Understanding how microbes assemble into communities is a fundamental open question in biology, with applications to human health, environmental sustainability, and metabolic engineering. Although it is known that the competition and exchange of nutrients (i.e., metabolic interactions) shape microbial community structure and dynamics, the ability to reliably predict the metabolic interactions and their effect on microbial communities is still being studied. This dissertation investigates how metabolism and environment shape microbial communities through the use of mathematical models, based on linear programming (LP) and mixed integer linear programming (MILP) methods.
The first system I studied is a synthetic microbial consortium composed of two species, Cellulomonas fimi and Yarrowia lipolytica, hypothesized to be able to jointly transform cellulose into biofuel precursors. I combined experimental data and flux balance analysis (FBA) to test our capacity to predict metabolic interactions between the two organisms, and explored a proof-of-concept method to monitor the growth dynamics of this coculture. I next explored the possibility of generalizing the design of synthetic communities through the implementation of a computational method that can design division of labor strategies. The algorithm finds consortia of engineered bacterial strains that can survive by exchanging with each other specific nutrients. By distributing functions, microbial consortia can perform tasks that are impossible for individual species to accomplish alone. In addition to highlighting the trade-off between metabolic self-reliance and mutualistic exchange, this approach suggests how division of labor may arise in Escherichia coli monocultures.
While mechanistic models are helpful for studying metabolism in microbes and microbial communities, it is interesting to ask whether increasingly cheaper high-throughput phenotypic data, can help achieve similar goals. To address this question, I developed a computational approach to investigate the relationship between growth profiles and microbial species, based on the identification of growth conditions that can best represent the whole dataset. This approach can help engineer microbial communities by identifying microbes that are more likely to engage in cross-feeding, rather than competition, based on their phenotypic profiles.
In general, this dissertation demonstrates how different types of metabolic modeling approaches, both mechanism-based and data-driven, can be used to predict metabolic interactions between members of microbial consortia, and to help engineer novel synthetic communities.
| Identifer | oai:union.ndltd.org:bu.edu/oai:open.bu.edu:2144/41030 |
| Date | 19 May 2020 |
| Creators | Thommes, Meghan |
| Contributors | Segre, Daniel |
| Source Sets | Boston University |
| Language | en_US |
| Detected Language | English |
| Type | Thesis/Dissertation |
| Rights | Attribution 4.0 International, http://creativecommons.org/licenses/by/4.0/ |
Page generated in 0.0018 seconds