Abstract
There is an urgent need to discover new therapeutic targets to treat infections. The rise in antimicrobial resistance exacerbates this need. If left unmitigated, antimicrobial resistance is estimated to claim over 10 million lives worldwide by 2050. However, despite this rise in resistance, very few new antibiotics have been brought to market. In fact, over the past thirty years, there has been a lack of discovery of new antibiotic classes as a whole.
There are several reasons for this lack of antibiotic discovery, many of which are scientific bottlenecks. For instance, current target identification platforms require extensive screening and downstream follow-up experiments that are very time-consuming. Frequently, all that work leads to low success rates because they identify targets of unknown function, requiring even more experiments, or they identify targets that actually promote resistance. Additionally, current platforms use whole bacterial population approaches and fail to capture heterogeneous subpopulations with unique susceptibilities.
Metabolic network modeling is emerging as a powerful tool for antimicrobial target discovery to overcome these limitations. Genome-scale metabolic network reconstructions (or GENREs) serve as knowledge-bases for everything we know to-date about the metabolism of an organism. These reconstructions are also tools that allow us to study the genotype-to-phenotype relationship within a cell. Ultimately, using these models, we can probe the capability of an organism in different environmental conditions. Importantly, we can use these models to identify essential processes for different organism objectives, such as growth or the production of metabolites of interest. By identifying these essential processes, we can suggest potential therapeutic targets.
In this work, I demonstrate that antimicrobial target discovery with metabolic network modeling overcomes challenges associated with current target identification platforms. Specifically, I show that metabolic network models (1) enable high-throughput target discovery, (2) delineate targets of known function, (3) determine targets that may mitigate resistance, and (4) identify targets for heterogeneous subpopulations. To do this, I applied a metabolic network model of the Gram-negative, multi-antimicrobial resistant pathogen Pseudomonas aeruginosa to antimicrobial target discovery in three different applications. In the first, I use the model to probe the interrelationship between growth and the synthesis of metabolites important for infection known as virulence factors (Chapter 2). In the second application, I reconcile conflicting high-throughput in vitro gene essentiality datasets and demonstrate the utility of contextualizing and interpreting these datasets with the model (Chapter 3). Finally, in the third, I generate condition-specific metabolic network models to identify targets for a specific subpopulation of bacteria, called persister cells, that is known to tolerate antimicrobial treatment (Chapter 4).
Together, this research demonstrates the unique ability of metabolic network modeling to facilitate the drug discovery pipeline and identify antibacterial targets that would be impossible to delineate without the use of computational models.
