Inferring metabolic regulation from high-throughput data

Abstract

In order to respond appropriately to a constantly changing nutritional environment, cells must sense nutritional cues and respond by modulating their metabolism. While decades of research have elucidated the biochemical reactions that underlie metabolism, many questions about its regulation remain: for instance, which metabolic cues are sensed, and how transcription of metabolic genes is subsequently regulated. The development of high-throughput technologies to allow the quantitation of hundreds of metabolites and thousands of gene products in parallel has great promise for interrogating these questions; however, we still lack methods for analyzing and interpreting the resulting data. This work presents three case studies, using the model eukaryote <italic>Saccharomyces cerevisiae</italic>, of how such high-throughput data can be analyzed to inform our understanding of metabolic regulation.

In the first study, we show that transcript and metabolite concentrations are coordinated in response to nutrient deprivation, and go on to use this coordination to predict interactions between genes and metabolites using machine learning. Second, using linear models, we identify intracellular metabolites that potentially limit cellular growth, as well as the transcriptional programs associated with specific nutrient limitations. We then expand these models to include not only slow growth but also the exit from the cell cycle, and show evidence both for distinct quiescent states as well as a limited quiescence program of gene expression. Finally, we use a large compendium of microarray data to identify conditions under which apparently redundant isozymes may actually be required, and find that a set of sixteen isozymes are associated with the diauxic shift, or the transition from growth on glucose to growth on other carbon sources. We then go on to show that while knockout strains of two "minor" isozymes, the aconitase <italic>ACO2</italic> and pyruvate kinase <italic>PYK2</italic>, have subtle or no growth defects under standard laboratory conditions, they in fact have large differences in fitness when grown on specific alternative carbon sources. These studies present data-driven approaches for discovery of metabolic regulation in eukaryotes, and provide insights into the question of how cellular metabolism can be geared to a changing environment.