Team:NUS Singapore/Model P

Awards Check

Measurement Modelling Medal Check List

Team

Team Member Attributions

Collaboration

Collaboration

Human Practice

Silver Up Gold Integrated

Modelling

Overview Methodology Kill Switch for Probiotics Kill Switch for BeeT

Projects

Description Interlab Study Parts Experiment Documentation Improvement Demonstrate Safety Future Works

Home

Back to Top

Kill Switch for Engineered Probiotics

Introduction

Why a kill switch for Engineered Probiotics?

Advancements in engineered probiotics have spurred a new generation of novel medical therapeutics and diagnostic solutions. Notable uses of this emerging technology include scientists using engineered microbes to successfully tumours in the gastrointestinal tract with minimal collateral damage. Yet, despite the developments in engineered probiotic medicine, the risk of having these engineered microbes leak into a non-designated environment remains an unresolved threat: a threat that could disrupt ecosystems and endanger patients [1]. Therefore, critical to the feasibility of a functional and commercial engineered probiotic are control systems called kill switches that allow for the spatial and temporal killing of the engineered microbe once removed from its specific target site. Although in recent years kill switches have undergone significant improvement, the feasibility of many existing kill switches in this field are limited because they rely on:

1. The limited specificity of a single input biosensor [2];
2. Artificial amino acids: chemicals that are often difficult and unsuitable for implementation in the human body [3]; and
3. A lack of measures that fail to protect against mutagenesis, leakage and system failure [4, 5, 6, 7]. Ultimately, since no kill switches have passed the standards set by regulatory environmental and, food and drug bodies, the vision of accessible consumer biotherapeutic products such as engineered probiotics will remain just that, a vision [8, 9].

We aim to improve current kill switches to meet such regulatory standards, starting with making the engineering of customised kill switches for engineered probiotics easier and more accessible.

For our full report, please click here.

Key findings

1. We have used modelling to study two kill switch designs, one generated automatically by CELLO (Automated Circuit) and one design by the team (Simple Circuit). The performance of kill-switch designs was evaluated based on the following criteria: number of parts required in a design, metabolic burden on cell, and response/lag time. Results showed that the kill-switch design by the team has faster response and lower metabolic burden. Hence, we proceeded to build and perform experiments with this design. Modelling helps!

2. The model has helped us to gain insights into how we can further improve the phosphate-temperature sensor. See below for more details.

Design and Specifications

Input Sensors

The Engineered Probiotic Kill Switch uses a Phosphate Promoter (BBa_K116401) and a Temperature Promoter (plTpA36). With more sensors, the user can increase specificity and develop more complex logic.

We choose to use these two sensors because when used together, they offer sensitivity between all the environments involved, and therefore allow controlled activation for killing the engineered probiotic in the correct wastewater conditions. (See Figure 1)

Output Killing Mechanism

To kill the engineered probiotic, we employ use of the E2 Toxin that is standardised in our proposed approach. We use the E2 Toxin because it can degrade Host DNA, whilst posing little threat to the gut microbiota.

We control killing by controlling the expression of the anti-toxin, IM2. With little or no expression of the IM2 anti-toxin, the constitutively expressed E2 toxin will kill the engineered probiotic.

State Diagram

It is important to consider the application lifecycle during the design stage. For our engineered probiotics application, we expect the engineered probiotic to travel through four different states.

1. In the Transportation stage, the engineered probiotic is stored in its packaging conditions, such that it is ready to be transported or consumed. It is stored in a high phosphate solution. The temperature in this condition is irrelevant.
2. In the Ingestion Stage, the engineered probiotic resides in the intestines. In the intestines, the probiotic is in a high phosphate concentration and high temperature (body temperature) condition.
3. In the Colon Stage, the engineered probiotic resides in the Colon. In the Colon stage, much of the phosphate has been absorbed by the small intestine. Hence, the engineered probiotic is in a low phosphate and high temperature condition (body temperature).
4. In the Wastewater Stage, or the Excretion stage, the probiotic exists in wastewater and has been exposed to outside low temperatures. The low phosphate and low temperature environment triggers activation of the killing mechanism.

Figure 1 The State Diagram of a kill switch for an engineered probiotic in application. Phosphate = 1 refers to a high concentration of phosphate, Phosphate = 0 refers to a low concentration of phosphate. Temperature = 1 refers to greater than 36°C conditions, Temperature = 0 refers to lower than 36°C conditions, Temperature = X refers to any temperature condition.

Timing Diagram

The Timing Diagram is a graphical representation of the different states. Both logic circuits will be following the timing diagram such that in the final Low Temperature and Low Phosphate stage, the killing mechanism will activate.

Figure 2 Timing Diagram of the kill switch for engineered probiotics. In the final state we want to stop producing antitoxin IM2 so we can kill the engineered probiotic

Logic Table

Depending on the method (Simple Circuit or Automated Circuit) used to implement the genetic circuit , the logic used may slightly vary. Despite these variations in logic, both circuits will activate the killing mechanism in the same stage.

Figure 3 Logic Table of the kill switch for engineered probiotics. Because of the different circuit designs, the logic table for the Simple Circuit and Automated Circuit slightly varies. However, both circuits obey the timing diagram and logic table and function the same way.

Simple Circuit

Genetic Circuit

The Simple Circuit design can be represented two biosensors cascaded together.This cascade mechanism refers to the Phosphate Sensor, pPhoB, controlling the repressor TlpA36. When the repressor TlpA36 is in a condition below 36 °C, it will repress the pTlpA36 Promoter. When the pTlpA36 promoter is repressed, anti-toxin IM2 production will decrease, and ultimately the probiotic will die. Since the pPhoB downstream expressions are linked with pTlpA downstream expressions, the system is said to be cascaded.

To generate the model of the genetic circuit we need to transfer the genetic circuit to a dynamic modelling software such as AdvanceSyn, a local startup providing design and modelling software for SynBio design. Using AdvanceSyn we can generate a functional model, apply sensitivity analysis and combinatorial analysis as well as implement our timing states into the model. The genetic circuit as displayed in AdvanceSyn is shown in Figure 4.

Figure 4 The genetic circuit for the Simple Circuit Design. (Below) The Simple Circuit as displayed in the AdvanceSyn Studio.

Functional Model

Simulating circuit for a total duration of 50 hours at timestep = 300s, with timing states implemented.
Creating the functional model is one of the most important steps when modelling a genetic circuit. This is because the functional model serves as a proof of concept and acts as a target for achieving the ideal performance of the genetic circuit. To generate the functional model for the Simple Circuit, we reduced the expression of toxin E2 by 20% and increased the binding affinity of the temperature promoter. The effect of these changes allows for us to generate the functional model.

Figure 5 The functional model illustrates the ideal performance of the Simple Circuit in all timing states. This serves as a proof of a concept that the Simple Circuit can successfully be used as a kill switch for engineered probiotics!

How can we aid the experimenter?

We first applied a sensitivity analysis to determine which kinetic values had the largest effect on the output response. Our sensitivity analysis determined that the binding affinity, and promoter strength of the temperature promoter, and the promoter strength of phosphate promoter were the most sensitive parts of the circuit. As a result, because the phosphate sensor by NYMU Taipei in 2008 (which we intended to use) was very weak, we decided to improve this part by increasing its expression level and range (more details at improvement page).

• Need to be mindful about phosphate sensor expression level


Figure 6 Low phopshate promoter strength delays the killing time and hence, renders the kill switch ineffective.



Figure 7 Decreasing the promoter strength of the phosphate promoter reduces the effect that leaky TlpA has on the output response, anti-toxin IM2

Figure 6 shows that if the Km of the temperature sensor is high (more repressors are required to repress the sensor), the circuit will not work if the phosphate sensor is weak (i.e., IM2 will not be repressed). This also suggests that we need to improve the phosphate sensor expression by NYMU Taipei in 2008. Further analysis using the model also revealed that we need to be mindful about not having too strong a phosphate promoter (figure 7) which might result in unintended leakiness and causing the circuit to malfunction, assuming that the temperature promoter has a low binding affinity.

• Dealing with a Strong Toxin


Figure 8 If E2 toxin is too strong, we recommend increasing the strength of pTlpA. Increasing pTlpA increases IM2 expression. However, if IM2 is produced in excess, this can cause increased metabolic stress for the host.



Figure 9 Ideal promoter strength for effective and reliable killing.

Another potential problem that can be elucidated during the construction of the circuit is the balancing of toxin and antitoxin expressions. In the genetic circuit, the toxin E2, is controlled by a constitutive promoter while the antitoxin IM2 is controlled under the temperature promoter. It is important to balance the expression of these molecules because 1. If there is more toxin than antitoxin, the probiotic will die in all stages; and, 2. If there is an excess of antitoxin expression, it places unnecessary metabolic stress on the probiotic host. To examine this issue, we modelled antitoxin IM2 production across different temperature promoter strengths. Assuming 1:1 binding, we can observe that low temperature promoter strengths do not produce enough antitoxin IM2 to keep the probiotic alive in all stages. On the other extreme, a very high promoter strength produces an excess of antitoxin IM2 and places unnecessary stress on the probiotic. Therefore, using modelling, we conclude that an ideal promoter strength is one which allows for the production of antitoxin IM2 to be slightly greater than that of E2 toxin production (See Figure 8). If the lab realises that the probiotic keeps on dying because of the strong toxin, we can propose increasing the promoter strength of the temperature promoter accordingly.

Automated Circuit

Genetic Circuit

Simulate the circuit for a total duration of 50 hours at timestep = 300s with timing states implemented.
To create the genetic circuit, we first translate the logic table proposed earlier to Verilog code. We then enter the Verilog code in the CAD software, Cello, which will then generate genetic circuits that fit the logic table. The genetic circuit generated by Cello is shown in Figure 10. We the then transfer the genetic circuit generated from cello and transfer it to the dynamic modelling software AdvanceSyn Studio. From AdvanceSyn we can generate the functional model, apply sensitivity analysis, apply combinatorial analysis and implement timing states.

Figure 10 The Automated Circuit generated by Cello CAD.

Figure 11 The Automated Circuit displayed in AdvanceSyn Studio

Functional Model

The functional model generated from Cello and modelled in AdvanceSyn fits the timing diagram most accordingly. The only changes made to generate the functional model required increasing the binding affinity of the temperature promoter. It should be noted that in the Automated Circuit functional model, the repression of antitoxin IM2 is significantly slower compared to the Simple Circuit Functional model. The reason there is a delay in killing is due to propagation delay. In the Automated Circuit, propagation delay occurs because of the complexity of the circuit. With each added part and its respective interaction, the time taken for changes in the environment to be recognised in the output response increases because of the increased processes. In other words, the more complex the genetic circuit is, the longer it takes to generate output. To rectify this issue, it is possible to increase the promoter strength and RBS of the parts in the circuit, however, this comes at the risk of increasing metabolic stress.
For more information about the complete model of the Automated Circuit including sensitivity analysis, and optimisation of the Automated Circuit, click here.

Figure 12 The Functional Model of the Automated Circuit. Note the time taken to repress IM2 in Wastewater (45-50hrs) is slower than in the Simple Circuit.

Comparison of Simple Circuit & Automated Circuit

Comparing the two designs, we can observe that both have their own advantages and disadvantages. The Automated Circuit offers increased response control, modularity, and buffering against some mutations and error. The Automated Circuit offers increased response control because each of its input sensors is independent of each other. Therefore, a user can not only develop complex logic to control output, but can also rely less on ensuring that all sensors work 100% since the output can be reliant on multiple input sensors. For example, unlike the Simple Circuit where the input promoters are cascaded, in the automated circuit, the input promoters are independent and therefore in the case of a leaky promoter, the effect it has on the output response is less in Automated Circuit than in the Simple Circuit. In addition, the Automated Circuit offers modularity because a user can simply replace the input sensors with other input sensors and make slight modifications to the kinetics involved. However, in the Simple Circuit, a user were to use different sensors, they may need to overhaul the genetic circuit design depending on the promoter types. Finally, the automated circuit offers buffering against some mutations and small errors. Because of the intermediate processes involved between input and output, some small errors caused by mutation are not significant in changing the output response of the genetic circuit. On the other hand, if mutations occurred in the Simple Circuit, because of the cascaded configuration and high dependency on parts, errors will have a large effect on the response output and can potentially cause failure.

However, there are limitations that exist in the Automated Circuit that are made up for in the Simple Circuit. One of the biggest drawbacks in the Automated Circuit is complexity. With increasing complexity comes issues surrounding propagation delay, metabolic stress, and construction difficulty. For example, when compared to the functional model of the Automated Circuit, the functional model of the Simple Circuit performed 33% faster killing thanks to its simple design. However, as mentioned earlier, because of its simple design, the Simple Circuit sacrifices modularity and control for faster response and less metabolic stress. This trade-off, as shown through modelling, is a reasonable justification as to why we construct the simple model in the lab. After all, we aim to generate a kill switch as a secondary function that can kill the probiotic host quickly and effectively too. Therefore, we can propose that for simple kill switches (less than two input sensors), a user can employ the Simple Circuit design, and for more complex systems (more than two input sensors), a user can employ the Automated Circuit design.

Conclusion

Using our modelling workflow, we successfully demonstrate in silico how a customised genetic circuit for engineered probiotics can be made easier. Our results from modelling can be used by the experimenter to guide them during construction. We have mentioned examples of using modelling to recommend solutions to problems that may be faced during the construction of the kill switch. Making the engineering of customised kill switches for probiotics easier enables anyone to use their own biosensors to develop their own kill switch for engineered probiotics. We hope that with the proliferation of different sensors and killing mechanisms, anyone will be able to utilise our workflow to develop kill switches and in turn, meet regulatory standards.

Reference

• [1] A. Maxmen, “Living therapeutics: Scientists genetically modify bacteria to deliver drugs,” Nature Medicine, vol. 23, no. 1, pp. 5-7, 2017.
• [2 ]D. I. Piraner, M. H. Abedi, B. A. Moser, A. Lee-Gosselin and M. G. Shapiro, “Tunable thermal bioswitches for in vivo control of microbial therapeutics,” Nature Chemical Biology, 2017.
• [3] D. Mandell, M. Lajoie, M. Mee, R. Takeuchi, G. Kuznetsov, J. Norville, C. Gregg, B. Stoddard and G. Church, “Biocontainment of genetically modified organisms by synthetic protein design,” Nature, vol. 518, pp. 55-60, 2015.
• [4 ]F. Stirling, L. Bitzan, J. Oliver, J. Way and P. Silver, “Rational Design of Evolutionarily Stable Microbial Kill Switches,” 21 04 2017. [Online]. Available: http://www.biorxiv.org/content/early/2017/04/21/129445.article-info.
• [5] B. Torres, S. Jaenecke, K. Timmis, J. Garcia and E. Diaz, “A dual lethal system to enhance containment of recombinant micro-organisms,” Microbiology, vol. 149, pp. 3595-3601, 2003.
• [6] C. T. Y. Chan, W. L. Jeong, D. E. Cameron, C. J. Bashor and J. J. Collins, “‘Deadman’ and ‘Passcode’ microbial kill switches,” Nature Chemical Biology, vol. 12, pp. 82-86, 2016.
• [7] G. H. Moe-Behrens, R. Davis and K. A. Haynes, “Preparing synthetic biology for the world,” Frontiers in Microbiology, vol. 4, p. 5, 2013.
• [8] G. N. Mandel and G. E. Marchant, “The Living Regulatory Challenges of Synthetic Biology,” Iowa Law Review, vol. 100, p. 155, 2014.
• [9] S. R. Carter, R. Michael, S. M. Garfinkel and R. M. Friedman, “Synthetic Biology and the U.S Biotechnology Regulatory System: Challenges and Options,” J. Craig Venter Institute, 2014.