Correlation of Gut Firmicutes/Bacteroidetes Ratio with Fibrosis and Steatosis in Patients with Non-alcoholic Fatty Liver Disease

Background : We investigated the gut microbiota in patients with non-alcoholic fatty liver disease (NAFLD) and its correlation with brosis and steatosis as reected in the controlled attenuation parameter and transient elastography values Methods : A cross-sectional study was performed on 37 patients with NAFLD at Cipto Mangunkusumo National General Hospital from December 2018 to March 2019. The gut microbiota was investigated in fecal samples with 16S RNA sequencing using the next-generation sequencing platform MiSeq (Illumina). Results : NAFLD was more common in patients with metabolic syndrome. Firmicutes, Bacteroidetes, and Proteobacteria were the predominant phyla. Bacteroides was more dominant than Prevotella, contrary to the results in previous studies on normal populations in Indonesia. Microbiota dysbiosis was observed in most samples. The gastrointestinal microbiota diversity was signicantly decreased in patients with NAFLD with high triglyceride levels and central obesity. The Firmicutes/Bacteroidetes ratio correlated with steatosis and obesity, whereas some other species in the lower taxonomy were mostly correlated with steatosis and obesity without brosis. Proteobacteria is the only phylum strongly correlated with brosis in patients with normal body mass index. Conclusions : The gut microbiota diversity was decreased in patients with NAFLD with high triglyceride levels and central obesity, and certain gut microbes were correlated with brosis and steatosis.

The relationship between the Firmicutes/Bacteroidetes ratio (as dysbiosis marker) and the severity of NAFLD, especially in Indonesia, is still unknown. The microbiota composition can be a roadblock in NAFLD management, especially intervention at the gut microbiota level. The goal of this study was to investigate the con guration of gut microbiota in patients with NAFLD and its correlation with brosis and steatosis condition as re ected in the controlled attenuation parameter (CAP) and transient elastography (TE) values. TE (Fibroscan ® Echosens) measures the velocity of the sound wave passing through the liver and then converts that measurement into a liver stiffness measurement; the entire process is often referred to as liver ultrasonographic (USG) elastography. The CAP speci cally targets liver steatosis using a process based on TE. It measures the degree of ultrasound attenuation by hepatic fat at the central frequency of the Fibroscan® M probe simultaneously with liver stiffness measurement.

Methods
This study used correlation tests to evaluate the relationship between two numerical variables. First, the gut microbiota were pro led in a cohort of patients with NAFLD. The correlation between the Firmicutes/Bacteroidetes ratio and CAP and TE values in patients with NAFLD were analyzed. Afterwards, the correlation of each microorganism with CAP and TE values were also analyzed to determine if certain microbes were correlated with the degree of brosis and steatosis in NAFLD. All the experiments were performed in accordance with relevant guidelines and regulations.

Subjects
The study samples were taken from patients at the Hepatobiliary Division Cipto Mangunkusumo Hospital Jakarta from December 2018 to March 2019. The sampling method used simple random sampling with a numeric correlation test formula, which determined that the number of samples needed was 37 subjects, including drop out possibilities. Patients with NAFLD between 18 and 60 years old, who were willing to participate in the study, were included, and informed consent from the patients was obtained. The exclusion criteria were pregnant or lactating patients, patients with other chronic liver diseases [hepatitis B, hepatitis C, autoimmune hepatitis, or with history of alcohol consumption (>40 g/day)], patients with a history of intestinal resection surgery, patients with chronic intestinal in ammation [in ammatory bowel disease (IBD)], patients with liver cell carcinoma, patients with a history of antibiotics and probiotics in the past month, or patients with a special diet in the past month as evidenced by a food recall questionnaire. The detailed ow chart of subject recruitment are shown in Figure 1.

Measurements
Patients lled out the informed consent form prior to examinations and underwent CAP and TE examinations using Fibroscan ® Echosens. The ultrasonographic probe was used to measure the level of brosis (in kPa) and steatosis (in dB/m).
Patients were also told to collect stool into the provided sterile tubes. The stool samples were initially stored at 2°C-8°C and then immediately taken to be stored at −80°C within 4 hours after sample collection. Afterwards, the bacterial genome was extracted at the laboratory of Child Health Department and sequenced by Biosains Medika (BioSM) Indonesia using the stool DNA kit Herculase II Fusion and the DNA Polymerase Nextera XT Index Kit V2 with 16S Metagenomic Sequencing Library Preparation Part #15044223 Rev. B protocols. The 16S rRNA sequencing used NGS platform MiSeq (Illumina) with four steps: sample preparation, library construction, sequencing and raw data collection.

Statistical Analysis
The data were analyzed using IBM SPSS Indonesia statistical program version 23. Numerical data normality was examined using a Shapiro-Wilk test. Data were not normally distributed if the p value < 0.05. The Spearman correlation test was used to evaluate the correlation between the Firmicutes/Bacteroidetes ratio and CAP and TE values. The correlation coe cient was interpreted with an r value and considered signi cant if p < 0.05 and with the following interpretation: 0-0.19 = very weak, 0.2-0.39 = weak, 0.4-0.59 = medium, 0.6-0.79 = strong and 0.8-1 = very strong correlation.
Upon correlation between variables, we continued to search for the intersecting value by creating a receiver operating curve. Results obtained as the area under the curve values can be interpreted as follows: 0.5-0.6 = very low accuracy, 0.6-0.7 = low accuracy, 0.7-0.8 = medium accuracy, 0.8-0.9 = high accuracy and 0.9-1.0 = very high accuracy. The microbiota diversity test between the group with signi cant brosis compared with non-signi cant brosis and signi cant steatosis compared with non-signi cant steatosis used the Mann-Whitney test. Dysbiosis was determined by an intersecting value between the quartile of the Firmicutes/Bacteroidetes ratio to brosis and steatosis and/or based on the intersecting value of the diversity value of brosis or steatosis group.

Patient Characteristics
The characteristics of the 37 subjects are presented in Table 1. Women dominated the group as 23 of them were female while only 14 were male with an average age of 50 ± 7.93 years old. The body mass index (BMI) assessment showed that 25 subjects were obese, seven were overweight, and ve had normal BMI. The average waist circumference was 96.65 cm, with 24 subjects experiencing dyslipidemia, and 30 subjects had type 2 diabetes mellitus. However, based on the HbA1c median value of 6.6 (4.8-14), we can see that the blood glucose was under control as also shown by the moderate value of fasting blood glucose and 2-hour postprandial blood glucose at 108 and 149 mg/dL, respectively. The lipid pro le, including triglyceride, low-density lipoprotein (LDL), and high-density lipoprotein (HDL) levels also showed moderate values, while the liver function test (aspartate aminotransferase, alanine aminotransferase, and serum albumin) also appeared normal.
We divided the subjects into three groups based on the BMI: normal (5 subjects), pre-obesity (7 subjects), and obesity (25 subjects). The characteristics of each group are presented in Table 2. In all three groups, most patients had central obesity (normal 4/5, pre-obesity 5/7 and obesity 25/25). Dyslipidemia was dominant in the pre-obesity and obesity groups (5/7 and 18/25, respectively), while 4/5 in the normal BMI group did not have dyslipidemia. Diabetes mellitus was dominant in all groups, even in the normal BMI group. The prevalence of diabetes mellitus in the pre-obesity and obesity group was 71.4% and 80%, respectively.

Characteristics of Intestinal Microbiota in Bacterial Mean Relative Abundance Based on Taxonomy
Intestinal microbiota were dominated by three major phyla: Firmicutes 55.35%; Bacteroidetes 29.94% and Proteobacteria 11.83% (Figures 2 and 3). The majority of subjects (20 of 37 subjects) had been dominated by Firmicutes, while 17 others were dominated by Bacteroidetes.
Microbiota Diversity Index Table 3 shows the Mann-Whitney test between alpha microbiota diversity index with other parameters, including brosis, steatosis, BMI, central obesity, triglyceride, HDL, LDL, and diabetes mellitus. The analysis presents the quantity of bacteria for each taxonomy level in its community. The diversity was determined by richness or amount (how numerous is one  Table 4 shows the Spearman correlation between the Firmicutes/Bacteroidetes ratio with each microbiota based on BMI. If divided into groups based on BMI, the medium positive correlation was between the Firmicutes/Bacteroidetes ratio with steatosis (r = 0.435; p = 0.030) and in the obesity group only. At the phylum level, the medium positive correlation (r = 0.528; p = 0.007) was also found between Firmicutes and steatosis. The only phylum strongly correlated with brosis in the normal BMI group was Proteobacteria (r = 0.921; p = 0.026). In the pre-obesity BMI group, there is a very strong positive correlation between steatosis with Lachnospiraceae (r = 0.883; p = 0.008) and Intestinimonas butyriciproducens (r = 0.847; p = 0.016).

Study Subject Characteristics
There were more women than men in this study, similar to a population-based study in Thailand by Summart (2017) [15].
Metabolic syndrome was dominant in the study subjects, which aligned with many other studies stating an apparent relationship among NAFLD, obesity, diabetes mellitus, and metabolic syndrome. In this study, 25 out of 37 subjects were obese, and 30 from 37 subjects had type 2 diabetes mellitus. Such characteristics were similar to other previous studies, which stated that there is a higher prevalence of NAFLD in adults with obesity (65.7%) and type 2 diabetes mellitus (74%) [16,17]. Individuals with NAFLD have a ve times higher risk of developing diabetes [18,19]. The association between NAFLD and type 2 diabetes mellitus can be explained through insulin resistance, dyslipidemia, and accumulation of liver triglyceride in NAFLD and β-cell defect in type 2 diabetes mellitus [20].
Although there were many previous studies on the microbiota, the results were inconsistent. A study by Raman et al. (2013) reported an increase in Firmicutes in obese patients with NAFLD compared with that of the patients without obesity and NALD [21]. Another study by Jiang et al. (2015) found no signi cant microbiota differences in NAFLD and normal control group [22]. Our study found that on average, Firmicutes was higher than Bacteroidetes. While at the genus level, Bacteroides (14.43%) were more numerous than Prevotella (9.14%), con rming that Bacteroides dominated other genera from the Firmicutes phylum.
Rahayu (2019) studied young Indonesian adult microbiota pro le [23] and reported that in numerous orders, the microbiota was Clostridium, Prevotella, Atopobium, Bi dobacterium, and Bacteroides. The results were quite similar to other local studies showing dominant Prevotella but different from our study. This may be due to the different study population as the subjects in this study lived in Jakarta and represented the urban population with high protein and animal fat in their diet [24].
We attempted to identify the prevalence of dysbiosis by looking at the diversity of microbiota and/or an increase in the ratio of Firmicutes/Bacteroidetes. Using dysbiosis criteria of Firmicutes/Bacteroidetes ratio, there were 26 out of 37 subjects with dysbiosis. There were 25 subjects ful lling the criteria if only based on the decrease of microbiota diversity. By combining the two criteria, we found dysbiosis in 18 subjects.

Microbiota Diversity Index in NAFLD
The microbiota diversity in our study was assessed using the alpha diversity index through OTUs, the Shannon index or the inverse Simpson index. The results showed signi cant differences in microbial diversity between the central obesity and non-central obesity group and between the high triglyceride and normal triglyceride group. Central obesity and high triglyceride groups showed a reduction in diversity compared with that of the other groups. This is similar with studies by Turnbaugh et al. (2009) [25] and Le Chatelier (2013) [26], which showed a total reduction in bacteria diversity in obesity. This is also in concordance with the dysbiosis theory commonly used in many studies as a marker for dysbiosis condition related to diseases.

Correlation of Firmicutes/Bacteroidetes with Fibrosis and Steatosis Based on Body Mass Index
We analyzed the correlation between the Firmicutes/Bacteroidetes ratio and microbiota in each taxonomy level with brosis and steatosis based on BMI. NAFLD analysis in each BMI group showed that the Firmicutes/Bacteroidetes ratio only had a positive correlation with steatosis in the obesity group. There was no signi cant correlation with brosis and also with steatosis in groups other than the obesity group. We also found that Firmicutes had a strong positive correlation with steatosis in the obesity group. This is similar to many previous studies that highlighted the role of Firmicutes in obesity [8,21]. Further analysis in lower taxonomy level in each microbiota revealed that only the group from phylum Proteobacteria was correlated with brosis in the obesity and normal BMI group. This is similar to a study by Loomba et al. (2017) [27]. The higher the brosis degree, the more numerous Proteobacteria and Bacteroidetes. Through this correlation study, we can see that Proteobacteria has a role in the process of liver brosis, although the exact mechanism is still unknown.
Most microbiotas had a positive correlation with steatosis, especially in obese patients. Some of which were from the order Clostridiales and Selemonodales in the Firmicutes phylum. While those correlated with steatosis in the normal BMI group were from the Actinobacteria phylum, the mechanism underlying steatosis by the microbiota were from several pathways especially related to fat metabolism [25]. Lactobacillus was very consistent in protecting steatosis, while the Enterobacteriaceae family in our study showed a very strong positive correlation with steatosis in the normal BMI group.
These ndings differed from those of previous studies [11]. However, in a study by Rahayu in healthy Indonesian population showed that the family of Enterobacteriaceae, especially Escherichia coli is part of the normal ora that increases in old age [23].
At present, there are very few studies that can show the direct cause and effect relationship between microbiota and NAFLD pathogenesis. However, some interventional studies in animals showed the important role of intestinal microbiota, especially in triggering a metabolic response. The intestinal microbiota from obese subjects can induce liver steatosis through modulation of fat metabolism. This is probably why most intestinal microbiota in our study correlated with steatosis but not with brosis. The process of turning steatosis into brosis needs more complex pathways and involving more factors aside from intestinal microbiota [28].
We acknowledge that the limitations of this study include the small sample size because of which we could not demonstrate that small variations in the bacterial counts were statistically signi cant. However, this cross-sectional single-center study was unable to view in detail the change in microbiota in relation to disease progression. We did not use normal healthy control because it was di cult to nd a population in urban settings that was absolutely healthy and free of metabolic disorders and was not affected by extreme diet.
In conclusion, we assumed that the bigger the difference between the subgroups studied, the stronger the potential effect of the bacteria on the phenotype. This is the rst study in Indonesia to thoroughly pro le the microbiota in patients with NAFLD using next-generation sequencing and tried to nd the correlation of each microbiota with brosis and steatosis. There was a strong positive correlation between the Firmicutes/Bacteroidetes ratio with steatosis in the obesity group. There were positive and negative correlations between some microbiota with brosis and steatosis. We suggest that future studies examine microbiota pro les in the general Indonesian population, the microbiota population in patients with NAFLD based on groups with other metabolic syndrome co-morbidities and the relationship between microbiota metabolism products and

Consent for Publication
Not applicable

Availability of data and materials
The data that support the ndings of this study are available on request with the corresponding author. The data are not publicly available because they contain information that could compromise research participants' privacy/consent.

Competing interests
The authors declare that they have no competing interests Funding This study received one third funding from the university (Medical Faculty Universitas Indonesia) and the rest from the private funding.

Author Contributions
COMJ proposed and conducted the study. AM, IH, MS, and IR performed the research and supervised the study. CRAL provided the idea for the rst draft of the manuscript. ASS, JK, KFK, SHN, and RAG collected and analyzed the data. All authors contributed to the design of the study, interpretation of the results, and in writing the nal manuscript. Chyntia Olivia Maurine Jasirwan is the guarantor of this study.   Characteristics of gut microbiota in bacterial mean relative abundance. a: Phylum; b: Class; c: Order; d: Family