Systematic approaches to deciphering genes and ecosystems

Abstract

In this thesis I investigate how the individual components of biological systems interact, and how the form of these interactions determines overall system behavior. The interactions I study occur at widely different scales: from whole ecosystems to individual genes. Despite these differences, the approaches I use share similarities: in both realms I apply systematic experimental perturbations, then utilize mathematical and computational tools to identify novel properties of these interactions and to interpret their importance.

In the first study, I examine how 3-way species interactions affect ecosystems dynamics. Ecological models typically assume that interactions occur among pairs of species. While higher-order interactions among greater numbers of species are thought to occur in natural ecosystems, the strength and overall importance of these interactions for ecosystem behavior has been unclear. My study focuses on species interactions mediated by antibiotic toxins, which either inhibit or kill antibiotic sensitive species. Here I develop a quantitative 3-way interaction assay to measure how the inhibition of a sensitive species by an antibiotic producer is affected by the presence of a third “modulator” species. Systematically testing combinations of species and antibiotics, I find that antibiotic degradation by the modulator species frequently attenuates the inhibition of the sensitive species. I then use simulations and mathematical models to show that such 3-way interactions can dramatically alter ecosystem dynamics. Ecosystems of antibiotic producing, sensitive and resistant species are thought to coexist only when they are spatially separated in the environment, but these conclusions are based on models that assume pairwise species interactions. Surprisingly, I find that the 3-way interactions created by the counteraction of antibiotic production and degradation enable coexistence even in well-mixed environments. These findings are robust to choices of parameters and modeling assumptions, and shed light on the role of antibiotic production and degradation in maintaining the diversity of natural microbial communities.

In the second study, I shift to the molecular scale and develop strategies for deciphering multiple protein and mRNA selective pressures that affect gene function. Recent technological advances in “Mutagenomics”, i.e. large-scale mutagenesis and phenotyping, have enabled the systematic mapping of fitness landscapes. Current challenges include the difficulty of applying such methods to essential genes, which control many core biological processes, and also the problem of interpreting high-dimensional fitness landscapes in terms of sensible biochemical properties. Here I present advances on both fronts: first by developing MAGE-seq, a high-throughput method that combines genome engineering with a DNA sequencing-based assay to enable rapid measurement of fitness landscapes anywhere on the Escherichia coli genome. Second, I describe methods of analysis to identify key properties that determine gene fitness through a case study of infA, an essential translation initiation factor (IF1). At the protein level, I find that selection is determined primarily by amino acid properties like hydrophobicity, flexibility, size and charge, and I relate these properties to protein folding and function; at the mRNA level, I show that selection is strongest in the early regions of the gene, where codon preferences are determined by the formation of RNA hairpins containing the start codon and by the avoidance of deleterious motifs such as out-of-frame start codons. Disruption of this optimal RNA structure determines codon preferences in later regions of the gene, suggesting that gene evolution is constrained by both mRNA and protein properties. Together these experimental and analytical methods make it possible to systematically identify and engineer the key properties of biological function genome-wide.