Are the dynamics of our microbial communities unique to us or does everyone's microbiota follow the same rules? The emerging insights into this question could be of relevance to health and disease. See Letter p.259
The composition of a body part's microbial community can differ substantially from one person to the next1,2,3,4,5. This is due to both host pressures and the dynamic behaviour of the microbes themselves. Understanding whether these interactions are consistent across hosts or whether each individual's microbiota follows its own rules has big implications. If the dynamics of an organ's microbial community are universal, we can use them to predict effective interventions for modulating the microbiota. If, however, microbial dynamics are host-specific, interventions must be designed separately for each person. Bashan et al.6 address this issue using a new approach and report their intriguing observations on page 259.
To find out whether community dynamics are universal, ideally we should study long and densely sampled time series from many individuals with different traits and backgrounds. Models of microbial communities should then be fitted to the varying proportions of microbial species, which may become challenging when going beyond the most dominant groups of species. Such large temporal data sets are currently gravely lacking.
Bashan and colleagues devised an indirect method to address the question of universality. They measured two independent aspects of community similarity: overlap, which compares species assemblies by quantifying the proportion of shared species; and dissimilarity, which assesses the difference in abundance profiles of the shared species between individuals. The dissimilarity is then plotted against the overlap for all sample pairs to create a dissimilarity–overlap curve (DOC). If microbiota dynamics are truly universal (host-independent), then having the same species present should lead to the same relative proportion of those species, because they would dynamically influence each other in the same way. Consequently, a larger proportion of shared species should increase the community similarity and result in the tell-tale negative slope of the DOC (Fig. 1).
The authors tested their method by simulating microbial communities computationally using what is known as the generalized Lotka–Volterra model7, to generate communities with the same and with different dynamics as positive and negative controls. In addition, they showed that randomizing data by shuffling microbial species across samples also removes the negative slope. These simulations confirm that the DOC detects universal dynamics and flattens in the absence of such dynamics. The curve even identifies strongly interacting species.
Most notably, the team detected negative slopes for the oral and gut communities in several human-microbiome data sets, including those of the Human Microbiome Project3 and two human-gut time series8,9. However, the skin microbiota displayed weakly negative or flat DOCs in some cases, suggesting that the microbial dynamics in the skin are host-specific at certain sites. Another interesting finding was that the DOC for the gut microbiota of people recurrently infected with the bacterial pathogen Clostridium difficile10 is flat, but gains a negative slope after faecal transplantation from people who have not been infected.
If the assumptions hold, the consistent negative slopes observed for the healthy cohorts and for people treated after infection with C. difficile point to universal gut microbial dynamics. This is good news for all modelling efforts aiming to predict the behaviour of the gut microbiota during interventions or in disease. It means that when parameters such as growth rates and interactions are determined for the gut microbial community of one healthy human, they are also valid for those of other individuals. Thus, the knowledge of such parameters can be combined across different studies and could, in the long term, allow a detailed, common microbial community model to be developed.
The DOC method has all the hallmarks of a powerful analytical tool. It is easy to implement, addresses a crucial question and may inspire applications beyond its intended use.
But, like all analyses, it makes a couple of assumptions — that the microbiota are in a steady state, and that having the same steady state implies that microbiota are governed by the same dynamics. The second assumption is the more risky: microbiota may end up in similar steady states not because of their intrinsic dynamics, but because of a strong environmental pressure that selects for a particular set of species. The authors rule out obvious host parameters such as diet, weight, age, race and transit time through the gut (measured by stool consistency) that may shape gut microbial communities. However, they do not account for all factors that may conceivably influence the gut microbiota2, and so cannot provide an entirely conclusive answer regarding the universality of the gut's microbial community dynamics.
The value of this work lies primarily in the importance of the question asked, the originality of the approach and the fact that it could spur a whole range of microbiome research. We expect it to spark fruitful discussions and lead to fresh ideas for analyses and experiments. For instance, it might be plausible to set up an artificial community under controlled conditions within a chemostat and then develop and define a model that describes its dynamics reasonably well. One could then compare the steady states reached by different subsets of the community to directly test the second assumption. If universal dynamics are confirmed, modelling efforts have a better chance of leading to more-effective clinical interventions. Bashan and colleagues' paper gives a glimpse of the deeper insights to be gained once we overcome the hurdles of controlled, high-throughput microbial community cultivation and manipulation.
About this article
The ISME Journal (2018)