The gut microbiome is implicated in a growing array of diseases, spanning asthma, allergies, obesity and autoimmunity. The need for interventions that return aberrant dysbiotic communities to symbiotic partners is clear. Yet, despite known sensitivities to diet, a conceptual framework through which to design rational interventions remains elusive.
We adopted an ecological perspective in building a conceptual model to reason about how diet impacts the microbiome. Microbes must satisfy their nutritional requirements for life. The most readily limiting resources are carbon, as an energy source, and nitrogen, a key element in all amino acids.
Microbes employ a diversity of strategies to satisfy these nutritional requirements. Some specialise on metabolising diet-derived substrates such as fibre. Others can grow on mucin, a host secretion, alone. We further envisage hybrid strategies between these two extremes, where carbon and nitrogen substrates either both originate from the diet or the host, or one from each.
We evaluated this ecological conceptualisation's capacity to explain observed microbiome responses to diet through the construction of an agent-based model. Host mucin secretion and dietary nutrient absorption dynamics have been estimated from literature, the remaining dietary nutrients being supplied to a one dimensional simulated gut environment that houses bacteria. Each bacteria cell agent internalises nutrients from its local environment, targeting a 5.2:1 carbon:nitrogen ratio. Nutrient-rich cells exhibit a higher probability of division, where nutrient-poor cells likely perish.
Our modelling work was closely aligned with an in vivo murine dietary study. Herein, 112 mice were administered diets systematically varying in protein:carbohydrate:fat distribution, and energy density by diluting these mouse-accessible nutrients with cellulose, which neither mouse nor its microbiome can metabolise. Their actual intake, and hence the energy they derive from each macronutrient, was recorded. Using generalised additive models we interpolated from these 112 samples across a wide dietary intake space. As such, we could ascertain how each microbe responds to a given diet. We observed two broad patterns of response: microbes that thrive under energy-rich or energy-poor diets. The dietary intakes of each of these 112 mice were simulated in our model, which broadly recapitulated these in vivo observations. This suggests that the balance of host- versus diet-derived substrates is a critical factor in shaping the microbiome.
Our model permits a comprehensive exploration of how multiple dietary factors, such as prebiotic administration, fasting regimes, macronutrient distribution and energy density, interact in impacting the microbiome. The model predicts that interventions such as fasting and prebiotic (fibre) administration do not have uniform effects, which, rather, depend on other dietary parameters also. For instance, prebiotic administration in low-protein diets can have a limited effect, as under these diets microbes are nitrogen-limited, an the carbon prebiotics supply is of little consequence. These results underscore the importance of taking an integrative approach to studying diet-host-microbiome interactions, and tools such as these offer a promising path to designing personalised dietary interventions to rationally manipulate a person's microbiome.