Berberine attenuates choline-induced atherosclerosis by inhibiting trimethylamine and trimethylamine-N-oxide production via manipulating the gut microbiome

To investigate whether TMAO inhibition by BBR under choline condition is mediated by gut microbial remodeling, we first explored bacterial populations in the cecal content samples from four groups of mice (chow, BBR-H, Choline, C + BBR-H) by 16S ribosomal RNA (rRNA) gene sequence analysis (Supplementary Fig. 1a). Compared to choline-fed mice, alpha diversity was significantly reduced in the C + BBR-H group as measured using the Shannon index (Fig. 1c) as well as the Sobs, Chao, and Simpson index (Supplementary Fig. 1b). The overall structural changes of the gut microbiota were then analyzed via principal coordinate analysis (PCoA). The result showed a dramatic difference in microbiota composition among the four groups at OTU level (Fig. 1d). Based on the linear discriminant analysis (LDA), some bacteria were significantly enriched in choline-fed mice, while others were enriched in BBR-treated mice (Fig. 1e). In order to investigate if the microbial composition changes correlated with the BBR-induced decrease of TMAO levels, we performed a redundancy analysis (RDA) (Fig. 1f). Most samples in the choline-fed mice (8/10) were positively correlated to TMAO levels, and a negative correlation was shown in the BBR-treated mice (10/10), suggesting that decreasing TMAO level by BBR under choline diet was directly associated with gut microbiota changes. The above results indicated that a high-choline diet coupled with BBR treatment could induce gut microbiota composition changes in C57BL/6J mice, which were directly related to the TMAO level.

To further investigate the TMAO formation in mice that were treated with choline and BBR for 4 months, we detected the d9-TMAO contents in serum after administration of a single dose of d9-choline (400 mg/kg of body weight) or d9-TMA (40 mg/kg) by intragastric gavage. First, TMAO contents in serum were measured at 0, 1, 2, 4, 6, and 8 h after administration of a single dose of choline (400 mg/kg) or TMA (40 mg/kg) in the C57BL/6J mice. It was found that peak concentrations of TMAO with the treatment of choline or TMA occurred 4 or 1 h after respective administration (Fig. 2c, d). These times points were used for subsequent experiments. Following a single dose of d9-choline, both serum d9-TMA and d9-TMAO levels decreased significantly in BBR-treated mice in the choline diet-fed group, while barely any d9-TMA and d9-TMAO were produced in Abs treated mice (Fig. 2e, f). These results suggested that BBR treatment inhibited the metabolism of choline to TMA and thus TMAO through gut microbiota in choline diet-fed ApoE KO mice.

In addition, when BBR-treated mice were given d9-TMA, d9-TMAO levels were also lower than those in the control group under choline condition, but not altered under chow condition (Fig. 2g). This suggests that BBR may inhibit the transition of TMA to TMAO in the liver of ApoE KO mice fed choline. We then determined the level of FMO3 mRNA in the liver, and consistently, the result showed that BBR decreased FMO3 expression in choline-fed ApoE KO mice but not in the chow groups (Fig. 2h).

In order to investigate the correlation between gut microbial composition remodeling and the BBR-induced phenotypic changes, we performed db-RDA analysis at OTU level. As shown in Fig. 4c, the cecal microbiota composition showed a notable correlation with both TMA and TMAO levels and with aortic plaque area. The Spearman’s correlation tests were applied at OTU level, and the results showed that more common OTUs were significantly correlated with plaque area and serum TMA/TMAO content, but not with TC, LDL-C, HDL-C, TBA or FMO3 mRNA levels (Fig. 4d). The abundance of OTU282, OTU256, OTU280, OTU78, OTU406, and OTU54 etc. were positively correlated with TMA and TMAO levels and with aortic plaque area, which was enriched in choline group and significantly decreased after BBR treatment (Fig. 4e). The abundance of OTU401, OTU245, OTU45, OTU130, OTU440, and OTU452 were negatively correlated with TMA and TMAO levels and with aortic plaque area, which was enriched after BBR treatment compared to the choline group (Fig. 4e). These results indicated that BBR played an important role in reducing serum TMA and TMAO levels and aortic plaque area by remodeling gut microbiota.

To further explore the mechanism of BBR to modulate gut microbiota, we performed metagenomic sequencing of cecal contents in 4 groups of C57BL/6J mice (chow, BBR-H, Choline, C + BBR-H, n = 21). An average of 49 million paired-end reads for each sample (ranging from 38 million to 60 million) was obtained, which led to a catalog of 3,118,451 nonredundant microbial genes. The C + BBR-H group (n = 5) had a notable reduction in gene counts (the number of genes identified per sample) compared to choline group (n = 6) (Supplementary Fig. 5a). This was consistent with the reduced community diversity of intestinal microbiota in 16S rRNA gene sequencing assays (Fig. 1c). On the basis of Bray-Curtis distances in PCoA assay, the overall structure of the gut microbiota from the four groups of mice showed significant alteration (Supplementary Fig. 5b). Sequences associated with four distinct domains were detected, and the majority microbial population were Bacteria (98.94–99.12%), with a low abundance of Eukaryota (0.31–0.48%), Archaea (0.06–0.08%) and Viruses (0.08–0.12%). Interestingly, the proportion of sequences assigned to Archaea and Viruses were significantly decreased in the C + BBR-H group compared to the choline group (Supplementary Fig. 5c–e). By LDA analysis at the genus level, the bacteria enriched in the choline group including Clostridium, Eubacterium, Lachnoclostridium, Roseburia, Odoribacter etc. and the C + BBR-H group significantly enriched Bacteroides, Prevotella, Parabacteroides, Alloprevotella etc. (Supplementary Fig. 5f), which is mostly consistent with the 16S rRNA gene sequencing results (Fig. 1e).

To explore the functional changes of the gut microbiota that might contribute to the improved phenotypic outcomes of BBR treatment, the non-redundant coding sequences were annotated based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. The result revealed that differential enrichment of microbial metabolic pathways by BBR treatment was linked to carbohydrate, amino acid and energy metabolism (Supplementary Fig. 6a). The genes for carbohydrate utilization—carbohydrate-active enzyme (CAZy) showed similar PCoA patterns as that of the total genes (Supplementary Fig. 6b), suggesting that the carbohydrate utilization changes in the gut microbiota contributed to BBR treatment-induced protection from atherosclerosis. The bile salt hydrolase (BSH), a key enzyme involved in bile acid deconjugation⁴⁰, decreased in the choline group and was restored by BBR treatment (Supplementary Fig. 6c). Accordingly, the ratio of T-β-MCA/β-MCA (conjugated/unconjugated β-muricholic acid) was significantly reduced in the BBR-treated mice compared to that in choline-fed mice, while the serum TBA level was not affected (Supplementary Fig. 6d).

To further confirm the influence of BBR on the potential microbial communities responsible for TMA generation, cecal contents from choline-fed ApoE KO mice with or without BBR treatment were collected and cultured for an ex vivo choline-TMA transformation assay. The result showed that d9-TMA production was markedly reduced and the remaining d9-choline content was significantly increased in the BBR group compared to that in the choline group at both time points, suggesting that the transformation of d9-choline to d9-TMA in the anaerobic fermentation of the cecal contents was significantly decreased in BBR group (Fig. 5b). These results demonstrated that BBR could remodel the gut microbiome for inhibition of TMA production and ultimate TMAO formation in mice, especially by decreasing the levels of cutC and cntA genes.

Furthermore, to mimic the complex physiological environments of the microbial community in the intestinal tract, four TMA-producing strains (Community A, Table 1), four non-TMA-producing strains (Community B, Table 1), and the above eight strains (Community A + B, Table 1) were mixed. Cultures were incubated anaerobically with different concentrations (8, 40, and 200 µg/ml) of BBR and d9-choline for 3 h, then d9-TMA generation was detected. While Community B group could not produce TMA from choline as expected, in Community A and Community A + B groups BBR effectively inhibited TMA production in a dose-dependent manner without dramatically influencing the bacterial growth (Fig. 6b). These results showed that BBR could suppress TMA production from choline by some human intestinal bacteria at their sub-MIC anaerobically in vitro. To further demonstrate the role of TMA-producing microbes in BBR reduced TMAO content, TMA-producing strains from Community A were intragastrically administrated into C57BL/6J mice fed choline-supplemented chow diet with or without BBR treatment for 8 days. The results showed that the Community A transplant significantly increased serum TMAO level which was decreased by BBR (Fig. 6c), providing a direct evidence of the critical role of BBR in the inhibition of gut microbe-induced TMA and TMAO formation.

We next examined the effect of BBR on gut microbes recovered from feces of C57BL/6J mice fed chow or high-choline diet for 6 weeks in an ex vivo assay. As previously reported, the CutC/D specific inhibitor bromomethylcholine (BMC) displayed a potent inhibition in TMA production (IC₅₀ = 160 nM (40 ng/ml) against human fecal cultures ex vivo)²⁶ and was used as a positive control. In our assay, TMA production was inhibited by both BBR and BMC in a dose-dependent manner, while BMC exhibited a much stronger suppression of TMA generation in gut microbes of mice under both diets (IC₅₀ = 0.4–21 ng/ml, Fig. 6d). However, BBR had a better TMA inhibitory effect in gut microbes from choline diet-fed mice (IC₅₀ = 43.5 and 185.4 µg/ml) than that from chow diet-fed mice (IC₅₀ > 512 µg/ml, Fig. 6d), implying that BBR may exert different inhibitory capacities on diverse intestinal microbiome under different conditions. Finally, to assess the inhibitory effect of BBR on choline-to-TMA transformation by human gut microbes, fresh stool samples of three volunteers were cultured with different concentrations of BBR for 6 h under anaerobic conditions. Although there are apparent differences in the conversion rate of choline to TMA by intestinal bacteria from different individuals, BBR efficiently inhibited the TMA production ex vivo at similar concentrations in choline diet-fed mice (Fig. 6e). Overall, the results from in vitro, ex vivo to in vivo studies demonstrated that BBR can suppress anaerobic production of TMA in both bacterial isolates and in the complex gut microbial community, which may be the main underlying mechanism of BBR in reducing serum TMA and TMAO levels.

C57BL/6J mice (8 weeks, female) and ApoE KO mice with a C57BL/6 genetic background (8-weeks old, female) were purchased from Vital River Laboratory Animal Technology Co., Ltd (Beijing, China), and maintained at 22 ± 2 °C, 55 ± 5% relative humidity, with a 12-h light/dark period. In the experiments, mice were fed a standard chow diet (containing 0.1% choline) or choline diet (chow diet with 1% additional choline) with or without additional BBR (Deze biotechnology, Nanjing, China) and libitum access to water. A mixture of Abs (200 mg/kg vancomycin, 400 mg/kg neomycin sulfate, 400 mg/kg metronidazole, 400 mg/kg ampicillin; Sigma-Aldrich, St. Louis, MO, USA) were given by gavage. During the experiment, C57BL/6J mice and ApoE KO mice were fed for 6 weeks and 16 weeks before sacrificed respectively. Mice were weighed every week. At the end of the experiment, mice were anesthetized with pentobarbital sodium anesthesia (50 mg/kg of body weight) to surgical procedures to minimize suffering, then the heart was perfused with PBS. At euthanasia, tissues and cecal content were harvested and stored at −80 °C. All animal experiments were carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Authority for Laboratory Animal Care of Institute of Medicinal Biotechnology.

For quantitation of TMA, TMAO, d9-TMA and d9-TMAO level, serum protein was precipitated by adding 3-fold volumes of 80% acetonitrile at room temperature for 30 min. Samples were centrifuged (14,000 × g, 15 min, 4 °C) and the supernatants were filtrated and analyzed by HPLC-MS/MS. Briefly, HPLC-MS/MS analysis was carried out using an Agilent 1100 series HPLC coupled to Agilent 6410 Triple Quadrupole mass spectrometer (Agilent Technologies, Wilmington, DE, USA) equipped with an electrospray ionization source. Separation of TMA and TMAO was performed on an XBridge™ HILIC column (150 × 2.1 mm, the internal diameter of 3.5 μm; WATERS, Milford, MA, USA) with a flex capillary XBridge™ HILIC guard column (10 × 2.1 mm, the internal diameter of 3.5 μm; WATERS). The column was eluted isostatically at a flow rate of 0.25 ml/min with a mobile phase that mixed by an equal volume proportion of acetonitrile (phase A) and water with 10 mM ammonium formate (phase B). The capillary voltage was set at +4000 V and heated to 350 °C. Analytes were monitored in positive-ion mode with multiple reaction monitoring (MRM) of precursor and characteristic product-ion transitions of TMA at m/z 60 → 44, TMAO at m/z 76 → 58, d9-TMA at m/z 69 → 49 and d9-TMAO at m/z 85 → 66. Calibration curves were prepared by spiking various concentrations of the standard into control sample for quantification of serum analytes. The standards were purchased from Sigma-Aldrich (St. Louis, MO, USA).

To evaluate the peak TMA and TMAO concentrations in serum, the 8-week-old female C57BL/6J mice were made to fast overnight and given a dose of choline (400 mg/kg) or TMA (40 mg/kg) via gastric gavage. Blood samples were collected from the retro-orbital plexus at 0, 1, 2, 4, 6, and 8 h after gavage. Serum was collected and subjected to HPLC-MS/MS detection TMA and TMAO level as mentioned above.

ApoE KO mice fed a chow diet or 1% choline diet with or without BBR (100 mg/kg) for 4 months (n = 10 per group) were administered a single dose of d9-choline (400 mg/kg) or d9-TMA (40 mg/kg). At 4 h after choline was given, or 1 h after TMA was given, the mice were euthanized and blood was collected. Serum d9-TMA and d9-TMAO levels were determined by HPLC-MS/MS as mentioned above.

Atherosclerosis lesions were quantified in aortas by staining for lipid depositions with Oil Red O (ORO; Sigma-Aldrich) as described previously⁶⁶. Briefly, for enface analysis of the aorta, aorta samples (from the proximal ascending aorta to bifurcation of the iliac artery) were dissected from mice and stored in 20% sucrose and cleaned from peri-adipose tissue after fixing in 4% paraformaldehyde for 48 h at 4 °C. Then the aortas were opened longitudinally and stained with ORO to determine the lesion area. For analyzing aortic root lipid deposition, the heart with ascending aorta was cut through below the aortic root, embedded in optimal cutting temperature (OCT) tissue freezing medium (Sakura Finetek, Torrance, CA, USA) following a snap frozen in liquid nitrogen and stored at −20 °C. The aortic sinus was sectioned by serial 7 µm slices (Leica CM1950; Leica Microsystems, Wetzlar, Germany), which were stained by ORO and pictured using Leica DM3000 Graphic Analysis System. All the pictures were analyzed and quantified using ImageJ software (U.S. National Institutes of Health, Bethesda, MD, USA). Atherosclerotic lesions were expressed as a percentage of total area, which was calculated by dividing the ORO-stained area over the total surface.

Immunohistochemical staining was performed on the cryosections from the aortic root to characterize the presence of macrophages in the aortic sinus plaque with mouse macrophage-specific antibody (Mac-3) (BD Pharmingen, CA, USA) and pictured by Leica DM3000 Graphic Analysis System. Immunofluorescence staining was used to characterize the presence of smooth muscle cells (SMCs), determined with SM22α antibody (Abcam, Cambridge, UK) and Cy3-conjugated secondary antibodies (Abcam). Nuclei were labeled using DAPI and images were recorded using Olympus IX71 microscope (Olympus, Tokyo, Japan). All the primary antibodies and secondary antibodies were used in 1:100 dilution. All the images were processed and quantified with Leica Application Suite X (LAS X) microscope software (Leica Microsystems).

Blood was collected by retro-orbital sampling. Serum TC, LDL-C, and HDL-C were enzymatically measured with the commercial cholesterol assay kit from Zhong Sheng Biotechnology (Beijing, China) according to the manufacturer’s instructions. Serum ALT and AST activity were also detected by kit from Zhong Sheng Bio-technology. All the detections were performed on Hitachi-7100 automatic analyser (Hitachi, Tokyo, Japan).

For β-muricholic acid (β-MCA) and tauro-β-muricholic acid (T-β-MCA) detection, 100 μl serum was added to 400 μl of methanol containing 0.1% formic acid. The sample was vortexed for 10 s and incubated at room temperature for 30 min. Then centrifuged at 14,000 × g for 10 min at 4 °C. The supernatant was transferred to a new tube and the organic phase was evaporated at 30 °C. The residue was reconstituted in 100 μl 50% methanol. Ten microliters of each reconstituted sample was injected for HPLC-MS/MS analysis. The chromatographic separation of bile acids was carried out on an Agilent 6410 Series Triple Quadrupole mass spectrometer. A Capcell Pak AQ C₁₈ column (4.6 × 250 mm; internal diameter of 5.0 µm, Shiseido) was used to separate the bile acids. The HPLC conditions were: 60% B for 2 min, a gradient increasing to 100% B over 40 min, 100% B for 2 min, a gradient decreasing to 60% B over 10 min, and finally, 60% B for 10 min (solvent A = 2.6 mM ammonium acetate in water, pH 6.5; solvent B = methanol). The flow rate through the column was 0.8 ml/min. The detection and quantification of β-MCA and T-β-MCA were accomplished by MRM mode with the transitions of β-MCA at m/z 391.1 → 355.2 and T-β-MCA at m/z 516.3 → 462.4. Different concentrations of standards β-MCA (Sigma) and T-β-MCA (Sigma) were added to the control sample to generate calibration allowing to quantification.

Total RNA was isolated from liver with the SV total RNA isolation system (Promega, Madison, WI, USA) and reverse transcription was preformed using the GoScript™ Reverse Transcription System (Promega) according to the manufacturer’s instructions. RNA and cDNA were quantified by Nanodrop spectrophotometry (Thermo Fisher Scientific, Waltham, MA, USA). Quantitative real-time PCR (qPCR) was performed in triplicate with the Bio-Rad CFX96 real-time system (Bio-Rad, Hercules, CA, USA) using the TaqMan^® Gene Expression Assays (Applied Biosystems, Foster City, CA, USA). The following primers and probes were used: flavin-containing monooxygenase 3 (FMO3; Fmo3; Mm01306345_m1), Glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Gapdh; Mm99999915_g1) was used as the internal standard.

Mice cecal content samples were collected and snap frozen at −80 °C. Total DNA was isolated and purified using the FastDNA® SPIN kit for Feces and the FastPrep® Instrument (MP Biomedicals, Santa Ana, CA, USA) according to the manufacturer’s instruction. The quality of DNA samples was assessed by gel electrophoresis and the concentration was measured by Nanodrop 8000 (Thermo Scientific, Wilmington, DE, USA). To investigate the microbiota community composition in the cecum, bacterial 16S ribosomal RNA (rRNA) gene sequencing assay was performed. Extracted DNA was used as a template to amplify the V4 region (for C57BL/6J mice) of the 16S rRNA gene with barcode-indexed primers 515F (5′-GTGYCAGCMGCCGCGGTAA-3′) and 806R1 (5′-GGACTACNVGGGTWTCTAAT-3′), and V3-V4 region (for ApoE KO mice) with barcode-indexed primers 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R2 (5′-GGACTACHVGGGTWTCTAAT-3′). PCR was carried out with initial denaturation at 95 °C for 3 min and subsequently 27 cycles of denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 45 s, and a final extension at 72 °C for 10 min. For each sample, three independent PCRs were performed to avoid bias. The PCR products for each sample were pooled and were extracted from a 2% agarose gel and further purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA). The amplicon libraries for high-throughput sequencing were subjected to the Illumina MiSeq platform according to standard protocols (C57BL/6J mice samples were performed by Promegene Technology Co., Ltd. (Shenzhen, China), and ApoE KO mice samples were performed by Majorbio Bio-Pharm Technology, Co., Ltd. (Shanghai, China). After removal of the barcodes and primers, the obtained sequences were clustered into the operational taxonomic units (OTUs) at a 97% similarity threshold and was aligned with SILVA128/16 S bacteria database for taxonomy information. Bioinformatics analysis of microbial diversity was analyzed on Majorbio I-Sanger Cloud Platform (www.i-sanger.com↗).

Metagenomic DNA was extracted from cecal content samples of C57BL/6J mice which were fed a chow or choline diet with or without BBR for 6 weeks by using the PowerSoil® DNA isolation kit (MoBio, Carlsbad, CA, USA) according to the manufacturer’s protocol. DNA was fragmented and sequenced in the Illumina HiSeq X Ten instrument at Promegene Technology Co., Ltd. Paired-end reads were decoded and trimmed. The remaining reads were filtered to eliminate host DNA based on the mice genome reference GRCm38.p5 and human genome reference GRCh38.p10. Clean reads were assembled into contigs using IDBA-UD software (http://i.cs.hku.hk/~alse/hkubrg/projects/idba_ud/↗) with default parameters. Genes were predicted by MetaGene (http://metagene.cb.k.u-tokyo.ac.jp↗). A non-redundant gene catalogue was constructed with CD-HIT (http://www.bioinformatics.org/cd-hit/↗), of which identity more than 95% with a minimum coverage over 90% were removed as redundancies. Gene taxonomy was annotated to NCBI NR databases by BLASTP search (version 2.2.28+, e-value cut-off 1e−5). Gene functional identification was annotated to KEGG (Kyoto Encyclopedia of Genes and Genomes, http://www.genome.jp/kegg/↗) by BLASTP search (version 2.2.28+, e-value cut-off 1e−5) and CAZy (Carbohydrate-active enzymes, http://www.cazy.org/↗) databases by hmmscan (e-value cut-off 1e−5), respectively. Principal coordinate analysis (PCoA), linear discriminant analysis (LDA) and relative abundance of domains/pathways/enzymes were analyzed on Majorbio I-Sanger Cloud Platform (www.i-sanger.com↗).

The cutC and cntA gene abundances were measured by qPCR using degenerate primers (cutC_qF: 5′-TTYGCIGGITAYCARCCNTT-3′, cutC_qR: 5′-TGNGGYTCIACRCAICCCAT-3′; cntA_qF: 5′-TAYCAYGCITGGRCITTYAARCT-3′, cntA_qR: 5′-RCAGTGRTARCAYTCSAKRTAGTTRTCRAC-3′) designed by Rath et al.⁴². Amplification was performed with 100 ng of metagenomic DNA from cecal contents of ApoE KO mice using FS Universal SYBR Green Master (Rox, Roche, Germany) according to the manufacturer’s instructions. The final primer concentration was 400 nM. An initial 95 °C for 10 min was followed by 40 cycles of denaturation at 95 °C for 30 s; annealing at 60 or 53 °C for cutC or cntA, respectively, for 30 s; and an extension step at 72 °C for 20 s. Reactions were performed on the Bio-Rad CFX96 real-time system (Bio-Rad, CA, USA). The relative fold change of abundance levels was analyzed via the 2^−ΔΔCt method normalized to 16S rRNA gene levels from each sample using degenerate primers (16S_q515F: 5′-GTGCCAGCMGCCGCGG-3′, 16S_q806R, 5′-GGACTACHVGGGTWTCTAAT-3′) with an annealing temperature of 57 °C. qPCR amplified products were analyzed by agarose gel electrophoresis (5 μl each line).

Eight bacterial strains (see Table 1) were used in the susceptibility test for BBR. The MICs of BBR were determined through the agar dilution method and broth microdilution method according to the CLSI guidelines⁴⁴. BBR was tested at concentrations of 1 to 512 µg/ml. Metronidazole was tested at concentrations of 0.125–64 µg/ml and used as positive control. B. fragilis and B. thetaiotaomicron were standard strains and served as quality controls. The MIC was defined as the lowest concentration of an agent that prevents turbidity after 48 h of anaerobic condition (10% CO₂–10% H₂–80 % N₂) in AW400SG anaerobic workstation (Electrotek, West Yorkshire, UK). The experiment was repeated three times.

TMA-producing bacteria, Clostridium asparagiforme DSM 15981, Clostridium sporogenes ATCC 19404, Anaerococcus hydrogenalis DSM 7454, and Escherichia fergusonii ATCC 35469 were inoculated in the 10 ml of Mega media⁴³ with 15 mM d9-choline at an initial concentration of 1 × 10⁶ CFU/ml. The final concentrations of BBR were 1/64, 1/16, 1/4 of the MIC of each strain. Cultures were incubated for 0, 6, 9, 12, 24 h at 37 °C in anaerobic condition (10% CO₂–10% H₂–80% N₂) using AW400SG anaerobic workstation (Electrotek). OD₆₀₀ values were measured using Ultrospec 10 Cell Density Meter (Biochrom, Cambridge, UK).

To determine the effects of BBR on TMA production in the above growth curves assay, cultures were harvested at 6 h (except for C. asparagiforme, which was harvested at 9 h).

To determine the effects of BBR on different bacteria mixture (see Table 1), four strains of TMA-producing bacteria (Community A), four of non-TMA-producing bacteria (Community B), and all eight (Community A + B) were mixed at 1 × 10⁶ CFU/ml of each strain and cocultured in 10 ml of Mega media with 15 mM d9-choline for 3 h anaerobically.

The production of d9-TMA in all culture medium samples was determined using HPLC-MS/MS according to the above-mentioned procedure.

For in vivo study, four strains of TMA-producing bacteria (Community A) were anaerobically grown in Mega Medium for 24 h at 37 °C. Each strain was adjusted to 2–8 × 10⁷ CFU per mice. The supernatant was removed by centrifugation at 12,000 × g for 5 min and the strains were resuspended in saline and were combined, intragastrically administrated into C57BL/6J mice with or without BBR treatment for 8 days under choline diet condition. The mice were then sacrificed, and serum TMA and TMAO levels were determined using HPLC-MS/MS according to the above-mentioned procedure.

Fresh cecal content samples from choline-fed ApoE KO mice were diluted 1:20 (wt/vol) in sterile Mega medium. The dilution was vortexed for 1 min and centrifuged (100 × g, 30 s) at room temperature. A 50-μl supernatant was transferred to a capped 2-ml 96-deep-well plate containing 1 ml of Mega medium supplemented with 15 mM d9-choline. Samples were incubated in duplicate for 12 h, 24 h at 37 °C in anaerobic condition (10% CO₂–10% H₂–80% N₂) using AW400SG anaerobic workstation (Electrotek). The concentration of d9-TMA in the culture medium was determined using HPLC-MS/MS according to the above-mentioned procedure. d9-choline was similarly quantified by HPLC-MS/MS using a Capcell-Pak C₁₈ ADME column (250 × 4.6 mm, the internal diameter of 5.0 µm; Shiseido, Tokyo, Japan) at a flow rate of 1 ml/min with 98% (V/V) aqueous solution of 0.2% acetic acid (phase A) and 2% (V/V) methanol (phase B). m/z 113 → 69 for d9-choline.

To investigate the dose-response curves of BBR on TMA-producing ability of different diet-fed mice, the intestinal microbiota of the feces from chow or choline diet-fed C57BL/6J mice (n = 5 for each cage, the feces of the same cage were pooled) were cultured under anaerobic condition ex vivo. 50 mg of fresh sample was vortexed to suspended in 500 μL of sterile Brucella broth (BD Biosciences) and centrifuged (200 × g, 10 s, 4 °C). Then 100 μL of supernatant was inoculated into a capped 96-deep-well plate containing 1 ml of Brucella broth (containing 5% horse blood) supplemented with 1 mM d9-choline and BBR in doses ranging from 1 to 512 µg/ml. BMC was tested at doses ranging from 0.32 ng/ml to 625 µg/ml and used as the positive control. The plate was subsequently sealed with sterile foil and incubated anaerobically for 14 h at 37 °C. The concentration of d9-TMA in the culture medium was determined using HPLC-MS/MS according to the above-mentioned procedure. Dose-response curves were fit using nonlinear regression to determine IC₅₀ values in Graphpad Prism software.

Similarly, the IC₅₀ values of BBR on TMA-production were determined in human gut microbiota from three volunteers. Human fecal samples were collected from volunteers had not received Abs within 2 months of donation and provided written informed consent. 125 mg of fresh sample was vortexed to suspended in 1 mL of sterile Mega medium and centrifuged (200 × g, 10 s, 4 °C). Then 100 μL of supernatant was added into 1 ml of Mega medium supplemented with 1 mM d9-choline and BBR in doses ranging from 1 to 512 µg/ml. After anaerobic incubation at 37 °C for 6 h, the production of d9-TMA was measured by HPLC-MS/MS and IC₅₀ values were calculated as described above-mentioned procedure.

C. sporogenes ATCC 19404 was inoculated in 100 ml of sterile Mega media supplemented with 1 mM choline chloride (to induce expression of the cut genes) and grown anaerobically at 37 °C overnight. Cells were harvested by centrifugation and resuspended in 10 mL of PBS buffer supplemented with one Mini Protease Inhibitor Tablet (Pierce). Cells were placed in a Lysing Matrix B tube (MP Biomedicals) and then were disrupted using FastPrep® Instrument (MP Biomedicals) following parameters of speed 6.0 m/s for 40 s, two cycles. The soluble protein fraction of the cell lysate was collected by centrifugation at 12,000 × g for 20 min at 4 °C. The protein concentration of the clarified cell lysates was determined by a BCA protein assay kit (Thermo Fisher Scientific Co.). Lysate concentrations were diluted to 1.5 mg/mL with PBS buffer. C. sporogenes lysate were incubated with 20 mM of d9-choline in the presence or absence of different concentrations of BBR for 13 h at 37 °C in anaerobic condition (10% CO₂–10% H₂– 80% N₂) in AW400SG anaerobic workstation (Electrotek). Sealed reaction mixtures were stopped by addition of three-fold volumes of 80% acetonitrile (containing 0.1% (v/v) formic acid). The production of d9-TMA in all reaction mixtures was determined using HPLC-MS/MS according to the above-mentioned procedure.

To determine the effects of BBR on bacteria choline uptake, C. sporogenes ATCC 19404 was inoculated in sterile Mega media until an OD₆₀₀ of around 0.5. Intact cells (1 ml) were incubated with different concentrations of d9-choline in the presence of different concentrations of BBR for 15 min at 37 °C in anaerobic condition (10% CO₂–10% H₂– 80% N₂) in AW400SG anaerobic workstation (Electrotek). Then the supernatant was collected by centrifugation and the extracellular d9-choline was quantified by HPLC-MS/MS according to the above-mentioned procedure.

Data from animal experiments were presented as means ± SEM. Statistical analysis among groups were tested by a two-tailed Student’s t-test or one-way analysis of variance, followed by Bonferroni’s correction as applicable (GraphPad Prism Software; Graph-Pad). p values are two-sided and < 0.05 denote statistical significance. Error bars denote SEM. p values for microbiota alpha diversity were conducted with the Kruskal–Wallis H test with false discovery rate (FDR) adjustment. p values for four groups’ comparation in the relative abundance of OTUs/domains/pathways/enzymes were conducted with Kruskal–Wallis H test with Tukey-kramer post hoc test.

Further information on research design is available in thelinked to this article. Nature Research Reporting Summary