The use of genome-scale metabolic network reconstruction to predict fluxes and equilibrium composition of N-fixing versus C-fixing cells in a diazotrophic cyanobacterium, Trichodesmium erythraeum

Abstract

BackgroundComputational, genome based predictions of organism phenotypes has enhanced the ability to investigate the biological phenomena that help organisms survive and respond to their environments. In this study, we have created the first genome-scale metabolic network reconstruction of the nitrogen fixing cyanobacterium T. erythraeum and used genome-scale modeling approaches to investigate carbon and nitrogen fluxes as well as growth and equilibrium population composition.ResultsWe created a genome-scale reconstruction of T. erythraeum with 971 reactions, 986 metabolites, and 647 unique genes. We then used data from previous studies as well as our own laboratory data to establish a biomass equation and two distinct submodels that correspond to the two cell types formed by T. erythraeum. We then use flux balance analysis and flux variability analysis to generate predictions for how metabolism is distributed to account for the unique productivity of T. erythraeum. Finally, we used in situ data to constrain the model, infer time dependent population compositions and metabolite production using dynamic Flux Balance Analysis. We find that our model predicts equilibrium compositions similar to laboratory measurements, approximately 15.5% diazotrophs for our model versus 10-20% diazotrophs reported in literature. We also found that equilibrium was the most efficient mode of growth and that equilibrium was stoichiometrically mediated. Moreover, the model predicts that nitrogen leakage is an essential condition of optimality for T. erythraeum; cells leak approximately 29.4% total fixed nitrogen when growing at the optimal growth rate, which agrees with values observed in situ. ConclusionThe genome-metabolic network reconstruction allows us to use constraints based modeling approaches to predict growth and optimal cellular composition in T. erythraeum colonies. Our predictions match both in situ and laboratory data, indicating that stoichiometry of metabolic reactions plays a large role in the differentiation and composition of different cell types. In order to realize the full potential of the model, advance modeling techniques which account for interactions between colonies, the environment and surrounding species need to be developed.Electronic supplementary materialThe online version of this article (doi:10.1186/s12918-016-0383-z) contains supplementary material, which is available to authorized users.

MultIscale MultiObjective Systems Analysis (MIMOSA): an advanced metabolic modeling framework for complex systems

“…We developed MIMOSA by integrating an updated version of the genome-scale metabolic model (3) (Table S1 for updated reactions) with nutrient diffusion, light diffusion, cell/cell interaction and cell/environment interactions (see Figure 1) using an agent based modeling framework. We have also implemented the use of multiobjective optimization to account for the dual cellular objective of producing biomass and the metabolite which is transacted between cells (glycogen or β-aspartyl arginine, depending on cell type).…”

“…We have also implemented the use of multiobjective optimization to account for the dual cellular objective of producing biomass and the metabolite which is transacted between cells (glycogen or β-aspartyl arginine, depending on cell type). Constraints were imposed on the model as reported previously (3) with two notable exceptions. First, the ultimate product of nitrogen fixation was changed from ammonium to β-aspartyl arginine, which is the monomer used to create cyanophycin, a nitrogen storage polymer in T. erythraeum and other diazotrophic cyanobacteria (38–40).…”

“…In order to test the predictive accuracy of the model, we predicted growth rate for a variety of different light intensities (Figure 3A) and compared to other published models for T. erythraeum (1, 2) as well as other experimentally measured growth rates (1, 3, 6, 9, 10, 41) exhibiting light saturation at higher light intensities. Ultimately, our model is a metabolic model, so it is important that it can also capture the metabolic changes that occur in response to changes in the environment.…”

“…The simplicity of this technique does come at a cost: typical model formulations are limited to modeling steady-state growth of axenic cultures assuming homogenous environmental conditions; while this works for traditional metabolic engineering efforts on single species, it cannot accurately predict the behavior of consortia. There have been a few attempts to expand the applicability of these models to communities (3, 14, 15), but these models require assumptions that oversimplify the system, such as no diffusional limitations and identical or static growth rates for the different organisms. The current benchmarks for constraints-based metabolic modeling of microbial consortia are OptCom (16) and d-OptCom (17) (OptCom’s dynamic version).…”

“…Unlike other diazotrophs, which either spatially or temporally separate the oxygen sensitive nitrogenase enzyme from the water splitting reaction of photosynthesis (oxygen production), T. erythraeum is unique because it simultaneously carries out nitrogen and carbon fixation during the day in different cells along the same filament (trichome). Therefore, it is the ideal model system for the development of MIMOSA: it has structurally identical cells that operate in two distinct metabolic modes (photoautotrophic and diazotrophic), a published genome scale model (3), transcriptome data, and a plethora of in situ and laboratory data to both train the model and validate predictions. We use this organism as a test-case for the modeling framework and illustrate how it can be used to develop a predictive model that can also be used to investigate cellular physiology by elucidating rules of behavior.…”