Team:Sorbonne U Paris/Model


CONTEXT

The synthetic strain of Chlamydomonas reinhardtii that we want to create is characterized by the addition of exogenous genes that encode for specific enzymes that are the FAT-A, the FAT-B2, the DGAT1-2 and the LPAAT-A from the palm tree (Elaeis guineensis). Those enzymes catalyze reactions that have a key role in the lipid biosynthesis pathway. In that context, we found it interesting to implement those modified synthetic pathways in an in silico genome-scale model of Chlamydomonas reinhardtii, thus allowing us to study the impact of metabolic engineering on our model through the predictive analysis of the fluxes variations in the reactions that lead to the production of triacylglycerol.

FLUX BALANCE ANALYSIS

Flux balance analysis (FBA) is a mathematical approach of fluxomics allowing the analysis of flow metabolites through biochemical networks. It is performed on genome-scale metabolic network reconstructions that contain detailed informations on the organism of interest such as the exact reaction stoichiometry, reversibility and its relation with genes and proteins, according to the insights gained from literature and experimental data. FBA is usually used to predict the growth rate or the production of a specific metabolite of interest.

The constraint-based reconstruction and analysis (COBRA) approach consists in representing each reaction (n) of a system with m compounds as a part of a stoichiometric matrix (S) of size m × n. Stoichiometric coefficients associated to each metabolite imposes flux balance constraints to the system. Consumed metabolites have a negative coefficient and metabolites that are produced have a positive coefficient. It takes a value of zero when the metabolite is not evolved in any particular reaction. Thus, each reaction in the system is given a flux value (mmol.gDW-1.h-1) when optimizing for a particular reaction, and is characterized by a lower and upper bound that define the range of allowable flux distribution at steady-state mass conservation (S×v=0, v as a vector containing the individual flux values for each reaction that are to be determined). 1

FBA allows us to maximize or minimize the flux for a chosen reaction, which is defined as the objective function subject to a set of constraints. To obtain a realistic behavior, the model needs to be updated with parameters and constraints that are as close as possible to what is observed in nature, using experimental data. But it has to be taken into account that the optimization process using FBA algorithm does not allow just one solution, and for every situation, many other solutions could exist.

METHOD

The study of the in silico Chlamydomonas reinhardtii model iCre1355 2 was conducted through MATLAB software using COBRA toolbox 3 and GLPK solver to perform the calculations. The model was uploaded as a SBML file 4 and then converted to an excel file containing all the data for reactions and metabolites included in the system, and a matlab file to perform the analysis and solve the model through a set of commands. 5

The first step consists in implementing to the model the new synthetic pathways created by the addition of the genes of interest. It requires the writing of the reactions that are catalyzed by the enzymes encoded by those genes. We have noticed that the reactions were already present in the model, because Chlamydomonas reinhardtii naturally produce some triacylglycerol (TAG) and have similar lipid pathways as in the palm tree organism. 6

Thus, we simulated the addition of our genes of interest by relatively increasing the fluxes of each reaction that is catalyzed by the enzymes of interest. The objective function that was chosen was to optimize the production of TAG under mixotrophic conditions by adding photons in the environment as a source of light and acetate as a source of carbon as it is usually made for Chlamydomonas reinhardtii culture. In addition, all nitrogen sources were removed from the environment to put the microalgae into a nitrogen starvation state, thus allowing the accumulation of TAG, according to the conditions that we want to create for the in vitro experiments. The variations in the fluxes and the amount of TAG produced were analyzed to determine in which way the addition of the genes of interest encoding respectively for DGAT1-2 and LPAAT-A in the updated model could optimize the final amount of TAG produced.

RESULTS

First, we have studied the evolution of TAG flux when fluxes of reactions catalyzed with DGAT1-2 (DGATRxn) and/or LPAAT-A (LPAATRxn) are increased. The impact of FAT-A and FAT-B2 enzymes was not studied here because the model did not allow the specific analysis for the production of palmitic acid-containing TAG and oleic acid-containing TAG. Thus, for the reactions catalyzed by these enzymes, fluxes were left unconstrained, so it does not restrict variations in fluxes for downstream DGATRxn and LPAATRxn. TAG flux represents the final amount of TAG metabolite produced. It was simulated by the creation of an exchange reaction (= secretion) to be highlighted.

As shown in Figure 1, we have noticed that when DGATRxn fluxes are increased, there is no variation in the final amount of TAG produced, while TAG flux is greatly increased when LPAATRxn flux values are enhanced, and it takes the same value of TAG flux as when both DGAT and LPAAT Rxn fluxes are increased. It means that the addition of LPAAT-A enzyme seems to be necessary and sufficient for maximizing final amount of TAG produced. Indeed, TAG flux seems to be directly related to LPAATRxn flux and takes the same value until it reaches an optimized amount of 0.058 mmol.gDW-1.h-1 (Figure 2).

Moreover, to better understand in which way the addition of DGAT1-2 and LPAAT-A enzymes could have an impact on TAG biosynthesis pathway, fluxes of all the reactions implicated were printed and analyzed. The main reactions implicated in TAG biosynthesis pathway occurring in endoplasmic reticulum - following the synthesis of Fatty Acyl-CoA in Chloroplast - were represented in Figure 3. On one hand, we have noticed that when flux of DGATRxn is doubled, the final value of TAG flux is not changing, and that an enzyme registered in the model as a triacylglycerol lipase uses the excess of TAG to produce diglyceride and fatty acid, probably to fill the lack of substrate upstream the reaction catalyzed by DGAT1-2 enzyme. On the other hand, increasing LPAAT-A enzyme seems to be sufficient to optimize the final amount of TAG produced. Diacylglycerol (DAG) obtained by LPAAT-A-catalyzing reaction is transformed into TAG by DGAT1-2 until there is not enough enzyme to take charge of DAG. Then, it seems that flow metabolite is directed toward another reaction that is catalyzed by a phospholipid:diacylglycerol acyltransferase (PDAT) and leads to the production of Lysophosphatidylethanolamine (LPE) and Triacylglycerol (TAG). In that way, final amount of TAG is maximized following the increase of flux through LPAAT-A-catalyzed reaction.









CONCLUSION AND PERSPECTIVES

In addition to what we have learnt from literature research, the modeling work helped us choosing the enzymes that will be added to our synthetic strain of Chlamydomonas reinhardtii for the optimization of the amount and composition of TAG produced. According to the results, FBA allowed us to confirm that LPAAT-A is a key enzyme in TAG production pathway, and that increasing its quantity could efficiently lead to an increase of TAG production. However, we have to notice that FBA has some limitation. It does not use kinetics parameters and neither considers the variability in transcriptional behavior inherent to each gene. So it has to be taken into account when analyzing our results. Thus, despite the observations made on the in silico model of C.reinhardtii, we decided to test the DGAT1-2 gene in addition to the LPAAT-A gene in the synthetic strain for the first experiments, according to what we have learnt from previous studies about the key role of DGAT1-2 enzyme in TAG production 7, as well a FAT-A and FAT-B2 enzymes, that were not studied here.

Therefore, the wetlab lab experiments that will be conducted following drylab studies will allow us to get more experimental data and gain new insights about TAG biosynthesis in our model. Various strains containing different genetic constructs will be built. On one hand, cells will be transformed with LPAAT-A-containing plasmid or with DGAT1-2-containing plasmid or with LPAAT-A-DGAT1-2-containing plasmid to test respectively the independent or simultaneous effect of the two enzymes encoded by those genes on the general amount of TAG produced. This would allow us to finally determine if it has an interest to add DGAT1-2 in addition to LPAAT-A. On the other hand, the same experiments will be conducted with the addition of FAT-A gene or FAT-B2 genes to check the specific production of palmitic acid and oleic acid in two independent strains.

Those experimental results will be analyzed and implemented again for in silico analysis to improve our strategy before going back again on wetlab experiments. Applying the design-build-test cycle strategy should lead us to get an optimized TAG-producing strain of Chlamydomonas reinhardtii. The advantage of using FBA to perform our in silico analysis is that it provides a better understanding of what is going on inside the organism of interest and its complexity. Therefore, it gives us large perspectives in metabolic engineering of this microalga for lipid compounds production.

As we have learned a lot from our modelling work on C. reinhardtii, we want to compete for the Best Model Award.



  1. J. D. Orth, I. Thiele, et B. Ø. Palsson, « What is flux balance analysis? », Nat. Biotechnol., vol. 28, no 3, p. 245-248, mars 2010.
  2. S. Imam et al., « A refined genome-scale reconstruction of Chlamydomonas metabolism provides a platform for systems-level analyses », Plant J., vol. 84, no 6, p. 1239-1256, déc. 2015.
  3. L. Heirendt et al., « Creation and analysis of biochemical constraint-based models using the COBRA Toolbox v.3.0 », Nat. Protoc., vol. 14, no 3, p. 639‑702, 2019.
  4. « baliga-lab/Chlamy_model_iCre1355 », GitHub. [En ligne] [Consulté le: 28-sept-2019].
  5. J. Nogales, « A Practical Protocol for Genome-Scale Metabolic Reconstructions », 2014, p. 1‑25.
  6. Y. Li-Beisson, F. Beisson, et W. Riekhof, « Metabolism of acyl-lipids in Chlamydomonas reinhardtii », Plant J., vol. 82, no 3, p. 504‑522, mai 2015.
  7. S. Dussert et al., « Comparative Transcriptome Analysis of Three Oil Palm Fruit and Seed Tissues That Differ in Oil Content and Fatty Acid Composition1[C][W][OA] », Plant Physiol., vol. 162, no 3, p. 1337‑1358, juill. 2013.