Bacteroides vulgatus and Bacteroides dorei Reduce Gut Microbial Lipopolysaccharide Production and Inhibit Atherosclerosis

All participants provided written informed consent on enrollment, and the study was conducted according to the guidelines of the Declaration of Helsinki. This study was approved by the Ethics Committee of Kobe University (approval No. 1595) and registered with the University hospital Medical Information Network Clinical Trials Registry (trial registration No. UMIN000015703).

Thirty patients with CAD and 30 controls without CAD with coronary risk factors were recruited at Kobe University Hospital between October 2014 and July 2015. The sample size was calculated using R software (power=0.9, significance level=0.05, mean difference=6, SD=7; n=30 per group.). The group with CAD included patients with stable angina pectoris and old myocardial infarction with preserved left ventricular ejection fraction (>40%) who underwent percutaneous coronary intervention or coronary artery bypass graft surgery ≥6 months before the present study. Patients with acute coronary syndrome were excluded. Single- or multivessel disease was defined in terms of the number of major coronary vessels demonstrating >75% stenosis on diagnostic coronary angiography.

Thirty patients without CAD with coronary risk factors, such as hypertension, diabetes mellitus, or dyslipidemia, but without present or past history of coronary or other vascular diseases, were recruited, and age and sex were matched in the control group. All patients without CAD were hospitalized to treat supraventricular tachycardia such as atrial fibrillation and atrial flutter, hypertension, or diabetes mellitus. History of coronary or other vascular disease was defined as documented vascular disease, symptoms indicating angina pectoris, abnormality on electrocardiography indicating old myocardial infarction or angina pectoris, or abnormality on chest x-ray film.

Patients with systemic disease, including hepatic disease, renal disease (serum creatinine levels >2.0 mg/dL), collagen disease, or malignancy, were excluded from the study. Patients treated with antibiotics were also excluded. Diabetes mellitus was defined as glycohemoglobin levels >6.5% (per the indications of the National Glycohemoglobin Standardization Program), use of oral antihyperglycemic, or use of insulin therapy. Hypertension was defined as blood pressure >140/90 mm Hg or use of antihypertensive drugs. Dyslipidemia was defined as low-density lipoprotein cholesterol levels >140 mg/dL, triglyceride levels >150 mg/dL, or use of antidyslipidemic drugs per the guidelines issued by the Japan Atherosclerosis Society.¹³

Fecal samples were collected while participants were hospitalized and thus consuming a hospital diet for >1 day. DNA extractions from human fecal samples were performed by Nihon Gene Research Laboratories Inc according to a previously established procedure.^14–16

Parts of the 16S rRNA genes (the V3–V4 region, corresponding to positions 342–806 in the Escherichia coli numbering system) were polymerase chain reaction–amplified using our nondegenerate universal primer set for 342F and 806R. A detailed description of the primer set and polymerase chain reaction conditions is available elsewhere.¹⁷ After addition of the sequencing adapters, the amplicons were sequenced (250-bp paired end) at Takara Bio Inc using an Illumina MiSeq platform (Illumina Inc) according to the manufacturer’s protocol.

To create the bacterial composition matrix, we used USEARCH version 10.0.240.¹⁸ Our previous protocol was also used to select high-quality 16S rRNA gene amplicon sequences generated using Trimmomatic¹⁹ version 0.33 with the parameters LEADING:17 TRAILING:17 AVGQUAL:25 MINLEN:100. The remaining reads were processed using the -fastq_mergepairs command of USEARCH, with default parameters. Next, we removed sequences without the primer region using Tagcleaner²⁰ version 0.16, with parameters -tag5 CTACGGGGGGCAGCAG -mm5 3 -tag3 AGATACCCCGGTAGTCC -mm³ 3 -nomatch 3. After the removal of the primer, sequences with N were removed using an in-house python script. To remove PhiX reads, we used the -filter_phix command of USEARCH. Finally, we removed short sequences using the USEARCH command -sort_by_length, with the parameter -minseqlength 300. Finally, we generated operational taxonomic unit (OTU) tables using UPARSE algorithms (-fastx_uniquecomand and otu_cluster command with the parameter -minsize 1).²¹ The representative sequences of each OTU were annotated to bacterial genus using RDP Classifier version 2.12, with a bootstrap value ≥0.5.²² Moreover, we annotated each representative sequence of each OTU to the reference database SILVA Living Tree Project version 123²³ using BLASTN version 2.2.25, with identity threshold ≥97% and coverage ≥80%.

OTUs were constructed using the raw data reads from fecal samples. To join 2 paired-end reads, we used fastq-join software with default options. The chimeric sequences were deleted with usearch6.1.²⁴ OTUs at the 97% similarity threshold were selected with the Green genes database in the QIIME 1.8.0 pipeline.²⁵ Functional analysis of gut microbiota based on the KEGG Orthology database (Kyoto Encyclopedia of Genes and Genomes)²⁶ was performed using PICRUSt 1.1.1.²⁷ Identification of KEGG modules with a significant difference in mean relative expression was performed using Statistical Analysis of Metagenomic Profiles (STAMP 2.1.3) software.²⁸ The lipopolysaccharide (LPS) biosynthesis pathway map (map00540) was cited from http://www.kegg.jp/kegg/kegg1.html26 with permission.

All apolipoprotein E–deficient mice were on C57BL/6 background. All mice were housed in a specific pathogen-free animal facility at the Kobe University Institute. The animals were fed normal chow (CLEA) and water ad libitum under a strict 12-hour light/dark cycle. Six-week-old mice were randomly divided into 3 treatment groups: mice in the control group were gavaged with culture medium, those in the live Bacteroides group were gavaged with live B. vulgatus and B. dorei, and those in the heat-killed Bacteroides group were gavaged with heat-killed B. vulgatus and B. dorei at a dose of 2.5×10⁹ cfu/100 μL of B. vulgatus and B. dorei 5 times per week. All experiments were performed according to the Guidelines for Animal Experiments in effect at Kobe University School of Medicine (guideline No. P160701).

B. vulgatus (No. 8482; American Type Culture Collection) and B. dorei (No. 17855; Deutsche Sammlung von Mikroorganismen) were cultured anaerobically in Difco-reinforced clostridial medium (No. 218081; BD Bioscience) at 37°C. An anaerobic chamber (Coy Laboratory Products) containing 10% CO₂, 10% H₂, and 80% N₂ was used for all anaerobic microbiology steps. B. vulgatus and B. dorei were heat-killed at 121°C (treatment duration, 15 minutes). The success of the heat treatment was confirmed by the absence of growth of plated heat-killed bacteria.

The same human fecal samples were used for LPS measurement and 16S rRNA gene sequencing. Mouse fecal samples were collected from 16-week-old controls and Bacteroides-treated mice at 24 hours after the last oral gavage. Fecal supernatant was obtained according to a previously described protocol with some modifications.^29,30 Briefly, the fecal samples were suspended in sterile PBS to a concentration of 1 g per 10 mL (human) or 50 mg per 500 μL (mouse) and vortexed mildly to avoid disruption of bacterial cells. After centrifugation for 15 minutes at 3000 rpm, the supernatant was collected, sterilized by filtration through a 0.45-μm filter followed by refiltration through a 0.22-μm filter, inactivated for 15 minutes at 90°C, and stored at −80°C.

The plasma and fecal LPS levels were determined using a limulus amebocyte lysate assay (No. K50-643J; Lonza Inc) according to the manufacturer’s instructions. The plasma was diluted 10-fold, and the fecal supernatant was diluted 10 000-fold in pyrogen-free water and inactivated for 15 minutes at 90°C. LPS measurements were performed in pyrogen-free glass tubes, Eppendorf tubes, and plates.

Statistical analyses were performed using R software, version 3.1.0, JMP version 10 (SAS Institute), and Prism version 7.0 (GraphPad Software). The Shapiro-Wilk test was used to determine whether the data were normally distributed. Results were expressed as mean±SEM or mean±SD for normally distributed data, and median ± interquartile range (25th to 75th percentiles) for nonnormally distributed data. The significance of differences between 2 groups was evaluated using the 2-tailed Student’s t test for normally distributed data or Mann-Whitney U test for nonnormally distributed data. The Fisher exact test or χ² test was used to compare categorical variables. For all tests, a value of P<0.05 was considered to indicate statistical significance. To discover the strength and direction of a link between 2 parameters, Spearman’s rank correlation coefficient was calculated. One-way ANOVA was used to detect significant differences among 3 groups. The q values were calculated using the Benjamini-Hochberg method to adjust the P values for multiple comparisons. The clustering of data from 30 patients with CAD and 30 controls was performed at the genus level, as described in detail elsewhere.³¹ The Shannon-Wiener index was calculated using the vegan package for R software. Principal component analysis was performed using JMP.

We recruited 30 patients with CAD and 30 age- and sex-matched controls without CAD with coronary risk factors, such as hypertension, diabetes mellitus, or dyslipidemia. We then performed a detailed comparison of gut microbial profile using 16S rRNA gene sequencing in fecal samples. The baseline characteristics of the participants are shown in Table 1. The α diversity, Bacteroidetes/Firmicutes ratio, and Gram-positive to Gram-negative strain ratio did not differ significantly between the 2 groups (Figure 1A and 1C). To analyze the gut microbial profiles of the study participants, the samples were clustered into 3 clusters at the genus level, according to the procedure followed in a previous report (Figure 1D).^8,31 Each cluster was characterized by a high abundance of a specific genus as follows: Bacteroides in cluster 1, Prevotella in cluster 2, and Faecalibacterium, Ruminococcus, or Bifidobacterium in cluster 3 (Figure 1E). The controls without CAD were significantly more likely to be categorized into cluster 1, whereas few patients with CAD were categorized into this cluster (Figure 1F). A comprehensive comparison between the patients with CAD and controls in terms of the abundance of genera with an average relative abundance >0.1% is provided in Table I in the online-only Data Supplement. We found that the relative abundance of the genus Bacteroides was visibly lower in patients with CAD compared with controls (Figure 1G). The associations of gut microbial genera with clinical indices are shown in Figure I in the online-only Data Supplement. Principal coordinate analysis showed that the 2 groups differed in the abundance of major species of the gut microbiome, as patients with CAD showed a relative depletion of B. vulgatus and B. dorei and enrichment in Faecalibacterium prausnitzii and Prevotella copri (Figure 1H). Considering that previous reports have shown a lower abundance of the genus Bacteroides in patients with atherosclerosis,^1,9,10 we focus on the Bacteroides species, B. vulgatus and B. dorei. Because these 2 species share similar 16S rRNA sequencing patterns,³² we were not able to distinguish them using our methodology. These species have been implicated to have an anti-inflammatory response³³ and were the 2 most abundant species within the genus Bacteroides in 16S rRNA gene sequencing analysis (Figure 1I). We thus designed a mouse model involving the gavage of atherosclerosis-prone mice with B. vulgatus and B. dorei to confirm the causal relationship between these species and atherosclerosis and to clarify its underlying mechanisms. We also confirmed the significantly decreased abundance of B. vulgatus and B. dorei in patients with CAD by quantitative polymerase chain reaction (Figure 1J and Table II in the online-only Data Supplement).

To determine the effect of B. vulgatus and B. dorei on the development of atherosclerosis, 6-week-old female apolipoprotein E–deficient mice were treated with live B. vulgatus and B. dorei by oral gavage 5 times per week for 10 weeks. Control mice received only a vehicle (culture medium) according to the same protocol. At 16 weeks of age, the mice were euthanized, and several analyses were performed to evaluate atherosclerosis. Compared with control mice, mice gavaged with live Bacteroides showed significantly reduced lesion size in the aortic root and en face analysis of thoracoabdominal aortas (Figure 2A and 2B) without significant difference in body weight (Figure 2F), plasma cholesterol levels (Figure 2G), or plasma glucose levels (data not shown). Furthermore, immunohistochemical staining followed by morphometric analysis of atherosclerotic lesions in the aortic sinus revealed a marked reduction in macrophage and CD4⁺ T-cell accumulation in mice gavaged with live Bacteroides compared with controls (Figure 2C and 2D). Subsequently, we performed quantitative reverse transcription–polymerase chain reaction assays to evaluate the mRNA expression of immune cell markers and chemokines/chemokine receptors in mouse atherosclerotic aortas. The mRNA expression of several proatherogenic immune cell markers and chemokines/chemokine receptors was decreased in mice gavaged with live Bacteroides (Figure 2E). These results suggested that replenishment of live B. vulgatus and B. dorei by oral gavage reduced plaque inflammation, attenuating atherosclerotic lesion formation.

Endotoxemia is known to cause systemic inflammation leading to the disruption of innate and adaptive immunity and the development of atherosclerosis.^34–36 Indeed, compared with control mice, Bacteroides-treated mice had significantly lower plasma levels of LPS (Figure 2H) and lower plasma levels of proatherogenic cytokines, such as interleukin (IL)–2, IL-4, IL-6, IL-17A, interferon-γ, and tumor necrosis factor–α (Figure 2I). Because LPS stimulates cells through TLR4 and upregulates costimulatory molecules on antigen-presenting cells,³⁷ we further examined the source of these cytokines via flow cytometry assay of splenocytes. We found that the percentages of CD11b^highF4/80^high macrophages, CD11b^highLy6G^high neutrophils, and CD11c^high dendritic cells were not significantly different between the 2 groups (Figure IIA in the online-only Data Supplement), but mice gavaged with Bacteroides exhibited decreased expression of major histocompatibility complex class II and costimulatory molecule CD86, as well as increased expression of the coinhibitory molecules programmed death ligand–1 and programmed death ligand–2 on splenic CD11c^high dendritic cells (Figure IIB in the online-only Data Supplement). Consistent with the higher phenotypic tolerogenicity of dendritic cells, the number of CD4⁺ T cells was also lower in live Bacteroides-treated mice (Figure IIC in the online-only Data Supplement). Mice gavaged with live Bacteroides had lower numbers of effector CD44^highCD62L^low CD4⁺ T cells and a significantly higher proportion of CD4⁺CD25⁺Foxp3⁺ regulatory T cells, with higher levels of intracellular CTLA4 (Figure IID and IIE in the online-only Data Supplement), indicating that immune balance was shifted toward suppression. Taken together, these data suggest that the dampening of systemic innate immune-cell activation and Th1-driven inflammation involved in the pathogenesis of atherosclerosis was caused by the live Bacteroides treatment–induced reduction in plasma LPS concentration.

We next examined the gut microbial composition. 16S rRNA gene sequencing of mouse fecal samples revealed changes in gut microbiota composition in response to gavage with live B. vulgatus and B. dorei (Figure IIIA and IIIB in the online-only Data Supplement). The Bacteroidetes/Firmicutes ratio, Gram-positive to Gram-negative strain ratio, and fecal water content did not differ significantly among the 3 groups Figure IIIC, IIID, and IIIG in the online-only Data Supplement). Copy number of 16S rDNA per gram feces was significantly increased in mice treated with live Bacteroides compared with control mice (Figure IIIE in the online-only Data Supplement). The abundance of B. vulgatus and B. dorei was considerably higher in feces from live Bacteroides-gavaged mice (Figure IIIF in the online-only Data Supplement). It is important to note that the prediction of bacterial gene functions based on 16S rRNA gene sequences using PICRUSt²⁷ revealed some functional diversity among the 3 groups in accordance with the difference in gut microbiota composition (Figure IIIH in the online-only Data Supplement). Although the expression of genes related to TMA synthesis (CutC, CutD) had no difference among the groups, the expression of genes involved in LPS biosynthesis, especially LpxA and LpxD, which are involved in lipid A biosynthesis and constitute the essential acyltransferases of the lipid A structure (Figure 3B),^38,39 was significantly decreased in mice gavaged with live Bacteroides (Figure 3A and 3C, Figure IV in the online-only Data Supplement). Therefore, we measured fecal LPS levels as an indicator of colon LPS concentrations produced by gut microbiota. It is surprising to note that the fecal LPS levels were dramatically lower in mice gavaged with live Bacteroides than in control mice (Figure 3D). Thus, endotoxemia was attenuated in live Bacteroides-gavaged mice (Figure 2H). The fecal LPS levels as well as atherosclerotic plaque in aortic sinus (P=0.74) and plasma LPS levels (P=0.93) in heat-killed Bacteroides gavaged mice were not decreased compared with control mice (P=0.92) (6 samples per group).

Given that leaky gut followed by alteration in gut microbiota composition promotes endotoxemia and atherosclerosis,³⁵ we next investigated whether live Bacteroides treatment had an effect on intestinal tight junction permeability. Mice gavaged with live Bacteroides showed lower intestinal permeability to fluorescein isothiocyanate dextran (Figure VA in the online-only Data Supplement) and significantly higher mRNA expression of tight junction gene Zo1 (Figure VB and Table III in the online-only Data Supplement), reflected in higher mean fluorescence intensity of ZO-1 in the colon (Figure VC in the online-only Data Supplement). Because LPS increases intestinal permeability in a TLR4-dependent manner,^40,41 we stimulated HT29 human colorectal adenocarcinoma cells, which showed responsiveness to LPS through TLR4, with fecal supernatant from mice. Consistent with in vivo results, we observed significantly higher expression levels of ZO-1 (mRNA and protein) in HT29 cells stimulated with fecal supernatant from live Bacteroides-treated mice than in cells stimulated with fecal supernatant from control mice (Figure VD and VE in the online-only Data Supplement). These results suggest that gut microbiota–induced changes in colon LPS concentrations may directly affect tight junction–mediated paracellular permeability.

We previously reported that atherosclerosis can be prevented by modulation of intestinal immunity.^42,43 Mice gavaged with live Bacteroides had substantially lower mRNA expression of Cd11c, Cd80, and proinflammatory cytokines in the colon. In addition, the mRNA expression of Ccr7, which is critically important for migration of antigen-presenting cells in the intestinal lamina propria to the mesenteric lymph nodes (MLNs),⁴⁴ was significantly lower in the colon of mice gavaged with Bacteroides (Figure VIA and VIB, Table III in the online-only Data Supplement). Flow-cytometric analysis of MLNs showed that Bacteroides-gavaged mice had significantly lower abundance of CD11c^high cells, as well as significantly reduced expression of TLR4, major histocompatibility complex class II, and CD80 on CD11c^high cells, whereas the number of CD4⁺ T cells and several markers on CD4⁺ T cells did not significantly change (Figure VIC and VIF in the online-only Data Supplement). Cytokine mRNA expression in MLNs tended to be lower in mice treated with live Bacteroides (Figure VIG and Table III in the online-only Data Supplement). Further, we obtained TLR4-knockdown RAW cells by small-interfering RNA transfection. We found that, on transfection with control small-interfering RNA, the secretion of proinflammatory cytokines IL-1β, IL-6, and tumor necrosis factor-α was markedly lower in RAW cells stimulated with fecal supernatant from live Bacteroides-treated mice than in RAW cells treated with fecal supernatant from control mice. RAW cells transfected with Tlr4 small-interfering RNA secreted significantly lower levels of proinflammatory cytokines than RAW cells transfected with control small-interfering RNA (Figure VIH in the online-only Data Supplement). Taken together, these findings suggest that colonic LPS concentration regulates colon inflammation via LPS/TLR4-dependent signaling.

Finally, to further investigate the relationship between CAD incidence and gut microbial production of LPS in humans, we predicted gut microbial function and measured fecal LPS levels in the same samples used for 16S rRNA gene sequencing. Although the expression of LpxA and LpxD did not differ between the patients with CAD and controls, the expression of LpxM, which yields hexa-acylated lipid A (a strong TLR4 agonist), tended to be higher in the group with CAD (Figure 4A and 4B). Notably, fecal LPS levels in patients with CAD were significantly higher than controls (Figure 4C). It is interesting to note that the abundance of genus Bacteroides tended to be negatively correlated with fecal LPS levels (r=−0.227, P=0.0838). The plot with fecal LPS levels and the abundance of B. vulgatus and B. dorei in patients with CAD (red dots) revealed a significant negative correlation between 2 parameters (r=−0.541, P=0.002) (Figure 4D). These results strongly support the idea that B. vulgatus and B. dorei regulate gut microbial production of LPS and that their activity can influence the progression of atherosclerosis in patients with CAD.