DIETARY TRIMETHYLAMINES, THE GUT MICROBIOTA, AND ATHEROSCLEROSIS ROBERT ALDEN KOETH. Submitted in partial fulfillment of the requirements

Transcription

1 DIETARY TRIMETHYLAMINES, THE GUT MICROBIOTA, AND ATHEROSCLEROSIS By ROBERT ALDEN KOETH Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Dissertation Adviser: Stanley L. Hazen, M.D., Ph.D. Department of Pathology CASE WESTERN RESERVE UNIVERSITY August, 2013

2 CASE WESTERN RESERVE UNIVERSITY SCHOOL OF GRADUATE STUDIES We hereby approve the thesis/dissertation of Robert Alden Koeth candidate for the Ph.D. degree *. (signed) Alan D. Levine, Ph.D. (chair of the committee) Stanley L. Hazen, M.D., Ph.D. Jonathan D. Smith, Ph.D. George R. Dubyak, Ph.D. Clive R. Hamlin, Ph.D. (date) 04/16/2013 *We also certify that written approval has been obtained for any proprietary material contained therein.

3 TABLE OF CONTENTS LIST OF TABLES... 9 LIST OF FIGURES ACKNOWLEDGEMENTS ABSTRACT CHAPTER1: Introduction to Dietary Trimethylamines, the Gut Microbiota, and Atherosclerosis Cardiovascular Disease and Atherosclerosis A History of the Gut Microbiota Location and Composition of the Gut Microbiota Normal Functions of the Gut Microbiota The Relationship Between the Gut Microbiota and Disease Gut Microbiota Mediated Metabolism of Phosphatidylcholine Promotes Cardiovascular Disease CHAPTER 2: Intestinal Microbiota Metabolism of L-Carnitine, a Nutrient in Red Meat, Promotes Atherosclerosis Authors Abstract Introduction Results Metabolomic studies link L-carnitine with CVD

4 Gut microbiota plays an obligatory role in forming TMAO from L- carnitine in humans Vegans and vegetarians produce substantially less TMAO from dietary L-carnitine Plasma TMAO levels significantly associate with specific human gut microbial taxa TMAO production from dietary L-carnitine is an inducible trait TMA / TMAO production associates with specific mouse gut microbial taxa Plasma levels of L-carnitine associate with CVD Dietary L-carnitine in mice promotes atherosclerosis in a gut microbiota dependent manner Gut microbiota dependent formation of TMAO inhibits reverse cholesterol transport TMAO promotes significant alterations in cholesterol and sterol metabolism in multiple compartments in vivo Discussion Acknowledgements Methods Materials and general procedures Research subjects

5 General statistics Metabolomics study Identification of L-carnitine and d9-carnitine preparation Quantification of TMAO, TMA, and L-carnitine Human microbiota analyses Mouse microbiota analysis Aortic root lesion quantification Human L-carnitine challenge test and d3-l-carnitine preparation Germ-free mice and conventionalization studies Metabolic challenges in mice Preparation of bone marrow derived macrophages for reverse cholesterol transport studies Reverse cholesterol transport studies Cholesterol absorption studies Bile acid pool size and composition Cholesterol efflux studies Effect of TMAO on macrophage cholesterol biosynthesis, inflammatory genes, and desmosterol levels RNA preparation and real time PCR analysis

6 CHAPTER 3: Carnitine, a Nutrient Found in Red Meat and a Frequent Additive by the Nutritional Supplement Industry, Can Induce the Human Gut Microbiota to Produce Proatherogenic TMAO Authors Intro Methods Results Comment CHAPTER 4: Intestinal Microbial Metabolism of Phosphatidylcholine and Cardiac Risk Authors Abstract Introduction Results Role of intestinal microbiota in metabolism of dietary phosphatidylcholine Correlation of plasma levels of trimethylamine-n-oxide with major adverse cardiovascular events Correlation of trimethylamine-n-oxide levels with risk in low-risk subgroups Discussion

7 Acknowledgements Methods Study patients and design Dietary phosphatidylcholine challenge Measurements of choline metabolites Statistical analysis for the clinical outcomes study CHAPTER 5: Intestinal Microbiota Metabolism of L-Carnitine, a Nutrient in Red Meat, Produces TMAO Via Generation of an Intermediate Gut Microbiota Metabolite γ-butyrobetaine Authors Abstract Introduction Results Gut microbiota metabolism of L-carnitine produces γbb γbb produces TMA/TMAO in a gut microbiota dependent manner TMA formation occurs in the cecum and γbb is the dominant gut microbiota product of L-carnitine gut microbiota metabolism Metabolism of γbb by the gut microbiota to TMA/TMAO promotes atherosclerosis Metabolism of γbb from L-carnitine is an inducible trait

8 γbb associates with a microbiome composition that differs from TMA/TMAO formation TMAO production from γbb associates with microbiome composition Mice on a γbb diet have significant decreased liver expression of Cyp7a1, but not Cyp27a Discussion Methods Materials and general procedures Mouse challenge and atherosclerosis studies Mouse microbiome studies d9-γ-butyrobetaine chloride preparation Quantification of TMAO, TMA, a γbb, and L-carnitine In vitro mouse cecum study RNA preparation and real time PCR analysis General Statistics CHAPTER 6: γ-butyrobetaine is a Gut Microbiota Dependent Product of L- Carnitine Introduction Results γbb associates with CVD prevalence

9 γbb is associated with MACE, but not after TMAO adjustment γbb is produced from carnitine in a gut microbiota dependent manner in humans TMAO is the major gut microbiota metabolite of L-carnitine in humans γbb does not associate with a omnivorous diet Red meat is an exogenous source of γbb, but is found at lower concentrations compared to carnitine Discussion Methods Research subjects Human L-carnitine challenge test Quantification of L-carnitine, γbb, and TMAO in plasma samples γ-butyrobetaine quantification in meat samples General statistics Chapter 7: Transcrotonobetaine, a Gut Microbiota Metabolite of Carnitine Metabolism, Promotes Atherosclerosis Introduction Results TC is a gut microbiota dependent product of L-carnitine TC is an abundant gut microbiota metabolite of L-carnitine in mice

10 TC produces both γ-butyrobetaine and TMA/TMAO in a gut microbiota dependent manner TC independently associates with cardiovascular disease, but not after multivariate model adjustment with TMAO Dietary TC promotion of atherosclerosis is gut microbiota-dependent manner Discussion Methods Materials and general procedures Research subjects Mouse challenge and atherosclerosis studies d9-tc and native TC preparation Quantification of TC, TMAO, TMA, γbb, and L-carnitine In vitro mouse cecum study General Statistics Chapter 8: Summary, Conclusions, and Future Directions Clinical implications A hypothetical role for the gut microbiota and TMAO in other disease states Summary REFERENCES

11 LIST OF TABLES Supp. Table 2-1 Characteristics of analyte m/z = 162 determined in LC/MS positive ion mode from plasma samples used in Validation and Learning cohorts (n = 150) of metabolomics study from Wang et. al., Nature, Supp. Table 2-2 Supp. Table 2-3 Supp. Table 2-4 Supp. Table 2-5 Subject characteristics, demographics, and laboratory values in the whole cohort (n = 2595), and across quartiles of plasma carnitine 108 Plasma levels of triglycerides, cholesterol, glucose, and insulin from mice on normal chow vs. carnitine supplemented diet 109 Liver levels of triglycerides and total cholesterol in mice on normal chow versus carnitine supplemented diet 110 Plasma levels of triglycerides, cholesterol, and glucose from mice on normal chow, carnitine, choline, and TMAO supplemented diets during the in vivo RCT studies 111 Table 4-1 Baseline characteristics 154 Table 4-2 Supp. Table 4-1 Table 5-1 Table 6-1 Table 6-2 Table 7-1 Table 7-2 Unadjusted and adjusted hazard ratio for risks of MACE at 3-years stratified by quartile levels of TMAO 155 Baseline characteristics of cohort according to TMAO quartiles values expressed in mean ± standard deviation or median (interquartile range) 159 Plasma and liver lipid levels in C57BL/6J, Apoe-/- female mice used in γbb atherosclerosis study 186 Baseline clinical characteristics of n = 1445 Genebank subjects used in analyses with γbb 217 Quantification of carnitine and γbb in beef, lamb, chicken, and perch samples 218 Baseline clinical characteristics of n = 836 Genebank subjects used in analyses with TC 238 Plasma levels of triglycerides, cholesterol, and glucose from mice on normal chow vs. transcrotonobetaine supplemented diet 239 9

12 LIST OF FIGURES Figure 1-1 Scheme of gut microbiota dependent metabolism of dietary PC and atherosclerosis 40 Figure 1-2 Metabolomics studies scheme and correlations 41 Figure 1-3 Production of TMAO from PC is gut flora dependent 42 Figure 1-4 Choline, TMAO and betaine are associated with CVD in humans 43 Figure 1-5 Dietary choline or TMAO enhances atherosclerosis 44 Figure 1-6 Hepatic FMOs associate with atherosclerosis 45 Figure 1-7 Figure 2-1 Figure 2-2 Figure 2-3 Dietary choline enhances atherosclerosis in a gut flora dependent manner 46 TMAO production from carnitine is a microbiota dependent process in humans 101 The formation of TMAO from ingested L-carnitine is negligible in vegans, and fecal microbiota composition associates with plasma TMAO concentrations 102 The metabolism of carnitine to TMAO is an inducible trait and associates with microbiota composition 103 Figure 2-4 Relation between plasma carnitine and CVD risks 104 Figure 2-5 Dietary carnitine accelerates atherosclerosis and inhibits reverse cholesterol transport in a microbiota dependent fashion 105 Figure 2-6 Effect of TMAO on cholesterol and sterol metabolism 106 Supp. Figure 2-1 Supp. Figure 2-2 Mass spectrometry analyses identify unknown plasma analyte at retention time of 5.1 min and m/z = 162 as carnitine 112 LC/MS/MS analysis of synthetic heavy isotope standard d9(trimethyl)carnitine spiked into human plasma sample confirms unknown peak at 5.10 min (m/z = 162) is carnitine

13 Supp. Figure 2-3 Supp. Figure 2-4 Standard curves for LC/MS/MS quantification of carnitine and d3-(methyl)-carnitine in plasma matrix 114 LC/MS/MS analyses of a subject s 24 hr urine samples demonstrate an obligatory role for gut microbiota in production of TMAO from carnitine 115 Supp. Figure 2-5 Plasma levels of carnitine and TMAO following carnitine challenge in a typical omnivorous subject 116 Supp. Figure 2-6 Supp. Figure 2-7 Supp. Figure 2-8 Supp. Figure 2-9 Plasma levels of carnitine and d3-carnitine following carnitine challenge (steak and d3-carnitine) in typical omnivore with frequent red meat dietary history and a vegan subject 117 Plasma levels of d3-carnitine following d3-carnitine challenge (no steak) in omnivorous (n = 5) versus vegan subjects (n = 5) 118 Human fecal microbiota taxa associate with plasma TMAO 119 Demonstration of an obligatory role of the commensal gut microbiota of mice in the production of TMA and TMAO from oral carnitine in germ-free and conventionalized mice 120 Supp. Figure 2-10 Demonstration of an obligatory role of commensal gut microbiota of mice in the production of TMA and TMAO from oral carnitine 121 Supp. Figure 2-11 Analysis of mouse plasma TMA and TMAO concentrations and gut microbiome composition can distinguish dietary status 122 Supp. Figure 2-12 Haematoxylin/eosin (H/E) and oil-red-o stained liver sections 123 Supp. Figure 2-13 Arginine transport in the presence of 100 µm trimethylamine-containing compounds 124 Supp. Figure 2-14 Expression levels of cholesterol synthesis enzymes, transporters, and inflammatory genes in the presence or absence of TMAO 125 Supp. Figure 2-15 Effect of TMAO on desmosterol levels in media of 11

14 cultured mouse peritoneal macrophages in the presence of increasing cholesterol and acetylated LDL (AcLDL) concentrations 126 Supp. Figure 2-16 Plasma concentrations of TMAO in mice undergoing in vivo reverse cholesterol transport studies 127 Supp. Figure 2-17 [ 14 C] Cholesterol recovered from mice on normal chow vs. TMAO diet enrolled in in vivo reverse cholesterol transport studies 128 Supp. Figure 2-18 Effect of TMAO on mouse peritoneal macrophages 129 Supp. Figure 2-19 Effect of TMAO on cultured macrophage cholesterol efflux 130 Supp. Figure 2-20 Liver expression of cholesterol transporters in mice examined during reverse cholesterol transport studies 131 Supp. Figure 2-21 Western blot analysis of liver scavenger receptor B1 (Srb1) expression 132 Supp. Figure 2-22 Small intestines expression profile of bile acid transporters in mice 133 Supp. Figure 2-23 Small intestines expression profile of cholesterol transporters in mice 134 Figure 3-1 Figure 4-1 Figure 4-2 Figure 4-3 Supp. Figure 4-1 Carnitine supplementation can induce the gut microbiota 138 Human plasma levels of phosphatidylcholine Metabolites (TMAO, choline, betaine) after oral ingestion of two hard-boiled eggs and d9- Phosphatidylcholine before and after antibiotics 156 Kaplan-Meier estimates of long-term major adverse cardiac events, according to TMAO Quartiles 157 Pathways linking dietary phosphatidylcholine, intestinal microflora (gut flora), and incident adverse cardiovascular events 158 Human plasma levels of phosphatidylcholine metabolites (TMAO, choline, betaine) after oral ingestion of two hard-boiled eeggs and d9-12

15 phosphatidylcholine before and after antibiotics 160 Supp. Figure 4-2 Supp. Figure 4-3 Figure 5-1 Figure 5-2 Figure 5-3 Figure 5-4 Human 24-hour urine levels of TMAO after oral ingestion of two hard-boiled eggs and d9- phosphatidylcholine before and after antibiotics 161 Risks of major adverse cardiac events (MACE) among patient subgroups, according to baseline TMAO levels 162 γbb is produced as a major gut microbiota metabolite of L-carnitine 187 γbb is produced from L-carnitine in a gut microbiota dependent manner 188 TMA/TMAO is a gut a microbiota dependent product of γbb metabolism 189 Confirmatory studies that TMA/TMAO is a gut a microbiota dependent product of γbb metabolism 190 Figure 5-5 γbb is the dominant gut microbiota metabolite of L- carnitine and is metabolized to TMA at a great equamolar capacity than L-carnitine 191 Figure 5-6 Figure 5-7 γbb promotes atherosclerosis in a gut microbiota dependent manner 192 Plasma trimethylamine concentrations of C57BL/6J, Apoe-/- female mice used in γbb atherosclerosis study 193 Figure 5-8 γbb production from L-carnitine is an inducible trait 194 Figure 5-9 Figure 5-10 Figure 5-11 Figure 5-12 γbb production from L-carnitine associates with microbiome composition 195 γbb production from L-carnitine and microbiome composition associate with dietary status 196 TMA/TMAO production from γbb associates with microbiome composition 197 TMAO production from γbb and microbiome composition associate with dietary status 198 Figure 5-13 Liver Expression of Bile acid enzymes

16 Figure 5-14 Figure 6-1 Scheme of endogenous and exogenous γbb production 200 Relationship between plasma γbb and CVD prevalence 219 Figure 6-2 Relationship between plasma γbb and CVD risks 220 Figure 6-3 Figure 6-4 Figure 6-5 Figure 6-6 Figure 7-1 Relationship between plasma γbb, plasma TMAO, and CVD risks 221 γbb production from carnitine is a gut microbiota dependent process in humans 222 TMAO is the major gut microbiota metabolite in human carnitine catabolism 223 The formation of γbb from ingested L-carnitine is similar in vegans and vegetarians compared to omnivores 224 Demonstration of an obligatory role of the commensal gut microbiota of mice in the production of TC from oral carnitine in germ-free and conventionalized mice 240 Figure 7-2 TC is the an abundant gut microbiota metabolite of L- carnitine 241 Figure 7-3 Proposed scheme of carnitine metabolism 242 Figure 7-4 Figure 7-5 Figure 7-6 Figure 7-7 Demonstration of an obligatory role of commensal gut microbiota of mice in the production of TMA,TMAO, and γ-butyrobetaine from oral TC challenge 243 Plasma TC is associated with MACE over a 3-year period 244 Plasma TC is not associated with MACE over a 3-year period after adjustment with other CVD risk factors in n = 836 subjects. 245 Dietary TC gut microbiota metabolism accelerates atherosclerosis 246 Figure 7-8 Plasma analytes from TC atherosclerosis study

17 Figure 8-1 Relationship of dietary trimethylamines, atherosclerosis, and homocysteine formation

18 ACKNOWLEDGEMENTS There are many people over the last several years that have been critical for my personal and professional development. I would like to thank first and foremost my wife, Kim, for her consummate support, advice, and patience through the trials and tribulations of the M.D./Ph.D. process. Thanks goes to my newborn son Alden Scott Koeth for helping to bring perspective to this process. I would like to also thank my friends and family for their support. I wish to acknowledge the mentors both informal and formal that have helped encourage and shape my scientific endeavors. Assistance provided by Hazen laboratory members, the Cleveland Clinic Cardiovascular Prevention Research Laboratories, and members of the Lerner Research Institute was also greatly appreciated. Specifically, I would like to acknowledge Bruce S. Levison, Zeneng Wang, and Jennifer Buffa for providing crucial training and scientific support. Thanks also go all collaborators for helping to add valuable scientific insight and crucial experimental data to my studies. Thank you to the Department of Pathology of Case Western Reserve University for support and the opportunity to pursue a Ph.D. I would like to offer my special thanks to my committee members, Drs. George R. Dubyak, Jonathan D. Smith, and Clive R. Hamlin, and my committee chair, Dr. Alan D. Levine. 16

19 Thanks also go to the Cleveland Clinic Lerner College of Medicine and the Medical Scientist Training Program of Case Western Reserve University for giving me this opportunity. Last, but certainly not least, I would like to acknowledge my graduate student mentor Dr. Stanley L. Hazen who challenged me both professionally and personally to develop skills essential to becoming a successful physician scientist. 17

20 Dietary Trimethylamines, the Gut Microbiota, and Atherosclerosis Abstract by ROBERT ALDEN KOETH The gut microbiota has critical roles in mammalian physiological processes and has been increasingly recognized to be a culprit in disease pathogenesis. We recently identified a pathway that links the consumption of dietary phosphatidylcholine, the major dietary source of choline, the gut microbiota, and atherosclerosis. Choline, a trimethylamine compound, is metabolized by the gut microbiota to produce an intermediate compound known as trimethylamine (TMA). TMA is oxidized by hepatic flavin monooxygenase 3 (FMO3) to form the proatherogenic metabolite trimethyl amine N-oxide (TMAO). The recognition that the gut mediated metabolism of choline to TMAO promoted atherosclerosis raised the possibility that carnitine, another dietary trimethylamine found in red meat, could contribute to TMAO formation. Studies in mice and humans confirm that the formation of TMAO from carnitine is gut microbiota dependent. Interestingly, omnivorous subjects have a greater capacity to metabolize TMAO from carnitine than vegans/vegetarians and demonstrate significant differences in gut microbiota composition. Plasma TMAO levels are independently associated with prospective major adverse cardiovascular events (death, MI, stroke) and can 18

21 promote atherosclerosis by causing dysfunction in forward and reverse cholesterol transport. Subsequent studies of carnitine metabolism by the gut microbiota demonstrate the production of two other gut microbiota metabolites, γ- butyrobetaine (γbb) and transcrotonobetaine (TC). γbb is the dominant gut microbiota metabolite of carnitine in mice and is an intermediate in the gut microbiota dependent metabolism of carnitine to TMAO. Remarkably, two separate bacterial taxa in the gut microbiota associate with the two step metabolism of carnitine to TMAO suggesting distinct populations in the gut microbiota working in cooperation. Although humans also have the capacity to generate TMAO from carnitine in a gut microbiota dependent manner, γbb is produced at a lesser amount than TMAO. Additionally, plasma γbb had no association with dietary status (omnivore vs. vegan/vegetarian). This data suggests that in humans the direct metabolism of carnitine to TMA/TMAO is the major gut microbiota mediated pathway of carnitine metabolism. Interestingly, a minor gut microbiota metabolite of carnitine, transcrotonobetaine, also promotes atherosclerosis in a gut microbiota dependent manner. Together these data provide a previously unrecognized link between the consumption of dietary trimethylamines, the gut microbiota, and atherosclerosis. 19

22 CHAPTER 1: Introduction to Dietary Trimethylamines, the Gut Microbiota, and Atherosclerosis Cardiovascular Disease and Atherosclerosis Cardiovascular disease (CVD) is the leading cause of morbidity and mortality in the developed world 1 and atherosclerotic sequelae accounts for greater than 50% of all CVD deaths 2. Atherosclerosis is a chronic systemic inflammatory disease characterized by the evolution and accumulation of large lipid laden plaques in the artery wall. Myocardial infarction (MI) remains one of the most deadly sequelae of atherosclerotic disease and is largely precipitated by rupture of the thin fibrous cap or endothelial cell erosion of the atherosclerotic lesion 2. The resulting thrombus formed within the coronary artery completely or partially occludes the vessel causing ischemia and eventual death of the myocardium. Cholesterol is the major culprit lipid associated with the progression of atherosclerosis. The foundation of this association is based on a combination of a vast number of clinical outcome studies and cellular based studies demonstrating the link between hypercholesterolemia and atheroma formation 2-5. Still today, measurement of cholesterol is the major mode of risk stratifying subjects for CVD, and intervention in lipid metabolism with pharmacological agents, like statins for example, remains a major preventive treatment. Overall, atherosclerotic disease can be generally viewed as a balance between forward and reverse cholesterol transport (RCT). Forward cholesterol transport is 20

23 characterized by the accumulation of lipid in cells, most notably, in the macrophage. The precipitating event of atherosclerotic plaque formation involves physiological stress to the vessel endothelium resulting in the adherence and transmigration of macrophages into the vessel intima 2. Within the intima macrophages engulf large amounts of lipids (principally cholesterol) creating a characteristic foam cell. Foam cells comprise the majority of fatty streaks, the earliest atherosclerotic lesions, which develop early in life. Indeed, a recent evaluation of coronary artery disease (CAD) in young adults and teenagers demonstrated a high prevalence of early atherosclerotic disease 6. Foam cells are not only involved in the initiation, but also the progression, and sequelae of atherosclerotic disease suggesting a central role of the macrophage foam cell in atherosclerosis 1,2,7. RCT is defined as the net movement of cholesterol from peripheral sources to the feces for elimination and is believed to be one of the major mechanisms by which high density lipoprotein (HDL) mediates its antiatherogenic effect. The vast majority of peripheral cells do not have the capacity to catabolize cholesterol making transport the only major way to eliminate cholesterol from the cell 8. RCT consists of movement of cholesterol through multiple compartments in the body. Transport begins when cholesterol is made available to move from cellular sources (e.g. macrophage foam cells) to apolipoprotein A1 (apoa-1) containing molecules (native apoa-1 and HDL). Mature HDL interacts with scavenger receptor b1 (SR-B1) and is taken up by the liver for further metabolism. In the 21

24 canonical pathway, cholesterol is secreted into bile or metabolized into bile acids for excretion into the gastrointestinal (GI) tract. Bile acids and cholesterol are reabsorbed in an enteroheptic pathway for return to the liver or ultimately eliminated in the feces. RCT was coined several decades ago by Glomset, but more recently, there has been an increased recognition of the importance of the RCT pathway in cholesterol metabolism 9. This is large part due to the development of an in vivo RCT assay by Rader and colleagues, and the recognition that HDL function may be a better measure of atheroprotection than absolute HDL cholesterol concentrations 8,10. Indeed, studies of macrophage RCT rates are associated atherosclerosis burden in mice 11. Mice deficient in apoa-1, the major lipoprotein of HDL, impairs RCT and apoa-1 overexpression increases total RCT 12,13. These data are consistent with the atheroprotective role of apoa-1 and HDL 8. Macrophage cholesterol transporters ATP-binding cassette sub-family G member 1(ABCG1) and ATP binding cassette transporter A1 (ABCA1) are critical for removing cholesterol from macrophage foam cells and the absence of these cholesterol transporters in macrophage foam cells impairs RCT 14. Finally, inflammation, a major contributor to the pathogenesis of atherosclerosis, has been shown to impair the overall RCT pathway 15. Together these data suggest important roles for net forward and reverse cholesterol transport in the atherosclerotic disease process. A History of the Gut Microbiota The human microbiota consists of trillions of bacteria that form a complex symbiotic relationship with the host. Anywhere from 500-1,000 different bacterial 22

25 species live in a human microbiome, and the total microbial cells are quantitatively approximately 10 fold the total number of eukaryotic cells in the host 16,17. The importance of the gut was first recognized by Hippocrates who noted that All diseases begin in the gut and that death sits in the bowels 18. In modern medicine, the focus has traditionally been on the pathologic invasion of the gut by various bacteria and viruses. Indeed, at the turn of the 20 th century diarrhea and gastroenteritis were the 3 rd leading causes of death accounting for almost 10% of all deaths in the United States 19. The advent of antimicrobial agents, development of vaccinations, improved nutrition, advancement of epidemiology, and recognition of the importance of sanitation and hygiene theory have effectively seen this and other infectious diseases be eclipsed by chronic disease as the major challenge in modern medicine 19. The emergence of chronic disease and, more recently, the increased recognition for the capacity of the commensal human microbiota to influence human physiology has led to renewed interest of the gut in human health and disease. Location and Composition of the Gut Microbiota The gut microbiota is heterogeneous in its composition and quantity throughout the GI tract. Overall, the amount and biodiversity of bacteria per gram of content in the gut increases from proximal to distal ends of the GI tract 20. The relative lack of stable colonization of a microbiome in the proximal GI tract has been attributed to the pulsatile contractions of the small bowel and the harsh environmental conditions in the GI lumen (bile acids, HCl, and pancreatic 23

26 enzymes) 16. In contrast, the large bowel contains a diversity of gut microbes culminating with to per gram of intestinal luminal contents 20,21. The vast majority of bacteria that inhabit the gut are anaerobes or facultative anaerobes with anaerobes dominating overall 22. Currently, anywhere from 300-1,000 different bacterial species are believed to colonize the human GI tract, but with advances in sequence technology this number could mushroom. Indeed some recent analyses have suggested that as many as 35,000 different species of bacteria may in fact colonize the human GI tract 23. Humans have a sterile gut in utero and begin to acquire a microbiome at time of birth. Passage through the vaginal canal exposes infants to both maternal vaginal and fecal flora initiating the development of the gut microbiome 24,25. One study suggested that the initial makeup of infant GI microbiota and maternal vaginal flora are closely aligned immediately after vaginal delivery 26. However, this appears to be a transient establishment as infants quantitatively have 10 9 CFU/g feces of enteric bacteria established by the end of the first day of life that expands to CFU/g feces of enteric bacteria by the first month of age 27. The initiation and establishment of the gut microbiome is the culmination of a complex interplay between a number of extrinsic and intrinsic factors including: mode of birth (delivery vs. cesarean section), maternal flora, host genetics, diet, exposure to antimicrobials, bile acids, peristalsis, drugs, host immunity, intestinal luminal ph, intermicrobial interactions, and the bacterial load in the environment 24. As a result, the infant microbiome remains immature and greatly variable becoming 24

27 more similar in quantity and composition to an adult microbiome by the first year of age 20. Normal Functions of the Gut Microbiota The gut microbiome has a mutualistic relationship with the mammalian host that has developed through millions of years of coevolution. The establishment of the microbiome from birth becomes critical for the normal maturation and development of GI structure and function. Insight into the importance of the gut microbiome and normal GI development has been demonstrated by mouse germ free (mice lacking any microbiome; GF) animals studies. Structurally, these animals have impaired peristalsis, a reduction in the villous capillary network, and decreased overall surface area 20. GF animals characteristically develop physically enlarged cecums that can often predispose the animal to both reproductive and GI dysfunction 20. The GI tract is the largest exposed mucosal surface and contains the largest number of immunocompetent cells in the human body. The GI tract therefore serves as a critical component of the development and maintenance of a normal immune system 16. Insights into the importance of the gut microbiota in the development of the immune system have been garnered by studies in gnotobiotic (GF) rodents. GF mice have less lymphoid tissue, lower numbers of immunocompetent cells, decreased expression of immune receptors such as Toll like receptors (TLRs), and overall decreased circulating immunoglobulin 25

28 concentrations compared to conventional mice 20,28. The subsequent rapid expansion and development of the GI immune system in GF mice upon exposure to luminal microbes suggests important roles of the gut microbiota in the development of both GI and systemic immunity 20,28. These defects most notably result in increased susceptibility to pathological infection 20,28. Germ free guinea pigs challenged with the gram-negative enteric pathogen Shigella flexneri showed increased mortality when compared to conventional guinea pigs 29. Additionally, infection of GF mice with Listeria monocytogenes resulted in decreased clearance, and infection with Salmonella enterica resulted in more severe gastroenteritis compared to conventional control mice 30,31. The microbiome also provides an important site for immune tolerance and modulation. Conventional mice challenged with oral ovalbumin antigen, for example, showed systemic tolerance to the same antigen for a 2-3 month period. In contrast, Germ free experimental mice showed a loss of tolerance between only a 7-21 day period 32. Additionally, gut mucosa epithelial cells constantly sample ingested and commensal microbiota antigen providing real time immunological adaptation to the environment by generation of cytokines and transmitting signals to submucosal inflammatory and immune cells 33. These data together suggest an important role of the gut in immune host defense. 26

29 Not only does the gut microbiota aid in the development of the immune system, it also provides a critical barrier function against invading pathogen microbes. There are several mechanisms that establish the resident microbiome as a barrier. Ostensibly, a barrier is established physically by the competitive exclusion of pathogens and opportunistic microbes by growth. There is also tight control over nutrient exchange between the host and microbiome, thereby preventing excess available nutrients for opportunistic and/or pathogen establishment 16. Although the exact molecular mechanisms are ill-defined, numerous studies have demonstrated that human fecal bacterial species have antimicrobial activities against specific invading enteric pathogens 20. Presumably, one of the major anti microbial mechanisms is widely produced proteinacious substances known as bacteriocins 16,20. Interestingly, bacteriocins often utilize host proteases for both activation and degradation reinforcing the mutualistic host-microbiome relationship 16,20. Other mechanisms include bacterial production of metabolites like lactic acid by Lactobacillus species that inhibit local bacterial growth 20. The gut microbiome also stimulates the host to synthesize antimicrobial peptides (AMPs) like defensins, cathelicidins, and C-type lectins that serve to prevent the gut microbiota from overgrowing and invading the epithelial cell barrier, but also will serve as protection against pathogens 20. Finally, the intestinal microbiota helps repair damaged epithelial cells, maintain tight junctions between epithelial cells, and maintain epithelial cells through interaction with surface epithelial receptors and stimulation of signaling cascades 34,35. 27

30 Over an average human lifetime approximately 60 tons of food will pass through the gastrointestinal tract implying an important relationship between the gut microbiota and diet 18. An increased recognition of the importance of the gut microbiota in energy harvest and metabolism has occurred over the last years. Indeed, the gut microbiome has a critical role influencing nutrition, energy harvest, and normal metabolism in mammals. Humans are not able to metabolize most complex carbohydrates or plant polysaccarrhides like cellulose, some starches, and xylan 36. These products are instead degraded by the gut microbiota into short chain fatty acids (SCFAs) such as acetate 36,37. SCFAs provide an energy source for the gut microbiota itself, the colonic epithelium, and peripheral tissues. Additionally, SCFAs can influence inflammation, wound healing, motility, and vessel vasoreactivity 36. Recently, a role for SCFA in normal protein homeostasis has been shown 37. The most abundant SCFA produced by the human microbiome, acetate, was demonstrated to contribute to pools of acetyl-coa, participate in lysine-ε-acetylation, and influence protein function 36,38. Insights into the importance of the gut microbiota in energy harvest are mostly garnered from studies in GF animals. GF rodents produce less SCFAs and excrete 2 fold more calories in the feces and urine compared to conventional mice 39,40. This results in decreased adiposity in germ free mice and the consumption of twice the caloric intake of 28

31 conventional mice Remarkably, germ free mice can normalize their adiposity after only 2 weeks post conventionalization 36,41,42. Further support of the importance of the gut microbiota in energy harvest is demonstrated in studies of ob/ob mice, a mouse model of obesity. ob/ob mice have more cecal SCFAs and less residual calories found in their feces 43. Together these studies suggest that the gut microbiota can influence energy harvest in mammals. GF rodents are also noted to have dysfunctional lipid metabolism. Overall, systemic cholesterol metabolism is reduced, but surprisingly GF mice develop increased cholesterol content in the liver and excrete more cholesterol in feces 44. Moreover, GF rodents also have exhibited dysfunction in bile acid metabolism. Primary bile acids are synthesized from cholesterol and excreted into the intestinal lumen where the gut microbiota further metabolizes them into secondary bile acids. The metabolism of bile acids by the gut microbiota allows bypass of the normal mechanisms of reuptake and excretion into the feces 36. However, in GF animals the absence of bacteria allows for unmetabolized bile acids to be taken up vastly expanding the bile acid pool size 36. Additionally, both primary and secondary bile acids function in signaling and regulation of normal host metabolism. Disruption of the normal metabolism of bile acids consequently could cause metabolic dysfunction 36. Many of the advances in understanding the role of the microbiota in normal human metabolism have coincided with and largely been driven by the 29

32 development of sophisticated (e.g. 16s ribosomal RNA surveys) tools to characterize the gut microbiome. As a result of these technological advancements, there has been increased recognition that long term diets fundamentally alter and determine the composition of the gut microbiota. A study by Ley et al. demonstrated that interspecies analysis of carnivorous, omnivorous, and herbivore mammals reveal closely aligned gut microbiota compositions suggesting dietary patterns have a had a great influence on the coevolution of the mammals and the gut microbiome 45. Indeed, a follow-up study showed that gut microbiota composition significantly aligned with the dietary components total protein, insoluble fiber, and carbohydrates 46. Moreover, comparison of the gut microbiota composition of children from rural Africa who predominantly consume a carbohydrate and plant based diet and European children where fats and protein constitute a larger part of the diet, show distinct differences 47. Overall, the human microbiome can be subdivided into five major known bacteria phyla: Firmicutes, Bacteroides, Actinobacteria, Proteobacteria, and Verrucomicorbia 36. More recently more refined classification systems have suggested subdividing major intermicrobial communities that work in a symbiotic relationship. There has also been a suggestion that the human gut microbiome can be stratified into 3 major enterotypes primarily composed of the genera Bacteroides, Prevotella, and Ruminococcus respectively 47. Remarkably, these enterotypes have also been found to associate with dietary habits (e.g. Bacteroides with a high protein, carnivorous diet; Prevotella with a carbohydrate 30

33 based diet) 48. Moreover, 10 day dietary interventions failed to significantly alter the composition of the gut microbiota suggesting that only long term dietary habits are important in determining its composition 49. Mice also have an altered microbiota based on diet. Mice consuming high fat diets, for example, have a gut microbiota that have an increased composition of Firmicutes and Proteobacteria 50. These descriptive microbiota differences imply that the gut microbiome plays an important role in dietary metabolism. The Relationship between the Gut Microbiota and Disease Like disruption of other normal physiological processes in our body, dysfunction of the gut microbiome can contribute to pathological processes 16,20,51. For example, the pathogenesis of inflammatory bowel disease (IBD) has been partly attributed to dysfunction in the interaction of host immunity with the gut commensal microbiota 1,5. Studies of the commensal gut microbiota have demonstrated that certain genera of bacteria (e.g. Bacteroides) are associated with the severity and presence of IBD 16. IBD patients often have a greater mass of commensal bacteria adhering to the epithelial cell layer and commensal bacteria have been found to invade into the epithelial cell layer 52. A role for bacteria has been further elaborated in mice by studies utilizing broad spectrum antibiotics that suppress the gut microbiota and result in decreased mucosal inflammation in rodents 53,54. These observations were recapitulated in humans by the demonstration that antibiotic treatment decreased mucosal inflammation in subjects with IBD to a greater extent than systemic steroid treatment 53,55. Host 31

34 immunity has also been implicated in the disease pathogenesis of IBD patients who characteristically contain IgG against a wide range of commensal microbiota species including relatively innocuous species 56. The generation of IgG promotes epithelial cell injury and inflammatory cascades that further damage the mucosal intestinal barrier 57. Moreover, an estimated 25% of patients with Crohn s disease have a loss of function mutation in the NOD2/CARD25 gene that is found primarily in leukocytes and recognizes the muramyl dipeptide (MDP) moiety in bacteria 58. These data together suggest a role of intestinal microbiota in the pathogenesis of IBD. A role for the gut microbiota in colon cancer has also been implicated. There are multiple studies demonstrating links between gut microbiota bacterial taxa and colon cancer. Bacterial genera including Bacteroides, Clostridium, and Bifodobacterium are associated with colon cancer; whereas species including Lactobaccilus and Eubacterium are inversely associated 59,60. The intestinal microbiota may also play a role in facilitating the metabolism of dietary nutrients into carcinogenic products like N-nitroso compounds 16. These observations were confirmed by a recent study that demonstrated a direct relationship between the gut microbiota and colon cancer 61. An enterotoxigenic species Bacteroides fragilis that can asymptomatically colonize the colons in a proportion of the human population can secrete the Bacteroides fragilis toxin (BFT). BFT has been known to cause human inflammatory diarrhea, but also can promote colon cancer via a TH17 (subtype of CD4+ T cells)-dependent pathway

35 A more expansive role of the gut microbiota and disease pathogenesis has been implicated in syndromes that are associated with breakdown of the normal epithelial mucosa that allow bacteria to translocate across the mucosal barrier 16. Gut microbiota bacteria translocation can lead to adverse severe sequelae such as sepsis, toxemia, multisystem failure, or death 16. Translocation is believed to be mediated by at least three major mechanisms which include bacterial overgrowth, gut immunological deficiencies, or increased permeability of the gut mucosa 62. Many disease states such as multi-system organ failure, pancreatitis, liver cirrhosis, and intestinal obstruction have demonstrated evidence of invading gut microbiota bacteria into the intestinal wall 16,62. A more recent notable example of immunological deficiency resulting in loss of the mucosal barrier derives from studies in HIV and pathogenic SIV infection. These studies show that chronic depletion of TH17 cells, a critical cell in the maintenance of the normal immunological gut mucosa barrier, of the GI mucosa are associated with progression of HIV pathogenesis 63. Depletion of TH17 cells and the subsequent loss of the secreted proinflammatory cytokine interleukin-17 (IL-17) led to the ability of opportunistic infections to advance 64. Presumably, the depletion of these cells may also allow for translocation of gut microbiota across the GI mucosa and adverse sequelae 63. High rates of positive culture of gut microbiota in mesenteric lymph nodes in diseases commonly associated with bacterial translocation such as IBD, 33

36 pancreatitis, or liver cirrhosis are expected. Indeed an estimated 16-40% of these patients have positive cultures 65. However, positive mesenteric lymph node cultures are also typically found an estimated 5% of the population 65. This observation demonstrates that bacterial translocation is a common occurrence and may have a role in less acutely insidious pathological processes such as CVD or diabetes. Together, these data demonstrate a role of the gut microbiota in cancer, infectious and inflammatory disease states. A relationship of the gut microbiota with complex metabolic diseases was first described by Gordon and colleagues that published seminal papers for the role of the gut microbiota and obesity. Most notably was a study where donation of an obese microbiota (from ob/ob mice) to a germ free animal significantly increased total body fat when compared to parallel transplantation of GF mice with conventional mouse microbiota 43. The mechanism of this effect is largely attributed to an increase in energy harvest from the microbiota of ob/ob mice. Indeed, the composition of the microbiota in mice fed a high fat diet is significantly different compared to mice on a chow diet 50,66. These data are further supported by studies in lean and obese twins 43. The obese individuals of the twin pairs were more associated with decreased gut microbiota diversity, phyla differences, and altered bacterial metabolic pathways 43. Overall, there have been many reports demonstrating significant changes in the gut microbiota composition in accordance with weight 43,67,68. Interestingly, the gut microbiota composition in patients undergoing gastric bypass operations is also significantly 34

37 altered post-op 36. The observation that these subjects are able to experience an antidiabetic effect even before significant weight loss occurs suggests that the gut microbiota may also play a role in this direct effect 36. Gut Microbiota Mediated Metabolism of Phosphatidylcholine Promotes a, 69 Cardiovascular Disease Recently, the role of the microbiota has been extended from complex metabolic diseases such as diabetes and obesity to atherosclerotic disease 69. The gut microbiota metabolism of phosphatidylcholine, the major source of dietary choline, produces a noxious intermediate compound known as trimethylamine (TMA) that is further metabolized by liver Flavin monooxygenase (FMO) to TMAO thereby promoting atherosclerotic disease (Fig. 1-1). The discovery of this pathway began with a search for novel pathways involved in cardiovascular disease pathogenesis by using an unbiased metabolomic study (Fig. 1-2). Small-analyte plasma profiles were acquired initially in a Learning Cohort that consisted of subjects undergoing elective coronary angiography who then experienced an ensuing major adverse cardiovascular event (MACE; MI, stroke, or death) over a 3-year period and age and gender matched controls that did not experience MACE. Direct comparison between diseased and control subject profiles using liquid chromatography with on-line spectrometry (LC/MS) demonstrated that 40 analytes out of >2,000 analyzed were associated with a From Gut flora metabolism of phosphatidylcholine promotes cardiovascular disease, v.472 Copyright (2011) Nature Publishing group. Reprinted with permission. 35

38 cardiac risks 69. A second Validation Cohort study was performed in which small-analyte plasma profiles were acquired for a cohort of completely independent subjects undergoing elective coronary angiography who also experienced MACE over a 3-year period and age or gender matched controls that did not experience MACE. Analysis of >2,000 possible analyses yielded significant differences in 24. When comparing the Learning and Validation cohorts 18 common unknown analytes significantly associated with cardiovascular disease (Fig. 1-2). Among these unknown analytes 3 had a common association with m/z 76, 104, 118 respectively suggesting they may be part of a common biochemical pathway. Further structural studies confirmed the identities of these unknowns as trimethyl amine N-oxide (TMAO; m/z 76), choline (m/z 104), and betaine (m/z 118) 69. Remarkably, the metabolism of choline to TMAO was a gut microbiota mediated process and suggested a role of the gut microbiota in atherosclerosis 69,70. The major dietary source of choline is in the form a member of the phospholipid class of lipids, phosphatidylcholine (PC). Interestingly, whereas the other two major classes of lipids, sterols and triglycerides, have been associated with CVD, a role for phospholipids has not. These data suggested a link between dietary phospholipids, the gut microbiota, and atherosclerosis 69. Challenge of mice with heavy stable isotope labeled d9-phosphatidylcholine (d9- PC) demonstrated production of both d9-tmao and d9-betaine (Fig. 1-3). 36

39 Following suppression of the gut microbiota with oral broad spectrum antibiotics, mice rechallenged with d9-pc showed complete absence of d9-tmao production, but still demonstrated production of d9-betaine. Reconventionalization of mice shows reacquisition of the gut microbiota production of d9-tmao from d9-pc demonstrating an obligatory role of the gut microbiota in TMAO production from d9-pc, but not betaine. These observations were also confirmed in germ free mouse d9-pc challenges studies 69. Next confirmatory studies of the relationship between plasma choline, TMAO, and betaine discovered in the unbiased metabolomics approach were performed. Quantification of plasma trimethylamines in n=1,876 sequential subjects undergoing elective coronary angiography at Cleveland Clinic showed increasing concentrations of choline, TMAO, and betaine were associated with increased, dose dependent, prevalence of CVD (Fig. 1-4). Moreover, adjustment for traditional risk factors with multivariate modeling demonstrates the PC metabolites are independently associated with CVD 69. Together these data raised the possibility that dietary supplementation of choline and TMAO may promote atherosclerosis. Atherosclerosis-prone female C57BL/6J, Apoe-/-mice were placed on increasing concentrations of dietary choline or TMAO at time of weaning (4 weeks of age) for 16 weeks before sacrifice. Quantification of aortic root plaque showed an increased amount of atherosclerotic plaque at the aortic root of mice fed a trimethylamine diet 37

40 compared to chow controls (Fig. 1-5) despite no significant increases in plasma lipid profiles or liver pathology 69. Interestingly, plasma TMAO levels significantly correlated with burden of plaque at the aortic root suggesting the terminal gut microbiota dependent product, TMAO is responsible for promotion of atherosclerosis. A similar trend was observed in humans when examining the burden of atherosclerotic disease (defined by the presence of coronary artery disease (CAD) in one, two, or three vessels) with plasma TMAO levels 69. Hepatic FMO3 is the enzymatic source of TMAO production in humans and loss of function mutations in the gene coding for FMO3 protein results in fish malodor syndrome that is characterized by an individual s inability to metabolize TMA, a noxious gas at room temperature that smells like rotting fish, to odorless TMAO 71,72. The end result for afflicted individuals is body odor that smells like rotting fish that is made worse by consuming dietary sources rich in trimethylamine containing compounds (e.g. dairy products and meats). The involvement of endogenous FMO3 in TMAO production raised the possibility of genetic regulatory involvement in atherosclerosis. Using integrative genetic approaches the role of FMO3 expression and regulation was investigated in murine atherosclerosis. Association studies between purified liver FMO3 mrna from a F2 intercross between an atherosclerotic resistant mouse strain (C3H/HeJ, Apoe-/-) and an atherosclerotic prone mouse (C57BL/6J, Apoe-/-) and atherosclerotic plaque burden showed a positive correlation (R = 0.29, P = 0.002). Moreover, hepatic FMO3 expression also had a significant positive 38

41 association with plasma TMAO levels and a negative association with mouse plasma HDL levels (R = 0.80, P < 0.001). In a next set of studies eqtl analysis was performed using mice from the same F2 intercross. A single nucleotide polymorphism on mouse chromosome1 that was in close proximity to the FMO3 gene, and that was simultaneously in a region that had previously been linked to atherosclerotic disease burden, showed a dose dependent relationship with atherosclerotic disease burden (Fig. 1-6). These data provide evidence of a relationship between FMO3 expression and atherosclerotic disease. In a final set of studies, a role for the gut microbiota in dietary choline induction of atherosclerosis was determined. C57BL/6J, Apoe-/- male and female mice were placed on a choline supplemented or control diet in the presence or absence of broad spectrum antibiotics (used to suppress the gut microbiota) for 16 weeks post weaning. Quantification of atherosclerotic plaque at the aortic root of these mice showed a significant increase in plaque area compared to chow controls. Importantly, this increase was also significant compared to mice supplemented with choline and with a concomitant suppressed gut microbiota. This confirms a gut flora dependence mechanism in choline induced atherosclerosis (Fig. 1-7). These data link together a previously unrecognized pathway between dietary lipids (in the form of phosphatidylcholine), the gut microbiota, and atherosclerosis. Moreover, these data also suggested that other dietary trimethylamine containing compounds may contribute to TMAO formation and atherosclerotic disease. 39

42 Figure 1-1. Scheme of gut microbiota dependent metabolism of dietary PC and atherosclerosis. Choline, a trimethylamine species that is found in food principally as phosphatidylcholine (PC), is metabolized by commensal gut flora to form the noxious intermediate compound trimethylamine (TMA). TMA is further oxidized by flavin monooxygenases (FMOs) to form TMAO promoting the formation of atherosclerotic plaque

43 Figure 1-2. Metabolomics studies scheme and correlations. a. Scheme of unbiased metabolomic study that identified plasma analytes associated with CVD. b. Significant correlations between analytes m/z 76, 104, 118 from metabolomics studies suggested a common biochemical pathway

44 . Figure 1-3. Production of TMAO from PC is gut flora dependent. LC/MS/MS plasma quantification of d9-choline, d9-tmao, d9-betaine after gastric gavage of d9-dppc in conventional mice, following suppression of the gut microbiota with broad spectrum antibiotics (3 weeks), and then following a reacquisition period (4 week housing with non-sterile mice (i.e. conventionalized )). Data are presented as mean ± SE from 4 independent replicates. d9-tmao production

45 Figure 1-4. Choline, TMAO and betaine are associated with CVD in humans. a-c. Logistic regression spline plots of the relationship between plasma analytes choline, TMAO, and betaine with cardiovascular disease (CVD) (with 95% CI) in n = 1876 subjects

46 Figure 1-5. Dietary choline or TMAO enhances atherosclerosis. C57BL/6J, Apoe-/- female mice at time of weaning (4 weeks) were placed on the respective choline diet, TMAO diet or a chow diet for 16 weeks. Atherosclerotic plaque was quantified at the aortic roots of mice at time of sacrifice

47 Figure 1-6. Hepatic FMOs associate with atherosclerosis. Association of the FMO3 genotype (SNP rs ) with both (C57BL/6J, Apoe-/-) and atherosclerosis resistant (C3H/HeJ, Apoe-/-) mice

48 Figure 1-7. Dietary choline enhances atherosclerosis in a gut flora dependent manner. C57BL/6J, Apoe-/- male and female mice at time of weaning (4 weeks) were placed on either a choline diet (1.0%) or a chow diet (0.08 % total choline) in the presence or absence of broad spectrum antibiotics (+ABS) in the drinking water for 16 weeks. Atherosclerotic plaque was quantified at the aortic roots of mice at time of sacrifice

49 CHAPTER 2 b : Intestinal Microbiota Metabolism of L-Carnitine, a Nutrient in Red Meat, Promotes Atherosclerosis 73 Authors: Robert A. Koeth, Zeneng Wang, Bruce S. Levison, Jennifer A. Buffa, Elin Org, Brendan T. Sheehy, Earl B. Britt, Xiaoming Fu, Yuping Wu, Lin Li, Jonathan D. Smith, Joseph A. DiDonato, Jun Chen, Hongzhe Li, Gary D. Wu, James D. Lewis, Manya Warrier, J. Mark Brown, Ronald M. Krauss, W. H. Wilson Tang, Frederic D. Bushman, Aldons J. Lusis, and Stanley L. Hazen Abstract Intestinal microbiota (i.e. "gut flora") - metabolism of choline/phosphatidylcholine produces trimethylamine (TMA), which is further metabolized to a proatherogenic species, trimethylamine-n-oxide (TMAO) 69. Herein we demonstrate that gut microbiota metabolism of dietary L-carnitine, a trimethylamine abundant in red meat, also produces TMA, TMAO and accelerates atherosclerosis. Omnivorous subjects to a far greater extent than vegans/vegetarians are shown to produce TMAO following ingestion of L-carnitine through a gut microbiota dependent mechanism. Specific bacterial taxa in human feces are shown to associate with both plasma TMAO and omnivore versus vegan/vegetarian status. Plasma L- carnitine levels in sequential stable subjects undergoing cardiac evaluation (n>2,500) predict increased risks for both prevalent cardiovascular disease (CVD) and incident major adverse cardiac events (MI, stroke or death), but only among subjects with concurrently high TMAO levels. Chronic dietary L-carnitine supplementation in mice is shown to significantly alter cecal microbial b From Intestinal microbiota metabolism of L-carnitine, a nutrient in red meat, promotes atherosclerosis, Copyright (2013) Nature Publishing group. Reprinted with permission 47

50 composition, augment synthesis of TMA/TMAO by over 10-fold, and increase aortic root lesion area, but not following suppression of intestinal microbiota. Dietary supplementation of TMAO in mice, or either carnitine or choline in mice with intact but not suppressed intestinal flora, significantly reduced reverse cholesterol transport in vivo. Gut microbiota may thus participate in the wellestablished link between increased red meat consumption and CVD risk. Introduction Cardiovascular disease (CVD) remains the leading cause of morbidity and mortality in western societies. The high-frequency consumption of meat products in the developed world is linked to cardiovascular disease risk, presumably due to the large content of saturated fats and cholesterol found in these foods 74,75. However, a recent meta-analysis of prospective cohort studies showed no association between dietary saturated fat intake and CVD, prompting the suggestion that other environmental exposures linked to increased dietary meat consumption are responsible 76. In fact, the suspicion that the cholesterol and saturated fat content of red meat may not be sufficiently high to account for observed risks has long stimulated the investigation of alternative sources of disease-promoting exposures that accompany dietary meat ingestion, such as the high content of salt or heterocyclic compounds generated during cooking 77,78. Of note, to date, such studies have largely focused on the biochemical content of meat itself before or following processing, and have yet to address the impact of 48

51 our commensal intestinal microbiota (i.e. gut flora) and their participation in modifying the diet-host interaction. Trillions of bacteria populate our digestive system, exceeding by approximately an order of magnitude the total number of cells in our body. Our gut microbiota has been linked to intestinal health, immune function, bioactivation of critical nutrients and vitamins, and more recently, complex disease phenotypes such as obesity and insulin resistance 43,79,80. We recently reported a novel pathway in both humans and murine models of atherosclerosis linking gut microbiota metabolism of dietary choline to CVD pathogenesis 69. Choline, a trimethylamine containing compound and part of the head group of phosphatidylcholine (PC), the major dietary source of choline, is metabolized by the action of gut microbiota to produce an intermediate gaseous compound known as trimethylamine (TMA) (Fig. 2-1a). TMA is rapidly further oxidized by one or more hepatic flavin monooxygenases (FMO) to form the metabolite trimethyl amine N-oxide (TMAO), which was shown to be proatherogenic. Atherosclerotic prone apolipoprotein E-/- mice treated with a diet supplemented with either choline or the downstream metabolite, TMAO, demonstrated enhanced aortic root atherosclerotic burden. In contrast, germ-free mice, or animals with suppressed intestinal microbiota through use of oral broad spectrum antibiotics, failed to make both TMA and TMAO following ingestion of either phosphatidylcholine or choline, and showed no increase in atherosclerosis from a high choline diet. Further, integrative genetics studies demonstrate the FMO gene cluster on chromosome 1 as an 49

52 atherosclerosis susceptibility locus in the rodent model. Finally, plasma levels of TMAO and choline in subjects were associated with CVD risks 69. These results collectively indicated both an obligatory role for gut microbiota in the production of TMAO from dietary choline and phosphatidylcholine, and that elevated levels of TMAO are mechanistically linked to accelerated atherosclerotic heart disease in rodent models and humans. The findings further raise the possibility that other dietary nutrients that possess a similar trimethylamine structure may also contribute to TMAO formation via gut microbiota, and consequently, accelerated atherosclerosis. How TMAO is mechanistically linked to development of accelerated atherosclerosis and which specific microbial species contribute to TMAO formation remain unknown. L-carnitine is an abundant dietary nutrient in red meat that contains a trimethylamine structure similar to choline (Fig. 2-1a). A hydrophilic quaternary amino acid, its name is derived from Latin carnis, meaning flesh. While ingestion of L-carnitine from diet is a major source of the compound in omnivores, the amino acid is also endogenously produced in mammals from lysine, and serves an essential function in the transport of fatty acids from the cell cytoplasm to the mitochondrial compartment 81,82. Recent changes in dietary habits in industrialized societies have included a tremendous growth in L- carnitine supplementation as a food or drink additive, particularly in many power or energy drinks, or nutritional supplements aimed at increasing muscle mass. Over the past few years L-carnitine also has become a common supplement 50

53 added to commercial beverages including coffees/espresso, flavored vitamin/water drinks, and other beverages widely consumed by the public. Whether there is a potential health risk for such pervasive and rapidly growing nutritional supplement practices has not been considered, much less explored. Herein we examine the gut microbiota-dependent metabolism of L-carnitine to produce TMAO in both rodents and humans (omnivore vs. vegans). Through a combination of isotope tracer studies, large scale clinical studies and animal model investigations employing both germ-free mice and mice with intact and suppressed intestinal microbiota, we demonstrate a role for gut microbiota metabolism of L-carnitine in atherosclerosis pathogenesis in the appropriate dietary setting (high carnitine ingestion). In addition to the upregulation of macrophage scavenger receptors potentially contributing to enhanced "forward cholesterol transport" 69, we further show that TMAO, and its dietary precursors choline and carnitine, suppress reverse cholesterol transport through gut microbiota dependent mechanisms in vivo. Finally, we define microbial taxa in humans and murine models associated with both TMAO production and dietary carnitine ingestion, and show dynamic microbial compositional changes that occur with carnitine supplementation, and consequent marked enhancement in TMAO synthetic capacity in vivo. Results Metabolomic studies link L-carnitine with CVD 51

54 Given the similarity in structure between L-carnitine and choline (Fig. 2-1a) we hypothesized that dietary ingestion of L-carnitine in humans, like choline and phosphatidylcholine, might produce TMA and TMAO in a gut microbiota dependent fashion, and be associated with atherosclerosis risk in humans. To test this we initially examined data from our recently published unbiased small molecule metabolomics analyses of plasma analytes and CVD risks 69. An analyte with identical molecular weight to L-carnitine (mass to charge ratio (m/z) 162) was not in the top tier of analytes that met the stringent P value cutoff for association with CVD after Bonferroni adjustment for multiple comparisons in both the initial Learning Cohort and the subsequent Validation Cohort 69. However, a hypothesis-driven examination of the data using less stringent criteria (no adjustment for multiple testing) did reveal an analyte with m/z 162 that showed a tendency toward positive association with major adverse cardiovascular events over a 3 year period in the combined cohort of patients (unadjusted Hazard Ratio 2.63; 95% CI( ); P = 0.04)(Supplementary Table 2-1). In further studies we were able to confirm the identity of the plasma analyte as L-carnitine by using multiple approaches, including demonstration of identical retention time under multiple chromatographic conditions during LC/MS analysis, identical collision-induced dissociation (CID) mass spectrum with that of an authenticate standard L-carnitine, and co-elution of multiple characteristic precursor product ion transitions in a plasma sample spiked with synthetic stable isotope labeled (d9-trimethyl)-carnitine standard (Supplementary Figs. 2-1, 2-2). Unbiased metabolomics data as performed 69 are semi-quantitative in 52

55 nature; however, they are hypothesis generating, and thus suggested that plasma levels of L-carnitine may associate with CVD risks. For all subsequent studies we therefore developed and used quantitative stable isotope dilution LC/MS/MS methods for measuring endogenous L-carnitine using a synthetic isotopologue of L-carnitine (d9-(trimethyl)-l-carnitine) as internal standard (Supplementary Fig. 2-3). Gut microbiota plays an obligatory role in forming TMAO from L-carnitine in humans The participation of gut microbiota in TMAO production from dietary L-carnitine in humans has not yet been shown. We therefore first sought to test the ability of human micro microbiota to help produce TMAO from ingested L-carnitine by developing a human L-carnitine challenge test. Since the bioavailability of dietary L-carnitine within an endogenous food source is reported to be substantially higher (estimated 4-fold) than the bioavailability of carnitine supplements 83, in initial subjects (omnivores), the L-carnitine challenge test incorporated a major source of dietary L-carnitine (8 ounce sirloin steak, corresponding to an estimated 180 mg L-carnitine) 84,85 and a capsule containing 250 mg of a heavy isotope labeled L-carnitine (synthetic d3-(methyl)-l-carnitine). At baseline (Visit 1), post-prandial increases in d3-tmao and d3-l-carnitine in plasma were readily detected, and 24 hour urine collections also revealed d3- TMAO (Fig. 2-1b-e; Supplementary Fig. 2-4, 2-5. Data shown in all panels of Fig. 2-1 and Supplementary Fig. 2-4 are tracings from a representative 53

56 omnivorous subject, of n=5 studied with complete serial blood draws post carnitine challenge). As previously observed 69, endogenous (non-labeled) fasting plasma TMAO levels showed wide variation in levels at baseline among subjects, suggesting wide inter-individual variations exist in gut microbiota capacity to generate TMAO (see below). In most subjects examined, despite clear increases in plasma d3-carnitine and d3-tmao over time, post prandial changes in endogenous (non-labeled) carnitine and TMAO were modest (Supplementary Fig. 2-5), consistent with a total body (and intravascular) pool of natural abundance carnitine and TMAO that are relatively vast in relation to the amount of carnitine ingested and TMAO produced from the carnitine challenge. To examine the potential contribution of gut microbiota to TMAO formation from dietary L-carnitine, volunteers were then placed on oral broad spectrum antibiotics to suppress intestinal microbiota for a week as described under Methods, and then another baseline sample collected, and repeat L-carnitine challenge performed (Visit 2). A remarkable complete suppression of measurable endogenous TMAO at baseline in both plasma and urine were noted (Fig. 2-1be; Supplementary Fig. 2-5). Moreover, in every subject examined with carnitine challenge following the course of oral antibiotics, virtually no detectable formation of either native or d3-labeled TMAO was observed in post prandial plasma or 24 hour urine samples, demonstrating that TMAO production from dietary L-carnitine in subjects has an obligatory role for gut microbiota(supplementary Fig. 2-4). In contrast, both d3-l-carnitine and unlabeled L-carnitine were readily detected 54

57 following their ingestion during carnitine challenge, and showed little change in the overall time course for post prandial changes in levels observed before (Visit 1) versus after antibiotic treatment (Visit 2; Fig. 2-1e, Supplementary Fig. 2-5). After discontinuation of antibiotics, subjects were invited back for a third visit after at least another three weeks. Examination of baseline and post L-carnitine challenge plasma and urine samples again showed TMAO and d3-tmao formation in both plasma and urine, consistent with intestinal re-colonization (Fig. 2-1b-e; Supplementary Fig. 2-4, 2-5). Collectively, these data clearly show that TMAO production from dietary L-carnitine in humans is gut microbiota dependent. Vegans and vegetarians produce substantially less TMAO from dietary L- carnitine As noted above, the capacity to produce native and d3-labeled TMAO following native and d3-l-carnitine ingestion was variable among individuals. A post-hoc nutritional survey performed amongst the volunteers suggested that the antecedent dietary habits (red meat consumption) may influence the capacity to generate TMAO from L-carnitine. To test this prospectively, we examined TMAO and d3-tmao production following the same L-carnitine challenge, first in a long term (>5 years) vegan who consented to the carnitine challenge (including both steak and d3-(methyl)-carnitine consumption). Figure 2-2a illustrates results from carnitine challenge in this vegan volunteer who was willing to ingest steak as part of the carnitine challenge. Also shown for comparison are data from a single 55

58 omnivore with reported common (near daily) dietary consumption of red meat. Post-prandially we noted that the omnivorous subject with common red meat consumption showed both an increase in TMAO and d3-tmao levels in sequential plasma measurements (Fig. 2-2a), and in a 24 hour urine collection sample (Fig. 2-2b). In contrast, the vegan subject showed nominal fasting plasma and urine TMAO levels at baseline, and virtually no capacity to generate d3-tmao or TMAO in plasma after the carnitine challenge, with approximately 1000-fold less d3-tmao produced from the same oral d3-l-carnitine load compared to the representative omnivore (Fig. 2-2a,b). The vegan subject also had lower fasting plasma levels of L-carnitine compared to the omnivorous subject (Supplementary Fig. 2-6). To confirm and extend these findings we examined additional vegans/vegetarians (n=23) and omnivorous subjects (n=51). Fasting baseline TMAO levels were significantly lower among vegan/vegetarian subjects compared to omnivores (Fig. 2-2c). In a subset of these individuals an oral d3(methyl)-carnitine challenge (but with no steak) was performed, confirming that long term (> 1 year) vegan/vegetarians have markedly reduced synthetic capacity to produce TMAO from oral carnitine (Fig. 2-2c,d). Interestingly, vegan/vegetarians challenged with d3-carnitine also had significantly more postchallenge plasma d3-carnitine compared to omnivorous subjects (Supplementary Fig. 2-7), a result that may reflect decreased intestinal microbial metabolism of carnitine prior to absorption. 56

59 Plasma TMAO levels significantly associate with specific human gut microbial taxa Dietary habits (e.g. vegan/vegetarian versus omnivore/carnivore) are associated with significant alterations in intestinal microbiota composition and function 45,46,86. We therefore examined fecal samples from both vegans/vegetarians (n=23) and omnivores (n=30) for analyses of the gene encoding for bacterial 16S ribosomal RNA, and in parallel, plasma TMAO, and carnitine and choline levels were quantified by stable isotope dilution LC/MS/MS. Global analysis of taxa proportions by combining both the weighted and unweighted Unifrac distances using PeranovaG revealed significant associations with plasma TMAO levels (P=0.03), but not plasma carnitine (P= 0.77) or choline (P =0.74) levels. Several bacterial taxa remained significantly associated with plasma TMAO levels after false discovery rate (FDR) adjustment for multiple comparisons (Supplementary Fig. 2-8). When subjects were classified into previously reported enterotypes 49 based upon fecal microbial composition, individuals with an enterotype characterized by enriched proportions of the genus Prevotella (n=4) demonstrated higher (p<0.05) plasma TMAO levels than subjects with an enterotype notable for enrichment of Bacteroides (n=49) genus (Fig. 2-2e). Examination of the proportion of specific bacterial genera and subject TMAO levels revealed several taxa (genus level) that simultaneously were significantly associated with both vegan/vegetarian versus omnivore status, and plasma TMAO levels (Fig. 2-2f). 57

60 TMAO production from dietary L-carnitine is an inducible trait Analyses of data from vegan/vegetarians versus omnivores thus far suggested that preceding dietary habits modulate gut microbiota composition, and the synthetic capacity to ultimately produce TMAO from dietary L-carnitine may be highly adaptable. We next investigated the ability of chronic dietary L-carnitine to induce gut flora-dependent production of TMA and TMAO in the murine model. Pilot LC/MS/MS studies first confirmed the presence of L-carnitine in plasma of conventional C57BL/6J and atherosclerosis-prone C57BL/6J, Apoe-/-mice on normal chow diet, which contains no L-carnitine per manufacturer (carnitine content of chow diet was also confirmed by LC/MS/MS analyses, data not shown). To confirm that mouse intestinal microbiota could produce TMA and TMAO from dietary L-carnitine, we also examined germ-free mice and observed no detectable plasma d3-(methyl)tma or d3-(methyl)tmao following oral (gastric gavage) d3-(methyl)carnitine challenge, but acquisition of capacity to produce both d3-(methyl)tma and d3-(methyl)tmao following oral (gastric gavage) d3-(methyl)carnitine after a several week period in conventional cages to allow for microbial colonization (i.e. conventionalization ) (Supplementary Fig. 2-9). Parallel studies with conventional C57BL/6J, Apoe-/- mice that were placed on a cocktail of oral broad spectrum antibiotics previously shown to suppress intestinal microflora 35,69 showed similar results as the germ-free mice (i.e., complete suppression of both TMA and TMAO formation; Supplementary Fig. 2-10), confirming in the mouse model an obligatory role for gut flora in both TMA and TMAO production from dietary L-carnitine. To examine the impact of 58

61 dietary L-carnitine on inducibility of TMA and TMAO production from intestinal microbiota, we compared the pre- and post-prandial plasma profile of C57BL/6J, Apoe-/- mice on normal chow diet versus a diet supplemented in L-carnitine for 15 weeks. The production of both d3-(methyl)tma and d3-(methyl)tmao following oral ingestion (gastric gavage) of d3-(methyl)carnitine was induced by approximately 10-fold in mice on the L-carnitine supplemented diet compared to normal chow diet fed controls (Fig. 2-3a). Further, plasma post-prandial d3- (methyl)carnitine levels in mice in the carnitine supplemented diet arm were significantly lower than that observed in mice on the carnitine free diet (normal chow), consistent with enhanced gut flora-dependent catabolism prior to absorption in the carnitine supplemented mice. TMA / TMAO production associates with specific mouse gut microbial taxa The marked effect of chronic dietary carnitine on enhanced TMA and TMAO production from a carnitine challenge (d3-(methyl)carnitine by gavage) suggested that carnitine supplementation may have significantly altered intestinal microbial composition with enrichment of taxa better suited for TMA production from carnitine. To test this we first identified the cecum as the segment of the entire intestinal tract of mice that shows the highest synthetic capacity to form TMA from carnitine (data not shown). We then sequenced 16S rrna gene amplicons from cecum of mice on either normal chow (n=10) or carnitine supplemented diet (n=11) and in parallel, quantified plasma levels of TMA and TMAO using stable isotope dilution LC/MS/MS (Fig. 2-3b,c). Global analyses of individual taxa 59

62 proportions reveals that in general, microbial genera that show increased proportions coincident with increased plasma levels of TMA also tend to show increased proportions coincident with plasma TMAO levels. Several bacterial taxa remained significantly associated with plasma TMA and/or TMAO levels after false discovery rate (FDR) adjustment for multiple comparisons (Fig. 2-3b). Further analyses examining the proportion of specific bacterial genera and mouse plasma TMA and TMAO levels revealed several taxa that significantly segregate with both mouse dietary groups and are associated with plasma TMA or TMA levels (Fig. 2-3c; Supplementary Fig. 2-11). Interestingly, a direct comparison of genera identified in humans versus mice that significantly associated with plasma TMAO levels failed to identify common genera, consistent with prior reports that microbes identified from the distal gut of the mouse represent genera that are typically not detected in humans 45,68. Plasma levels of L-carnitine associate with CVD We next investigated the relationship of fasting plasma levels of L-carnitine with CVD risks in an independent large cohort of stable subjects (n=2,595) undergoing elective cardiac evaluation. Patient demographics, laboratory values, and clinical characteristics are provided in Supplementary Table 2-2. A significant dose dependent association between L-carnitine levels and risk of prevalent coronary artery disease (CAD), peripheral artery disease (PAD), and overall CVD was noted (Fig. 2-4a-c). Moreover, the association of plasma L- carnitine levels with CAD, PAD and CVD remained significant following 60

63 adjustments for traditional CVD risk factors, including age, sex, history of diabetes mellitus, smoking, systolic blood pressure, and lipoproteins/lipids. In further analyses, plasma levels of L-carnitine were observed to be increased in subjects with significant ( 50% stenosis) angiographic evidence of CAD, regardless of the extent (e.g. single versus multi-vessel) of CAD, as revealed by diagnostic cardiac catheterization (Fig. 2-4d). Next, the relationship between baseline fasting plasma levels of L-carnitine and incident (3 year) risk for major adverse cardiac events (MACE = composite of death, MI, stroke, and revascularization) was examined. Elevated levels of L-carnitine (4 th quartile) remained an independent predictor of MACE even after adjusting for traditional CVD risk factors (Fig. 2-4e). After further adjustment for both TMAO and a larger number of comorbidities that might be known at time of presentation (extent of CAD, ejection fraction, medications, and estimated renal function), the significant relationship between carnitine and MACE risk was completely attenuated (Model 2) (Fig. 2-4e). Notably, the significant association between carnitine and incident cardiovascular event risks was observed in Cox regression models after multivariate adjustment, but only among those subjects with concurrent high plasma TMAO levels (Fig. 2-4f). Thus, while plasma levels of carnitine appear to be associated with prevalent and incident cardiovascular risks, the present results are consistent with TMAO, and not the dietary precursor carnitine, which serves as the primary driver of the association with cardiovascular risks (i.e. it is TMAO that may be the pro-atherogenic species). 61

64 Dietary L-carnitine in mice promotes atherosclerosis in a gut microbiota dependent manner We therefore next sought to investigate whether dietary L-carnitine had any impact on the extent of atherosclerosis in the presence vs. absence of TMAO formation in animal models. C57BL/6J, Apoe-/- mice were initially fed normal chow diet versus the same diet supplemented with L-carnitine from time of weaning. Aortic root atherosclerotic plaque quantification revealed approximately a doubling in disease burden compared to normal chow fed animals (Fig. 2-5a, b). Importantly, parallel studies in mice placed on oral antibiotic cocktail to suppress intestinal microflora showed marked reductions in plasma TMA and TMAO levels (Fig. 2-5c), as well as complete inhibition of the dietary L-carnitinedependent increase in atherosclerotic lesion burden (Fig. 2-5b). Analysis of plasma revealed that the increase in atherosclerotic plaque burden noted with dietary L-carnitine supplementation occurs in the absence of significant proatherogenic changes in plasma lipids, lipoproteins, glucose, or insulin levels; moreover, both biochemical and histological analyses of livers in the mice failed to demonstrate significant steatosis (Supplementary Table 2-3, 2-4; Supplementary Fig. 2-12). Quantification of plasma levels of L-carnitine in the mice revealed a significant increase in the L-carnitine fed animals versus the normal chow fed controls (Fig. 2-5c). Interestingly, an even higher increase in plasma L-carnitine levels was noted in mice supplemented with L-carnitine on the antibiotic arm of the study, which failed to show enhanced atherosclerosis. These results parallel what was observed with mice on carnitine supplemented diet 62

65 following the d3-(methyl)carnitine challenge (Fig. 2-3a), and are consistent with a major role for gut microbiota in the catabolism of dietary L-carnitine in the setting of chronic carnitine ingestion. They also suggest that it is not L-carnitine itself, but a down-stream (presumably) gut flora-dependent metabolite that promotes the increased atherosclerosis burden (Fig. 2-5c). Consistent with this hypothesis, whereas plasma levels of L-carnitine in the mice had no significant association with atherosclerotic disease burden (R=0.09(Spearman), P=0.59), plasma levels of both of its gut microbiota-dependent monitored metabolites, TMA (R=0.30, P<0.01) and TMAO (R=0.45, P<0.01), showed significant dose dependent associations with atherosclerotic plaque burden. Gut microbiota dependent formation of TMAO inhibits reverse cholesterol transport In recent studies we showed that TMAO can promote macrophage cholesterol accumulation in vivo in a gut microbiota dependent manner by increasing surface expression of scavenger receptors CD36 and SRA1 69, receptors previously shown to participate in atherosclerosis in murine models 87,88. In an effort to identify additional mechanisms through which TMAO may promote a proatherosclerotic phenotype, additional experiments were performed. We first noted that TMAO and its trimethylamine nutrient precursors are all quaternary amines, and thus have the potential to compete with the amino acid arginine for cellular uptake via cationic amino acid transporters. TMAO might thus hypothetically limit arginine bioavailability and hence, nitric oxide synthesis, 63

66 under conditions of elevated plasma TMAO and its dietary precursors. However, direct testing of this hypothesis in bovine aortic endothelial cells through competition studies using [ 14 C]arginine and (patho)physiologically relevant levels of TMAO and other trimethylamine containing compounds demonstrated no significant decrease in [ 14 C]arginine transport (Supplementary Fig 2-13). Taking note of the enhanced cholesterol accumulation within macrophages recovered from mice in the presence of either dietary (directly) or gut microbiotagenerated TMAO 69, we decided to next focus on the impact of TMAO on various aspects of cholesterol metabolism in vivo. Taking a "black box" approach, one can in general envision three non-exclusive mechanisms through which cholesterol can accumulate within cells of the artery wall such as a peripheral macrophage: (i) enhanced rate of flux in (as noted above, a mechanism already shown to occur with TMAO-induced increased surface levels of macrophage scavenger receptors SRA1 and CD36) 69 ; (ii) enhanced synthesis; or (iii) diminished rate of flux out. To test whether TMAO might alter the canonical down regulation of cholesterol biosynthesis genes attendant with macrophage cholesterol loading 5, several different sources of macrophages were loaded with cholesterol and suppression of cholesterol synthesis related genes and LDL receptor was confirmed. Concomitant addition of TMAO to media at physiologically relevant concentrations (corresponding to those observed in the top 1 percentile in patient plasma), however, failed to alter mrna levels of the LDL receptor or cholesterol synthesis genes (Supplementary Fig. 2-14). Parallel 64

67 studies examining desmosterol levels and macrophage inflammatory gene expression in the presence vs. absence of cholesterol loading, processes recently linked 89, failed to show any effect of TMAO within tissue culture media (Supplemental Figs. 2-14, 2-15). Turning our attention next to potential mechanisms of cholesterol removal from macrophages (i.e. diminished rates of cholesterol efflux), we sought to test the hypothesis that dietary sources of TMAO inhibit reverse cholesterol transport (RCT) in vivo using an adaptation of the model system first described by Rader and colleagues 90. Mice were placed on normal chow or diets supplemented with either carnitine or choline. After several weeks of diet, [ 14 C]cholesterol-loaded peripheral macrophages were injected subcutaneously into the different groups of mice and RCT quantified by counting fecal radiolabel cholesterol, as described under Methods. Remarkably, a significant (~30%, P<0.05) decrease in RCT was observed in mice on either the choline or carnitine supplemented diets compared to normal chow controls (Fig. 2-5d, left panel). Furthermore, suppression of gut microbiota (and plasma TMAO levels) with oral broad spectrum antibiotics completely blocked the diet-dependent (for both choline and carnitine) suppression of RCT in vivo (Fig. 2-5d, middle panel), suggesting that a gut flora-generated product (e.g. TMAO) inhibits RCT in vivo (Supplementary Fig. 2-16). To test this hypothesis, in a separate series of studies mice were placed on either normal chow versus a diet supplemented with TMAO, and after several weeks, [ 14 C]cholesterol-loaded peripheral macrophages were injected 65

68 subcutaneously for RCT quantification. Analyses of fecal [ 14 C]sterol levels showed significant reduction in mice on the TMAO containing diet (35% decrease relative to normal chow, P<0.05) (Fig. 2-5d, right panel). Further examination of plasma, liver and bile compartments in the normal chow versus TMAO supplemented diet mice demonstrated significant reduction in [ 14 C]cholesterol recovered within plasma of the TMAO fed mice (16%, p<0.05), but no significant changes in counts recovered within the liver or bile (Supplementary Fig. 2-17). TMAO promotes significant alterations in cholesterol and sterol metabolism in multiple compartments in vivo In an effort to better understand potential molecular mechanism(s) through which TMAO reduces RCT in vivo, we examined candidate genes and biological processes in multiple compartments (i.e. macrophage, plasma, liver, intestine) known to participate in cholesterol/sterol metabolism and RCT. Mouse peritoneal macrophages recovered from C57BL/6J mice were exposed to TMAO in vitro and mrna levels of cholesterol transporters ATP-binding cassette, sub-family A (ABC1), member 1 (Abca1), scavenger receptor class B, member 1(Srb1) and ATP-binding cassette, sub-family G, member 1 (Abcg1) were examined. Modest but statistically significant increases in expression of Abca1 and Abcg1 were noted (P < 0.05) (Supplementary Fig. 2-18). Parallel examination of plasma recovered from both groups of mice showed no significant differences in total cholesterol and HDL cholesterol concentrations (Supplementary Table 2-5). To assess the potential biological significance of the modest TMAO-induced 66

69 changes in macrophage cholesterol transporter mrna levels observed, parallel studies were performed quantifying cholesterol efflux from [ 14 C]cholesterolloaded macrophages cultured in media in the absence vs. presence of TMAO. Modest but statistically significant increases in cholesterol efflux to apolipoprotein A1, but not to HDL, as cholesterol acceptor were noted in the macrophages cultured in vitro in the presence of TMAO (Supplementary Fig. 2-19). Collectively, these results show that TMAO promotes modest changes in macrophage cholesterol transporter expression and function both in vitro and in vivo; however, the directionality of the modest changes induced by TMAO are opposite to what one would expect inasmuch as an increase in these transporters cannot account for the observed significant global reductions in RCT in vivo induced by TMAO. In parallel studies, we examined the expression levels of cholesterol transporters (i.e. Sr-b1, Abca1, Abcg1, Abcg5, Abcg8, and Shp) within mouse liver between the normal chow vs. TMAO dietary groups. No significant differences were noted (Supplementary Fig. 2-20, 2-21). In contrast, liver expression of the key bile acid synthetic enzymes Cyp7a1 and Cyp27a1 showed significant reductions in mice supplemented with dietary TMAO (p<0.05 each; Fig. 2-5e;). Interestingly, dietary supplementation of TMAO did not decrease expression of Shp, an upstream regulator of Cyp7a1, suggesting another upstream target for TMAO (Supplementary Fig. 2-20). Bile acid transporters in the liver (e.g. Oatp1, Oatp4, Bsep, Mrp2, Ephx1/mEH, and Ntcp) also showed a dietary TMAO-induced 67

70 decrease in expression (p<0.05 each; Fig. 2-5f). Despite these TMAO-induced changes in mouse liver, no significant differences in bile acid transporter expression in the gut were noted between dietary groups (Supplementary Fig. 2-22). Taken together, these data suggest that the gut flora dependent metabolite TMAO fosters significant alterations in a major pathway for cholesterol elimination from the body, the bile acid synthetic pathway. To confirm these findings, the total bile acid pool size was examined. Mice supplemented with TMAO showed significant decreases in the total bile acid pool size (Fig. 2-6a). Dietary supplementation with TMAO also markedly reduced expression of both intestinal cholesterol transporters Npc1L1 (transports cholesterol into enterocyte from the gut lumen 91 ; and Abcg5/8 (transports cholesterol out of enterocyte into gut lumen 91 ; (Supplementary Figure 2-23). Previous studies with either Cyp7a1 or Cyp27a1 null mice have demonstrated a reduction in cholesterol absorption. In separate studies, dietary TMAO supplementation similarly promoted a reduction (26%, P <0.01) in total cholesterol absorption (Fig. 2-6b). Discussion L-carnitine has been studied for more than a century since its initial discovery in 1905 from muscle extracts 92. Although eukaryotic organisms can endogenously produce L-carnitine, only prokaryotic organisms have known metabolic pathways that can catabolize L-carnitine 82. While a role for gut microbiota in TMAO production from dietary carnitine has been suggested from studies in rats, and TMAO production from several dietary trimethylamine containing compounds has 68

71 been suggested in humans, a role for gut microbiota in production of TMAO from dietary L-carnitine in humans has not yet been demonstrated The present studies reveal an obligatory role of gut flora in the production of TMAO from orally ingested L-carnitine in humans (Fig. 2-6c). They also reveal an additional potential nutritional basis in the pathogenesis of CVD that involves dietary L- carnitine, an abundant nutrient in meat, the intestinal microbial community, and production of the recently identified pro-atherosclerotic down-stream metabolite, TMAO. Finally, they show that the gut flora dependent metabolite, TMAO, impacts multiple distinct compartments and processes involving cholesterol and sterol metabolism in vivo, with net increase in atherosclerosis in vivo through a combination of both enhanced forward cholesterol transport into macrophages and reduced reverse cholesterol transport (Fig. 2-6c). The present studies also suggest a mechanistic rationale for the observed relationship between dietary red meat ingestion and accelerated atherosclerosis. Although L-carnitine is endogenously produced in all mammals, consuming foods rich in L-carnitine (predominantly red meat and to a lesser extent dairy products) can significantly increase fasting human L-carnitine plasma levels 96. Meats and full fat dairy products are abundant foods in the Western diet and excess consumption of these is commonly cited as a major contributor to CVD morbidity and death worldwide. Moreover, numerous studies have suggested a decrease in atherosclerotic disease risk in vegan/vegetarian individuals when compared to omnivorous subjects Together, L-carnitine and choline containing lipids can 69

72 constitute up to 2% 84,85,100 of these foods, suggesting that gut flora dependent production of TMAO may have a significant contributory role in the pathogenesis of atherosclerosis, particularly in omnivorous subjects. Despite the elaboration of this new diet - gut microbiota- host interaction as it relates to CVD pathogenesis and TMAO formation, the molecular mechanism(s) accounting for how TMAO promotes acceleration of atherosclerosis in vivo are only partially illuminated. As shown in the present studies, one potential mechanism is through reduction in RCT. Both dietary carnitine and choline each promoted a significant reduction in RCT in vivo, but only in the presence of intact intestinal microbiota when TMAO was produced (Fig. 2-5b). Importantly, suppression of intestinal microbiota and TMAO production completely eliminated the diet-dependent inhibition in RCT with both choline and carnitine supplementation, and dietary supplementation with TMAO directly promoted a similar ~30-35% reduction in RCT in vivo. These results are thus consistent with a gut microbiota dependent mechanism in the setting of specific dietary exposures (such as a diet rich in carnitine and total choline) whereby generation of TMAO impairs RCT in vivo and contributes to a pro-atherosclerotic phenotype. One additional mechanism through which TMAO may contribute to accelerated atherosclerosis is by influencing macrophage cholesterol metabolism, leading to cholesterol deposition and foam cell formation, since macrophages from TMAO supplemented mice also demonstrate significant increases in scavenger receptor expression (SRA1 and CD36) 69 (Fig. 2-6c). Within the macrophage, TMAO does 70

73 not appear to alter desmosterol levels, cholesterol biosynthetic enzyme expression levels, or LDL receptor expression levels, and thus does not appear to directly impact either the regulation of cholesterol biosynthetic and uptake pathways initially reported by Brown and Goldstein 4,5, or the more recently described regulatory role of desmosterol by Glass and colleagues in integrating macrophage lipid metabolism and inflammatory gene responses 89. Within the liver, a consistent finding observed to be associated with elevated TMAO levels is decreased bile acid pool size and altered composition, as well as reduction in key bile acid synthesis and transport proteins (Figs. 2-5, 2-6). However, whether these changes contribute to the reductions in RCT in vivo that accompany TMAO supplementation are unclear. They are consistent with reports that human genetic variants in the Cyp7a1 gene, the major bile acid synthetic enzyme and rate limiting step in the catabolism of cholesterol, are associated with reduced bile acid synthesis, elevated plasma cholesterol levels refractory to statin therapy, decreased bile acid secretion into the intestines and enhanced atherosclerosis Further, up- (as opposed to down-) regulation of Cyp7a1 is reported to lead to an expansion of the bile acid pool size, increased RCT, and reduced atherosclerotic plaque in susceptible mice Moreover, an overall increase in bile acid secretion via alternative mechanisms has been reported to be associated with reduced atherosclerosis and an increase in reverse cholesterol transport 105. Within the intestines, TMAO again was associated with marked changes in cholesterol metabolism (Fig. 2-6), but the significant reductions in cholesterol absorption observed, while consistent with the reduction 71

74 in intestinal Npc1L1 107 (and hepatic Cyp7a1 and Cyp27a1 108,109 ), cannot explain the reproducible reduction in RCT observed in mice supplemented with TMAO. Thus, the molecular mechanisms through which the gut microbiota TMAO pathway inhibits RCT in vivo are not entirely clear, and whether there exist additional mechanisms through which TMAO exerts a pro-atherosclerotic effect remains to be determined. Finally, it is not known whether TMAO interacts with a specific receptor(s) directly to promote the many observed biological effects, or whether it acts to alter signaling pathways indirectly by altering protein conformation (i.e., via allosteric effects). A small quaternary amine with some aliphatic character, TMAO is reportedly capable of directly inducing conformational changes in proteins, including both stabilization of protein folding, and functioning as a small molecule protein chaperone 110,111. It is also of interest that recent studies show that TMA can influence signal transduction by direct interaction with a family of G protein-coupled receptors 112,113. It is thus conceivable that TMAO may potentially alter a multitude of signaling pathways without directly acting at a TMAO receptor. One of the more remarkable finding of the present studies is the magnitude with which long term preceding dietary habits impacts TMAO synthetic capacity in both humans (i.e. vegan/vegetarian vs. omnivore) and mice (normal chow vs. chronic carnitine supplementation). Microbial composition analyses from both humans (fecal) and mice (cecal) revealed specific taxa that segregated with both preceding dietary status and plasma TMAO levels. Recent studies have shown 72

75 that significant global changes in gut microbial composition, or "enterotype" (i.e. the clustering of microbial communities), are associated with long-term dietary changes 49, and indeed, we observed that plasma TMAO levels were significantly different within subjects segregated according to prior reported enterotypes (Fig. 2-2e). Using a combination of studies involving germ-free mice, as well as in both humans and mice before vs. following suppression of intestinal microflora using a cocktail of poorly absorbed antibiotics, an obligatory role for gut microbiota in TMAO formation from dietary carnitine was shown. The marked differences observed in TMAO production following an "L-carnitine challenge" within omnivore versus vegan subjects (Fig. 2-2) is striking, consistent with the observed differences in microbial community composition. Recent reports have shown significant differences in microbial communities among vegetarians and vegans versus those who commonly consume animal proteins in their diet 114. Of note, we observed a significant increase in baseline plasma TMAO concentrations in what historically was called enterotype 2 (Prevotella), a relatively rare enterotype that previously in one study was associated with low animal fat and protein consumption 49. Notably, in our study, 3 of the 4 individuals classified into enterotype 2 are self-identified omnivores suggesting more complexity in the human gut microbiome perhaps than anticipated with only a few enterotypes. Indeed, other studies have demonstrated variable results in associating human bacterial genera, including Bacteroides and Prevotella, to omnivorous and vegetarian eating habits 86,115. This complexity is no doubt in part attributed to the fact that there are many species within any genus and distinct 73

76 species within the same genus may have different capacity to use carnitine as a fuel and form TMA. Indeed, prior studies have suggested that multiple bacterial strains can metabolize carnitine in culture 116, and by analogy, comparison of distinct species within the genus Clostridium reveals some that are capable and others not of using choline as the sole source of carbon and nitrogen in culture 117. A search of the Prevotella genus, for example, reveals ~250 known species (NCBI search ). The present studies, coupled with the demonstration of both inducibility of enhanced L-carnitine metabolism and TMAO production with antecedent L-carnitine feeding, and the association of bacterial taxa that associate with a carnitine enriched diet (Figs 2-2, 2-3), suggests that multiple proatherogenic (i.e. TMA/TMAO producing) species, likely exist. Consistent with this supposition, others have reported that several bacterial phylotypes are associated with a history of atherosclerosis, and that the human gut flora biodiversity may at least in part be influenced by carnivorous eating habits 45,49,118. The association observed between plasma carnitine levels and both prevalent and incident cardiovascular risks further supports the potential physiological importance of the carnitine gut microbiota TMA/TMAO atherosclerosis pathway. Of note, the association remained significant even following adjustments for traditional cardiovascular risk factors and comorbidities. The significance of this association was only attenuated (becoming completely nonsignificant) following addition of plasma TMAO levels to the model. These 74

77 findings are consistent with the proposed mechanism whereby the association of carnitine with atherosclerotic cardiovascular disease risks is mediated via the gut microbiota metabolite TMAO, and not the dietary nutrient carnitine itself. Further, we are tempted to speculate that similar to the increased sensitivity observed with use of an oral glucose tolerance test versus a fasting plasma glucose level in the diagnosis of diabetes, it is possible that use of a provocative challenge test involving a defined oral load of isotope labeled L-carnitine alone or in combination with other trimethylamine precursor nutrients like phosphatidylcholine or choline, has a greater potential to identify those at increased risk for cardiovascular disease over fasting plasma TMAO levels alone. A provocative oral challenge test following isotope labeled L-carnitine administration may also better allow one to characterize and identify microbial communities most likely to promote disease. There are several reports of specific intestinal anaerobic and aerobic prokaryotic bacterial species that can utilize L- carnitine as a carbon nitrogen source 81,82,119. Based upon the present studies, one might speculate that a microbial composition that has adapted to produce more TMA/TMAO may equip the host with greater potential to develop enhanced atherosclerotic disease burden in the setting of a diet rich in trimethylamine containing nutrients. It logically follows, but remains to be proven, that development of a prebiotic or probiotic intervention that alters microbial compositions associated with enhanced TMAO production may serve as an alternative therapy for the treatment or prevention of atherosclerotic disease. 75

78 L-carnitine has indispensable roles in animal metabolism. It is essential in the import of activated long chain fatty acids from the cytoplasm into mitochondria for β-oxidation. It also participates in the transport of intermediate and short chain organic acids from peroxisomes into mitochondria, functions as a reservoir of activated acetyl groups, and impacts upon crucial steps of intermediate metabolism 81,82. As a consequence of these critical roles, L-carnitine supplementation has been widely studied and therefore merits some comment. There are case reports of L-carnitine supplementation showing benefit in terms of symptomatic improvement for individuals with inherited primary and acquired secondary L-carnitine deficiency syndromes 83. L-Carnitine has also been suggested for treatment of subjects with end stage renal disease undergoing hemodialysis, as they commonly acquire a secondary L-carnitine deficiency that may participate in several dialysis-related symptoms including muscle weakness and diminished exercise capacity. While some of these studies have shown improvement following supplementation, others have yielded conflicting results, possibly in part because of heterogeneity in the route of administration of L- carnitine amongst other factors 120,121. Oral treatment with L-carnitine (1gram) in end stage renal disease patients undergoing hemodialysis over a brief period has been shown to raise plasma L-carnitine (pre-dialysis) to normal levels, but with accompanying substantial increases in plasma TMAO to supraphysiological levels. A broader potential therapeutic scope for L-carnitine and two related metabolites, acetyl-l-carnitine and propionyl-l-carnitine, has also been explored in the treatment of acute ischemic events including myocardial infarction and 76

79 stroke, as well as for chronic treatment of a multitude of cardio-metabolic disorders like PAD, congestive heart failure and diabetes 121. Here too, results from studies are conflicting. One potential explanation for the discrepant findings of various intervention studies may be explained in part by variations in the route of administration and the length of time of L-carnitine dosing. Many studies have provided one of the L-carnitines over short intervals of treatment, and often in part via parenteral route, bypassing the gut flora. The obligatory role of the gut flora in the promotion of TMAO production and atherosclerotic disease enhancement observed in the present studies likely also explains the apparent contradictory report from Sayed-Ahmed et al. that showed intraperitoneal administration of L-carnitine reduced atherosclerotic lesion in the hypercholesterolemic rabbit model through unclear mechanisms 122. There are also a number of studies showing long term treatment with Mildronate, an inhibitor of L-carnitine synthesis, can both reduce atherosclerosis and promote cardio-protective effects 123,124. Carnitine metabolism is clearly complex, and administration of the L-carnitine vs. acetyl or proprionyl carnitine forms may not elicit the same responses. Discovery of a link between oral carnitine ingestion and cardiovascular disease risks has broad health related implications. The results of the present studies underscore the need for further examination of the safety of chronic oral L- carnitine supplementation. They also argue for careful attention to route of administration when designing and comparing carnitine intervention 77

80 studies/strategies. Lastly, the present studies raise the possibility that chronic ingestion of high amounts of carnitine through either supplements and/or carnivorous eating habits may under some conditions prime our gut microbiota to become proatherogenic. Further studies on the long term health impact of increased levels of carnitine ingestion are needed. Acknowledgements We thank L. Kerchenski and C. Stevenson for assistance in performing the clinical studies; A. Pratt, S. Neale, M. Pepoy, and B. Sullivan for technical assistance with human specimen processing and routine clinical diagnostic testing; E. Klipfell, F. McNally, and M. Berk for technical assistance, and the subjects who consented to participate in these studies. Mass Spectrometry instrumentation used was housed within the Cleveland Clinic Mass Spectrometry Facility with partial support through a Center of Innovation by AB SCIEX. Germ free animals used were obtained from the University of North Carolina Gnotobiotic Facility, which is supported by P30-DK and P40- RR Methods Materials and general procedures C57BL/6J, Apoe / and C57BL/6J mice were obtained from Jackson Laboratories. All animal studies were performed under approval of the Animal Research Committee of the Cleveland Clinic. Mouse plasma total cholesterol and 78

81 triglycerides, and human fasting lipid profiles, glucose, creatinine, and high sensitivity C-reactive protein levels were assayed using the Abbott ARCHITECT platform model ci8200 (Abbott Diagnostics, Abbott Park, IL). Mouse HDL cholesterol was determined enzymatically (Stan bio, Houston, TX) from mouse plasma HDL isolated using density ultracentrifugation. Mouse plasma insulin measurements were performed using the Mercodia Mouse Insulin Elisa Kit (Uppsala, Sweden). Human plasma MPO levels were measured using the US Food and Drug Administration-cleared CardioMPO test (Cleveland Heart Lab, Inc., Cleveland, OH). Liver triglyceride content was quantified by GPO reagent (Pointe Scientific, Canton, MI) and normalized to liver weight in grams as described 125. Liver cholesterol was quantified in liver homoginates in which coprostanol (Steraloids, Inc, Newport, RI) was added as an internal standard, lipids extracted by the Folch method (chloroform:methanol (2:1, v/v)), and then cholesterol quantified as its trimethylsilane (TMS) derivative (Sylon HTP, Sigma- Aldrich, Sigma St. Louis, MO) by GC/MS (Agilent 5973N model, Santa Clara CA) on a DB-1 column (12m x 0.2 mm diameter x 0.250um film thickness) 126,127. Research subjects Two cohorts of subjects were used in the present studies. The first group of volunteers had extensive dietary questioning, and stool, plasma and urine collection. A subset of subjects with stool collected also underwent oral carnitine challenge testing (n = 5 omnivores and n = 5 vegans), consisting of d3(methyl)carnitine (250 mg within a veggie capsule). Where indicated, 79

82 additional omnivores, and an individual vegan, also underwent carnitine challenge testing with combined ingestion of the synthetic d3-carnitine capsule (250 mg) and an 8 ounce steak (within 10 minutes). Male and female volunteers gave written informed consent and included individuals of at least 18 years of age. Volunteers were excluded if they were pregnant, had chronic illness (including a known history of heart failure, renal failure, pulmonary disease, gastrointestinal disorders, or hematologic diseases), an active infection, received antibiotics within 2 months of study enrollment, used any over the counter or prescriptive probiotic or bowel cleansing preparation within the past 2 months, ingested yogurt within the past 7 days, or had undergone bariatric or other intestinal (e.g. gall bladder removal, bowel resection) surgery. The second cohort of subjects (n = 2,595) were from GeneBank, a research registry of sequential consenting stable subjects undergoing elective cardiac evaluation and who were subsequently followed longitudinally for incident cardiovascular disease (CVD) outcomes 69,128. Patients with a recent (< 4 weeks) clinical history of myocardial infarction or elevated troponin I (> 0.03 md dl 1 ) at enrollment were excluded from the study. CVD was clinically defined as having a previous history of coronary artery disease (CAD), peripheral artery disease (PAD), and/or cerebral vascular disease (history of a transient ischemic attack or cerebrovascular accident), history of revascularization (coronary artery bypass graft, angioplasty, or stent) or significant angiographic evidence of CAD ( 50% stenosis) in at least one major coronary artery. Subjects with CAD included patients with diagnoses of stable or unstable angina, myocardial infarction, 80

83 history of coronary revascularization, or angiographic evidence of 50% stenosis of one or more major coronary arteries. PAD was defined as subjects having clinical evidence of extra-coronary atherosclerosis. Medications were documented by patient interview and chart review. All subjects gave written informed consent. The Institutional Review Board of the Cleveland Clinic approved all study protocols. General Statistics The Student s t-test or the Wilcoxon Rank-Sum test for continuous variables was used for two-group comparison. The analysis of variance (ANOVA, if normally distributed) or Kruskal-Wallis test (if not normally distributed) was used for multiple group comparisons of continuous variables and a Chi-square test was used for categorical variables. Odds ratios for cardiac phenotypes (CAD, PAD, and CVD) and corresponding 95% confidence intervals were calculated using logistic regression models. Kaplan Meier analysis with Cox proportional hazards regression was used for time-to-event analysis to determine Hazard ratio (HR) and 95% confidence intervals (95%CI) for adverse cardiac events (death, MI, stroke, and revascularization). Adjustments were made for individual traditional cardiac risk factors (age, gender, diabetes mellitus, systolic blood pressure, former or current cigarette smoking, low-density lipoprotein cholesterol, highdensity lipoprotein cholesterol), extent of CAD, left ventricular ejection fraction, history of MI, baseline medications (aspirin, statins, β-blockers, and ACE inhibitors), and renal function by estimated creatinine clearance. A robust 81

84 Hotelling T 2 test was used to examine the difference in the proportion of specific bacterial genera along with subject TMAO levels between the different dietary groups. All data was analyzed using R software version 2.15 and Prism (Graphpad Software, San Diego, CA). Metabolomics study In a previous study we reported results from a metabolomics study in which small molecule analytes were sought that associated with cardiovascular risks 69. The metabolomics study design used a two stage screening strategy. In the first phase, totally unbiased metabolomics studies were performed on randomly selected plasma samples from a Learning Cohort generated from Genebank subjects that was comprised of 50 cases, defined as those that experienced a major adverse cardiovascular event (defined as non-fatal myocardial infarction, stroke, or death) in the 3 year period following enrollment, versus age- and gender-matched controls (n = 50) that did not experience an event. A second phase (Validation Cohort) of unbiased metabolomics analyses were performed on an independent (non-overlapping) cohort of cases (n = 25) and age- and gender-matched controls (n = 25) using identical inclusion / exclusion criteria 1. Analytes were only known for their m/z and retention time, with identities unknown. Analytes considered of interest in the metabolomics studies were selected based on the following criteria: (i) the unknown analyte had a significant difference (P < 0.05) between cases and controls in the Learning and Validation Cohorts after a two sided Bonferroni adjusted t-test; (ii) the unknown analyte had 82

85 a significant (P < 0.05) dose-response relationship between analyte peak area and major adverse cardiovascular event risk using an unadjusted log-rank test of trend; and (iii) to facilitate future quantification and structural identification efforts, analytes had to have a signal-to-noise ratio of 5:1 in at least 75% of subjects within cases and controls of both the Learning and Validation Cohorts as previously described 69. While an ion with m/z of 162 and retention time identical to carnitine was not among the top analytes identified in the above metabolomics studies, we attempted for the present studies to examine the original data, this time using less stringent criterion, with the hypothesis-generated focus of examining just the single ion that had chromatographic and mass spectral characteristics observed identical with standard L-carnitine: namely, m/z = 162 and appropriate retention time. Examination of Supplementary Table 2-1 shows that an analyte with appropriate m/z and retention time in the Learning and Validation Cohorts was observed that failed to meet significance using the originally used stringent criteria (the more strict Bonferroni adjusted p value). However, reexamination of the combined Learning and Validation cohorts (n = 75 cases, n = 75 controls) without adjustment for multiple testing (since only one analyte here was being screened for) showed the unknown analyte giving rise to an ion at m/z = 162 with retention time identical to carnitine was associated with atherosclerotic disease outcomes. Identification of L-carnitine and d9-carnitine preparation 83

86 Matching CID spectra of the unknown metabolite of interest with a precursor ion at m/z = 162 and authentic L-carnitine standard were examined using a Cohesive Technologies Aria HPLC interfaced to a AB Sciex API 5000 triple quadrupole mass spectrometer (Applied Biosystems) in positive ion mode by a method described previously 69. Mouse and human samples were also spiked with synthetic d9(trimethyl)-carnitine as internal standard. Samples were analyzed using a similar system as above except a Shimadzu (Columbia, MD) dual gradient HPLC system was interfaced to the AB Sciex API 5000 tandem mass spectrometer. Multiple distinct parent product ion transitions specific for the natural abundance and d9-isotopologue of carnitine were monitored simultaneously in the spiked sample, to determine if multiple characteristic MRM channels for each isotopologue of carnitine were present and cochromatographed. Synthesis of the d9-carnitine standard for the above experiment, and for use as internal standard in stable isotope dilution LC/MS/MS analyses of carnitine and synthetic d3-carnitine following a carnitine challenge, were prepared and characterized as follows: First, 3-hydroxy-4-aminobutyric acid (Chem-Impex Intl.) was dissolved in methanol and reacted with d3-methyl iodide (Cambridge Isotope Labs, Boston, MA) in the presence of potassium hydrogen carbonate to give d9-carnitine, as per Chen and Benoiton 129. The d9-l-carnitine was isolated by passing the reaction mixture directly over a silica gel column rinsing with additional methanol, and then eluting the heavy isotope labeled L- carnitine with 30% v/v water in methanol. The product was dried via azeotropic 84

87 distillation of absolute ethanol and subsequently recrystallized from ethanol and acetone. The white to off-white crystalline product was dried over P 2 O 5 in vacuo and stored refrigerated. TLC on silica gel eluted with methanol plus 0.2%v/v formic acid visualized by iodine staining showed one spot with the same Rf as L- carnitine. The mass spectrum of the compound dissolved in 50% v/v methanol / water (5 mm formic acid) to a concentration of 50 µg ml 1 exhibited a base peak at m/z = 171 in the positive ion mode corresponding to [M]+. CID fragments peaks were observed at m/z = 111,103, 85, 69, 57, and 43. Mass spectral fragmentation patterns and m/z ratios are consistent with the L-carnitine except for those fragment ions that contain the trimethylammonium group; these ions exhibit fragments 9 atomic mass units (amu) higher than the corresponding signals from L-carnitine due to the incorporation of 9 deuterium atoms on the methyl groups attached to the nitrogen. Quantification of TMAO, TMA, and L-carnitine Stable isotope dilution LC/MS/MS was used to quantify TMAO, TMA and L- carnitine from acidified plasma samples in positive MRM mode. Precursor product ion transitions at m/z 76 to 58, m/z 162 to 60 and m/z 60 to 44 were used for TMAO, L-carnitine, and TMA respectively. As internal standards, d9(trimethyl)tmao (d9-tmao), d9(trimethyl)carnitine (d9-carnitine), and d9(trimethyl)tma (d9-tma) were added to mouse and human plasma samples for their respective native compounds. Increasing concentrations of L-carnitine, TMA and TMAO standards with a fixed amount of internal standard were added 85

88 to human control plasma to generate calibration curves for determining plasma concentrations of each analyte, using methods similar in approach to that previously described 69, with samples run on an AB Sciex API 5000 triple quadrupole mass spectrometer. Human microbiota analyses Stool samples were stored at 80 o C and DNA for the gene encoding 16SrRNA was isolated using the MoBio PowerSoil kit (Carlsbad, CA) according to the manufacturer's instructions. DNA samples were amplified using V1-V2 region primers targeting bacterial 16S genes and sequenced using 454/Roche Titanium technology. Sequence reads from this study are available from the Sequence Read Archive (CaFE: SRX037803, SRX021237, SRX021236, SRX020772, SRX020771, SRX020588, SRX020587, SRX020379, SRX (metagenomic). COMBO: SRX020773, SRX020770). Overall association between TMAO measurements and microbiome compositions was assessed using PermanovaG 130 by combining both the weighted and unweighted UniFrac distances. Associations between TMAO measurements and individual taxa proportions were assessed by Spearman's rank correlation test. False discovery rate (FDR) control based on the Benjamini Hochberg procedure was used to account for multiple comparisons when evaluating these associations. Each of the samples was assigned to an enterotype category based on their microbiome distances (Jensen-Shannon distance) to the medoids of the enterotype clusters as defined in the COMBO data 131. Association between enterotypes and TMAO 86

89 level was assessed by Wilcoxon rank sum test. Student s t-test was used to test the difference in means of TMAO level between omnivores and vegans. A robust Hotelling T 2 test 132 was used to examine the association between both the proportion of specific bacterial taxa and TMAO levels in groups using R software version Mouse microbiota analysis Microbial community composition in mouse cecal contents was assessed by pyrosequencing 16S rrna genes derived from the mice of normal chow diet (n = 11) and L-carnitine diet (n = 13). DNA was isolated using the MoBio PowerSoil DNA Isolation Kit (Carlsbad, CA). The V4 region of the 16S ribosomal DNA gene was amplified using bar-coded fusion primers (F515/R806) with the 454 A Titanium sequencing adapter. The barcoded primers were achieved following the protocol described by Hamady et al 133. Sample preparation was performed similarly to that described by Costello et al Each sample was amplified in triplicate, combined in equal amounts and cleaned using the PCR clean-up kit (Mo Bio, Carlsbad, CA). Cleaned amplicons were quantified using Picogreen dsdna reagent (Invitrogen, Grand Island, NY) before sequencing using 454 GS FLX titanium chemistry at the EnGenCore Facility at the University of South Carolina. The raw data from the 454 pyrosequencing machine were first processed through a quality filter that removed sequence reads that did not meet the quality criteria. Sequences were removed if they were shorter than 200 nucleotides, longer than 1,000 nucleotides, contained primer mismatches, 87

90 ambiguous bases, uncorrectable barcodes, or homopolymer runs in excess of six bases. The remaining sequences were analyzed using the open source software package Quantitative Insights Into Microbial Ecology (QIIME 135,136 ). A total of quality filtered reads were obtained from 23 samples (1 sample was removed due to low number of sequences). Individual reads that passed filtering were distributed to each sample based on bar-code sequences. De-multiplexed sequences were assigned to operational taxonomic units (OTUs) using UCLUST with a threshold of 97% pair-wise identity. Representative sequences were selected and BLASTed against a reference Greengenes reference database. For each resulting OTU, a representative sequences were selected by choosing the most abundant sequence from the original post-quality filtered sequence collection. The taxonomic composition was assigned to the representative sequence of each OTU using Ribosomal Database Project (RDP) Classifier The relative abundances of bacteria at each taxonomic level (e.g., phylum, class, order, family and genus) were computed for each mouse. For tree-based analyses, a single representative sequence for each OTU was aligned using PyNAST 138, then a phylogenetic tree was built using FastTree. The phylogenetic tree was used to measure the β-diversity (using unweighted UniFrac) of samples 139. Two-way ANOVAs were conducted to evaluate the effects of diet with P values corrected for multiple comparisons. Spearman correlations between relative abundance of gut microbiota and TMA and TMAO levels and association testing were performed in R. False discovery rates (FDR) of the multiple comparisons were estimated for each taxon based on the P- 88

91 values resulted from correlation estimates. A robust Hotelling T 2 test 132 was used to examine the association between both the proportion of specific bacterial taxa and mouse plasma TMA/TMAO levels in groups using R software version Aortic root lesion quantification Apolipoprotein E knockout mice on C57BL/6J background (C57BL/6J, Apoe / ) were weaned at 28 days of age and placed on a standard chow control diet (Teklad 2018). L-carnitine was introduced into the diet by supplementing mouse drinking water with 1.3% L-carnitine (Chem-Impex Intl.), 1.3% L-carnitine and antibiotics, or antibiotics respectively. The antibiotic cocktail dissolved in mouse drinking water has previously been shown to suppress commensal gut microbiota, and included 0.1% Ampicillin sodium salt (Fisher Scientific), 0.1% Metronidazole, 0.05% Vancomycin (Chem Impex Intl.), and 0.1% Neomycin sulfate (Gibco) 35,69. Mice were anaesthetized with Ketamine/Xylazine before terminal bleeding by cardiac puncture to collect blood. Mouse hearts were fixed and stored in 10% neutral buffered formalin before being frozen in OCT for sectioning. Aortic root slides were stained with Oil-red-O and counterstained with Haematoxylin. The aortic root atherosclerotic lesion area was quantified as the mean of sequential sections of 6 microns approximately 100 microns apart 69. Human L-carnitine challenge test and d3-l-carnitine preparation Consented adult men and women fasted overnight (12 hours) before performing the "L-carnitine challenge test", which involved baseline blood and spot urine 89

92 collection, and then oral ingestion (T = 0 at time of initial ingestion) of a veggie caps capsules containing 250 mg of a stable isotope labeled d3-l-carnitne (under Investigational New Drug exemption). Where indicated, for a subset of subjects, the carnitine challenge also included a natural source of L-carnitine (an 8 ounce sirloin steak cooked medium on a George Forman Grill) in a 10 minute period concurrent with taking the capsule containing the d3-carnitine. After combined ingestion of the steak and d3-l-carnitine, sequential venous serial blood draws were performed at respective time points, and a 24 hour urine collection was performed. An ensuing 1 week treatment period of oral antibiotics (Metronidazole 500 mg bid, Ciprofloxacin 500 mg bid) was given to suppress intestinal microbiota that use carnitine to form TMA and TMAO before repeating the L-carnitine challenge. After at least 3 weeks off of all antibiotics to allow reacquisition of intestinal microbiota, a third and final L-carnitine challenge test was performed. Dietary habits (vegan vs ominivore) were determined using a questionnaire assessment of dietary L-carnitine intake, similar to that conducted by the Atherosclerotic Risk in Community (ARIC) study 140. Synthesis of d3-l-carnitine for carnitine challenge tests was prepared and characterized as follows: L-Norcarnitine (3-hydroxy-4-dimethylaminobutyric acid) was prepared from L-carnitine (Chem Impex International, Woodale, IL) with thiophenol (Sigma Aldrich Milwaukee, WI) in N,N-dimethylaminoethanol (Sigma Aldrich Milwaukee, WI) and subsequently converted to its sodium salt with sodium hydroxide by the method of Colucci, et. al Sodium L-norcarnitine was 90

93 recrystallized three times from ethanol and 3 volumes of ethyl acetate prior to the subsequent conversion to d3-l-carnitine. TLC on silica gel eluted with methanol plus 0.2%v/v formic acid visualized by iodine staining showed one major spot with a higher Rf (> 0.1) than L-carnitine. 600MHz 1H-NMR (10 mg ml 1 in D 2 O): δ 2.1ppm (singlet, 6H), δ 2.2ppm (complex multiplet, 3H), δ 2.3ppm (complex multiplet, 1H) δ 4.0ppm (complex multiplet, 1H). The mass spectrum of the compound dissolved in 50% v/v, methanol/water (5 mm formic acid) to a concentration of 50 µg ml 1 exhibited a base peak at m/z = 148 in the positive ion mode corresponding to [M+H]+. CID fragments peaks were observed at m/z = 130,112, 94, 88, 85, 84(base) 82, 71, 69, 58, 57, 56, and 43. Sodium L- norcarnitine was dissolved in methanol and reacted with d3-methyl iodide (Cambridge Isotope Labs, Boston, MA) in the presence of potassium hydrogen carbonate to give d3-l-carnitine as per Chen and Benoiton 129. The d3-l-carnitine was isolated by passing the reaction mixture directly over a silica gel column rinsing with additional methanol and then eluting the heavy isotope labeled L- carnitine with 30% v/v water in methanol. The product was dried via azeotropic distillation of absolute ethanol and subsequently recrystallized from ethanol and acetone. The white to off-white crystalline product was dried over P 2 O 5 in vacuo and stored refrigerated. Upon analysis, the d3-l-carnitine was found to be > 98% pure by LC/MS, NMR and TLC. TLC on silica gel eluted with methanol plus 0.2%v/v, formic acid visualized by iodine staining shows one spot with the same Rf as L-carnitine. 600MHz 1 H-NMR (10 mg ml 1 in D 2 O): δ 2.3ppm (complex multiplet 2H), δ 3.1ppm (singlet 6H), δ 3.3ppm (complex multiplet, 2H) δ 4.5ppm 91

94 (complex multiplet, 1H), which is consistent with the spectrum obtained for L- carnitine under the same conditions and concentration except for the singlet peak at 3.1 ppm corresponding to 9 protons on the trimethylammonium group on L- carnitine integrates for 6 protons (three protons less) due to the incorporation of 3 deuterium atoms on one of the methyl amino groups in this compound. The only impurity peaks observed corresponded to residual ethanol and acetone in the product (integrated area less than 1% of total), and these were removed by placement in vacuum dessicator. 13 C-NMR (10 mg ml in D 2 O): δ 64.3ppm (multiplet, 1C), δ 70.2ppm (multiplet, 1C), δ 54.2ppm (muliplet, 2C), δ 43.1ppm (multiplet, 1C), δ 179.0ppm (singlet 1C). The mass spectrum of the compound dissolved in 50% v/v, methanol/water (5 mm formic acid) to a concentration of 50 µg ml 1 exhibits a base peak at m/z = 165 in the positive ion mode, corresponding to [M]+. CID fragments peaks were observed at m/z = 105,103, 85, 63, 57, and 43. Mass spectral fragmentation patterns and m/z ratios are consistent with the L-carnitine except for those fragment ions that contain the trimethylammonium group; these ions exhibit fragments 3 atomic mass units (amu) higher than the corresponding signals from L-carnitine due to the incorporation of 3 deuterium atoms on one of the methyl groups attached to the nitrogen. Germ-free mice and conventionalization studies 10-week-old female Swiss Webster germ-free mice (SWGF) underwent gastric gavage with the indicated isotopologues of L-carnitine (see below for details of L- 92

95 carnitine challenge) immediately following removal from the germ-free microisolator shipper. After performing the L-carnitine challenge, germ-free mice were conventionalized by being housed in cages with non-sterile C57BL/6J female mice. Approximately 4 weeks later, the L-carnitine challenge was repeated on the conventionalized Swiss Webster mice. Quantification of natural abundance and isotope labeled carnitine, TMA and TMAO in mouse plasma was performed using stable isotope dilution LC/MS/MS as described above. Metabolic challenges in mice C57BL/6J female or C57BL/6J, Apoe / female mice were provided via gastric gavage d3-l-carnitine (150 µl of 150 mm stock) d3-l-carnitine (synthetically prepared as above) dissolved in water using a 1.5-inch 20-gauge intubation needle. Plasma was collected from the saphenous vein at baseline and at the indicated time points. C57BL/6J, Apoe / female mice were used in the study examining the inducibility of microbiota to generate TMA and TMAO following carnitine feeding. For these studies, animals were placed on an L-carnitine supplemented diet (1.3% L-carnitine in mouse drinking water) for 10 weeks. Quantification of natural abundance and isotope labeled forms of carnitine, TMA and TMAO in mouse plasma was performed using stable isotope dilution LC/MS/MS as described above. Preparation of bone marrow derived macrophages for reverse cholesterol transport studies 93

96 Femur bone marrow from C57BL/6J mice was collected and cultured in PFA bags (Welch Fluorocarbon, Dover, NH) with RPMI-640 supplemented with L-cell conditioned media, β-mecaptoethanol, penicillin/streptomycin, and glutamine for 6 days. Each PFA bag of bone marrow derived macrophages were then loaded with 40 µci [ 14 C] cholesterol preincubated with carbamylated LDL for 48 hours. Carbamylated LDL was prepared as described previously 128. At the end of 48 hours bone marrow derived macrophages were collected for injection into reverse cholesterol transport mice. Reverse cholesterol transport studies Adult (> 8 weeks of age) C57BL/6J, Apoe / female mice were placed on diets for 4 weeks prior to beginning of reverse cholesterol transport experiments. Mice were individually placed into single ventilated cages with wire rack inserts (Ancare, Spring Valley, Illinois) for a hour acclimatization period. Mice were injected subcutaneously in the back with 300ul of labeled bone marrow derived macrophages as described above. Feces were collected every 24 hours, processed, and analyzed by a modified method previously described 142. Briefly, each 24 hour feces collection was extracted with 3:2 chloroform/methanol and back extracted with 1:5 0.88% KCl. The organic phase was collected dried, dissolved in scintillation fluid, and counted on a Beckman Coulter LS6500 liquid scintillation counter. Total 72 hour reverse cholesterol transport studies were calculated as a sum of each 24 hour period. Percent reverse cholesterol transport is expressed as the percentage of [ 14 C] DPM recovered from feces 94

97 versus [ 14 C] counts injected into each mouse. At the end of the 72 hour period animals were fasted for 3 hours and then sacrificed for collection of blood, liver, bile, and intestine. [ 14 C] was counted in aliquots of plasma and bile dissolved in scintillation fluid and counted on a Beckman Coulter LS6500 liquid scintillation counter. [ 14 C] was quantified in liver by extraction with 3:2 chloroform methanol and back extraction with 2:5 0.88% KCl. Both the aqueous and organic phases were dried, dissolved in scintillation fluid, and counted on a Beckman Coulter LS6500 liquid scintillation counter. The percent injected was calculated as the percentage of [ 14 C] DPM recovered from feces versus [ 14 C] counts injected into each mouse normalized by liver weight analyzed. Cholesterol absorption studies Cholesterol absorption experiments were performed as previously described 143. Briefly, adult (> 8 weeks of age) C57BL/6J, Apoe / female mice were placed on the indicated diets for 4 weeks prior to beginning of cholesterol absorption experiments. Mice were individually placed into single ventilated cages with wire rack inserts (Ancare, Spring Valley, Illinois) for a hour acclimatization period. Animals were fasted 4 hours before gavage with olive oil supplemented with [ 14 C] cholesterol/ [3H] β-sitostanol. Feces were collected over a 24 hour period. Feces samples and cholesterol absorption rates were calculated as previously described 143. Briefly, feces were extracted with 3:2 chloroform/methanol and back extracted with 1:5 0.88% KCl. The organic phase was collected dried, dissolved in scintillation fluid, and counted on a Beckman 95

98 Coulter LS6500 liquid scintillation counter. The percent cholesterol absorption was calculated as the ratio of ([ 14 C] DPM in the feces: [3H] β-sitostanol) / the ratio of [ 14 C] DPM: [3H] β-sitostanol gavaged subtracted from 1. Bile acid pool size and composition Total bile acid pool size was determined in female C57BL/6J, Apoe / as the total bile acid content of the combined small intestine, gallbladder, and liver, which were extracted together in ethanol with Nor-Deoxycholate (Steraloids Newport, RI) added as an internal standard. The extracts were filtered (Whatman paper #2), dried and resuspended in water. The samples were then passed through a C18 column (Sigma St. Louis, MO) and eluted with methanol. The eluted samples were again dried down and resuspended in methanol. A portion of this was subjected to HPLC using Waters Symmetry C18 column ( mm No. WAT054275, Waters Corp., Milford, MA) and a mobile phase consisting of methanol: acetonitrile: water (53:23:24) with 30 mm ammonium acetate, ph 4.91, at a flow rate of 0.7 ml min 1. Bile acids were detected by evaporative light spray detector (Alltech ELSD 800, nitrogen at 3 bar, drift tube temperature 40 0 C) and identified by comparing their respective retention times to those of valid standards (Taurocholate and Tauro-β-muricholate from Steraloids (Newport, RI); Taurodeoxycholate and Taurochenodeoxycholate from Sigma (St. Louis, MO); Tauroursodeoxycholate from Calbiochem (San Diego, CA). For quantitation, peak areas were integrated using software Chromperfect Spirit (Justice laboratory software, Denville, NJ) and bile acid pool size was expressed as 96

99 µmol/100 g body weight (bw) after correcting for procedural losses with nordeoxycholate. Cholesterol efflux studies RAW mouse macrophages were cultured in a 48 well plate. Macrophages were labeled with cholesterol using 1 µci ml 1 [ 3 H] cholesterol preincubated with AcLDL for 24 hours. In wells examining Abca1 dependent efflux, Abca1 was induced with 0.3 mm 8Br-cAMP as previously described 144. Cells were washed and chased with serum free media containing 8Br-cAMP and 10 µg ml 1 (final) human APOA1 for 6 hours (for pretreated wells) or isolated human HDL (50 µg protein ml 1 final) in serum free media. Media was counted directly using Beckman Coulter LS6500 liquid scintillation counter. Cells were washed and extracted with 3:2 hexane:isopropanol. Dried extracts were then counted using a Beckman Coulter LS6500 liquid scintillation counter. Total Cholesterol efflux was determined as total media DPM/ (total media DPM and Total extract DPM). Abca1 efflux was determined as the difference between cholesterol efflux in the presence of 8Br-cAMP compared to the absence of 8Br-cAMP. Effect of TMAO on macrophage cholesterol biosynthesis, inflammatory genes, and desmosterol levels The effect of cholesterol loading on macrophage cholesterol biosynthetic and inflammation genes, LDL receptor expression levels, and desmosterol levels, were performed by a modified method as previously described 89. Briefly, mouse peritoneal macrophages (MPMs) were thioglycollate elicited 4 days prior to 97

100 harvest and were subsequently cultured in RPMI 1640 supplemented with 10% FCS and penicillin/streptomycin overnight. MPMs were then lipoprotein starved in RPMI 1640 supplemented with 10% lipoprotein deficient serum (LPDS) and penicillin/streptomycin for 24 hours and then further cultured in the same media for 18 hours in the presence of increasing cholesterol, AcLDL concentrations or vehicle (carrier for AcLDL and cholesterol (Sigma, St. Louis, MO) with (+) or without ( ) 300 µm TMAO dehydrate (Sigma, St. Louis, MO)). AcLDL was prepared as previously described 145. RNA was prepared and analyzed as described below. Desmosterol in the cholesterol loading studies was quantified by stable isotope dilution GC/MS analysis. Briefly, desmosterol was extracted from 400 µl medium by 1 ml isopropanol/hexane/2 M acetic acid (40/10/1, vol/vol/vol) with 100 ml of 10 mg ml 1 deuterated internal standard, cholesterol-2,2,3,4,4,6-d6 (Sigma) in isopropanol added beforehand. After adding 1 ml hexane, the mixture was vortexed and spun down, desmosterol and cholesterol-2,2,3,4,4,6-d6 were extracted to the hexane layer. The medium was re-extracted by the addition of 1 ml hexane, followed by vortexing and centrifugation. The hexane layer was collected and combined with the previous hexane extract. The extract was dried under N ml Sylon HTP (HMDS+TMCS+Pyridine, 3:1:9) (Supelco) was added to the dried desmosterol preparative and trimethylsilyl (TMS) ethers were achieved in 1 hour at 90 o C. Calibration curves were prepared using varying desmosterol levels and a fixed amount of stable isotope-labeled internal 98

101 standard, d6(2,2,3,4,4,6) cholesterol undergoing derivatization to TMS ethers. 1 ml of the TMS ethers was injected onto a 6890/5973 GC/MS equipped with an automatic liquid sampler (Agilent Technolgies) using the positive ion chemical ionization mode with methane as the reagent gas. The source temperature was set at 230 C. The electron energy was 240 ev, and the emission current was 300 µa. The cholesterol TMS ethers were separated on a J&W Scientific (Folsom, CA) DB-1 column (20 m, 0.18 mm inner diameter, 0.18-µm film thickness). The injector and the transfer line temperatures were maintained at 250 C. The initial GC oven temperature was set at 230 C and increased at 20 C/min to 270 C then increased at 4 C/min to 300 C. The GC chromatograms extracted at m/z = 327 and 335 corresponding to desmosterol and cholesterol- 2,2,3,4,4,6-d6, were extracted and the peak area were integrated, respectively. RNA preparation and real time PCR analysis RNA was first purified from tissue (macrophage, liver, or gut) using the animal tissue protocol from the Qiagen Rneasy mini kit. Small bowel used for RNA purification was sectioned sequentially in 5 equal segments from the duodenum to illeum before RNA preparation. Purified total RNA and random primers were used to synthesize first strand cdna using the High Capacity cdna Reverse Transcription Kit (Applied Biosystems, Foster City, CA) reverse transcription protocol. Quantitative real-time PCR was performed using Taqman qrt-pcr probes (Applied Biosystems, Foster City, CA) and normalized to tissue β-actin by 99

102 the C T method using StepOne Software v2.1 (Applied Biosystems, Foster City, CA). 100

103 Figure 2-1. TMAO production from carnitine is a microbiota dependent process in humans. (a) Structure of carnitine and scheme of carnitine and choline metabolism to TMAO. L-Carnitine and choline (are both dietary trimethylamines that can be metabolized by microbiota to TMA. TMA is then further oxidized to TMAO by flavin monooxygenases (FMOs). (b) Scheme of human carnitine challenge test. After a 12 hour overnight fast, subjects received a capsule of d3-carnitine (250 mg) alone, or in some cases (as in data for subject shown) also an 8 ounce steak (estimated 180 mg L- carnitine), whereupon serial plasma and 24h urine collection was obtained for TMA and TMAO analyses. After a weeklong regimen of oral broad spectrum antibiotics to suppress the intestinal microbiota, the challenge was repeated (Visit 2), and then again a final third time after a three week period to permit repopulation of intestinal microbiota (Visit 3). Data shown in (panels c-e) are from a representative omnivorous subject who underwent carnitine challenge. Data is organized to vertically correspond with the indicated visit schedule above (Visit 1, 2 or 3). (c,d) LC/MS/MS chromatograms of plasma TMAO or d3-tmao in an omnivorous subject using specific precursor product ion transitions indicated at T = 8 hour time point for each respective visit. (e) Stable isotope dilution LC/MS/MS time course measurements stable isotope (d3) labeled TMAO and carnitine, in plasma collected from sequential venous blood draws at noted times

104 Figure 2-2. The formation of TMAO from ingested L-carnitine is negligible in vegans, and fecal microbiota composition associates with plasma TMAO concentrations. (a-b) Data from a vegan in the carnitine challenge consisting of co-administration of 250 mg d3-carnitine and an 8 ounce sirloin steak, and a representative omnivore. (a) Plasma TMAO and d3-tmao were quantified post carnitine challenge, and in a (b) 24 hour urine collection. (c) Baseline fasting plasma concentrations of (n = 26) vegans and vegetarians and (n = 51) omnivores. Boxes represent the 25th, 50th, and 75th percentile and whiskers represent the 5th and 95th percentile. (d) Plasma d3-tmao levels in male and female (n = 5) vegan/ vegetarian versus (n = 5) omnivores participating in a d3-carnitine (250 mg) challenge. P value shown is for comparison between area under the curve (AUC) of groups using Wilcoxon nonparametric test. (e) Baseline plasma concentrations of TMAO associates with Enterotype 2 (Prevotella) subjects with a characterized gut enterotype. (f) Plasma TMAO concentrations (x axes) and the proportion of taxonomic operational units (OTUs, Y axes) were determined as described in Methods. Subjects were grouped as vegan/vegetarian (n = 23) or omnivore (n = 30). P value shown is for comparisons between dietary groups using a robust Hotelling T 2 test

105 Figure 2-3. The metabolism of carnitine to TMAO is an inducible trait and associates with microbiota composition. (a) d3-carnitine challenge of mice on either a carnitine supplemented diet (1.3%) at 10 weeks and age versus age-matched normal chow controls. Plasma d3-tma and d3- TMAO were measured at the indicated times following d3-carnitine administration by oral gavage using stable isotope dilution LC/MS/MS. Data points represents mean ± SE of 4 replicates per group. (b) Correlation heat map demonstrating the association between the indicated microbiota taxonomic genera and TMA and TMAO levels (all reported as mean ±SE in µm) of mice grouped by dietary status (chow, n = 10 (TMA,1.3±0.4; TMAO, 17±1.9); and carnitine, n = 11 (TMA, 50±16; TMAO, 114±16). Red denotes a positive association, blue a negative association, and white no association. A single asterisk indicates a significant false discovery rate adjusted (FDR) association of P 0.1 and a double asterisk indicates a significant FDR adjusted association of P (c) Plasma TMAO and TMA concentrations were determined by stable isotope dilution LC/MS/MS (x axes) and the proportion of taxonomic operational units (OTUs, Y axes) were determined

106 Figure 2-4. Relation between plasma carnitine and CVD risks. (a-c) Forrest plots of odds ratio of CAD, PAD, and CVD and quartiles of carnitine before (closed circles) and after (open circles) logistic regression adjustments with traditional cardiovascular risk factors including age, sex, history of diabetes mellitus, smoking, systolic blood pressure, low density lipoprotein cholesterol, and high density lipoprotein cholesterol. Bars represent 95% confidence intervals. (d) Relationship of fasting plasma carnitine levels and angiographic evidence of CAD. Boxes represent the 25th, 50th, and 75th percentile of plasma carnitine and the whiskers represent the 10th and 90th percentile. The Kruskal- Wallis test was used to assess the degree of coronary vessel disease on L-carnitine levels. (e) Forrest plot of hazard ratio of MACE (death, non fatal-mi, stroke, and revascularization) and quartiles of carnitine unadjusted (closed circles), and after adjusting for traditional cardiovascular risk factors (open circles), or traditional cardiac risk factors plus creatinine clearance, history of MI, history of CAD, burden of CAD (one, two, or three vessel disease), left ventricular ejection fraction, baseline medications (ACE inhibitors, statins, β-blockers, and aspirin) and TMAO levels (open squares). Bars represent 95% confidence intervals. (f) Kaplan Meier plot (graph) and hazard ratios with 95% confidence intervals for unadjusted model, or following adjustments for traditional risk factors as in panel e. Median levels of carnitine (46.8 µm) and TMAO (4.6 µm) within the cohort were used to stratify subjects as high ( median) or low (< median) concentrations

107 Figure 2-5. Dietary carnitine accelerates atherosclerosis and inhibits reverse cholesterol transport in a microbiota dependent fashion. (a) Representative Oil-red-O stained (counterstained with hematoxylin) aortic roots of 19 week old C57BL/6J, Apoe / female mice on the indicated diets in the presence versus absence of antibiotics (ABS). (b) Quantification of mouse aortic root plaque lesion area of 19 week-old C57BL/6J, Apoe / female mice on respective diets. (c) Carnitine, TMA, and TMAO were determined using stable isotope dilution LC/MS/MS analysis of plasma recovered from mice at time of sacrifice. (d) Reverse cholesterol transport (RCT) in female C57BL/6J, Apoe / mice on normal chow versus diet supplemented with either carnitine or choline, as well as following suppression of microbiota using cocktail of antibiotics (+ ABS). Also shown are RCT results in female C57BL/6J, Apoe / mice on normal chow versus diet supplemented with TMAO. (e,f) Relative mrna levels (to β-actin) of mouse liver candidate genes involved in bile acid synthesis or transport. Ephx1, epoxide hydrolase 1, microsomal

108 Figure 2-6. Effect of TMAO on cholesterol and sterol metabolism. Measurement of (a) total bile acid pool size and composition, as well as (b) cholesterol absorption in adult female (> 8 weeks of age) C57BL/6J, Apoe / mice on normal chow diet versus diet supplemented with TMAO for 4 weeks. (c) Summary scheme outlining pathway for microbiota participation in atherosclerosis via metabolism of dietary carnitine and choline forming TMA and TMAO, as well as the impact of TMAO on cholesterol and sterol metabolism in macrophages, liver and intestines. FMOs, flavin monooxygenases; TMA, trimethylamine; TMAO, trimethylamine-n-oxide; OST-α, solute carrier family 51, alpha subunit; ASBT, solute carrier family 10, member

109 SUPPLEMENTARY MATERIAL Supplementary Table 2-1: Characteristics of analyte m/z = 162 determined in LC/MS positive ion mode from plasma samples used in Validation and Learning cohorts (n = 150) of metabolomics study from Wang et. al., Nature, Plasma samples used in the metabolomics study described in Wang et al 69 were from GeneBank, a large clinical repository of patients undergoing elective diagnostic cardiac evaluation. The original study utilized a Learning cohort of 50 cases (randomly selected GeneBank subjects who experienced death, non-fatal MI, or stroke in the ensuing 3 year follow up period) and 50 age and gender matched controls (subjects with no ensuing history of death, non-fatal MI or stroke in the 3 year period after enrollment). Peaks within LC chromatograms from the metabolomics analyses that exceeded a signal to noise ratio of greater than 5 were integrated. Bonferroni adjusted two sided T-tests were calculated and adjusted logp value > 1.3 were considered significant. An odds ratio (OR) between the highest and lowest quartile was calculated for each unknown analyte. Only analytes with 95% confidence intervals not crossing unity were considered significant. Additionally, Cochran-Armitage trend tests across the quartiles were performed with P < 0.05 being considered significant. A similar analysis was performed in a non-overlapping Validation cohort consisting of 25 additional cases and controls from GeneBank. In both Learning and Validation cohorts only 18 plasma analytes met this strict set of validation criterion, and an analyte with m/z =162 (same as carnitine) was not among them 1. Results for an analyte with m/z = 162 and retention time similar to that of authentic L-carnitine in the Learning and Validation cohorts are shown. In a new hypothesis-generated analysis that did not adjust for multiple sampling (since only an analyte was being examined) and that used the combined data set (Learning + Validation cohorts, n = 75 cases and n = 75 age-gender matched controls), the plasma analyte with m/z = 162 and retention time identical to carnitine was significantly associated with cardiovascular risks (bottom table, P = 0.04). These results suggested that plasma levels of an analyte with m/z = 162, perhaps L-carnitine, may be associated with cardiovascular risks

110 Patient Characteristics Whole cohort Q1 Carnitine Q2 Quartiles Q3 Q4 P < 31.7 µm µm > 45.2 µm (n = 2595) (n = 649) µm (n = 649) (n = 650) (n = 647) Age (years) 62 (54-71) 63 (54-72) 62(54-71) 63(54-71) 61 (53-71) < 0.01 Male (%) < 0.01 Smoking (%) < 0.01 Diabetes mellitus (%) Hypertension (%) Hyperlipidemia (%) < 0.01 Prior CAD (%) < 0.01 CAD (%) < 0.01 PAD (%) CVD (%) < 0.01 BMI (kg/m 2 ) 29 (25-33) 28 (24-31) 29 (2-32) 29 (25-32) 29 (26-34) < 0.01 LDL cholesterol (mg dl 1 ) 96 (78-117) 94 (74-111) 100 (80-122) 97 (80-117) 96 (77-117) < 0.01 HDL cholesterol (mg dl 1 ) 34( 28-41) 35 (30-43) 34 (28-41) 32 (27-39) 32 (27-38) < 0.01 Total cholesterol (mg dl 1 ) 160 ( ) 159 ( ) 164 ( ( ) 161 ( ) < ) Triglycerides (mg dl 1 ) 117 (85-167) 103 (76-148) 110 (84-159) 124 (88-170) 129 (96-192) < 0.01 hscrp (mg l 1 ) 2.3 ( ) 2.3 ( ) 2.1 ( ) 2.2 ( ) 2.5 ( ) < 0.01 MPO (pmol l 1 ) 113 (76-230) 122 (75-267) 109 (73-205) 110 (76-215) 114 (79-234) < 0.01 egfr (ml min/1.73/m 2 ) 83 (70-96) 86 (73-99) 85 (73-98) 83 (70-95) 79 (64-93) < 0.01 Carnitine (µm) 38 (32-45) 28 (25-30) 35 (33-36) 41 (39-43) 51 (48-56) < 0.01 Baseline medications (%) ACE inhibitors < 0.01 Beta-blockers < 0.01 Statin Aspirin Supplementary Table 2-2: Subject characteristics, demographics, and laboratory values in the whole cohort (n = 2595), and across quartiles of plasma carnitine. Values are expressed in mean ± SD for normally distributed variables, or median (interquartile range) for non-normally distributed variables. The P value represents a Kruskal Wallis test for continuous variables and Chi-square test for categorical variables across quartiles of carnitine. Abbreviations: ACE, angiotensin converting enzyme; ATP III, Adult Treatment Panel III guidelines; BMI, body mass index; CAD, coronary artery disease; CVD, cardiovascular disease; ctni = cardiac Troponin I; HDL, high-density lipoprotein; hscrp, high-sensitivity C-reactive protein; LDL, low-density lipoprotein; MPO, myeloperoxidase; PAD, peripheral artery disease

111 Supplementary Table 2-3: Plasma levels of triglycerides, cholesterol, glucose, and insulin from mice on normal chow vs. carnitine supplemented diet. C57BL/6J, Apoe / female mice at time of weaning were placed on the indicated diets until time of sacrifice for aortic root quantification of atherosclerosis (19 weeks of age). Parallel groups of animals were also provided an antibiotics cocktail in drinking water as described under Methods. Lipid profiles, glucose, and insulin levels shown were determined in plasma isolated at time of organ harvest at conclusion of study. Data shown are mean ± SD for each of the indicated feeding groups. Student t-test comparisons are between chow and carnitine (1.3%) supplemented diets with the noted antibiotic (ABS) treatment status

112 Supplementary Table 2-4: Liver levels of triglycerides and total cholesterol in mice on normal chow versus carnitine supplemented diet. Liver was harvested from female C57BL/6J, Apoe / mice on the indicated diets at time of sacrifice for aorta harvest for aortic root quantification (19 weeks of age and 15 weeks on diets). Liver was homogenized and the content of triglycerides and total cholesterol determined as described under Methods. Data are presented as mean ± SD for each of the indicated groups of mice. A student t-test comparison was performed between chow and carnitine groups on or off a cocktail of oral broad spectrum antibiotics (+ ABS) as described in Methods. No significant increases in liver lipid levels were noted in the carnitine supplemented mice compared to the respective chow controls

113 Supplementary Table 2-5: Plasma levels of triglycerides, cholesterol, and glucose from mice on normal chow, carnitine, choline, and TMAO supplemented diets during the in vivo RCT studies. C57BL/6J, Apoe / female mice were enrolled in two separate studies to quantify in vivo reverse cholesterol transport (RCT) by placement on the indicated diets at time of weaning ("TMAO RCT" study, and "Carnitine and Choline RCT" study). Following 4 weeks of diet, [ 14 C]cholesterol loaded macrophages were injected subcutaneously, and in vivo RCT quantified as described under Methods. Lipid profiles and glucose levels shown were determined in plasma isolated at time of organ harvest at conclusion of study (72h post injection of [ 14 C]cholesterol loaded macrophages). Data shown are mean ± SD for each of the indicated dietary groups

114 Supplementary Figure 2-1: Mass spectrometry analyses identify unknown plasma analyte at retention time of 5.1 min and m/z = 162 as carnitine. a) Extracted ion chromatograms at m/z = 162 from human plasma sample (top), and authentic L-carnitine standard (bottom). Identical retention times under multiple chromatographic conditions during LC/MS analysis were demonstrated for analyte m/z =162 and L-carnitine standard. b) Collision-induced dissociation (CID) spectra from 5.10 min peak in human plasma and L-carnitine standard. This data demonstrate that the analyte at 5.10 min with m/z = 162 from human plasma possesses identical CID mass spectrum and retention time to authenticate synthetic L-carnitine standard

115 Supplementary Figure 2-2. LC/MS/MS analysis of synthetic heavy isotope standard d9(trimethyl)carnitine spiked into human plasma sample confirms unknown peak at 5.10 min (m/z = 162) is carnitine. a) Human plasma was spiked with synthetic d9(trimethyl)-carnitine. The sample was then analyzed by LC/MS/MS using multiple distinct precursor product ion transitions in multiple reaction monitoring (MRM) mode that are characteristic for L-carnitine and its d9(trimethyl)-isotopologue. Note that multiple characteristic precursor product transitions demonstrate identical retention times for both the plasma analyte with m/z = 162, and synthetic d9(trimethyl)-carnitine standard. b) Precursor product ion transitions were determined from CID spectra of both authentic L-carnitine and synthetic d9(trimethyl)carnitine. Insets: Shown are proposed fragmentation overlay on structure from positive ion electrospray analyses of L-carnitine and d9- carnitine

116 Supplementary Figure 2-3. Standard curves for LC/MS/MS quantification of carnitine and d3-(methyl)-carnitine in plasma matrix. We used synthetic d9(trimethyl)carnitine as internal standard to quantify d3(methyl)carnitine, and natural abundance carnitine isotopologues in plasma recovered from mice and humans following carnitine challenge. To generate standard curves for each isotopologue in plasma matrix, a fixed amount of d9-(trimethyl)carnitine as an internal standard was added to dialyzed human plasma, and increasing concentrations of L-carnitine (a) and synthetic d3-(methyl)carnitine (b) were spiked into the samples. Plasma proteins were precipitated with a methanol at 0 C. Aliquots of the supernatant solution were analyzed by LC with on-line tandem mass spectrometry using electrospray ionization in positive ion mode on an AB SCIEX 5000 triple quadrupole mass spectrometer. Unique precursor product ion transitions were selected for carnitine and its d3- and d9- isotopologues. Areas of peaks from multiple reaction monitoring (MRM) were divided by the peak area from m/z transition from d9-carnitine. Standard curves of peak area ratio versus known concentrations are plotted on the same axis for carnitine (a) and d3-carnitine (b). For quantification, precursor product transitions of and were typically used to measure carnitine and d3-carnitine, respectively, and if needed, alternative indicated transitions used to confirm results

117 Supplementary Figure 2-4. LC/MS/MS analyses of a subject s 24 hr urine samples demonstrate an obligatory role for gut microbiota in production of TMAO from carnitine. (a) Scheme of overall study. There were 3 visits where carnitine challenge (following overnight fast, ingestion of carnitine in form of 8 oz steak (where indicated) and 250 mg d3-(methyl) carnitine) occurred with serial plasma and 24h urine collection. Visit 1 served as baseline. Subjects then took a cocktail of oral antibiotics for 1 week as described in Methods to suppress intestinal microbiota, and repeat carnitine challenge was performed at Visit 2. A third and final Visit was performed after at least 1 month of being off of antibiotics. (b) Data shown are chromatographic peaks from analysis of urine samples (aliquot of 24 hour collections) from a typical omnivorous subject (from n > 10 who underwent carnitine challenge and had complete serial blood draws performed) following carnitine challenge at the indicated visit shown above. The top row of chromatograms is from LC/MS/MS analyses of the indicated precursor product transition specific for TMAO, and the bottom panel represents similar analyses using precursor product transitions specific for d3-tmao. Note that TMAO and d3-tmao are readily detected at Visit 1 and 3 after d3-carnitine ingestion, but not Visit 2 where intestinal microbiota is suppressed by oral broad spectrum antibiotics, consistent with a requirement for gut microbiota involvement in both TMA and TMAO formation. Data is organized to vertically correspond with the indicated visit schedule above (Visit 1, 2 or 3)

118 Supplementary Figure 2-5. Plasma levels of carnitine and TMAO following carnitine challenge in a typical omnivorous subject. (a) Scheme of human carnitine challenge test. After an overnight fast, subjects were challenged with a capsule of d3- carnitine (250 mg) alone and with an 8 ounce steak (estimated 180 mg L-carnitine). This was followed with serial plasma and a 24h urine collection for TMAO and carnitine analyses. Visit 2 occurred after a weeklong regimen of oral broad spectrum antibiotics to suppress the intestinal microflora. The challenge was repeated a third time after a three week period off antibiotics (Visit 3). Data shown (b) are from stable isotope dilution LC/MS/MS time course measurements of natural abundance TMAO and carnitine in plasma collected from sequential venous blood draws at noted times from a representative omnivorous subject of n > 10 who underwent carnitine challenge. Data is organized to vertically correspond with the indicated visit schedule above (Visit 1, 2 or 3)

119 Supplementary Figure 2-6. Plasma levels of carnitine and d3-carnitine following carnitine challenge (steak and d3-carnitine) in typical omnivore with frequent red meat dietary history and a vegan subject. Plasma was isolated at baseline (T = 0) and the indicated times points following carnitine challenge (8-ounce steak mg of d3-carnitine) in an omnivore who reported near daily consumption of red meat, and in the one vegan subject who agreed to consume 8 ounces of steak with the d3-carnitine. Plasma levels of endogenous (natural abundance) carnitine (left panel) and the d3- carnitine isotopologue (right panel) were determined by stable isotope dilution LC/MS/MS analysis using synthetic d9-carnitine as internal standard as described in Methods. The data shown for natural abundance carnitine are typical for the omnivore, where nominal changes in plasma levels are noted following consumption of a steak, but increases from typically relatively lower levels (for vegans/vegetarians) are noted in the vegan subject shown. Substantial increases in the isotope labeled d3-carnitine were found in both vegan and omnivore alike. Also note the greater extent of increase in d3- carntine within the vegan observed compared to the omnivore following ingestion of the d3-carnitine containing capsule, consistent with more intestinal microbiota-mediated catabolism of the d3-carnitine in the omnivore, blunting the amount of carnitine absorbed relative to that observed in the vegan

120 Supplementary Figure 2-7. Plasma levels of d3-carnitine following d3-carnitine challenge (no steak) in omnivorous (n = 5) versus vegan subjects (n = 5). Similar studies to that shown in Supplementary Figure 2-6 where carnitine challenge did not include ingestion of steak, but only d3-carnitine (250 mg) in a capsule. Plasma was isolated at baseline (T = 0) and the indicated times points following d3-carnitine ingestion in both omnivorous (n = 5) and vegan (n = 5) subjects. Plasma levels of d3- carnitine were determined by stable isotope dilution LC/MS/MS analysis using synthetic d9-carnitine as internal standard as described in Methods. Statistical analysis was performed by a Wilcoxon rank-sums test between the mean area under the curve between subjects grouped by omnivorous versus vegan status. A significant increase in plasma d3-carnitine occurs in both vegan and omnivore alike over baseline values, but to a greater extent in vegans, following ingestion of the d3-carnitine containing capsule (P < 0.05). This is consistent with more intestinal microbiota-mediated catabolism of the d3-carnitine in the omnivore, blunting the amount absorbed relative to that observed in vegans. * P < 0.05 for difference between vegan and omnivore subjects

121 Supplementary Figure 2-8. Human fecal microbiota taxa associate with plasma TMAO. Human fecal samples were collected from vegan/vegetarians (n = 23) and omnivores (n = 30) and microbiota gene encoding for 16S rrna was analyzed as described under Methods. Associations between plasma TMAO and taxa proportions were assessed as described under Methods. False discovery rate (FDR) control based on the Benjamini Hochberg procedure was used to account for multiple comparisons. Asterisked taxa met a FDR adjusted P value < 0.1. Further details of the preparation and analysis of human fecal samples can be found in Methods

122 Supplementary Figure 2-9. Demonstration of an obligatory role of the commensal gut microbiota of mice in the production of TMA and TMAO from oral carnitine in germ-free and conventionalized mice. d3-carnitine challenge (oral gavage of d3- carnitine) in germ-free female Swiss Webster mice before and after ensuing conventionalization ( 3 weeks in conventional cages with conventional mice). Each point represents mean ± SE of 4 independent replicates. Plasma levels of d3-carnitine, d3-tmao and d3-tma were determined by stable isotope dilution LC/MS/MS analysis using synthetic d9-(trimethyl)carnitine, d9-(trimethyl)tma, and d9-(trimethyl)tmao as internal standards. Note that there is an obligatory role for gut microbiota in generation of TMA and TMAO from orally ingested carnitine, as reflected by the absence of these metabolites in the germ-free mice, but their formation within the conventionalized mice

123 Supplementary Figure Demonstration of an obligatory role of commensal gut microbiota of mice in the production of TMA and TMAO from oral carnitine. Left panel - C57BL/6J, Apoe / female mice (n = 5) in conventional cages were given oral d3- carnitine via gavage at T = 0, and then serial blood draws were obtained at the indicated times. Plasma levels of d3-carnitine, d3-tmao and d3-tma were determined by stable isotope dilution LC/MS/MS analysis using synthetic d9-(trimethyl)carnitine, d9- (trimethyl)tma, and d9-(trimethyl)tmao as internal standards. Middle panel - Mice were then treated with a cocktail of oral broad spectrum antibiotics to suppress intestinal microbiota as described in Methods. Repeat gastric gavage with d3-carnitine was performed, and serial testing of plasma for quantification of d3-carnitine, d3-tma and d3-tmao levels were determined. Right panel - Antibiotics were stopped and mice allowed to reacquire ( 3 weeks) their intestinal microbiota in conventional cages. Repeat gastric gavage with d3-carnitine was performed, and d3-carntine and its metabolites d3-tma and d3-tmao were then quantified by LC/MS/MS in serial plasma samples. Results shown are mean ± SE for 5 animals

124 Supplementary Figure Analysis of mouse plasma TMA and TMAO concentrations and gut microbiome composition can distinguish dietary status. C57BL/6J, Apoe / female mice were maintained either on normal chow (n = 10) or a carnitine supplemented (1.3%) diet (n = 11) as described under Methods. At sacrifice, blood and intestines were harvested, microbial DNA for the gene encoding 16S-rRNA was isolated from cecal contents, and microbiota composition analyzed as described under Methods. Plasma TMAO and TMA concentrations were determined by stable isotope dilution LC/MS/MS (plotted on x axes) and the proportion of taxonomic operational units of indicated taxa (OTUs, plotted on Y axes). Analyses and P values shown are for comparisons between dietary groups, and were determined as described in Methods

125 Supplementary Figure Haematoxylin/eosin (H/E) and oil-red-o stained liver sections. Representative liver sections from female C57BL/6J, Apoe / mice used in atherosclerosis study on the indicated diets collected at time of aorta harvest (19 weeks of age and 15 weeks on diets). Liver was stained by H/E (left column) or oil-red-o and counterstained with Haematoxylin (right column). Mice on these diets exhibit no obvious hepatosteatosis or other pathology. As a positive control for comparison showing fatty liver, C57BL/6J mice fed a high fat diet (16 weeks of age and 6 weeks on the diet) is shown in bottom row

126 Supplementary Figure Arginine transport in the presence of 100 µm trimethylamine-containing compounds. Bovine aortic endothelial cells (BAEC) were incubated in DMEM medium supplemented with glutamine, 10% FCS and penicillin/streptomycin and with 100 µm of the indicated trimethylamine-containing cationic compounds. BAEC cell arginine uptake studies were performed in Krebs Henseleit buffer by the addition of 50 µm L-[ 3 H] arginine (1 µci ml 1 ). The samples were incubated for 30 min at 37 C and chased with cold 10 mm L-Arg. After washing with Krebs Henseleit, the samples were solubilized with 0.1 M NaOH, transferred into plastic liquid scintillation vials and mixed with 4 ml scintillation fluid prior to counting in a Beckman Coulter LS6500 liquid scintillation counter. Data represented mean ± SE from 6 independent replicates. No significant reduction in arginine uptake is noted, suggesting TMAO, carnitine and choline, cationic amino acids, do not compete with arginine for uptake into BAEC

127 Supplementary Figure Expression levels of cholesterol synthesis enzymes, transporters, and inflammatory genes in the presence or absence of TMAO. Elicited mouse peritoneal macrophages (MPMs) were cultured in RPMI 1640 supplemented with 10% FCS and penicillin/streptomycin overnight. MPMs were then lipoprotein starved in RPMI 1640 supplemented with 10% LPDS and penicillin/streptomycin for 24 hours and then further cultured in the same media for 18 hours in the presence of increasing cholesterol or AcLDL concentrations or vehicle (carrier for AcLDL and cholesterol) with (+) or without ( ) 300 µm TMAO (the upper 1% of plasma levels of TMAO noted in the cohort examined in the present study). RNA was then purified, cdna amplified, and relative (to β-actin) expression of the indicated genes quantified by RT-PCR as described in Methods. Data are expressed as the mean ± SE of n = 3 replicates. Differences between conditions + versus TMAO were evaluated using a student s t-test. Note that no consistent significant effects on candidate gene expression within MPMs in the presence or absence of TMAO are noted. Hmgcr, 3- hydroxy-3-methylglutaryl-coenzyme A reductase; Srebp2, sterol regulatory element binding factor 2; Ldlr, low density lipoprotein receptor; Dhcr24, 24-dehydrocholesterol reductase; Cxcl9, chemokine (C-X-C motif) ligand 9; Cxcl10, chemokine (C-X-C motif) ligand

128 Supplementary Figure Effect of TMAO on desmosterol levels in media of cultured mouse peritoneal macrophages in the presence of increasing cholesterol and acetylated LDL (AcLDL) concentrations. Elicited mouse peritoneal macrophages (MPMs) were cultured in RPMI 1640 supplemented with 10% FCS and penicillin/streptomycin overnight. MPMs were then lipoprotein starved in RPMI 1640 supplemented with 10% LPDS and penicillin/streptomycin for 24 hours and then further cultured in the same media for 18 hours in the presence of increasing cholesterol and AcLDL concentrations or vehicle (carrier for AcLDL and cholesterol) with (+) or without ( ) 300 µm TMAO. Media was harvested and the content of desmosterol was determined as described under Methods. Data represented mean ± SE from 3 independent replicates

129 Supplementary Figure Plasma concentrations of TMAO in mice undergoing in vivo reverse cholesterol transport studies. Adult (> 8 weeks of age) C57BL/6J, Apoe / female mice were placed on normal chow or either carnitine (1.3%) or choline (1.3%) supplemented diets. Where indicated, some groups of mice also had addition of a cocktail of antibiotics to their drinking water as described under Methods throughout the duration of the dietary feeding period and RCT study. TMAO concentration was determined by stable isotope dilution LC/MS/MS analysis of plasma recovered from mice at time of sacrifice in the reverse cholesterol transport studies. A Wilcoxon nonparametric test was used to assess the difference in plasma TMAO between animal diets. Data shown are mean ± SE

130 Supplementary Figure [ 14 C] Cholesterol recovered from mice on normal chow vs. TMAO diet enrolled in in vivo reverse cholesterol transport studies. Adult (> 8 weeks of age) C57BL/6J, Apoe / female mice were placed on either normal chow or a TMAO (0.12%) supplemented diet for 4 weeks before performing in vivo reverse cholesterol transport studies as described under Methods. Mice were sacrificed 72 hours post injection with [ 14 C]cholesterol-loaded bone marrow-derived macrophages and counts within plasma, liver, and bile were determined as described in Methods. Results shown are mean ± SE

131 Supplementary Figure Effect of TMAO on mouse peritoneal macrophages. Thioglycollate elicited mouse peritoneal macrophages (MPMs) from C56Bl/6J mice were cultured in RPMI media supplemented with 5% lipoprotein deficient serum, glutamine, and penicillin and streptomycin. MPMs were then further incubated in the same media for an additional 20 hours with the indicated levels of TMAO. RNA was then purified, cdna amplified, and relative (to β-actin) expression of the indicated genes quantified by RT-PCR as described in Methods. Data are expressed as the mean ± SE

132 Supplementary Figure Effect of TMAO on cultured macrophage cholesterol efflux. RAW264.7 macrophages were cultured in DMEM media supplemented with 10% FBS and penicillin/streptomycin until 75% confluence. Cells were then further incubated in DMEM media supplemented with 12.5 g l 1 glucose, 200 mm glutamine, 1.25 g l 1 BSA, and penicillin/streptomycin + or cyclic AMP (for Abca1 expression induction) for an additional 16 hours with the indicated levels of TMAO. Abca1-dependent and total cholesterol efflux were then determined using lipid free isolated human apolipoprotein A1 (APOA1), or isolated human HDL, as cholesterol acceptor, as described in Methods. Data are expressed as the mean and ± SD of replicates (n = 4). A student s t-test was used to assess the relative increase in cholesterol efflux relative to a PBS (no exposure) control. While a statistically significant increase in Abca1-dependent cholesterol efflux in macrophages exposed to TMAO is noted (P < 0.01), the biological significance is unclear given the modest level of the effect, even at the highest levels of TMAO used

133 Supplementary Figure Liver expression of cholesterol transporters in mice examined during reverse cholesterol transport studies. Livers from C57BL/6J, Apoe / female mice on the indicated diets in the reverse cholesterol transport experiments were collected at time of sacrifice. The relative expression levels of the indicated genes were determined by RT-PCR as described in Methods. Data are presented as mean ± SE

134 Supplementary Figure Western blot analysis of liver scavenger receptor B1 (Srb1) expression. Female C57BL/6J, Apoe / mice were placed on either normal chow or diet supplemented with TMAO (0.12%) at time of weaning, and then lever harvested at time of sacrifice (20 weeks of age). Mouse liver lysate (30 µg protein) was run on SDS PAGE and then transferred to PVDF membrane. The membranes were probed with antibodies against Srb1 (Novus, Littleton, CO) and β-actin (Sigma, St. Louise, MO), and intensity of bands quantified by densitometry using ImagePro Plus software. Data are expressed as means ± SE

135 Supplementary Figure Small intestines expression profile of bile acid transporters in mice. Intestines from C57BL/6J, Apoe / female mice on the indicated diets enrolled in the in vivo reverse cholesterol transport experiments were harvest at completion of the study en-block. The small intestines were resected, extended lengthwise, and divided into 5ths. Tissue RNA was isolated from each segment and relative expression (to β-actin) of the indicated genes determined by RT-PCR as described in Methods. Data are expressed as mean ± SE. Note that there is no statistically significant differences noted in the expression pattern of the monitored genes along the length of the small intestines when comparing the pattern in chow vs. TMAO dietary groups of animals, as assessed by ANOVA. Ost-α; solute carrier family 51, alpha subunit; Asbt, solute carrier family 10, member

136 Supplementary Figure Small intestines expression profile of cholesterol transporters in mice. Intestines from C57BL/6J, Apoe / female mice on the indicated diets enrolled in the in vivo reverse cholesterol transport experiments were harvest at completion of the study en-block. The small intestines were resected, extended lengthwise, and divided into 5ths. Tissue RNA was isolated from each segment and relative expression (to β actin) of the indicated genes determined by RT-PCR as described in Methods. Data are expressed as mean ± SE. P values for differences in the distribution of expression patterns of the monitored genes along the length of the small intestines when comparing the chow vs. TMAO dietary groups of animals were assessed by ANOVA

137 CHAPTER 3: Carnitine, a Nutrient Found in Red Meat and a Frequent Additive by the Nutritional Supplement Industry, Can Induce the Human Gut microbiota to Produce Proatherogenic TMAO Authors: Robert A. Koeth, Bruce S. Levison, PhD, Zeneng Wang, PhD, Jill Gregory, Stanley L. Hazen, MD, PhD Intro: The pathogenesis of cardiovascular disease has been linked to gut flora metabolism of carnitine to trimethylamine N-oxide (TMAO) 73. Dietary production of TMAO promotes atherosclerosis and plasma concentrations independently associate with cardiovascular disease 69,73. Dietary carnitine is principally found in red meat and recently has become a frequent additive to the multi-billion dollar energy drink industry. We recently reported that omnivorous subject gut microbiota has a greater capacity to metabolize carnitine to TMAO compared to vegan/vegetarians, and mice placed on a chronic carnitine diet also had an increased capacity to metabolize carnitine to TMAO 73. This raised the possibility that chronic carnitine ingestion in humans can induce the gut microbiota to produce the atherogenic gut microbiota metabolite TMAO. Methods: Volunteer subjects with no history of chronic disease, recent infection, or recent antibiotic/probiotic use were enrolled to perform an oral carnitine challenge (250 mg d3-carnitine synthesized as previously described 73. Subjects were then placed on oral carnitine supplement (500 mg daily (L-Carnitine capsules) and rechallenged in 2 follow-up visits. The interval time between the baseline visit and visit 1 was 1 month and 2-3 months from the baseline visit to 135

138 visit 2. TMAO/d3-TMAO measurements were quantified in sequential venous blood draw plasma by stable isotope dilution LC/MS/MS at the indicated times as described way ANOVA analysis was performed on composite study of d3- carnitine challenge and a 1-way ANOVA was performed on the baseline TMAO plasma levels. The P values represent overall differences between groups. Results: Quantification of d3-tmao in plasma from serial venous blood draws demonstrates an increase in d3-tmao production post carnitine challenge at each subsequent visit (Fig. 3-1a). We noted great variability in the both the kinetics and capacity of individual gut microbiota to metabolize d3-carnitine. However, increases in d3-tmao production from all five individual gut microbiota studied were noted at subsequent visits and the composite analysis of all subjects challenged revealed a significant increase in the gut microbiota capacity to metabolize d3-carnitine (Fig. 3-1a). Baseline plasma measurements of TMAO at each visit also revealed significant increases in fasting TMAO levels that are comparable to concentrations in mice supplemented with dietary carnitine with accelerated atherosclerosis (Fig. 3-1b) 73. Remarkably, the dosage subjects received in this study is comparable to the mass of carnitine in energy drinks found on today s market and the total content of carnitine found in a 8 ounce steak 84. Comment: These data demonstrate that chronic carnitine supplementation can increase the capacity of the gut microbiota to produce TMAO. The important physiologic role of carnitine in fatty acid metabolism has led to the pervasive belief (and use by the nutritional supplement industry) that oral consumption of 136

139 carnitine is beneficial in energy expenditure, when, in fact, there is no compelling evidence suggesting any enhancement in healthy individuals. This is particularly concerning as the energy drink industry frequently adds carnitine to beverages and markets to adolescents and young adults 146. These studies demonstrate that frequent consumption of dietary carnitine can induce gut flora capacity to produce TMAO and may be priming our gut flora to become proatherogenic at an alarmingly young age. 137

140 Figure 3-1. Carnitine supplementation can induce the gut microbiota. a) Composite plasma tracings of d3-tmao in sequential venous blood draws post oral d3-carnitine challenge in n=5 subjects at baseline, visit 1 (V1), and visit 2 (V2). 2-way ANOVA analysis reveals that plasma d3-tmao production is significantly higher after carnitine supplementation. Points represent means + SE at T=0, 2, 4, 6, 8, and 24 hours. b) Baseline plasma TMAO measurements of subjects in d3-carnitine challenge. Bars represent means + SE. One way-anova analysis was used to assess differences between groups. 138

141 CHAPTER 4 c,d : Intestinal Microbial Metabolism of Phosphatidylcholine and Cardiac Risk 147 Authors: W. H. Wilson Tang, MD, Zeneng Wang, PhD, Bruce S. Levison, PhD, Robert A. Koeth, Earl B. Britt, MD, Xiaoming Fu, Yuping Wu, PhD, Stanley L. Hazen, MD, PhD Abstract Background: Recent animal studies show a mechanistic link between intestinal microbial metabolism of the choline moiety in dietary phosphatidylcholine and coronary artery disease pathogenesis via production of a pro-atherosclerotic metabolite, trimethylamine-n-oxide. In this study we investigated the relationship between intestinal microbiota-dependent metabolism of dietary phosphatidylcholine, trimethylamine-n-oxide levels, and adverse cardiac events in humans. Methods: We quantified plasma trimethylamine-n-oxide, choline, betaine, and urine trimethylamine-n-oxide levels by liquid chromatography with online tandem mass spectrometry following phosphatidylcholine challenge (ingestion of stable isotope (d9)-labeled phosphatidylcholine and two hard-boiled eggs) in healthy individuals before and following intestinal microflora suppression with oral broadspectrum antibiotics. We further examined the relationship between fasting plasma levels of trimethylamine-n-oxide and incident major adverse cardiac c Reproduced with permission from (Tang, W.H., et al. Intestinal microbial metabolism of phosphatidylcholine and cardiovascular risk. N Engl J Med 368, (2013).), Copyright Massachusetts Medical Society. d This chapter was drafted for submission to NEJM with my being the fourth author. After a joint writing effort, lead by the primary author Dr. Wilson Tang, the final version was agreed upon, as it appears as Chapter 4 in this dissertation. I wish to thank Dr. Wilson Tang, Dr. Stanley Hazen, and my coauthors (listed above) for their contributions. 139

142 events (death, myocardial infarction, or stroke) over 3-year follow-up in 4,007 stable patients undergoing elective coronary angiography. Results: Time-dependent increases in levels of both trimethylamine-n-oxide and its d9 isotopologue, as well as other choline metabolites, were detected following phosphatidylcholine challenge. Plasma levels of trimethylamine-n-oxide were markedly suppressed following antibiotics, and reappeared after cessation of antibiotics. Higher levels of plasma trimethylamine-n-oxide were associated with increased risk of major adverse cardiovascular events (Hazard ratio for highest versus lowest quartile, 2.5; 95% confidence interval ; p<0.001). Elevated trimethylamine-n-oxide levels predicted risk of major adverse cardiovascular events following adjustments for traditional risk factors (p<0.001), as well as in lower-risk subgroups. Conclusion: Trimethylamine-N-oxide production from dietary phosphatidylcholine in humans is dependent on metabolism by the intestinal microbiota. Higher trimethylamine-n-oxide levels are associated with higher risk of incident major adverse cardiovascular events. Introduction The phospholipid phosphatidylcholine (lecithin) is the major dietary source of choline, a semi-essential nutrient that is part of the B-complex vitamin family 148,149. Choline has various metabolic roles ranging from its essential involvement in lipid metabolism and cell membrane structure, to serving as a precursor for synthesis of the neurotransmitter acetylcholine. Choline and some 140

143 of its metabolites, like betaine, can also serve as a source of methyl groups that are required for proper metabolism of certain amino acids, such as homocysteine and methionine 150. There is a growing awareness that intestinal microbial organisms, collectively termed "microbiota", participate in global metabolism of their host 42,151,152. We recently demonstrated a potential role of a complex phosphatidylcholine/choline metabolism pathway involving gut microbiota in contributing to the pathogenesis of atherosclerotic coronary artery disease in animal models 69. We also reported an association between history of prevalent cardiovascular disease and elevated fasting plasma levels of trimethylamine-noxide, an intestinal microbiota-dependent metabolite of the choline headgroup of phosphatidylcholine that is excreted in the urine 69, Herein, we examine the relationship between oral intake of phosphatidylcholine and the involvement of the intestinal microbiota in formation of trimethylamine-n-oxide in humans. We also further examine the relationship between fasting plasma levels of trimethylamine-n-oxide and long-term risk for occurrence of incident major adverse cardiac events. Results Role of intestinal microbiota in metabolism of dietary phosphatidylcholine For the 40 participants in the phosphatidylcholine challenge study, plasma levels of trimethylamine-n-oxide are shown in Fig. 4-1, and plasma levels of choline and betaine in Supplementary Fig Endogenous (non-labeled) 141

144 trimethylamine-n-oxide (Fig. 4-1c), choline, and betaine (Supplementary Fig. 4-1c) were present in fasting plasma at baseline. Both trimethylamine-n-oxide and d9-trimethylamine-n-oxide were readily detected in plasma following the dietary phosphatidylcholine challenge at Visit 1 (Fig. 4-1a,b, left panels). Timedependent increases in both the natural isotopes (Fig. 4-1c, left panel) and d9- tracer forms (Fig. 4-1d, left panel) of trimethylamine-n-oxide were also observed postprandially. Examination of 24-hour urine specimens following the phosphatidylcholine challenge also showed the presence of trimethylamine-noxide and d9-trimethylamine-n-oxide (Supplementary Fig. 4-2, left panels). A strong correlation was observed between plasma and both absolute urine trimethylamine-n-oxide concentrations (Spearman s R=0.58, P<0.001) and urinary trimethylamine-n-oxide-to-creatinine ratio (Spearman s R=0.91, P<0.001). Time dependent increases in the plasma levels of both the natural isotopes and d9-tracer forms of choline and betaine also increased following ingestion the phosphatidylcholine challenge (Supplementary Fig. 4-1c,d, left panels). Suppression of intestinal microflora by the administration of oral broadspectrum antibiotics for one week (in six of the participants) resulted in nearcomplete suppression of detectable trimethylamine-n-oxide in fasting plasma (during Visit 2), as well as both trimethylamine-n-oxide and d9-trimethylamine-noxide following phosphatidylcholine challenge in both plasma (Fig.4-1, center panels) and urine (Supplementary Fig. 4-2, center panels). In parallel analyses, post-prandial elevations in plasma trimethlyamine and d9-trimethylamine were 142

145 observed following phosphatidylcholine challenge, but were completely suppressed to non-detectable levels following antibiotics (data not shown). In contrast, the time courses for post-prandial changes in free choline or betaine (naturally-occurring and d9-isotopologues) were not altered by suppression of intestinal microflora (Supplementary Fig. 4-1, center panels). Following cessation of antibiotics and reacquisition of intestinal microflora over the ensuing one month or longer, phosphatidylcholine challenge (at Visit 3) again resulted in readily detectable and time-dependent changes in trimethylamine-n-oxide and d9-trimethylamine-n-oxide in plasma (Fig. 4-1, right panels) and urine (Supplementary Fig. 4-1, right panels). Consistent with recent reports observing variable recovery of intestinal microbiota composition after antibiotic cessation 48,158, the extent to which trimethylamine-n-oxide levels in plasma at Visit 3 returned to pre-antibiotic levels was variable. Correlation of plasma levels of trimethylamine-n-oxide with major adverse cardiovascular events The baseline characteristics of the 4,007 participants in the clinical outcomes study are shown in Table 4-1. The mean age of the participants was 63 years, and two-thirds were male; the prevalence of cardiovascular risk factors was high and most had at least single-vessel coronary disease. Participants with incident major adverse cardiovascular events during three years of follow-up had higher risk profiles than those without events, including greater age, higher rates of 143

146 diabetes, hypertension, and prior myocardial infarction, and higher fasting glucose levels. As noted in Table 1, participants with major adverse cardiovascular events at three years of follow-up also had higher baseline levels of trimethylamine-noxide (median(interquartile range) 5.0( ) µm versus 3.5( ) µm, P<0.001). Compared to participants in the lowest quartile level of trimethylamine- N-oxide, the highest quartile had a 2.5-fold increased risk of an event (HR 2.5, 95% CI ; P<0.001, Table 4-2 and Supplementary Table 4-1). After adjusting for traditional risk factors and other baseline covariates, elevated plasma levels of trimethylamine-n-oxide remained a significant predictor of risk of major adverse cardiovascular events (Table 4-2). We observed a graded increase in the risk of major adverse cardiovascular events associated with increasing levels of trimethylamine-n-oxide, as illustrated in the Kaplan-Meier analysis shown in Figure 4-2. A similar graded increase in risk was observed when levels of trimethylamine-n-oxide were analyzed as a continuous variable in increments of one standard deviation (unadjusted HR 1.4, 95% CI ; P<0.01; adjusted HR 1.3, 95% CI , P<0.01). When components of the composite primary outcome (major adverse cardiovascular events) were analyzed separately, higher levels of trimethylamine-n-oxide remained significantly associated with higher risk of death (HR 3.2, 95%CI ; P<0.001) and non-fatal myocardial infarction or stroke (HR 2.3, 95%CI ; P<0.001). Inclusion of trimethylamine-n-oxide as a covariate resulted in a significant improvement in risk estimation over traditional 144

147 risk factors (net reclassification improvement 8.6%, P<0.001; integrated discrimination improvement 9.2%, P<0.001; C-statistic 68.3% vs. 66.4%, P=0.01). In a separate analysis, we excluded all participants who underwent revascularization within the 30 days following enrollment in the study. In this subcohort (n = 3,475), trimethylamine-n-oxide remained significantly associated with risk of major adverse cardiovascular events [highest quartile versus lowest quartile, unadjusted HR (95% CI), 2.47 ( ); adjusted HR (95% CI) 1.79 ( ); both P<0.001]. Correlation of trimethylamine-n-oxide levels with risk in low-risk subgroups The prognostic value of elevated plasma levels of trimethylamine-n-oxide remained significant in various subgroups associated with reduced overall cardiac risks (Supplementary Fig. 4-3). Subgroups examined included those who were younger, females, those without known history of coronary artery disease or coronary disease risk equivalents, those with lower-risk lipid and apolipoprotein levels, those with normal blood pressure, non-smokers, and those with lower levels of other known risk markers such as C-reactive protein, myeloperoxidase, or white blood cell count. Discussion Recent animal model studies with germ-free mice suggest a role for the intestinal microbial community in the pathogenesis of atherosclerosis in the setting of a diet 145

148 rich in phosphatidylcholine via formation of the metabolite trimethylamine and conversion to trimethylamine-n-oxide (Fig. 4-3) 69,70. Herein we demonstrate the generation of the pro-atherogenic metabolite trimethylamine-n-oxide from dietary phosphatidylcholine in humans through use of stable isotope tracer feeding studies. We further demonstrate a role for the intestinal microbiota in production of trimethylamine-n-oxide in humans via both its suppression with oral broadspectrum antibiotics, and then reacquisition of trimethylamine and trimethylamine-n-oxide production from dietary phosphatidylcholine following cessation of antibiotics and intestinal recolonization. Finally, we demonstrate the potential clinical prognostic significance of this intestinal microbiota-dependent metabolite by showing that fasting plasma trimethylamine-n-oxide levels predict development of incident major adverse cardiovascular events independent of traditional cardiovascular risk factors, presence or extent of coronary artery disease, and within multiple lower risk subgroups, including both primary prevention subjects and subjects with lower-risk lipid and apolipoprotein levels. The present findings suggest that intestinal microbial organism-dependent pathways may contribute to the pathophysiology of atherosclerotic coronary artery disease in humans, and suggest new potential therapeutic targets. The intestinal microflora have previously been implicated in complex metabolic diseases like obesity 42,151,152, However, involvement of microflora in the inception of atherosclerosis in humans has only recently been suggested 69,162. The ability of oral broad-spectrum antibiotics to temporarily suppress the production of trimethylamine-n-oxide is a direct demonstration that 146

149 intestinal micro-organisms play an obligatory role in trimethylamine-n-oxide production from phosphatidylcholine in humans. Intestinal microbiota convert the choline moiety of dietary phosphatidylcholine into trimethylamine, which is subsequently converted into trimethylamine-n-oxide by hepatic flavin-containing mono-oxygenases (Fig. 4-3) 71,163. The requirement for trimethylamine to be converted into trimethylamine-n-oxide by hepatic flavin-containing monooxygenases 164 may help to explain the observed delay in the detection of plasma d9-trimethylamine-n-oxide levels following oral ingestion of d9- phosphatidylcholine, since separate analyses monitoring trimethylamine and d9- trimethylamine production show a time course consistent with a precursor-toproduct relationship (not shown). Interestingly, trimethylamine-n-oxide has been identified in fish as an important osmolite, 165 and fish ingestion raises urinary trimethylamine-n-oxide levels. Nevertheless, the high correlation between urine and plasma levels argues for effective urinary clearance of trimethylamine-noxide. Hence, an efficient excretion mechanism may be protective in preventing the accumulation of trimethylamine-n-oxide and does not undermine the mechanistic link between trimethylamine-n-oxide and cardiovascular risk. While an association between infectious organisms and atherosclerosis has previously been postulated, studies looking at the role of antimicrobial therapy in preventing disease progression have been disappointing 166,167. It is important to recognize that the choice of antimicrobial therapy in prior intervention trials was largely based on targeting postulated organisms rather than modulating intestinal microflora composition or their metabolites. Further, 147

150 even if an antibiotic initially suppressed trimethylamine-n-oxide levels, the durability of that effect with chronic intervention remains unknown. Indeed, in unpublished studies we observed that chronic use (e.g. half year) of a single antibiotic (ciprofloxacin) that initially fully suppressed plasma TMAO levels in a rodent model completely lost its suppressive effect, consistent with expansion of antibiotic resistant intestinal microflora (Z. Wang and S.L. Hazen, unpublished). Thus, instead of suggesting that intestinal microbes should be eradicated with chronic antibiotics, the present findings imply that plasma trimethylamine-n-oxide levels may potentially identify a pathway within intestinal microflora amenable to therapeutic modulation. For example, our data suggest that excessive consumption of dietary phosphatidylcholine and choline should be avoided; a vegetarian or high-fiber diet can reduce total choline intake 159. It also should be noted that choline is a semi-essential nutrient and should not be completely eliminated from the diet, as this can result in a deficiency state. However, standard dietary recommendations, if adopted, will limit phosphatidylcholine- and choline-rich foods since these are also typically high in fat and cholesterol content 148. An alternative potential therapeutic intervention is targeting intestinal microbial organism composition or biochemical pathways, either with a functional food such as a probiotic 160, or even a pharmacologic intervention. This latter intervention hypothetically could take the form of either an inhibitor to block specific microbial metabolic pathways, or even a short course of nonsystemic antibiotics to reduce the burden of trimethylamine-n-oxide-producing microbes, as seen in the treatment of irritable bowel syndrome 168. Further 148

151 studies are warranted to establish whether antimicrobial targeted therapies can significantly reduce cardiovascular risk. In summary, we demonstrated that intestinal microbes participate in phosphatidylcholine metabolism to form circulating and urinary trimethylamine-noxide in humans. We also established a correlation between high plasma levels of trimethylamine-n-oxide and higher risk of incident major adverse cardiovascular events independent of traditional risk factors, even in lower-risk cohorts. Acknowledgements We thank Linda Kerchenski and Cindy Stevenson for assistance in subject recruitment, and Amber Gist and Naomi Bongorno for assistance in the preparation of figures and the manuscript. Mass spectrometry instrumentation used was housed within the Cleveland Clinic Mass Spectrometry Facility with partial support through a Center of Innovation by AB SCIEX. Methods Study patients and design We designed and performed two prospective clinical studies, which were funded by the National Institutes of Health and approved by the Cleveland Clinic Institutional Review Board. All participants gave written informed consent. The first study (the phosphatidylcholine challenge study) enrolled 40 healthy volunteers 18 years of age or above, who were without chronic illness (including 149

152 known history of heart, renal, pulmonary, or hematologic disease), without active infection and not currently (or within preceding month) taking antibiotics or probiotics. Participants underwent a dietary phosphatidylcholine challenge (see below) during Visit 1. Among these study participants, six were then given metronidazole 500 mg twice daily plus ciprofloxacin 500 mg once daily for one week, and a repeat phosphatidylcholine challenge was performed after antibiotics (Visit 2). A third and final phosphatidylcholine challenge was performed one month or longer following cessation of antibiotics and reacquisition of gut flora (Visit 3). After each challenge, choline metabolites were measured in plasma and urine as described below. The second study (the clinical outcomes study) enrolled 4,007 stable adults 18 years of age or older, who were undergoing elective diagnostic cardiac catheterization with cardiac troponin I less than 0.03 ug/l and no evidence of acute coronary syndrome. History of cardiovascular disease was defined as a documented history of coronary artery disease, peripheral artery disease, coronary or peripheral revascularization, 50% or greater stenosis of one or more vessels during coronary angiography, or remote history of either myocardial infarction or stroke. Fasting blood samples were obtained at the time of cardiac catheterization on all participants. Routine laboratory tests were measured on the Abbott Architect platform (Abbott Laboratories, Abbott Park IL) except for myeloperoxidase, which was determined using the CardioMPO test (Cleveland Heart Labs, Inc., Cleveland, OH). Creatinine clearance was estimated by the Cockcroft-Gault equation. Trimethylamine-N-oxide was measured in plasma as 150

153 described below. Major adverse cardiovascular events (defined as all-cause mortality, non-fatal myocardial infarction, and non-fatal stroke) were ascertained and adjudicated for all participants over the ensuing three years following enrollment. Dietary phosphatidylcholine challenge A simple dietary phosphatidylcholine-choline challenge test was administered to all participants in the first study. For each participant, baseline blood and spot urine samples were obtained following an overnight (12 hours or longer) fast. At baseline, participants were provided two large hard-boiled eggs including yolk (containing approximately 250 mg of total choline each) to be eaten within a 10- minute period together with 250 mg of deuterium-labeled phosphatidylcholine [(d9-trimethyl)-dipalmitoylphosphatidylcholine, d9-phosphatidylcholine] contained in a gelatin capsule as a tracer (administered under an Investigational New Drug exemption). Serial venous blood sampling was performed at 1, 2, 3, 4, 6 and 8 hours post-baseline, along with a 24-hour urine collection. The high-purity d9-(trimethyl)-phosphatidylcholine (greater than 98% isotope enrichment) provided was synthesized from 1-palmitoyl,2-palmitoyl,snglycero-3-phosphoethanolamine following exhaustive methylation with d3- methyliodide (Cambridge Isotopes Laboratories Inc, Andover MA). The d9- phosphatidylcholine was isolated by sequential preparative thin layer chromatography and high performance liquid chromatography, and crystallized 151

154 and dried under vacuum. Its purity (greater than 99%) was confirmed by both multinuclear nuclear magnetic resonance spectroscopy and mass spectrometry. Measurements of choline metabolites Plasma aliquots were isolated from whole blood collected into ethylenediaminetetraacetic acid tubes, maintained at 0 to 4 C until processing within four hours, and stored at -80 C. An aliquot from each 24-hour urine collection was spun to precipitate any potential cellular debris, and supernatants were stored at -80 C until analysis. Trimethylamine-N-oxide, trimethylamine, choline, betaine and their d9-isotopologues were quantified using stable isotope dilution high-performance liquid chromatography (HPLC) with on-line electrospray ionization tandem mass spectrometry on an AB SCIEX QTRAP 5500 mass spectrometer, using d4(1,1,2,2)-choline, d3(methyl)-trimethylamine- N-oxide, and d3(methyl)-trimethylamine as internal standards. For measurement of trimethylamine in plasma, a sample aliquot was acidified (60 mm HCl final) prior to storage at -80ºC. Concentrations of trimethylamine-n-oxide in urine were adjusted for urinary dilution by analysis of urine creatinine concentration. Statistical analysis for the clinical outcomes study Student s t-test, the Wilcoxon rank-sum test for continuous variables, and the chisquare test for categorical variables were used to examine the differences between participants in the clinical outcomes study who had major adverse cardiovascular events during follow-up and those who did not. For most 152

155 analyses of outcomes, plasma trimethylamine-n-oxide levels were divided into quartiles. Where indicated, trimethylamine-n-oxide was also analyzed as a continuous variable with hazard ratio (HR) determined per standard deviation change in trimethylamine-n-oxide level. Kaplan Meier analysis with Cox proportional hazards regression was used for time-to-event analysis to determine HR and 95% confidence intervals (95% CI) for major adverse cardiovascular events. Logistic regression analyses were performed adjusting for traditional cardiac risk factors (age, gender, systolic blood pressure, history of diabetes mellitus, low-density and high-density lipoprotein cholesterol, triglycerides, and smoking history) with log-transformed high-sensitivity C-reactive protein, both alone and with myeloperoxidase, log-transformed estimated glomerular filtration rate (GFR), total leukocyte count, body mass index (BMI), medications, and angiographic extent of coronary artery disease. For subgroup analyses, logistic regression analyses were performed by adjusting for traditional cardiac risk factors and log-transformed high-sensitivity C-reactive protein. Improvement in model performance introduced by the inclusion of trimethylamine-n-oxide was evaluated using net reclassification improvement. The C-statistic was calculated using the area under the receiver-operating-characteristic (ROC) curve. Threeyear predicted probabilities of a major adverse cardiovascular event were estimated from the Cox model. All analyses were performed using R version (Vienna, Austria). P values <0.05 (two-sided) were considered statistically significant. 153

156 Variable Whole cohort (n=4,007) Without Events With P value (n=3,494) Events Age (years) 63±11 62±11 68±10 <0.001 Male Gender (%) Body mass index 28.7( ) 28.7( ) 28.1 ( ) Diabetes mellitus (%) <0.001 Hypertension (%) <0.001 History of MI (%) <0.001 Number of CAD vessels* Smoking (%) LDL-c (mg/dl) 96 (78-117) 96 (78-117) 96 (75-116) HDL-c (mg/dl) 34(28-41) 34(28-41) 33(28-40) Triglycerides (mg/dl) 118 (85-170) 118 (85-169) 124 (86-173) ApoB (mg/dl) 82 (69-96) 82 (69-96) 82 (68-96) ApoA1 (mg/dl) 116 ( ) 117 ( ) 114 ( ) Fasting glucose 102 (93-119) 102 (92-117) 106 (94-135) <0.001 hscrp (ng/l) 2.4 (1-5.9) 2.3(1-5.5) 3.9( ) <0.001 MPO (pm) ( ) ( ) ( ) <0.001 GFR(ml/min/1.73m 2 ) 82 (69-95) 83 (71-96) 75 (56-89) <0.001 Total leukocyte count (WBC, x10 9 ) Baseline drugs (%): 6.1 ( ) 6.1 (5-7.5) 6.4 ( ) Aspirin ACE inhibitor/arb <0.001 Statin Beta blockers TMAO (µm) 3.7 ( ) 3.5 ( ) 5.0 ( ) <0.001 Values expressed in mean ± standard deviation or median (interquartile range). Abbreviations: ACE, angiotensin converting enzyme; ARB, angiotensin receptor blocker; ApoA1, apolipoprotein A1; ApoB, apolipoprotein B; CAD, coronary artery disease; GFR, estimated glomerular filtration rate; HDL-c, high-density lipoprotein cholesterol; hscrp, high sensitivity C-reactive protein; LDL-c, low-density lipoprotein cholesterol; MI, myocardial infarction; MPO, myeloperoxidase; TMAO, trimethylamine N-oxide; WBC, white blood cell *Defined as a coronary stenosis of 50% or greater. Table 4-1. Baseline characteristics

157 TMAO (range) Quartile 1 Quartile 2 Quartile 3 Quartile 4 Range < Major adverse cardiac events (Death, myocardial infarction, stroke) Unadjusted HR ( ) 1.5 ( )** 2.5 ( )** Adjusted HR Model ( ) 1.3 ( ) 1.9 ( )** Model ( ) 1.2 ( ) 1.6 ( )** Model ( ) 1.1 ( ) 1.4 ( )* ** p<0.01; HR, * p<0.05. Cox Proportional Hazards analyses variables were adjusted to +1 standard deviation increment for continuous variables. Model 1: Adjusted for traditional risk factors (age, gender, smoking, systolic blood pressure, low density lipoprotein cholesterol [LDL], high-density lipoprotein cholesterol [HDL], and diabetes mellitus), plus log-transformed hscrp Model 2: Adjusted for traditional risk factors, plus log-transformed hscrp, myeloperoxidase, log-transformed estimated GFR, total leukocyte count, body mass index, aspirin, statins, ACE inhibitor/arb, and beta blockers Model 3: Adjusted for traditional risk factors, plus log-transformed hscrp, myeloperoxidase, log-transformed estimated GFR, total leukocyte count, body mass index, aspirin, statins, ACE inhibitor/arb, beta blockers, and angiographic extent of disease. 147 Table 4-2. Unadjusted and adjusted hazard ratio for risks of MACE at 3-years stratified by quartile levels of TMAO

158 Figure 4-1. Human plasma levels of phosphatidylcholine Metabolites (TMAO, choline, betaine) after oral ingestion of two hard-boiled eggs and d9-phosphatidylcholine before and after antibiotics. At the top of the figure, the visit sequence is shown. All 40 study participants (healthy volunteers) participated in the first dietary phosphatidylcholine challenge (Visit 1). Six participants were then administered broad-spectrum antibiotics for one week, followed by a second phosphatidylcholine challenge (Visit 2). These same participants returned again at least one month after discontinuing antibiotics for a third challenge (Visit 3). The panels in Rows A and B show the results of assays for trimethylamine-n-oxide (Panel a) and d9-trimethylamine-n-oxide (Panel b) after the phosphatidylcholine challenge, using stable isotope dilution high-performance liquid chromatography with on-line electrospray ionization tandem mass spectrometry. Panels c and d show the time course of plasma concentrations of betaine, choline and trimethylamine-n-oxide (Panel c) and of their d9 isotopologues (Panel d). Note that, in Panel d, the concentrations of d9-trimethylamine-n-oxide are multiplied by 4; in Panel c, the concentrations of d9-trimethylamine-n-oxide are multiplied by 12, and those of choline are multiplied by 4. All left panels show data from Visit 1; center panels, from Visit 2; and right panels, from Visit

159 Figure 4-2. Kaplan-Meier estimates of long-term major adverse cardiac events, according to TMAO Quartiles

160 Figure 4-3. Pathways linking dietary phosphatidylcholine, intestinal microflora (gut flora), and incident adverse cardiovascular events. Ingested phosphatidylcholine (lecithin), the major dietary source of total choline, is acted on by intestinal lipases to form a variety of metabolic products including the choline-containing nutrients glycerophosphocholine, phosphocholine, and choline. Choline-containing nutrients that reach the cecum and large bowel may serve as fuel for intestinal microbiota (gut flora), producing trimethylamine (TMA). TMA is rapidly further oxidized to trimethylamine-n-oxide (TMAO) by hepatic flavin-containing monooxygenases (FMOs). TMAO enhances macrophage cholesterol accumulation, foam cell accumulation in the artery wall and atherosclerosis 69, and incident risk of heart attack, stroke, and death. Choline can also be oxidized to betaine in both liver and kidney 169. Dietary betaine can also serve as a substrate for bacteria to form TMA 117 and presumably TMAO

161 Supplementary Tables and Figures Characteristic (TMAO, µm) Quartile 1 (n=1001) (<2.4) Quartile 2 (n=998) ( ) Quartile 3 (n=1003) ( ) Quartile 4 (n=1005) (>6.2) P-value Age (years) 59±11 62±11 65±10 66±10 <0.001 Male Gender (%) Body mass index ( ) ( ) ( ) ( ) Diabetes mellitus (%) <0.001 Hypertension (%) <0.001 History of MI (%) # of diseased vessels < <0.001 Smoking (%) LDL-c (mg/dl) 97 (79-117) HDL-c (mg/dl) 34 (29-42) Triglycerides (mg/dl) 115 (84-166) ApoB (mg/dl) 82 (70-97) ApoA1 (mg/dl) 116 ( ) Fasting glucose 100 (91-114) hscrp (ng/l) 2.3 ( ) MPO (pm) ( ) egfr(ml/min/1.73m 2 ) 92 (81-103) Total leukocyte count (WBC, x10 9 ) 6.2 ( ) Baseline drugs (%): 98 (81-120) 35 (29-42) 115 (84-163) 83 (69-98) 117 ( ) 101 (92-116) 2.2 (1-5.6) ( ) 86 (75-96) 6.1 ( ) 95 (78-116) 34 (29-41) 121 (86-178) 82 (69-95) 117 ( ) 102 (93-119) 2.3 (1.1-5) ( ) 79 (67-91) 6.1 (5-7.5) 92 (74-114) 33 (27-40) 123 (86-180) 80 (68-94) 115 ( ) 107 (95-134) 3.1 ( ) ( ) 69 (52-83) 6.1 (5-7.5) <0.001 < <0.001 < < Aspirin ACE inhibitor/arb <0.001 Statin Beta blockers Supplementary Table 4-1. Baseline characteristics of cohort according to TMAO quartiles values expressed in mean ± standard deviation or median (interquartile range). Abbreviations: MI, myocardial infarction; LDL-c, low-density lipoprotein cholesterol; HDL-c, high-density lipoprotein cholesterol; ApoB, apolipoprotein B; ApoA1, apolipoprotein A1; hscrp, high sensitivity C-reactive protein; MPO, myeloperoxidase; WBC, white blood cell; ACE, angiotensin converting enzyme; ARB, angiotensin receptor blocker; TMAO, trimethylamine N-oxide

162 Supplementary Figure 4-1: Human plasma levels of phosphatidylcholine metabolites (TMAO, choline, betaine) after oral ingestion of two hard-boiled eeggs and d9-phosphatidylcholine before and after antibiotics. At the top of the figure, the visit sequence is shown. All 40 study participants (healthy volunteers) participated in the first dietary phosphatidylcholine challenge (Visit 1). Six participants were then administered broad-spectrum antibiotics for one week, followed by a second phosphatidylcholine challenge (Visit 2). These same participants returned again at least one month after discontinuing antibiotics for a third challenge (Visit 3). Panels a and b show the time course of plasma concentrations of betaine, choline and trimethylamine- N-oxide (Panel a) and of their d9 isotopologues (Panel b). Note that, in Panel a, the concentrations of choline are multiplied by 4, and the concentration of trimethylamine-noxide are multiplied by 12; in Panel b, the concentrations of d9-trimethylamine-n-oxide are multiplied by 4. All left panels show data from Visit 1; center panels, from Visit 2; and right panels, from Visit

163 Supplementary Figure 4-2. Human 24-hour urine levels of TMAO after oral ingestion of two hard-boiled eggs and d9-phosphatidylcholine before and after antibiotics. Numbers in label represent the mass-to-charge ratios for the precursor product ion transitions monitored for TMAO and d9-tmao

164 Supplementary Figure 4-3: Risks of major adverse cardiac events (MACE) among patient subgroups, according to baseline TMAO levels. Hazard ratios compare top to bottom quartiles. Significant interactions were observed between plasma trimethylamine-n-oxide and cigarette smoking (P=0.027) as well as plasma trimethylamine-n-oxide and plasma myeloperoxidase (P=0.012)

165 CHAPTER 5: CHAPTER 5: Intestinal Microbiota Metabolism of L-Carnitine, a Nutrient in Red Meat, Produces TMAO Via Generation of an Intermediate Gut Microbiota Metabolite γ-butyrobetaine Authors: Robert A. Koeth, Bruce S. Levison, Zeneng Wang, Jennifer A. Buffa, Elin Org, Miranda Culley, Yuping Wu, Lin Li, Jonathan D. Smith, Joseph A. DiDonato, W. H. Wilson Tang, Aldons J. Lusis, and Stanley L. Hazen Abstract Carnitine, an abundant nonessential nutrient found in red meat, was recently shown to promote atherosclerosis via generation of a gut microbiota dependent compound trimethylamine-n-oxide (TMAO) 73. These studies did not explore the possibility of an intermediate compound formed in the gut microbiota metabolism of carnitine to TMAO. Further analysis of plasma from mice fed a carnitine supplemented diet demonstrates the production of a second gut microbiota dependent trimethylamine metabolite, γ-butyrobetaine (γbb) that is both a quantitatively dominant product of gut microbiome dependent carnitine metabolism and an inducible trait. γbb is endogenously produced as the terminal precursor in the synthesis of carnitine, but serves no other known physiological function 81,82. Atherosclerotic prone mice supplemented with a γbb diet develop significantly more aortic root lesion area compared to mice on a normal chow diet and gut flora suppressed animals supplemented with γbb. Importantly, the increase in atherosclerotic disease burden cannot be attributed to abnormalities 163

166 in plasma or liver lipid metabolism. These data suggest the terminal gut flora product TMA/TMAO is promoting the atherosclerotic disease process and that γbb is a gut flora intermediate in carnitine metabolism to TMA/TMAO. Remarkably, microbiome analysis of mice fed carnitine diet and γbb diets demonstrate differing bacterial compositions suggesting cooperation of two distinct microbiota populations in the sequential 2 step gut microbiota dependent metabolism of carnitine to TMAO (e.g. carnitine to γbb to TMAO). γbb further extends our knowledge on gut flora metabolism of carnitine by defining a second, dominant pathway for TMAO formation and revealing further complexity of the proatherogenic microbiome. Introduction The gut microbiome participates in the pathogenesis of complex disease phenotypes such as diabetes, obesity, and more recently atherosclerosis 36,43,67,69. The gut microbiome promotes atherosclerosis by the gut microbiota metabolism of phosphatidylcholine, the major dietary source of choline, and resulting production of the proatherogenic metabolite TMAO (Gut Pathway 1 ) 69. TMAO increases scavenger receptor expression (CD36 and SRA) resulting in enhanced foam cell formation and causes gut microbiota dependent dysfunction in the reverse cholesterol transport pathway by disruption of bile acid synthesis 73. Moreover, plasma levels of TMAO associate with cardiovascular disease and independently predict prospective near and long-term Major Adverse Cardiovascular Events (MACE) 147. Together these data suggest a like between dietary trimethylamines, the gut microbiota, and atherosclerosis. 164

167 Microbiota Gut Pathway 1: Choline TMA FMOs TMAO Recently, we demonstrated that L-carnitine, a dietary trimethylamine found principally in red meat, can also be metabolized by the gut microbiota to produce TMA/TMAO 73. Animals fed a L-carnitine supplemented diet developed gut microbiota dependent accelerated atherosclerosis that associates with TMA/TMAO production. Further analyses reveal that TMAO significantly associates with carnivorous eating habits and together with microbiome composition can distinguish dietary patterns in humans and mice. This suggests gut microbiome metabolism of carnitine can partly explain the commonly observed association between a carnivorous diet and atherosclerosis 73 (Gut Pathway 2). Microbiota Gut Pathway 2: L-carnitine TMA FMOs TMAO Although the metabolism of L-carnitine to TMA/TMAO is clear, the pathways and enzymes bacteria use to metabolize are not. Our previous studies did not consider the possibility that carnitine may produce other gut microbiota metabolites. Studies in a rat model using a radioactive isotope of L-carnitine suggested that another trimethylamine, γ-butyrobetaine (γbb), is produced from L-carnitine (Gut Pathway 3) 94. Gut Pathway 3: L-carnitine Microbiota γ-bb 165

168 γ-butyrobetaine (γbb) is a trimethylamine containing compound that is used as a dietary supplement and endogenously produced as the terminal precursor in the production of endogenous carnitine (Endogenous Pathway) 81,82. Little is known regarding the relationship between γbb and the gut microbiota. Endogenous Pathway: Lysine TML HTML TMABA γbb L-carnitine The gut microbiota metabolism of carnitine to γbb raised the possibility that direct formation of TMA/TMAO from L-carnitine is principally mediated an intermediate metabolite γbb formation (Hypothesized Gut Pathway). Microbiota? Hypothesized Gut Pathway: L-carnitine γ-bb TMA FMOs TMAO Herein, we demonstrate that the gut microbiota metabolite, γbb, is the dominant metabolite of gut microbiota metabolism of carnitine. γbb produces TMA/TMAO and promotes atherosclerosis in a gut microbiota dependent fashion and production of γbb from L-carnitine is an inducible trait. Moreover, gut microbiome characterization studies reveal the production of γbb from L-carnitine and the concordant production of TMA/TMAO from γbb associate with entirely separate bacterial taxa suggesting a more complex microbiome than previously anticipated. 166

169 Results Gut microbiota metabolism of L-carnitine produces γbb Survey of the literature revealed that the production of γbb by the gut microbiota may be a gut microbiota product of L-carnitine 94. This observation raised the possibility that γbb could be another gut microbiota metabolite contributing to the atherosclerotic disease process directly or by further metabolism into TMA/TMAO. Quantification of γbb in plasma by LC/MS/MS from mice on a L- carnitine supplemented diet demonstrated an almost 100 fold increase in plasma concentration compared to chow fed control or antibiotic controls (Figure 5-1). Remarkably, plasma concentrations of γbb exceeded the concentration of plasma TMA or TMAO in L-carnitine supplemented mice by approximately 2-fold suggesting its production was also the major gut microbiota metabolite produced from L-carnitine (Figure 5-1). The gut microbiota dependent production of γbb from L-carnitine was confirmed by performing a L-carnitine challenge (d3-lcarnitine direct gastric challenge) in female Germ Free Swiss Webster mice. Post challenge measurements of d3-γbb in Germ Free mice demonstrate absolutely no production of d3-γbb. However, following acquisition of the gut microbiota, mice rechallanged demonstrated the ability to produce d3-γbb (Figure 5-2). γbb produces TMA/TMAO in a gut microbiota dependent manner The production of γbb from L-carnitine raised the possibility that γbb could contribute to TMA/TMAO formation by serving as an intermediate in the gut 167

170 microbiota metabolism of L-carnitine. C57BL/6J 12 week old female mice were challenged with d9-γbb chloride and followed with serial venous bleeds for 12 hours. Quantification of d9-containing trimethylamines in plasma by LC/MS/MS revealed that both d9-tma and d9-tmao were produced (Figure 5-3, first panel). Interestingly, d9-l-carnitine was also produced. Gut microbiota suppression with broad spectrum antibiotics and rechallenge demonstrated the complete absence of d9-tma/tmao production confirming gut microbiota dependence (Figure 5-3, second panel). In contrast, d9-carnitine was produced in a similar concentration compared to the initial conventional challenge establishing that L-carnitine is not a gut microbiota product of γbb. Instead, d9- γbb appears to be absorbed and shuttled through the endogenous L-carnitine synthetic pathway. This is supported by the early peak of d9-γbb in plasma followed by the gradual increase in d9-l-carnitine concentration in plasma over the 12 hour period. Conventionalization of the experimental mice demonstrated reacquisition of the capacity of the gut microbiota to produce d9tma/tmao while d9-l-carnitine levels remained similar to concentrations in the two previous challenges (Figure 5-3, last panel). The gut microbiota dependent formation of d9tma/tmao from d9-γbb was confirmed by challenging Germ free female Swiss Webster mice immediately upon receipt (Figure 5-4). As observed in antibiotic suppressed C57BL/6J female mice, both d9-tma/tmao was absent and d9 L-carnitine was produced. Conventionalization of germ free mice with inhouse mice, demonstrate the acquisition of the capacity for the gut microbiota to 168

171 produce d9-tma/tmao while the amount of d9-carnitine production remained the same(figure 5-4). TMA formation occurs in the cecum and γbb is the dominant gut microbiota product of L-carnitine gut microbiota metabolism Incubation of equal molar amounts of d3-l-carnitine or d9-γbb with segments of mouse intestines demonstrate the production of TMA mainly in the bacterial rich cecum of mice for both d3-l-carnitine and d9-γbb suggesting the cecum is the major site of TMA generation from dietary trimethylamines (Figure 5-5). In contrast, the production of γbb was more uniformly distributed along the distal intestinal tract (Figure 5-5). d9-γbb was also more readily metabolized to TMA than d3-l-carnitine by approximately 2-fold (Figure 5-5). Moreover, quantitative comparison of the production d3-γbb and d3-tma from in vitro cecal studies show a remarkable 1000 fold increase in d3-γbb production over d3-tma demonstrating that d3-γbb is the dominant gut microbiota metabolite produced from the intestinal gut microbiota (Figure 5-5; lower panel). Metabolism of γbb by the gut microbiota to TMA/TMAO promotes atherosclerosis The gut microbiota dependent production of TMA/TMAO from γbb and the dominant production of γbb from L-carnitine raised the possibility that γbb was contributing to the development of atherosclerosis either indirectly by serving as an intermediate between the terminal metabolism of L-carnitine to TMA or as a 169

172 direct gut microbiota L-carnitine metabolite. C57BL/6J, Apoe-/- female mice were placed on a chow diet or γbb (1.3%) diet with respective gut microbiota suppression controls (+ABS) at weaning for 15 weeks before necroscopy (Figure 5-6). Quantification of atherosclerotic plaque at the aortic root revealed an approximately 1.5 fold increase in total area of plaque in γbb animals compared to controls (Figure 5-6). Importantly, there was no increase in total plaque area in mice on a γbb diet with gut microbiota suppression (+ABS) suggesting that TMA and TMAO was the gut microbiota product promoting atherosclerotic disease and not γbb. These data were confirmed by quantification of TMA/TMAO in terminal mouse plasma samples by LC/MS/MS. Both TMA/TMAO were produced in a quantitatively dominant amount in mice fed a γbb diet compared to the respective chow or gut microbiota suppressed controls (Figure 5-7). Additionally, +ABS, γbb supplemented mice failed to demonstrate an increase in atherosclerotic aortic root plaque yet had the highest plasma γbb concentrations among mice in the study (Figure 5-7). Metabolism of γbb from L-carnitine is an inducible trait We previously reported that metabolism of TMA/TMAO from L-carnitine by the gut microbiota is an inducible trait. This suggested that the production of γbb from L-carnitine may also be inducible. To test this possibility, a L-carnitine challenge (d3-l-carnitine) was performed on mice on a L-carnitine supplemented diet and control chow fed mice respectively over a 12 hour period. LC/MS/MS 170

173 analysis of plasma from serial venous draws from mice reveal that production of plasma d3-γbb is an inducible trait (Figure 5-8). γbb associates with a microbiome composition that differs from TMA/TMAO formation We previously demonstrated that production of TMA/TMAO associated with microbiome genera of mice on a L-carnitine diet 73. The gut microbiota dependent production of γbb from L-carnitine naturally raised the possibility that production of γbb may also associate with the gut microbiome. Compositional analysis of the microbiota of mice on a carnitine diet with plasma γbb demonstrated significant associations even after adjustment for multiple testing (Fig. 5-9,5-10). Remarkably, analysis of the gut microbiome composition revealed distinctly different microbiome associations than previously reported with TMA/TMAO 73. γbb associates with the bacterial genera of Parasutterella, Prevotella, and Bacteroides and we previously had demonstrated that TMA/TMAO production from L-carnitine associates with Prevotella, Anaeroplasma, and Mucispirillium 73. Although these associations overlapped with the genus Prevotella, TMA/TMAO production generally had a negative association with bacteria from the Bacteroides and Proteobacteria phyla suggesting that distinct microbiota participate in the metabolism of L-carnitine. Combined analysis of plasma γbb (xaxis) and bacterial operational taxonomic units (OTUs; y Axis) further demonstrate mice can be distinguished by dietary status. 171

174 TMAO production from γbb associates with microbiome composition The dietary contribution of γbb and the production of γbb from L-carnitine data suggested that γbb can also be involved in shaping the gut microbiota. Compositional analysis of mice on a γbb diet reveal that gut microbiota metabolite TMAO associates with gut microbiome composition of microbiota from the phyla Verrucombria. This appears to be mostly driven by an association with the genus Akkermansia (Figure 5-11,5-12). Remarkably, plasma γbb metabolized from L-carnitine and microbiota composition studies demonstrate virtually no association with this genus. Together these data suggest cooperation between bacterial microbiota species in the sequential production of TMA from L- carnitine. Mice on a γbb diet have significant decreased liver expression of Cyp7a1, but not Cyp27a1 We previously demonstrated that mice on trimethylamine supplemented diets (e.g. carnitine or choline) have a significant gut microbiota dependent reduction in total reverse cholesterol transport that may be attributed to a decrease in total bile acid pool size and related bile acid synthetic enzyme 73. Thus, in a final set of studies liver expression of two key bile acid producing enzymes (Cyp7a1 and Cyp27a1) were examined (Figure 5-13). Consistent with previous studies the expression of liver Cyp7a1, the rate-limiting enzyme in the classic pathway of bile acid synthesis from cholesterol, was significantly decreased in γbb liver compared to chow fed control liver. In contrast to previous studies, expression of 172

175 Cyp27a1, an important enzyme in the classic pathway and the alternative acidic pathways of bile acid synthesis from cholesterol, was not significantly different 170 (Figure 5-13). Discussion Early studies demonstrate that mammals lack the capacity to catabolize L- carnitine 81,82,119. We previously demonstrated that catabolism of L-carnitine directly to TMA/TMAO is a gut microbiota dependent pathway. The present studies unambiguously show that gut microbiota metabolism to γbb, mainly thought to be involved only in the endogenous L-carnitine synthesis, is the dominant metabolite in L-carnitine degradation by the gut microbiota. As our data shows, even in the proximal small bowel where the bacterial load is relatively small compared to more distal parts of the GI tract L-carnitine catabolism to γbb begins. This pathway (e.g. L-carnitine catabolism to γbb) is kinetically favored over a 1,000 fold compared to direct metabolism of L-carnitine to TMA/TMAO (Figure 5-14). The importance of γbb in mammalian physiology has traditionally centered on its role in L-carnitine synthesis. γbb serves as the terminal substrate in endogenous L-carnitine production that begins with metabolism of lysine and methionine. Indeed, the majority of studies on γbb in mice and humans have been in the context of understanding endogenous L-carnitine production and its effect on L- carnitine levels Here, we demonstrate an important role for γbb in gut 173

176 microbiota metabolism of L-carnitine by demonstrating for the first time the microbiota mediated metabolism of γbb to TMA. Not only does γbb serve as the dominant metabolite of gut microbiota dependent L-carnitine metabolism, but more TMA is produced from γbb than L-carnitine on an equamolar basis. As with dietary supplementation with choline and carnitine, supplementation of γbb also increased atherosclerotic plaque area in a gut microbiota dependent manner. Importantly, these data also clarify that the gut microbiota metabolism of γbb to TMA/TMAO and not γbb directly promote atherosclerosis. This conclusion is supported by the following evidence. First, mice supplemented with γbb and cosuppression of the microbiota with antibiotics did not have an increase in total plaque area at the aortic root. Secondly, mice with the highest plasma concentrations of γbb (e.g. γbb, +ABS) did have increased atherosclerotic plaque compared to the chow control. Finally, mice with the highest plasma concentrations of TMA/TMAO had the most plaque at the aortic root compared to chow and antibiotic controls. Together these data further suggest the terminal microbiota production of TMA/TMAO is responsible. Interestingly, there was a significant decrease in plasma L-carnitine concentration in mice fed γbb diet compared to chow controls suggesting that high plasma γbb concentrations have a suppressive effect on endogenous L- carnitine production. The difference in plasma γbb in mice fed γbb to the respective gut microbiota suppressed control would suggest that γbb is being 174

177 metabolized by the gut microbiota into TMA and not being readily absorbed. This may also explain why there is also a significant decrease in plasma L-carnitine when comparing the γbb and γbb gut microbiota suppressed control (e.g. the more plasma γbb there is the less plasma L-carnitine is observed). The metabolism of γbb from L-carnitine appears to be more evenly distributed between the cecum and colon. However, the site of proatherogenic TMA production from both L-carnitine and γbb appears to be primarily localized in the cecum. The observation that both the production of γbb from carnitine and TMA/TMAO production from γbb is inducible indicates a more complex microbiome than previously anticipated. Microbiome analysis of cecums from L-carnitine supplemented mice demonstrate a gut microbiome that associates with plasma γbb. Remarkably, this analysis revealed that γbb associated with a different microbiome composition than TMA/TMAO from L-carnitine supplemented mice. Close examination between the trends of γbb and TMA/TMAO association with the gut microbiota reveal that only Prevotella commonly associated taxa between both studies (γbb and TMA/TMAO). Moreover, the microbiota taxa that associated with γbb production tended to have no association or inversely associate with TMA/TMAO production. 175

178 To further understand the role of the gut microbiota in L-carnitine metabolism, we performed compositional microbiota studies of animals on a γbb diet and analyzed it with corresponding mouse plasma TMA /TMAO concentrations. Interestingly, correlation heat maps between the gut microbiota composition of TMAO from mice on a γbb diet and TMAO from mice on a L-carnitine diet are different suggesting more complexity in the gut microbiome metabolism of carnitine than previously anticipated. Together, these data suggest the participation of two distinct microbiomes in the 2 step metabolism of L-carnitine to TMA. It also suggests cooperation between microorganisms metabolizing L- carnitine to γbb and γbb to TMA/TMAO. These observations suggest that there are multiple potential clinical therapeutic targets. Disruption of the enzymes involved in the metabolism of L-carnitine to γbb and/or inhibition of γbb to TMA, for example, may be more important potential therapeutic targets than enzymes involved in the direct production of TMA from L-carnitine. One would anticipate that a probiotic could be used that would allow one to consume steak without metabolizing carnitine to TMAO. However, further studies in human subjects to confirm the dominance of gut microbiota metabolism of carnitine to γbb are needed. The exact microorganisms responsible for metabolizing carnitine to γbb remain unclear. There are reports demonstrating that microorganisms found in the GI tract have the capacity to metabolize L-carnitine to γbb 81. Initially, the recognition 176

179 that γbb was a gut metabolite of L-carnitine was demonstrated from Escherichia coli, purified from rat intestine 174. Subsequent studies using other gut microorganisms in the Enterobacteriaceae family also have been found to metabolize L-carnitine to γbb (e.g. Escherichia coli, Proteus vulgaris, and Salmonella typhimurium). Interestingly, these microorganisms are contained with the Proteobacteria phyla. In the present studies many of the taxa positively associating with γbb production are also found in this same phyla. Although none of the associated taxa contained these bacterial genera, it does raise the possibility that this phyla of bacteria have the capacity to L-carnitine to γbb (e.g.. Proteobacteria). In contrast, there are no reports of microorganisms metabolizing TMA from γbb. Previously, we have demonstrated that the mechanisms accounting for the contribution of the terminal gut microbiota end products TMA/TMAO in the pathogenesis of cardiovascular disease are multifactorial. Mice fed a trimethylamine diet can promote atherogenesis by increasing foam cell formation by upregulation of scavenger receptors CD36 and SRA 69. Follow-up work demonstrated that dietary trimethylamines also promote atherogenesis by reducing reverse cholesterol transport by decreasing the bile acid pool size. This dysfunction was apparently being mediated by the downregulation of enzymes involved in cholesterol metabolism into bile acids and bile acid transporters 73. Here we demonstrate that like mice fed a TMAO supplemented diet, animals on a γbb diet also have dysfunctional bile synthesis evidenced by the parallel 177

180 decrease in Cyp7a1, the rate limiting enzyme in bile acid synthesis. The exact molecular explanation for the role of TMA/TMAO in dysfunctional bile acid metabolism and upregulation of SRA and CD36 remains unclear, but could be mediated by a TMA/TMAO interaction with a TMAO specific receptor. In summary, we have discovered that γbb, a trimethylamine, serves as a gut microbiota intermediate compound in the metabolism of L-carnitine to TMA/TMAO and provides an important mechanistic link in understanding the L- carnitine gut microbiota dependent promotion of atherosclerosis. Methods Materials and general procedures Mice and/or breeders were obtained from Jackson Laboratories. All animal studies were performed under approval of the Animal Research Committee of the Cleveland Clinic. Mouse plasma total cholesterol, triglycerides, and glucose were measured using the Abbott ARCHITECT platform model ci8200 (Abbott Diagnostics, Abbott Park, IL). HDL cholesterol concentration in mice used for the γbb atherosclerosis study was enzymatically determined (Stan bio, Houston, TX) from plasma HDL isolated using density ultracentrifugation as previously described 73. Liver triglyceride content was measured using the GPO reagent (Pointe Scientific, Canton, MI) and normalized to liver mass (g) grams as previously described (millard). Liver cholesterol was quantified in liver homoginates with added coprostanol (Steraloids, Inc, Newport, RI) internal 178

181 standard. Liver were lipids extracted by the Folch method (chloroform:methanol (2:1, v/v)), and then cholesterol quantified as its trimethylsilane (TMS) derivative (Sylon HTP, Sigma-Aldrich, Sigma St. Louis, MO) by GC/MS (Agilent 5973N model, Santa Clara CA) as previously described 126. Gut microbiota suppression studies were performed by dissolving antibiotics in mouse drinking and included 0.1% Ampicillin sodium salt (Fisher Scientific), 0.1% Metronidazole, 0.05% Vancomycin (Chem Impex Intl.), and 0.1% Neomycin sulfate (Gibco) as previously described 35,69. Mouse challenge and atherosclerosis studies An oral γbb or L-carnitine challenge in mice consisted of a gastric gavage of d9- γbb (prepared as described below) or d3-l-carnitine (Cambridge Isotope Laboratories; Andover, MA) dissolved in water respectively. 10-week-old female Taconic Swiss Webster germ-free mice were γbb challenged immediately upon arrival in a microisolater. The mice were then conventionalized by being housed in cages with non-sterile C57BL/6J female mice for approximately 1 month before the γbb challenge was perform again. The γbb challenge was also performed on 12-week old C57BL/6J female mice in the native state, after gut microbiota suppression with broad spectrum antibiotics for 1 month, and finally, after being housed with native mice for an approximately 3 month conventionalization period 69. Gut microbiota inducibility studies were completed by performing L-carnitine or the γbb challenge on 12 week old C57BL/6J, Apoe- /- mice on a chow diet or an L-carnitine supplemented diet for at least a 10 week 179

182 period. For the atherosclerosis study, C57BL/6J, Apoe-/- were placed on a standard chow control diet (Teklad 2018) or γbb supplemented diet (mouse drinking water with 1.3% γbb; BOC scientific) with and without antibiotics at time of weaning for a 15 week duration. The antibiotic regimen used was provided to the mouse in the drinking water as described above. Mouse aortic root plaque was prepared and quantified as previously described 69. Quantification of natural abundance and isotope labeled forms of carnitine, γbb, TMA and TMAO in mouse plasma was performed using stable isotope dilution LC/MS/MS as described below. Mouse microbiome studies Microbial community composition was assessed by pyrosequencing 16S rrna genes derived from the mice cecal samples of normal chow diet (n=16), transcrotonobetaine (n=11) and γ-butrobetaine (n=12). DNA was isolated using the MoBio PowerSoil DNA Isolation Kit according to the manufacturer s instructions. The V4 region of the 16S rrna gene was amplified using bar-coded fusion primers (F515/R806) with the 454 a Titanium sequencing adapter. The barcoded primers were achieved following the protocol described by Hamady et 133. Sample preparation was performed similarly to that described by Costello et al Each sample was amplified in triplicate, combined in equal amounts and cleaned using the PCR clean-up kit (Mo Bio). Cleaned amplicons were quantified using Picogreen dsdna reagent (Invitrogen) before sequencing using 454 GS FLX titanium chemistry at the GenoSeq Facility at the University of California, 180

183 Los Angeles. The raw data from the 454 pyrosequencing machine were first processed through a quality filter that removed sequence reads that did not meet the quality criteria. Sequences were removed if they were shorter than 200 nucleotides, longer than 1,000 nucleotides, contained primer mismatches, ambiguous bases, uncorrectable barcodes, or homopolymer runs in excess of six bases. The remaining sequences were analyzed using the open source software package Quantitative Insights Into Microbial Ecology (QIIME) 135,136. A total of 49,458 quality filtered reads were obtained from 39 samples (three samples were removed due to low sequence count). Individual reads that passed filtering were distributed to each sample based on bar-code sequences. Demultiplexed sequences were assigned to operational taxonomic units (OTUs) using UCLUST with a threshold of 97% pair-wise identity. Representative sequences were selected and BLASTed against a reference Greengenes reference database. For each resulting OTU, a representative sequences were selected by choosing the most abundant sequence from the original post-quality filtered sequence collection. The taxonomic composition was assigned to the representative sequence of each OTU using Ribosomal Database Project (RDP) Classifier The relative abundances of bacteria at each taxonomic level (e.g., phylum, class, order, family and genus) were computed for each mouse. Correlations between relative abundance of gut microbiota and TMA and TMAO levels and association testing were performed in R. False discovery rates (FDR) of the multiple comparisons were estimated for each taxon based on the P- values resulted from correlation estimates. 181

184 d9-γ-butyrobetaine chloride preparation (3-Carboxypropyl)trimethyl(d 9 )ammonium Chloride (d9-γ-butyrobetaine Chloride, d9-γbb Cl) was prepared from γ-aminobutyric (GABA) acid (Sigma #A2129) in methanol with potassium hydrogen carbonate and d3-methyl iodide by the method of Cain Morano, Xin Zhang, and Lloyd D. Fricker 175. After 72 hours, the entire reaction mixture was quantitatively transferred onto a short silica gel column (grade 60, mesh) equilibrated in methanol in a coarse fritted Buchner funnel. Non-polar material was removed by elution of the column with the 1.25 column volumes of methanol. The product d9-γbb was eluted in 2.5 column volumes of 30%v/v water in methanol. Rotary evaporation of this second eluate gave the crude product as an oily semisolid which was dissolved in water and titrated to ph 7.2 with dilute hydrochloric acid. The water was removed by rotary evaporation and final traces of moisture were removed azeotropically by distillation of absolute ethanol from the residue. The white to off-white solid was dissolved in absolute ethanol and filtered to remove residual inorganic salts. The material was concentrated to dryness and dissolved in excess diluted hydrochloric acid (3M). The resulting straw colored solution was concentration to dryness by rotary evaporation was followed by re-dissolution of the semicrystalline light amber colored salt in a minimal amount of methanol. This methanolic solution was treated with 5 volumes of acetone; the resulting almost clear solution was allowed to sit at room temperature for several hours. The resulting plate-like crystals were isolated by suction filtration, transferred to a clean container, and dried under vacuum at 60 o C. This material darkened slightly 182

185 to an off white, slightly amber free flowing powder which was stored refrigerated over desiccant. Concentrations of stock solutions of this material were determined relative to a standard curve of authentic γbb Cl, by LC/MS/MS as described below. ESI positive ion mode mass spectrum for d9-γ-butyrobetaine chloride (5µM in 50% v/v Methanol in water plus 0.1% v/v formic acid) shows a base peak at m/z [M] +, a peak at m/z corresponding to [M+Na] + and a peak at m/z corresponding to [M+K] +, MS 2 positive ion mode for m/z (collision energy 20) shows a base peak at m/z 87.2 corresponding to [M- N(CD 3 ) 3 ] + and a peak at 69.3 corresponding to [HN(CD 3 ) 3 ] + a peak at m/z 45.1 corresponding to [CO 2 H] +, and a peak at m/z 43.2 corresponding to [C 2 OH 3 ] + Quantification of TMAO, TMA, a γbb, and L-carnitine Stable isotope dilution LC/MS/MS was used to quantify trimethylamine compounds from mouse plasma samples in positive MRM mode using the supernatant from methanolic plasma precipitation. Precursor product ion transitions at m/z 76 to 58 (TMAO), m/z 60 to 44 (TMA), m/z 146 to 60 (γbb), m/z 162 to 60 (carnitine) and were used. d9(trimethyl)tmao (d9-tmao), d9(trimethyl)tma (d9-tma), d9 (trimethyl) γbb, and d9(trimethyl)carnitine (d9- carnitine), were added to mouse plasma to quantify native compound concentrations. d4-choline was used to quantify d9-γbb and d9 gut microbiota mouse products (d9-tma, d9-tmao) from d9-γbb-challenge studies. Increasing concentrations of the trimethylamines with a fixed amount of internal standard 183

186 were added to control plasma to generate calibration curves for determining plasma concentrations of each respective analyte as previously described 69. In vitro mouse cecum study C57BL/6J female mouse (n=3) cecums were harvested, sectioned longitudinally into 2 halves, and placed into 10mM Hepes PH 7.4 containing either a 150 µm d9-γbb or d3-l-carnitine respectively. Samples were placed into a sealed falcon tubes under anaerobic (in the presence of Argon) and acidic conditions (in the presence of 0.1% formic acid) for a 16 hour incubation at 37 o C. Reactions were halted by the mixing of the reaction mixture and 0.1% formic acid. A methanolic precipitation was performed and the supernatant of samples were analyzed by LC/MS/MS using d4-choline as internal standard as described above. RNA preparation and real time PCR analysis RNA was first purified from liver using the animal tissue protocol from the Qiagen rneasy mini kit. Purified total RNA and random primers were used to synthesize first strand cdna using the High Capacity cdna Reverse Transcription Kit (Applied Biosystems, Foster City, CA) reverse transcription protocol. Quantitative real-time PCR was performed using Taqman qrt-pcr probes (Applied Biosystems, Foster City, CA) and normalized to tissue β-actin by the C T method using StepOne Software v2.1 (Applied Biosystems, Foster City, CA). 184

187 General Statistics The Wilcoxon Rank-Sum test was used for two-group comparison and Spearman associations were performed for correlation studies. A robust Hotelling T 2 test was used to assess differences between dietary groups (chow or 1.3% γbb) by utilizing the proportion of specific bacterial genera and mouse plasma TMA or TMAO concentrations 132. All data was analyzed using R software version 2.15, JMP (SAS Inc, Cary NC), and Prism (Graphpad Software, San Diego, CA). 185

188 Table 5-1. Plasma and liver lipid levels in C57BL/6J, Apoe-/- female mice used in γbb atherosclerosis study. Data is expressed as means +SD. Plasma Lipids Chow γ-butyrobetaine (1.3%) P (n = 20) (n = 17) Triglyceride (mg/dl) <0.01 Total Cholesterol (mg/dl) HDL (mg/dl) Total Glucose (mg/dl) Lipids Chow,+ABS γ-butyrobetaine (1.3%), +ABS P (n = 7) (n = 18) Triglyceride (mg/dl) Total Cholesterol (mg/dl) HDL (mg/dl) Total Glucose (mg/dl) Liver Lipids Chow γ-butyrobetaine (1.3%) P (n=20) (n=17) Triglyceride (mg/g liver) Cholesterol (mg/g liver) <0.01 Lipids Chow,+ABS γ-butyrobetaine (1.3%), +ABS P (n=7) (n=18) Triglyceride (mg/g liver) Cholesterol (mg/g liver) Table 5-1. Plasma and liver lipid levels in C57BL/6J, Apoe-/- female mice used in γbb atherosclerosis study. Data is expressed as means +SD. 186

189 Figure 5-1. γbb is produced as a major gut microbiota metabolite of L-carnitine. Stable isotope dilution of LC/MS/MS of plasma γbb, carnitine, TMA, and TMAO in female terminal plasma of C57BL/6J, Apoe-/- mice on respective diets. Data is expressed as means + SE. 187

190 Figure 5-2. γbb is produced from L-carnitine in a gut microbiota dependent manner. Female Swiss Webster Germ Free mice challenged with d3-l-carnitine before and after conventionalization. Post challenge measurement of d3-l-carnitine and d3- γbb was performed in serial venous blood draws by stable isotope dilution LC/MS/MS. 188

191 Figure 5-3. TMA/TMAO is a gut a microbiota dependent product of γbb metabolism. C57BL/6J female mice (n=5) challenged with d9-γbb gastric gavage (left panels; upper (d9-carnitine and d9- γbb) and lower (d9tma/tmao)) followed with serial blood venous blood draws and quantification of plasma deuterated analytes by stable isotope dilution LC/MS/MS. Middle panels-repeat gastric gavage with d9-γbb after 1 month gut suppression with a cocktail of broad spectrum antibiotics as described in Methods. Right Panels- A final d9-γbb gastric challenge and sequential measurement of deuterated plasma compounds was performed after a month long reconventionalization period. 189

192 Figure 5-4. Confirmatory studies that TMA/TMAO is a gut a microbiota dependent product of γbb metabolism. Female Swiss Webster Germ Free mice (n=5) were challenged with d9- γbb before and after conventionalization. Post challenge measurement of d9-carnitine, d9-γbb (upper panels), d9-tma, and d9-tmao (lower panels) was performed in serial venous blood draws by stable isotope dilution LC/MS/MS. 190

193 Figure 5-5. γbb is the dominant gut microbiota metabolite of L-carnitine and is metabolized to TMA at a great equamolar capacity than L-carnitine. C57BL/6J Female mouse intestine (n=3) was sectioned into two complementary pieces for incubation with equamolar amounts of d3-l-carnitine or d9-γbb under anaerobic conditions at 37 o C for 12 hours. Deuterated trimethylamine analytes were quantified by stable isotope dilution LC/MS/MS as detailed in Methods. d9/d3-tma production by the gut microbiota from d9-γbb and d3-l-carnitine respectively occurs primarily in the cecum (top and middle panels) whereas d3-γbb production from d3-l-carnitine is more evenly distributed in the cecum and colon. d9-γbb produced more d9-tma on an equamolar basis than d3-l-carnitine produced d3- TMA (top and middle panels). d3-γbb production from d3-l-carnitine is approximately 1000 fold higher (bottom panel) than d3-tma production (middle panel). 191

194 Figure 5-6. γbb promotes atherosclerosis in a gut microbiota dependent manner. (A) Oil-red-O stained and hematoxylin counterstained representative aortic roots slides of 19 week old C57BL/6J, Apoe-/- female mice on the respective diets in the presence versus absence of gut microbiota suppression (+ antibiotics (ABS)) as described under Experimental Procedures. (B) Quantification of mouse aortic root plaque lesion area of 19 week-old C57BL/6J, Apoe-/- female mice. Mice were started on the indicated diets at the time of weaning (4 weeks of age). Lesion area was quantified as described under Methods. 192

195 Figure 5-7. Plasma trimethylamine concentrations of C57BL/6J, Apoe-/- female mice used in γbb atherosclerosis study. Carnitine, TMA, γbb, and TMAO were determined using stable isotope dilution LC/MS/MS analysis of terminal plasma recovered from γbb atherosclerotic mice. Data is expressed as means + SE. 193

196 Figure 5-8. γbb production from L-carnitine is an inducible trait. d3-l-carnitine challenge of mice on a L-carnitine supplemented diet (1.3%) at 10 weeks and age or age-matched normal chow controls. Plasma d3- γbb was measured in sequential venous blood draws at the indicated times post d3-lcarnitine oral gavage. 194

197 Figure 5-9. γbb production from L-carnitine associates with microbiome composition. A correlation heat map demonstrating the association between the indicated gut microbiota taxonomic taxa and γbb plasma levels of mice grouped by dietary status (chow, n=10 and L-carnitine, n=13). Red denotes a positive association, blue a negative association, and white no association. The single asterisk indicates a significant FDR adjusted association of P

198 Figure γbb production from L-carnitine and microbiome composition associate with dietary status. Plasma γbb concentrations were determined by stable isotope dilution LC/MS/MS (plotted on x axes) and the proportion of taxonomic operational units (OTUs, plotted on Y axes) were determined as described in Experimental Procedures. The P value shown is for comparisons between dietary groups using a robust Hotelling T 2 test. 196

199 Figure TMA/TMAO production from γbb associates with microbiome composition. A correlation heat map demonstrating the association between the indicated gut microbiota taxonomic taxa and TMA/TMAO plasma levels of mice grouped by dietary status (chow, n=15 and γbb, n=11 ). Red denotes a positive association, blue a negative association, and white no association. The single asterisk indicates a significant FDR adjusted association of P

200 Figure TMAO production from γbb and microbiome composition associate with dietary status. Plasma TMAO concentrations were determined by stable isotope dilution LC/MS/MS (plotted on x axes) and the proportion of taxonomic operational units (OTUs, plotted on Y axes) were determined as described in Experimental Procedures. The P value shown is for comparisons between dietary groups using a robust Hotelling T 2 test. 198

201 Figure Liver Expression of Bile acid enzymes. Relative mrna levels (to β-actin) of mouse bile acid synthetic enzymes liver Cyp7a1 and Cyp27a1. 199

202 Figure Scheme of endogenous and exogenous γbb production. γbb is endogenously produced as part of the L-carnitine synthetic pathway, but can also be produced by the metabolism of sources of TMA production by the gut microbiota. 200