The goal of our modelling was to optimize the yield of a target product in our Komagataella pastoris (formerly Pichia Pastoris) strains through knockouts or gene overexpression.
We used the model iMT1026v3, which is an updated version of the iMT1026 model. The model's accuracy was improved by updating biomass equations and validated with flux balance analysis (FBA). The updated model has improved predictive capabilities over a wider range of substrates, providing valuable insights for bioprocess engineering in Komagataella pastoris. (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5743807/)
The primary target we selected was the increase in production of Acetyl CoA. There were multiple reasons for this selection.
OptKnock is designed to identify gene knockouts within genome scale models that will lead to overproduction of a desired product, while still allowing for fast growth. Optknock is also designed to facilitate growth coupling. Growth coupling is when the production of the desired compound is tightly coupled to the growth of the microorganism. (https://onlinelibrary.wiley.com/doi/abs/10.1002/bit.10803)
OptGene operates similarly to OptKnock but builds on the concept by using Evolutionary Algorithms and Simulated Annealing to optimize sets of gene deletion for the desired objective function. Simulated annealing is a probabilistic optimization strategy that iteratively explores the solution space by accepting worse solutions with a decreasing probability, controlled by a temperature parameter. Evolutionary algorithms are optimization techniques inspired by natural selection. They maintain a population of candidate solutions, iteratively applying genetic operators like selection, crossover, and mutation to evolve and improve solutions over generations. Both techniques eventually converge towards an optimal or near-optimal solution. The strains are then simulated using FBA or Minimization of Metabolic Adjustment (MOMA). (https://core.ac.uk/download/pdf/55609541.pdf) In our case MOMA was used.
In FSEOF you first enforce a flux directed toward your target product and then search for variations in the metabolic fluxes caused by this change. Fluxes that increase because of this change are assumed to be possible overexpression targets. The flux through your target is increased iteratively so that you can determine the relationships between the fluxes. (https://www.researchgate.net/publication/230713107_Flux_variability_scanning_based_on_enforced_objective_flux_for_identifying_gene_amplification_targets)
The Cameo package was used to perform all three of these metabolic modelling techniques. (https://github.com/biosustain/cameo)
We ran the OptKnock algorithm on the iMT1026v3 model with Acetyl CoA cytosolic and Acetyl CoA mitochondrial.
We ran the OptGene algorithm on the iMT1026v3 model with Acetyl CoA cytosolic and Acetyl CoA mitochondrial.
No soloutions found for both mitochondrial and cytosolic Acetyl-CoA
Here we used the iMT1026v3 model to perform FSEOF with mitochondrial Acetyl CoA as a target.
| Target: AcetylCoA m | Flux begining | Flux end | Change |
|---|---|---|---|
| H2Ot H2O transport via diffusion PAS_chr3_0763 | -278.174 | -356.926 | 0.787521 |
| GLUDxi glutamate dehydrogenase (NAD) PAS_chr2-1_0311 | 0.507641 | 0.055312 | 0.452329 |
| PDHcm pyruvate dehydrogenase (dihydrolipoamide dehydrogenase) PAS_chr2-1_0089 or PAS_chr2-2_0048 or PAS_chr1-3_0094 | 0.605326 | 0.329448 | 0.275878 |
| FRDO frdo PAS_chr3_0058 | -0.03704 | -0.30093 | 0.263891 |
| ICDOXSUCm Isocitrate:NADP+ oxidoreductase PAS_chr2-1_0120 and PAS_chr4_0580 | 0.283293 | 0.063642 | 0.219651 |
| OSUCCCLm oxalosuccinate carboxy-lyase (2-oxoglutarate-forming) PAS_chr2-1_0120 and PAS_chr4_0580 | 0.283293 | 0.063642 | 0.219651 |
| SUCD1m succinate dehydrogenase PAS_chr3_0225 | 0.230071 | 0.057793 | 0.172278 |
| SUCD3_u6m succinate dehydrogenase (ubiquinone-6), mitochondrial PAS_chr3_1111 and PAS_chr2-2_0283 and PAS_chr4_0733 and PAS_chr1-4_0487 and PAS_chr3_0424 | 0.230071 | 0.057793 | 0.172278 |
| AKGDH1 2-Oxogluterate dehydrogenase PAS_chr2-1_0089 | 0.22385 | 0.05711 | 0.16674 |
| AKGDH2 2-Oxogluterate dehydrogenase PAS_chr2-1_0089 0.22385 0.05711 | 0.22385 | 0.05711 | 0.16674 |
| 10FTHFtm 10-Formyltetrahydrofolate mitochondrial transport via diffusion | -0.03699 | -0.20197 | 0.164986 |
| GLUDyi glutamate dehydrogenase (NADP) PAS_chr1-1_0107 | 0.836535 | 0.685176 | 0.151359 |
| Ht H+ diffusion | -0.41045 | -0.53987 | 0.129425 |
| PGK phosphoglycerate kinase PAS_chr1-4_0292 | -0.70111 | -0.80223 | 0.101124 |
| GAPD glyceraldehyde-3-phosphate dehydrogenase PAS_chr2-1_0437 | 0.701107 | 0.802231 | -0.101124 |
| GLNS glutamine synthetase PAS_chr4_0785 | 0.088692 | 0.207654 | -0.118962 |
| FUM fumarase PAS_chr3_0647 0.038788 0.20217 | 0.038788 | 0.20217 | -0.163382 |
| MTHFCm methenyltetrahydrifikate cyclohydrolase, mitochondrial PAS_chr1-4_0632 | 0.036985 | 0.201971 | -0.164986 |
| MTHFDm methylenetetrahydrofolate dehydrogenase (NADP), mitochondrial PAS_chr1-4_0632 | 0.036985 | 0.201971 | -0.164986 |
| THFtm 5,6,7,8-Tetrahydrofolate transport, diffusion, mitochondrial 0.036985 0.201971 | 0.036985 | 0.201971 | -0.164986 |
| PPA inorganic diphosphatase PAS_chr1-3_0028 or PAS_chr1-3_0070 or PAS_chr2-1_0101 | 0.204803 | 0.41832 | -0.213517 |
| PGCD phosphoglycerate dehydrogenase PAS_chr2-1_0657 | 0.06652 | 0.304171 | -0.237651 |
| PSERT phosphoserine transaminase PAS_chr3_0566 | 0.06652 | 0.304171 | -0.237651 |
| PSP_L phosphoserine phosphatase (L-serine) PAS_chr4_0285 | 0.06652 | 0.304171 | -0.237651 |
| PIt2r phosphate reversible transport via symport PAS_chr3_0495 or PAS_chr4_0337 or PAS_chr2-1_0235 | 0.043446 | 0.301635 | -0.258189 |
| ADK1 adenylate kinase PAS_chr3_0257 | 0.105745 | 0.407435 | -0.30169 |
| NH4t ammonia reversible transport PAS_chr1-4_0394 or PAS_chr2-2_0391 | 0.401186 | 0.736762 | -0.335576 |
We have successfully identified potential target genes for both overexpression and knockout. It could potentially be beneficial to do modelling on a similar organism like Saccharomyces cerevisiae that has newer and more refined models and compare these results to the results we generated.
More experimental work is necessary to validate our results and further optimize the yield.