Hydrogen peroxide (H2O2) is a toxic molecule utilized by the immune system in the first line of defense against infections. After being engulfed by phagocytic cells of the innate immune response, pathogens are exposed to reactive oxygen species (ROS) including superoxide (O2−•) and H2O2. O2−• cannot cross cell membranes, but H2O2 freely diffuses into the bacterium, where it can damage sensitive biomolecules, react with ferrous iron to form highly reactive hydroxyl radicals (•OH), or be detoxified by specialized enzymes. Accordingly, H2O2 detoxification is an important aspect of bacterial virulence, and deletion of one or more detoxification systems affects the ability of a variety of pathogens to establish or sustain an infection. While the importance of these H2O2 detoxification systems is known, a quantitative understanding of how they interact under different levels of oxidative stress is lacking. For instance, Escherichia coli utilizes an alkyl hydroperoxidase (AHP) and two separate catalases (HPI and HPII) to clear H2O2. AHP requires one NADH molecule per reaction cycle, affecting the maximum rate of clearance that can be achieved, whereas H2O2 is the only substrate required by HPI and HPII. The simultaneous activity of multiple detoxification systems and broad reactivity of H2O2 necessitates the use of quantitative kinetic modeling to understand how H2O2 distributes throughout its reaction network.
Thus, we have constructed an H2O2 detoxification network for E. coli consisting of 60 metabolites and enzymes, as well as 75 rate equations describing spontaneous and enzymatic reactions, transcriptional regulation, and degradation of key enzymes. Structural and parametric uncertainty with regard to enzyme degradation and the existence of an H2O2 gradient across the cell membrane led us to consider ten distinct structures while building the model. Using a strategic method of iteratively gathering data, optimizing uncertain parameters, and assessing comparative model performance, we found that only one model structure was able to consistently predict how H2O2 distributes amongst the different pathways. This observation suggested that bimolecular degradation of detoxifying enzymes is an important aspect of the H2O2 clearance network, and that under our conditions diffusion of H2O2 across the membrane was not limiting. Using a Monte Carlo method to explore the viable parameter space, we identified an ensemble of parameter sets that could all fit the data, and used this model ensemble to gain a better understanding of the importance of protein synthesis and reducing equivalents under oxidative stress. We have developed a platform to quantitatively explore H2O2 stress in E. coli which enhances understanding of the impact of mutations and environmental alterations on detoxification, and our method provides a template for translation to other organisms.