Stan Hazen, director of the Center for Cardiovascular Diagnostics and Prevention at the Cleveland Clinic Lerner Research Institute, pored over page after page of mass spectrometry data. As he combed through the measurements of more than 6000 compounds from 50 victims of heart disease, he excitedly noted a set of metabolites related to phosphatidylcholine or lecithin digestion that showed an overwhelming association with the risk for heart attack, stroke, or death.
Hazen’s studies would lead to the first evidence of a dramatic difference in metabolism between omnivores and vegans or vegetarians. He also would show for the first time that certain bacteria in the human gut can promote heart disease.
Researchers have long known that the human body is colonized by millions of microbes, most of which can't be cultured in the lab. These bugs were commonly believed to be harmless, maybe even helpful. But Hazen’s work and other studies are altering this perception, suggesting now that the human microbiome is a vastly more complex landscape with important influences on human biology and disease.
Out of Africa
While Hazen tried to find a vegan willing to eat meat, Mark Manary and Indi Trehan from the Department of Pediatrics at the Washington University Medical School in St. Louis were busy recruiting their own volunteer group from a vastly different population. Manary and Trehan sought out newborn twins from Malawi who were at risk for developing kwashiorkor, an enigmatic form of severe acute malnutrition characterized by tremendous capillary leakage and swelling.
Since 1994, Manary had exhausted all existing methods for determining the cause of this disease, conducting prospective dietary surveys of 2500 children and double blind placebo controlled randomized trials testing the predominantly accepted explanations for what leads to kwashiorkor, but came up empty handed.
"You think of malnutrition in a simplistic way as having something to do with a deficiency in your diet, as something you're not eating," Manary said, "but the more we looked into this over 15 years, we thought this was not about what they were eating at all. They were eating the very same thing in the very same quantities in the very same places as children who weren't getting kwashiorkor. What else is in their environment that could be causing this condition?"
Trehan emailed Manary that same question after encountering a set of twins where only one had kwashiorkor. Trehan and Manary began to wonder if the condition might have something to do with gut bacteria and turned to their colleague Jeff Gordon at Washington University who had recently shown differences between the microbiomes of obese and non-obese women in the United States. “That’s a form of malnutrition too. Overnutrition is just as bad in many ways as malnutrition, so we took a look at the population that we knew, the malnourished kids,” Trehan said.
Working through diesel shortages, power outages, civil unrest, and washed out roads during the rainy season, Trehan and Manary regularly traveled to 20 clinics, collecting stool samples from 317 pairs of twins over a course of 3 years. Situations where both twins remained healthy or where one of the twins developed kwashiorkor were of particular interest, and those samples were flash frozen in liquid nitrogen and shipped to Gordon’s lab for 16S rRNA and multiplex shotgun sequencing.
Where's the Beef?
Back at the Cleveland Clinic Lerner Research Institute, Hazen’s first volunteers ate hard boiled eggs or swallowed capsules containing deuterium-labeled phosphatidylcholine, allowing him to follow the labeled compound through metabolism. TMAO levels in the volunteers’ plasma and urine before and after taking a broad spectrum antibiotic showed that the sharp increases in TMAO seen after consuming eggs or phosphatidylcholine was dependent upon intestinal microbes.
Those hours spent poring over mass spectrometry data would pay off for Hazen when he noticed that carnitine, a compound found in large quantities in red meat, was a significant hit in the original metabolomics data. Hazen didn’t focus on it until he recognized that a portion of its chemical structure was identical to choline. “There is no question that a diet rich in red meat—at least epidemiologically—is associated with increased cardiovascular risk,” Hazen said. He thought this bacteria-dependent metabolism just might provide a long sought after explanation for the connection.
Although it took eight months to find a vegan willing to eat a beef steak, several were willing to swallow labeled carnitine capsules. Hazen’s team, to their surprise, found that not only were baseline levels of TMAO significantly lower in vegan and vegetarian volunteers, but those individuals also had a reduced capacity to synthesize TMAO from carnitine.
“Bacteria, whether they live in a Petri dish or in your intestines or in someone else’s intestines or in the intestines of a mouse, they are still the same. If you feed them what they like, they will grow. If you don’t feed them what they like, their proportions will be less,” Hazen explained.
What microbes thrive on carnitine? To find out, Hazen also turned to 16S rRNA sequencing.
What Does It Mean?
Comparing sequences from thousands of bacteria is a complicated process. But to differentiate between vegetarian and omnivorous diets or nourished and malnourished children over the course of three years and then determine the roles of the microbes corresponding to those sequences seems an almost insurmountable task.
When overwhelmed by the vastness of metagenomic sequencing data, many researchers turn to Rob Knight at the University of Colorado, Boulder, a leader in the field who has collaborated on projects as diverse as looking at differences in the guts of Burmese pythons before and after eating or sampling surfaces in public restrooms. Knight performed the data analysis for the Malawian twins at risk for kwashiorkor and is currently working on a similar study of malnutrition in Bangladesh.
16S rRNA marker gene analysis relies on the fact that differences between bacterial genomes are highly correlated with differences between ribosomal RNA sequences. Focusing on this area of the genome allows deeper sequencing for the same cost as shotgun metagenomic sequencing, facilitating taxonomic assignment of even rare organisms present in the sample.
"In general, knowing the phylogenetic affiliation of what species or higher taxa the data fit is very useful for getting a general idea of function," Knight explained. With enough samples, "you can figure out which particular bacteria are co-varying with the clinical or environmental variable. Once you know who is changing, you can start to develop hypotheses about why they're changing."
Currently, computation is slow; there’s a need for improved methods for homology searches and assembly. Annotation databases also need to be expanded to better assign functions to the genes or taxa identified. But even when faced with these shortcomings, Knight and his team can design new algorithms to accomplish their goals, such as the one they created for the Malawi kwashiorkor study to follow how the microbial community in the gut changed in a systematic way between individuals over time.
Groups of bacteria that differed with cardiovascular disease and kwashiorkor were identified, but establishing causality can't be done with metagenomic data alone. "If you can transfer the microbes that you think are responsible to a mouse that you have raised without any microbes, and that mouse shares the phenotype only in the presence of those microbes, then you can be pretty sure that the microbes are involved," Knight said.
Both Hazen’s and Manary’s groups did precisely that. Mice who received microbes from affected individuals developed the diseases as well (1-3). Bacteria acting as pathogens is a familiar story, but these are unique cases. Receipt of the microbes alone was not enough: microbiomes associated with kwashiorkor produced the disease only when the mice were fed Malawian diets and microbiomes associated with cardiovascular disease increased TMAO levels only in the presence of carnitine or lecithin.
"Obviously microbiome analysis takes a long time, but if it was a real time thing, you could almost predict which child was going to end up with kwashiorkor," Trehan said. Right now, sequencing technologies are too expensive to use in the developing world for predicting disease, but Trehan believes that one day we will be able to modify the microbiome with antibiotics or probiotics, incorporating these treatments into periodic therapies given yearly to children in the developing world.
Knight also anticipates that understanding the microbiome will lead to preventive measures in the future. He compares this public health potential to that of vaccines, where the influence of maintaining a healthy microbiome would be almost invisible. Rather than bringing about a cure, many diseases would simply be avoided by careful microbiome maintenance.
Hazen questions the use of antibiotics for modifying the microbiome since microbes will simply develop a resistance and then carry on as before. Manary states simply that probiotics will not work since they are only present in stools for a fleeting period of time. "I think instead, the approach is going to be developing a specific inhibitor against the enzyme activity responsible," Hazen said. And he pointed out that adopting healthier diets will reduce risk naturally.
"We need a lot more data from a lot more people. We need to track people over time. We need to understand how the host behavior and phenotype influences the microbiome, and we also need prospective longitudinal studies so we can figure out the trajectories in the microbiome that lead to disease," Knight said. Only then can some of the speculation progress to testable hypotheses, clinical trials, and therapeutic options.
1. Tang, W. H. W., Z. Wang, B. S. Levison, R. A. Koeth, E. B. Britt, X. Fu, Y. Wu, and S. L. Hazen. 2013. Intestinal microbial metabolism of phosphatidylcholine and cardiovascular risk. N Engl J Med 368(17):1575-1584.
2. Koeth, R. A., Z. Wang, B. S. Levison, J. A. Buffa, E. Org, B. T. Sheehy, E. B. Britt, X. Fu, Y. Wu, L. Li, J. D. Smith, J. A. Didonato, J. Chen, H. Li, G. D. Wu, J. D. Lewis, M. Warrier, J. M. Brown, R. M. Krauss, W. H. Tang, F. D. Bushman, A. J. Lusis, and S. L. Hazen. 2013. Intestinal microbiota metabolism of l-carnitine, a nutrient in red meat, promotes atherosclerosis. Nature medicine advance online publication(April).
3. Smith, M. I., T. Yatsunenko, M. J. Manary, I. Trehan, R. Mkakosya, J. Cheng, A. L. Kau, S. S. Rich, P. Concannon, J. C. Mychaleckyj, J. Liu, E. Houpt, J. V. Li, E. Holmes, J. Nicholson, D. Knights, L. K. Ursell, R. Knight, and J. I. Gordon. 2013. Gut microbiomes of malawian twin pairs discordant for kwashiorkor. Science (New York, N.Y.) 339(6119):548-554.