---

Summary

Summary: We have identified potential users and designed a circuit with easy implementation and little safety concerns.

Proposed users: Using research from our human practices we identified robustness as core challenge in synthetic biology. An integral controller would be particularly important for applications in metabolic engineering, biomanufacturing and microbial drug delivery.

Project implementation: We have created a three step process to facilitate simple implementation. First, a circuit is identified based on features of what the user wants to express. Second, their protein of interest is easily knocked-in to our circuit. Finally, the recombinant circuit is transformed and used directly in the given application. We have provided an FAQ to assist implementation challenges.

Safety concerns: Throughout our design process we have considered whether our project would be safe to implement. Generally, our circuit would add no additional safety problems and would be instead dependent on protein of interest and the applications.

Applications: Who would our project benefit?

The basis of our project was to help address the fundamental challenge within synthetic biology of stochastic, molecular variability and consequently sensitivity of the system to perturbations. Provided it is possible to mitigate these problems, how would our project be used and who would use it? A summary of the applications is provided in the Description page.

The integral controllers we have designed have two distinct beneficial properties. Firstly, as described by control theory, an integral controller allows production of the desired product at a reference steady state point. This allows the user to express their protein consistently at a set level with reduced steady state error compared to constitutive expression or negative feedback loop. An ideal application would be one in which a ‘Goldilock’ gene expression is necessary i.e. gene expression is wanted to be not too high and not too low.

Secondly, the integral controller allows enhanced ability to homeostatically react in response to intrinsic and extrinsic perturbations. These perturbations might be changes in intrinsic controller protein X or extrinsic factors such as temperature or antibiotics.

Metabolic engineering and biomanufacturing (Briat, Codexis, Biosyntia, Cambrium)

Metabolic engineering describes the genetic optimisation of biochemical reactions within a cell to facilitate and maximise a cells production of a useful product. The resulting circuits and cells are used to express important proteins in biomanufacturing processes. Biomanufacturing refers to the process of producing a biological molecule on an industrial scale, typically through a fermentation process of a microorganism such as bacteria, E. coli. Biomanufacturing is used in a wide variety of applications from biofuel production, pesticides, cosmetics, food production (wine and cheese) or pharmaceuticals (biologics and small molecules).

There are a number of advantages with using an integral controller in metabolic engineering. Firstly, provided the metabolic load does not significantly reduce production rate the yield of the produced product will be more consistent between bacteria within the population. The principle that the metabolic load, on a population level, might even be decreased was discussed with Dr Marucci and Salzano. In some cases this could lead to a total greater yield.

Secondly, the population will be able to cope much better with extrinsic differences such as mixing problems (Heins et al., 2018) with oxygen, temperature or nitrogen/carbon sources, provided they don’t damage the cells (see Human Practices). This becomes more important the larger the fermentation tank since mixing becomes harder. The different mixing problems was discussed with Codexis. In particular, attention was paid to the timescales in mixing problems and how long circuit recovery times would take. Often mixing problems take place on the order of minutes whilst our circuit would work on a slightly longer timescale. This is not to say that an integral controller wouldn’t see act to reduce the cell-cell heterogeneity produced by poor mixing. Equally, in time of mixing failures the cells would be able to better adapt. This could be particularly important in smaller scale, decentralised or less well controlled bioreactors perhaps in remote areas. An example of this was described by Codexis in which a small-scale bioreactor used unheated rainwater all year round. Essentially, the integral controller acts to move the control system inside the cell. Therefore a combination of extrinsic and intrinsic control systems would only enhance cellular robustness.

Thirdly, integral controllers might be particularly important in continuous production. Continuous production in biotechnology refers to the production of the desired protein product in a contiguous fashion. This is opposed to the more commonly used batch production method where the desired product is produced at the maximum possible rate all the time. This leads to toxic levels of the desired product and will kill the bacterial population whilst harvesting the product. Continuous processing is advantageous in that it is far more efficient allowing higher volumes produced in shorter time periods. Typically, the processing is less labour intensive and cheaper to run. This principle was demonstrated in Briat et al., 2018. We discussed this application further with Briat. He described a model in which a biofuel exporter was influenced by an integral controller. Too much exporter would lead to leakage of the cell membrane and cell death whilst too little exporter would lead to build up of toxic intracellular butanol.

Alternatively, proteins may need to be expressed at a particular level rather than maximally. This is particularly important when engineering metabolic networks or considering the ratios by which two genetic factors interact. For example, with Cambrium we discussed how our project could be used when considering complex biomolecules i.e. specific cofactors concentrations or where excess products lead to aggregation/toxicity. Additionally, with Biosyntia we considered that these circuits could be useful for high value compounds such as pharmaceuticals where quality is more important then quantity.

Microbial drug delivery (Frei, Baetica)

Secondly, there is also a potential for implementation of an integral controller in drug delivery. Microbial drug delivery is a novel type of drug administration in which genetically modified bacteria release medications. This has been previously used for chronic diseases in humans such as inflammatory bowel disease but is also hypothesised to treat cancers. Microbial-based drug delivery could be considered to be safer, site-specific and with less side effects. Using an integral controller with the drug release circuit would allow bacteria to produce the drug at a specific and constant rate. This would be important as too much drug could be dangerous whilst too little would not treat the disease. Dr Frei uses a similar principle within his lab. Instead he envisions future usage as a cell therapy with integral controllers in mammalian cells. There remain significant biosafety concerns with microbial drug delivery. In a landmark study, Braat et al., 2006 had to control drug release through dietary thymidine. The consistencies of drug release could be partially overcome by the integral controller. More recently, synlogic developed two engineering E. coli Nissel strains which encode phenylalanine ammonia lyase (PAL) (Adolfsen et al., 2021). PAL breaks down Phe, lowering Phe levels and allows consumption of dietary protein for patients with phenylketonuria. Placing PAL under an integral controller would ensure that a safe amount of Phe is broken down.

Alternative applications

Finally, whilst metabolic engineering and microbial drug delivery are the predominant applications we also considered a few alternative applications in which an integral controller could be used. Firstly, the circuit could be used within a biological biosensor. For example, a cell could be producing a protein which detects a particular compound. Ideally, detection would be only dependent on the compounds concentration. If the protein cannot be in excess to the chemical concentration then the only way this could occur is if the expression is at a fixed level. Secondly, the circuit may have value in co-cultures. A co-culture describes a beneficial interaction between two or more different populations of cells. Interestingly, we aimed to build on the work from Imperial iGEM 2016 where they built a STAR:anti-STAR mRNA controller for co-cultures.

How would our project be practically implemented?

To implement our project we have set out a simple three step process:

1. Identification of the key properties you might want for your POI (protein of interest)
2. Synthesis/order corresponding circuit
3. Transformation and usage of the circuit

1. Identification of the key properties you might want for your POI

To aid in the calculation of what circuit would be desirable, please follow the before questions and flow chart:

At what relative level would you want your protein to be expressed at?

• Use circuit 1 for low expression, 2 for medium, 3 for high

Is your protein highly toxic?

• Use circuit 1 for low expression

Is the circuit metabolic load important?

• Use circuit 1 for low expression

Would you need the steady state to recover quickly to perturbations?

• Use circuit 3 for high expression

This is just meant to act as a rough guide since a more detailed calculator could not be provided due to challenges in cloning. Given this, the high, medium and low circuits are determined by a simple 3 ODE model. As expression from Z1 increases and/or Z2 decreases the rate at which steady state increases (see modelling). So out of the 10 level 2 ECF20 circuits designed: ID6 has highest metabolic rate (circuit 3), ID3 has the lowest metabolic rate (circuit 1) and ID9 might represent an intermediate/medium circuit (circuit 2).

2. Synthesise the genetic circuit

a. We will ship you the plasmids directly

1. If the POI needs to be synthesised. Synthesis the POI with SAP1 adapter sequences either side (see Fig 7).
2. If the POI does not need to be synthesised, order the SAP1 adapter sequences (based design on adapters attached below). Different primer sequences for amplification and identification will vary based on the lab. Importantly, the part must have CIDAR MoClo overhangs and RE site for BsaI for this to work with our linear golden gate assembly protocol (see protcol).

Using a golden gate assembly (Start-Stop Assembly, Figure 8), insert the POI directly into the level 2 circuit (Taylor et al., 2019). Transform the plasmid and the resulting colourless colonies should be picked, liquid cultured and miniprepped (see protocols)

b. If the cost of long gene synthesis is NOT prohibitive

1. Order the level 2 sequence. In the place of module C CDS replace this with the POI CDS (see registry for link)

c. If the cost of large gene synthesis is prohibitive

1. Order all the parts online as either level 0 or level 1 parts and assemble via JUMP cloning (see parts and protocols).

Restriction enzyme

To allow users of our circuit to replace the reporter protein as the protein of interest for their application, we have created a new MegaT that with SapI recognition sites that allow users to replace the reporter protein in the dropout cassette with a useful protein they want to express. This Start-Stop assembly allows scarless assembly (Taylor et al., 2019).

The dropout cassette consists of a transcription unit that produces RFP. RFP is chosen because the dropout cassette of the JUMP vectors are superfolder GFP, red thus allow the ability to select for the cassette with colour contrast (Red/Green screening).

On the 5’ end, we have created a SapI recognition site just before the transcription unit with RFP, this allows SapI to cut from inside to the outside at the sequence ATG. On the 3’ end, we have created a SapI recognition site just after the transcription unit so SapI cuts at TAA, again cutting from inside to outside. Cutting from inside to outside is the key to the success of high yield in golden gate assembly. As the SapI cuts the drop out cassette off, the recognition site for SapI is removed together with the cassette and is not on the backbone. Now, as we put in the protein of interest with SapI site flanking its 5’ and 3’ ends to allow for SapI to cut inwards instead of outwards, the protein of interest inserted into the location of the original dropout cassette will not contain the SapI recognition site. As a result, once the product is formed, SapI will not digest it again, leading to very high efficiency golden gate assemblies. Since both JUMP plasmid level 0 and 1 acceptors contain SapI sites, replacement of RFP transcription unit drop-out cassette can only be achieved for the level 2 assemblies (don’t contain SapI sites).

3. Transformation and usage of circuit

• Once the circuit has been designed and build, transform and liquid culture the circuit to obtain a high concentration for miniprep.
• Directly transform the circuit into the cell of interest. Given that the circuit was build with proteins exogenous to E.coli, we recommend using an E.coli strain. For metabolic engineering use expression strains such as BL2/Rosetta or probiotic strains such as Nissle. This is dependent on the application.

Frequently asked questions (FAQ)

Q: At what relative level would you want your protein to be expressed at?
A: This is ultimately determined by a cost of factors; what you are using it for and toxicity and load etc

Q: Is my protein highly toxic?
A: Start with literature search but can be tested by expression your protein under a strong constitutive promoter and looking at cell health.

Q: Do I need my steady state to recover quickly to perturbations?
A: This depends on what the bacteria are used for or whether the cell is particularly sensitive to perturbations

Q: Is metabolic load important?
A: This depends on the protein/ additional proteins on the plasmid you want to express

Q: How I choose a reference point?
A: Refer to the flow chart above to choose a weak, medium or strong set point

Demonstration of implementation: Bath iGEM 2022

Our work in collaboration with the Bath iGEM 2022 directly demonstrates an example of implementation using our project. Briefly, their project involves release of fertiliser chemicals for soil based Bacillus bacteria. Specifically, malate induction drives expression of enzymes GlpQ and PhoD which are secreted. These act to break down and release phosphate for the plant.

A key challenge is that within their lab design they are testing their circuit at 37oC in a highly controlled nutrient conditions. However this poorly models the highly fluctuating environmental conditions of soil outside. To accommodate this we have collaborated to model how our circuit could be used as a homeostatic buffer reducing the impact of extrinsic fluctuations. In particular putting the secreting enzymes under control of our circuit should allow consistent and robust release of phosphate.

We initially considered what the circuit would look like, and the design requirements. Given that our circuit was made using an E.coli system, the circuit can’t be immediately applied to Bacillus since the controller species would not be exogenous. As such we began to identify possible controller species exogenous to Bacillus.

Safety aspects that need to be considered within the project

One of the major concerns in synthetic biology with many complex circuits that aren’t found in nature is that of escape. The engineered E.coli cells are self-replicating and the plasmids could be passed to native species through horizontal gene transfer. This could result in unpredictable and undesired consequences. Ideally, a well designed synthetic circuit would ensure that the host has to use the circuit and that the circuit could only be used within the circuit.

We believe that our circuits are unlikely to have a particularly high risk associated with them within our lab. Firstly, the circuit contains a certain metabolic load that would likely not confer any selective advantage. Secondly, it would be highly unlikely that any harmful proteins would substitute for our reporter gene. As such both the likelihood of conferring an advantage over a wild type population as well as the hazards associated with that danger are small.

This is not to say that there may be some biosafety considerations with their usages. Firstly, in metabolic engineering any danger would be dependent on what protein is being expressed. There are some biosafety strategies to prevent plasmid transfer/ bacterial spread that could be used such as re-factoring translation machinery so only the synthetic cargo can be expressed. Alternatively, the selection could be autotrophic for a key amino acid such as cysteine (Wright et al., 2013). Alternatively, the experimental nature of microbial drug delivery confers a number of biosafety risk but none dependent on the circuit we have designed.

References

1. Heins AL, Weuster-Botz D: Population heterogeneity in microbial bioprocesses: origin, analysis, mechanisms, and future perspectives. Bioprocess Biosyst Eng 2018,


2. Taylor GM, Mordaka PM, Heap JT. Start-Stop Assembly: a functionally scarless DNA assembly system optimized for metabolic engineering. Nucleic Acids Res. 2019 Feb 20;47(3):e17. doi: 10.1093/nar/gky1182. PMID: 30462270; PMCID: PMC6379671.


3. Wright O, Stan GB, Ellis T. Building-in biosafety for synthetic biology. Microbiology (Reading). 2013 Jul;159(Pt 7):1221-1235. doi: 10.1099/mic.0.066308-0. Epub 2013 Mar 21. PMID: 23519158.


4. Braat H, Rottiers P, Hommes DW, Huyghebaert N, Remaut E, Remon JP, van Deventer SJ, Neirynck S, Peppelenbosch MP, Steidler L. A phase I trial with transgenic bacteria expressing interleukin-10 in Crohn's disease. Clin Gastroenterol Hepatol. 2006 Jun;4(6):754-9. doi: 10.1016/j.cgh.2006.03.028. Epub 2006 May 22. PMID: 16716759.


5. Briat C, Khammash M. Perfect Adaptation and Optimal Equilibrium Productivity in a Simple Microbial Biofuel Metabolic Pathway Using Dynamic Integral Control. ACS Synth Biol. 2018 Feb 16;7(2):419-431. doi: 10.1021/acssynbio.7b00188. Epub 2018 Jan 30. PMID: 29343065.


6. Adolfsen KJ, Callihan I, Monahan CE, Greisen PJ, Spoonamore J, Momin M, Fitch LE, Castillo MJ, Weng L, Renaud L, Weile CJ, Konieczka JH, Mirabella T, Abin-Fuentes A, Lawrence AG, Isabella VM. Improvement of a synthetic live bacterial therapeutic for phenylketonuria with biosensor-enabled enzyme engineering. Nat Commun. 2021 Oct 28;12(1):6215. doi: 10.1038/s41467-021-26524-0. PMID: 34711827; PMCID: PMC8553829.