Integrative analysis of metabolome and gut microbiota in diet-induced hyperlipidemic rats treated with berberine compounds

BBR, γ-oryzanol, pyridoxine (vitamin B₆), urease, pyridine, methoxylamine hydrochloride, L-phenylalanine-¹³C₉-¹⁵N, dulcitol, L-leucine-¹³C₆, L-isoleucine-¹³C₆-¹⁵N, L-valine-¹³C₅-¹⁵N, L-alanine-¹³C₃-¹⁵N and N,O-bis(trimethylsilyl)-trifluoroacetamide (BSTFA) with 1 % trimethylchlorosflane (TMCS) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Pure water was obtained from Millipore Alpha-Q water system (Bedford, MA, USA). Methanol and ethanol for HPLC grade were purchased from Merck Chemicals (Darmstadt, Germany). Chloroform for analytical grade from Sinopharm Chemical Reagent Company (Shanghai, China). KAPA HiFi polymerase (KAPA Biosystems, USA), Agencourt AMPure XP beads (Beckman Coulter, USA), Qubit dsDNA HS assay kit (Invitrogen, CA), Agilent high sensitivity DNA Kit (Agilent, USA), Ion PGM template OT2 200 kit, Ion PGM sequencing 200 kit v2 and Ion 316 chip kit v2 (Life Technology, CA) and QIAamp^® Fast DNA stool mini kit (Qiagen, Germany) were obtain for fecal microbiota analysis.

Male Wistar rats (8-week old, weighting 180–220 g), purchased from Shanghai SLAC Laboratory Animal Co., Ltd (Shanghai, China), were housed in a standard 12 h light/12 h dark cycle. The animal room temperature and relative humidity were 22 ± 2 °C and 60 ± 5 %, respectively. After 1-week acclimatization, rats were divided into chow diet group (n = 10) and HFD (10 % lard, 20 % sucrose, 2 % cholesterol, 1 % bile salt and 67 % standard chow) group (n = 20). Blood samples were collected randomly from tail vein of 6 rats in each group after 4 week feeding, and TC, TG, LDL-c, HDL-c and free fatty acid (FFA) level in serum were quantified. The hyperlipidemic rats were further divided into two groups, namely, untreated group (continued to supply HFD, n = 10) and BC treated group (with HFD and BC supplementation, n = 10). BC (150 mg/kg of BBR, 24 mg/kg of oryzanol and 10 mg/kg of vitamin B₆) were suspended with 0.5 % sodium carboxy methylcellulose (CMC-Na) solution, and then administered by oral gavage for the next 4 weeks. The chow diet and HFD groups (untreated) rats were received an equal volume 0.5 % CMC-Na solution. Food intake and body weight were recorded weekly. At the end of experiment, all rats were fasted overnight, and sacrificed with 2 % sodium pentobarbital anesthesia (0.5 mL/100 g). Blood samples were drawn from the abdominal aorta, and the serum was obtained by centrifugation at 3000 rpm for 15 min at 4 °C. To collect urine and feces samples, six rats in each experimental group were housed in metabolic cages, and the samples accumulated in the metabolic cages were immediately transferred into sterile tubes. The liver samples were immediately dissected, weighted, washed with icy normal saline, snap-frozen in liquid nitrogen, and then stored at −80 °C.

All the experimental protocols were approved by the animal committee of Shanghai University of Traditional Chinese Medicine.

For each serum sample (20 μL), 80 μL of ice-cold methanol containing 10 μL of internal standards (0.02 mg/mL of L-phenylalanine-¹³C₉-¹⁵N, 0.05 mg/mL of dulcitol, L-leucine-¹³C₆ and L-isoleucine-¹³C₆-¹⁵N, 0.1 mg/mL of L-valine-¹³C₅-¹⁵N and L-alanine-¹³C₃-¹⁵N) was added. For each urine sample (15 μL), 15 μL urease solution were blended and catalyzed at 37 °C for 60 min, and then 180 μL of ice-cold ethanol containing 10 μL of internal standards (0.05 mg/mL of L-phenylalanine- ¹³C₉-¹⁵N and dulcitol) was added. For each liver tissue or feces sample (40 ± 5 mg), 800 μL chloroform/methanol/water (v/v/v, 2:5:2 in ice water) mixture were added and homogenized. After centrifuged at 16,000g for 15 min, the supernatant (100 μL) was added to a GC vial, containing 10 μL of internal standards (0.05 mg/mL of L-phenylalanine-¹³C₉-¹⁵N and dulcitol).

The respective serum, urine, liver and feces samples were dried under gentle nitrogen stream. The glass vial with dry residue was added with 30 μL of 20 mg/mL methoxylamine hydrochloride in anhydrous pyridine. The resultant mixture was vortex-mixed vigorously for 30 s and incubated (37 °C for 90 min). A 30 μL of BSTFA (with 1 % TMCS) was added into the mixture and derivatized (70 °C for 60 min).

Each derivatized sample (1 μL) was injected using the splitless mode with an Agilent Technologies 7890A chromatograph equipped with a HP-5MS column (30 m × 0.25 mm × 0.25 μm) and Agilent Technologies 5975C inert MSD detector. The initial oven temperature was held at 70 °C for 2 min, increased to 160 °C with 6 °C/min, and then to 240 °C with 10 °C/min, and finally increase to 300 °C with 20 °C/min, constant for 6 min, with He as carrier gas (1 mL/min) and MS detection. The temperatures of injector, transfer line, and electron impact ion source were set to 250, 290, and 230 °C, respectively.

The extraction, alignment, deconvolution, and further processing of raw GC/MS data were converted into NetCDF format via DataBridge (Perkin-Elmer, USA). The data was normalized against total peak intensities before performing univariate and multivariate statistics. Multivariate data analysis was carried out using SIMCA-P 11.0 software (Umetrics AB, Umeå, Sweden) to perform principal component analysis (PCA) where general clusters and outliers were observed. Prior to PCA, all data were mean-centered and unit variance-scaled. Subsequently, the data were subjected to partial least squares-discriminant analysis (PLS-DA) and orthogonal partial least squares-discriminant analysis (OPLS-DA) where a model was built and utilized to identify and reveal differential metabolites accountable for the separation between identified groups. The differential metabolites were determined by cross-referencing with the Golm Metabolome Database. In addition, metabolic pathway interpretation of differential metabolites was performed using the KEGG database.

For the extraction of fecal DNA, 180–220 mg each stool sample were weighed on ice and operated based on the protocol for “Isolation of DNA from stool for human DNA analysis” in the handbook provided by QIAamp DNA stool mini kit. DNA yields are determined from the concentration of DNA in the eluate, measured by absorbance at 260 nm. Purity is determined by calculating the ratio of absorbance at 260 to 280 nm, measured by NanoDrop microvolume quantitation of nucleic acids. pure DNA has an A260/A280 ratio of 1.7–1.9.

The DNA extractive from each stool sample was used as a template for the amplification of V3 region of 16S rDNA genes. The bacterial genomic DNA was PCR amplified with the forward primers (5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG) and the reverse primers (5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC) for the V3 hypervariable regions of the 16S rDNA gene.

The PCR condition were 95 °C for 3 min, followed by 25 cycles of 95 °C for 30 s, 55 °C for 30 s and 72 °C for 30 s, and then 72 °C for 5 min on an Eppendorf thermocycler. The PCR products were verified on a Bioanalyzer DNA 1000 chip (Agilent), and the expected size on a Bioanalyzer trace after the Amplicon PCR step is ~550 bp. Then we use AMPure XP beads to purify the 16S V3 amplicon away from free primers and primer dimer species, and quantified the sample with fluorometric quantification method that used dsDNA binding dyes. The concentrated final library was diluted to 4 nM using resuspension buffer (RSB) or 10 mM Tris pH 8.5. 5 μL of diluted DNA from each library was mixed for pooling libraries with unique indices. After samples were loaded, the MiSeq system provided instrument secondary analysis using the MiSeq Reporter software (MSR). The MSR provided several options for analyzing MiSeq sequencing data. All operations for fecal microbe analysis were performed under sterile conditions.

Raw pyrosequencing reads generated from the sequencer were quality filtered by FASTX Toolkit 0.0.13 (http://​hannonlab.​cshl.​edu/​fastx_​toolkit/​index.​html). The high-quality valid reads were clustered into operational taxonomic units (OTUs) using Mothur (http://​www.​mothur.​org/​). The representative sequences of OTUs were used to analyze α-diversity (rarefaction curve analysis and Shannon diversity index) on the basis of their relative abundance. A heatmap was generated according to the relative abundance of OTUs by R software (http://​www.​R-project.​org). Phylogenetic β-diversity measures such as unweighted unifrac significance test, principal coordinate analysis (PCoA) and nonmetric multidimensional scaling (NMDS) were performed using representative sequences of OTUs for each sample by the Mothur program to analyze community and phylogenesis. Taxonomy-based analyses were performed to classify taxonomically using the ribosomal database project (RDP) classifier with a 60 % bootstrap score.

Male Wistar rats developed hyperlipidemia after 5 weeks of HFD feeding. Thus, the serum TC and LDL-c concentrations were increased whereas HDL-c concentration decreased as compared to chow diet controls (Additional file 1: Figure S1). A trend of increase in serum TG and FFA was also observed in HFD-fed group, although the difference was statistically insignificant. Treatment of the hyperlipidemic rats with BC for 4 weeks significantly attenuated the serum concentrations of TG, TC, LDL-c, whereas the serum HDL-c concentration was restored as compared to that in untreated (i.e. HFD alone) rats (Fig. 1a–d). The serum FFA concentration was unchanged (Fig. 1e).

Rats fed HFD for 9 weeks developed hepatomegaly (Table 1). The body weight between HFD-fed and chow diet-fed rats showed no difference (Table 1). After 4-week BC treatment, both body weight and liver/body weight ratio were decreased compared with those untreated (i.e. HFD alone) rats (Table 1). Food intake was identical between HFD and BC-treated groups (Table 1), thus the weight loss upon BC treatment was unrelated to satiety.

Analysis of serum metabolites between HFD-fed (i.e. hyperlipidemia model) and chow diet-fed rats revealed changes in 31 serum metabolites (Additional file 1: Table S1). Of which, metabolites related to fatty acid synthesis or cholesterol synthesis were enhanced, whereas metabolites related to TCA cycle were decreased (Additional file 1: Table S1) upon high fat dieting.

We next determined the effect of BC treatment on serum metabolites in the hyperlipidemic rats. The R ² X and Q ² of PCA score analysis were 0.595 and 0.101, respectively; the R ²Y and Q ² of PLS-DA score analysis were 0.995 and 0.802, respectively. These data indicated the validity of current models.

Distinct clustering of serum metabolites was apparent between control and BC treated groups (Fig. 2a, b). The set of identified metabolites were systematically searched for Pearson’s correlations, and the correlations were indicated in different colors (Fig. 2c). Fold change (FC) of each metabolite was used to construct a heat map (Fig. 2d). Compared with the metabolic profiles of untreated rats, 15 metabolites were increased and 16 decreased in BC treated rats, and the profiles are summarized in Fig. 2e. Among the increased metabolites, oxaloacetic acid, pyruvic acid and malic acid are involved in glycolysis and the TCA cycle (Fig. 2f), glutamic acid, aspartic acid, and alanine are involved in the metabolism of gluconeogenic amino acids (i.e. alanine, aspartate and glutamate) (Fig. 2g). Decreased metabolites, such as cis-11,14-eicosadienoic acid, stearic acid, margaric acid, myristic acid, arachidic acid, trans-oleic acid, palmitoleic acid, and palmitic acid, are involved in fatty acid biosynthesis (Fig. 2h).

A total of 51 urine metabolites showed difference between HFD-fed and chow diet-fed rats (Additional file 1: Table S2), of which 15 were increased and 41 were decreased. Alterations in these metabolites suggested suppression in the TCA cycle, the pentose phosphate pathway, and purine/pyrimidine metabolism (Additional file 1: Table S2).

We next determined the effect of BC treatment on urine metabolites in the hyperlipidemic rats. The R ² X and Q ² of PCA score analysis were 0.557 and 0.204, respectively; the R ²Y and Q ² of OPLS-DA score analysis were 0.995 and 0.824, respectively (Fig. 3a, b), indicating the classifications was well suited for the models, and the BC untreated and treated groups were classified clearly. The set of identified metabolites was systematically searched through for Pearson’s correlations (Fig. 3c), and FC of each metabolite was used to construct a heat map (Fig. 3d). The observed changes in metabolites in the rat urine samples were summarized in Fig. 3e. Notably, metabolites pyridoxine and 4-pyridoxic acid that are related to vitamin B₆ metabolism were remarkably increased (Fig. 3f) upon BC treatment. In addition, metabolites involved in phenylalanine metabolism, such as succinic acid and 3-hydroxyphenylpropionic acid, were decreased, whereas phenylacetic acid and phenyllactic acid were increased (Fig. 3g).

For liver samples, we identified 36 metabolites in hyperlipidemic rats that distinctly different from those in chow diet group (Additional file 1: Table S3), 16 metabolites were increased and 20 metabolites were decreased in hyperlipidemic rats. These altered metabolites were supposed to be involved in the TCA cycle, the pentose phosphate pathway, fatty acid metabolism, and amino acid metabolism (Additional file 1: Table S3).

We next determined the effect of BC treatment on liver metabolites in the hyperlipidemic rats. The R ² X and Q ² of PCA score analysis were 0.539 and 0.226, respectively; the R ²Y and Q ² of OPLS-DA score analysis were 0.996 and 0.564, respectively (Fig. 4a, b), validating the classification for the chosen model. The set of identified metabolites was systematically searched through for Pearson’s correlations (Fig. 4c), and FC of each metabolite was used to construct a heat map (Fig. 4d). Metabolites from the BC treated liver samples that were significantly different are summarized in Fig. 4e. Glucuronic acid, lyxonic acid, 3-hydroxyadipic acid and galacturonic acid showed increase, whereas phytosterol, γ-aminobutyric acid, methionine, nonanoic acid, hypotaurine, 1-octadecanol, lyxosylamine and N-acetylgalactosamine were decreased (Fig. 4f).

For the fecal samples, 27 metabolites were up-regulated and 16 metabolites were down-regulated in hyperlipidemic rats (Additional file 1: Table S4). Changes in these metabolites suggested the enhancement in fatty acid and cholesterol synthesis, and the suppression in TCA cycle, tryptophan metabolism, and purine/pyrimidine metabolism (Additional file 1: Table S4).

We next determined the effect of BC treatment on fecal metabolites in the hyperlipidemic rats. The R ² X and Q ² of PCA score analysis were 0.548 and 0.262, respectively; the R ²Y and Q ² of PLS-DA score analysis were 0.989 and 0.912, respectively (Fig. 5a, b), indicating the classification was well suited for the models, and the profiles between BC treated rats and hyperlipidemic rats were classified clearly. The set of identified metabolites was systematically searched through for Pearson’s correlations (Fig. 5c). FC of each metabolite was used to construct a heat map (Fig. 5d). The observed changes in endogenous metabolites in rat feces were summarized in Fig. 5e. It was noted that the BC increased the metabolites (tyrosine, hydrocaffeic acid, 4-hydroxycinnamic acid and indole-3-lactic acid) involved in the tyrosine metabolism (Fig. 5f), the metabolites (pantothenic acid, γ-aminobutyric acid and β-alanine) involved in the β-Alanine metabolism (Fig. 5g) and the metabolites (ornithine, proline and putrescine) involved in the arginine and proline metabolism (Fig. 5h), and decreased the metabolites (deoxycholate and lithocholate) involved in the secondary bile acid biosynthesis (Fig. 5i).

We used high quality pyrosequencing technology to determine the structure of fecal microbiota. A total of 363,615 usable raw sequences (average of 20,201 sequences per sample) and 34,866 operational taxonomic units (OTUs) were generated by pyrosequencing method from 18 samples. Rarefaction and Shannon diversity curves demonstrated that most of the diversity had already been captured (Additional file 1: Figure S2). The thetayc logarithm showed the gut microbiota could be divided into independent clusters on the basis of the community composition (Additional file 1: Figure S3A, S3B), and the relative abundance of bacterial class also showed clearly distinct (Additional file 1: Figure S3C, S3D). PCoA and NMDS were applied to visualize the similarities or dissimilarities of these data (Additional file 1: Figure S4). The weighted and unweighted unifrac significance test demonstrated that the microbiota structure of the BC treated group was distinct from that of the untreated group (Additional file 1: Table S5).

We detected 8272 OTUs through the three groups. By plotting the ranked abundance of all 8272 OTUs according to their occurrence, we chose the highest 50 OTUs in species abundance, most of which distributed across such families as Bacteroidaceae (12 OTUs), S24-7 (5 OTUs), Enterobacteriaceae (5 OTUs), Prevotellaceae (5 OTUs), Ruminococcaceae (4 OTUs), Paraprevotellaceae (4 OTUs), Lachnospiraceae (3 OTUs), Desulfovibrionaceae (2 OTUs), Verrucomicrobiaceae (2 OTUs) and Lachnospiraceae (2 OTUs). According to frequency within each sample, the heat map showed the genus level clustering (Fig. 6a), 31 of the 50 identified key OTUs were eliminated or decreased in HFD group, namely Bacteroides (3 OTUs), Prevotella (4 OTUs) and one OTU to each of the following genera: Escherichia, Sutterella, Parabacteroides, Clostridium and Blautia. Meanwhile, hyperlipidemic rats showed higher abundance of Akkermansia. 27 of the 50 identified key OTUs were eliminated or decreased with BC intervention, whereas the rest of identified OTUs were enriched. The identified OTUs belonging to Prevotella (4 OTUs) and one OTU to each of the following genera: Escherichia, Clostridium and Sutterella, which were markedly decreased at genus level in the BC treated group. Meanwhile, higher abundances of Bacteroides (10 OTUs), Parabacteroides (1 OTU) and Blautia (1 OTU) were observed in the BC treated group (Fig. 6b).

Analysis of serum ALT and AST showed that HFD-fed rats exhibited increase in these liver enzymes, suggesting compromised liver function was associated with hyperlipidemia (Table 2). However TBIL or TBA in the hyperlipidemic rats were normal (Table 2). Treatment with BC normalized serum AST to control level, and also had a trend in reducing ALT in the hyperlipidemic rats (Table 2). Overall, the BC treatment showed no adverse effect on any of the parameters examined (Table 2), suggesting the dosage of BC was well tolerated in these animals. The food intake and body weight gain were not affected by the BC treatment (Table 1), nor was dry stool or diarrhea observed in BC treated rats (data not shown).