Metabolic modeling and experimental studies on carbon and energy metabolism of Mycobacterium tuberculosis

Tuberculosis is an ancient human disease affecting millions of people every year. The common causative agent, Mycobacterium tuberculosis, is a peculiar pathogen that can persist in the host tissues for decades. Many of the key attributes of M. tuberculosis can be traced to its metabolism. Therefore, novel insights into the metabolism of this bacterium could lead to the identification of new drug targets as well as better methods to understand the mechanism of action of anti-tuberculosis drugs. Analysis of the metabolic network of this bacterium using a genome-scale model offers an unconventional approach to understanding utilization of metabolic pathways of this pathogen during growth and non-replicating persistence. Metabolic adaptation of a pathogen is central to its life style in vivo. To understand the capacity and limitations of M. tuberculosis, we developed an updated and thermodynamically curated genome scale model – iOS783. This model has been verified with respect to the observed phenotypes for targeted gene deletion studies reported previously. Evaluation of iOS783 with thermodynamics based flux balance analysis (TFBA) improved overall reliability of the model results and led to identification of possible bottleneck reactions. In addition, several new hypotheses were generated with iOS783, involving the role of isocitrate lyase enzymes, propionate toxicity, and energy metabolism of this pathogen. Analysis of iOS783 for conditional gene essentiality emphasizes the importance of the carbon source available to the pathogen. Several genes show a conditional impact on biomass production rate of the model, depending on the availability of carbohydrates or fatty acids as the carbon substrate. Conditional essentiality is even more pronounced when ATP replenishment capacity is used as a measure of a pathogen’s ability to survive under stress in a non-growing state. As expected, ATP synthase plays a major role in estimations of ATP replenishment capacity based on iOS783. Curiously, computational analysis of the metabolic network with ATP synthase activity blocked indicates that the damaged network can use substrate-level phosphorylation reactions to sustain ATP production to fulfill non-growth-associated ATP requirements when glucose is available. In order to establish the dependence of M. tuberculosis on ATP synthase, we attempted to construct a deletion of the chromosomal atp operon. Our results show that the native copy of these genes can be deleted only when a second set of genes are introduced elsewhere in the genome, indicating essentiality of the ATP synthase for growth. Although our results indicate that ATP synthase is essential for growth, we observed carbon source specific efficacy of an ATP synthase inhibitor, bedaquiline, on M. tuberculosis. These experimental results are consistent with ATP turnover estimations based on modeling results. Furthermore, analysis of iOS783 suggests a novel role in energy metabolism for recently described glycine dehydrogenase (GDH) on energy metabolism. Deletion of ald gene responsible for GDH activity reduces the ATP replenishment capacity of iOS783, especially during metabolism of fatty acid derivative carbon sources. Consistent with these modeling results, wetlab experiments designed to test this hypothesis show that bedaquiline has a stronger bactericidal effect on an M. tuberculosis ∆ald mutant, compared to the wild-type parental strain, when cells are cultured in medium containing acetate as the carbon source. Metabolic modelling is already well established as an invaluable tool for engineering of strains with desired characteristics. For pathogenic species, however, the application of this approach has so far been quite limited. The work presented in this thesis exemplifies an interdisciplinary approach combining computational modeling and experimental studies of M. tuberculosis metabolism. First, iOS783 was utilized to interpret existing experimental results and to generate novel hypotheses, thereby providing the impetus for new wetlab experiments. Conversely, results from iOS783-inspired wetlab experiments were used to further refine and expand the iOS783 model. This cyclic exchange between computational modeling and wetlab experimentation illustrates how a close interdisciplinary partnership can create a positive feedback loop that enriches both disciplines.