Personalizing nutrition based on the gut microbiome

“You are what you eat” is a popular phrase to describe how diet influences health, but not everyone responds in the same way to the same foods. “If I feed ten people a banana, different chemicals come out the other side,” said Sean Gibbons, a microbiome scientist at the Institute for Systems Biology.

Researchers in Sean Gibbons laboratory wear white lab coats and pose for a photo between the benches in his laboratory.

Sean Gibbons and his laboratory at the Institute for Systems Biology study the gut microbiome using experimental and computational models.

Credit: Institute for Systems Biology (ISB)

These chemicals are short-chain fatty acids (SCFAs). “The gut microbiome is a short chain fatty acid producing machine,” said Gibbons. “It takes in fibers and ferments them into SCFAs.” SCFAs have important roles in regulating metabolism and modulating the immune response, so understanding their production can help link diet to health.

To predict SCFA production on the individual level, Gibbons and his team recently designed and validated a microbial community-scale metabolic model (1). This model integrated existing genome-scale metabolic models of individual bacterial taxa with an individual’s microbiome and dietary inputs to predict SCFA production on a personalized level. “When we make a prediction, we actually know how it's made,” said Gibbons. “We know which bug and which biochemical pathway are actually responsible for the production of a specific thing.” This contrasts with existing methods that use black-box machine learning approaches where it’s difficult to understand how or why the algorithm made that prediction.

SCFAs are hard to measure in the body as they are quickly absorbed by the gut. Therefore, to validate the model, Gibbon’s team turned to experimental systems outside of the body. The researchers found that their metabolic model accurately predicted the production of one of the most common SCFAs, butyrate, using both in vitro co-cultures of gut microbes as well as stool samples. Using existing datasets, they found that their metabolic model identified expected correlations between SCFA production and blood-based biomarkers, including those associated with inflammation, insulin resistance, and blood pressure.

If I feed ten people a banana, different chemicals come out the other side.
- Sean Gibbons, Institute for Systems Biology

Emanuel Canfora, a human physiologist studying the relationship between gut microbiota and metabolic diseases at Maastricht University Medical Center+, noted how difficult it can be to see correlations between SCFAs and metabolic outcomes. “What was really impressive to see [in this study] is that [predicted SCFAs are] more correlated or associated with metabolic outcomes than, for example, if you just look at butyrate levels in the plasma,” said Canfora, who was not involved in the study.

To address the implications for personalized nutrition, the researchers tested whether they could use their model to design personalized interventions — either in diet, prebiotics, and/or probiotics — to optimize SCFA production. When individuals switched from a low-fiber to a high-fiber diet, the researchers found that not everyone responded to the diet shifts in the same way. Some produced more butyrate, while others produced less or had no change. In terms of probiotic or prebiotic supplementation, they found that no single intervention was beneficial across the board. “That gives very good leads for the future for individualization of the intervention,” said Canfora.

Gibbons added, “Looking forward, we want to try to implement these types of models in clinical trials where we prove that a precision intervention that is designed using the model is better than either a standard of care that a nutritionist might suggest or a random intervention.”