Measurement of serum lipids and glucose
The serum levels of lipids and glucose at 8, 14, and 17 weeks of different treatment are displayed in Fig. 1. Glucose, LDL-C, TC and TG in the HFD group are higher than those in the C group at 8, 14, and 17 weeks, which indicates that the hyperlipidemia hamster model was successfully established. Moreover, fenofibrate-treated could significantly reduce the serum level of lipid in hyperlipidemia hamsters.
Fenofibrate alters the gut microbiota composition
We obtained an average of 36 million paired-end reads for each sample (from 30 to 52 million; Supplementary Table 1). MetaGeneMark software was used to distinguish coding regions in the genome. Cd-hit software was used to edit the nonredundant genome, the similarity threshold was set to 95%, and the coverage threshold was set to 90%. Fenofibrate treatment altered the gut microbiota composition of high-fat diet-treated hamsters, from phylum to species. The three groups were clearly separated in the PCoA (Figures 2a-b). At the genus level, Clostridium, Oscillibacter, Bacteroidetes, Ruminococcus were the predominant microbial divisions (Figure 2c). At the genus level, Firmicutes, Bacteroidetes, Proteobacteria and Actinobacteria were the predominant microbial divisions (Figure 2d). Compared with C, HFD decreased the relative abundance of Prevotella. Compared with HFD, F increased the relative abundance of Bacteroides, Parabacteroides, Alistipes and Mycoplasma genera. The difference in the abundance of taxa within groups was also identified using the linear discriminant analysis (LDA) effect size (LEfSe) method using non-parametric factorial Kruskal-Wallis and Wilcoxon rank-sum tests. The histogram of the distribution of LDA values (>2) shows the typical microbiota of the groups (Supplementary Figures 1a-b). The F group showed selective phyla enrichment of Tenericutes, Thermodesulfobacteria, et al., and showed decreased phyla of Planctomycetes and Basidomycota (Supplementary Figures 1c).
The results showed that 14450 bacterial species were annotated. Compared with the C group, the HFD group showed a significant increase (P < 0.05) in the levels of 853 bacteria species including Bacteroides acidifaciens, Faecalibaculum rodentium, and Listeria monocytogenes, whereas levels of 380 bacterial species were decreased including Bacteroides ovatus, Prevotella veroralis, and Prevotella saccharolytica (Supplementary Table 2). However, compared with the HFD group, the F group showed a significantly (P < 0.05) higher relative abundance of 695 bacterial species (P < 0.05) including Bacteroides ovatus, Bifidobacterium animalis and Bacteroides intestinalis et al. Furthermore, there was a significant reduction in the abundance of 276 bacteria species (P < 0.05), such as Lactobacillus johnsonii, Lachnoclostridium sp. An14 et al (Figure 2f and Supplementary Table 3). Random forest regression showed that Firmicutes bacterium CAG:582, Flavobacteriales bacterium, Bifidobacterium animalis et al were the main biomarkers for HFD, and F (Figure 2e). And the relative abundance of these species was showed as figure 2f.
Fenofibrate changed metabolic pathway of gut microbiota
Principal coordinate analysis (PCoA) based on KEGG modules revealed differences in microbial functions within groups (Figure 3a-b). The number of carbohydrate metabolism-related genes was highest at KEGG level 1. Based on the KEGG level 2 analysis, the HFD group demonstrated a higher level of carbohydrate metabolism and signal transduction, but lower metabolism of cofactors and vitamins than the C group. However, the F group showed higher levels of replication and repair than the HFD group (Supplementary Figures 2). At KEGG level 3, The one-way ANOVA showed that 44 pathways were significantly different within groups (Figure 3d). Compared with the C group, the HFD group showed significantly lower levels (P < 0.05) in the activity of 24 metabolic pathways, including glycosphingolipid biosynthesis-ganglio series, biosynthesis of siderophore group nonribosomal peptides, atrazine degradation, limonene and pinene degradation, naphthalene degradation. In contrast, 12 metabolic pathways showed higher activity in the HFD group than in the C group, including valine, leucine and isoleucine degradation, taurine and hypotaurine metabolism, beta-alanine metabolism, lipopolysaccharide biosynthesis, riboflavin metabolism. However, compared with the HFD group, the F group showed a significantly higher level (P < 0.05) of activity in five pathways, including biosynthesis of unsaturated fatty acids, base excision repair, glycerophospholipid metabolism, pyrimidine metabolism, mismatch repair, decreased 5 kinds pathways including glyoxylate and dicarboxylate metabolism, tyrosine metabolism, tryptophan metabolism, nonribosomal peptide structures, biosynthesis of siderophore group nonribosomal peptides (Figure 3d). The LDA score analysis determined that the most effective pathways for C, HFD, and F were peptidoglycan biosynthesis, lipopolysaccharide biosynthesis, and riboflavin metabolism and pantothenate and CoA biosynthesis, respectively (Figure 3c). The LDA biomarker pathways correlation analysis showed that lipopolysaccharide biosynthesis was positively correlated to pantothenate and CoA biosynthesis but negatively related to peptidoglycan biosynthesis and biosynthesis of antibiotics (Figure 3e, Supplementary Table 4).
Fenofibrate changed gene expression profiles of gut microbiota
Based on the KEGG annotation analysis, we identified 6,572 genes from 18 samples, 3 groups shared 6135 genes, and the one-way ANOVA showed that 1347 genes were significantly different within groups. The LDA showed that k06147 (ABCB-BAC, ATP-binding cassette), k07667 (kdpE, two-component system, KDP operon response regulator KdpE), k03088 (rpoE, RNA polymerase sigma-70 factor) were the predominant genes of the HFD, F, and C groups (Figure 4a). Furthermore, the Welch’s T-test of these genes showed that compared with the C group, in the HFD group, the relative abundance of 28 genes was upregulated, whereas this was downregulated for 11 others (Figure 4b). However, compared with the HFD group, the relative abundance of the following 3 genes in the F group k04567 (KARS, lysS), k02864 (RP-L10, MRPL10, rplJ), k06409 (spoVB) was upregulated. Furthermore, the abundance of 4 genes k05366 (mrcA), k01867 (WARS, trpS), k21071 (pfk, pfp), k13292 (lgt, umpA) was downregulated (Figure 4b). Moreover, the results of the Spearman correlation analysis of 7 genes and significantly different gut microbial species determined 3 genes negatively and 2 genes positively related to Firmicutes bacterium CAG:582, the bacteria identified as the most predominant species between the HFD and F group (Figure 4c, Supplementary Table 5).
Further comparative analysis of the CAZy database also found that murein polymerase (GT51), sucrose synthase (GT4), β-xylosidase (GH43) were dominant enzymes in the HFD, F, and C groups the after LDA analysis, respectively (Figure 5a). After the Welch’s T-test, the HFD group showed higher PL21, GT51, CBM51, GT5 expression levels than those of the C group, whereas levels of the other 8 enzymes were lower. However, the F group showed relatively lower levels of the 6 enzymes, including GT51, GT5, GH96, GH77, GT8, GH140 than those of the HFD group (Figure 5b). Furthermore, the distribution of eggNOG functions was also compared among the three groups, which were clearly separated based on the PCoA (Figure 5c). According to the LDA analysis, the dominant eggNOG pathway of C, F and HFD are transcription, replication recombination and repair, and defense mechanisms pathway (Figure 5d).
We also analyzed the distribution of antibiotic-resistant genes (ARGs) in the experimental groups. The ARG annotation in the CARD database results showed that 779 antibiotic resistant organisms (ARO) were annotated. The Venn diagram illustrates that the significant changed ARO between C and HFD, F and HFD groups had 71 and 48, respectively, and 23 common ARO (Figure 6a). The LDA showed that ARO:3002368 (PER-6), ARO:3002651 (APH(3')-Vc), ARO:3002817 (carA) were the predominant genes of the HFD, F, and C groups (Figure 6b). The HFD group demonstrated higher levels of APH (3')-VIIa, AAC (2')-Ia, chrB, fexA, vanTC, TriC, baeR, baeS, amrB, AAC (6')-Iih, tet (J), CMY-38, lsaE and Bacillus clausii chloramphenicol acetyltransferase, and lower levels of carA and adeC, compared with C group. However, the F group showed significantly higher levels of vgbA, MexT, vanD, clbA, carA and adeC than those in the HFD group (Figure 6c). The spearmen correlation analysis demonstrated that the above mentioned 6 ARO were positively related with decreased bacteria in HFD group, and negatively related with increased bacteria species in HFD group (Figure 6d, Supplementary Table 6).
Overall, the analysis of the microbiome-associated genes revealed that the HFD group displayed a dysfunction in several pathways of the KEGG, CAZy, and eggNOG databases. The metabolism of energy, carbohydrates, lipids, vitamins, and nucleotides was shown to be more disrupted in the HFD group than the C group. However, these metabolic pathways and their related gene expression were regulated in the F group, which established a healthy intestinal condition.