Introduction

Theory and experimental design

Results

Binding Sites	A	B	C	D	E	F	G	H
nTet_Lac	1	1	1	1	2	2	2	2
nTet_GFP	1	1	2	2	1	1	2	2
nCI_GFP	2	2	1	1	2	2	1	1
nCI_Tet	2	1	2	1	2	1	2	1
nLac	1	2	1	2	1	2	1	2

Introduction

The BananaGuard system is designed with the assumption that our engineered Pseudomonas putida is not outcompeted by other bacteria and fungi found in the rhizosphere of banana roots. However, our system consists of multiple constitutively expressed genes controlling the kill switch and toxin-antitoxin system. Moreover, when the system is switched to the active state by sensing fusaric acid, Pseudomonas putida produces multiple antifungal enzymes and compounds. The integrated synthetic pathways use metabolic resources that would otherwise be dedicated to cellular maintenance or growth. Thereby, the synthetic pathways reduce the bacteria’s potential to sustain themselves in the rhizosphere. Therefore, we wondered to what degree P. putida metabolism was affected by our system and, if need be, how the system could be changed to lessen the impact. To approach this problem a genome-scale metabolic model (GSMM) is used. Our GSMM predicts the interaction of our integrated plasmids with genes originating from endogenous pathways. With this indication of the metabolic stress we get an insight in the metabolic burden caused by the integration of our plasmids and whether the engineered Pseudomonas putida strain is losing the capability to compete with other rhizosphere-populating microorganisms in the resting state.

Theory and experimental design

In order to solve our problem with a GSMM we need a model that:

• contains a comprehensive representation of P. putida metabolism
• is capable of predicting growth rates in various environments
• contains the most reactions that are necessary for our system

For this reason a comparison was made between various GSMM for P. putida. There are multiple genome-scale metabolic models available for Pseudomonas Putida but only two of them are up-to-date, published and verified: iJP962 and iJN746. For the necessary work to model the metabolic stress in Pseudomonas Putida, the best starting model needs to be chosen which has the most potential for accurately modeling the metabolic stress of our system.

Genomic Model	iJP962	iJN746
Reactions	1070	957
Metabolites	959	685
Compartments	1	2
Focus of the model (based on metabolites and reactions)	This model has a broad spectrum of reactions in every class, but is lacking the in-depth reactions.	This model has a focus on the production of interesting compounds for use in biotechnology (aromatic compounds, alkenes, alkanoates).

Taking the above arguments in consideration, model iJP962 is most suitable for modeling our system. Even though the reactions and metabolites in iJN747 are divided in 2 compartments this is not an advantage. Many of the reactions are over the internal membrane and not functional for the core intermediate metabolites. On top of that iJN747 has a focus on compounds that are in a different class of metabolites then the toxins we chose for our system. Therefore iJP962 is more representative of the real system due to more reactions and more metabolites in a broader spectrum. The primary category of concern for toxin production is the amino acid metabolism and iJP962 has the most reactions in this category. Both models did not include all the specific reactions, metabolites and genes that are necessary for modeling the toxin metabolites. These reactions were added manually.

Metabolic modeling

Traditionally, metabolic models only predict metabolic fluxes. Although this is directly applicable to the metabolites that we want to produces, such as DMDS, this does not suffice to answer our questions regarding the impact of the whole system. Therefore, we added the production of all the proteins of interest as if they are metabolites. For this the protein sequences are used to determine the correct stoichiometric ratio of amino acids. The necessary energy and other side-substrate/products are in the same ratio as described for production of proteins for the biomass reaction in the GSMM. From the dynamic model the protein production rate is estimated.

As our metabolites of interest (DAPG, DMDS, DMTS) were not yet included in the GSMM, we had to deduce their production pathways based on databases and literature information(link notebook with all reactions). Furthermore, we took special care that all added pathways are stoichiometrically balanced, which is essential for proper and reliable integration in a GSMM. To calculate the metabolic stress as accurately as possible, the correct constraints have to be set in the model to be certain that all the calculated production rates are biological relevant. To find which constraints are biological relevant in vivo data validation is important. Using this information an estimation can be made about the production range of the toxins which are relevant for P. putida.

Programs and toolboxes

Matlab (matlabsource) was used in combination with the Cobra toolbox(cobrasource) for handling the model. The Cobra toolbox is designed for working with genome-scale metabolic models. To solve the flux balance analysis a linear equation solver is needed. Gurobi(gurobisource) was chosen because of its compatibility with Matlab and the Cobra toolbox.

Flux Balance Analysis

Flux Balance Analysis (FBA) is a framework for investigating the theoretical capabilities of a stoichiometric metabolic model S and can be solved by using a linear solver program. This is possible because of linear objective function and linear constraints (described below). FBA is based on the following assumptions to ensure its linearity:

1. steady state: The fluxes are considered to have attained a static equilibrium value and do not change through time.
2. Thermodynamic constraints: (ir)reversibility of Reactions
3. No enzyme saturation: The maximum flux for every reaction is not dependent on the concentration of the enzymes used for that reaction.
4. There is no accumulation of intermediate metabolites in the model.

The FBA method uses a representation of the metabolic reaction network in the form of a stoichiometry matrix (S) where :

• Each column corresponds to a reaction R_i
• Each row corresponds to a metabolite C_j

The definition of S is :

The FBA problem is then formulated as a maximization or minimization problem under the determined constraints:

Where :

• v is the vector of unknown reaction fluxes
• c is a vector of constants defining the objective function
• S is the stoichiometry matrix
• lowerbound and upperbound are vectors of constraints (minimal and maximal flux values for each reaction)

Media composition

For metabolic modeling the constraints of your exchange reactions are important for the results. With these exchange reactions you can dictate what’s available in the model, you need arbitrary bounds as actual uptake rates are unknown. This also enforces the later decision to not include all root components but only the most prevalent. You don’t want that rare compounds have a huge impact on the model predictions. This does imply that the model predictions could be slightly negative: with including the other available compounds the system may work even better.

In order to set the medium constraints of the models all lower and upper bounds are first set to the value of 0 and 1000 arbitrary units, respectively. Then one by one the lower and upper limit of the reaction bounds are set to -1000 for one reaction at a time to your composition of interest. If the lower bound is set to a value higher than zero a minimum flux of that value is forced in a FBA. To get the most realistic approximation for our engineered P. putida in the soil the M9 media composition was used with carbon as limiting nutrient. The carbon source is based on the exudates of banana roots and consists of 34% sucrose and 66% divided over multiple amino acids. We have neglected other secondary metabolites in banana roots exudates because they are exudated in relatively small amounts.

Results

P. putida is able to grow in the rhizosphere with a doubling time of 3 hours resulting in a growth rate of 0.23 h^-1. Because of the linearity of the growth rate and carbon uptake rate in this model we expect that the carbon uptake rate is ≈4.5 mmol gDW^-1hr^-1 in the rhizosphere of the banana roots. Figure 1 depicts how the growth rate of BananaGuard compares to the growth rate of the wild type P. putida depending on the carbon uptake rate of the organism in the resting state and active state of the kill switch. To get the best case scenario glucose was chosen as carbon source, because of the efficient degradation of glucose. To get an insight of the growth rate in the rhizosphere a carbon uptake rate between 4-4.5 mmol gDW^-1hr^-1 is expected. Our engineered P. putida is still able to grow at almost max growth rate (>99%) compared to the wild type in the resting state. We can conclude out of this that P. putida is not getting a huge disadvantage to compete with other bacteria. When looking at the active state we are not interested in the capability to compete with other bacteria but the capability to produce around the roots all the toxins with a constant speed, without overloading a metabolic pathway and causing shortage of intermediates. In this case our engineered strain is still able to growth with 50-70% of the maximum growth of the wild type. This indicates that metabolic stress is not a bottleneck for the production of toxins in our activated system.

To get a more realistic result for the soil in banana farms the carbon source is changed to have the same ratio of carbohydrates and amino acids as found in the exudate's of banana roots. This new carbon composition causes a reduced growth rate because of less efficient degradation of these compounds. In figure 1 the relative growth rate is shown for the resting and active state of the kill switch with the new carbon source. These changes in the carbon composition do not change our conclusion previously made. The range of the relative growth rate has changed from 50-70% to 45-60% indicating it still has the capacity to grow and maintain the constant toxin production with one condition that our P. putida needs to stay close to the roots to maintain its carbon uptake rate.

Oxygen necessity in the soil

P. putida is a strict aerobic bacteria. This means that without oxygen available the bacteria cannot grow. To inhibit the growth of the Fusarium our engineered Pseudomonad must grow in the rhizosphere of the banana roots. If there is not enough oxygen present in the rhizosphere it might not be possible anymore to sustain anti-fungals production or other anaerobic bacteria can outcompete our P. putida. In figure 2 the growth rate in the active state is plotted as a function of the oxygen and carbon availability. If there is little oxygen available the growth rate drops and so will the production of the toxins. However the depletion of oxygen should not be a problem because there is a layer of oxygen present closely to the roots. This supports the task of our P. putida to protect the banana roots from Fusarium invasion and will contain our engineered P. putida close to the banana roots.

Parameter analysis results

Every parameter in this metabolic model is an estimation based on literature growth experiments except for the pyoverdin production rate. This parameter has been determined from growth experiments with P. putida (Link notebook walter pyoverdin) . So every parameter is biological relevant, but because different studies report different values a mean has to be taken between the different data points. This means that the parameter can fluctuate in a small range. To compensate this a parameter analysis has been performed. Every parameter that could have a significant impact on the results has been checked over a range, dependent on the accuracy of the calculated parameter. The same analysis has been performed for all the fluctuations in the parameter set combined. This analysis shows the possible fluctuations of the modeled system and which parameters has greatest impact on the system.

Introduction

Theory and experimental design

Results

Table of parameters

References