Genome Scale Metabolic (GSM) Models contain the known set of all metabolic reactions that exist in an organism. These reactions are represented by a stoichiometric matrix that contains all coefficients and reaction directionalities. Using additional components such as a scaled biomass equation and gene-protein-reaction (GPR) relationships, GSM models can be used to predict phenotypes and propose intervention strategies.

Flux Balance Analysis (FBA) is used to predict possible flux distributions that achieve optimal flux through the objective reaction. FBA assumes the organism is at pseudo-steady state and at the logarithmic growth phase. When the objective reaction is biomass production, FBA can be used to predict growth rate at different conditions.

^Source

The figure above shows FBA as a pipeline. Biomass production can be achieved through various reactions (Product A, B, and C) and knocking out a reaction (closing a valve) forces the organism to obtain biomass through a specified reaction.

Electron Donor Modeling

Initially, a source of ferredoxin was needed. In order to determine the best possible source, we identified two reactions, knocked those out, and replaced them with equivalent reactions that used ferredoxin as an electron carrier. Two options were found: a reaction that converted pyruvate to acetyl- CoA and one that converted 2-oxoglutarate to succinyl- CoA. These were both knocked out and replaced with a ferredoxin equivalent. Replacing the original pyruvate to acetyl-CoA reactions allowed for ferredoxin production, but there was less biomass production than wild type. When the wild type 2-oxoglutarate to succinyl-CoA reaction was knocked out and replaced with the ferredoxin equivalent reaction, wild type biomass was retained but there was no ferredoxin production. This indicated that there are likely several competing pathways through the TCA cycle and knocking out all of these was not a viable option for cell growth. Thus, the pyruvate to acetyl-CoA reaction was replaced with the ferredoxin equivalent to achieve ferredoxin production.

These discoveries lead to two key wet lab approaches:

1. The pyruvate to acetyl-CoA reaction was knocked out by obtaining a knockout strain for aceE. aceE codes for a subunit of pyruvate dehydrogenase (PDH), a key enzyme in the pyruvate to acetyl-CoA reaction. By knocking it out, we were able to knock out the reaction. This ΔaceE strain was then used in conjunction with our electron donor plasmids to further increase reduced electron donor production
2. We were able to replace the PDH reaction by overexpressing pfo. pfo codes for pyruvate ferredoxin oxioreductase and is an enzyme that reduces electron donors. We modeled its overexpression, which predicts that there should be an increase in the production of reduced electron donors. In the lab, we constructed new electron donor plasmids that had pfo overexpressed along with the electron donor genes.

ATP Modeling

ATP could be further maximized by ensuring higher minimum ATP flux levels. In order to do this, reactions that had very high ATP minimums could be knocked out. In doing this, there could not be negative fluxes when minimizing the flux values (a negative value means the reaction is in the reverse direction) so reversible reactions were split. The binary reactions were then added to the model to ensure that all flux ranges were positive. Flux was minimized over the subset of metabolic reactions that produced ATP and possible knockouts were identified by identifying the reactions that had minimums significantly higher than wild type.

Unlike electron donor modeling, we did not have time to test these modeling predictions in the lab.

Proof of Concept Modeling

Ferredoxin yield can be estimated by the ratio of biotin to ferredoxin flux. When running FBA, the biotin synthesis reaction carried flux, indicating that under the specified conditions, reduced ferredoxin was being produced. Biotin was not found to be consumed otherwise, thus it was a strong candidate for indicating the presence of reduced ferredoxin.

After designing our Super Cells and improving them via modeling, we conducted experiments to get quantitative results.