Expanding Computational Metabolic Modeling Methods for Novel Metabolic Engineering Applications

Public Deposited

Computational models greatly benefit metabolic engineering efforts by helping to elucidate experimental observations and predict engineering targets for improved cellular performance. Additionally, supplementing experimental efforts with computational modeling can reduce the loss of time and resources in the lab by narrowing down testing conditions. In optimal cases, computational models can be continuously improved as more experimental feedback is applied, lending to the success of iterative experimental and computational testing cycles. In this dissertation, existing metabolic modeling paradigms are expanded for use on previously untested systems. First, constraint-based modeling methods are used to predict essential gene knockouts leading to metabolically active, non-growth states in Escherichia coli cells. Thirty of our predicted candidates were screened in the lab and predicted metabolite auxotrophies were confirmed. The goal of this effort is to identify ways to turn off growth in cells without shutting down metabolic activity (i.e. carbon uptake). The initial modeling work described here provides a foundation for uncovering the governing objectives of cells during non-growth conditions. These factors are currently unknown as most constraint-based modeling methods have been developed exclusively to predict optimal growth conditions where maximizing flux toward biomass production is always assumed. In a non-growing cell, this assumption no longer holds and the best identifier of cell health to optimize under these conditions is unclear. However, being able to accurately predict the flux distribution of non-growth metabolism is an essential step toward enabling the development on non-growth, high-carbon yield biosynthetic processes where cells will no longer divert fed substrates toward growth. Second, we look toward expanding the application of kinetic models of metabolism to predict engineering targets for increased product formation. We specifically focus on improving the limitations of the computationally intensive kinetic ensemble modeling (EM) framework. EM is a Monte Carlo-based modeling method used to sample many, possible kinetic parameter sets of metabolism from a previously defined reference state and then screening them against additional phenotypic datasets. In its original form, the framework is prohibitively slow when applied to large metabolic networks and often results in non-stable solution sets. To alleviate these challenges, we implemented three acceleration strategies, each providing increased computational efficiency. Furthermore, by screening for locally stable parameter sets, we greatly reduce the sample space and generate more biologically representative solutions. Lastly, we applied our accelerated EM framework to develop a novel kinetic representation of Clostridium autoethanogenum which accurately predicts intracellular metabolite concentrations and engineering targets for increased ethanol production. Specifically, our average ensemble predictions fall within demonstrated experimental error ranges for sixty percent of observed metabolite species. Additionally, we were able to demonstrate the experimental observation of a limiting acetyl-CoA pool with increasing biomass concentration and confirm the production of ethanol from acetate to increase adenosine triphosphate (ATP) generation. Finally, through sensitivity analysis, we have identified several enzyme targets for improving ethanol production. Encouragingly, we show that two of the enzymes we have identified as potential down-regulation targets, phosphate acetyltransferase (PTA) and carbon monoxide dehydrogenase/acetyl-CoA synthase (CODH_ACS), have previously shown increased ethanol production when knocked out in similar clostridia strains. Ultimately by demonstrating expanded applications for existing modeling methods, this dissertation highlights the expansive opportunities to improve metabolic engineering outcomes through creative computational design. These results will improve efforts to harness and optimize non-growth metabolism as well as increase access to kinetic exploration of metabolic pathways

Last modified
  • 02/15/2019
Date created
Resource type
Rights statement