Intestinal microbiota in patients with nonalcoholic fatty liver disease

Authors

  • Marialena Mouzaki,

    1. Department of Pediatrics, Hospital for Sick Children, Toronto, Canada
    2. Toronto General Hospital, University Health Network, Toronto, Canada
    Search for more papers by this author
  • Elena M. Comelli,

    1. Department of Nutritional Sciences, University of Toronto, Toronto, Canada
    Search for more papers by this author
  • Bianca M. Arendt,

    1. Toronto General Hospital, University Health Network, Toronto, Canada
    Search for more papers by this author
  • Julia Bonengel,

    1. Toronto General Hospital, University Health Network, Toronto, Canada
    Search for more papers by this author
  • Scott K. Fung,

    1. Toronto General Hospital, University Health Network, Toronto, Canada
    2. Department of Medicine, University of Toronto, Toronto, Canada
    Search for more papers by this author
  • Sandra E. Fischer,

    1. Toronto General Hospital, University Health Network, Toronto, Canada
    2. Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Canada
    Search for more papers by this author
  • Ian D. McGilvray,

    1. Toronto General Hospital, University Health Network, Toronto, Canada
    2. Department of Medicine, University of Toronto, Toronto, Canada
    Search for more papers by this author
  • Johane P. Allard

    Corresponding author
    1. Toronto General Hospital, University Health Network, Toronto, Canada
    2. Department of Nutritional Sciences, University of Toronto, Toronto, Canada
    3. Department of Medicine, University of Toronto, Toronto, Canada
    • University Health Network, Toronto General Hospital, 585 University Ave., Suite 9-NU-973, Toronto, Ontario, Canada M5G 2N2===

    Search for more papers by this author
    • fax: 416-348-0065


  • Potential conflict of interest: Nothing to report.

  • Funded by the Canadian Institutes of Health Research (CIHR), Grants NMD-86922, MOP-89705, MOP-123459.

Abstract

Despite evidence that the intestinal microbiota (IM) is involved in the pathogenesis of obesity, the IM composition of patients with nonalcoholic fatty liver disease (NAFLD) has not been well characterized. This prospective, cross-sectional study was aimed at identifying differences in IM between adults with biopsy-proven NAFLD (simple steatosis [SS] or nonalcoholic steatohepatitis [NASH]) and living liver donors as healthy controls (HC). Fifty subjects were included: 11 SS, 22 NASH, and 17 HC. One stool sample was collected from each participant. Quantitative real-time polymerase chain reaction was used to measure total bacterial counts, Bacteroides/Prevotella (herein referred to as Bacteroidetes), Clostridium leptum, C. coccoides, bifidobacteria, Escherichia coli and Archaea in stool. Clinical and laboratory data, food records, and activity logs were collected. Patients with NASH had a lower percentage of Bacteroidetes (Bacteroidetes to total bacteria counts) compared to both SS and HC (P = 0.006) and higher fecal C. coccoides compared to those with SS (P = 0.04). There were no differences in the remaining microorganisms. As body mass index (BMI) and dietary fat intake differed between the groups (P < 0.05), we performed linear regression adjusting for these variables. The difference in C. coccoides was no longer significant after adjusting for BMI and fat intake. However, there continued to be a significant association between the presence of NASH and lower percentage Bacteroidetes even after adjusting for these variables (P = 0.002; 95% confidence interval = −0.06 to −0.02).

Conclusion:

There is an inverse and diet-/BMI-independent association between the presence of NASH and percentage Bacteroidetes in the stool, suggesting that the IM may play a role in the development of NAFLD. (HEPATOLOGY 2013 )

Nonalcoholic fatty liver disease (NAFLD) is a term used to describe a spectrum of hepatic pathology that is closely linked to obesity and ranges from simple steatosis (SS), to steatosis with inflammation and/or fibrosis (nonalcoholic steatohepatitis; NASH) and even cirrhosis.1 NAFLD is considered the hepatic manifestation of the metabolic syndrome. The pathophysiology of NAFLD is not entirely understood but it has been proposed to be the result of multiple “hits.”2 These include signals from the adipose tissue (e.g., fatty acids, cytokines), the diet (e.g., fructose, fatty acids), as well as the immune system.2 Insulin resistance (IR) is frequently found in patients with NAFLD and is thought to contribute to its pathogenesis partly by enhancing lipolysis within the adipose tissue, subsequently increasing the flux of free fatty acids into the liver.2 Research in the field of obesity has provided preliminary evidence that the intestinal microbiota (IM) may play a role in the development of obesity and the metabolic syndrome, suggesting a potential role in the pathogenesis of NAFLD.3

The intestinal lumen is populated by trillions of microorganisms that carry 150-fold more genes compared to the host, collectively referred to as the microbiome.4, 5 The IM is composed of bacteria, Archaea, yeasts, and viruses.6, 7 Despite significant interindividual variations in the IM of humans at lower taxonomical levels, the dominating phyla are Bacteroidetes and Firmicutes.8 Several human studies in the field of obesity have suggested differences in the IM between obese and lean individuals. Ley et al.9 showed an increased fecal Firmicutes-to-Bacteroidetes ratio in obese subjects but subsequent studies have shown inconsistent results, likely due to the uncontrolled effects of factors, such as diet and environment, as well as methodological issues that include variations in sample size and use of different techniques for the determination of the IM composition.10-12 Very few studies have explored the role of IM in NAFLD and, to our knowledge, there are no studies directly assessing the IM composition of adults with nonexperimental SS or NASH.13-15

Animal studies have shown that the IM can contribute to all the histological components of NAFLD: hepatic steatosis, inflammation, and fibrosis.3, 16-18 The IM have the potential to increase intrahepatic fat through mechanisms such as altered appetite signaling, increased energy extraction from the diet, altered expression of genes involved in de novo lipogenesis or β-oxidation, or by way of inflammation-driven steatosis.11, 16, 19-21 Hepatocellular inflammation may be secondary to altered intestinal permeability and translocation of either intact bacteria or microbial cell components (such as lipopolysaccharide [LPS] derived from the cell wall of gram-negative bacteria) to the circulation.22-26 Bifidobacteria specifically have been associated with enhanced integrity of the intestinal barrier of animals by way of increased expression of tight junction proteins.27 Lastly, the IM may contribute to hepatic fibrosis by way of direct activation of hepatic stellate cells by LPS or by way of stimulation of profibrotic pathways by Toll-like receptor (TLR)-9-dependent recognition of certain bacteria by Kupffer cells in the liver.28

Despite the amount of evidence providing pathogenetic links between the IM and various components of NAFLD, there are no published studies focused on assessing IM composition of adults with this condition. Recently, Zhu et al.29 reported differences in the IM of children with NASH compared to obese and normal-weight children. In that study, NASH was associated with higher levels of ethanol-producing bacteria, as well as increased serum ethanol levels.

The aim of our study was to assess if there are any differences in the IM of adults with biopsy-proven SS, NASH, and healthy controls (HC), taking into account potential confounders, such as dietary intake and body mass index (BMI).

Abbreviations

ALT, alanine aminotransferase; BMI, body mass index; BMR, basal metabolic rate; HC, healthy control; IM, intestinal microbiota; IR, insulin resistance; LPS, lipopolysaccharide; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; SS, simple steatosis.

Materials and Methods

General Protocol.

This was a cross-sectional study performed at the University Health Network, Toronto, Canada. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the appropriate Institutional Review Committee.

Patients referred to the hepatology clinics for persistently elevated liver enzymes and clinical suspicion of NAFLD were initially assessed as per standard of care to rule out other causes of liver disease. After 3-6 months of persistently elevated alanine aminotransferase (ALT) levels, patients underwent a liver biopsy to confirm the diagnosis of NASH and to assess its severity. During the initial visit, patients were invited to participate in this study. After providing written informed consent, they were instructed on how to collect and transport the stool sample and complete 7-day food records and 7-day activity logs. They were asked to return the stool sample and the food and activity records the morning of their liver biopsy. On the day of but prior to liver biopsy, a blood sample was taken for metabolic, nutritional, and hepatic parameters, as explained below.

Healthy subjects undergoing assessment for living donation by the Living Donor Liver Transplant Program at the University Health Network were invited to participate as controls. These subjects were rigorously assessed as per the protocol of the Transplant Program to ensure that they had no significant medical comorbidities. After obtaining informed consent, subjects were given the same instructions for stool sample and food record/activity log collection as the NAFLD patients. Samples were returned 1 week prior to liver donation. Blood samples were also collected at that time. Histology was obtained either during a predonation biopsy (done to verify the healthy state of the liver) or during hepatectomy.

Inclusion criteria were age >18 years and biopsy-confirmed NAFLD or healthy liver. Exclusion criteria were: liver disease other than NAFLD, anticipated need for liver transplantation within a year, or complications of endstage liver disease such as variceal bleeding or ascites; concurrent medical illnesses, contraindications for liver biopsy; use of medications known to cause or exacerbate steatohepatitis (such as corticosteroids) or antibiotics, pre- or probiotics in the preceding 6 months; consumption of more than 20 g of alcohol/day; use of vitamin E or fish oil supplements; chronic gastrointestinal diseases, previous gastrointestinal surgery modifying the anatomy; pregnancy or lactating state.

Clinical Data.

Patients provided information regarding medication use, alcohol consumption, and smoking history. Past medical and surgical history was recorded and, in addition, data on ethnicity were collected. Height and weight were measured and BMI was calculated.

Nutritional and Activity Assessment.

Subjects were asked to complete the 7-day food records the week prior to liver biopsy (or liver donation). The stool sample was collected at the end of this week and within 24 hours preceding the biopsy. Portion sizes were estimated using the 2D Food Portion Visual chart (Nutrition Consulting Enterprises, Framingham, MA). Food records were analyzed for macro- and micronutrient content using Diet Analysis Plus v. 7.0.1 (Thomson Wadsworth, Stamford, CT).

The participants also recorded their physical activities for 7 days during the week preceding the biopsy. They listed the type of activity, duration, and level of difficulty (mild, moderate, strenuous, very strenuous). Units of exercise were used to estimate physical activity as follows: 1 unit = 30 minutes mild, 20 minutes moderate, 10 minutes strenuous, or 5 minutes very strenuous activity.30

Basal metabolic rate (BMR) was calculated with the Harris-Benedict equation [men: BMR = 66.5 + (13.75 × weight in kg) + (5.003 × height in cm) − (6.755 × age in years); women BMR = 655.1 + (9.563 × weight in kg) + (1.850 × height in cm) − (4.676 × age in years)] and the estimated energy expenditure (EER) was calculated according to Health Canada Guidelines (http://www.hc-sc.gc.ca/fn-an/nutrition/reference/index-eng.php).

Biochemistry.

Fasting plasma glucose was measured by the enzymatic hexokinase method on an Architect c8000 System (Abbot Laboratories, Abbot Park, IL). Serum insulin was determined by radioimmunoassay (Immulite 2500, Siemens Diagnostics, Los Angeles, CA). IR was calculated using the Homeostasis Model Assessment (HOMA)-IR. Hemoglobin A1c in plasma was measured by ion exchange HPLC (Variant II analyzer, Bio-Rad Laboratories, Montreal, QC, Canada). ALT, aspartate aminotransferase (AST), and alkaline phosphatase (ALP) in plasma as well as serum triglycerides and total cholesterol were measured using the Architect c8000 system (Abbot Laboratories). Serum biochemistry was performed by the diagnostic testing laboratory at University Health Network, Toronto.

Histology.

A pathologist blinded to clinical characteristics assessed liver histology for presence of steatosis, inflammation, and fibrosis. The presence of NASH was determined using the Brunt scoring system, which is a validated and reproducible tool for the evaluation of NASH.31

Stool Samples and Quantification of IM.

The stool collection kit included a plastic collection/storage container with a tightly closing lid, an insulated bag, and cooling elements. Patients were asked to collect one sample within 24 hours of their next clinic appointment. The samples were immediately frozen in the patients' home freezer and transported to the hospital using the cooling elements and the insulated bag, similar to previously published methods.5 Stools were then stored at −80°C until analysis.

The stool was thawed, immediately homogenized with a masticator blender, and 0.1 g was used for DNA extraction using the E.Z.N.A. stool DNA Isolation Kit (Omega, Norcross, GA), as per the manufacturer's protocol. The extraction protocol was modified to include a lysozyme digestion step (incubation at 37°C for 30 minutes). DNA concentration and purity were measured using ThermoScientific Nanodrop 1000 Spectrophotometer (ThermoScientific, Rockford, IL). DNA samples were subsequently stored at −20°C. Fifty nanograms of the extracted DNA were used for the quantification of fecal bifidobacteria, Bacteroides/Prevotella, Clostridium leptum, C. coccoides, Escherichia coli, as well as total bacteria and Archaea, by quantitative polymerase chain reaction (qPCR) using a 7900HT thermocycler from Applied Biosystems (Foster City, CA) under default thermocycling conditions. Custom-made TaqMan primers for total bacteria,32 C. coccoides,32 C. leptum,32 Bacteroides/Prevotella,32, 33 bifidobacteria,32 and Archaea34 were used. Real-time PCR for E. coli was done using SYBR Green Gene Expression master mix (Applied Biosystems) and the specific forward and reverse primer.32 Number of cells of each microorganism in fecal samples was calculated by interpolation from standard curves and expressed as log cell counts/g feces. Bacteroides/Prevotella counts were considered representative of the Bacteroidetes phylum (as previously33) and will herein be referred to as Bacteroidetes.

Statistical Analyses.

Results are expressed as median (range) as the data were not normally distributed. Kruskal-Wallis test was used to compare the three groups for demographic, dietary, and laboratory data (Stata v. 12, College Station, TX). Nonparametric tests were used for statistical comparisons of the results of the fecal analyses as well (Kruskal-Wallis; Stata v. 12 and GraphPad Prism v. 4.0, GraphPad Software, La Jolla, CA). For the microbiota, high and low outliers were defined as numbers higher than the third quartile plus 1.5 times the interquartile range (IQR) and lower than the first quartile minus 1.5 times the IQR, respectively.35 Spearman correlation coefficients were used to assess bivariate relationships between variables. Analysis of covariance (ANCOVA) was used to investigate the associations between IM and liver disease state after controlling for variables that were found to be significant at the univariate level (these nonnormally distributed variables had been normalized using their logarithmic value, prior to performing the ANCOVA). Statistically significant α was considered any value lower than 0.05. Considering the paucity of literature assessing the IM of patients with NAFLD, determination of the sample size was based on studies in the field of obesity, where differences at a phylum level have been detected between the groups with as few as 14 subjects in total.9

Results

A total of 50 patients were enrolled in this study: 17 HC, 11 SS, and 22 NASH. The demographic and laboratory data are summarized in Table 1.

Table 1. Demographic and Laboratory Results
VariablesHC (n =17 )SS (n = 11)NASH (n = 22)
  • P = 0.003;

  • +, *

    P = 0.001;

  • **

    P < 0.001.

  • The values are expressed as medians (range). For each comparison, identical letters indicate the groups between which the statistical difference was significant.

  • Abbreviations: ALP, alkaline phosphatase; ALT, alanine transaminase; AST, aspartate transaminase; BMI, body mass index; HOMA-IR, Homeostasis Model of Assessment for Insulin Resistance.

Age (years)+36 (23-58)a,b48 (22-55)b47 (29-68)a
Gender (% male)593645
BMI (kg/m2)*26.0 (18.8-30.5)a28.6 (23.5-44.2)32.4 (24.2-49.5)a
ALT (U/L)**17 (6-41)a, b32 (14-54)b69 (22-168)a
AST (U/L)*22 (12-31)a21 (16-40)44 (18-114)a
ALP (U/L)65 (49-98)64 (42-105)77 (37-114)
Glucose (mmol/L)5.0 (4.1-6.5)5.1 (4.6-6.5)5.9 (4.1-7.6)
Insulin (pmol/L)21 (15-61)33 (15-465)97 (29-437)
HOMA-IR**0.60 (0.5-2.0)a2.15 (0.5-13.7)4.15 (1.0-34.6)a
Hemoglobin A1c0.054 (0.046-0.057)0.055 (0.050-0.072)0.061 (0.050-0.074)
Triglycerides (mmol/L)0.94 (0.58-3.32)1.10 (0.62-3.97)1.58 (0.28-3.29)
Total cholesterol (mmol/L)4.4 (2.78-7.46)5.05 (3.75-6.88)4.56 (2.65-6.94)

Demographics and Anthropometrics.

Patients with NASH and SS were older compared to HC. The gender distribution was not statistically different between groups. The majority of subjects in each group were Caucasians: 86% of the HC and 67% of both the SS and NASH patients. Patients with NASH had higher BMI when compared to HC (Table 1).

Laboratory Data.

Transaminases (ALT, AST) were higher in NASH compared to SS and HC. HOMA-IR was higher in patients with NASH compared to HC. No differences were found in ALP, glucose, hemoglobin A1c, cholesterol, or triglyceride levels. All patients had normal liver synthetic function as determined by albumin and International Normalized Ratio levels (data not shown).

Histology.

The median steatosis of the SS group was 12.5% (range: 5%-35%) and 40% in the NASH group (range: 5%-90%). Eighty percent of NASH patients had a variable degree of fibrosis (ranging from F1-F4). The median NAFLD activity score was 4 (range: 2-8).

Dietary Data.

The dietary data are summarized in Table 2. The total energy intake per day and the percentage of energy from carbohydrate and fat was not different among the groups. Adjusting the caloric intake for weight (total kcal/day divided by weight) revealed that HC were consuming more calories per kg compared to patients with NASH. The BMR was similar among subjects of all groups, as was the EER. The reported energy intake was lower than the EER in all three groups.

Table 2. Dietary Data and Calculations
VariablesHC (n = 17)SS (n = 11)NASH (n = 22)P Value
  1. The values are expressed as medians (range). For each comparison, identical letters indicate the groups between which the statistical difference was significant.

  2. Abbreviations: BMR, basal metabolic rate; EER, estimates energy requirement.

Energy intake (kcal/day)2094 (1103-2796)1715 (1248-2348)1617 (1025-2653)0.23
Fat-derived energy %37 (29-45)33 (22-39)34 (18-43)0.06
Carbohydrate-derived energy %49 (4-57)53 (39-74)47 (4-65)0.42
Weight-adjusted energy intake (kcal/kg/day)26.9 (13.8-64.1)a21.7 (15.3-27.9)17.4 (10.7-31.6)a<0.01
BMR (kcal/day)1541 (1223-2145)1655 (1252-2224)1635 (1329-2052)0.99
EER (kcal/day)2683 (1675-4115)2639 (2042-3681)2418 (1768-3899)0.62
Energy intake/BMR1.18 (0.6-2.3)1.15 (0.7-1.2)0.88 (0.6-1.6)0.40
% fat intake/BMR0.023 (0.016-0.035)a, b0.019 (0.012-0.470)b0.020 (0.013-0,031)a0.04

When dividing the percentage fat intake by BMR, to adjust for factors such as age, the HC group was found to consume more energy from fat compared to patients with SS and NASH.

IM.

Patients with NASH had higher fecal C. coccoides levels compared to those with SS, as depicted in Fig. 1. There were no differences between the groups for bifidobacteria, Bacteroidetes, C. leptum, E. coli, and total bacteria (P > 0.05). There were no differences in the Firmicutes-to-Bacteroidetes ratio between the groups (P > 0.05; Supporting Fig. S1). Archaea were only detectable in five HC, two SS, and two NASH, which limited the statistical power for any comparisons (Supporting Fig. S2).

Figure 1.

Comparisons for fecal C. coccoides between the groups. Patients with NASH have higher fecal C. coccoides compared to patients with SS. N = 17 HC, 11 SS, 19 NASH.

The relative abundance of each microbe of interest (ratio of bacterial number to total bacteria) was also compared between the groups. The percentage of Bacteroidetes was significantly lower in patients with NASH, compared to both SS and HC (Fig. 2). There were no differences between the groups in the percentages of the other microorganisms assessed.

Figure 2.

Comparisons for percentage Bacteroidetes (Bacteroidetes to total bacteria ratio) in stool between the groups. Patients with NASH have lower fecal percentage Bacteroidetes compared to both patients with SS and HC. N = 17 HC, 11 SS, 22 NASH.

Exploring for potential relationships between dietary intake and bacteria counts, we found no statistically significant correlations between total caloric intake, percentage fat or carbohydrate consumption, and fecal Bacteroidetes, C. leptum, C. coccoides, bifidobacteria, or E. coli (P > 0.05). Performing the same correlations for the NAFLD cohort only (SS and NASH combined), there was a statistically significant negative association between total daily caloric intake and fecal Bacteroidetes counts (Spearman r = −0.43, P = 0.038).

Taking into consideration that BMI and percentage of fat intake could be contributing to the association between the percentage of Bacteroidetes and NASH, ANCOVA was performed to control for these potential confounders. There was an independent association between the percentage of Bacteroidetes and the presence of NASH (P = 0.002; 95% confidence interval [CI] = −0.06 to −0.02). This was not the case with C. coccoides, which was no longer associated with NASH once BMI and percentage fat intake were taken into account (P > 0.05).

We also assessed whether the percentage of Bacteroidetes was associated with IR, controlling for BMI. There was a trend (r = −0.31; P = 0.06) towards a negative association between the percentage of Bacteroidetes and HOMA-IR.

Discussion

To our knowledge, this is the first study assessing the IM of adults with nonexperimental NAFLD and specifically comparing the IM composition of subjects classified as HC, SS, or NASH based on histological data. We found a lower relative abundance of Bacteroidetes in NASH, which was independent of BMI and energy intake from fat in the diet.

The importance of classifying patients based on liver histology is significant, as one of the most challenging aspects in the pathophysiology of NAFLD is understanding the differences between mechanisms causing simple hepatic steatosis versus those that lead to steatohepatitis. Since bacteria are known to play a pathogenetic role in the development of inflammation, comparisons between all groups (HC, SS, and NASH) allow for further elucidation of the effects of the IM on the liver.

Along with Firmicutes, Bacteroidetes comprise the majority of the human IM.36, 37 In our cohort, the relative abundance of Bacteroidetes in the stool was lower in NASH compared to both SS and HC. This finding is in agreement with previously published literature in the field of obesity that has demonstrated lower Bacteroidetes in patients with higher BMI.9, 37 The novelty of our study is the suggestion of a BMI-independent association between Bacteroidetes and liver disease state. Interestingly, our findings contrast those of Zhu et al.,29 who showed increased Bacteroidetes in children with obesity or NASH, compared to those with a normal BMI. These results may have been, in part, reflective of the imbalance in BMI between the groups in that study, as well as differences in environmental and dietary factors between the two study cohorts. In addition, the lack of classification of patients in each group based on liver histology may have also affected the results of the study by Zhu et al.29 Lastly, differences in age may have also played a role, as discussed below.

There are various theories to support an inverse correlation between Bacteroidetes and steatohepatitis. First, Bacteroidetes carry 45% of the lean metabolic potential in a study comparing the microbiome of lean and obese adults.21 A lower percentage of Bacteroidetes could have affected energy balance by facilitating metabolic dominance of other bacteria that are more efficient in extracting energy from the diet. Jumpertz et al.11 showed that a 20% increase in fecal Bacteroidetes is associated with a 150 kcal decrease in energy harvest from the diet. A second theory is that an initial hit causes the cell death of Bacteroides leading to LPS release from their cell wall and subsequent endotoxemia.19 The latter leads to the development of NASH.19, 27 It is not clear what would cause the death of these microorganisms. Changes in diet could play a role, as shown by studies in obese subjects, whose baseline lower fecal Bacteroidetes increase when placed on a hypocaloric diet or after bariatric surgery.38, 39

There is literature supporting an increase in intestinal permeability of subjects with IR, such as in obesity40 and diabetes.41 Recently, Zhu et al.29 reported higher serum ethanol levels in children with NASH, which was thought to be bacterially derived and, hence, potentially also contributing to increased intestinal permeability. In addition, there is scientific evidence linking endotoxemia with states of glucose intolerance, such as NAFLD.41, 42, 43 Animals and humans exposed to low levels of endotoxin develop IR.19, 22 Exploratory analysis from our study also suggested a potential link between the intestinal microbiota and IR by showing a trend toward a negative association between Bacteroidetes and IR when controlling for BMI. This requires further studies.

We did not find lower bifidobacteria counts or higher Firmicutes-to-Bacteroidetes ratio in NASH compared to SS and HC. This is in contrast to some of the previously published literature in the field of obesity9, 14, 44 and the recent study by Zhu et al. on children with NASH.29 The inability to show differences in these bacteria may have been due to the sample size; however, the size of our cohort was similar to that of other cross-sectional studies on IM in obesity9, 11, 12, 37 and NAFLD.29 Our results on Firmicutes-to-Bacteroidetes ratio are in line with other smaller projects, which also failed to replicate the findings of Ley et al.10-12 Compared to previous reports, our study provides stronger evidence supporting these results, as it did not simply assess the differences in the IM composition between the groups, but it also carefully addressed the potential confounding effects of diet and BMI.

Escherichia coli was not statistically different between the groups. Zhu et al. not only found E. coli to be higher in children with NASH compared to those who were obese without NASH, but also proposed that these bacteria may be contributing to the synthesis of ethanol with subsequent hepatotoxic effects.29 In our cohort there was a low overall abundance of E. coli in the stool, which may have contributed to the difficulty in detecting potential differences between the groups. Ours is the first study addressing the presence of Archaea in the stool of adults with NAFLD. These organisms were only found in a small proportion of study subjects overall, limiting the power of statistical comparisons. Further studies are required to elucidate the role of E. coli and Archaea in the development of NASH in both children and adults.

We assessed the intestinal microbiota by using qPCR, which is the gold-standard technique for bacterial enumeration.45 It is currently employed for the compositional analysis of the gut microbiota in humans and animals and was therefore ideal to quantify, in this study, fecal microbes that are known to play a role in obesity. Because qPCR does not allow for the identification of novel species,45 future studies could include metagenomic approaches, such as those based on 16S rRNA gene sequencing, potentially leading to the discovery of additional microbes associated with NAFLD. Moreover, a combination of these approaches with qPCR would provide an assessment of microbial diversity in healthy versus patients with NAFLD.

In our cohort, patients with NASH were older than HC. While the IM of infants and elderly patients appear to differ from that of adults, within the adult spectrum it is unlikely that there are significant, age-dependent variations in the IM composition.33 For that reason, age was not considered as a confounder and was not included in the ANCOVA. This factor, however, may in part explain the differences between the results of our study and those of Zhu et al.,29 who assessed the IM of children with NASH.

The median BMI of HC was at the lower spectrum of the overweight range (Table 1). This is unlikely to have influenced the results of this study, as all subjects had had a biopsy-proven unaffected (nonsteatotic, noninflamed) liver. In addition, the higher BMI in the control group allowed for smaller differences in BMI between the groups overall, theoretically limiting the potential confounding effect of this factor.

As dietary intake contributes to the fecal microbial composition, all subjects provided a 7-day food record. The reported caloric intake was not different between the groups, similar to the study by Zhu et al.29 In addition, there were no differences in calculated energy requirements, as expressed by BMR and EER. Considering the differences in BMI between the groups, dietary under-reporting may have influenced these results. A direct correlation between dietary underreporting and BMI has been previously shown in the literature.46

In summary, this is a novel study providing evidence for a link between percentage Bacteroidetes and the presence of NASH, which is independent of diet and BMI. Future research should address this topic, considering that the IM may serve as a potential therapeutic target in NASH, which is currently primarily managed by recommending weight loss and increased physical activity, which are notoriously difficult to sustain.

Acknowledgements

We thank Drs. David Wong, Gideon Hirschfield, Hemant Shah, Jordan Feld, and George Therapondos for assistance with patient recruitment, as well as Dr. Thomas Wolever, Kervan Rivera-Rufner, Wen Su, and Natasha Singh for support during the laboratory work.

Ancillary