Abstract

Key summary

1. Summarize the established knowledge of the subject:

   • Intestinal microbiota is one component of the “multiple hits” in the pathogenesis of nonalcoholic fatty liver disease (NAFLD).

   • Short-chain fatty acids (SCFAs) are the most abundant bacterial products derived from commensal bacterial fermentation of dietary fibers in intestines.

   • Fecal SCFAs have been proposed to influence the development of NAFLD, but their exact role in NAFLD pathogenesis remains unclear.

2. What are the significant and/or new findings of the study?

   • Intestinal microbiota was characterized by higher abundance of Fusobacteria and Fusobacteriaceae in nonalcoholic steatohepatitis (NASH) patients compared to nonalcoholic fatty liver (NAFL) and healthy controls.

   • Higher fecal SCFA concentrations and domination of SCFA-producing bacteria were observed in NAFLD patients.

   • SCFA concentrations were associated with immunological features of NAFLD disease progression.

   • Intestinal microbiota and fecal SCFA concentrations may be important targets for future personalized medicine in NAFLD patients.

Introduction

Nonalcoholic fatty liver disease (NAFLD) has growing clinical and socioeconomic impact. The pathogenesis of NAFLD is described by “multiple hits” on a nonalcoholic fatty liver (NAFL) that progress to nonalcoholic steatohepatitis (NASH). The intestinal microbiota is one of the “multiple hits” and is the focus of recent studies in NASH patients with different technical and methodical approaches leading to some inconsistent findings. In one study NASH patients showed a lower percentage of Bacteroidetes (Bacteroides/Prevotella) compared to NAFL patients and healthy controls (HCs),¹ whereas in another study higher abundance of Bacteroides and lower abundance of Prevotella were observed in NASH patients.² Loomba et al. detected higher abundance of Firmicutes in mild/moderate NAFLD and higher abundance of Proteobacteria in NAFLD patients with advanced fibrosis (no change for Bacteroidetes).³ Different mechanisms are proposed for the microbial contribution in NAFLD including regulation of gut barrier and inflammatory responses as well as bacterial metabolites, e.g. short-chain fatty acids (SCFAs).⁴

SCFAs such as acetate, propionate and butyrate are the most abundant bacterial products derived from commensal bacterial fermentation of dietary fibers in intestines; their role in obesity development is still controversial. On the one hand, SCFAs provide additional energy to the host and play a pivotal role in lipogenesis and gluconeogenesis, but on the other hand SCFAs increase satiety by activation of free fatty acid receptors and weight loss can thereby be sustained.⁵ In human studies fecal SCFA concentrations were higher in obese volunteers compared to lean individuals.^6,7 In animal and in vitro studies, SCFAs affect proinflammatory cytokine production,⁸ promote conversion toward regulatory T-cells (Tregs) under specific conditions⁹ and enhance Foxp3 expression of colonic T-cells via G protein coupled receptor (GPCR) 43 activation,¹⁰ but also facilitate conversion of naïve T-cells into Th1 or Th17 cells depending on cytokine milieu and histone deacetylase inhibitor activity.¹¹

In this study, we analyzed for the first time fecal SCFA concentrations together with gut microbiota composition in obese patients with histology-proven NASH and (NAFL) for a better understanding of functional changes of the fecal microbiota. Further integrative analysis of peripheral immune cell changes as characteristics of disease progression (NASH)¹² were performed and interpreted together to gain insight into a potential interplay between intestinal bacteria composition, their metabolites and peripheral immune cells.

Methods

Patient characteristics

The study was approved by the local ethics committee (University of Würzburg: EK 96/12, 5 September 2012) and conforms to the ethical guidelines of the 1975 Declaration of Helsinki. After written informed consent was obtained, 32 NAFLD patients (14 NAFL, 18 NASH) and 27 HCs were enrolled in this prospective study. An overview of study participants is depicted in Figure 1. Study participants were all >18 years old. Twenty-four patients were morbidly obese and underwent bariatric surgery, and eight obese NAFLD patients were included without bariatric surgery, but with histologically diagnosed NAFLD. Significant alcohol consumption (women >20g/d and men >30g/d) and other liver diseases such as chronic hepatitis B virus, hepatitis C virus infections as well as autoimmune liver diseases (hepatitis B surface antigen, anti-HCV, antinuclear antibodies) were excluded. Patients had no standardized diet before surgery. In general, a low-carbohydrate diet is recommended two weeks prior to surgery. Patients without bariatric surgery had no specific diet. Blood samples were taken at the day of inclusion while fecal samples were harvested immediately before bariatric surgery. Bariatric surgery followed up to 14 days after study inclusion. Patients had not received any antibiotic treatment the last six months prior to study inclusion. Biopsies were taken from subcapsular liver tissue during bariatric surgery or percutaneously from the right liver lobe in participants without bariatric surgery. All liver biopsies were analyzed by an experienced pathologist and classified using the Steatosis, Activity and Fibrosis (SAF) score.¹³ HCs were all individuals without evidence or history of liver or metabolic diseases. In HCs liver ultrasound and liver stiffness as determined by transient elastography measurement (FibroScan®) were normal. In HCs liver enzyme levels were measured and laboratory analysis showed no elevation for Glutamate oxalacetate tranasminase (GOT) and Glutamate pyruvate transaminase (GPT).

Open in a separate window

Figure 1.

Flow diagram with study participants including NAFLD and healthy controls. NAFL: nonalcoholic fatty liver; NAFLD: nonalcoholic fatty liver disease; NASH: nonalcoholic steatohepatitis.

Results

Changes in fecal microbiota in NAFLD patients and HCs

First, we analyzed α-diversity of the respective study groups. The Shannon diversity index showed a significantly reduced diversity in NAFL (p<0.01) and NASH (p<0.01) patients compared to HCs (Figure 2(a)). At the phylum level, NASH patients had a significantly higher abundance of Fusobacteria compared to NAFL patients (p=0.022) and to HCs (p^uncorr=0.027) (Table 1). On the family level substantial and significant differences between the study groups were observed (Table 1, Figure 2(b)). NASH patients were characterized by a higher frequency of Fusobacteriaceae (p^uncorr=0.002) in comparison to NAFL patients and higher abundance of Fusobacteriaceae (p=0.009) and Prevotellaceae (p^uncorr=0.009) compared to HCs. NAFL patients had a higher frequency of Prevotellaceae (p^uncorr=0.034) in comparison to HCs. Further detailed results on phylum and family level are listed in Table 1.

Open in a separate window

Figure 2.

(a) Reduced microbial diversity in NAFLD patients depicted by Shannon index. *p<0.05. (b) Overview of microbial composition on family level in NAFL, NASH and HCs. HCs: healthy controls; NAFL: nonalcoholic fatty liver; NAFLD: nonalcoholic fatty liver disease; NASH: nonalcoholic steatohepatitis.

Table 1.

Differences on phylum and family levels.

		Normalized count mean		SE	logFC	p uncorr	p corr
Phylum
NASH vs HCs		HCs	NASH
	k__Bacteria_unclassified	953	67	0.564	–2.320	0.000	0.000
	Fusobacteria	19	203	1.109	2.461	0.027	0.087
	Cyanobacteria	1176	126	0.991	–2.139	0.031	0.087
	Tenericutes	783	44	0.832	–1.756	0.035	0.087
NAFL vs HCs		HCs	NAFL
	k__Bacteria_unclassified	953	28	0.657	–2.251	0.001	0.006
	Verrucomicrobia	4194	213	0.782	–1.833	0.019	0.096
	Tenericutes	783	3	1.301	–2.749	0.035	0.115
NASH vs NAFL		NAFL	NASH
	Fusobacteria	2	203	1.249	3.823	0.002	0.022
Family
NASH vs HCs		HCs	NASH
	Odoribacteraceae	1133	200	0.300	–1.208	0.000	0.002
	k__Bacteria_unclassified	953	67	0.476	–1.862	0.000	0.002
	Fusobacteriaceae	19	203	1.043	3.609	0.001	0.009
	o__RF32_unclassified	971	25	0.863	–2.715	0.002	0.020
	Prevotellaceae	6904	9177	0.955	2.494	0.009	0.087
	Shewanellaceae	18	36	0.643	1.471	0.022	0.177
	Halomonadaceae	43	77	0.479	0.994	0.038	0.261
NAFL vs HCs		HCs	NAFL
	k__Bacteria_unclassified	953	28	0.555	–1.495	0.007	0.157
	p__Firmicutes_unclassified	4510	27	0.947	–2.541	0.007	0.157
	Prevotellaceae	6904	7696	1.085	2.298	0.034	0.490
	Odoribacteraceae	1133	135	0.350	–0.687	0.049	0.530
NASH vs NAFL		NAFL	NASH
	Fusobacteriaceae	2	203	1.253	3.949	0.002	0.070
	Barnesiellaceae	589	403	0.639	–1.309	0.040	0.869

Open in a separate window

HCs: healthy controls; NAFL: nonalcoholic fatty liver; NASH: nonalcoholic steatohepatitis.

Normalized count mean, standard error (SE) and fold changes (logFC) are given on the log2 scale of gut microbiome phyla and families in NAFL, NASH and HCs. p^corr=p value adjusted for multiple testing and p^uncorr=p value not adjusted for multiple testing. Only phyla and families are listed if p^uncorr≤0.05.

For the analysis on species level, 126 species were applied. Table 2 depicts detailed results, including all species with p^uncorr≤0.05. On the species level, NASH patients were characterized by a significantly higher abundance of Fusobacterium (p=0.033) as well as of Prevotella, Eubacterium biforme and Enterobacteriacae_unclassified (p^uncorr=0.003, p^uncorr=0.016, p^uncorr=0.016) compared to NAFL patients. Further differences on species levels between NASH/NAFL and HCs are listed in Table 2.

Table 2.

Differences on species levels.

Species		Normalized count mean		SE	logFC	p uncorr	p corr
NASH vs. HCs		HCs	NASH
	Acidaminococcus	4	1473	1.280	5.788	0.000	0.000
	Lachnospira	3556	539	0.353	−1.582	0.000	0.000
	Odoribacter	942	139	0.365	−1.492	0.000	0.002
	Bacteria_unclassified	953	67	0.494	−1.761	0.000	0.008
	Megasphaera	255	2442	1.076	3.817	0.000	0.008
	Fusobacterium	19	203	1.048	3.713	0.000	0.008
	Ruminococcus.torques	212	879	0.491	1.600	0.001	0.019
	Parabacteroides	2283	200	0.496	−1.546	0.002	0.027
	Eubacterium.biforme	117	559	1.010	2.922	0.004	0.050
	Prevotella.copri	4945	8450	0.998	2.801	0.005	0.054
	RF32_unclassified	971	25	0.922	−2.584	0.005	0.054
	Alistipes.finegoldii	542	19	0.707	−1.829	0.010	0.094
	Escherichia.coli	3988	6990	0.699	1.638	0.019	0.172
	Enterobacteriaceae_unclassified	117	784	0.662	1.496	0.024	0.188
	Shewanella.algae	18	36	0.602	1.358	0.024	0.188
	Halomonas	42	76	0.470	0.945	0.045	0.326
	Bifidobacterium.longum	1732	579	0.505	−0.979	0.053	0.354
	Enterococcus	9	107	0.996	1.917	0.054	0.354
NAFL vs HCs		HCs	NAFL
	Prevotella.copri	4945	7455	1.066	3.433	0.001	0.050
	Acidaminococcus	4	1750	1.592	5.080	0.001	0.050
	Bacteria_unclassified	953	28	0.551	−1.707	0.002	0.050
	Firmicutes_unclassified	4510	27	0.908	−2.813	0.002	0.050
	Odoribacter	942	89	0.361	−1.104	0.002	0.050
	Lachnospira	3556	307	0.419	−1.258	0.003	0.050
	Rikenellaceae_unclassified	2816	420	0.432	−1.110	0.010	0.165
	Ruminococcus.torques	212	391	0.555	1.300	0.019	0.270
	Dialister	3103	4181	0.757	1.715	0.024	0.296
	Enterobacteriaceae_unclassified	117	5	0.887	−1.792	0.043	0.491
	Anaerotruncus	161	16	0.521	−1.031	0.048	0.491
NASH vs NAFL		NAFL	NASH
	Fusobacterium	2	203	1.241	4.494	0.000	0.033
	Prevotella	20	466	0.842	2.465	0.003	0.192
	Roseburia	328	273	0.455	−1.127	0.013	0.333
	Eubacterium.biforme	20	559	1.036	2.492	0.016	0.333
	Enterobacteriaceae_unclassified	5	784	1.287	3.091	0.016	0.333
	Parabacteroides	744	200	0.551	−1.308	0.018	0.333
	Dialister	4181	1707	0.650	−1.332	0.040	0.653
F0/F1 vs≥F2		F0/F1	≥F2
	Alistipes.onderdonkii	285	31	0.778	−2.566	0.001	0.092
	Clostridium.citroniae	87	12	0.851	−2.378	0.005	0.246
	Acidaminococcus	219	1254	1.492	3.331	0.026	0.665
	Prevotella	71	395	1.212	2.662	0.028	0.665

Open in a separate window

HCs: healthy controls; NAFL: nonalcoholic fatty liver; NASH: nonalcoholic steatohepatitis.

Normalized count mean, standard error (SE) and fold changes (logFC) are given on the log2 scale of gut microbiome species in NAFL, NASH and HCs as well as in advanced diseases (F0/F1 and ≥F2 fibrosis). p^corr=p value adjusted for multiple testing and p^uncorr=p value not adjusted for multiple testing. Only species are listed if p^uncorr≤0.05.

ß-diversity analysis showed different profiles of gut microbiome data

PCA on operational taxonomic unit (OTU) level separates NAFLD patients (including NAFL and NASH patients) from HCs on the first axis as depicted in Figure 3(a). No significant difference was observed in the PCA on the OTU level between bariatric patients and obese patients without bariatric surgery to exclude preoperative diet as possible bias (Supplemental Figure 1). In our previous work NASH patients were successfully differentiated and characterized by increased Th17 (CD4+IL17+)/resting Treg (rTreg/CD4+CD45RA+CD25++) ratio in peripheral blood compared to NAFL patients.¹² Here we observed a significant correlation (R²=0.171, p value<0.01) of Th17/rTreg ratio, with the ordination configuration of the first two axes (Figure 3(a)). In Figure 3(b) PCA of microbiome data separate clearly the study cohort into high- and low-Th17/rTreg ratio subgroups (Th17/rTreg low<0.38, Th17/rTreg high≥0.38). There was thus a good correlation between the observed changes in the gut microbiome and peripheral T-cells. In a supervised ordination approach (RDA), NAFL, NASH and HCs could be successfully differentiated as well (Figure 3(c)). A permutation test based on 10,000 random permutations yields a highly significant effect (p<0.001) of the study group factor.

Open in a separate window

Figure 3.

(a) Principal component analysis of gut microbiome data in NAFL, NASH and HCs with positive correlation to peripheral immune cell ratio Th17/rTreg. (b) Principal component analysis of gut microbiome data in patients with high- and low-Th17/rTreg ratio. (c) Redundancy analysis of gut microbiome data based on the study group (NAFL, NASH and HCs). HCs: healthy controls; NAFL: nonalcoholic fatty liver; NAFLD: nonalcoholic fatty liver disease; NASH: nonalcoholic steatohepatitis; rTreg: resting regulatory T-cell; Treg: regulatory T-cell.

Higher fecal SCFA concentrations as well as SCFA-producing bacteria in NAFLD patients

Acetate, propionate as well as butyrate concentrations were increased in feces of NAFLD patients. NAFL as well as NASH patients had significantly higher fecal concentrations of acetate and propionate compared to HCs. A trend toward higher fecal butyrate levels in NASH patients compared to HCs was observed (Figure 4(a)).

Open in a separate window

Figure 4.

(a) Fecal acetate, propionate and butyrate concentrations in HCs, NAFL and NASH patients. *p<0.05. Median±IQR; acetate: HCs 17.4±7.6mM, NAFL 29.5±17.1mM, NASH 25.6±9.2mM; propionate: HCs 4.4±2.2mM, NAFL 10.4±8.7mM, NASH 9.0±4.5mM; butyrate: HCs 3.2±2.2mM, NAFL 3.6±3.7mM, NASH 4.1±4.2mM. (b) Fecal acetate, propionate and butyrate concentrations in F0 of NAFLD, NAFLD patients with no fibrosis (F0/F1) and moderate/advanced fibrosis (F≥2). *p<0.05. Median±IQR. HCs: healthy controls; IQR: interquartile range; NAFL: nonalcoholic fatty liver; NAFLD: nonalcoholic fatty liver disease; NASH: nonalcoholic steatohepatitis.

Bacterial communities were analyzed regarding their functional capacity of SCFA production based on previously published data.^19–22 At the phylum and family levels, higher abundance of Fusobacteria and Fusobacteriaceae and higher frequency of Prevotellaceae were observed. Fusobacteriaceae as well as Prevotellaceae are bacterial families characterized by SCFA-producing capacity.^19,20,22 Differences in species levels between NAFLD (NAFL/NASH) patients and HCs showed a significantly higher abundance of species such as Prevotella copri, Megashpaera, Fusobacterium, Ruminococcus torques and Eubacterium biforme (Table 2). Interestingly, these species are well described as SCFA-producing bacteria or belong to families with high capacity of SCFA production.^19,20,22 Thus the observed differences in family and species level in gut microbiome analysis of NAFLD patients are functionally relevant and correlate well with higher fecal concentrations of acetate and propionate in these patients (Figure 4(a)).

Different microbial composition and SCFA production in patients with moderate/advanced fibrosis

Microbial composition and fecal SCFA concentrations were different in NAFLD patients with advanced disease. Patients with significant fibrosis had higher frequency of Acadaminococcus and Prevotella (p^uncorr=0.026, p^uncorr=0.028) (Table 2). No significant differences were observed in family level between F0/F1 patients and F≥2 patients (Supplemental Table 3).

NAFLD patients with significant fibrosis (F≥2 fibrosis) had reduced fecal acetate as well as a trend toward reduced propionate concentrations (Figure 4(b)). No significant differences were observed for butyrate between F0/F1 and F≥2 patients (Figure 4(b)). In a supervised PCA, F0/F1 patients were successfully separated from F≥2 patients (Figure 5).

Open in a separate window

Figure 5.

Redundancy analysis of gut microbiome data for patients with no fibrosis (F0/F1) and moderate/advanced fibrosis (F≥2), (p<0.05, permutation test based on 10,000 permutations). PC: principal component; RDA: redundancy analysis.

Peripheral immune cell markers in NASH patients are associated with fecal SCFA levels

As a unique feature of this study, data of fecal SCFA concentrations were analyzed together with peripheral immune cell changes as depicted in Figure 6. A significant negative correlation between rTregs in peripheral blood and fecal propionate as well as acetate levels was observed. Peripheral Th17/rTreg ratio showed a positive correlation for fecal acetate and propionate concentrations (Figure 6).

Open in a separate window

Figure 6.

As previously described, NASH patients were characterized by an increased Th17/rTreg ratio in peripheral blood as well as in hepatic tissue compared to NAFL patients.¹² These immune cell changes correlated to fecal SCFA concentrations in this study. Correlations between fecal propionate, acetate and peripheral rTregs (propionate: Spearman rho –0.33, p<0.05; acetate: Spearman rho –0.34, p<0.05) as well as Th17/rTreg ratio (propionate: Spearman rho 0.36, p<0.01; acetate: Spearman rho 0.34, p<0.05). Spearman rho correlation coefficient is depicted in each graph. NAFL: nonalcoholic fatty liver; NASH: nonalcoholic steatohepatitis; rTreg: resting regulatory T-cell; SCFA: short-chain fatty acid.

Discussion

In this study, gut microbiome was analyzed in 32 NAFLD patients and 27 HCs together with fecal SCFA as functional bacterial metabolites. Differences in microbial composition were rigorously corrected for multiple testing. In our study, NASH patients could be differentiated from NAFL patients based on their intestinal microbiome by higher abundance of Fusobacteria (phylum), Fusobacteriaceae (family), Fusobacterium (species) as well as Prevotella, Eubacterium biforme and Enterobacteriacae_unclassified (Tables 1 and ​and22).

In past years, different studies showed partly controversial results about changes in the intestinal microbiome of NAFLD patients.^1,2,23 These differences could be explained by the complexity of obesity in the context of metabolic syndrome as well as the heterogeneity of methodological data assessment and data interpretation of published studies (including inconsistent correction for multiple testing).²⁴

An increase in the phylum Fusobacterium and/or of some species in this phylum has recently been described in Asian obese patients.²⁵ Fusobacterium is associated with a higher expression and activation of inflammatory signals and has been identified as an important pathogen in patients with acute and chronic periodontitis.^26,27 As a functional link between microbiome and active metabolites, higher prevalence of Fusobacterium and correlation to SCFA concentrations in root canals was observed in subgingival biofilm samples from periodontal pockets of type 2 diabetic patients.²⁷ Recent studies indicated an interrelationship between periodontitis and systemic diseases (e.g. cardiovascular disease, type 2 diabetes and NAFLD) through bacterial translocation, bacteremia and increase in circulating inflammatory mediators.^28,29 Thus, the higher abundance of Fusobacteria in gut microbiota in our study and the impact of Fusobacteriaceae in the context of metabolic diseases could be attributed to direct or indirect activation of inflammatory signals as well as to production of bacterial metabolites such as SCFAs.

Following recent research fecal SCFA concentrations were analyzed and NAFLD patients were characterized by significantly higher fecal acetate and propionate in our study (Figure 4(a)). Rahat-Rozenbloom et al. observed higher fecal SCFA concentrations in overweight and obese individuals with no difference in rectal SCFA absorption or dietary changes.⁶ Increased abundance of Fusobacteria, Fusobacteriaceae as well as Prevotellaceae and the above-described changes in species level in NAFLD patients in our study are well in line with higher fecal SCFA concentrations due to predominance of SCFA-producing families and bacteria.^19,20,22 In a metagenomic analysis Loomba et al. revealed a different abundance of SCFA production proteins in NAFLD patients with advanced and mild/moderate fibrosis.³ The observed higher predicted abundance of acetate-producing enzymes in the microbiota of mild/moderate NAFLD patients matches higher fecal acetate concentrations in our study population (F0–F1) vs≥F2 (Figure 4(b)).

Despite increasing evidence of significant changes in various inflammatory diseases, the particular functional role of SCFAs in NAFLD remains to be defined. In light of increased fecal SCFAs in this study, NAFLD patients were characterized by a remarkable positive correlation between fecal propionate/acetate concentrations and peripheral Th17/rTreg ratio as well as a negative correlation to peripheral rTregs as two immunological features of progressive disease (Figure 6). Even though many studies support the idea that SCFAs induce an anti-inflammatory profile of T-cells, there is also evidence that SCFAs induce pro-inflammatory T-cells (e.g. Th1 and Th17) under specific conditions and depending on disease model.^11,30

The mechanistic role of SCFAs in the context of obesity and metabolic diseases in humans remains open since intestinal immune cell analysis was not possible in our study given the limitations of a human study of NAFLD and has never been performed elsewhere. Furthermore, in this study a limited number of NAFLD patients were analyzed because of a complex and simultaneous analysis of ex vivo PBMCs, liver histology and analysis of fecal samples. Finally, it is important to note that the present cohort includes a combination of patients undergoing bariatric surgery as well as those without bariatric surgery.

This study explored for the first-time the link between gut microbial changes in NAFLD patients and higher SCFA concentrations in feces as well as the significant correlation to peripheral immune cells as a marker of disease progression in NAFLD. Our data suggest that higher prevalence of SCFA-producing bacteria in feces of NAFLD patients could contribute to disease progression by maintaining low-grade inflammatory processes with influence on circulating immune cells and impact on peripheral target organs such as the liver. The present study further contributes to elucidate the still controversial role of SCFAs in the context of obesity in humans and targets the field for future studies of intestinal immune cell activation to better understand the final impact of fecal SCFAs in humans. Furthermore, increased permeability in the gut-liver axis as well as other bacterial metabolites (e.g. endotoxins, bacterial ethanol) and fecal bile acids are important targets influencing disease progression in NAFLD and should be the focus of future studies together with fecal SCFAs.^31,32

Supplemental Material

Acknowledgments

References