Establish the Problem
Cybergenetics
Problem Evaluation (Silver)
Integrated Human Practices (Gold)
Implementation
Future Developments
Dr Corentin Briat
Dr Stephanie Aoki
Dr Fulvio Forni
Dr Timothy Frei
Dr Ania-Ariadna Baetica
Dr Noah Olsman
Dr Lucia Marucci and Mr Davide Salzano
Evonetix
Codexis
Cambrium
Biosyntia
References
Summary
Summary (Silver): Through a series of interviews from academics and industry experts we established that our project was valuable and responsible.
Values of project: Our ideation process considered principles of effective altruism to assess what would produce the most amount of good and effect the most amount of people whilst being feasible. We decided to target the challenge of genetic circuit robustness. In addition to scientific, other principles such as medical, industrial or environmental were also important.
Evidence that the project is responsible: To ascertain that our project was responsible and good for the world we spoke to a number of academics, industry experts and potential users. Within this our project had to clearly answer a series of questions. Through this framework we established that our project would have value.
Summary (Gold): Our silver human practices helped “close the loop” to ensure our project matched what was essential for users and what was superfluous. Furthermore, they helped establish our approaches of education & outreach.
Desire into design: Throughout our project design we considered the advice from the silver human practices. In particular we altered our project design to better match what would be wanted from potential users.
- Changing how the circuit would be perturbed: temperature, nitrogen and chloramphenicol
- Energetic consideration: E. coli as the model organism and assembly in low copy number plasmids
- Enhanced usability: combinatorial design which allows consideration of users needs and restriction enzyme to enhance facilitate modular implementation
Furthemore, the advice provided help to optimise our wet lab and dry lab success from experimental set-up, cloning tips, to mathematical modelling.
Responses to human practice: Our surveys led us to realise that control theory, synthetic biology and in particular cybergenetics remain poorly understood amongst peers and 6th formers. As such we endeavour to communicate the values of cybergenetics in a simple to understand way through collaborating with the Kazakhstan high school iGEM team.
Compromises in design: Throughout the design process we had to make a series of compromises. Through the advice of our interviews, these were often made based on prioritisation of the simplest and most time/cost efficient design and experimentation whilst providing the essential information.
Establishing what types of problems wanted to tackle (Silver)
Throughout our iGEM journey, human practices were a driving force. In particular, we wanted to combine our love for synthetic biology with a real world change. This would elevate a summer internship to something with real-world value. The basic workflow of our project consisted of first initial ideation and human practices. This instructed the design element of the design, build, test, learn (DBTL) cycle. Once the DBTL circuit had been completed the genetic circuit could be implemented.
As such we wanted to target a fundamental problem within synthetic biology and hence provided a solution that could impact many different people with many different problems. Therefore we applied to the fundamental track. In our early ideation, planning and reading of the literature we identified that one major challenge was variability within a system (see our Description page). Designing a biological integral controller could help address this challenge.
Framework for evaluating the values of our project (Silver)
An important feature of the ideation process was that the project had clear utility. We needed to clearly establish when, where, why and to whom our project would be useful to. Would our project actually reach as many different people as we would have hoped using effective altruism as a basis? To answer these questions we endeavoured to speak to as many different academics, industrial professionals and potential users who are involved in the field. Once we established a clear utility and value associated with the project we wanted to understand how the project could become practically usable. Here we tried to implement as many features requested by our interviews with the design of our project. Finally, we used our human practices and contacts as a source to make this project into a reality. Throughout we used this framework of questions to guide our human practices.
- Evaluation of what the problem and solutions are?
a) What is the problem? How big is the problem? Who does this affect?
b) How could this be fixed? Would our project be able to achieve this?
c) What core values would it provide? How could this be evaluated?
d) What are the possible applications of our project?
e) What are the challenges associated with the field (cybergenetics) - Integrating desire to design
a) Would our project be useful?
b) What features would be necessary for it to be useful?
c) What features would be desirable but non-essential?
d) Who and how would our project be useful for?
e) Are there any limitations or concerns within our project? Are there any concessions that we would have to make in our design?
f) How could our project be successfully implemented? - How can this be made into a reality?
a) Technically, how could our project be built?
b) What are some challenges with our design?
c) How could this project be achieved in the lab?
d) How could this project be achieved through modelling?
e) Any further advice?
Evaluation of the problems within synthetic biology and values associated (Silver)
Given we wanted to tackle a fundamental problem we wanted to identify what are some of the key limitations affecting synthetic biology. Kwok 2010 outlined ‘Five hard truths’: (1) many parts are undefined, (2) circuitry is unpredictable, (3) complexity is unwieldy, (4) many parts are incompatible and (5) variability crashes the system. Furthermore, biological circuits are subjected to intrinsic (stochastic gene expression, genetic variation, abundance of endogenous resources or copy number variation) and extrinsic (pH, nutrient availability, temperature) (Wang et al., 2019). To accommodate this, biology utilises homeostatic mechanisms known as perfect adaptation to maintain internal components at an unperturbed level. Designed synthetic genetic circuits are notoriously fragile to disturbance (Sun et al., 2022). This is important as the circuits designed in the lab often fail in real-world and industrial applications. To overcome these fundamental challenges the field of cybergenetics ensures robustness. In particular feedback controller modules such as integral controllers have been developed to set the output of interest at a strict target level.
Using these core challenges of synthetic biology in mind we wanted to design an integral controller to demonstrate robustness and to reduce the impact of variability within a genetic circuit. Throughout we attempted to clearly test all components used within the system to better define the parts and understand what parts were compatible. Furthermore, our factorial and combinatorial design aimed to better define the circuitry.
Whilst our project would have value to the synthetic biology world in a general sense we wanted to also establish clear functional applications in the real-world. Through talking to those working in academia and industry we established that our project would have an important and functional impact for applications such as such as metabolic engineering & biomanufacturing (Dr Briat, Codexis, Cambrium, Biosyntia), microbial drug delivery (Dr Frei, Dr Baetica). See implementation for applications and interviews for more details. Throughout the design process we tried to implement as many desired features outlined through our human practices research. Additionally, through their advice we design a project that would be purpose build and could be achieved with the iGEM timeline.
2. Integrating desire into design (Gold Medal)
Once we established whether cybergenetics and our project could be useful we wanted to speak to those who would actually use what we have created. With this in mind we wanted to identify what features would be essential and useful for proper implementation. Through this we identified and implemented three key features into our design: (A) Methods of circuit perturbance, (B) energetic considerations, (C) usability through factorial design.
A. Circuit perturbance: Most useful way to perturb the circuit
We wanted to test our circuit in two different ways (to see more information see description). Firstly, we wanted to test that the circuit acts as an integral controller in the way that it was designed. This is achieved through positive perturbation. However, we also wanted to test that our circuit can robustly adapt to a global perturbance that might affect the production of the desired protein. For our global perturbation, we initially planned to decrease temperature from 37
Firstly, we discussed ‘mixing problems’ observed in biomanufacturing with Codexis. ‘Mixing problems’ describe how in large fermentation tanks found in biomanufacturing the liquid culture is often very poorly mixed and attempts to do so effectively are expensive and difficult. Most commonly there is an unequal distribution of nitrogen and carbon sources, oxygen, pH and temperature. This can result in significant heterogeneity in cell survival and in particular the cell's ability to produce the desired protein. However, oxygen is difficult to control within a typical iGEM lab and glucose is difficult to make independent of cell growth. As such, Codexis validated our rationale of using temperature and they also suggested that nitrogen (via ammonia sulfate) would be a useful global perturbance. More generally, this would be useful since the cell via our circuit would be able to produce the desired protein at a consistent level regardless of external environmental factors within a given survival range. To accomodate for temperature in our design, we used mVenus as our reporter instead of GFP. This is because unlike mVenus, the maturation time of GFP is temperature sensitive so would effect the reliability of fluorescent measurement at different temperatures tested.
Secondly, we considered a global perturbation that could theoretically take place in microbial drug delivery with Dr Timothy Frei. One possible application is if the microbial drug delivery population may subsist in the gut for extended periods of time or coincides with antibiotics given to treat a second ailment. Antibiotics would otherwise act to harm the microbial drug delivery population and therefore reduce the amount of drug delivered.
Our conversations with Dr Frei and Codexis validated use of temperature as a perturbation but also introduced nitrogen and antibiotic concentration changes as additional perturbations
B. Energetic considerations: Organism of choice, speed of response and metabolic burden - Aoki, Frei, Codexis and Biosyntia
When we were designing our circuit we considered different model organisms and their resulting different advantages and disadvantages as well as applications. Initially, we were considering a range of different organisms to test our circuit in including E. coli, B. subtilis, S. cerevisiae or chlamydomonas. Speaking with Aoki, she suggested that E. coli would be the best model organism to use. It is the best studied both experimentally and computationally, with a fast replication rate, easy to manipulate and importantly is used widely in industrial metabolic engineering. Furthermore, we were initially considering an mRNA based system in S.cerevisiae but it was suggested that a protein based system would better match the properties (i.e. half-life) of the protein expression we were aiming to control. Finally, expression strains such as E. coli BL21 or Rosetta strain have high protein expression rate. Biosyntia mentioned that if our circuit is to have value in industry it should have fast recovery rate to steady state 0. These high expression strains would have faster protein turnover and therefore increased ability to recover to perturbation.
A further consideration that was suggested by Codexis was the requirement to reduce metabolic burden. This is most important within biomanufacturing where even small amounts of additional energetic costs would be selected against. Initially, we were planning to use high copy number plasmids for our level 2 assembly for experimental ease but after consulting Codexis they suggested that low copy number plasmids would significantly reduce the metabolic yield.
A further compromise when designing our circuits that had to be considered, was that higher production resulting in faster recovery from perturbations also resulted in increased metabolic load. As such in our design we designed multiple different circuits allowing for both reduced metabolic load as well as increased homeostatic control dependent on the users needs.
Through our conversations with Dr Frei, Dr Aoki we changed our model organism from S.cerevisiae to E. coli. Additionally, based on the advice from Codexis and Biosyntia we changed our design from using high to low copy number plasmids in our testing strains.
C. Usability: Factorial combinations of circuits- Bath iGEM, Timothy Frei, Cambrium
A final consideration is the usability of the circuit. When speaking to Cambrium they suggested that if our integral controller were to be actually useful, we would need to consider how the circuits could be altered to accommodate different users' needs. Firstly, we wanted to establish what different features of a circuit different users would actually want. Through our collaboration with Bath iGEM they talked about some of their needs. These included speed of recovery after perturbation, metabolic load concerns or what level their protein needs to be expressed at. For example, if the desired protein was toxic, then the reference expression point should be comparatively lower to a different user to with a non-toxic protein. Alternatively, one user might value a reduced metabolic load whilst another might prioritise speed of recovery. To accommodate for this we decided to go for a factorial design whereby if we made and tested many different circuits using different controller proteins and using different set levels a given user could choose what level they would want their protein to be expressed at.
Furthermore with Bath iGEM we discussed the need to make our system as usable and modular as possible. To easily implement the users protein we discussed a design in which the user would, in a one-step restriction enzyme reaction, known in their protein of choice.
Through discussions with Cambrium, Dr Frei and Bath iGEM we first identified what different features users want for their integral controller circuit. Using this we planned a factorial design to allow selectability. Moreover, to enable easy implementation we discussed a one-step restriction knock-in reaction.
Human practices informing communication decisions
Control theory and its role in synthetic biology through cybergenetics was a new field to many of our iGEM team. Only through extensive background research and interviews could we fully appreciate the importance of this burgeoning field. To establish whether the general public or even other iGEM teams were aware of cybergenetics we created a survey to identify what was and wasn’t known (see communications). From these results we found that only about ⅓ to ½ of our participants knew anything about synthetic biology or control theory terminology.
Firstly, this helped educate us on how best to communicate the principles of cybergenetics. In particular, what level of understanding we could describe these fields at. Secondly, we realised that majority of the public did not know much about cybergenetics. As such this drove us to try and educate as many different people from across the public, backgrounds and ages. This range from the general public (YouTube videos, information leaflets) young children (CHaOS) to secondary school students (Kazakhstan team collaboration and sixth form talks) to undergraduates (Freshers Fair).
How can this be made into a reality?
Once the value of our project had been established and we had understood and implemented key features necessary for our project to be useful we then relied on the expertise of academic and industrial experts to help with our project. In many cases their suggestions fundamentally changed our project design considered practical and experimental problems. This is described in more detail in design, attributions and modelling.
The initial inspiration of our project was based upon work from Aoki et al., 2019. As such we endeavoured to talk with Dr Aoki to understand the principles and success of her build design. Firstly, Dr Aoki helped us identify the optimal permutations of sigma:anti-sigma factors acting as the controller proteins. For example, she recommended against using AraC as a controller protein as it was toxic at high concentrations and was reliant on AraE exporter proteins. Furthermore, unlike Aoki et al., 2019, the steady state reference points were not set by inducer concentrations and instead set by using distinct circuits in the combinatorial design. Inducers are experimentally cumbersome, expensive and unwieldy for modulating different steady state levels. Additionally, unlike her paper tuning with inducers was cumbersome and expensive for modulating different steady state levels. This further validated our combinatorial design as a mechanism to allow tuning of the different circuits.
Talking with Dr Osman we discussed the possibility of designing a positive rather then negative perturbation of controller X in order to prove our circuit works as an integral controller. In the literature, typically only negative perturbations have been described. He helped us realise a relatively simple design in which a positive perturbation would be easy to implement. Furthermore, we initially planned to record the fluorescence from our circuits using a plate reader. Dr Osman highlighted the importance of instead using single cell recording devices such as a mother machine. This would allow a better method to introduce the perturbation (i.e. quick and simple introduction of chloramphenicol antibiotics) as well as data which statistical power would not be reduced by averaging. Furthermore, we discussed the importance of proper characterisation of all the parts within our circuit. Only when you established which parts work and whether they work as expected could we expect to understand the properties of more complex circuits. This led to us prioritise proper characterisation of all our sigma and anti-sigma factor circuits. From this we could perform factorial design to get different circuits which enable different properties.
Compromises and future developments
Throughout the design process there were a variety of key features suggested by our human practices that could not be implemented due to time, cost and feasibility constraints. So for future development of this project we would suggest these possible developments.
Oxygen as a circuit perturbationArguably one of the most useful perturbations to our circuit would have been oxygen concentrations. Within both humans and in biomanufacturing, oxygen levels are not constant. This leads to regions of anoxic conditions which reduce gene expression. Unfortunately, anoxic conditions would be hard to test for two reasons. Firstly, all experiments would need to be tested in an oxygen controlled environment with constant and controllable oxygen concentrations which the lab at BMS did not have. Secondly, our reporter system cannot operate in anoxic conditions as fluorescent proteins require oxygen to mature. Creating bioluminescence tags as a reporter is possible to solve the problem.
Metabolic load analysis and plasmid integrationWithin synthetic biology, metabolic burden represents a major problem. This describes the burden caused by redistribution of resource availability from the host cell to the expression protein. If metabolic load is high then the cell's biochemistry may be significantly impacted, reducing protein expression. In the case of plasmids this could also lead to a selective advantage of those not expressing the expressed system. Within our human practices this appeared as a major concern particularly in the biomanufacturing applications.
Our circuit consists of 3 transcription units (above that of a constituently expressed protein) which adds increased costs. These factors need to be replicated and for the CDS transcribed and translated. Furthermore, the translated components may result in toxicity especially if expressed at a high level. The integral controller works by producing Z1 and Z2 which constantly annihilate each other which is energetically wasteful. Our hope is that on a population level consistent control of the protein expression would globally reduce metabolic load. However due to time constraints, we were unable to properly characterise the burden of the circuits and analyse the energy cost. This could be done by comparing the circuits in different copy number plasmids/ origins. Moreover, searching for appropriate controller proteins which minimise metabolic load/ toxicity remains a game of trial and error and literature searching.
Furthermore, if plasmids carry a metabolic load they can be selected against and over time since the population will lose the plasmids and expression of our desired proteins. One suggestion from Biosyntia, provided more time, would be to integrate these circuits on the chromosomal backbone or implement within the backbone.
Proportional Integral (PI) controller:The antithetic integral controller circuit we have designed is an integral controller. However, using an integral controller on its own theoretically creates large variance compared to both the open loop and proportional feedback system despite leading to 0 steady state error. Therefore, if given the chance to improve our circuit design again in the future, we would design a proportional controller to negatively regulate the antithetic species Z1. Theoretically, this will lead to both a lower variance and like the antithetic integral controller can approach 0 steady state error. A PI controller was implemented in mammalian cells using mRNA as antithetic species in Dr Frei’s lab but to our knowledge it has not been implemented in bacteria yet.
Our project focuses on the design of antithetic integral controllers in a single cell. However, it is possible to implement this motif in a population level, allowing the realisation of this motif in a multicellular approach.
The main limitation of the antithetic integral controller is that all the circuit components, both the controller and the process, are confined into one cell. As many circuit components are expressed in the same cell which leads to a high metabolic burden. Another limitation is that modularity is hard where if we want to control another pathway, a new pathway has to be reengineered for the purpose.
One way to solve this problem is to use a multicellular approach where the controller species is produced by one population of cells while the process is produced in another population of cells. This allows the process species population to be changed easily to adapt for production of other products, as long as the 2 strains can still recognise each other. This essentially allows the production of different target products controlled with the same controller.
Interviews
Below we have documented further details of the interviews conducted for our human practices and integrated human practices:
The details of these interviews are summarised as follows:
1. Context of the interview and why we approached this individual
2. The insights we gained
3. How this acquired knowledge influenced our project
Furthermore, key research papers, relevant to the context of the interview, are highlighted.
Dr Corentin Briat
PhD in Systems and Control Theory
1. Context:
Dr Briat is the lead author on multiple papers that we took interest in when formulating our project. He has spent much of his academic career researching integral control feedback mechanisms including the antithetic feedback circuit. We took inspiration and guidance from his extensive theoretical work, including his modeling of the behavior of the antithetic feedback control circuit in the context of biofuel production.
We approached Dr Briat to ask more about his modeling approach, the application of the antithetic feedback controller to biofuel production, his thoughts on other applications, and how his theoretical work can guide in-vivo application of the antithetic feedback controller.
2.Insights:
- Discussed single cell vs population level; Dr Briat explained that the single cell control does translate to the population level. He also suggested the use of multiple control loops; one for control on a single cell level and one on a population level.
- Discussed how best to incorporate stochasticity into the model and the relative importance of stochastic considerations.
- Dr Briat explained the benefits of batch production vs continuous production and how the antithetical feedback controller can improve these production methods.
- We discussed the energetics associated with the implementation of our circuit; Dr Briat explained how we needed to consider the responsiveness of our circuit and energy consumption.
- We discussed how best to perturb our system to demonstrate robustness and what perturbation he had in mind when modeling the antithetical feedback system in biofuel production.
- We discussed other applications of the antithetical feedback control system other than in the context of biofuel production. He explained that the system is well suited to applications in which there is a clear steady state optimum; in the context of biofuel production ,this arises because too much production of biofuel is toxic yet too little would lead to a lower yield.
3. Influence:
- Following his insights on how to model the single cell vs population dynamics, we decided to treat the population level as the time average of single cell level in our models.
- Following his insights on stochasticity we incorporated stochastic effects in to our simulations using a stochastic solver for the equivalent deterministic set of ODEs and further theoretical consideration of stochastic effects was not required as stochasticity does not affect the operation of the controller.
- Following his insights on how to perturb the system we decided to perturb our system both by a global mechanism (temperature) and a local mechanism (changing degradation rate of a specific species). To then demonstrate our circuit is more robust to these perturbations we decided to directly compare the response of our system to that of an equivalent closed loop system since even if our system is not perfectly robust we can hopefully demonstrate that it is more robust than it would be with out the control system.
Key publications:
"Perfect Adaptation and Optimal Equilibrium Productivity in a Simple Microbial Biofuel Metabolic Pathway Using Dynamic Integral Control"
"Antithetic Integral Feedback Ensures Robust Perfect Adaptation in Noisy Biomolecular Networks"
"Antithetic proportional-integral feedback for reduced variance and improved control performance of stochastic reaction networks"
Dr Stephanie Aoki
PhD in Systems Biology
1. Context:
Dr Aoki is the lead author on multiple papers that we took interest in when formulating our project. Specifically, she is the lead author of the paper that details her work in implementing the first antithetic integral feedback controller to achieve robust adaptation in-vivo and is to date the only instance of successful implementation of such in a bacterial system. This work was incredibly valuable to us when designing our genetic circuits and planning our experimental work. As a team, we found Dr Aoki’s work to be inspiring and it gave us a solid foundation on which to build to realize our project goals.
2.Insights:
- Dr Aoki discussed with us the best mechanism of integration in our system. Specifically, we discussed the benefits and negatives of using sigma-antisigma factors vs RNA-based mechanisms.
- We discussed the best applications of the antithetic feedback control system, including metabolic engineering, biofuel production, and regulating hormone levels in the context of therapeutics and expression libraries.
- We discussed the most appropriate organism to implement the circuit in
- We discussed the implications of different types of perturbation, including the effects on the symmetry of Z1 and Z2 and how different perturbations would affect our measurements (e.g. via the maturation rate of GFP), and how this would need to be taken into account when interpreting our results.
- We discussed practical considerations and challenges in her experimental set up
3. Influence:
- We decided to use sigma-antisigma factors as our integrator species. This is because the rapid degradation of mRNA would mean the steady state error would be higher and the integrator would essentially be too leaky.
- We decided to explore more applications where a clear steady state optimum is required (i.e. due to a trade-off between yield and toxicity) rather than applications where instead constitutive expression at high levels would give rise to high yield with no trade-off considerations.
- We decided to put our Z1 species under constitutive expression, rather than under an inducible promoter as she had done in her research since this would lead to greater symmetry between the Z1 and Z2 species which in theory leads to greater performance.
- We decided to perturb our system in such a way that affected both Z1 and Z2 identically as this was required to maintain the symmetry of the two species; this could be done via a global perturbation or one that specifically targeted our X species.
- Although yeast and the mammalian system can be beneficial in the fact that the effects of dilution are mitigated, we decided to initially implement our circuit in E. coli; this is because Dr Aoki explained how the tuning process of the system took a long time and therefore, given the time constraints of the project, opted for a system with a shorter doubling time. We would then take into account the effects of dilution by including extra degradation terms in the ODEs that we use to model our system.
Key publications:
"A universal biomolecular integral feedback controller for robust perfect adaptation"
Dr Fulvio Forni
PhD in Computer Science and Control Engineering
1. Context:
Dr Forni is a control theory engineer and is an expert on the engineering principles that are crucial to understanding our project. He has published papers on the control theory aspects of the antithetic integral feedback motif that we are interested in and understanding his work was important to us in order to appropriately model our system. The theoretical considerations his work described were fundamental to us understanding how to successfully implement the antithetic feedback motif.
2.Insights:
- Dr Forni explained some of the concepts behind oscillatory networks and we discussed whether this was something that we could incorporate into our system (i.e. manipulate the system to allow for oscillatory dynamics).
- Dr Forni discussed with us the best ways to study the stability of our system from a control theory perspective.
- Dr Forni discussed with us the best ways to model the effects of different temperatures; namely the ways it could affect the kinetic parameters and noise.
3. Influence:
- Following Dr Forni’s discussion about stability analysis we decided it would be important and insightful to carry out an extensive stability analysis on our system and this would furthermore benefit our experiment design by finding the most suitable parameters.
Specifically, we decided to use the Jacobian of the system and investigate the stability in the Laplace domain using one node. We would use the stability analysis to ensure oscillations are damped (as we decided this would be undesirable in our system).
We decided that we needed to calculate a robustness margin in order to fully characterize the success in achieving a robust adaptation.
Furthermore, in discussing sensitive analysis, Dr Forni gave us direction on which approximations we should make in our model. - After discussing the timescales of each of the components of our circuit, we decided that a time constant perturbation would allow us to demonstrate the robust adaptation best, as opposed to a fluctuating perturbation.
- In considering the ways in which temperature affects the system we decided that if using temperature change to perturb the system then we should use a positive temperature perturbation since lowering the temperature would reduce the sequestration parameters and in order to achieve a wide robustness margin we would ideally need high sequestration rates.
Key publications:
Dr Timothy Frei
PhD student in Systems Biology
1. Context:
Dr Frei is the lead author on multiple papers that we took interest in when formulating our project. Specifically, he is the lead author of the paper that details his work in implementing the antithetic integral feedback controller to achieve robust adaptation in-vivo and is to date the only instance of successful implementation of such in a mammalian system.
2.Insights:
- We discussed the metabolic burden of using the circuit and how using a mammalian system reduced the relative burden.
- We discussed the possible applications of the project in both mammalian, yeast and bacterial systems.
- We discussed his future work and his outlook on the field.
3. Influence:
- Following a discussion on metabolic burden and our previous decision to use implement our circuit in bacteria, we decided to use low copy numbers in order to reduce metabolic burden.
- Following a discussion about what species worked best in the circuit, Dr Frei confirmed that it would be most appropriate to use protein-protein annihilation as opposed to using protein-RNA based annihilation as this would introduce an unwanted asymmetry in to the circuit.
- Following a discussion on improving experimental design we concluded that experiments to characterize stability could be very helpful. Specifically, we planned to put the controller in an open loop to show the characteristic response, then we would set Z1 somewhere, titrate/induce Z2 and should see Z1 activated after a threshold, and should be able to change this threshold by level of Z1. This would give a good indication the circuit is behaving as in theory before continuing with experiment.
- Following discussions on how to tune the set point in mammalian cells vs bacteria, we decided to do so by using different promoter and RBS strengths as doing so with changing the inducer of a promoter introduces difficulties with burden.
- Following discussions on the best way to measure the response of the system, we decided to use plate readers and then if time permits flow cytometry could be used (and would also be useful in verifying the plate reader measurements).
Key publications:
"A genetic mammalian proportional–integral feedback control circuit for robust and precise gene regulation"
"Characterization and mitigation of gene expression burden in mammalian cells"
Dr Ania-Ariadna Baetica
Post-doctoral scholar in Biophysics and Biochemistry
1. Context:
Dr Baetica authored multiple papers that our team studied extensively when designing our project. Her work included research on the hard limits and performance trade offs associated with the antithetic feedback control mechanism, of which are important things to consider when aiming for a successful in vivo realization of the circuit. She also published a paper detailing guidelines for designing the antithetic feedback motif from a theoretical perspective, of which was, again, critical work for us to consider when planning our project.
2.Insights:
- We discussed possible applications and Dr Baetica provided valuable suggestions including bio manufacturing and cell based therapeutics.
- Dr Baetica shared her thoughts on the extent to which these controls motifs could be modular.
- Dr Baetica shared her thoughts regarding the future outlook on control theory in synthetic biology.
3. Influence:
- Following a discussion on modelling considerations we decided to extend our modelling efforts to further explore the hill functions activation and the parameters associated with this.
- Following a discussion on modelling we decided to include mRNA in our two node model by using time delays as well as explicitly including the relevant ODEs.
- Following a discussion on modelling we decided to use the concept of a leaky integrator to account for dilution, degradation and instability.
Key publications:
"Guidelines for designing the antithetic feedback motif"
"Hard Limits and Performance Tradeoffs in a Class of Antithetic Integral Feedback Networks"
"Control theoretical concepts for synthetic and systems biology"
"Design Guidelines For Sequestration Feedback Networks"
"Hard Limits and Performance Tradeoffs in a Class of Antithetic Integral Feedback Networks"
Dr Noah Olsman
PhD Control and Dynamical Systems
1. Context:
Dr Olsman is a well-established and influential scientist in the field of systems and synthetic biology and specializes in the study of how cells use feedback control to guarantee that their behavior is robust to the inevitable uncertainties in living systems. His work has, therefore, unsurprisingly, been greatly influential to us from the project ideation stage through to the planning of the implementation of our project. His work provided crucial insights for us on the fundamental limits of biomolecular feedback control and we were very eager to discuss how his research and extensive knowledge could benefit our project and for him to scrutinize our plans.
2.Insights:
- We discussed the justification of using linear activation instead of using Hill dynamics, whereas for the repression case hill dynamics were used.
- We discussed the best ways to measure the response of our circuit accurately.
- We discussed the best ways in which to reliably disturb our system.
- We discussed the consequences of bimodality for toggle switches.
- We discussed the positives and negatives of using low vs high plasmid copy numbers.
- We discussed the differences between an activation-based and repression-based antithetic circuit.
3. Influence:
- Provided methods of verification of the assumptions of linear activation (as opposed to Hill dynamics.
- Following a discussion on distributions, we now considered the importance of understanding the distributions underlying data and how this needs to be understood in order to correctly interpret the data. Distribution fitting algorithms could be used on our data to help ensure we do not incorrectly interpret data.
- Following a discussion on repression-based antithetic circuits we became aware that we would need to pay greater attention to possible promoter saturation in the repression case; repressor binding is highly cooperative and so saturation would be harder to regulate.
Key publications:
"Hard Limits and Performance Tradeoffs in a Class of Antithetic Integral Feedback Networks"
"Architectural Principles for Characterizing the Performance of Antithetic Integral Feedback Networks"
"Modeling and analysis of modular structure in diverse biological networks"
"A universal control system for synthetic gene networks"
Dr Lucia Marucci and Mr Davide Salzano
Associate Professor in Systems and Synthetic Biology, PhD student in Engineering Mathematics applied to personalized therapies
1. Context:
Dr Marucci’s research is dedicated to developing quantitative tools to understand cell dynamics and engineering-inspired methodologies for controlling living system functions.
Her approach combines tools from different disciplines: control engineering, systems, synthetic biology, and computer science. She is therefore an expert in the disciplines that our project is routed in and we were eager to reach out for her thoughts and expert opinion.
Supervised by Dr Marucci, Davide Salzano's research deals with the application of control theory for the development of synthetic microbial consortia for use in biomedical applications therefore conversing with Mr Salzano would enhance our understanding of our proposed feedback controller and importantly be useful when considering its applications, specifically microbial drug delivery.
2.Insights:
- Discussed possible applications of the antithetic circuit for robust perfect adaptation, including in being used as a module that would give a reliable signal, in bioproduction, and in complex production processes requiring specific concentrations of intermediates and products.
- Discussed the limitations of our proposed circuit; specifically in the context of metabolic burden; Dr Marucci has considered this in her research.
- The academics provided insight on how the antithetic motif could be realized in a population setting including in the control of endogenous pathways in cells that cannot be synthetically engineered.
- Discussed the performance tradeoffs and design principles that these academics have previously researched, in the context of our project and design of the experiment.
- Provided guidance on how best to identify parameters pre-experiment and then fit the parameters of our system post-experiment, then how to incorporate this into our models to gain meaningful insights and understanding about the system.
- Provided guidance on how to deal with the linearization of the hill function associated with the activation of species in our network.
3. Influence:
- Following the discussions with Dr Marucci and Mr Salzano, we used their advice to outline a clear method for which to appropriately obtain the parameters of our system pre-experiment, incorporate them into our models and then fit them with experimental data. This included suggestions of what parameter ranges were appropriate for initial models and how accurate the parameters ought to be pre-experiment, and which optimization algorithms would be appropriate.
- Following the discussions with Dr Marucci and Mr Salzano, we used their advice on how to handle the non-linearity of the hill function in our models. Like many theoretical papers discussed, we decided to assume linearity with the understanding that this is only true if the promoter works far from saturation. We planned to test the validity of assuming the promoter is far from saturating by considering the open loop response experimentally.
Key publications:
"Hard Limits and Performance Tradeoffs in a Class of Antithetic Integral Feedback Networks"
"Architectural Principles for Characterizing the Performance of Antithetic Integral Feedback Networks"
"Modeling and analysis of modular structure in diverse biological networks"
"A universal control system for synthetic gene networks"
Evonetix
1. Context:
Evonetix is a Cambridge based biotechnology company. They have designed novel semiconductor technology capable of synthesizing thousands of independent sequences across the surface of a single chip, re-engineered phosphoramidite chemistry to enable thermal control of the synthesis cycle, developed new ways to assemble DNA into genes and established novel approaches to enzymatic DNA synthesis. As one of our sponsors, the team were lucky enough to be invited to meet with a variety of the researchers, computer scientists and leaders of the Evonetix company to discuss their technology and synthetic biology solutions.
2.Insights:
- We discussed the current limitations of DNA synthesis methods and how the novel technology being developed at Evonetix could push these limits.
- We discussed the positive implications this would have on synthetic biology as a whole.
- We discussed how expertise from a wide range of fields and disciplines is required at a company like Evonetix to achieve it’s goals; skills from life sciences, economics and business, computer science, physics and engineering, leadership etc. are all fundamental to the success of a start up company. From this we found a greater appreciation of the diversity in skills and education that we have on our team!
3. Influence:
- It inspires us to pursue a project rooted in foundational advance.
- Seeing the novel methods of synthesising DNA sequences, we see how exciting synthetic biology is.
- Evonetix is a company working on foundational advances, making it clear to us the merits of having a project which was not restricted in the positive applications it could have.
- Following our visit and discussionss with Evonetix, we decided to commit to the integral controller for robust adaptation project idea and follow the foundational advance track.
Website:
"Evonetix"
Codexis
Dr Michael Miller- Director
1. Context:
Codexis, Inc. is a protein engineering company that develops enzymes for pharmaceutical, food and medical applications. Codexis use a range of technologies to achieve a variety of goals. We wanted to know if feedback control motifs were something that would enhance their biotechnology endeavors and if so specifically which of their focus areas it would be most suited to improve. We met with Dr Michael Miller; a molecular biologist from Codexis leading a group producing a variety of enzyme products.
2.Insights:
- We discussed the different contexts in which steady state output is desirable over trying to maximize output. Although in his group they are aiming for maximal production rather than steady state, Dr Miller appreciated the use of feedback control to maintain steady state in metabolic engineering applications.
- We discussed the metabolic burdens associated with implementing feedback control motifs and how this would affect industrial costs and operation.
- We discussed oxygen and nutrient variability within tanks; Dr Miller spoke about how there is large variation with in tanks and to engineer cells that would produce the same output regardless of position in tank would be useful and could even reduce overhead cost compared to methods of stabilizing outputs using controls on the population level.
- We discussed different test cases in terms of output species and types of perturbation that would reflect the perturbations industry face. Examples include temperature and nutrient perturbations where in large bioreactors it is difficult to get things well mixed (this requires lots of additional energy input) and to have the cells be engineered in such a way to reduce the need for well mixed reactors would be desirable.
- We discussed the limits of different industrial facilities; e.g. how reliable the nutrient valves in a bioreactor are, how reliable their water cooling systems are.
3. Influence:
- Following discussions on metabolic burden, we decided to ensure that the most simple version of the antithetic feedback motif would be used (1 node) as additional, unnecessary complexity would increase metabolic burden.
- Following discussions on ways to perturb the system that would reflect perturbations that are relevant to real-world applications, we concluded that temperature or nutrient perturbations would be best. It was interesting to consider the problem of nutrient valves being unreliable or breaking and we could simulate this form of perturbation by adding an influx of ammonia (or other chemical/ nutrient) into the system. Ammonia perturbation would furthermore be more practical as opposed to glucose or oxygen perturbations since glucose would be hard to make independent of cell growth and changing and monitoring oxygen concentration would be difficult.
- Following suggestions from Dr Miller we considered using test cases that we had not previously considered.
- Dr Miller further provided reassurance of the value of achieving success in a stripped down operational system and instead of targeting a specific end product we could target a specific time-frame or concentration to make it work in, ones which would be relevant for industry and determined by a facilities limits.
Website:
"Codexis"
Cambrium
Dr Charlie Cotton- Chief Scientific Officer
Dr Pierre Salvy- Head of Engineering
1. Context:
Cambrium is a company that uses microbes as cell factories to produce designer proteins at scale. We were eager to learn more from the experts at Cambrium about their technologies, the challenges they face using microbial cell factories, and whether synthetic control feedback motifs could help to address any of these issues or improve other aspects of their production process.
Moreover, as a team and individuals, we were inspired by Cambrium's platform; the way they use microbes to positively impact the environment and society (e.g. by producing novel sustainable materials) is very exciting!
We were lucky to meet with Dr Charlie Cotton and Dr Pierre Salvy; chief scientific officer and head of engineering, respectively.
2.Insights:
- We discussed new applications to consider for the antithetic integral controller; this included applications where protein expression needs a specific cofactor or when excess product leads to protein aggregation. However, it was noted that pharmaceutical production would be one of the best applications.
- We discussed applications in the realm of quality control and using the circuit as a sensor to say when the cell is stressed.
- We discussed the difficulties of controlling the temperature in large-scale fermentation facilities.
- We discussed the implications in terms of cost and energy saving that improving temperature control would have.
- We discussed the problems associated with scaling up in biotechnology.
3. Influence:
- Following discussions of their costs and bottlenecks we decided to look for applications in which the bottlenecks are in the fermentation process as opposed to the post-processing of product stages. The majority of the costs in the production of biomaterials using microbial cell factories are in the post-processing of the materials themselves rather than in the fermentation procedure.
- We gained further appreciation of the merits of demonstrating robustness against temperature perturbation and the consequences this could have in increasing cell metabolism without inducing stress responses.
Website:
"Cambrium"
Biosyntia
Dr Carlos Acevedo-Rocha- Principal Scientist
1. Context:
Biosyntia is a biotechnology company focused on the production of natural ingredients using microbial fermentation processes. They use their unique and proprietary microorganisms and integrate them into full-scale manufacturing processes to produce a variety of products in an environmentally sustainable way.
As a team, we were interested in the microbial fermentation processes they use, since this is an area that we identified as being an ideal application for our controller motif. We also found the positive environmental impact that Biosyntia is making to be very inspiring.
We were lucky to meet with Dr Carlos G. Acevedo-Rocha who is the principal scientist at Biosyntia. He gave us an engaging introduction to the company and described comprehensively their production process including the bottlenecks and challenges that they face. We then had an informative discussion about various aspects of our project and our aims; Dr Acevedo-Rocha was incredibly insightful, thorough and knowledgeable.
2.Insights:
- We discussed the problems associated with using microbes to produce products. This included socio-economic issues and upscaling issues.
- We discussed how closely the parameters of the system need to be controlled in their processes. Our discussions about the company’s journey and how it has evolved over the past years were exciting and interesting.
- The discussion provided additional insights into what it means to work in a biotech start-up and we found it inspiring as the members of our team consider their future careers.
3. Influence:
- Our discussions lead to a greater appreciation of the economical considerations needed when identifying applications; our control motif would likely be best suited for the production of high-value compounds with the industries of pharmaceuticals and cosmetics being suggested.
- This discussion lead to us consider how our work in the lab may feasibly be realized in industry; specifically, how using plasmids in E. coli may be problematic (plasmids lose productivity as they have elements that depend on how they replicate and need to add antibiotic for selection). A long-term solution would be to integrate the sequence into the chromosome.
Website:
"Biosyntia"
References
- Aoki SK, Lillacci G, Gupta A, Baumschlager A, Schweingruber D, Khammash M. A universal biomolecular integral feedback controller for robust perfect adaptation. Nature. 2019 Jun;570(7762):533-537. doi: 10.1038/s41586-019-1321-1. Epub 2019 Jun 19. PMID: 31217585.
- Kwok, R. Five hard truths for synthetic biology. Nature 463, 288–290 (2010). https://doi.org/10.1038/463288a
- Heins, AL., Weuster-Botz, D. Population heterogeneity in microbial bioprocesses: origin, analysis, mechanisms, and future perspectives. Bioprocess Biosyst Eng 41, 889–916 (2018). https://doi.org/10.1007/s00449-018-1922-3
- Wang T, Dunlop MJ. Controlling and exploiting cell-to-cell variation in metabolic engineering. Curr Opin Biotechnol. 2019 Jun;57:10-16. doi: 10.1016/j.copbio.2018.08.013. Epub 2018 Sep 24. PMID: 30261323.
- Zhi Sun, Weijia Wei, Mingyue Zhang, Wenjia Shi, Yeqing Zong, Yihua Chen, Xiaojing Yang, Bo Yu, Chao Tang, Chunbo Lou, Synthetic robust perfect adaptation achieved by negative feedback coupling with linear weak positive feedback, Nucleic Acids Research, Volume 50, Issue 4, 28 February 2022, Pages 2377–2386, https://doi.org/10.1093/nar/gkac066