the role played by the IM in the sensing of nutrients and hormones via the gastrointestinal nervous system (110–112). The increase in intestinal permeability and dysbiosis in alcoholic patients was associated with higher scores of depression, anxiety, and alcohol craving (113). Alcoholic patients with gut permeability harbored a lower abundance of genera belonging to the Ruminococcaceae family (Ruminococcus, Faecalibacterium, Subdoligranulum, Oscillibacter, and Anaerofilum), as well as clostridia, than patients with less gut permeability. Conversely, the abundance of Dorea (Lachnospiraceae), Blautia, and Megasphaera was increased in these patients. VOC produced by these members of the IM were different in alcoholic patients harboring dysbiosis and gut leakiness, especially several indole and phenol species. This suggests that metabolites produced by the IM are involved not only in cell dysfunction and liver disease but also in the psychological symptoms of alcohol dependence (113).

Prebiotics, Probiotics, or Fecal Microbiota Transfer To Treat ALD

Several probiotics have been tested in rodent models and several human trials to improve ALD progression. The first use of probiotics in ALD was performed in a rodent model of alcohol intake by using Lactobacillus sp. strain GG (Lactobacillus GG), which improved gut leakiness and liver inflammation (114–116). The addition of oat fiber or a supernatant of Lactobacillus GG induced similar results, suggesting that bacterial products were partially involved in the protective mechanisms (94, 117, 118). VSL#3, a mixture of eight probiotic strains, also improved liver lesions in ALD in humans and rodents (119, 120). Several other probiotic strains have been tested and improved liver lesions in rodents. They are well summarized in a recent review (121). Prebiotic use was also tested and showed that FOS improved rodent alcohol-induced liver damage in a mouse model of alcoholic liver disease (95).

Altogether, these studies highlight the promise of controlling the IM by pre- or probiotic treatments, but further studies are needed to decipher the mechanisms involved in bacterium-related gut and liver damage.

ADVANCED LIVER DISEASE

Fibrosis

The composition and functions of the gut microbiota are altered in patients with fibrosis (122). Patients with a fibrosis score of 2 or more harbor more Bacteroides and Ruminococcus and fewer Prevotella bacteria than patients with a fibrosis score of 0 or 1. Moreover, genes related to carbohydrate, lipid, and amino acid metabolism are overrepresented in the microbiome of patients with stage 2 fibrosis or greater, indicating a shift in the metabolic functions of the gut microbiota (122).

The bile duct ligation mouse model in combination with an HFD was used to evaluate the consequences of fibrosis on the IM. These mice developed higher hepatic stellate cell (major producers of the fibrotic matrix) activity, resulting in increased liver fibrosis. This was associated with dysbiosis, characterized by a marked increase in the abundance of Enterobacteriaceae (Gram negative) and the complete disappearance of Bifidobacterium (Gram positive) (123). The authors further selected the Gram-negative and Gram-positive fractions of the gut microbiota from the fibrogenic HFD/bile duct ligation mice and transplanted them into control mice. Mice receiving the Gram-negative fraction displayed increased liver injury, showing that dysbiosis may contribute not only to steatosis and inflammation but also to fibrogenesis in the liver (123).

Fibrosis has also been induced by injections of carbon tetrachloride (CCl₄), showing that the IM was modified during the process of fibrosis development. There was a specific decrease in the abundance of Clostridium spp. and, more generally, an increase in the ratio of aerobic to anaerobic bacteria, which correlated with a higher fibrosis score (124). Bacterial DNA found in the mesenteric lymph nodes, reflecting disruption of the gut barrier, was modified in rats with chemically induced fibrosis (125). Moreover, the absence of an IM worsened fibrosis development in germ-free mice, but the immaturity of their immune system may have also played a role in the evolution of the liver damage (126). Probiotic treatments, such as use of Saccharomyces cerevisiae subsp. boulardii, were associated with a slower progression of liver fibrosis (127), and VSL#3 prevented bacterial translocation (128). However, it is not possible to conclude whether these treatments played a role early in the inflammation stage to limit the progression to fibrosis.

Cirrhosis

Patients with cirrhosis have distinct fecal microbial communities relative to healthy individuals. Members of the phylum Bacteroidetes were less abundant in cirrhotic patients (24 hepatitis B virus [HBV] and 12 alcohol-related cirrhosis patients), whereas Proteobacteria and Fusobacteria were more abundant (129) (Table 3). Among cirrhotic patients, the abundance of Prevotellaceae was greater in alcoholic patients than those infected with HBV. Moreover, the prevalence of potentially pathogenic bacteria, such as Enterobacteriaceae and Streptococcaceae, and the reduction of beneficial populations, such as Lachnospiraceae, in patients with cirrhosis may affect prognosis (129). In a more recent study, the functional diversity of the IM was shown to be significantly lower in patients with hepatitis B-related cirrhosis than in controls (130). Moreover, 54% of the cirrhotic patient enriched taxonomically assigned species are of buccal origin, suggesting an invasion of the gut from the mouth in liver cirrhosis (130).

TABLE 3.

Comparison of healthy microbiotas and microbiotas associated with cirrhosis^a

Phylum	Family	Genus	Reference
Bacteroidetes ⇩			Healthy (n = 24)Cirrhosis (n = 36: HBV = 24, Alc = 12)Chen et al., 2011 (129)
[Firmicutes]	Veillonellaceae ⇧
	Streptococcaceae ⇧
	Lachnospiraceae ⇩
Proteobacteria ⇧	Enterobacteriaceae ⇧
Fusobacteria ⇧
[Bacteroidetes]	Rikenellaceae ⇩		Healthy (n = 14)Cirrhosis (n = 47)Kakiyama et al., 2013 (101)
[Firmicutes]	Veillonellaceae ⇧	Blautia ⇩
	Ruminococcaceae ⇩
	Lachnospiraceae ⇩
[Proteobacteria]	Enterobacteriaceae ⇧
[Bacteroidetes]	Porphyromonadaceae ⇩		Healthy (n = 25)Cirrhosis (n = 175: HCV = 75, Alc = 31, others = 71)Bajaj et al., 2014 (132, 137)
[Firmicutes]	Veillonellaceae ⇩
	Ruminococcaceae ⇩
	Lachnospiraceae ⇩
	Clostridiales XIV ⇩
	Enterococcaceae ⇧
	Staphylococcaceae ⇧
[Proteobacteria]	Enterobacteriaceae ⇧
Bacteroidetes ⇩	[Bacteroidaceae]	Bacteroides ⇩	Healthy (n = 83)Cirrhosis (n = 98: HBV = 79, Alc = 10, others = 9)Qin et al., 2014 (130)
	[Prevotellaceae]	Prevotella ⇧
	[Rikenellaceae]	Alistipes ⇩
[Firmicutes]	[Veillonellaceae]	Veillonella ⇧
	[Streptococcaceae]	Streptococcus ⇧
	[Clostridiaceae]	Clostridium ⇧
	[Lachnospiraceae]	Eubacterium ⇩
Proteobacteria ⇧
Fusobacteria ⇧

^a Microbiota were analyzed using 16S deep sequencing, except in the study of Qin et al., which used metagenomics. Additional taxonomic information for discriminating taxa is in brackets. Taxa in boldface are those identified in more than one study.

The gut microbiome has been assessed in cirrhotic patients and compared to that of healthy individuals by metagenomic studies. The modified metabolic pathways showed an enrichment of genes involved in gluconeogenesis and lipid metabolism in the IM of cirrhotic patients and a decrease of genes involved in BA metabolism (131). Another study revealed specific clusters representing cognate bacterial species; 28 were enriched in cirrhotic patients and 38 in control individuals. Among bacterial genes, those involved in denitrification, assimilation or dissimilation of nitrate, gamma-aminobutyric acid biosynthesis, heme biosynthesis, and phosphotransferase systems were found to be associated with liver cirrhosis (130). Genetic and functional biomarkers specific for liver cirrhosis were revealed by a comparison with those for type 2 diabetes and inflammatory bowel disease. Fifteen biomarkers were specifically associated with cirrhosis. However, generalization of these results is difficult, as the number of patients was low and the cause of cirrhosis was restricted mainly to HBV infection (130).

Most cirrhotic patients are asymptomatic until they develop decompensated cirrhosis. The survival rate of cirrhotic patients is dependent on the associated complications, including episodes of hepatic encephalopathy (HE), ascites, or infections, which have been associated with specific bacterial communities (132).

In cirrhotic patients, HE was associated with a specific dysbiosis: a high level of Enterobacteriaceae, already associated with infection in decompensated cirrhosis, was also associated with HE (133). Alcaligenaceae, bacteria that are specifically involved in urea metabolism, were also more abundant in HE patients. In a prospective clinical trial, VSL#3 was found to be efficient in preventing hepatic encephalopathy in cirrhotic patients, suggesting that the dysbiosis in HE may have a causative role (134). Conversely, this treatment did not affect the portal pressure (135).

A cirrhosis dysbiosis ratio (CDR) has been developed to evaluate the severity of cirrhosis (132). The CDR is a ratio between the abundance of Clostridiales (Clostridium cluster XIV), Ruminococcaceae, and Lachnospiraceae and that of Bacteroidaceae and Enterobacteriaceae. Healthy patients have a CDR of approximately 2, decreasing to 0.9 in patients with compensated cirrhosis and to 0.3 in cirrhotic patients with an infection. Enterobacteriaceae is the most prevalent family modified in the CDR and is specifically associated with the complications of cirrhosis due to the production of potent endotoxins. The level of fecal Bacteroidaceae and Clostridium cluster XIV may predict the risk of hospitalization in patients with cirrhosis, independently of other classical clinical predictors, such as the MELD (model for end-stage liver disease) score, HE, or the intake of proton pump inhibitors (136). These data prove the relationship between dysbiosis and the severity of cirrhosis. Moreover, a randomized clinical trial using Lactobacillus GG in cirrhotic patients demonstrated that altering the composition of the IM through the use of probiotics improved dysbiosis and endotoxemia (137).

Viable bacteria have been found in the ascites of patients with decompensated cirrhosis, including several species of Propionibacterium, Pseudomonas, and Staphylococcus recognized as commonly colonizing the skin (138). Thus, access to the peritoneal cavity is not limited to the translocation of bacteria from the gut. The quantity of bacteria in the ascites is not sufficient to induce spontaneous bacterial peritonitis, but there is a relationship between the bacterial species and the severity of the disease (138).

Bacterial metabolites are related to the abundance of the bacterial species and may have beneficial or deleterious effects. Secondary BAs are exclusively produced by the IM. Cirrhosis has been recently shown to be associated with the abundance of specific bacterial families and to be linked with a decrease in the amount of secondary BAs in the feces (101, 132). Moreover, BA levels can discriminate between patients with severe alcoholic hepatitis associated with cirrhosis and alcoholic patients without liver lesions (97). Further studies are needed to decipher the spectra of metabolites that mediate the deleterious effects of bacterial dysbiosis.

Hepatocellular Carcinoma

HCC occurs in the presence of chronic liver inflammation, typically in patients with cirrhosis (139). To understand the contribution of NAFLD or ALD to the development of HCC, other causes of liver disease need to be excluded, especially HBV or HCV infection. TLR4 deficiency resulted in decreased promotion of HCC but did not affect HCC initiation in chemically induced HCC in adult mice, either by diethylnitrosamine (DEN) alone or by DEN and chronic CCL4 injections (140, 141). Sterilization of the gut by antibiotherapy decreased both HCC initiation and progression. Accordingly, germ-free mice showed an approximately 80% reduction of HCC (140). Altogether, these data show that the inhibition of the LPS/TLR4 pathway prevents HCC progression. However, opposite results were obtained in another model of HCC, using TLR4-deficient mice in which HCC was induced by the unique injection of DEN, 15 days after birth (142). These discrepancies can be explained by the high frequency of DNA mutations in this DEN model relative to DEN injections in adult rodents associated with the development of a chronic liver inflammation. In humans, only indirect evidence has been published and showed that TLR2, -4, and -9 overexpression was associated with a poor prognosis for HCC progression (143).

The role of the IM was recently assessed in obesity-associated HCC. HCC was chemically induced in mice using DMBA [7,12-dimethylbenz(a)anthracene] at the neonatal stage. The association of DMBA and HFD induced tumor development in the mice (65). The analysis of serum metabolites showed an increase in DCA levels in mice that developed tumors. The production of the secondary BA DCA by the IM was associated with the presence of Clostridium species that harbored a 7α-dehydroxylase function to metabolize primary BAs (144). DCA levels were lowered using UDCA or by chemical inhibition of 7α-dehydroxylase. Moreover, vancomycin, which targets Gram-positive bacteria, especially Clostridium, also decreased serum DCA levels. Under these conditions, lowering serum DCA levels reduced HCC development.

These findings provide new evidence that the IM could be involved in the initiation and progression of HCC.

CONCLUSION

The complex community of the intestinal bacteria influences both the normal and pathological states of the liver, and we are only beginning to understand the mechanisms involved in this relationship. In dysbiosis of the gut microbiota associated with metabolic or alcoholic liver diseases, the changes in bacterial gene expression or metabolite production may be more important than the changes in bacterial composition. Understanding the clinical significance of gut microbiota dysbiosis in the context of liver disease remains a challenge and must consider the possible association between host genetic or/and epigenetic changes that favor liver injury. Further studies are needed, combining metagenomics, metatranscriptomics, and metabolomics with longitudinal studies, along with standardization of these techniques, to better comprehend the role of microbiota changes in the development of liver disease.

Many studies already strongly suggest that targeting the IM through the use of pro-, pre-, and symbiotics, and possibly fecal transfer, may be an effective treatment for NAFLD and ALD. However, larger studies incorporating liver biopsies, standardized dose administration and duration, and an analysis of the impact on the gut microbiota are clearly needed. This should help to open new treatment strategies based on the modulation of the gut microbiota.