Metabolic Network Reconstruction and Modeling of Microbial Communities.

Abstract

In nature, most microorganisms live in synergistic communities performing important biological functions and ecological roles, such as polysaccharide utilization in the gastrointestinal tract of mammalian hosts. In the past decade, metagenomics has advanced rapidly, providing detailed information of population structures and genetic sequences for microbial communities. In this dissertation, my goal is to develop community-wide metabolic network models to understand the cellular metabolic properties and inter-species relationships in microbial communities.

Metabolic network reconstruction at the (meta)genome scale requires tremendous efforts and time. To address this challenge, we started by developing a bioinformatic pipeline for automated reconstruction of high-quality genome-scale metabolic networks using annotated genomes. It was tested with model bacterium Escherichia coli. The results agreed well with a benchmark network manually curated for over a decade. Furthermore, we applied the pipeline to twelve strains of the most abundant cyanobacterium on earth, Prochlorococcus marinus, and defined pan and core metabolic networks of the species, demonstrating the utility of the tool.

Next, we extended our bioinformatic pipeline to community-wide metabolic network reconstruction and investigated two types of microbial communities. First, we studied the metagenomes of acid mine drainage biofilms, which cause water pollution in many mining areas. Both individual metabolic networks and community-wide metabolic networks were reconstructed to study the metabolism and inter-species interactions related to biofilm formation. Several essential interactions were predicted. For example, Leptospirillun Gp III was predicted to fix nitrogen for the whole community, which was supported by experimental data.

Second, we examined two synthetic gut microbiomes to explore their metabolic capabilities and microbe-microbe-host interactions. For each system, we reconstructed community-wide metabolic networks considering all the species using annotated genomes and transcriptomes through a three-step curation process. With these metabolic networks, we could explain mechanistically metabolic phenotypes and predict inter-species interactions. For instance, for a ten-species microbiome, a number of molecules, including urea, citrate and agmatine, were revealed to be cross-fed.

This dissertation demonstrates that metagenome-scale metabolic network reconstruction and analysis is a promising tool for studying intracellular metabolism and inter-species interactions of microbial communities, which can advance fundamental understanding and provide valuable hypothesis for experimental testing.