Computational Model Improvement: 2+ Iterations of the Engineering Success Cycle
Here we present our engineering success whose main objective was to develop a genome-scale metabolic model of Bacillus subtilis able to support simulations and optimizations using PET and PE - the major constituents of microplastics - as the carbon sources to grow.
Several genome-scale metabolic models of B. subtilis already exist and could be used as the origin of our project. However, none of them included the PET or PE degradation pathways, so these pathways had to be inserted as a part of our project regardless of the chosen model. Additionally, the majority of the available models were not in the ideal SBML format, hampering the application of the required modifications and further simulations. Some of the candidate models to be converted into SMBL were too time intensive for the span of the project and were discarded. Therefore, a small set of possible models to be converted to SBML were then analyzed. We finally opted for the iBsu1103 model , which was adapted to be the basis of our work, despite being one of the oldest models available, it was manually annotated and covers the majority of the metabolism, only differing slightly in size and complexity from more recent models.
Our main goal was to incorporate all of the reactions found in the PET and PE degradation pathways into our model and use it to simulate the consumption of these polymers as the only sources of carbon for our modified B. subtilis strain. Thus, supporting the simulations and optimizations to find an optimum genotype which would allow our GMO to survive and possibly thrive in PET and/or PE rich environments. We also intend, in later stages of the project, to use the optimized model for the production of innocuous (e.g. ethanol) or possibly high/added value compounds (e.g. surfactin) from these carbon sources. Optimization tools can be used to identify the best suited candidates for metabolic engineering (i.e upregulation, downregulation and knockout) by maximizing ethanol production whilst maintaining biomass growth in an environment containing PET or PE as a carbon source. This project must be developed in tandem with the WetLab group in hopes of aiding in the metabolic target discovery process.
Nevertheless, to achieve this stage, a functional model integrating PET and PE degradation pathways is required.
Once the most suitable model was chosen (iBsu1103) as a starting point, new metabolites and reactions specific to our project had to be added. After understanding the degradation pathway of PET, we realized that we could link the metabolite Ethylene Glycol (EG) to the cell's metabolism as an alternative carbon source. For this reason, EG and the respective reactions that link it to the cell's endogenous metabolism were first added.
After these inputs, we used the software Optflux (version 3.4.0) to run simulations with the cell using EG as the only carbon source. Unfortunately, the biomass growth rate obtained during these simulations was null. Since biomass growth rate value determines the cell's efficiency and likelihood of survival, the results indicate that either cell growth is not feasible using EG as the sole carbon source or there is an error in the metabolic model.
A drain is a hypothetical reaction that represents the consumption or production of metabolites by the system. This can be done by transforming a non-existent hypothetical metabolite (cpd99999_b) into an external metabolite (cpd99999_e). By adding this reaction to the SBML model as presented in Figure 1, we were able to run simulations and get the result seen in Figure 2.
As depicted in Figure 2, the biomass growth rate was 140.71 h-1 in our simulations which possibly validates cell growth when using EG as the sole carbon source.
We decided to first add the entire PET degradation pathway and lastly the PE degradation pathway. Once again, the biomass growth rate was null when using PET as the only carbon source. All reactions until EG enters the cell are extracellular, as well as every metabolite involved in those reactions. Therefore likely EG, it was also necessary to add drains for all those metabolites, representing their presence on the surrounding medium.
After adding these reactions, we were able to simulate cell growth when using PET as the only carbon source (Figure 3). In this simulation, contrary to the one with EG, PET intake was limited to better represent real life conditions, which is why there is a smaller growth rate (0.90481 h-1).
Since, alongside PET, PE is the major constituent of microplastics, we wanted to perform simulations where both PET and PE are used as the carbon sources by our cells. Therefore, and learning from the previous design problems, the PE degradation pathway was inserted into the model as well. However, simulations with only PE as carbon source showed null growth rate. These results are further explained in the Project Modelling page.
New round of optimization
Our model is still incomplete as it does not include the degradation pathway of the TPA metabolite, leading to an overproduction and secretion in the simulation results. Consequently, the incorporation of the TPA degradation pathway must be the subject of our third iteration on the Engineering Cycle. We plan on adding this pathway to our model and re-test it so we could determine if any further improvements are required or if the resulting model is already ready to support reliable simulations and optimizations. In Figure 4 we have represented our Imagine-Design-Test-Learn-Improve work cycles used for the construction of our model so far. In the future, similar cycles should be used for the next engineering problems.
- Henry, C. S., Zinner, J. F., Cohoon, M. P., & Stevens, R. L. (2009). iBsu1103: A new genome-scale metabolic model of Bacillus subtilis based on SEED annotations. Genome Biology, 10(6), R69. https://doi.org/10.1186/gb-2009-10-6-r69