Bacteroides fragilis may worsen high-fat diet liver disease by changing fat metabolism and gut bacteria

After being given 1 week to adjust to their circumstances, the experimental mice were subjected to an HFD for 16 weeks in order to induce liver steatosis. To investigate the effects of B. fragilis on NAFLD, we assigned four groups of mice: the C group received a chow diet with phosphate buffer solution (PBS), the CB group received a chow diet with B. fragilis, the H group received an HFD with PBS, and the HB group received an HFD with B. fragilis for the last 8 weeks (Fig. 1A). Mice fed an HFD developed obesity-related symptoms, as evidenced by an increase in body weight in comparison to mice on a control diet. Moreover, B. fragilis further exacerbated HFD-induced obesity (Fig. 1B). The weight growth in the H group began to significantly exceed that of the C group starting from the 5th week (P < 0.05). The weight growth in the HB group was obviously higher compared with that observed in the H group, starting from the 12th week (P < 0.05). At 17 weeks, in comparison to the C group, the H group demonstrated significant increases in liver weight (1.346 vs 0.9047), liver index (4.129 vs 3.259), epididymal fat (0.8837 vs 0.4088), perirenal fat (0.4462 vs 0.2676), and kidney (0.4010 vs 0.3327) (P < 0.01; Fig. 1C through F; Fig. S1A). Meanwhile, supplementation of B. fragilis further exacerbated the weight of these tissues in HFD-fed mice (P < 0.05).

The liver, being the principal organ responsible for glucolipid metabolism, exhibits impaired functionality as a significant abnormality. The mice in the H group had a more severe lipid metabolic problem than the C group, as shown by significantly higher serum total cholesterol (TC) (3.520 vs 2.012), triglyceride (TG) (0.3089 vs 0.2376), and low-density lipoprotein cholesterol (LDL-C) levels (1.985 vs 1.075) (P < 0.05; Fig. 2A through C). In comparison to the H group, the HB group showed elevated levels of serum TC (4.399 vs 3.520), TG (0.3749 vs 0.3089), and LDL-C (3.009 vs 1.985) levels in mice (P < 0.01). Although high-density lipoprotein cholesterol (HDL-C) levels in the H group were significantly lower than those in the C group, there was no statistically significant difference (Fig. S1B). Mice fed with an HFD exhibited exacerbated liver dysfunction, as evidenced by a significant increase in serum levels of aspartate aminotransferase (AST) (47.90 vs 22.61) and alanine aminotransferase (ALT) (60.30 vs 30.85), compared with the C group (P < 0.05; Fig. 2D and E). Supplementation with B. fragilis significantly increased ALT levels (103.9 vs 60.30; P = 0.0014), while having little influence on AST levels (57.09 vs 47.90; P = 0.3469), compared with the H group. With regard to glucose metabolism, in comparison to the C group, HFD-fed mice had a higher fasting blood glucose (FBG) level (6.900 vs 4.933) and that level was drastically higher in the HB group than that in the H group (8.700 vs 6.900) (P < 0.05, Fig. 2F).

The levels of TC, TG, LDL-C, ALT, and AST in the liver of mice given an HFD were significantly higher than those in the C group (P < 0.05; Fig. S2A through E). Surprisingly, compared with the H group, supplementation of B. fragilis dramatically raised the ALT level in the liver (41.58 vs 35.49; P = 0.0313; Fig. S2D). Although the levels of TG, TC, LDL, and AST in the liver of the HB group were higher than those of the H group, the differences were not statistically significant (Fig. S2A through C and E). Furthermore, the H group had higher levels of lipopolysaccharide (LPS) in the serum (17.89 vs 7.458) and liver (9.531 vs 4.473) than the C group (P < 0.001; Fig. S1C and S2F). B. fragilis significantly increased serum (22.31 vs 17.89) and liver (14.16 vs 9.531) LPS levels in comparison to the H group (P < 0.001; Fig. S1C and S2F). In general, the findings showed that B. fragilis exacerbated liver dysfunction, including the accumulation of lipids in the liver and disturbances in glucolipid metabolism.

We subsequently conducted a histopathological examination of the liver. In hematoxylin-eosin (HE) staining, B. fragilis showed a significant increase in the occurrence of lipid droplets and hepatocyte ballooning degeneration as compared with the H group (Fig. 3A). According to the NAFLD activity score (NAS) histological system, the HB group demonstrated a significantly higher NAS compared with the H group (P < 0.05; Fig. 3C). Oil red O (ORO) staining demonstrated a higher presence of red-stained lipids in the H group, which was enhanced further in the HB group, as well as an increase in oil red-positive regions (Fig. 3B and D). HFD increased total TG and TC concentrations in the liver, as well as lipogenesis-related gene expression, including stearoyl coenzyme decarboxylase 1 (SCD1), sterol regulatory element binding protein 1 (SREBP1), fatty acid synthase (FAS), acetyl-CoA carboxylase 1 (ACC1), peroxisome proliferator-activated receptor γ (PPARγ), and fatty acid translocase (CD36), which were exacerbated by B. fragilis treatment (P < 0.05; Fig. 3F through J). Compared with the C group, the H group had no significant difference in sterol regulatory element binding protein 1 (SREBP1) levels, whereas an increase was observed in the HB group (P < 0.05; Fig. 3E). HFD downregulated fatty acid β-oxidation genes including peroxisome proliferator-activated receptor α (PPARα), carnitine palmitoyl transferase-1 (CPT-1), and peroxisome proliferator-activated receptor 1α (PGC-1α), and fatty acid decomposition genes including hormone-sensitive lipase (HSL) and adipose triglyceride lipase (ATGL) (P < 0.05; Fig. 3K through O). Compared with the H group, CPT-1, HSL, and ATGL were reduced by B. fragilis treatment (P < 0.05; Fig. 3L, N and O).

Gut microbiota might have a role in the development of NAFLD. We investigated the impacts of commensal B. fragilis on gut microbiota by applying high-throughput sequencing of the 16S rRNA gene. All rarefaction curves approached a plateau, suggesting that the sequencing depth was adequate for subsequent analysis (Fig. S3A). Alpha diversity including the ACE index (P = 0.045), Chao1 index (P = 0.037), Shannon index (P = 0.0091), and Simpson index (P = 0.011) in the H group were substantially less than that in the C group (Fig. 4A). These findings demonstrated that the richness and diversity of the gut microbiota were substantially different from the C group to the H group. With the B. fragilis supplement, the Shannon index was significantly decreased (P = 0.046), while the ACE index (P = 0.39), Chao1 index (P = 0.51), and Simpson index (P = 0.075) were decreased with no significant difference, compared with the H to the HB groups (Fig. 4A). Beta diversity was used to measure the variability of the intestinal microbiota structure, and samples were characterized using principal component analysis (PCA) and non-metric multidimensional scaling (NMDS) to explore differences in the intestinal microbiota community composition. In the PCA and NMDS analyses, the samples showed distinct clustering patterns among the four groups (Fig. 4B and C). An HFD strongly influenced the community composition of the gut microbiota. Meanwhile, B. fragilis treatment showed an extensive effect on the diversity of bacteria.

In order to assess the compositional changes in the gut microbiota among the four groups, we carried out a systematic analysis at different levels of structures. At the phylum level, Bacteroidota, Firmicutes, and Desulfobacterota were identified as the top three most abundant microbial taxa (Fig S3B). The results showed that the relative abundances of Bacteroidota (62.68%) in the C group proved to be higher compared with those in the H group as well as the HB group (31.51% and 28.79%, respectively), while the C group exhibited a lower relative abundance of Firmicutes (32.46%) and Desulfobacterota (1.54%) compared with the H group (41.40% and 16.14%), respectively. Significantly, B. fragilis increased the relative abundance of Desulfobacterota (20.01%) (Fig S3B). These indicated that HFD and B. fragilis could change the gut microbiota community structure in mice. At the family level, Muribaculaceae, Lachnospiraceae, and Desulfovibrionaceae were the top three most abundant microorganisms. Compared with the H group (20.00% and 20.40%, respectively) and HB group (17.35% and 17.10%, respectively), the relative abundances of Muribaculaceae (51.89%) and Lachnospiraceae (24.40%) were higher in the C group. Similar to Desulfobacterota, Desulfovibrionaceae had a lower relative abundance in the C group (1.54%) in comparison with the H group (16.14%) and the HB group (20.01%) (Fig. S3C). At the species level, compared with the C group, the relative abundances of Bacteroides_acidifaciens (13.45%), Paraprevotella_clara (2.56%), Muribacter_muris (0.05%), Lactobacillus_gasseri (0.02%), Dialister_sp (0.05%), Alistipes_timonensis (0.03%), and Clostridium_perfringens (0.04%) were increased substantially in the CB group, while the relative abundances of Firmicutes_bacterium (3.58%), Alistipes_sp (0.28%), Rikenella_microfusus (0.20%), Lactobacillus_johnsonii (0.62%), Dorea_sp (0.27%), and Metamycoplasma_sualvi (0.02%) were significantly decreased in the CB group. The H group mice were noted at the species level, including higher Anaerofustis_stercorihominis (0.02%), Adlercreutzia_caecicola (0.02%), Alistipes_indistinctus (0.02%), Holdemania_massiliensis (0.67%), Mucispirillum_sp (1.11%), Alistipes_finegoldii (4.68%), and Clostridioides_difficile (2.01%) correlated to the C group. B. fragilis significantly increased the relative abundance of Desulfovibrio_fairfieldensis (0.28%), Adlercreutzia_mucosicola (0.04%), Agathobacter_ruminis (0.04%), Bacteroides_fragilis (4.81%), Bacteroides_vulgatus (1.68%), Butyricimonas_synergistica (0.64%), Enterococcus_faecalis (0.11%), Flavonifractor_plautii (0.13%), Helicobacter_sp (1.21%), Holdemania_filiformis (0.21%), Morganella_morganii (0.07%), Mucispirillum_schaedleri (0.56%), and Parabacteroides_sp (0.27%) in HFD mice (Fig. 4D).

To identify the distinct bacterial groups within various categories, we used the linear discriminant analysis effect size (LEfSe) technique (Fig S3D). Surprisingly, the findings revealed a distinct difference in gut microbiota in the four groups. The H group showed a higher abundance of Firmicutes within the phylum level, Clostridia within the class level, and Oscillospirales within the order level compared with the C group, whereas the CB group showed an increased prevalence of Bacteroidales within the order level. Similarly, B. fragilis changed gut microbiota, especially enriched in Desulfovibrionales in HFD mice (Fig. S3D). All of these results indicated that commensal B. fragilis changed gut microbiota.

To examine the variations in SCFA levels associated with gut microbiota in mice, we used gas chromatography-mass spectrometry (GC-MS) to analyze the levels of seven SCFAs found in mouse fecal samples: acetic acid (AA), propionic acid (PA), butyric acid (BA), isobutyric acid (IBA), valeric acid (VA), isovaleric acid (IVA), and caproic acid (CA). We observed comparable overall levels of SCFAs in the stool samples from all four groups, predominantly comprising AA, PA, and BA. AA was the most multiple of the three SCFAs, next to PA and BA. Following that, the SCFAs present in the stool samples were subjected to analysis. IBA and IVA levels were considerably higher in the H group compared with the C group, but AA and CA levels were significantly lower. The commensal B. fragilis increased BA and VA levels compared with the C group. Although the levels of AA, PA, BA, IBA, IVA, and CA were increased with commensal B. fragilis treatment compared with the H group, significant differences were not found (Fig. 5A through G). These results suggested that an HFD has a significant regulatory effect on SCFAs, which were metabolites of gut microbiota. B. fragilis had a regulatory effect on some SCFAs under a normal diet but had little effect on SCFAs under an HFD.

To investigate the potential association between alterations in gut microbiota and SCFAs, Pearson correlation analysis was performed to assess the correlation between different bacteria and SCFAs (Fig. 5H). Microbial abundance and SCFA content of correlation analysis showed that Bifidobacteriaceae and Muribaculaceae had significantly negative correlations with AA. Bifidobacteriaceae and Muribaculaceae had clear positive correlations with PA, while Helicobacteraceae, Erysipelotrichaceae, Oscillospiraceae, Ruminococcaceae, and Desulfovibrionaceae had clear negative correlations with PA. Interestingly, there was no obvious correlation between BA and gut microbiota in this study. Lactobacillaceae had clear positive correlations with CA.

The correlation between serum metabolism and gut microbiota was further analyzed (Fig. S4A). The Pearson correlation analysis showed that Desulfovibrionaceae had clear positive correlations with TC, LDL-C, AST, and LPS levels in the serum. Erysipelotrichaceae had clear positive correlations with LDL-C levels in the serum. In addition, Prevotellaceae had an extremely negative correlation with serum levels of TG, TC, LDL-C, AST, and LPS. Muribaculaceae had an extremely negative correlation with serum levels of TC, LDL-C, and LPS. Bifidobacteriaceae had an extremely negative correlation with serum levels of LDL-C, AST, and LPS.

We then examined the correlation between gut microbiota and liver metabolism (Fig. S4B). The Pearson correlation analysis showed that Bacteroidaceae had clear positive correlations with TC, LDL-C, AST, and LPS levels in the liver. Desulfovibrionaceae had clear positive correlations with the liver level of TG. Prevotellaceae had clear negative correlations with TG TC, ALT, AST, and LPS levels in the liver. Muribaculaceae had clear negative correlations with the levels of TC in the liver. Among the genes involved in liver lipid metabolism, Pearson’s correlation analysis showed that ATGL had clear positive correlations with Bifidobacteriaceae, Prevotellaceae, and Muribaculaceae, while having clear negative correlations with Erysipelotrichaceae and Desulfovibrionaceae. HSL had clear negative correlations with Erysipelotrichaceae. ACC1 had clear positive correlations with Desulfovibrionaceae. CD36 had a clear positive correlations with Desulfovibrionaceae, while had a clear negative correlation with Prevotellaceae.

B. fragilis (ATCC 25285) was derived from the China Center for Type Culture Collection. B. fragilis was cultured on Columbia blood agar at 37°C with anaerobic conditions for 24–48 hours. Subsequently, B. fragilis was cultivated in a brain heart infusion medium at 37°C with anaerobic circumstances for 24–48 hours. All mediums were purchased from Qingdao Hope Bio-Technology Company (Qingdao, China). B. fragilis was harvested and resuspended in PBS for measurement of optical density (OD) at 600 nm, as described previously (15).

Six-week-old C57BL/6J inbred specific pathogen-free (SPF) male mice were acquired from Hunan Slake Jingda Experimental Animal Co. Ltd. (Changsha, China). Mice were housed in a regulated environment, providing unrestricted availability of food and water, under a 12-hour daylight cycle at 25°C ± 2°C. Mice were assigned at random into four groups after a week of adaptation (C, CB, H, and HB group, n = 8). The mice in the C and CB groups were provided with a standard chow diet (Jiangsu Xietong), while the mice in the H and HB groups were provided with an HFD (Jiangsu Xietong) for a duration of 16 weeks. Mice in CB and HB groups were administered intragastrically with B. fragilis at an oral dosage of 2 × 10⁸ CFU per 200 µL every alternate day for the last 8 weeks (51). Mice in the C and H groups were administered with the equivalent amount of PBS. All animals have been cared for in compliance with the guidelines of the Institutional Animal Care and Use Committee at Huazhong University of Science and Technology (IACUC Number: 3326).

The total weight of the mice was measured weekly during the course of the experiment. In the last week before euthanasia, the FBG levels were assessed following an 8-hour period of fasting. The mice were subjected to tail vein blood collection. An adaptable blood glucose meter and blood glucose test sheets were utilized to measure blood glucose levels. The excrement of mice was freshly gathered and preserved in a refrigerator at −80°C. The mice were starved overnight, weighed, and euthanized under anesthesia in accordance with guidelines for experimental animals. Blood was taken and allowed to coagulate for 30 minutes in microcentrifuge tubes. The upper serum was centrifuged next, and then, it was kept for further examination at −80°C. Weigh liver, kidneys, epididymal fat, and perirenal fat weight after euthanasia. The tissue samples were washed with pre-cooled saline solution to eliminate any residual blood or impurities adhering to the surface of the tissue and preserved in a refrigerator at −80°C.

The tissue blocks were weighed, cut into pieces, and homogenized by ultrasound in a proportion of 1:9 (tissue weight to PBS volume). The homogenate was transferred to a centrifuge tube and centrifuged for 10 minutes at 4°C at 1,000 g. To prevent recurrent cycles of freezing and thawing, the supernatant was gathered and stored at −80°C. Assay reagents (Nanjing Jiancheng Bioengineering Institute, China) were used to detect serum and liver levels of TG, TC, AST, ALT, LDL-C, and HDL-C in accordance with the manufacturer’s directions (52).

The RNA-Easy Isolation Reagent was used for extracting total RNA (Vazyme, China). The cDNA was then reverse transcribed using HiScript II QRT SuperMix for qPCR (Vazyme, China), according to the manufacturer’s directions (53). The mRNA levels of SREBP1, FAS, SCD1, ACC1, PPARγ, ACC1, CD36, PPARα, CPT-1, PGC1α, HSL, and ATGL were measured. The primers were presented in Table 1.

The hepatic tissues were fixed in a neutral fixative containing 4% paraformaldehyde for a minimum of 24 hours. After undergoing dehydration and transparency processes, the tissue was embedded in wax to form a block. Thin sections measuring 4 μm were then dewaxed, hydrated, stained, dehydrated again, and covered with a coverslip. An optical microscope (Olympus) was utilized for the observation and photography of the sections. The liver histological score was determined by using the NASs. The scoring system contains steatosis (0–3), lobular inflammation (0–2), hepatocellular ballooning (0–2), and fibrosis (0–4). Liver fibrosis was not assessed during the limited duration of the modeling period, as described previously (54).

To identify liver lipid droplet accumulation, an ORO kit (Servicebio, Wuhan, China) was used. With a frozen slicer, liver tissue sections were made at a thickness of 10 µm and dyed according to the directions of the ORO staining kit. Lipid droplets emerged as red-stained structures, while nuclei were dark-blue stained. ImageJ software was used to quantify the lipid droplets.

LPS concentrations among liver tissues and serum were measured using an enzyme-linked immunosorbent test (Bioswamp, China), as directed by the manufacturer.

The genomic DNA of bacteria was isolated from frozen stool samples by using the CTAB/SDS technique, following the manufacturer’s guidelines (55). The Qubit assay was employed for quantifying DNA concentration, while using 1% agarose gel electrophoresis, the integrity of isolated genomic DNA was assessed. Primer set 341F-806R was used to amplify the 16S rRNA V3/V4 region. 341F (5′-CCTAYGGGRBGCASCAG-3′) was the forward primer and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) was the reverse primer. The amplified PCR products were mixed to build the library, and the sequencing adapter was added. After PCR amplification, the small fragments such as primer dimer were removed by magnetic bead sorting, and the library index information was recorded. Samples exhibiting a unique amplification outcome were chosen for subsequent investigation. After preparing DNA libraries, sequencing was conducted using an Illumina NovaSeq platform, following the standard protocols provided by Baiaoweifan Biotechnology Co. Ltd. (Wuhan, China). Amplicon sequencing data, primer removal, quality control, ASV inference, double-end splicing, and ASV feature table generation for further analysis are the data analysis process. For quality control of the original sequencing sequence, use the FLASH software. Select a window size of 50 bps, and filter readings with a tail mass value less than 20. Based on the overlap connection between PE reads, the paired reads are combined into a sequence with a minimum overlap length of 10 bp. There is no match between the screening and the sequence, and the maximum permitted mismatch ratio in the concatenated sequence’s overlap area is 0.2. Utilizing the barcode and the primers at the start and finish of the sequence, distinguish the sample; if required, change the direction of the sequence. A maximum of two primer mismatches are allowed, while zero mismatches are allowed for the barcode. Using default parameters, the DADA2 method in the Qiime2 process is employed to denoise and optimize the sequence after quality control stitching. The sequences obtained after DADA2 denoising are commonly referred to as ASVs. The number of denoised sequences for each sample was normalized to 4,000 in order to reduce the effect of sequencing depth on the outcomes of following analyses, such as alpha diversity and beta diversity. Even after flattening each sample, the average coverage of Good’s may still be 97.90%. Based on the Sliva 16S rRNA database, a Naive Bayes classifier from Qiime2 was used to conduct the taxonomic analysis of ASVs.

GC-MS was used to measure the amounts of SCFAs in stool samples. As in the prior investigation, samples were pretreated (56). The supernatant was immediately subjected to GC-MS analysis after being passed through a 0.22-µm syringe filter. The conditions for GC analysis were as follows: the capillary column size was 30 mm × 0.25 mm × 0.25 µm. The carrier gas, helium, had a flow rate of 1.2 mL/min. The temperature of the front injection was 200°C. The oven temperature ramp was as follows: 90°C for 1 min, then 100°C at a rate of 25°C/min, then 150°C at a rate of 20°C/min, held for 0.6 min, and 200°C at a rate of 25°C/min, maintained for 0.5 min, after operating for 3 min. The transfer line temperature was 230°C, the ion source temperature was 230°C, and the quad temperature was 150°C. An electron bombardment ionization (EI) source with full scan and sim scanning modes at an electron energy of 70 eV was used in the mass spectrometry analysis. An Agilent chemical workstation was used to process all of the data. Based on the standard, the MWDB (Metware Database) was created to qualitatively interpret mass spectrometry data.

GraphPad Prism 8.0 and SPSS 22.0 statistical software were used for data analysis, with assessment of normality (Kolmogorov-Smirnov test) and variance homogeneity (Fisher’s F-test). One-way or two-way ANOVA was employed to compare data across multiple groups, followed by Tukey’s multiple comparisons test or LSD test. Mean ± SD values were reported. Statistical tests included Kruskal-Wallis Pairwise test for alpha diversity analysis, PERMANOVA for beta diversity analyses, and non-parametric Pearson test to examine the correlation between gut microbiota and host metabolism.

The raw sequencing data were submitted to the National Center for Biotechnology Information (NCBI) database and given the accession number PRJNA1013139↗.