Over

view

To anticipate module selection and predict experimental results in wet experiments, we modeled the sensing switch system, MEL fermentation system, semi-continuous fermentation system, and medium-chain fatty acid production system in this project. The modeling module mainly contains four parts.

In modeling the sensing switch system, we identified defects in the benzo[a]pyrene receptor and vetoed the proposal, employing a red light switch as the element controlling the initiation of fermentation. Then, we predicted the efficiency of the red light sensing system to determine the optimal red light irradiation intensity during fermentation. Ultimately, we simulated the arrangement of red light sources during fermentation, providing an efficient and energy-saving arrangement method.

In the MEL fermentation system, we conducted kinetic modeling of the metabolic pathway of Moesziomyces, calculated the changes of various substrates and products, and compared the production of MEL and intracellular lipids before and after simulated gene knockout. The model reflects actual changes well and can provide a reference for teams using Moesziomyces or conducting MEL fermentation. Next, we analyzed the data of MEL obtained from fermentation and fermentation substrates in the experiment, used correlation analysis models and random forest models for data processing, and provided the relationship between different kinds of MEL produced by fermentation substrates, assisting us in the customized production of medium chain fatty acid.

In the medium-chain fatty acid production system, we analyzed the structure of the fusion protein constructed by alkT and GDH, screened out the best connection method, and successfully reduced the interaction between the connector and the protein itself through simulated point mutation.

In the semi-continuous fermentation system, we modeled semi-continuous fermentation, calculated the impact of the initial inoculation amount of strains, fermentation substrate supplementation rate, product separation interval, etc., on fermentation yield, to assist in adjusting various fermentation indicators in the experiment to achieve the highest fermentation efficiency.

1.Receptor of benzopyran

1.1 Polycyclic aromatic hydrocarbon receptor

Due to the presence of benzo(a)pyrene in kitchen fee oil, we initially intended to use benzo(a)pyrene receptors to control the fermentation of the engineered bacteria. Since the dissociation equilibrium constant and Hill's constant of benzo(a)pyrene pyrene's dissociation equilibrium constant and Hill's constant have not been studied before, we determined the possible ranges of their dissociation equilibrium constants and Hill's constants by reviewing the literature and igem's previous team by summarizing the existing PAH substances.



Assumptions:
(1) The dissociation equilibrium constants of benzo(a)pyrene to switch and Hill's constant are in the range of common PAHs.
(2) the complex of AHR with ARNT is constant within the fine
(1)

It can be found that the Hill coefficient n will mainly affect the sensitivity of the switch, while the dissociation equilibrium constant k mainly affects the minimum detected concentration of the switch.

1.2 Proliferation of benzo(a)pyrene

We determined the content of benzo(a)pyrene in kitchen waste oil, and considering that it takes a certain time for benzo(a)pyrene to enter the cell, we established a diffusion equation for benzo(a)pyrene from the solution to the cells of the engineered bacteria.



Assumptions:
(1) The volume of the fermentation broth is large relative to the bacteria, therefore the concentration of
benzo(a)pyrene in the fermentation broth is considered to be constant during the sensory phase.
(2) The cell volume is small, therefore the concentration of benzo(a)pyrene in the cell is considered equal everywhere.
(3) Cells do not autonomously efflux or degrade benzo(a)pyrene during diffusion



By calculations of diffusion, the concentration of intracellular benzo(a)pyrene will reach a maximum after about 600 seconds of addition, and when the system is at equilibrium, the concentration of intracellular benzo(a)pyrene is equal to the concentration of extracellular benzo(a)pyrene.

1.3 Switch Viability Analysis

Since the content of benzo(a)pyrene was found to be very low in the kitchen waste oil, we did the relationship between the binding rate at a specific benzo(a)pyrene concentration and its KA and n values.



It was concluded that when the KA and n values of benzo(a)pyrene are smaller, the more favorable the switch is for benzo(a)pyrene detection, but because the concentration of benzo(a)pyrene in the solution is too low, if we want the binding rate P to reach more than 0.5, we need to mutate the binding site of benzo(a)pyrene to make its KA close to the order of magnitude of 10^-7, and this is almost impossible. We therefore decided to replace the switch.

2.Receptor of Red light

In the yeast system, all fusion proteins are constitutively nuclear-localized due to the presence of the natural nuclear localization sequence (NLS) present in the GBD tag or the SV40 NLS motif fused to the GAD fusion mate. Thus, the only light-dependent event in this system is the interaction of the photosensitive pigment with its corresponding protein partner.



Assumptions:
(1) the expression and degradation of relevant proteins, and the uptake and emission of substances reach equilibrium, and the total concentrations of PhyA and FHY1 are constant.
(2) all photosensitive pigments are in an inactive (Pr) state prior to the light pulse.
(3) the dark inversion rates of PhyA and PhyA-FHY1 complexes are the same.
(4) The luciferase-luciferin subsystem approximates the steady state before light treatment.
(5) The reaction system is a closed system with no material exchange with the outside world.
(6) Assuming that the concentration of the bacterium remains constant throughout the process.


2.1 Simulation of red light switching dynamics

Since realistic fermentation experiments are more complicated, we used a computer to perform a pre-experiment of fermentation simulation to provide a reference for selecting experimental conditions. We constructed kinetic equations for the red light system, using the law of mass action to describe changes in substance binding and the Hill equation to describe the transcription process. Predictions were made about the effects on the expression of target proteins when the light intensity of the main control factor, red light, was changed.

(1)

(2)

(3)

(4)

(5)



The results show that the changes of each substance under different red light intensities are as expected, indicating that the model itself can well describe the kinetic process of red light switching.

2.2 Relationship between light intensity and optimal separation time

In order to further investigate the variation of protein product concentration with time under different red light intensities RL to better improve our production process, we plotted the variation of protein product concentration cP with time t for each red light intensity.



It can be seen that for the same downstream gene, the concentration of the protein product increases faster when the light intensity of red light is higher, although a small decrease in its maximum concentration was found, which was found to be due to the parameter Ki_R by further analysis. Thus for continuous fermentation, one can simply increase the red light intensity to increase the expression of the target protein, whereas for semi-continuous fermentation, the relationship between the red light intensity and the time of isolation of the fermentation product needs to be balanced in order to achieve maximum fermentation efficiency.

We further let plot the relationship between the time required for the cP concentration to reach the maximum and the growth rate of cP concentration to reach the maximum and the red light intensity RL, in order to help how to control the separation time in semi-continuous fermentation, assist the selection of light intensity in wet experiments, and narrow down the selection interval of light intensity. The results show that the expression amount of the target protein and the expression rate to reach although the time has a strong marginal effect with the increase of light intensity, in the light intensity of 50 umol/m^2/nm, it has almost reached the optimal one, so we choose 50-60 umol/m^2/nm light intensity range for the subsequent test.



2.3 Red light system arrangement

Due to the fact that the fermentation of MEL-producing sap is thicker, it may cause the red light system to not work well. We therefore used a simulation based on Rombauer's law to model the propagation of light through the sap during fermentation to determine what lighting arrangement scheme we could use to maximize fermentation with minimal energy consumption in order to reduce the expense of subsequent experiments and increase their success.



I = I0 * e^(-αx)



The fermentation tank we used is a cylindrical vessel with a radius of approximately 10cm and a height of approximately 40cm. Considering the actual fermentation process, most of the fermentation culture is located in the upper layer, while the lower layer is mainly occupied by the product MEL. After continuous adjustment of the number and positioning of the lights, we finally determined that wrapping 8 light bulbs with an intensity of 180 umol/m^2/nm at distances of 4cm, 12cm, and 20cm from the top of the tank would ensure that the upper 2/3 region within the fermentation tank reaches a light intensity of 50 umol/m^2/nm.

3.Suicide switch

Due to the inducers are relatively expensive, we did statistics on the relationship of sterilization efficiency with the concentration of the inducer, hoping to find a concentration of the inducer that would be effective in inhibiting the suicide switch but not very high.



Fitting using experimental fermentation data assumes that the sterilizing efficiency equation is:

Y(%)= [(N_i- N_f) / N_i] × 100%

The fitted function yields;

Y = 11.288966 * (X+7.717510).^-1 - 0.849

P = (0.95082-(-Y+0.6021))/ 0.95082

fitting error:0.005990



From this we can get that when using 10g/L of inducer, it is possible to inhibit the suicide switch effect by about 50% and keep the strain growing well. The concentration of the inducer required to increase the switching efficiency increases dramatically.

Conclusions:

The modeling of this module analyzes the performance of three sensory switches in the system: benzo(a)pyrene sensory switch, red light switch, and suicide switch. In the benzo(a)pyrene sensory switch module, we established an evaluation model for the switch activated by substrate binding, analyzed the effects of substrate concentration, Dissociation constant, Hill constant on the efficiency of the switch, and negated the feasibility of the switch due to the low benzo(a)pyrene concentration, to avoid the failure of the switch design affecting the experiments conducted. In the red light switch module, we performed kinetic simulation of the red light switch to evaluate its expression effect, and found the optimal red light intensity of 50-60 umol/m^2/nm, and then designed an optimal arrangement of the light around the fermenter to provide a reference for the subsequent fermentation experiments. In the suicide switch module, we calculated the relationship between the concentration of the inducer and the sterilization efficiency, and determined that using an inducer concentration of 10 g/L was a cost-effective solution.

1.Enzymatic kinetic analysis

The system uses Rubredoxin-NAD(+) reductase (AlkT) from Pseudomonas oleovorans, with Glucose 1-dehydrogenase (gdh) from Bacillus subtilis (strain 168), which originally consumes NADH by AlkT to produce electrons, and AlkG transfers the electrons to AIkB, which ultimately introduces an oxygen atom from the molecular oxygen in the terminal position to achieve the light radicalization reaction.



We first performed kinetic simulations of the reaction system and found that increasing the electron transfer efficiency can effectively increase the rate of product production in the early stages without changing the relevant enzyme activities. It can therefore be determined that the construction of fusion proteins can indeed increase the production of medium-chain fatty acids.



Assumptions:
(1) The electron concentration is never saturated during the reaction and the electron concentration is one of the factors affecting the reaction rate.
(2) The catalytic efficiency of each catalytic site of the fusion protein is the same as that of its monomer.
(3) The fusion protein and its monomer have the same concentration for the enzymatic reaction.
(4) The fusion protein only changes the efficiency of electron transfer.




2.Construction and optimization of fusion protein

2.1 Screening of fusion protein splicing patterns

In order to further provide the catalytic ability of the system, we intended to construct a fusion protein of AlkT and GDH to increase the efficiency of electron transfer. Through structure prediction and screening[2], we finally decided to construct an intracellular self-assembling multi-enzyme complex designed to assemble cofactors GDH and AIKT using the SpyCatcher-SpyTag system.



ASGAGGSEGGSEGG



GGGGGGGG



AlkT-SpyCatcher



GDH-SpyTag



AlkT-SpyCatcher- SpyTag- GDH



* The yellow portion is the protein itself and the blue portion is the SpyCatcher-SpyTag system or linker

The results show that the accuracy of protein prediction is high, the structure of the protein itself does not change much, and the SpyCatcher-SpyTag system are located on the surface of the protein, which theoretically does not affect its own catalytic efficiency as well as the docking of the joints.

2.2 Validation of alkT activity

Because alkT is a key enzyme in the production of electrons, its activity is critical to the success or failure of the entire experiment. To verify that alkT with the addition of the SpyCatcher-SpyTag system performs its original function, we molecularly docked alkT with its substrate, NADH, and performed the same test on the original alkT and various alkT-DGHs connected with linkers to compare and analyze their substrate-binding abilities.




* alkT without tag (left), alkT with added SpyTag (right) molecularly docked with NADH





Docking parameter table



* Smaller values for all data indicate better docking results

In order to minimize the effect of the SpyCatcher-SpyTag system on the structure and catalytic activity of alkT, we finally decided to link the smaller SpyTag to alkT. For the fusion proteins linked by linker, because of their lower score of predicted structure and no significant improvement in the stability of substrate binding compared to the SpyTag-SpyCatcher system, and considering that their larger proteins may lead to difficulties in expression or transportation, etc., we finally chose to use the SpyTag-SpyCatcher system as the linkage structure of the fusion protein.

2.3 SpyTag and AlkT structure optimization

Since the simulated structures show that SpyTag and SpyCatcher will interact with the AlkT and GDH proteins causing them to move closer to them, this may cause the SpyTag-SpyCatcher system to not work properly or have an effect on the structure and activity of the protein itself. By analyzing the structure of the AlkT-SpyCatcher with the CDH-SpyTag protein, we know the site of its interaction with the protein to which it is attached.



* The cyan color is the amino acid that interacts with the protein on SpyTag and SpyCatcher.

Considering that the linkage site of the SpyTag-SpyCatcher system is ASP on SpyTag and Lys on SpyCatcher, we decided to substitute glycine for some of the amino acids[4].(The His-263、TYR-270、THR-273、LYS-274 in SpyTag. The MET386、Tyr388、His390、His391、His394、His395 in SpyTacher.)



Structural simulations were performed after replacing amino acids in the relevant sites, and SpyTag no longer interacted with the protein itself. This method can be used as a direction to optimize the performance of fusion proteins in the future, and the specific binding efficiency and enzyme activity changes need to be determined by subsequent experiments.

Conclusions:

This module is all about optimizing the alkT-GDH system to improve the production efficiency of medium-chain fatty acids. In this module, we first analyzed the alkT-CDH system for kinetic simulations to find the key steps that affect the production of medium-chain fatty acids, and then found that increasing the electron transfer efficiency by constructing fusion proteins through the linker or SpyTag-SpyCatcher system can significantly increase the yield of the products. Subsequently, we carried out structural simulation and screening of various fusion proteins, and finally decided to use the SpyTag-SpyCatcher system to construct the fusion protein.

As we found that there were some structural problems, we replaced some of the amino acids, and successfully solved the problem of interaction between the SpyTag-SpyCatcher system and the protein itself, which provided a good basis for the construction and subsequent optimization of fusion proteins in the experiment construction and subsequent optimization.

1.Metabolic model of Moesziomyces aphidis

When producing MEL, strain XM01 produces a large amount of intracellular lipids, and the content of intracellular lipids is more than 60%. Compared with MEL secreted to the outside of the cell, intracellular lipids mainly contain long-chain fatty acids (triacylglycerol and sterol esters), and due to the coexistence of different fatty acid pathways in strain XM01, a large amount of MEL is formed and the accumulation of intracellular lipids is caused at the same time. The synthesis of intracellular lipids and MEL, energy, fatty acid precursors and acetyl coenzyme A are all necessary for metabolic processes, and the synthesis of intracellular lipids will inevitably impede the large amount of MEL synthesis to a certain extent. Therefore, by identifying the key genes for intracellular oil synthesis in strain XM01 and knocking them out, the synthesis of MEL can be enhanced. Therefore, we measured the relative expression of various genes in strain XM01 and modeled them according to the catalytic rates associated with specific enzymes, and observed the changes in intracellular lipid production by down-regulating the expression of different genes in the intracellular lipid metabolic pathway, in order to identify the key genes for intracellular oil production and to knock them out.



2.Carbon flow reconstruction

Construct a metabolic model based on the above figure, in which each substance changes in the form shown below.



Assumptions:
(1) Knockout changes only the gene expression, not the activity of the relevant enzyme.
(2) The relative expression of each gene will change after gene knockdown, but the total expression will remain unchanged.

The results of bacterial growth and intracellular lipid production were observed by adjusting the expression of individual genes in the model, initially setting the expression efficiency to 100%, and for knockdown setting it to 20% of the original. This can help the subsequent modification of the strain to increase the efficiency of MEL production by adjusting the expression of each gene and identifying the key genes for MEL production and intracellular lipid production by knocking out/over-expressing them.



Before simulated knockout



After simulated knockout

It can be seen that simulating the pre-knockout and post-knockout, the growth of the knockout bacteria was accelerated by about 20% (yellow line) intracellular lipid production was reduced by about 80% (gray line), which is basically consistent with the results obtained in the experiment. Finally, we performed double knockdown of ARE1 gene and DGA1 gene of Aphis mossambicus XM101. The model can help us to get the results after modification of its metabolic pathway in a simple way, which can help to improve the yield further in the following, and can also provide a reference for other teams using Aphidium molluscum black powder fungus. However, due to some parameter settings, the model can only analyze the results qualitatively at present, after which we will use more experimental data for fitting to achieve quantitative results.

3.Customization of medium chain fatty acid

Since the composition of MEL produced by fermentation of different substrates varies greatly, and the economic value of medium-chain fatty acids of different chain lengths among MELs is different, the analysis of the relationship between substrate and product compositions can help to realize the production of specific medium-chain fatty acids.



*The closer the data is to 1, the stronger the positive correlation is, and vice versa for the negative correlation.

It can be seen that the specificity of different substrates will cause changes in the structure of MEL synthesis, the ratio of oleic acid and linoleic acid determines the proportion of the composition of C10 acid among the fatty acids synthesized by MEL, and similarly the composition of hexadecanoic acid will have a great impact on the composition of C8 acid in the synthesis of MEL. The relationship between linoleic acid and 4-decenoic acid implies that we can utilize natural vegetable oils rich in unsaturated polyfatty acids to produce MEL rich in unsaturated medium-chain fatty acids, and then obtain medium-chain fatty acids rich in unsaturated medium-chain fatty acids.

The results of linear correlation analysis showed that a few substrate and product linear correlation is very strong, which still has more correlation is low or non-linear correlation of the amount, so we fermentation data for random forest analysis, after learning the experimental data, to generate a matrix from 1 to 100, every 2% for a step, traversing all the composition of the substrate as a new input, and the output of the prediction results, and the results. The proportion of the composition of each medium-chain fatty acid in the product and the total yield of the product can be obtained for any substrate composition case, which can help to customize the production of medium-chain fatty acids in order to increase the production efficiency and reduce the cost of production and purification.

Conclusions:

In this module, we modeled the metabolic network of Aphanizomenon molestans and predicted the effect of gene knockout on the strain in order to reduce intracellular oil production and increase MEL fermentation yield. We used the constructed metabolic network to simulate the gene knockout, and then compared it with the experimental data, and obtained similar results, which proved that the metabolic network could reflect the actual situation of the strain to a certain extent, and provided a reference for the subsequent experiments to modify the strain. After that, we processed the data of MEL produced by fermentation, analyzed the relationship between its fermentation substrate and product species and established a prediction model to help how to ferment high-value and high-purity MEL products easily and efficiently.

1.Simulation of MEL fermentation process

Due to the characteristics of MEL, the semi-continuous fermentation method we adopted, firstly, the mixed fermentation is carried out in fermenter 1, and the extracted fermentation broth is settled in fermenter 2. Due to the mixing of sugar in the MEL, it has the highest density, and it will be settled to the bottom, and the supernatant will be refluxed back to the fermentation fermenter 1, and the MEL is diverted to the fermenter 3. In the end, fermenter 3 is the initially separated MEL product, and the fermentation can be maintained all the way in the fermenter 1. The fermentation state can be maintained in fermenter 1.



Assumptions:
(1) The bacteria in fermenter 2 die at the same rate as the bacteria in fermenter 1 without growing, producing products, or producing intracellular oils.
(2) All of the MEL in fermenter 1 is separated into fermenter 2, and 30% of s and x are separated into fermenter 2.
(3) The next separation is carried out when the previous reflux and precipitation are just about complete.




During fermentation in fermenter 1:

s_1=s0+si-rs*t

u=(umax*s)/(Ks+s)

x_1=x0*e^(μ*t)-x_d*x_1*t

rs=(μ*x_1/ Yx_s)+(rmel/ Ymel_s)+(roil/Yoil_s) +b

MEL_1=rmel*x*t

Oil=roil*x_1*t – roil_d *Oil*t

Separation:

MEL_2=MEL_2 + MEL

Changes in fermenter 2

x_2=x_1_f-x_d*x_2*t

s_2=s_2-(x_2/ Yx_s)*t

MEL_2c=0+K*MEL_2*t

returning flow:

MEL_3=MEL_3_0 +MEL_2c

MEL_1=MEL_1+MEL_2

s_1=s_1+s_2

x_1=x_1+x_2



From the simulation results, we can see that the growth of the bacteria on the reagent is mainly in fermenter 1, and the bacteria in fermenter 2 will die quickly after consuming the substrate due to the lack of continuous substrate input, so the significance of refluxing the supernatant to fermenter 1 needs to be further determined to see if this step can be omitted. Overall, the model can help us simulate the actual fermentation process and thus adjust the values of the inputs as well as the time of isolation to achieve a maximum MEL yield.

[1] Sorokina, O., Kapus, A., Terecskei, K. et al. A switchable light-input, light-output system modelled and constructed in yeast. J Biol Eng 3, 15 (2009).

[2] Chen X, Zaro JL, Shen WC. Fusion protein linkers: property, design and functionality. Adv Drug Deliv Rev. 2013 Oct;65(10):1357-69. doi: 10.1016/j.addr.2012.09.039. Epub 2012 Sep 29.

[3] Beck A, Vogt F, Hägele L, Rupp S, Zibek S. Optimization and Kinetic Modeling of a Fed-Batch Fermentation for Mannosylerythritol Lipids (MEL) Production With Moesziomyces aphidis. Front Bioeng Biotechnol. 2022 May 17;10:913362.

[4] Zakeri B, Fierer JO, Celik E, Chittock EC, Schwarz-Linek U, Moy VT, Howarth M. Peptide tag forming a rapid covalent bond to a protein, through engineering a bacterial adhesin. Proc Natl Acad Sci U S A. 2012 Mar 20;109(12):E690-7.