Changes in gut bacteria during atherosclerosis progression around menopause linked to blood fat metabolites

After HFD-fed ApoE^−/- mice undergoing bilateral ovariectomy, they were intragastrical administered with 0.13 mg/kg estrogen for 90 days, the levels of TC, TG, and LDL-c were significantly reduced whereas HDL-c was significantly increased in them, when compared to those of HFD-fed ApoE^−/- mice without receiving surgery (Figure 1(a)). Furthermore, estrogen supplementation relieved the atherosclerotic lesions aggravated by HFD and ovariectomy, manifested as the reduction of lipid deposits in the thoracic aorta (Figure 1(b,c)) and aortic root (Figure 1(d,e)) together with intima-media damage (Figure 1(f)).

Orthogonal partial least squares discriminant analysis (OPLS-DA) was also used to analyze metabolomics data, allowing the visualization and depiction of general metabolic variations between two groups. R2X, R2Y, and Q2Y represent the interpretability of independent variables, the interpretability of dependent variables and the predictability of OPLS-DA, respectively (Figure 3(b,d,f). The permutation test can be used to evaluate whether OPLS-DA is overfitting, with R2Y and Q2 representing the goodness-of-fit coefficients. No comparisons between every two groups showed overfitting (Figure 3(c,e,g)). The interpretabilities of OPLS-DA between normal diet-fed C57BL/6 mice and HFD-fed ApoE^−/- mice, between HFD-fed ApoE^−/- mice and ovariectomized HFD-fed ApoE^−/- mice, as well as between ovariectomized HFD-fed ApoE^−/- mice and estrogen-treated ovariectomized HFD-fed ApoE^−/- mice exceeded 0.5, and the discrimination between every two groups was larger. There were remarkable separations between the indicated two groups (Figure 3(b,d,f)).

At the phylum level, CE18:2 in CE was significantly positively correlated with Verrucomicrobia and Tenericutes, but negatively correlated with Actinobacteria, the same as CE18:3. CE22:5 was significantly negatively correlated with Bacteroidetes. A variety of TG showed positive correlations with Bacteroidetes, Proteobacteria and Deferribacteres. PC (P-14:0/2:0) and PC (P-17:0/2:0) in phospholipids showed significant positive correlations with Tenericutes. PC (P-16:0/26:4), PC (P-16:0/2:0) and PC (P-14:0/24:0) were negatively correlated with Verrucomicrobia (Figure 11).

At the family level, most kinds of CE were significantly positively correlated with Streptococcaceae, Enterobacteriaceae, Leuconostocaceae, Aerococaceae, Nocardiaceae, Moraxellaceae, Staphylococcaceae, Bacillaceae, Peptostreptococcaceae, Dermacoccaceae, Carnobacteriaceae and Bradyrhizobiaceae, whereas negatively correlated with Bifidobacteriaceae. Some TG were positively correlated with Enterobacteriaceae and negatively correlated with Enterococcaceae, but the remaining showed negative correlation. Almost all kinds of phospholipids were significantly positively correlated with Streptococcaceae, Enterococcaceae, Aerococaceae, Leuconostocaceae, Bacillaceae, Peptostreptococcaceae, Dermacoccaceae, Microbacteriaceae, Carnobacteriaceae, Pseudomonadaceae, Bradyrhizobiaceae and Mycobacteriaceae, and negatively correlated with Erysipelotrichaceae. The other types of lipids (including free fatty acids, acylcarnitine, sphingomyelins and ceramides) showed significant positive correlations with Rhizobiaceae, Moraxellaceae, Staphylococcaceae, Enterobacteriaceae, Nocardiaceae, Aerococaceae, Streptococcaceae, Enterococcaceae, Carnobacteriaceae, Bradyrhizobiaceae, Leuconostocaceae, Microbacteriaceae, Bacillaceae, Peptostreptococcaceae and Dermacoccaceae (Figure 12).

In comparison, the correlations between the gut microbiota at the genus level and lipid metabolomics were more complicated and detailed than those at the family level owing to existence of more species (Figure 13).

Firstly, the correlations of estrogen deficiency and supplementation with the gut microbiota and lipid metabolites were not validated for normally fed ApoE^−/- mice. Secondly, the composition of the fecal gut microbiota is more like that of the large intestine, but the expressions of lipid-related genes in our study were only detected in the small intestine responsible for lipid absorption and transport. Further in-depth studies are needed to clarify the similarities and differences of genes related to lipid absorption and transport between large and small intestines. Thirdly, human subjects with or without AS in premenopausal and postmenopausal periods were not tested.

Eight-week-old female C57BL/6 and ApoE^−/- mice were purchased from Beijing Hua Fukang Biological Technology Co., Ltd. (China). Four individuals were housed in each cage, with free access to food and water. Sixteen ApoE^−/- mice received bilateral ovariectomy and 90 days of HFD (including 0.3% cholesterol and 20% pork fat; Beijing Hua Fukang Biological Technology Co., Ltd., China). Eight ApoE^−/- mice were subjected to sham operation (needle threading, without ovariectomy) and maintained on HFD for 90 days. Eight C57BL/6 mice, which were used as a control group, were given sham operation, and maintained on a normal diet. Eight of the ApoE^−/- mice undergoing bilateral ovariectomy were intragastrical administrated with estrogen (0.13 mg/kg β-estradiol; Sigma-Aldrich, USA) daily for 90 days. Other mice were intragastrical administered with sterile carboxymethyl cellulose sodium (1%) daily.

In another independent experiment, twenty ApoE^−/- mice received bilateral ovariectomy and HFD for 90 days. These twenty mice also received fecal microbiota transplantation (FMT) every 3-day pre time for 90 days. Five of them received feces from C57BL/6 mice with sham operation and normal diet, five of them received feces from ApoE^−/- mice with sham operation and HFD, five of them received feces from ApoE^−/- mice with bilateral ovariectomy and HFD, and the last five of them received feces from ApoE^−/- mice with bilateral ovariectomy, HFD, and estrogen supplementation. Before these mice received FMT, broad-spectrum antibiotic was added to the drinking water (vancomycin 0.5 g/L and cefixime1g/L) to suppress the intestinal flora for two weeks. After that, fresh feces of donor mice were collected every three days. There were three mice in each group of the donors, and all donor mice were treated (sham operation or bilateral ovariectomy, normal diet or HFD, and estrogen supplementation) on the same time as the mice (bilateral ovariectomy and HFD) receiving fecal microbiota transplantation to ensure the feces were fresh. After the feces of each donor mouse were collected separately, the feces of the same group of donor mice were mixed and put into one sterile tube. After collecting, the feces were mixed in sterile PBS (1 g feces/ml PBS) and centrifuged at 500 rpm/min for 5 minutes. The supernatant was intragastrical administered to the recipient mice (0.1 ml/10 g).

All experimental procedures were performed in accordance with the national and international guidelines and regulations, and approved by Nanjing University of Chinese Medicine Animal Care and Use Committee (approval number: ACU-40(20141226) and ACU-02(20200425)).

Blood was collected from the inner canthus after 90 days of administration and left still at room temperature for over 30 min, from which serum was separated by centrifugation at 1500 rpm for 10 min. Afterward, 100 µL of serum was collected from every sample. The serum TG, TC, LDL, and HDL levels were tested using biochemical kits (Jiancheng Bioengineering Institute, Nanjing, China) by HITACHI 7020 Chemistry Analyzer.⁵⁴

After blood collection, all the mice were anesthetized with isoflurane before sacrifice. The liver and aorta were separated and fixed in 4% paraformaldehyde or 2.5% glutaraldehyde for 12–24 h. The atherosclerotic lesions of the aorta were evaluated by oil red O staining⁵⁵ of thoracic aortic root cross-sections,⁵⁶ and by scanning electron microscopy of aortic root cross-sections,⁵⁷ The lipid deposition damage of the liver was evaluated by oil red O staining, hematoxylin-eosin (HE) staining, and transmission electron microscopy of liver tissues.⁵⁸ Different histological examination methods were briefly described as follows.

For oil red O staining, liver tissues were fixed in 4% paraformaldehyde for over 12 h and rinsed with deionized water for 1 h. Then, the tissues were dehydrated with saturated sucrose solution for over 12 h, OCT-embedded, and cut into 10 μm-thick sections with a freezing slicer. Then, the sections were washed with water and 70% ethanol solution, stained by oil red O staining solution, washed with 70% ethanol solution, mounted with glycerin-gelatin jelly, and observed and photographed under an optical microscope. In contrast, the thoracic aorta was only dehydrated, washed, stained by the same reagent, and photographed by using a digital camera.

For HE staining, the liver tissues were fixed in 4% paraformaldehyde, rinsed by tap water for 1 h, dehydrated with different concentrations of ethanol solutions (70%, 80%, 90%, 95%, and 100%), transparentized by using xylene for 30 min and immersed in paraffin at 65°C for 45 min to fill the interstitial space. The resulting paraffin block was then cut into 5 μm-thick sections by using a slicer, heated at 65°C for 30 min, immersed in xylene for 20 min and rehydrated with different concentrations of ethanol solutions (100%, 95%, 90%, 80%, and 70%) and tap water. Subsequently, the sections were stained with hematoxylin staining solution for 3 min, rinsed with tap water for 20 min, stained with eosin staining solution for 30 s, washed with tap water, dehydrated by different concentrations of ethanol solutions (70%, 80%, 90%, 95%, and 100%), transparentized by xylene, mounted with neutral resin and photographed under an optical microscope

For electron microscopic imaging, the thoracic aorta and liver were fixed in 2.5% glutaraldehyde overnight. Afterward, the samples were immersed in ultrapure water for 1 h, re-fixed with osmic acid for at least 4 h, and dehydrated with different concentrations of tertiary butanol solutions (30%, 50%, 70%, 80%, 90%, and 100%), from which the remaining liquids were removed by a supercritical extractor. Then, each sample was treated with a gold-plated instrument and placed in a conductive copper mesh to be observed and photographed by scanning electron microscopy or transmission electron microscopy.

a. Sample preparation

Blood was collected from mice after 90 days of administration and left still at room temperature for over 30 min, from which plasma was separated by centrifugation at 1500 rpm for 10 min. Samples were prepared based on liquid–liquid MTBE extraction to analyze lipid metabolism. Briefly, plasma (20 μL) was added into a 1.5 mL centrifuge tube, mixed with 225 μL of ice-cold methanol solution and internal standard (lysoPE (17:1), SM (17:0) for positive ion mode and PE (17:0/17:0) for negative ion mode; concentration: about 5 μg/mL), and vortexed for 10 s. Next, 750 μL MTBE was added, and the mixture were shaken for 10 min at 4°C. After 188 μL of deionized water was added, the mixture was vortexed for 10 s and then centrifuged at 14,000 rpm at 4°C. Lipids in the upper (organic) phase were transferred to clean tubes and dried by a vacuum centrifuge. Finally, the upper phase lipids were reconstituted with 110 μL of methanol: toluene (9:1) for LC-MS.

b. Untargeted lipidomic analysis

To detect lipids, 2 μL aliquots of sample solution were injected into a reversed-phase Waters Acquity UPLC CSH C18 column (100 mm × 2.1 mm, 1.7 μm) maintained at 60°C for gradient elution. Mobile phase A was water: ACN (6:4), and mobile phase B was isopropanol: ACN (9:1), both containing 10 mM ammonium formate and 0.1% formic acid. The flow rate was 0.3 mL/min. The elution gradient was as follows: 0–4.0 min, 15% B; 4.0–5.0 min, 15–48% B; 5.0–22.0 min, 48%–82% B; 22.0–23.0 min, 82–99% B; 23.0–24.0 min, 99% B; 24.0–24.2 min, 99%–15% B; 24.2–30.0 min, 15% B.

The spray voltage was 3.5 kV in the positive ion mode and 3.0 kV in the negative ion mode. For both ion modes, the sheath gas, aux gas, capillary temperature, and heater temperature were maintained at 35, 15 (arbitrary units), 325°C and 300°C, respectively.

c. Data processing

After lipid annotation, a small-scale database was set up with lipid name, retention time and accurate mass/charge ratio (m/z). Raw data files acquired from Xcalibur 2.2 software (Thermo Scientific, USA) were converted to the ABF format using ABF converter (accessible at: http://www.reifycs.com/AbfConverter). For data processing, MS-DIAL (v. 2.78) software program was used. In this study, only a lipid feature defined as an m/z – retention time pair can be aligned for an identical lipid. The resulting output data tables of high-quality time-aligned lipids together with the corresponding retention time, m/z and peak area were subjected to further statistical analysis. The screening conditions for relevant lipid metabolites were p value of ≤0.05 in combination with fold change of ≥1.5 or ≤0.667. PCA and OPLS-DA were performed according to the lipid metabolites. A heat map of serum lipid metabolites was also derived from the original data of lipid metabolism.

a. Fecal sample processing and microbial DNA extraction

The fecal samples of mice were frozen immediately in liquid nitrogen prior to euthanasia and stored at −80°C under sterile conditions until analysis. Fecal genomic DNA was extracted from the fecal samples with QIAamp® DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol. The DNA concentration and purity were detected by Nanodrop 1000 spectrophotometer (Thermo Fisher, Waltham, MA), and the DNA integrity was tested by 0.8% agarose gel electrophoresis.

b. 16SrRNA gene library preparation and high-throughput sequencing

Bacterial genomic DNA was used as the template to amplify the V3–V4 hypervariable region of 16SrRNA gene with high-fidelity Phusion polymerase (M0530S; Thermo Fisher, USA), forward primer (5’-CCTACGGGNGGCWGCAG-3’) and reverse primer (5’-GACTACHVGGGTATCTAATCC-3’). The universal sequence of the illumina adapter were added to the 5’ ends of each primer. Each sample was amplified by three repeated PCR experiments. The PCR products were tested by agarose gel electrophoresis. Then the products from the same sample were pooled, and purified using Agencourt AMPure XP Kit (Beckman Coulter, CA, USA). Then, using primers with Index sequence, a specific tag sequence compatible with the Illumina platform was introduced through high-fidelity PCR to construct the final complete library structure. The amplified products were purified with Agencourt AMpure XP magnetic beads to obtain an original library of samples. The library quality was assessed with Qubit@2.0 Fluorometer (Thermo Scientific, USA) and Agilent Bioanalyzer 2100 system (USA). The V3-V4 hypervariable regions of the 16S rRNA gene were sequenced using Illumina MiSeq Sequencer, at least 5 M 2x250bp pair-end raw reads were generated.

c. Sequence quality control and microbial community analysis

Raw reads were quality-filtered and merged with the following steps. First, the raw reads at any site with an average quality score of <20 were truncated and the adapter sequence and reads with lengths of <100 were removed by TrimGalore. Second, the paired reads were merged into tags by using Fast Length Adjustment of Short reads (FLASH, v1.2.11). The reads with ambiguous base (N base) and homopolymer of >6 bp were removed by Mothur. Third, the reads with low complexity were removed by using USEARCH to obtain clean reads for further bioinformatics analysis. The remaining reads were chimera-checked according to the gold.fa database (http://drive5.com/uchime/gold.fa↗) and clustered into OTUs by UPARSE with 97% similarity cutoff. Bioinformatics analysis was performed by Genesky Biotechnology Inc. (Shanghai, China).

All OTUs were classified based on Ribosomal Database Project. Alpha-diversities indicating within-sample richness were analyzed by Mothur. Sample tree cluster by Bray-Curtis distance matrix, unweighted pair-group method with arithmetic means and Jaccard principal coordinate analysis based on OTUs were conducted by R Project (Vegan package, V3.3.1). Redundancy was analyzed by Canoco for Windows 4.5 (Microcomputer Power, NY, USA) and assessed by MCPP with 499 random permutations. Linear discriminant effect size analysis was carried out to identify the microorganism features distinguishing fecal microbiota specifically for biomarker discovery. The two independent sample t-test and Mann – Whitney U test was performed to detect the significant differences of abundances among taxa.

Data were represented as mean ± standard error of mean (mean ± SEM) if not indicated in another way. The four groups were compared using the Kruskal-Wallis test with the Dunn’s multiple comparisons test. The Mann–Whitney test was used to compare the differences of two groups. Differences were considered significant when p < .05. All statistical analyses were conducted with SPSS 23.0 for Windows (SPSS Inc., USA)

The correlations between the high-throughput 16SrRNA sequencing results of gut microbiota and plasma lipid metabolites were studied by analyzing those between species and environmental factors. Relevant original files, including Supplementary Table 3 and Table 4, were uploaded to the data analysis server (available at: http://cloud.geneskybiotech.com/login.html↗). The correlation coefficients (Pearson correlation coefficient) between the selected species and environmental factors were calculated. Finally, the correlation matrix was visualized through a heat map.