Atherosclerosis amelioration by allicin in raw garlic through gut microbiota and trimethylamine-N-oxide modulation

Cardiovascular disease (CVD) is strongly associated with the gut microbiota and its metabolites, including trimethylamine-N-oxide (TMAO), formed from metaorganismal metabolism of ʟ-carnitine. Raw garlic juice, with allicin as its primary compound, exhibits considerable effects on the gut microbiota. This study validated the benefits of raw garlic juice against CVD risk via modulation of the gut microbiota and its metabolites. Allicin supplementation significantly decreased serum TMAO in ʟ-carnitine-fed C57BL/6 J mice, reduced aortic lesions, and altered the fecal microbiota in carnitine-induced, atherosclerosis-prone, apolipoprotein E-deficient (ApoE−/−) mice. In human subjects exhibiting high-TMAO production, raw garlic juice intake for a week reduced TMAO formation, improved gut microbial diversity, and increased the relative abundances of beneficial bacteria. In in vitro and ex vivo studies, raw garlic juice and allicin inhibited γ-butyrobetaine (γBB) and trimethylamine production by the gut microbiota. Thus, raw garlic juice and allicin can potentially prevent cardiovascular disease by decreasing TMAO production via gut microbiota modulation.


Introduction
Cardiovascular disease (CVD) is a primary cause of global mortality responsible for approximately onethird of all deaths worldwide. 1,2 Consumption of an unhealthy diet, excessive alcohol use, smoking, and lack of physical activity are considered traditional health risk factors for CVD. 3 Recently, an insidious CVD risk factor has been explored -the gut microbiota, now considered an endocrine organ that communicates with the human body through the gut-systemic axis and extensively modi able by foods. 4,5,6 The gut microbiota utilizes nutrients from undigested foods for growth and produces miscellaneous metabolites that regulate host-microbe homeostasis, which may be involved in the pathogenesis of cardiometabolic diseases. 7,8,9 For example, phosphatidylcholine, choline, and -carnitine from egg, dairy products, and red meat can be metabolized by speci c gut bacteria to produce trimethylamine (TMA), which is subsequently oxidized to trimethylamine N-oxide (TMAO) by hepatic avin monooxygenase. 10,11,12 Increased concentrations of blood TMAO has demonstrated a strong link to increases in major adverse cardiovascular events and all-cause mortality. 13,14 The mechanisms underlying the atherogenic and thrombogenic effects of TMAO include enhanced foam cell formation, reduced reverse cholesterol transport, and induction of platelet aggregation. 9,10,12,15 TMAO has been reported to enhance platelet aggregation and induce thrombosis in both in vitro and human studies. 16,17 The following multistep microbial pathway was recently proposed as the principal route for TMAO production through carnitine metabolism by the gut microbiota: -carnitine → γ-butyrobetaine (γBB) → TMA → TMAO. 18,19 A novel gut bacterium, Emergencia timonensis, was found to be capable of transforming γBB to TMA anaerobically and was reported to partially explain the phenotype of high-TMAO producers among human beings. 18,20 Investigations on how different food components interact with the gut microbiota have improved our understanding of modifying our dietary behaviors to achieve a health-promoting state. 8,21 Certain approaches were proposed for reducing TMAO production by the gut microbiota such as antibiotic use, fecal microbiota transplantation, and administration of TMA lyase inhibitor. 7,9,22,23,24 However, these approaches might exhibit safety concerns and require validation in human studies. Garlic (Allium sativum L.) has a long history of use as a spice in human food. 25 Traditionally, the purpose of adding garlic to meat in processed foods is not only to improve avor but also to extend shelf life. In herbal medicine, garlic has been used as a dietary therapy against cardiovascular and other metabolic diseases. 26,27,28,29 It has been widely used as a natural antibiotic on the basis of its broad-spectrum antimicrobial property. 30 Allicin is the primary compound in raw garlic puree; it is produced through the conversion of alliin by alliinase when the clove is crushed. 31 Allicin has been shown to demonstrate various antimicrobial activities in both in vitro and in vivo studies. 31,32,33 Recently, allicin was reported to exhibit a modulatory effect on the gut microbiota and reduced hepatic steatosis. 34,35 Additionally, our previous pilot study showed that allicin supplementation shaped the gut microbiota composition and reduced TMAO production by the gut microbiota in mice subjected to moderate carnitine consumption (0.02 %carnitine). 36 However, the role of raw garlic with allicin in improving cardiovascular phenotypes via gut microbiota modulation has not been fully elucidated. Therefore, in this study, we aimed to investigate the effect of allicin and raw garlic in modulating both the function and composition of the gut microbiota for cardiovascular protection. We investigated the TMAO-reducing and gut microbiota-modulating effects of allicin in raw garlic in both rodent models and humans with higher carnitine consumption. 37 We additionally compared the anti-TMAO and anti-atherosclerotic e cacies of allicin with that of the novel TMA lyase inhibitor 3,3-dimethyl-1-butanol (DMB), 24 which has been previously shown to reduce serum TMAO in carnitine-treated mice, 17 through studies in both wild-type and apolipoprotein E-de cient (ApoE −/− ) mice. Finally, we examined the impact of allicin-containing raw garlic juice on the TMAproducing activity of the human gut microbiota and speci c gut bacteria responsible for the multistep TMAO production.

Results
Allicin signi cantly reduced TMA and TMAO in mice, with less impact on the gut microbiome compared to -carnitine.
To investigate the protective effect of allicin against CVD, we subjected C57BL/6J mice to increased daily consumption of -carnitine (1.3 % in water) to examine the TMAO-reducing e cacy of allicin at a higher carnitine dosage. Additionally, we compared the TMAO-reducing effect of allicin with DMB (Fig. 1a). There was no signi cant change in the bodyweight of the mice (Supplementary Fig. 1). Serum TMA and TMAO were signi cantly reduced in the allicin-supplemented groups than in the carnitine treatment group (P = 0.011 and P = 0.0008, respectively) ( Fig. 1b-c). Additionally, serum γBB was reduced in the allicinsupplemented group (Fig. 1d). According to the microbiome analysis, shift in the overall microbial composition was driven primarily by -carnitine treatment (adonis: P < 0.001) (Fig. 1e), whereas the gut microbiota composition did not vary considerably among the carnitine, carnitine + allicin, and carnitine + DMB groups. Additionally, the observed operational taxonomic units (OTUs) and Chao1 index were decreased in the -carnitine treatment groups (P = 0.0034; P = 0.0401, respectively), whereas supplementation with allicin and DMB did not reverse the alpha diversity ( Fig. 1f and Supplementary  Fig. 2b). The family-level relative abundance bar plot and OTU-level heatmap supported that -carnitine primarily shaped the gut microbiota composition with respect to speci c taxa ( Fig. 1g and Supplementary  Fig. 2c).
Raw garlic juice reduced TMAO-producing capacity and modi ed the gut microbiome by enriching bene cial gut bacteria in humans.
On the basis of the TMAO-reducing effects of allicin in previous murine studies, we conducted a human pilot study to investigate whether raw garlic juice at an allicin-equivalent dosage could reduce the TMAOproducing capacity in the human body. We utilized an oral carnitine challenge test (OCCT) to evaluate TMAO-producing capacity in humans. 37 Fig. 3a) and were allowed to proceed to the garlic juice intervention study. We rst converted the allicin dose used in the murine study into a dose for humans and calculated the amount of blended garlic juice containing the equivalent dose of allicin for the interventional study. The blended raw garlic juice used in this pilot study contained 0.89 mg/mL allicin, and the participants consumed 55 mL of the prepared garlic juice per day for a week ( Supplementary Fig. 4). All participants received a second OCCT after the one-week intervention. For the comparison of TMAO-producing capacity before and after the garlic juice intervention, a trend of reduced plasma TMAO MAX and TMAO AUC was observed (P = 0.051; P = 0.081) (Fig. 2b). Notably, both urine TMAO MAX and TMAO AUC were signi cantly reduced after the oneweek garlic juice intervention (P = 0.0252; P = 0.0248) (Fig. 2c). The strong correlation between plasma and urine TMAO concentrations ( Supplementary Fig. 5a) supports the use of urine samples for measuring TMAO-producing capacity. Interestingly, plasma levels of γBB in the OCCT were signi cantly increased (P < 0.01) (Fig. 2d), suggesting that the bioactive compounds in garlic may prevent the microbial transformation of γBB into TMA, thus decreasing TMAO formation. Our study showed a substantial reduction in platelet aggregation following the garlic juice intervention. Additionally, the maximum amplitude of platelet aggregation tended to exhibit a positive correlation with the TMAO levels ( Supplementary Fig. 5b-c). Intervention with garlic juice did not lead to toxicity based on plasma biochemistry results (Supplementary Table 1). Only aspartate aminotransferase (AST) and alanine aminotransferase (ALT) were slightly increased but remained within normal range.
Overall, the gut microbial composition was not altered by the one-week consumption of garlic juice, as demonstrated through principal coordinate analysis (PCoA). The intra-individual dissimilarity in the gut microbiome before and after treatment was considerably lower than the inter-individual index (P < 0.01), suggesting that a relatively minor portion of the gut microbiota was altered by the garlic juice intervention ( Fig. 2e and Supplementary Fig. 6a). Regarding the alpha diversity, garlic juice intervention signi cantly increased the Shannon index (P = 0.0304) (Fig. 2f). At the family level, the fecal microbiota was altered by garlic juice; the constituent families included Acidaminococcaceae, Akkermansiaceae, Burkholderiaceae, and Erysipelotrichaceae (Fig. 2g). At the genus level, after the garlic juice intervention, the fecal microbiota of high-TMAO producers was comparatively enriched in Akkermansia, Desulfovibrio, Lachnospiraceae UCG − 008, UBA1819, and Christensenellaceae R − 7 group (Supplementary Fig. 7). The volcano plot showed that 13 taxa were signi cantly altered by the garlic-juice intervention (Fig. 2h). Interestingly, certain bene cial and anti-in ammatory gut commensal bacteria, including Faecalibacterium prausnitzii and Akkermansia spp., were signi cantly enriched after the one-week garlicjuice intervention (q < 0.05) (Fig. 2i).
Our human study demonstrated that raw garlic juice potentially reduced TMAO production and regulated the gut microbiome by enriching bene cial gut bacteria. Thus, we further evaluated the plaque-inhibiting effects of allicin and compared it with DMB in the carnitine-induced atherosclerotic ApoE −/− murine model (Fig. 3a). The ApoE −/− mice fed -carnitine water showed more advanced aortic plaques than the control group (p < 0.0001). Both allicin and DMB signi cantly reduced aortic plaques (P = 0.008 and p < 0.0001) (Fig. 3b-c). We additionally performed an isotope-labeled d 9 -carnitine challenge test to speci cally evaluate the effects of allicin on d 9 -TMA/d 9 -TMAO producing ability from the gut microbiota.
The levels of both d 9 -TMA and d 9 -TMAO in the carnitine group were considerably elevated, and their area under the curve (AUC) values were signi cantly higher compared with that in the control group (P = 0.0006 and P = 0.0008, respectively). Both serum d 9 -TMA and d 9 -TMAO were substantially reduced in mice of both carnitine + allicin and carnitine + DMB groups compared with mice fed carnitine alone. Additionally, the control + allicin group showed lower d 9 -TMA and d 9 -TMAO levels than the control group ( Fig. 3d-e). Although the carnitine + allicin treatment in the ApoE −/− mice showed signi cantly reduced atherosclerotic plaques compared with carnitine treatment alone, the reduction in TMAO-producing capacity and serum TMAO did not exhibit the expected signi cant difference. As this is the rst time that allicin was administered to ApoE −/− mice over a 16-week period, we speculate that the mouse strain and treatment duration of allicin may in uence the effects of allicin on TMAO reduction. Additionally, d 9 -γBB, an intermediate of d 9 -carnitine metabolism by gut microbiota, increased signi cantly in the carnitine group compared with that in the control group (P = 0.047), which is consistent with the result from our previous human study ( Fig. 3f and Fig. 2c). Interestingly, the carnitine + allicin and carnitine + DMB groups showed greater elevation in serum d 9 -γBB than the carnitine group, which indicates that the process of d 9 -γBB transformation into downstream TMA might be inhibited, leading to the buildup of d9-γBB.
We subsequently investigated the alteration in gut microbiome composition in different treatment groups through 16S rRNA sequencing analysis with the QIIME pipeline. Regarding the beta-diversity, the PCoA showed that carnitine water induced a signi cant microbiome shift (especially at the PCoA2 axis) that was partially reversed by treatment with allicin and DMB (adonis: P < 0.001) (Fig. 3h). Regarding the -diversity, carnitine water alone exhibited decreased observed OTUs, the Shannon index, and the Chao1 index (P = 0.0004; P < 0.001; P = 0.0063 respectively), whereas additional treatment with allicin or DMB showed favorable alpha-diversity indices (Fig. 3g). A heatmap of the gut microbiome demonstrated differences in the fecal microbiome (Fig. 3i); the carnitine + DMB group exhibited enrichment in certain bene cial bacteria, including Roseburia and Akkermansia, which have been reported to correlate inversely with atherosclerotic lesions and protect against atherosclerosis. 38,39 However, in the -carnitine + allicin group, the relative abundances of these bacteria were reduced compared to that in the -carnitine + DMB group, suggesting that allicin might exert bactericidal activity against these bene cial bacteria and limit improvement in atherosclerotic plaques compared to the -carnitine + DMB group.
Raw garlic juice inhibited microbial carnitine → γBB → TMA pathways in vitro Our study showed that raw garlic juice reduced TMAO-producing capacity through carnitine metabolism in TMAO-producing participants, C57BL/6J mice, and aortic lesion-exhibiting ApoE −/− mice. However, it is unknown whether it can speci cally inhibit microbial pathways of carnitine metabolism. Thus, we subsequently conducted an in vitro study to examine the inhibitory effects of raw garlic juice on the following pathway: carnitine → γBB → TMA (Fig. 4a). In functional assays, we used fecal samples from the high-TMAO producers and documented strain types responsible for the following pathways: carnitine → γBB and γBB → TMA.
We standardized the functional assay to examine inhibition of the following pathway: carnitine → γBB → TMA. For this purpose, we established a simpli ed model by co-culturing type strains that have been reported to metabolize carnitine to form γBB, including Proteus penneri, Escherichia fergusonii, and Edwardsiella tarda. The three co-cultured bacterial strains were inoculated in Wilkins-Chalgren (WC) broth supplemented with d 9 -carnitine. The concentration of garlic juice was titrated to a potent inhibitory level in the incubation broth. Raw garlic juice at 5 % and 10 % exhibited potent inhibition of the transformation of d 9 -carnitine into d 9 -γBB ( Fig. 4b). Single type strain experiments showed consistent results ( Supplementary Fig. 10). E. timonensis is a novel bacterium isolated from the human gut; it was recently reported to metabolize γBB to TMA. We observed that raw garlic juice exhibited potent inhibition of d 9 -TMA production by E. timonensis using d 9 -γBB as a substrate at 0.6 % garlic juice, which is a relatively low concentration of the prepared garlic juice (Fig. 4c).
To test the inhibitory effect of garlic juice in a complex microbiome community, we collected the feces from high-TMAO producing participants and cultured in WC broth supplemented with d 9 -carnitine or d 9 -γBB. WC broth supplemented with d 9 -carnitine inoculated with feces from high-TMAO producers showed that the d 9 -γBB level was the highest at 6 h while d 9 -TMA was not detected at that time point. At 12 h, the d 9 -γBB level was reduced with simultaneous formation of d 9 -TMA. Raw garlic juice at 5 % and 10 % concentrations in the culture medium showed optimal inhibition of TMA formation with moderate bacterial growth inhibition (Fig. 4d). d 9 -γBB started to be utilized to form d 9 -TMA over 6 to 12 h in broth supplemented with d 9 -γBB. Similarly, 5 % and 10 % garlic juice were determined to be an optimal concentration for preventing fecal microbiota from utilizing d 9 -γBB to form d 9 -TMA (Fig. 4e).

Discussion
Allicin supplementation in C57BL/6J mice fed a high concentration (1.3 %) of -carnitine demonstrated that the primary compound in raw garlic juice, allicin, potentially inhibited the gut microbiota-and liverderived TMA and TMAO in circulation, which is consistent with the results of our previous study, in which mice were administered a daily dosage (0.02 %) of -carnitine. 36 In the murine study, -carnitine intake was the principal factor affecting the overall gut microbiota composition compared with allicin or DMB intervention. The TMAO-reducing effect of allicin is probably based on its broad-spectrum antimicrobial activities. 31 Additionally, previous studies reported that garlic and its bioactive compounds exhibited lipid-lowering and fatty liver-protective effects via gut microbiota modulation. 35,40 Following allicin intake, it can be metabolized to various sulfur-containing bioactive compounds, such as diallyl sul de, diallyl disul de, and diallyl trisul de for modulating the microbiome composition in the gut. 41,42,43,44 Previously, in a human study, we developed the OCCT, which can robustly distinguish the low-and high-TMAO producers. 37 In this study, high-TMAO producers were de ned as individuals exhibiting OCCT plasma TMAO MAX level ≥ 10 µM 20 . There is currently a lack of studies on the effect of garlic on the gut microbiota in the human body; this study is the rst to demonstrate the impact of garlic on TMAO production by the gut microbiota. In this human study, after the intake of garlic juice for one week, both plasma and urine TMAO were reduced, suggesting that the bioactive compounds in garlic inhibited gut microbial carnitine utilization and TMA-formation ability, resulting in reduction of TMAO. The current study demonstrated that plasma and urine TMAO exhibited a high positive correlation that was consistent with the results from previous studies. 20,37 Currently, the primary genes responsible for converting carnitine to TMA by the gut microbiota remains unclear. A Rieske-type microbial CntA/B enzyme was reported to convert carnitine to TMA. 15 However, it is doubtful whether CntA/B is responsible for TMA formation in the gut anaerobic environment because oxygen is required for its reaction. 18,37 A recent study showed that -carnitine can be metabolized to the intermediate γBB, followed by transformation into TMA via a low-abundance anaerobic bacterium called E. timonensis; this could be the major pathway for TMA generation by the gut microbiota in the oxygen-limited environment. 18 This study showed an increase in plasma γBB following the garlic-juice intervention, suggesting that garlic juice might block the conversion of γBB to TMA, resulting in accumulation of γBB.
The gut microbiome analysis showed that the garlic-juice intervention increased the Shannon index, indicating that a more diverse and balanced microbial community was produced by garlic-juice consumption. The overall gut microbiome pro le for each individual was moderately affected by the garlic-juice intervention; however, the relative abundances of certain bene cial gut bacteria were increased following garlic-juice intake, including Akkermansia, Lachnospiraceae, Christensenellaceae, and Faecalibacterium prausnitzii. Akkermansia muciniphila, a mucin-degrading bacterium that resides in the mucus layer, has been demonstrated to be inversely associated with obesity and type 2 diabetes. Prebiotic administration with Akkermansia improves the metabolic pro le and reduces metabolic disorders, obesity, endotoxemia, adipose tissue in ammation, and insulin resistance. 45 A. muciniphiladerived extracellular vesicles improve gut permeability through the regulation of tight junctions, as well as decrease body weight gain and reduce glucose tolerance in high-fat diet-induced diabetic mice. 46 Furthermore, it shields against atherosclerosis by blocking metabolic endotoxemia-induced in ammation in ApoE −/− mice. 39 The garlic intervention increased the level of Faecalibacterium pranusnitzii, which has been reported to be an anti-in ammatory commensal bacterium that secretes a metabolite blocking NF-κB activation and IL-8 secretion. 47 These results suggested that the garlic-juice intervention modulates the gut microbiota of high-TMAO producers in a favorable direction. However, drinking raw garlic juice on an empty stomach may cause abdominal discomfort, and thus it should be consumed with a meal to prevent undesirable stomachache.
In ApoE −/− mice, 1.3 % -carnitine induced atherosclerosis that was ameliorated by allicin supplementation, indicating the potential anti-atherosclerotic activity of allicin. However, we observed that TMAO inhibition by allicin was more effective with short-term treatment in C57BL/6J mice than with longterm treatment in ApoE −/− mice, which suggested that long-term allicin administration could result in the resistance of TMA-producing bacteria against the anti-microbial effect of allicin. However, long-term allicin administration resulted in more differences in the overall gut microbiota in ApoE −/− mice. The allicin and DMB group with -carnitine group in the long-term study exhibited a similar gut microbiota composition. Additionally, the in vitro study suggested that garlic inhibited the formation of both d 9 -γBB and d 9 -TMA in the culture medium inoculated with the co-culture of bacteria producing γBB (P. penneri, E. fergusonii, and E. tarda) and TMA (E. timonensis), as well as inhibited TMA formation by the complex bacterial consortium of the TMAO-producing microbiota.
Here, we demonstrated that supplementation with allicin inhibited TMAO production through carnitine metabolism by the gut microbiota in both mice and humans and prevented carnitine-induced atherosclerosis in ApoE −/− mice. This was supported by in vitro assays using speci c bacteria that metabolize -carnitine to γBB and γBB to TMA (Fig. 5). These data provide valuable evidence that raw garlic, which contains allicin, shifts the gut microbiota composition and modulates the following gut microbial pathway: -carnitine → γBB → TMA. In summary, this study suggests that garlic might serve as a potential dietary intervention for CVD prevention.

Methods
Allicin preparation for the animal study.
Allicin synthesis, puri cation, and identi cation were based on our laboratory protocol. 35,36,48 The purity of allicin was more than 97 %; it was dissolved in 0.5 % carboxymethyl cellulose (CMC) and stored at −80°C before administration to the mice.

The handling of animals complied with the guidelines of the Institutional Animal Care and Use
Committee of National Taiwan University (approval number: NTU107-EL-00170 and NTU107-EL-00084). Six-week-old male C57BL/6J mice were purchased from Taiwan National Laboratory Animal Center. The mice were housed in a room with a 12 h light-dark cycle at 22 ± 2 °C after an adaptation period of 2 weeks. The mice were randomly divided into experimental groups (i) control; (ii) control + allicin; (iii) carnitine; (iv) carnitine + allicin; and (v) carnitine + DMB. For the carnitine intake group, 1.3 % -carnitine in water was supplied. 10 Allicin treatment groups were orally administered 10 mg/(kg bw day) allicin in 0.5 % CMC. 36 The DMB group was treated with 1 % DMB in water. 24 The mice were sacri ced by CO 2 asphyxiation after two weeks. Blood was collected from the major arteries of the abdomen using a syringe.
The ApoE −/− mice were initially purchased from the Jackson Laboratory (Bar Harbor, ME) and bred in our laboratory animal room according to the guidelines of the Institutional Animal Care and Use Committee of National Taiwan University. Six-week-old female ApoE −/− mice were utilized for the experiment after an adaptation period of 2 weeks. The mice were grouped and treated similarly to the short-term C57BL/6J mouse model, as described previously. The ApoE −/− mice were sacri ced after 15 weeks. Blood was collected from the major arteries of the abdomen using a syringe. Phosphate-buffered saline (PBS) was used to gently ush the aorta to remove blood clots, and the aorta was removed from the body.
Oral d 9 -carnitine challenge test in the mouse.
On the day of the sacri ce, the mice were gavaged with 200 μL (150 mM) of d 9 -carnitine, followed by serial blood collection at 0, 4, 8, and 10 h. 36 Subsequently, the mice were subjected to carbon dioxide asphyxiation. Blood was collected from the major arteries of the abdomen using a syringe at 12 h. The blood samples were centrifuged at 1,000 × g for 15 min. The supernatants were collected and stored at −80 °C before further analysis.
Oil red staining of the aorta Aortic specimens were rinsed with PBS to remove excess blood residue. Dissection was performed under the microscope using micro-dissection scissors and forceps to remove the adipose tissue. The aorta was xed with 10 % formalin and stained with oil red O dye 49 as follows: it was (1) washed with deionized water for 5 min; (2) soaked in propylene glycol for 10 min; (3) stained with oil red O dye; (4) soaked in 85 % propylene glycol for 3 min; and (5) washed twice with deionized water. We captured images of the stained aorta and analyzed the aortic lesion area using the Image J software (Version 1.8.0).

Mouse blood biochemistry analysis
The serum was extracted via centrifuging the blood at 1,000 × g for 15 min at 4 °C. We analyzed serum biochemical parameters, including total cholesterol, total triglyceride, high-density lipoprotein (HDL-c), AST, and ALT, using commercial test strips (Spotchem II reagent strips; Arkray Inc., Kyoto, Japan) in an automatic blood analyzer (Spotchem EZ). 50 High-TMAO producer screening by OCCT. This research was approved by the Research Ethics Committee of National Taiwan University Hospital (201712031RIND), and the study has been registered on ClinicalTrials.gov as NCT04545879. Healthy participants were recruited (n=9), under the following criteria: (1) age ≥ 20 years old; (2) no exposure to antibiotics, probiotics, or carnitine supplements within the previous month; (3) no history of chronic diseases including, diabetes mellitus, myasthenia gravis, chronic renal disease, hyperparathyroidism, epilepsy, and severe anemia; (4) Participants were excluded from the study if they reported recent gastrointestinal discomfort (such as abdominal pain or diarrhea).
We used the OCCT, which was previously shown to exhibit better e cacy than fasting plasma TMAO, to identify the TMAO producer phenotype. 37 All participants fasted at least 8 h before the OCCT. The participants were orally administered 1,500 mg of -carnitine (3 tablets, General Nutrition Centers, Inc., USA). The blood and urine of the participants were collected at 0, 24, 48, and 72 h after carnitine intake. The plasma and urine samples were centrifuged at 1,000 × g for 15 min and stored at −80 °C before the following analyses. Participants with plasma TMAO ≥ 10 μM after the OCCT were de ned as high-TMAO producers and allowed to proceed for the garlic juice intervention. 20 Garlic juice preparation and allicin content quanti cation Raw garlic was peeled to remove the skin; 100 g of garlic and 300 mL of water (ratio 1:3) were mixed and blended into garlic juice using a blender. After ltration, the raw garlic juice was placed into a glass bottle (55 mL for each), ready to be consumed by the subjects. The garlic juice required was prepared at once and frozen at −20 °C before use.
The allicin content in the raw garlic juice was determined using HPLC (JASCO LC-NetII/ADC/JASCO UV-2075 Plus) with Shiseido C18 (5 µm, 4.6 mm × 250 mm). The mobile phase was deionized water and methanol, and the ow rate was 1 mL/min. The gradient program was run as follows: 0-100 % methanol, 0-30 min and 100 % methanol, 10 min. The absorbance was detected at 254 nm. The allicin concentration in raw garlic juice was calculated against the allicin standard curve at concentrations of 0.2, 0.4, 0.6, 0.8, and 1.0 mg/mL.

Garlic Juice Intervention
High-TMAO producers (n=7) were asked to consume 55 mL of raw garlic juice (48 mg of allicin equivalent) once a day during dinner for one week. High-TMAO producers were suggested to consume the garlic juice with a meal to decrease stomach irritation caused by quick drinking of raw garlic juice. The high-TMAO producers were free to choose their diet with no restriction on the type of food. After one week of raw garlic juice intervention, the second OCCT was performed.

Gut microbiota inoculum preparation from human feces for in vitro study
We mixed 1 g of raw fecal sample with 10 mL of PBS supplement with 0.05 % -cysteine in a tube containing glass beads, homogenized the mixture by vortexing for 1 min in an anaerobic chamber (Whitley DG250 Workstation, Don Whitley Scienti c Limited, UK), ltered the homogenized fecal liquid through a Falcon® 100 µm cell strainer, and subsequently used the fecal ltrate for the study.

Media selection study for maximal TMA production
To investigate the suitability of the culturing media for producing maximal TMA concentration, we compared three media: gut microbiota medium (GMM), 51 MEGA, 52 and WC medium. 53 Furthermore, we modi ed these media as a carbon-reduced medium. The media were supplemented with 100 μMcarnitine. The media recipes are provided in the Supplementary Tables 2-4. Two gut microbiota inocula from the feces of high-TMAO producers were prepared by mixing 1 g of feces with 15 mL of PBS with 0.05 % cysteine and ltering the mixture through a 100 μm Falcon cell strainer, following which 100 μL of the stool ltrate was transferred to 1.9 mL of medium and subsequently incubated at 37 °C under anaerobic conditions (80 % N 2 , 10 % CO 2 , and 10 % H 2 ) for 0, 6, 12, and 24 h. The cultures were collected, their growth measured at optical density (OD) 600 nm (Spectronic Helios Gamma UV-Vis Spectrophotometer, Thermo Fisher Scienti c, UK), centrifuged at 6,000 rpm for 1 min, and the supernatant collected for measuring TMA, TMAO, γBB, carnitine, and choline concentration using LC-MS.

Gut microbiota inoculum and bacteria
The medium used in this study was WC broth supplemented with d 9 -carnitine or d 9 -γBB. We used the WC broth because it produces the highest TMA level compared with GMM, the MEGA medium, and its carbonreduced medium (Supplementary Fig. 9). The high-TMAO producer stool for the inhibition study was recollected from the participant (GJ07), which exhibited the highest TMAO MAX during the OCCT and possessed the E. timonensis 16S sequence in the sample. The bacteria used in this study include the bacteria responsible for converting -carnitine to γBB P. penneri ATCC33519, E. fergusonii ATCC35469, and E. tarda ATCC15947. 54 E. timonensis DSM101844 transforms γBB to TMA. 18 A single colony on the WC agar plate of the bacteria was inoculated into the WC broth for 24 h under anaerobic conditions before being used in the inhibition study.
Inhibition study of raw garlic juice against high-TMAO producer gut microbiome and γBB/TMA-producing bacteria Garlic juice was prepared as described above, following which it was sterilized by ltration through 0.22 μm polyvinylidene uoride membranes. We inoculated 50 μL of the gut microbiota inoculum from the high-TMAO producers, which was prepared as described above, in 1 mL of WC broth with 50 μM d 9 -carnitine or d 9 -γBB. Subsequently, the sterilized raw garlic juice was added at nal concentrations of 0, 0.3, 0.6, 1.3, 2.5, 5.0, and 10.0 % and incubated at 37 °C under anaerobic conditions (80 % N 2 , 10 % CO 2 , and 10 % H 2 ) for 0, 6, 12, and 24 h. We collected the cultures, estimated the growth OD 600 , centrifuged the cultures at 6,000 rpm for 1 min, and collected the supernatant for measuring d 9 -TMA, d 9 -γBB, and d 9carnitine using LC-MS. For evaluating the inhibition of bacteria related to CVD, the γBB-producing bacteria P. penneri, E. fergusonii, and E. tarda were mixed before the inhibition analysis. We inoculated 100 μL of the γBB-producing bacterial mixture in 1 mL of WC broth with 100 μM d 9 -carnitine, added garlic juice, cultured the bacteria, and performed the analysis as mentioned above. For analyzing the inhibition of E. timonensis, we inoculated 100 μL of the inoculum in 1 mL of WC broth with 100 μM d 9 -γBB, added the garlic juice, cultured the bacteria, and performed the analysis as mentioned above.
Sample preparation and quanti cation of TMA, TMAO, gBB, carnitine, d 9 -TMA, d 9 -TMAO, d 9 -gBB, d 9carnitine in mice For sample preparation, 50 μL of mouse serum was added to 450 μL of methanol containing isotopically labeled internal standards (d 3 -carnitine, d 9 -TMAO, and 13 C 3 -TMAO). Subsequently, the mixture was centrifuged at 12,000 × g, 4 °C for 5 min, and the supernatants were collected for LC-MS. For MS, the positive electrospray ionization mode was used with the following parameters: drying gas temperature was set at 325 °C, ow rate 7 L/min, nebulizer pressure at 45 psi, sheath gas temperature at 325 °C, 11 L/min sheath gas of ow rate, the capillary voltage at 3,500 V, and nozzle voltage at 500 V.
The mass spectrometer was con gured in multiple reaction monitoring mode. The monitored transitions for carnitine were: m/z 162.1 → 43. Sample preparation and quanti cation for TMA, TMAO, gBB, carnitine, choline, d 9 -TMA, d 9 -TMAO, d 9 -gBB, d 9 -carnitine in human and culture medium For sample preparation. 1 μL of plasma, urine, or culture medium was mixed with 199 μL of isotopically labeled internal standards (d 3 -carnitine, d 9 -TMAO, and 13 C 3 -TMAO) in 0.1 % formic acid acetonitrile solution. The solution was subsequently centrifuged at 15,000 rpm, 0 °C for 5 min. For urine sample preparation, the urine was rst centrifuged at 3,000 × g, 22 °C for 15 min. We mixed 5 μL of the 1st supernatant obtained with 45 μL of isotopically labeled internal standards (d 3 -carnitine, d 9

Platelet Preparation and Aggregometry Assays
Platelet preparation and aggregometry assays were performed according to the methods used by Zhu et al. (2016) with modi cations. 16 Whole blood was collected from study participants using 3.2 % sodium citrate anticoagulant. Platelet-rich plasma (PRP) was obtained by centrifugation at 100 × g, 22 °C for 6 min. Platelet poor plasma (PPP) was prepared by further centrifugation at 11,000 × g for 2 min. Platelets were counted using Sysmex K1000 Hematology Analyzer, and the CHRONO-LOG490-4D system was used for aggregometry assays. Platelet concentrations were adjusted to 2 × 108/mL with PPP. 5 μM ADP was used as indicated to initiate aggregation with constant stirring (600 rpm). Both PRP and PPP samples were kept at 37 °C before the aggregometry assays were performed.
Fecal DNA Extraction and 16S rRNA amplicon sequencing and sequencing analysis Fecal sample collection and genomic DNA extraction Mouse fecal samples were collected from the large intestine during a sacri ce from both the long-term study in -carnitine-induced ApoE −/− mice and short-term study in the C57BL/6J experimental mouse model; the samples were snap frozen using liquid nitrogen and stored at −80 °C before use. Human fecal samples were collected before and after the garlic juice intervention using a stool collection kit. 55  agarose gel electrophoresis was used to visualize the ampli ed products. Dual index and Illumina sequencing adapters were attached by using a Nextera XT Index Kit via PCR according to the manufacturer's instructions. PCR product cleanup was conducted using AMPure XP beads to purify the V3-V4 amplicon. The sizes of PCR products were con rmed using the Bioanalyzer DNA 1000 chip. Library quanti cation was conducted for quality control before sequencing using the Agilent Technologies 2100 Bioanalyzer. The pooled libraries were subjected to paired-end sequencing (2 × 300 bps) using the Illumina MiSeq platform.

Sequencing Analysis
The raw sequences were processed following the instructions in the 16S Bacteria/Archaea SOP v1 of the Microbiome Helper work ows (https://github.com/mlangill/microbiome_helper). 56 The paired-end reads were ltered as a sequence length of over 400 and a quality score of 90 % at a Phed score 30, followed by the removal of the chimeric sequences using VSEARCH v2. 1.2. 57 The high-quality reads were subsequently analyzed using QIIME v.1.9.1. 58 OTUs were produced using the UCLUST algorithm and a closed-referenced OTU approach against the SILVA database (version 132) with 97 % of sequence identity, followed by rarefaction. The vegan package in R was used to calculate α-diversity indices, including the observed OTUs, Shannon index, and Chao1 index, PCoA based on the Bray-Curtis distance.
We performed permutation multivariate analysis of variance (ANOVA) using distance matrices (adonis) to determine the heterogeneity of the fecal microbiota among the groups. A heatmap was plotted using the heatmap3 package.

Statistical Analysis
Data are represented as the mean ± standard error of the mean (SEM) or mean ± standard deviation. An unpaired two-tailed Student's t-test or one-way ANOVA with Tukey's range test was used to compare group means in an animal study. For the human study, data are expressed as the mean ± SEM; paired two-tailed Student's t-test was used to compare the difference between before and after garlic juice intake. The Kruskal-Wallis test, Wilcoxon signed-rank test with or without false discovery rate, unpaired/paired two-tailed Student's t-test, and one-way ANOVA with Tukey's range test were used for the analysis of the fecal microbiome data based on the data set. All statistics were analyzed using Graphpad (f) observed OTUs α-diversity indices; and (g) relative abundance of fecal microbiota at family level. Scatter plots are expressed as the mean±SD; Statistical analyses were performed using an unpaired twotailed Student's t-test control vs control + allicin group (#, p < 0.05), control vs carnitine group (*, p < 0.05; and **, p < 0.01); one-way ANOVA with Tukey's range test for comparisons, carnitine vs carnitine + allicin group ( †, p < 0.05; and † † †, p < 0.001). The relative abundance bar plot statistical analyses were performed using an unpaired Wilcoxon signed-rank test with the false discovery rate (FDR), control vs control + allicin group (#, q < 0.1); control vs carnitine group (*, q < 0.1;); carnitine vs carnitine + allicin group ( †, q < 0.1); carnitine + DMB group ( ‡, q < 0.1).

Figure 4
Garlic juice inhibited the formation of d9-TMA and d9-γBB in vitro after inoculation of TMA/γBBproducing bacteria or the feces from high-TMAO producers. (a) Schematic diagram of the bacteria responsible for metabolite d9-carnitine → d9-γBB → d9-TMA; (b) inhibitory effect of garlic juice on carnitine → γBB bacteria (co-culture of Proteus penneri, Escherichia fergusonii, and Edwardsiella tarda) in WC medium supplemented with d9-carnitine; (c) inhibitory effect of garlic juice on the bacteria responsible for converting γBB to TMA (Emergencia timonensis) in WC medium supplemented with d9-γBB. (d) Inhibitory effect of garlic juice on the high-TMAO-producing gut microbiome in WC medium supplemented with d9-carnitine and (e) d9-γBB.

Figure 5
The role of allicin and raw garlic juice on CVD prevention and atherosclerosis amelioration through gut microbiota and TMAO modulation. (1) Allicin decreased gut microbiota and host-derived TMA and TMAO in mice supplemented with -carnitine in drinking water; (2) Raw garlic juice reduced plasma and urine TMAO and increased bene cial gut bacteria in high-TMAO producing humans; (3) Allicin ameliorated aortic lesions by reducing TMA/TMAO production and altered the gut microbiome in -carnitine-induced atherosclerosis ApoE−/− mice; (4) Raw garlic juice inhibited microbial pathway of carnitine → γBB → TMA conversion in speci c bacteria and high-TMAO producer's gut microbiome.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Supplementarynpjbio lmsandmicrobiomes.docx