Gut bacteria and butyrate may link estrogen loss to fatty liver disease before menopause

Female NAFLD patients and normal controls, aged between 35–45 years, attending Qilu Hospital of Shandong University between February 2019 and December 2019 were included in our experiment. Fresh feces were centralized collected in December 2019 and frozen immediately at -80°C. We conducted DNA extraction right after the samples collection. Informed consent was signed. We had access to information that could identify individual participants during or after data collection. The diagnostic criteria for NAFLD had to meet the following requirements: 1) No history of drinking, or an ethanol intake of less than 70g/week (female) for at least one year prior to the experiment; 2) Participants were without any of the following: viral hepatitis, drug-induced liver disease, total parenteral nutrition, hepatolenticular degeneration, autoimmune liver disease or any other specific diseases that can lead to fatty liver disease; 3) Liver imaging manifestations conformed to the diagnostic criteria for diffuse fatty liver disease. Patients who had used antibiotics or other special drugs in the past three months for systemic diseases such as diabetes and hypertension and those with unqualified fecal DNA were excluded from participation in the study.

Female C57BL/6 mice aged at 6–8 weeks were purchased from Jinan Pengyue animal company. All animals received humane care following the Guide for the Care and Use of Laboratory Animals approved by Shandong University. Throughout the study period, all animals had access to food and water ad libitum and were maintained on a 12 h light/dark cycle (21 ± 2°C with a relative humidity of 45 ± 5%). After one week of adaptation, animals were randomly divided into the following four groups: normal diet (C group), sham-operated+ high fat diet (HFD, 60 kcal% Fat, research diets, New Brunswick, Canada) (SH group), OVX+HFD (OH group), OVX+HFD+FMT (OHF group). Two weeks after OVX, the mice were given HFD. FMT was performed based on an established protocol [16]. Control animals received the same volume of saline solution. After HFD administration for 4 weeks, mice were anesthetized using 2% Isoflurane followed by cardiac puncture and the blood was collected. Whole livers were immediately prepared for histological analysis; some were stored at -80°C for RNA and western blot preparation, and blood samples were collected for biochemical analysis. Liver function was assayed by the serological activities of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) by using the Hitachi 7600–020 clinical analyzer (Hitachi, Tokyo, Japan).

The small intestines were incubated with 5 mM EDTA at 37°C for 40 min. Cells were collected by passing the supernatant through a 100-μm cell strainer (BD Falcon, Fremont, CA). After washing with PBS, IEC were separated from a 20/40% Percoll interface (Amersham Pharmacia Biotech, Amersham, UK).

Fresh feces were collected into individual sterile Ependorf tubes and then frozen immediately at -80°C until DNA extraction. The total bacterial community DNA was extracted using the TIANamp Stool DNA Kit (Tiangen Biotech Co., Ltd., Beijing, China) according to the manufacturer’s instructions.

After DNA extraction from the feces samples, we used polymerase chain reaction (PCR) amplification and pyrosequenced the V3 and V4 regions of the bacterial 16S ribosomal RNA gene (forward primer, 5-ACTCCTACGGGAGGCAGCAG-3; reverse primer, 5-GGACTACHVGGGTWTCTAAT-3). Barcoded V3-V4 amplicons were sequenced using the pair-end method by an Illumina MiSeq sequencing platform at the Shanghai Biotree Biotechnology Company (Shanghai, China). Sequences with an average Phred score lower than 30, containing ambiguous bases, with homopolymer runs exceeding 6 bp, having mismatches in primers, or having a length shorter than 100 bp were removed. Only sequences with an overlap longer than 10 bp and without any mismatches were assembled according to their overlap sequence. Reads that could not be assembled were discarded. Barcode and sequencing primers were trimmed from the assembled sequence [17]. All sequence files are available from the SRA NCBI database (https://www.ncbi.nlm.nih.gov/sra/PRJNA766510↗).

Liver and colon sections (5-μm) were prepared for paraffin-embedded cross-sectioning and stained with hematoxylin and eosin (H&E). Lesion images were captured with a microscope (Olympus, Tokyo, Japan).

The levels of estrogen, insulin and leptin in the serum were analyzed with Enzyme-Linked Immunoassay (ELISA) kits (Neobioscience Technology Co., Ltd., Shenzhen, China) according to the manufacturer’s instructions.

According to the reported method [18] with some modifications, the contents of SCFAs including acetic, propionic, n-butyric, i-butyric, n-valeric and i-valeric acids were determined with Agilent 6890 N gas chromatography (GC) equipped with a HP-INNOWAX column (30 m × 0.25 mm × 0.25 m; J & W Scientific, Agilent) and a flame ionization detector (FID). Briefly, the fecal content (50 mg) was mixed with 250 mL of deionized water; 150 mL of the slurry was then mixed with an equal volume of 0.3 mg/mL of 2-ethylbutyric acid (an internal standard). The mixture was centrifuged at 8000 rpm for 5 min, and the supernatant was filtered through a 0.45 m membrane for SCFA analysis. The operating conditions were set as follows: the initial column temperature was kept at 100°C for 1 min, then increased to 180°C at a rate of 5°C /min and kept at 180°C for 4 min. The flow rates for N2, H2, and air makeup gas were set at 30, 30 and 260 mL/min, respectively.

Total RNA was extracted with TRIzol (Life Technologies) and quantified for cDNA synthesis. Quantitative real-time PCR reactions were performed by using SYBR Green Gene Expression Assays (Bio-Rad, Hercules, CA, USA). Quantification was undertaken using the 2^-ΔΔCt method (Livak and Schmittgen,2001). All data were normalized to GAPDH mRNA expression.

Protein concentrations were determined with a bicinchoninic acid (BCA) assay kit (Pierce Biotechnology, Rockford, IL). Primary antibodies against Reg3ϒ, β-defensins 1(Affinity, Cincinnati, USA), Occludin-5, PPAR-ϒ, VLDLR and PPAR-α (all at 1:1000), and anti-β-actin (at 1:5000), all from Proteintech (Proteintech, Chicago, USA) were used in this study. The relative density of protein bands was quantitated with ImageJ version 1.5s (http://imagej.nih.gov/ij/↗) provided in the public domain by the National Institutes of Health, Bethesda, MD.

All values are presented as mean ± SD for each group. The data were analyzed by one-way ANOVA followed by a least significant difference (LSD) post hoc test; the level of significance was set at p < 0.05. SPSS 20.0 software (Chicago, IL, USA) and GraphPad Prism 6.0 software (La Jolla, CA) were used to perform the statistical analysis.

According to the lab design, females were divided into the NAFLD and control group depending on the ultrasound results for fatty liver. General characteristics of the enrolled people have been listed as Table 1. The results showed that there was no significant difference in age between the two groups. Also as shown in Fig 1A, the NAFLD group had a higher BMI and waist circumference, which are characteristics of obesity. However, blood glucose levels showed no statistical difference, while triglyceride, LDL, insulin, and HOMA-IR were higher in the NAFLD group, indicating the NAFLD group were insulin resistant. The average estrogen level of the NAFLD group was lower than that in the control group (74.7 ± 10.3 pg/mL vs 140.6 ± 15.7 pg/mL).

To investigate whether the gut microbiota composition differed between the two groups, we performed Illumina HiSeq 16S rRNA gene sequencing, which produced 8557224 clean reads from 45 samples. The extent of the OTUs (operational taxonomic units) shared between the two groups are summarized in the Venn diagram (Fig 1B). 617 OTUs were common, while 143 were unique to the control and 99 to the NAFLD group, revealing less OTU diversity in the NAFLD group. Rarefaction curves indicated that most gut microbial organisms in each sample were captured with the current sequencing depth (Fig 1C). As illustrated in Fig 1D, Bacteroidetes, Firmicutes, and Proteobacteria were the three dominant phyla. The abundance of Bacteroidetes and Proteobacteria exhibited significant differences between the two groups. The differences in the genera level are shown in Fig 1E. The abundance of Bacteroides, Alistipes, and Unassigned was much higher in the NAFLD group than control. LEfSe (LDA Effect Size) was used to explore significant changes and relative richness in the bacterial communities of the two groups, and five genera were identified with LDA scores (Fig 1F). Collectively, these result showed that estrogen affected the composition of the gut microbiota.

To investigate whether estrogen deficiency is a promoter of NAFLD, mice were administrated with OVX followed by an HFD to induce NAFLD. At the same time, FMT was administered to some of the OH mice to observe whether FMT would attenuate the NAFLD (OHF group). As shown in Fig 2A, following OVX, the OH group had lower levels of estrogen than the SH group, indicating that OVX successfully reduces estrogen in animal models. Fig 2B shows liver weight was increased after mice were treated with HFD, but liver weight in SH, OH and OHF groups showed no differences. Body weight was obviously increased in OH mice but was much lower in SH and OHF groups. Contrarily, liver/body weight ratio was decreased in OH mice. In addition, an increase in abdominal fat was found in the OH group, however, the OHF group had significantly less.

According to the HE staining, the OH group showed more hepatic steatosis than the SH group, in that it had more lipid droplets and inflammatory cell infiltration. But the OHF group showed greater mitigation in hepatic steatosis than the OH group (Fig 2C). The Oil Red O staining results showed that lipid droplets were further enhanced in OH mice compared with the SH mice, yet lipid accumulation within the liver tissues were alleviated in OHF mice (Fig 2D). These results indicated that estrogen reduction would exacerbate NAFLD, but that FMT could inhibit the NAFLD in the OH mice. In addition, the NAFLD condition was assessed by the serum biochemical index. Compared with the OH group, the SH and OHF groups both had lower total cholesterol (TC) and leptin than the OH group. Serum glucose, insulin and HOMA-IR level in OH group were higher than in the SH and OHF group (Fig 2E). Although the amounts of ALT, AST and triglyceride (TG) in the OH group were slightly higher than in the SH group, no statistical differences were found. Put all together, these results indicated that estrogen reduction worsened NAFLD, while FMT could significantly relieve the condition.

We then investigated whether OVX results in alterations in the composition of the gut microbiota. In total, 2,164,126 useable reads and 359 OTUs were obtained from 26 samples. The Shannon indexes were lower in the SH group compared with C groups (Fig 3A). Moreover, a Venn diagram of the three groups revealed that 312 OTUs overlapped among the groups: 322 OTUs were present in both the C and SH groups; 322 in both the C and OH groups; and 323 in both the SH and OH groups (Fig 3B). PCoA (principal co-ordinates analysis) showed that gut microbiota in the C group were different from those in the SH and OH groups (Fig 3C). The system clustering tree also indicated significant differences between each group, although the distance between the SH group and the OH group was small (Fig 3D).

We further investigated the gut microbiota species and their relative abundance. As shown in Fig 3E, at the phylum level, 10 phyla could be found in all samples and the most abundant phyla in all samples were: Firmicutes, Bacteroidetes, Proteobacteria, and Epsilonbacteraeota. Conversely, Verrucomicrobia were detected in the SH and OH groups but not in the C group. Furthermore, at a genus level classification, Faecalibaculum and Helicobacter were more common in the SH and OH groups but less abundant in the C group. Interestingly, results from Fig 3E show that the presence of Muribaculaceae and Lactobacillus in the SH and OH groups were significantly decreased relative to those in the C group. Furthermore, the presence of Firmicutes was significantly lower in the OH group compared with that in the SH group; whereas the relative abundance of Epsilonbacteraeota was remarkably higher. The relative abundance of Butyricicoccus, Eubacterium, and Roseburia were diminished in the OH group compared with the SH group (Fig 3F).

LefSe analysis was performed with the pooled data to identify specific taxa that could be used as biomarkers and dominant microbiota for each group. A cladogram for family and genus level abundance is shown in Fig 4A. In total, 41 genera were identified with LDA scores > 3.5 (Fig 4B). For correlation analysis, genera level correlations with SCFA and serological indicators such as Estrogen, TC, and leptin are shown in Fig 4C and 4D. Peptococcus and Romboutsia were positively, while Ruminiclostridiun-6 and Muribaculum were negatively correlated with SCFA. The occurrence of the genera exhibited a positive correlation with TC, leptin and body weight. Surprisingly, estrogen showed negative correlations with most of the genera detected. Put all together, our results showed that estrogen could modulate gut microbiota composition in NAFLD model mice.

To determine whether SCFA was changed in NAFLD patients with decreased estrogen, the concentration of their SCFA was examined. The butyrate content was much lower in the NAFLD patients than that in the controls (Fig 5A). However, the concentration of other SCFAs showed no difference between the two groups. Butyrate was also significantly decreased in the OH mice compared with the SH group (Fig 5B). Consistently, the other SCFAs showed no change. The content profile of SCFAs in OH mice was similar to that found in the NAFLD patients with decreased estrogen. These results indicate that the butyrate content is affected by estrogen in the NAFLD model.

To determine whether decreased butyrate causes the NAFLD according to estrogen reduction, the OH group was given butyrate (200mg/kg) for 4 weeks (named the OHB group). Results showed that the OHB group had no statistical difference in body and abdominal fat mass compared with the OH group, but compared with the OH mice, liver weight and liver/body weight ratio were clearly decreased in OHB mice (Fig 5C). HE staining also demonstrated that hepatocyte steatosis was reduced in the OHB group (Fig 5D). More importantly, the Oil Red O staining showed mice feeding with butyrate were apparently associated with attenuation of hepatic lipid deposition (Fig 5E), Furthermore, serum biochemical indicators also showed lower serum TC, insulin and HOMA-IR in OHB mice, but no statistical differences were shown between the groups for ALT and AST (Fig 5F).

We then investigated whether the altered microbiota and butyrate levels in estrogen reduced mice affect IEC function. AMP produced by IEC cells play a critical role in regulation of host and microbiota interaction. In order to study AMP production, IEC were collected to determine their AMP expression. Based on the results from qRT-PCR, the OH group had decreased expression of Reg3γ, β-defensins 1 and 3 than the SH group at mRNA levels (Fig 6A), and protein level of Reg3γ and β-defensins 1 were also decreased (Fig 6B and 6C). However, no difference was detected between the two groups for β-defensins 4. These results confirmed that AMP production was decreased in mice with NAFLD caused by estrogen reduction.

The mechanism by which hepatic steatosis induced by estrogen reduction was studied. Because the integrity of the intestinal epithelium plays a crucial role in maintaining a balance in fatty acid intake, damage to the intestinal epithelial barrier leads to an increase in fatty acid intake and potentially to fatty liver disease. The barrier function of the intestinal epithelium were investigated. The expression of the ZO-1 and Occludin-5 genes were lower in the OH group than in the SH group (Fig 7A), and expression of the protein ZO-1 and Occludin-5 had significantly declined in the OH group (Fig 7B). In addition, compared with the OH mice, ZO-1 and Occludin-5 were clearly increased in butyrate feeding mice but not changed in OHF mice (Fig 7C). Likewise, compared with the SH mice, HE staining showed that the OH and OHF group had reduced number of epithelium cells nucleus, with thickening of the intestinal wall. However, butyrate treatment mice seemed to block the reduction of intestinal epithelium cells (Fig 7D). The results confirmed that the intestinal epithelial barrier function was damaged in OH mice, but was relieved when mice were treated with butyrate.

As the central organ in the metabolic reactions, the liver plays an important role in maintaining metabolic balance. NAFLD is a disease expressed through multiple metabolic disorders and especially in FA metabolism. We hypothesize that mice with an estrogen reduction have a disorder of FA metabolism pathway. Therefore, genes and protein related to FA metabolism were tested in these mice. The expression of the lipid synthesis-related genes, SREBP1, PPAR-ɣ, FAS and CHREB were found to be significantly increased with estrogen reduction (Fig 8A). Moreover, in comparison with the SH group mice, the OH group had a significantly greater expression of the lipid intake related gene VLDLR (Fig 8B), while the lipid oxidation related genes, PPAR-ɑ and ACAA, had lower expression in the OH group (Fig 8C). The Western blot shown that the OH group mice had increases in PPAR-ɣ and VLDLR proteins but PPAR-ɑ expression had decreased (Fig 8D). In summary, estrogen was shown to be a crucial factor in FA metabolism regulation in the liver.