Team:XJTLU-CHINA/Model

Modeling



Bioinformatics methods play irreplaceable roles in biology science explorations nowadays. Various theoretical models can be used to simulate, construct, and predict the evolutions of microworld in a visualized way, thus assisting scientists with better understanding of biological processes.


In our project, the design of ExoCar is an extensive of last year's project and we made a considerable progress. We found an ideal protein EAAT2 that may contribute to effective treatment of neurodegenerative diseases. Additionally, a 3 plasmids regulation system is designed to facilitate and improve the exosome yield in the primary procedure of drug production. Among these designs, model is supposed to be the core and linkage of the whole story.


Model & project


You may ask: “Why do you so emphasize the importance of modeling? "

Modeling is one of the axes of the project of XJTLU-CHINA iGEM, 2019. This year, our models not only anticipate the results of our core experiments, and simplify the task of exploring experimental conditions, but also confirm the feasibility and rationality of our project design. The project design now we present on our wiki was adopted and implemented after our modeling results shows enough consistency with our expectations.

  • EAAT2 Functional Verification model (EFV model)

    To be Specific, when we found EAAT2 protein as a potential candidate to used in our project, we had no idea whether this protein can be expressed in our cell lines and intake excessive L-glutamate in culture medium with high efficiency. After building our EAAT2 Function Verification (EFV) model with the help of Matlab and online literatures, we predicted that by utilizing EAAT2 protein as the tool, we could achieve our purpose since our results provided a positive feedback for our idea, which suggested the EAAT2-induced L-glutamate intake rate can reach up to our expectation. Furthermore, from our model, we also predicted that the rate of L-glutamate rate was in the unit of μmol/min. This detail greatly reduced the work of exploration of experiments condition, shortened the experiment duration and ensured the progress of the project.

  • Riboswitch Optimization model (RO model)

    Besides the attribution to project evaluation and implementation, our models also facilitate us to design the part which we expected to improve and provide to the whole iGEM community. We use thermodynamic model to predict the translation strength of different Ribosomal Binding Site (RBS) regions when they are connected with our theophylline riboswitch sequence. By this way, we perfectly improved our reporting part, enabled it with higher sensitivity to its ligand, made it more detectable and stabilized its reporting function.

  • 3 Plasmids Regulation System model (PRS model)

    Not only for this year’s project, our models also achieved a long-term impact on the design of our project in the future. Specifically, our 3 plasmids regulation system (PRS) model is built up to mimic the process of therapeutic exosome production. By building up this model, we successfully predict the producing condition when cells are used to generate exosome drugs. More importantly, we added regulation parts to control gene expression intracellularly. This model approves our project design about cell production and can be extracted as a method to build up engineered cell lines. Our model provides guidance and some predicted details for cellular precisely-control production and it may have an impact on pharmaceutical industries.


Figure 1. Interactions of project design with modeling & experiment with modeling


Model Blocks


The main models we make for this project can be divided into such 3 parts


  • EAAT2 Functional Verification model (EFV model)


  • 3 Plasmids Regulation System model (PRS model)


  • Riboswitch Optimization model (RO model)


EAAT2 Functional Verification model
(EFV model)


Why this model?


When EAAT2 was suggested to be our treatment RNA drug, which was not involved in last year’s project. There was coming a problem that how the function of EAAT2 improves and qualifies glutamate absorption into the neural system. Therefore, we design experimentsto measure the glutamate absorbing ability of EAAT2 under certain condition.

Before operating the wet experiment, EAAT2 functional verification model (EFV model) was built to preverify the function of EAAT2. At the same time, details of wet experiment was modified to adapt to the real condition under the guidance of this model.


Figure 2. EAAT2 proteins absorb L-glutamate in N2A extracellular environment.


Methods


This model is a multi-step mathematical model based on Central Dogma Principle, and will predict the overall glutamate absorbing ability of N2A cell under our experimental condition. Following the gene expression process, we divide this model into 5 subparts, including plasmids transfection, gene transcription, mRNA translation, EAAT2 protein translocation and glutamate absorption. Each of the steps is described by the specific mathematic and statistic method.

However, Due to high level complexity of cell physiology. We make following assumptions to optimize and simplify the process.

Assumption
  • Refer to literatures and experimental design, we assume that the cell physiology has reached a relatively steady state, thus the number of EAAT2 proteins will keep in a fixed level when the absorption of glutamate begins.

  • Once plasmids are transfectedd into cells, they will be transcribed soon and then mRNAs will be generated. Simultaneously, mRNAs will start decay when they enter into cytoplasm. The generation and degradation of mRNA occur at the same time, with their own speed. Finally, the copy number of intracellular mRNA would approach to a constant at equilibrium.

  • Like mRNAs, protein translation and degradation also occur at the same time. However, based on our results, we predict that the time for protein to approach to a stable level is comparatively negligible. Therefore, we build our model mainly based on the number of proteins at the equilibrium stage.


Figure 3. Main procedure of EFV model.


1. Transfection

The first part of our model is transfection. To mimic cell transfection process in vitro experiments, the transfection efficiency can be expressed with a proportion constant F.


2. Transcription

Then in the process of transcription, mRNA of EAAT2 will be transcribed in cells once plasmids are transfected. Meanwhile, generated mRNAs are degrading simultaneously. Taken together the process of mRNA generation and degradation, according to assumption 1, overall number of mRNAs can be represented with a constant Nr.


3. Translation

The next step is translation, like the transcription, proteins will be produced and degrade at the same time in cells. Here, Law of Mass Action kinetics (LMA) is applied to describe this process.

Parameter Table

Parameter

Value

Unit

Description

Source

Nr

600

no unit

the mean number of mRNA per cell

[5]

R

122.3

s-1

translation rate

[8]

D

6*10-3

s-1

protein degradation rate

[1]

Variable Table

Variable

Unit

Description

P(t)

protein

the number of mRNA

t

second

time


With proper parameter, the model is solved by the ODE45 MATLAB function in file EFV_Trans_main.m and file EFV_Trans_function.m. The simulation results are indicated in the following figure.

Figure 4. The number of EAAT2 in N2A cell versus time. When the simulation exceeds 1000s, the number of EAAT2 reminds in a stable level Ne (~105)


4.Translocation

After translation, EAAT2 proteins will translocate to cell membrane. For this part, we read from online literature that most EAAT2 proteins will be engulfed into cell membrane immediately after translation. And only few proteins will be sent into proteasome and be degraded. Since protein translation and degradation have been taken into consideration in previous equation. We make a deduction that the number of membrane-valid EAAT2 equals to Ne (from Translation part figure 4).

Figure 5. The distribution of EAAT2 in astrocyte cell. This figure shows that most of the EAAT2 proteins are engulfed into cell membrane


5.Absorption

Finally, EAAT2 proteins anchored on cell membrane will absorb L-glutamate acid from extracellular medium. This process followes first-order Michaelis-Menten kinetics and Eadie-Hofstee transformation. Combined with the data from literatures, Vmax the maximum absorption rate of glutamate in our experimental system, can be calculated.

Parameter Table

Parameter

Value

Unit

Description

Source

Vmax

1.41

μmol min-1

maximum reaction rate

[2]

Kglum

52.7

μmol

km of EAAT2

[2]

Variable Table

Variable

Unit

Description

[Glu]

μmol

L-glutamate acid concentration in medium

Vabs

μmol min-1

glutamate absorption rate


Result


By running the equation with Matlab (EFV_Absor_main.m) , we get Figure 6 which predicts the absorption of L-glutamate acid in different concentration of glutamate in medium. When glutamate concentration approaches to Vmax, glutamate are absorbed in a comparative high speed. EAAT2-induced L-glutamate intake rate can reach up to the unit of μmol/min, which is considered to be efficiently enough for our project. In conclusion, induced up-regulated expression of EAAT2 protein can absorb glutamate significantly and we can achieve our therapeutic purpose by utilizing EAAT2 proteins.


Figure 6. The rate of EAAT2 absorbing extracellular glutamate versus extracellular glutamate concentration.


Finally, EAAT2 wet experiments can be performed with the insurance from modeling. Here, what should be mention is that an important adjustment is supported by EFV model. the initial concentration of glutamate used to treat N2A cell is adjust to 80μmol/L since this concentration mimic the real condition of early neurodegenerative diseases and also is a comparatively reasonable concentration for absorption test.

The result of EAAT2 wet experiment, as we expected, shows a high feasibility in absorbing glutamate. The absorption rate, consistent with our predictions, reaches up to the unit of μmol/min. More experiment details can be find by clicking this.


3 Plasmids Regulation System model
(PRS model)


Why this model?


Following the project from last year, we are facing a problem that how to engulf and pack the anchor protein and EAAT2-C/D Box mRNA effectively so the cell can produce the exosomes with therapeutic function as more as possible.

To solve this problem, we create to put riboswitches in front of each coding sequence on the plasmids. When being transcribed, those riboswitches can control the expression of proteins so the proteins and therapeutic mRNA can bind with high affinity and hopefully the production of therapeutic exosomes will be increased.


About 3 plasmids regulation system


Based on the last year model, we build up a genetic circuit which contains three riboswitches and 2 RNA-binding proteins to control the expression level of different proteins and save energies for cells. These elements are located on different locus of 3 plasmids and we named these e plasmids as GC1, GC2 and GC3. The map of these 3 genetic circuits are shown as below:

Figure 7. 3 plasmids regulation system genetic circuits graph.

MS2 riboswitch stat “OFF” and MS2 riboswitch state “On”) and proteins (MS2 and PP7) are designed aims at building up genetic circuits in cells to regulate the expression of gene(s) with different function. All coding proteins are connected by 2A peptides which are not shown in graph.


The mechanism of these 3 GCs (genetic circuits) are as below.

  • Initially when these 3 GCs are transcribed, the riboswitch state on GC1 is open and Proteins on GC1 can be translated. On GC2, initial state of MS2 riboswitch is off so the proteins on GC2 cannot be translated.

  • However, once MS2 proteins is translated from GC1, they will specifically bind with MS2 riboswitch on GC2 and GC3. Switch on GC2 will be turned on and translation is initiated. Subsequently, When PP7 proteins are translated, they will bind with PP7 riboswitch which provides a negative feedback to GC1 so the overall production of MS2 proteins will be inhibited. In term, PP7 proteins production will also decrease as MS2 protein level lowers down.

  • By building these genetic circuits which form a negative feedback, the expression of booster proteins and membrane proteins can stay in a stable condition at equilibrium. As for GC3, it also contains a MS2 riboswitch. However, different from GC2, the initial state of its MS2 switch is On. When MS2 binds with them, translation of GC3 is inhibited and this will save cell’s energy for protein production and more EAAT2 coding mRNAs can be packaged into exosomes and be carried to pathological cells.


Methods


We establish differential equation system to describe the number change of substance in the 3 plasmids circuit with time change, and utilize ODE45 algorithm in MATLAB to solve them. These two files run in Matlab: PRS_Method_main.m and PRS_Method_function.m .

Before talking about the equation, we raise reasonable assumptions to simplify and optimize the whole system.

  • The process of regulation protein (MS2 & PP7) binding to riboswitch is treated as a second-order reaction. Moreover, the binding process is ignored from the reduce of regulation proteins. This is because the consumption of regulation proteins by degradation is far beyond the consumption by protein-RNA binding.

Equations for 3 plasmids regulation system

Varible Table

Varible

Unit

Description

n1

mRNA

the number of mRNA1on

n2

protein

the number of MS2

n3

mRNA

the number of mRNA2off

n4

mRNA

the number of mRNA2on

n5

protein

the number of PP7

n6

mRNA

the number of mRNA3on

n7

mRNA

the number of mRNA3off

n8

mRNA

the number of mRNA1off

Parameter Table Of Regulation Model

Parameter

Value

Unit

Description

Source

dmRNA1on

0.104

min-1

cleavage rate of dmRNA1on

[4]

dmRNA2on

0.708

min-1

cleavage rate of dmRNA2on

[6]

dmRNA3on

0.197

min-1

cleavage rate of dmRNA3on

[6]

dmRNA1off

0.085

min-1

cleavage rate of dmRNA1off

[4]

dmRNA2off

0.073

min-1

cleavage rate of dmRNA2off

[6]

dmRNA3off

0.015

min-1

cleavage rate of dmRNA3off

[6]

dMS2

1.67*10-3

peptide chain*min-1

degradation rate of MS2

[1]

dPP7

1.67*10-3

peptide chain*min-1

degradation rate of PP7

[1]

pMS2

27.660

min-1

translation rate of MS2

[8]

pPP7

11.277

min-1

translation rate of PP7

[8]

Cr1

14.400

min-1

transcription rate of mRNA1

[7]

Cr2

0.144

min-1

transcription rate of mRNA2

[7]

Cr3

2.400

min-1

transcription rate of mRNA3

[7]

k1

6*10-6

no unit

constant of the second order reaction MS2 and mRNA2off

[4]

k2

7.22*10-6

no unit

constant of the second order reaction MS2 and mRNA3on

[6]

k3

6.8*10-6

no unit

constant of the second order reaction PP7 and mRNA1on

[6]


Results


The PRS model indicates 2 stages after 3 plasmids are transfected into mammalian cell lines (HEK293T):

  • At the beginning, booster proteins will initially be translated more than membrane proteins. The number of On-state mRNA3s (EAAT2/other therapeutic mRNA) will raise up in an extremely short time and then the number will turn to be flat. Meanwhile, Off-state mRNA3s(closed by produced MS2 proteins) begin to generate and accumulate consistently.

  • With the accumulation of MS2 proteins in cells, the production of PP7 protein begins to burst because MS2 turns on the OFF-state MS2 riboswitches and translation is initiated. Once when PP7 concentration (produced together with booster protein) exceeds MS2 concentration (produced together with membrane protein), engineered cells will begin to produce more membrane proteins. Almost at the same time, Off-state mRNA3s are largely accumulated in cells and the number of untranslated mRNA3s begin to be flat (MS2-riboswitch complexes prohibit these mRNAs to be translated). Soon the system reaches to equilibrium.


Figure 8. The number of two regulation protein (MS2 & PP7) and two state of mRNA3 change with time (in HEK293T cell).


Our in silico results have achieved our intended purpose:

This model mimics the real condition when we apply the 3 plasmids regulation system in host human cells to produce therapeutic exosomes in the future.

We theoretically achieved our purpose to predict the designed regulation system in mammalian (HEK293T) cell lines according to the results we received from PRS model. Exosomes released from equilibrium-stage cells are predicted have a higher possibility to be full-functional. The membrane proteins are highly-produced, the exosomes are generated in a lower rate. Furthermore, more mRNA3s have accumulated in cells. This condition may increase the abundance of membrane protein and mRNA3s in exosomes because cells provide theoretically better condition for mRNA packaging. Also, accumulated Off-state mRNA3s (EAAT2 coding sequence/therapeutic mRNA) will not be translated in host cells, which preserve more mRNA3s to be expressed in pathological cells. The turning-off of mRNA3s will also save energy for host cells to survive and replicate. This may increase the production of therapeutic exosomes in another way.

Above all, we predicted that therapeutic exosomes production in regulated cell lines are possibly achievable, based on our modeling results. Engineered Cells perhaps can be regulated and controlled for producing certain products, theoretically. This method has the potential to be utilized in the production of many drugs since from the results we received, the yields of cell products probably can be quantitively-controlled. Thus, we consider our model as a theoretically-supported example to provide guidance and predictions for future projects related to drug manufacture.


Riboswitch Optimization Model
(RO model)


Why this model?


This year, after our PRS model was built, we realized the great potential of utilizing regulation system in cells to manipulate cell activity. Since we had already designed the genetic circuit in PRS model, we had some background knowledge about ribozyme switch, so we wanted to optimize switches shared with similar mechanism to provide new switches with better functionality for iGEM communities.

Therefore, we search the whole part list and we find in 2011, PKU-R used to submit a part (coded as BBa_K598009) which is also a ribozyme switch. As a theophylline- responsive ribozyme switch, it is applied in E. Coli to test the function of the ribozyme sequence. We decide to optimize this part by utilizing bioinformatic method to predict the results of our designed improvement


Methods


1.Original part design:

The original part from PKU-R 2011 (BBa_K598009) is a theophylline-responsive ribozyme switch. To test the function of that sequence, they place a pBAD promoter at the upstream of the sequence and put an eGFP protein sequence at the downstream of the theophylline sequence. Inside the theophylline sequence, there is a Ribosomal binding site(RBS) region.

The whole sequence will form a unique secondary sequence after mRNAs are transcribed by the induration of L-arabinose. When theophylline is absent, the formed specific base-pair structure will hind the RBS region and the translation initiation process is inhibited. At the presence of theophylline, the secondary structure will undergo transformation and the RBS region is exposed, which means the ribosomes can bind with RBS and translation is initiated.

2.Our improvement:

Based on the mechanism of ribozyme-constructed switch, we found some switch sequences with similar ribozyme homologous regions through literature research. Some of these sequences show higher affinity to theophylline and the expression eGFP fluorescence intensity is higher than BBa_K598009. However, we still found that the expression of eGFP protein is limited, to accordance with these literatures. We wanted to improve the expression of the eGFP protein which enables the theophylline switch with higher sensitivity to its ligand. We utilized the algorithm and equation from Suess et al. (2004) to improve the RBS of BBa_K598009. Also, we replace the ribozyme region on BBa_K598009 with a new region which is read to have a higher sensitivity to theophylline, according to literatures. Finally, we got one candidate and we calculated its predicted translation level based on the equation:

This equation is derived from Suess et al. (2004). ΔGtot represents the predicted expression level. When ΔGtot is smaller, the tested RBS is predicted to have a higher expression in cells.

  • ΔGmRNA: rRNA: represents the energy released when the last 15 nucleotides of 16S ribosomal rRNA binds with 15 nucleotides exactly at the upstream of RBS.

  • ΔGstart: the energy released when start codon and anti-start codon binds.

  • ΔGspacing: the minimum energy required to deform the secondary structure of the gap sequence between RBS and downstream eGFP protein.

  • ΔGstandby: the minimum energy required to deform the secondary structure of the RBS and the 4 nucleotides at its the upstream.

  • ΔGmRNA: the energy needed to deform all secondary structure of the 35 nucleotides before and after the RBS (containing the start codon).

According to Suess et al. (2004), this equation can predict the translation strength with high accuracy. Therefore, we utilized it and calculated the ΔGtot energy of our candidate. We got ΔGmRNA:rRNA, ΔGstart, ΔGspacing from Suess’ thesis and we calculated the ΔGStandby and ΔGmRNA through applying NuBACK.

We also used the same equation and method to calculate the ΔGtot energy of the original part BBa_K598009, as a comparison to our improved candidate. After getting the results, we conpared them on the table below.


Result


Name

ΔGmRNA:rRNA

ΔGstart

ΔGspacing

ΔGstandby

ΔGmRNA

ΔGtot

BBa_K598009 (From PKU-R)

-6.2

-1.194

0.67

-1.6

-6.5

1.376

BBa_K3030017 (We improved)

-6.8

-1.194

0

-0.1

-0.1

-1.594

All numbers shown above are in the unit of Kcal/mol.

As the results shown above, the ΔGtot value of our improved switch is -1.594, and the original part BBa_K598009 has a ΔGtot value of 1.376. According to Suess at al. (2004), the value of the ΔGtot has an almost linear correlation with the translation strength. When ΔGtot is smaller, the tested RBS is predicted to have a higher expression in E. Coli. Based on results we received in silico, we believe that our candidate will perform with a higher sensitivity and expression level than BBa_K598009.


Experiment Validation


Based on our prediction, we believe that the sequence we designed is highly possible to have a bet-ter performance on sensitivity and expression intensity than the original part BBa_K598009 from PKU-R 2011. Then we sent our designed plasmids construction map to synthesize in company and later experimental group tested the function of our improved part BBa_K3030017 as well as the orig-inal part BBa_K598009.

The experiment results show that consistent with our modeling prediction, our improved part shows significant improvements of ligand-sensitivity and eGFP protein expression intensity, compared with original part.

Firstly, the eGFP fluorescent intensity/ OD600 level varies significantly. Our part has much greater eGFP fluorescent intensity/OD600 level than BBa_K598009. This indicates that our improved part is more sensible to low theophylline concentration than original part. Secondly, compared with BBa_K598009, the eGFP fluorescent intensity/OD600 level of our part also differentiates within theophylline gradients more significantly



Above all, we consider that our model facilitates us to improve theophylline-sensitive ribozyme switch successfully. We use thermodynamic model to estimate the expression improvement and the results is highly consistent with our expectations. Our improved part shows higher expression level of eGFP protein under the same concentration of theophylline and arabinose. This provides our improved part with higher sensitivity and accuracy.

For more details about the experimental details of our improved parts BBa_K3030017 and previous part BBa_K598009, please go to BBa_K3030017 and look further.


References


[1] Wu,B. et.al (2016) 'Translation dynamics of single mRNAs in live cells and neurons', Science (New York, N.Y.), vol. 352,6292 (2016): 1430-5. doi:10.1126/science.aaf1084 [Online]. Available from:
https://www.ncbi.nlm.nih.gov/ pmc/articles/PMC4939616/ #SD15



[2] August Krogh Institute. et.al (2002) '26 S proteasomes function as stable entities', J Mol Biol. 2002 Jan 25;315(4):627-36 [Online]. Available from:
https://www.ncbi.nlm.nih.gov/ pubmed/11812135? dopt=Abstract



[3] Dunlop,J et.al (1999) 'Inducible expression and pharmacology of the human excitatory amino acid transporter 2 subtype of L-glutamate transporter', Br J Pharmacol. 1999 Dec; 128(7): 1485–1490. [Online]. Available from:
https://www.ncbi.nlm.nih.gov/ pmc/articles/PMC1571787/



[4] Stenovec, M. et.al (2008) 'EAAT2 density at the astrocyte plasma membrane and Ca(2 + )-regulated exocytosis', Mol Membr Biol, 2008 Apr;25(3):203-15. [Online]. Available from:
https:// www.ncbi.nlm.nih .gov/pmc/articles/ PMC4231745/ bin/ supp_42_19_123 06__index.html



[5] Lim,F & Peabody,D (2002) 'RNA recognition site of PP7 coat protein', Nucleic Acids Res, 2002 Oct 1; 30(19): 4138–4144. [Online]. Available from:
https:// www.ncbi.nlm.nih .gov/pmc/articles/ PMC140551/



[6] Swinburne IA & Silver PA (2008) 'Intron delays and transcriptional timing during development', Dev Cell., 2008 Mar;14(3):324-30. [Online]. Available from:
https://www.ncbi.nlm.nih.g ov/pubmed/18331713? dopt=Abstract



[7] Kennedy, A. B., Vowles, J. V., d'Espaux, L., & Smolke, C. D. (2014). 'Protein-responsive ribozyme switches in eukaryotic cells', Nucleic acids research, 42(19), 12306–12321. [Online]. Available from:
https:// www.ncbi.nlm.nih.g ov/pubmed/18331713? dopt=Abstract



[8] Marcinowski, L. et.al (2012). 'Real-time Transcriptional Profiling of Cellular and Viral Gene Expression during Lytic Cytomegalovirus Infection', Real-time Transcriptional Profiling of Cellular and Viral Gene Expression during Lytic Cytomegalovirus Infection., PLoS Pathog 8(9): e1002908. [Online]. Available from :
https://journals.plos.org/plospathogens/article?id=10.1371/journal.ppat.1002908