Microbes, mathematics, and models

Classifying Interactions in a Synthetic Bacterial Community Is Hindered by Inhibitory Growth Medium

Predicting the fate of a microbial community and its member species relies on understanding the nature of their interactions. However, designing simple assays that distinguish between interaction types can be challenging. Here, we performed spent medium assays based on the predictions of a mathematical model to decipher the interactions among four bacterial species: Agrobacterium tumefaciens, Comamonas testosteroni, Microbacterium saperdae, and Ochrobactrum anthropi. While most experimental results matched model predictions, the behavior of C. testosteroni did not: its lag phase was reduced in the pure spent media of A. tumefaciens and M. saperdae but prolonged again when we replenished our growth medium. Further experiments showed that the growth medium actually delayed the growth of C. testosteroni, leading us to suspect that A. tumefaciens and M. saperdae could alleviate this inhibitory effect. There was, however, no evidence supporting such "cross-detoxification," and instead, we identified metabolites secreted by A. tumefaciens and M. saperdae that were then consumed or "cross-fed" by C. testosteroni, shortening its lag phase. Our results highlight that even simple, defined growth media can have inhibitory effects on some species and that such negative effects need to be included in our models. Based on this, we present new guidelines to correctly distinguish between different interaction types such as cross-detoxification and cross-feeding. IMPORTANCE Communities of microbes colonize virtually every place on earth. Ultimately, we strive to predict and control how these communities behave, for example, if they reside in our guts and make us sick. But precise control is impossible unless we can identify exactly how their member species interact with one another. To find a systematic way to measure interactions, we started very simply with a small community of four bacterial species and carefully designed experiments based on a mathematical model. This first attempt accurately mapped out interactions for all species except one. By digging deeper, we understood that our method failed for that species as it was suffering in the growth medium that we chose. A revised model that considered that growth media can be harmful could then make more accurate predictions. What we have learned with these four species can now be applied to decipher interactions in larger communities.

Proteomic traits vary across taxa in a coastal Antarctic phytoplankton bloom

Production and use of proteins is under strong selection in microbes, but it is unclear how proteome-level traits relate to ecological strategies. We identified and quantified proteomic traits of eukaryotic microbes and bacteria through an Antarctic phytoplankton bloom using in situ metaproteomics. Different taxa, rather than different environmental conditions, formed distinct clusters based on their ribosomal and photosynthetic proteomic proportions, and we propose that these characteristics relate to ecological differences. We defined and used a proteomic proxy for regulatory cost, which showed that SAR11 had the lowest regulatory cost of any taxa we observed at our summertime Southern Ocean study site. Haptophytes had lower regulatory cost than diatoms, which may underpin haptophyte-to-diatom bloom progression in the Ross Sea. We were able to make these proteomic trait inferences by assessing various sources of bias in metaproteomics, providing practical recommendations for researchers in the field. We have quantified several proteomic traits (ribosomal and photosynthetic proteomic proportions, regulatory cost) in eukaryotic and bacterial taxa, which can then be incorporated into trait-based models of microbial communities that reflect resource allocation strategies.

Exploring modularity in biological networks

Network theoretical approaches have shaped our understanding of many different kinds of biological modularity. This essay makes the case that to capture these contributions, it is useful to think about the role of network models in exploratory research. The overall point is that it is possible to provide a systematic analysis of the exploratory functions of network models in bioscientific research. Using two examples from molecular and developmental biology, I argue that often the same modelling approach can perform one or more exploratory functions, such as introducing new directions of research, offering a complementary set of concepts, methods and algorithms for individuating important features of natural phenomena, generating proofs of principle demonstrations and potential explanations for phenomena of interest and enlarging the scope of certain research agendas. This article is part of the theme issue 'Unifying the essential concepts of biological networks: biological insights and philosophical foundations'.