Gut microbiota in liver disease: too much is harmful, nothing at all is not helpful either

Abstract

The intestinal microbiome plays a major role in the pathogenesis of liver disease, with a hallmark event being dysbiosis, or an imbalance of pathobionts and beneficial bacteria with the associated deleterious effects on their host. Reducing the number of intestinal bacteria with antibiotic treatment is generally advantageous in experimental liver diseases. Complete absence of intestinal microbiota as in germ-free rodents can be protective in autoimmune hepatitis and hepatic tumors induced by chemicals, or it can exacerbate disease as in acute toxic liver injury and liver fibrosis/cirrhosis. In alcoholic liver disease, nonalcoholic fatty liver disease, and autoimmune cholangiopathies, germ-free status can be associated with worsened or improved hepatic phenotype depending on the experimental model and type of rodent. Some of the unexpected outcomes can be explained by the limitations of rodents raised in a germ-free environment including a deficient immune system and an altered metabolism of lipids, cholesterol, xenobiotics/toxins, and bile acids. Given these limitations and to advance understanding of the interactions between host and intestinal microbiota, simplified model systems such as humanized gnotobiotic mice, or gnotobiotic mice monoassociated with a single bacterial strain or colonized with a defined set of microbes, are unique and useful models for investigation of liver disease in a complex ecosystem.

INTRODUCTION

There is a strong symbiotic relationship between gut microbiota and its host. The human body harbors ~1–10 intestinal bacteria for each human cell (60, 93). Although each person’s microbial profile is distinct, the relative abundance and composition of bacterial species are similar among healthy individuals (119). Microbiota has protective effects on the human body, as it helps with digestion and synthesis of vitamins and furthers resistance to intestinal colonization of pathogens (27, 85). On the other hand, microbiota can become pathogenic. Alterations of the intestinal microbiome have been reported to affect the pathogenesis of many diseases, including liver diseases (14). Intestinal dysbiosis [or an imbalance of pathobionts and beneficial bacteria with its associated deleterious effects on their host (40)] plays a critical role in the development of alcoholic liver disease (88), nonalcoholic fatty liver disease (NAFLD; 1, 87), nonalcoholic steatohepatitis (NASH; 1, 87), acute (toxic) liver injury, fibrosis/cirrhosis (88), hepatic tumors (13), autoimmune hepatitis (AIH; 58), and autoimmune cholangiopathies (78, 86). Highlighting the significance of intestinal dysbiosis in the pathogenesis of liver diseases, a study published in 1990 reported development of liver inflammation and fibrosis following surgical creation of a small intestinal loop causing bacterial overgrowth (57). Gut microbiota affects liver disease via changing the metabolism and caloric yield of diet; it can also perturb gut permeability with resulting bacterial translocation and exposure of the liver to microbial components with subsequent inflammation (14, 88).

Although gaining popularity again today because of the mounting evidence of the significant effect of microbiota on its host’s health, germ-free rodents were employed to study liver disease already in the 1960s (8, 30, 39, 53, 54, 75). Germ-free rodents do not harbor any microbiota from birth on as they are born by caesarean section in sterile conditions and immediately separated from their biological mothers to be raised in isolators. Because of the lack of exposure to bacteria, germ-free animals display an aberrant maturation of the immune system with an abnormal intestinal epithelial cell morphology characterized by longer villi and shorter crypts, diminished antimicrobial peptides, and underdeveloped lymphoid tissue including the spleen, thymus, lymphoid follicles, Peyer’s patches, and mesenteric lymph nodes (7, 35, 37). A disrupted gut barrier contributes to the pathogenesis of liver disease in conventional rodents; however, intestinal dysfunction does not appear to be crucial for disease in germ-free mice, as, for example, germ-free mice can show intestinal permeability similar to or even higher than that of conventional rodents on control or high-caloric diet but are protected against diet-induced obesity and NAFLD (11, 45). This can be partially explained by the absence of intestinal microbiota in germ-free mice and hence lack of translocation of bacteria-derived lipopolysaccharide (LPS) into the bloodstream with associated hepatotoxic and hepato-modulating effects (see below). By understanding these hallmark events and examining disease development in germ-free rodents, scientists can have a better idea how microbiota affects human health. For this review, we screened the PubMed library for the terms “liver germ-free,” “liver disease germ-free,” “hepatitis germ-free,” “steatohepatitis germ-free,” “hepatoma/hepatocellular carcinoma germ-free,” and “liver disease antibiotics” to obtain a large set of articles related to liver disease in germ-free animals and selected articles on the basis of their title and/or abstract. In this review, we discuss the effects of germ-free status on the development of liver diseases and possible mechanisms (Table 1). We review implications of antibiotics in experimental liver disease (Table 2) and discuss the use of “humanized” rodents (Fig. 1).

Table 1.

Phenotype of germ-free rodents in various liver diseases compared with controls

Type of Liver Disease	Experimental Model	Animals	Phenotype of Germ-Free Animals Relative to Conventional Controls	Reference
Alcoholic liver disease	Single gavage of alcohol	C57BL/6 mice	Exacerbated	Chen et al. (21)
Alcoholic liver disease	Lieber-DeCarli alcohol diet over 7 days plus binge	Swiss mice	Protected	Canesso et al. (18)
NAFLD	High-carbohydrate diet (57% carbohydrates and 5% fat)	C57BL/6J mice	Protected	Bäckhed et al. (4)
NAFLD	High-fructose diet	C57BL/6J mice	Protected	Kaden-Volynets et al. (45)
NAFLD	High-fat diet	C57BL/6J mice	Protected	Rabot et al. (79)
NAFLD	Western diet	C3H/HeJ mice	Protected	Fleissner et al. (34)
NAFLD	Choline-deficient diet	Fischer rats	Protected/exacerbated depending on cystine and cholesterol content in diet	Levenson et al. (53)
NAFLD	Choline-deficient diet	Lobund rats	Exacerbated	Levenson and Tennant (54)
HCC	No intervention until euthanasia at 12 mo	C3H/He mice	Protected	Mizutani and Mitsuoka (68)
HCC	No intervention until euthanasia at 19-41 mo	Lobund-Wistar rats	Similar	Pollard et al. (73)
HCC	MAM-GlcUA	Sprague-Dawley rats	Protected	Laqueur et al. (50)
HCC	DMBA	C3H mice	Protected	Grant and Roe (39), Roe and Grant (83)
HCC	Phenobarbital	C3H mice	Protected	Mizutani and Mitsuoka (66)
HCC	DEN and CCl4	C57BL/6 mice	Protected	Dapito et al. (26)
AIH	ConA	BALB/c mice	Protected	Wei et al. (109)
AIH	ConA	Swiss Webster (TAC:SW) mice	Protected	Celaj et al. (20)
Autoimmune cholangiopathies
PSC	Abcb4−/− (Mdr2−/−) mice	FVB/N Abcb4−/− (Mdr2−/−) mice	Exacerbated	Tabibian et al. (99)
PBC	NOD.c3c4 mice	NOD.c3c4 mice	Protected	Schrumpf et al. (90)
Acute (toxic) liver injury	Acetaminophen-induced liver injury	C3H/HeH mice	Similar	Possamai et al. (76)
Acute (toxic) liver injury	DMA plus sodium nitrite	Carworth Farms Swiss Webster mice	Exacerbated	Pollard et al. (74)
Acute (toxic) liver injury	DMN or DMA plus sodium nitrite	Wistar rats	Exacerbated	Sumi and Miyakawa (95)
Acute (toxic) liver injury	Chlortetracycline	ICR mice	Exacerbated	Einheber et al. (30)
Acute (toxic) liver injury	Butter yellow (3′-methyl-4-dimethylaminoazobenzene)	Fischer rats	Exacerbated	Bauer et al. (8)
Acute (toxic) liver injury	Partial hepatectomy	BALB/c mice	Exacerbated	Cornell et al. (24)
Liver fibrosis/cirrhosis	TAA or CCl4	C57BL/6 mice	Exacerbated	Mazagova et al. (65)
Liver fibrosis/cirrhosis	CCl4	ICR mice	Exacerbated	Popper et al. (75)

Abcb4^−/− (Mdr2^−/−), ATP-binding cassette subfamily B member 4 (also known as multidrug resistance protein-2) knockout; AIH, autoimmune hepatitis; ConA, concanavalin A; CCl₄, carbon tetrachloride; DEN, diethylnitrosamine; DMA, dimethylamine; DMBA, 7,12-dimethylbenz(a)anthracene; DMN, dimethylnitrosamine; HCC, hepatocellular carcinoma; MAM-GlcUA, methylazoxymethanol-β-d-glucosiduronic acid; NAFLD, nonalcoholic fatty liver disease; PBC, primary biliary cholangitis; PSC, primary sclerosing cholangitis; TAA, thioacetamide.

Table 2.

Phenotype of antibiotic-treated rodents in various liver diseases compared with controls (representative articles selected)

Type of Liver Disease	Experimental Model	Animals	Antibiotics	Phenotype of Antibiotic-Treated Animals Relative to Controls	Reference
Alcoholic liver disease	Lieber-DeCarli alcohol diet for 8 wk	C57BL/6 mice	Polymyxin B and neomycin	Improved	Chen et al. (22)
Alcoholic liver disease	Intragastric alcohol for ≤3 wk	Wistar rats	Polymyxin B and neomycin	Improved	Adachi et al. (2)
NAFLD	High-fat diet	C57BL/6N mice	Bacitracin, neomycin, and streptomycin	Improved	Jiang et al. (44)
NAFLD	High-fructose diet	C57BL/6J mice	Polymyxin B and neomycin	Improved	Bergheim et al. (9)
NASH	High-fat diet	C57BL/6J wild-type mice cohoused with C57BL/6J ASC−/− mice	Ciprofloxacin and metronidazole	Improved	Henao-Mejia et al. (43)
NASH	High-fat diet (STHD-01)	C57BL/6J mice	Ceftazidime and metronidazole	Improved	Yamada et al. (114)
HCC	DEN and CCl4	C3H/HeOuJ mice	ANMV	Improved	Dapito et al. (26)
HCC	DMBA	C57BL/6 mice	ANMV or vancomycin only	Improved	Yoshimoto et al. (115)
AIH	ConA	BALB/c mice	ANMV	Improved	Celaj et al. (20)
AIH	ConA	C57BL/6NCrlCrlj mice	Rifampicin	Improved	Kataoka et al. (48)
Autoimmune cholangiopathies
PSC	Abcb4−/− (Mdr2−/−) mice	C57BL/6J Abcb4−/− (Mdr2−/−) mice	Polymyxin B and neomycin	Improved	Peng et al. (72)
PBC	NOD.c3c4 mice	NOD.c3c4 mice	Ampicillin and neomycin	Improved	Schrumpf et al. (90)
Acute (toxic) liver injury	Acetaminophen-induced liver injury	BALB/c mice	ANMV	Improved	Gong et al. (38)
Acute (toxic) liver injury	d-Gal	Crlj:CD(SD) rats	Rifampicin	Improved	Kataoka et al. (48)
Liver fibrosis/cirrhosis	Bile duct ligation	Tlr4-wild-type C3H/HeOuJ mice	ANMV	Improved	Seki et al. (91)
Liver fibrosis/cirrhosis	CCl4 plus ethanol	Wistar rats	Ciprofloxacin	Improved	Zhang et al. (117)

Abcb4^−/− (Mdr2^−/−), ATP-binding cassette subfamily B member 4 (also known as multidrug resistance protein-2) knockout; AIH, autoimmune hepatitis; ANMV, ampicillin, neomycin, metronidazole, and vancomycin; ASC^−/−, apoptosis-associated speck-like protein containing a COOH-terminal caspase recruitment domain knockout; ConA, concanavalin A; CCl₄, carbon tetrachloride; DEN, diethylnitrosamine; d-Gal, d-galactosamine; DMBA, 7,12-dimethylbenz(a)anthracene; HCC, hepatocellular carcinoma; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; PBC, primary biliary cholangitis; PSC, primary sclerosing cholangitis; STHD, steatohepatitis-inducing high-fat diet; Tlr4, Toll-like receptor 4.

Fig. 1.

Use of germ-free mice in preclinical liver disease models. Germ-free mice have an immature immune system and a significantly altered metabolism related to lipids, cholesterol, xenobiotics/toxins, and bile acids. They can be colonized with only one known bacterial strain (monoassociated), a specific set of known bacterial strains, or human feces. Monoassociation and colonization with several bacteria are unique models to investigate the effect of single or multiple bacterial strains on experimental disease. Transmissibility of disease can be investigated by microbiota transfer into gnotobiotic mice. Humanized gnotobiotic mice allow the testing of microbiome-centered therapies to treat disease. Gnotobiotic rodents oftentimes have a more normalized immune system and metabolism than germ-free mice. Gnotobiotic mice can be subjected to preclinical models of liver disease to examine the impact of colonized microbes on the disease process.

ALCOHOLIC LIVER DISEASE

Chronic alcoholic liver disease may progress from simple steatosis to steatohepatitis, liver fibrosis, liver cirrhosis, and hepatocellular carcinoma (HCC; 42). Alcoholic liver disease is a leading cause of liver-related morbidity and mortality worldwide.

Factors that link ethanol to the onset and progression of liver disease are poorly understood. The susceptibility of patients with alcohol use disorder to alcoholic liver disease is variable. With similar alcohol intake, some people can develop severe forms of alcoholic hepatitis, whereas others remain healthy (103). A specific intestinal microbiota composition confers a higher risk of alcohol-induced liver injury (33), underlying the key role of gut microbiota in alcoholic liver disease. Intestinal dysbiosis has been described in patients with alcohol use disorder (51) and cirrhosis (77), and deteriorating dysbiosis is associated with cirrhosis progression (6). Severe alcoholic hepatitis was associated with higher fecal proportions of bifidobacteria, streptococci, and enterobacteria (59). Thus, intestinal microbiota is central in the onset and progression of alcoholic liver disease.

Germ-free rodents are a tool to investigate the function of the intestinal microbiome. In a model of acute alcohol exposure that mimics binge drinking, germ-free C57BL/6 mice exhibited significantly more pronounced liver injury, inflammation, and steatosis than their conventional counterparts (21). One reason for the increased susceptibility to liver injury might be that germ-free mice have a more efficient xenobiotic metabolism compared with conventional mice (12). Hepatic and intestinal expression of genes involved in xenobiotic metabolism was overall higher in germ-free mice than in conventional mice, including alcohol dehydrogenase, aldehyde dehydrogenase, and cytochrome P450 superfamily members (12, 21). Although the detailed mechanism is still unknown, it is possible that the induction of these genes resulted from accumulation of constitutive androstane receptor ligands such as bilirubin, bile acids, and steroid hormones due to the absence of intestinal bacteria metabolizing these molecules in germ-free mice (12). Upregulated genes were associated with more rapid ethanol elimination from the blood and increased production of toxic alcohol metabolites in the liver, further exacerbating liver injury in germ-free mice (21). Besides, elevated baseline levels of inflammatory mediators such as macrophage markers CD68 and F4/80 were also likely to contribute to increased liver damage in germ-free mice (21). Moreover, the increased hepatic steatosis observed in germ-free C57BL/6 mice was associated with upregulated expression of genes involved in fatty acid and triglyceride synthesis, including sterol regulatory element-binding protein-1 (SREBP-1), stearoyl-coenzyme A desaturase 1 (SCD-1), fatty acid synthase (FASN), and acetyl-CoA carboxylase-1 (ACC-1; 21).

On the other hand, germ-free Swiss mice were found to exhibit reduced liver injury after chronic alcohol consumption relative to their conventional controls (18). This was accompanied by a (not significant) reduction in gut permeability to ~50% and a significantly lower hepatic neutrophil accumulation in germ-free mice compared with conventional mice after ethanol administration (18).

In summary, germ-free C57BL/6 mice exhibit higher susceptibility to acute alcohol-induced liver injury relative to conventional mice, which may be related to a more efficient xenobiotic metabolism, elevated baseline levels of inflammatory mediators, and/or upregulated expression of genes involved in fatty acid and triglyceride synthesis. In contrast, germ-free Swiss mice appear to be resistant to liver damage secondary to alcohol compared with their conventionally raised counterparts. The discrepancy between these two studies could also be due to different mouse strains. Multiple rodent studies in alcoholic liver disease have shown that strains can differ significantly with regard to liver injury, immune response (TNF-α), alcohol-metabolizing enzymes [alcohol dehydrogenase, aldehyde dehydrogenase, and cytochrome P450 2E1 (CYP2E1)], endoplasmic reticulum stress [78-kDa glucose-regulated protein (GRP78) and C/EBP-homologous protein (CHOP)], and fat metabolism [nuclear SREBP-1 (nSREBP-1); 29, 104]. Germ-free status further modulates these differences and could hence potentially confer some protection in one strain and exacerbate alcoholic liver disease in another. The different durations of the alcohol-feeding models could further explain exacerbation of liver injury in germ-free mice (after a 1-time binge) versus an alleviated hepatic phenotype (after alcohol feeding over 1 wk plus binge). Animal facility-related factors should also be considered such as slightly different diets and vivarium temperature (49). Finally, it is unclear whether the germ-free condition was maintained during the feeding of a Lieber-DeCarli diet in a chronic model of alcoholic liver disease, as it is difficult to sterilize/irradiate the liquid diet reliably and as it has to be changed frequently. Further studies are required to investigate potential mechanisms to explain the dissimilar phenotype between different mouse strains in that type of liver injury.

NONALCOHOLIC FATTY LIVER DISEASE

NAFLD is considered the hepatic manifestation of metabolic syndrome and is commonly associated with insulin resistance (15). It is defined as a spectrum of liver diseases ranging from simple steatosis to NASH, advanced fibrosis, cirrhosis, or even HCC (16).

The pathophysiology of NAFLD involves many factors, including ecological, genetic, and metabolic factors such as limited physical activity and an imbalanced diet (41, 112). Gut microbiota is a new environmental factor contributing to obesity and NAFLD (1, 28, 87, 88). Compositions of gut microbiota are disturbed in patients with NAFLD. Members of the phylum Firmicutes were detected at elevated levels in individuals with NAFLD relative to nonobese controls (80), whereas contributions of Bacteroidetes were decreased in adult patients with NASH compared with subjects with simple steatosis and healthy individuals (69). Conversely, Bacteroidetes, Proteobacteria, and Enterobacteriaceae/Escherichia coli were found to be increased in pediatric patients with NASH compared with healthy controls, whereas Firmicutes were detected at lower levels (120). Probiotics can protect against the onset of NAFLD and attenuate NAFLD (41, 81, 108). An intervention with Lactobacillus casei strain Shirota protected against experimental NASH induced by methionine-choline-deficient diet, with reduced serum LPS concentrations and suppression of inflammation and fibrosis in the liver (71). VSL#3 probiotics (a mixture of bifidobacteria, lactobacilli, and Streptococcus thermophilus) reduced high-fat diet (HFD)-induced steatosis, inflammation, and insulin resistance in a rodent model (61). Daily gavage of fecal microbiota from control C57BL/6 mice for 8 wk for the second half of a 16-wk HFD-feeding experiment protected against NASH with lower hepatic fat, inflammatory cytokines, and NAFLD Activity Score relative to mice that did not receive a fecal transplant (118). Fecal transplant was also associated with increased intestinal butyrate concentrations, increased expression of the tight junction protein zonula occludens-1 (ZO-1), and lower serum LPS concentrations (118).

Interestingly, germ-free C57BL/6J mice gained less weight (4, 5, 79) and had a lower hepatic fat concentration (4, 79) than conventional mice when given a high-carbohydrate diet (57% carbohydrates and 5% fat; 4), an HFD (60% fat; 79), or a Western diet (41% carbohydrates and 41% fat; 5), suggesting that those germ-free mice are protected against obesity and NAFLD. Similarly, germ-free C3H/HeJ mice fed a Western diet were protected from fatty liver compared with their conventional counterparts (34). Germ-free C57BL/6J mice were also found to be protected against hepatic steatosis after a high-fructose diet (45). Conversely, germ-free Fischer rats fed a choline-deficient, low-cystine, low-cholesterol diet exhibited less liver fat content than their controls but worsened hepatic steatosis relative to conventional controls after a choline-deficient, high-cystine, high-cholesterol diet (53). These findings suggest that the liver injury is ameliorated in germ-free rats in the low-cystine-low-cholesterol environment likely secondary to absent bacterial utilization of choline and methionine and is much worse in the high-cystine-high-cholesterol environment possibly because of a missing reduction of cystine/cholesterol content by bacterial metabolization and/or the lacking effects of hepatoprotective bacterial metabolites. On the other hand, more than 50 years ago, germ-free Lobund rats were shown to develop NAFLD-related liver cirrhosis due to choline-deficient diet more rapidly relative to their conventional counterparts (54).

The protective effect observed in the germ-free C57BL/6J mice might be the consequence of a lower consumption of calories and higher excretion of fecal lipids as observed in one study (79). However, the protective effect was also documented when the same number of calories was consumed and the same fecal energy content noted in germ-free and conventional mice (5). Furthermore, the beneficial outcome might also result from sufficient choline levels in germ-free mice, as they harbor no gut microbiota to regulate concentrations of choline (41, 53, 88). Additionally, the protection in germ-free rodents might be due to altered inflammasome activity with an abolished caspase-1 autocleavage and associated reduced IL-18 concentrations, indicating insufficient macrophage activity (56, 63). Consistent with this idea, fecal transplantation from conventional mice on high-carbohydrate diet (4) or HFD (52) to germ-free mice resulted in gain of fat mass (4, 52) and NAFLD (52) in colonized mice.

Moreover, germ-free status enhances insulin sensitivity with improved glucose tolerance, reduced fasting and nonfasting insulinemia, and increased phosphorylation of Akt in adipose tissue (5, 79). Germ-free mice are thought to be protected against diet-induced obesity because of increased AMP-activated protein kinase (AMPK) activity and increased fatty acid oxidation in their peripheral tissues (5). AMPK is a key regulator of energy balance, and loss/attenuation of its kinase activity leads to several metabolic disorders (70). Additionally, increased fatty acid oxidation in germ-free mice is secondary to increased intestinal levels of fasting-induced adipose factor (Fiaf) via higher expression of peroxisome proliferator-activated receptor-γ coactivator-1 (Pgc-1; 70, 107). Increased Fiaf in germ-free mice also suppresses lipoprotein lipase (LPL) resulting in reduced uptake of fatty acids and triglycerides in adipocytes (4).

Finally, alteration of cholesterol metabolism appears to be another important reason. Germ-free mice on HFD display reduced plasma cholesterol levels and higher hepatic cholesterol contents and excrete more fecal cholesterol compared with their conventional counterparts (79). This may be related to upregulated hepatic nSREBP-2, upregulated cholesterol biosynthesis genes such as the rate-limiting enzyme in cholesterol biosynthesis 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR), and upregulated membrane transporters for cholesterol excretion in liver and small intestine ATP-binding cassette subfamily G members 5 and 8 (ABCG5/8; 31, 79, 82). Besides, cholesterol is converted to coprostanol by the gut microbiota in conventional rodents and in humans (25, 32, 36). The conversion of cholesterol to coprostanol therefore does not occur in germ-free mice and is another explanation for elevated levels of fecal cholesterol.

In summary, germ-free rodents are either protected from or more susceptible to NAFLD depending on their strain and on the specific composition of the diet. Germ-free C57BL/6J and C3H/HeJ mice are resistant to NAFLD, partially secondary to reduced energy uptake, increased lipid excretion, enhanced insulin sensitivity due to increased AMPK activity, upregulated intestinal expression of Fiaf, and altered cholesterol metabolism. Conversely, germ-free Fischer rats can be protected from or more susceptible to NAFLD depending on the contribution of specific dietary ingredients such as choline, cystine, and cholesterol. Germ-free Lobund rats are more prone to liver cirrhosis due to choline-deficient diet than their conventional counterparts.

LIVER CANCER/HEPATOCELLULAR CARCINOMA

The incidence/mortality of liver cancer is influenced by sex, age, race/ethnicity, location (3), level of dietary protein (94), caloric intake (101), and the intestinal microbiota (13).

Liver tumorigenesis in C3H/He male mice that were gnotobiotic, i.e., mice with only one known bacterial strain or a specific set of known bacterial strains, was promoted by a bacterial combination of E. coli, Streptococcus faecalis, and Clostridium paraputrificum (67, 68). Monoassociated mice with Bifidobacterium longum, Lactobacillus acidophilus, or Eubacterium rectale exhibited an improved cancer burden; nevertheless, significantly more rodents with liver tumors, significantly more tumor nodules per liver, and significantly higher tumor sizes were found in these mouse groups than in germ-free mice (68).

Aging conventional and germ-free Lobund-Wistar rats (19–41 mo) had a similar incidence of HCC (73). However, germ-free Sprague-Dawley rats were protected from liver cancer induced by methylazoxymethanol-β-d-glucosiduronic acid (MAM-GlcUA; 50). Similarly, germ-free C3H mice were less susceptible to the induction of hepatoma by 7,12-dimethylbenz(a)anthracene (DMBA) or phenobarbital (39, 66). This holds true in particular with regard to early development of liver tumors (before 40 wk) in germ-free C3H mice given DMBA shortly after birth; yet, germ-free mice of older age (>80 wk) lose that protective effect (83). Germ-free C57BL/6 mice were also protected against liver cancer induced by diethylnitrosamine (DEN) and carbon tetrachloride (CCl₄; 26). Germ-free mice had a significantly lower number of tumor cells, with smaller tumor size and liver weight-to-body weight ratio, than conventional mice (26, 66, 67), which was secondary to absent LPS and lack of Toll-like receptor 4 (TLR4) signaling in germ-free rodents (26). Consistent with this, mice treated with broad-spectrum antibiotics as well as TLR4-inactivated mutant mice were overall protected against carcinogenesis due to DEN and CCl₄ and displayed a significant increase in cleaved caspase-3-positive cells in hepatocytes, and the number of cleaved caspase-3-positive cells in these mice was inversely correlated with tumor number, tumor size, and liver weight-to-body weight ratio (26). Additionally, treatment with low-dose LPS over 12 wk led to increased tumor number, tumor size, and ratio of liver weight to body weight (26). These results indicate that the LPS-TLR4 pathway promotes hepatocarcinogenesis. This pathway, among others, induces IL-1β, which promotes inflammation, stimulates hepatic stellate cells to produce fibrogenic mediators, and activates human liver progenitor cells, eventually leading to cirrhosis and HCC (13). Besides LPS, there are also other bacterial products or pathogen-associated molecular patterns that activate the immune system but are rarely measured, such as peptidoglycan, lipoproteins, lipoteichoic acid, double-stranded RNA, and unmethylated DNA (40). Measuring these pathogen recognition receptor ligands and their downstream signaling could broaden the understanding of how gut-derived products and metabolites affect HCC onset and progression. The intestinal microbiome also impacts liver tumorigenesis by other means: Alterations of gut microbiota induced by obesity lead to increased concentrations of deoxycholic acid, which induces secretion of various inflammatory and procarcinogenic factors in the liver and thus promotes HCC development (115).

Hence, germ-free rodents were less susceptible to the induction of hepatic tumors by chemicals, which was in part secondary to an absent metabolism of gut microbiota and lack of gut-derived LPS and related TLR4 activation.

AUTOIMMUNE HEPATITIS

The etiology of AIH is poorly understood. Immune tolerance breakdown, genetic predisposition, and environmental factors may all contribute to triggering the autoimmune process (62, 116). Recently, the intestinal microbiome has been shown to be involved in AIH development in patients with decreasing fecal bifidobacteria and lactobacilli concentrations as well as increasing plasma LPS in later stages of AIH (58).

Germ-free mice are protected against experimental AIH, as leukocyte infiltration and levels of inflammatory cytokines including IFN-γ, TNF-α, IL-4, monocyte chemotactic protein-1 (MCP-1), granulocyte colony-stimulating factor (G-CSF), C-X-C motif chemokine 1 (CXCL1) [or keratinocyte chemoattractant (KC)], granulocyte-macrophage colony-stimulating factor (GM-CSF), eotaxin, macrophage inflammatory protein-1β (MIP-1b), and MIP-1a were significantly lower, whereas apoptosis was nearly undetectable in the liver of concanavalin A (ConA)-treated germ-free BALB/c mice relative to controls (109). This may be due to the deficient activation of natural killer T cells in germ-free mice, resulting from lower levels of presented bacterial glycolipid antigens compared with controls (109). Additionally, bacteria-derived systemic LPS was significantly increased in ConA-treated conventional mice but undetectable in their germ-free counterparts (109), which could be another reason for the alleviated liver damage in germ-free mice, as LPS can activate hepatic Kupffer cells and lead to liver injury (105). Similarly, germ-free Swiss Webster mice also showed lower alanine aminotransferase (ALT) levels and liver histopathological scores after ConA treatment relative to conventionalized controls, which were previously germ-free but orally gavaged with fecal contents of conventional mice (20).

In summary, germ-free mice are resistant to ConA-induced liver injury, and this may be explained by lack of bacterial translocation including LPS and deficient activation of natural killer T cells.

AUTOIMMUNE CHOLANGIOPATHIES

Primary sclerosing cholangitis (PSC) and primary biliary cholangitis (PBC) are clinically heterogeneous diseases with genetic, immunological, environmental, and other contributory components (19, 55, 97, 110). In addition, intestinal microbiota also plays a pathophysiological role in PSC (64, 86, 97, 110) and PBC (64, 78, 100). Bile acid metabolism is central for the pathogenesis of PSC and PBC. Bacteria metabolize, dehydrogenate, and deconjugate primary to secondary bile acids and degrade taurine. Germ-free rodents thus have markedly lower concentrations of unconjugated and higher concentrations of taurine-conjugated bile acids in liver, kidney, heart, and plasma as well as higher free taurine levels in the liver relative to their conventional counterparts. They furthermore exhibit a markedly different metabolism with significantly altered farnesoid X receptor (FXR)/retinoid X receptor (RXR) activation, hepatic cholestasis, steroid synthesis, and fatty acid metabolism pathways (96).

In a model of experimental PSC, germ-free ATP-binding cassette subfamily B member 4 (also known as multidrug resistance protein-2) knockout [Abcb4^−/− (Mdr2^−/−)] mice displayed more pronounced biochemical and histological features of PSC than their conventional counterparts (99). Serum biochemical parameters, including alkaline phosphatase, aspartate aminotransferase, ALT, and bilirubin, were significantly higher in germ-free Abcb4^−/− mice, with more severe fibrosis, ductular reaction, and ductopenia (99). This susceptibility is related to increased cholangiocyte senescence and the absence of ursodeoxycholic acid, which reduces insult-induced cholangiocyte senescence, as shown in vitro (99). Senescent cells can transition to a potentially pathological state referred to as senescence-associated secretory phenotype (17, 98, 102). Cells of senescence-associated secretory phenotype can, in turn, alter their microenvironment, secrete proinflammatory cytokines and growth factors, and thereby induce senescence in neighboring cells (17). Conversely, in a murine model of PBC, germ-free NOD.c3c4 mice exhibited less portal inflammation and smaller liver weights in relation to their conventionally raised counterparts while displaying a similar level of fibrosis and dilated intrahepatic bile ducts (90).

Hence, germ-free Abcb4^−/− mice exhibit a worse PSC phenotype compared with conventional Abcb4^−/− mice, which is likely related to an altered bile acid metabolism (96), whereas germ-free NOD.c3c4 mice show some biliary/hepatic protection relative to conventional NOD.c3c4 mice. The discrepancy could possibly be explained by the fact that the disease process in Abcb4^−/− mice is driven by changes in bile composition, whereas biliary disease in NOD.c3c4 mice is mainly immune driven (90, 99). Correspondingly, there was no difference in the inflammatory markers CD4⁺ and CD8⁺ cells and IL-10 in the Abcb4^−/− mouse model between the groups, but there was lower inflammation [decreased CD3⁺, lymphocyte antigen 6 complex, locus G-positive (Ly6G⁺), and galectin-3-positive (Mac-2⁺) cells] in germ-free NOD.c3c4 mice relative to their conventional counterparts (90, 99). Further studies are necessary to clarify the role and mechanisms of the PSC and PBC phenotypes under germ-free conditions.

ACUTE (TOXIC) LIVER INJURY AND LIVER FIBROSIS/CIRRHOSIS

Germ-free C3H/HeH mice showed not significantly different ALT levels and liver necrosis after acetaminophen-induced liver insult compared with conventional controls (76). On the other hand, studies in Carworth Farms Swiss Webster mice (74) and Wistar rats (95) showed that germ-free rodents developed more pronounced liver necrosis and higher ALT/aspartate aminotransferase levels within 48 h after administration of dimethylnitrosamine (DMN) or dimethylamine (DMA) plus sodium nitrite than conventional controls and are hence more prone to that type of acute toxic liver insult. Similarly, germ-free ICR mice subjected to treatment with the antibiotic chlortetracycline over 4 days developed a higher rise of hepatic lipids compared with their conventional counterparts (30). Treatment with butter yellow (3′-methyl-4-dimethylaminoazobenzene; ip injections twice weekly for 6 wk) resulted in more liver necrosis and hepatocyte degeneration in germ-free Fischer rats as well as fibrosis, whereas conventional rats did not show any fibrosis (8). Hepatic regeneration following partial hepatectomy was markedly reduced in germ-free BALB/c mice in relation to conventional controls, which may be due to the lack of endogenous LPS in germ-free mice, as LPS has been shown to stimulate liver regeneration (24).

If the toxic liver injury recurs, liver fibrosis can develop and is characterized by excessive deposition of extracellular matrix proteins. It can result from any chronic liver disease including alcoholic liver disease, NASH, viral hepatitis, cholestatic liver disease, autoimmune liver disease, and toxin-related liver disease (89, 92). In experimental toxin-induced liver fibrosis secondary to administration of thioacetamide (TAA) in drinking water over 21 wk or 12 intraperitoneal injections of CCl₄ over 4 wk, germ-free C57BL/6 mice were found to be more susceptible to liver fibrosis relative to conventional mice, with more toxin-induced hepatic oxidative stress and liver cell death (65). Similarly, in studies performed more than 50 years ago, liver cirrhosis was observed in germ-free female ICR mice after 12 wk of twice weekly subcutaneous CCl₄ injections, whereas none was detected in their conventional female counterparts after 14 wk (cirrhosis was present in both germ-free and conventional male mice after 12 wk of treatment; 75). Exposure to some microbial products and metabolites might hence preserve liver homeostasis. The more pronounced liver damage in germ-free mice may be due to lower levels of P450 enzymes such as the phase I xenobiotic-metabolizing enzyme CYP26A1 (possibly allowing an accumulation of toxic products, increased oxidative stress, and resulting hepatotoxicity) and absent indole-3-propionic acid (IPA; 65). IPA is a deamination product of dietary tryptophan, is a powerful antioxidant, and is neuroprotective and hepatoprotective (23, 47). Production of IPA depends on the presence of the gut microflora, in particular Clostridium sporogenes (111).

In experimental cirrhosis with hepatic encephalopathy due to gavage with CCl₄, germ-free C57BL/6 mice exhibited less systemic inflammation, neuroinflammation, and microglial/glial activation in relation to their conventional counterparts (46). However, they showed a significantly higher increase in serum ammonia levels after cirrhosis development compared with germ-free controls than did conventional mice with cirrhosis compared with conventional controls (46). The elevated ammonia level may be due to upregulated small bowel glutaminase activity (46, 84), which has been shown to affect the occurrence of hepatic encephalopathy in humans (84). The reason for the increased small bowel glutaminase activity in germ-free mice is still unclear but could be related to disinhibition due to absent microbial metabolites.

Overall, the majority of germ-free rodents appear to be more susceptible to acute (toxic) liver injury and liver fibrosis/cirrhosis, which could be due to lack of metabolism of the toxic substances by the gut microbiome, a different intestinal absorption, hepatic uptake, or microsomal activation (95). The different phenotype could also be due to absent LPS in germ-free mice, as it was already shown decades ago (106) that LPS has multiple effects on the liver such as modulation of bile secretion and hepatic blood flow and could thereby result in a different resistance of the liver to toxins.

ANTIBIOTICS VERSUS GERM-FREE STATUS IN EXPERIMENTAL LIVER DISEASE

The complete absence of microbes confers sometimes protection, sometimes exacerbation of certain liver diseases. Does reducing the number of intestinal microbes using antibiotics lead to the same liver findings as reported with a complete absence of microbiota in germ-free mice? Although antibiotics can be associated with hepatotoxicity (10, 30), antibiotic treatment generally results in an improved hepatic phenotype in a multitude of experimental liver diseases including alcoholic liver disease (2, 22), NAFLD (9, 44), NASH (43, 114), HCC (26, 115), AIH (20, 48), PSC (72), PBC (90), toxic liver injury (38, 48), and liver fibrosis/cirrhosis (91, 117; Table 2).

Thus, chronic overstimulation via toxins, metabolites, and TLR ligands present in conventional rodents is harmful. However, no exposure to bacteria and their components and metabolites, as seen in germ-free animals, might not be helpful either.

GNOTOBIOTIC MICE AS MODEL TO STUDY HUMAN MICROBIOTA

Germ-free animals are somewhat artificial systems and might hence not produce physiologically meaningful results as exemplified by discordant findings in germ-free (Table 1) and in antibiotic-treated rodents (Table 2). How can we use them to mirror reality better? We can create gnotobiotic rodents colonized with only one known bacterial strain (monoassociated) or a specific set of known bacterial strains to examine the impact on the disease process; this allows investigation of the contribution of single or a specified set of bacteria to disease pathogenesis. The latter is sometimes hard to achieve, as humans and mice are colonized by multiple and often poorly defined microbes.

We can also “humanize” germ-free mice, or colonize them with human feces, to reproduce human disease better in rodents (Fig. 1). This could be done in cases where only suboptimal rodent disease models exist such as in alcoholic hepatitis: Germ-free mice were transplanted with human microbiota from patients with alcoholism and no hepatitis, nonsevere alcoholic hepatitis, or severe alcoholic hepatitis, which conferred susceptibility to more severe liver disease in mice transplanted with microbiota from the sickest patients (59). The link between intestinal dysbiosis and alcoholic liver disease was further strengthened by subsequent transplantation of microbiota from patients with no alcoholic hepatitis to mice previously transplanted with microbiota from patients with severe alcoholic hepatitis as these mice exhibited markedly improved ALT levels compared with mice that did not receive the subsequent fecal transplant (59). Studies as such could not be executed in humans because of ethical concerns, nor could they achieve such clinical relevance without humanizing rodents with human microbiota. Humanizing animals can also help us to understand carcinogenesis: After administration of the carcinogen azoxymethane, germ-free mice/antibiotic-treated mice were gavaged with microbiota from patients with colorectal cancer and were found to be more susceptible to tumorigenesis than mice gavaged with microbiota from healthy controls or gavaged with no microbiota (113). Again, these studies cannot be performed in humans, nor would an animal study have as much relevance for human disease as humanized models can have. Humanized models can also be used for precision medicine approaches such as precise editing of the intestinal microbiome: Mice transplanted with intestinal microbiota from patients with inflammatory bowel disease showed significantly decreased concentrations of Enterobacteriaceae after treatment with tungstate that selectively blocked metabolic pathways in Enterobacteriaceae; this was associated with markedly improved intestinal inflammation (121).

Hence, gnotobiotic or humanized models, despite being more costly than germ-free or conventional rodents, appear to be alternatives to overcome oftentimes dissimilar phenotypes between rodents and humans, in particular in microbiome research. However, posttransplant analyses of the intestinal microbiome should be performed to confirm successful transplantation, as not all bacteria are being transferred 1:1 because of a different microenvironment.

SUMMARY AND CONCLUSIONS

Germ-free rodents fulfill a central function in elucidating the role of the intestinal microbiota in the pathogenesis of liver diseases. Germ-free status can occasionally have divergent effects in different mouse or rat strains and even within the same strain can show completely opposite outcomes in different liver diseases. Antibiotic treatment is usually associated with ameliorated liver disease in preclinical models. The different phenotypes of germ-free rodents in various liver diseases in relation to conventional rodents can be explained by an immature immune system and a markedly altered metabolism with regard to lipids, cholesterol, xenobiotics, toxins, and bile acids. To advance understanding of the interactions between host and intestinal microbiota, simplified model systems such as gnotobiotic mice monoassociated with a single bacterial strain or colonized with a defined set of microbes constitute unique models for investigation of this complex ecosystem. In addition, humanized gnotobiotic mice allow us to study bacteria not present in mice and to exclude bacteria found in mice but not in humans. Humanized gnotobiotic mice will be very useful when microbiome-centered therapies are tested to treat human disease.

GRANTS

This study was supported in part by National Institutes of Health Grants R01-AA-020703, R01-AA-24726, U01-AA-021856, and U01-AA-026939 and by Award I01BX002213 from the Biomedical Laboratory Research & Development Service of the Veterans Affairs Office of Research & Development (to B. Schnabl).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

P.H. and H.C. drafted manuscript; P.H., H.C., Y.D., and B.S. edited and revised manuscript; P.H., H.C., Y.D., and B.S. approved final version of manuscript.