Safety by design
Safety is an integral aspect of the Cattlelyst project that we took into account from the very start. Our genetically engineered bacteria should be exclusively present inside our biofilter and unable to survive outside of it. This means that our task is to make our bacteria dependent on the controlled conditions inside the biofilter. This can be achieved both via the mechanical design of the biofilter and the biological design of the bacteria. This section focuses on the latter: how we use synthetic biology to ensure the safety of Cattlelyst.
To ensure biocontainment of our GMOs, we implemented three layers of safety:
The first two layers are based on unique features of the biofilter internal conditions: the high methane gas concentration (mechanism 1) and the high culture density (mechanism 2) of the bacterial biofilm. Additionally, the use of a co-culture allowed us to implement a third layer of safety based on co-dependency (mechanism 3). These mechanisms ensure total dependency of the two bacterial species on the biofilter conditions and on each other.
Methane-based kill switch
In the stalls were Cattlelyst is installed, the methane-rich gasses from the cows’ breath are directed into the biofilter from a specifically designed 'hood system' (see Project Description or Implementation page for a description of this system). The result is that the concentration of methane is higher inside the biofilter than in the surrounding environment. We used this difference in methane concentration to design our first safety strategy. By coupling the concentration of methane to a 'kill switch', E. coli can only survive in the ‘high methane’ environment of the biofilter and is killed outside of it.
We have modeled and engineered a toxin-antitoxin system in E. coli that senses the concentration of formaldehyde inside the cells, a direct intermediate in the metabolism of methane. As formaldehyde is a toxic compound, the native FrmR transcription factor is able to accurately sense its concentration inside the cell, and negatively regulate the Frm promoter (Pfrm). By binding formaldehyde, FrmR is released from Pfrm, lifting its repressing activity and thus activating transcription. We coupled the Pfrm activity to the hok/sok toxin-antitoxin system. This system contains the toxin hok, that can kill the cell from within by disrupting its membrane structure. The long-lived hok mRNA is regulated by a small short-lived antisense RNA, the sok antitoxin, that binds hok RNA and prevents its translation. Therefore, only cells that ongoingly possess the hok/sok system will survive .
To convert the repressing activity of FrmR on Pfrm into an activating signal for toxin production, LacI is placed under the control of Pfrm and toxin production is repressed by the presence of LacI and its corresponding promoter. Simultaneously, FrmR is directly used to activate sok production in high methane concentrations, and repress it when methane is low and hok is produced. A schematic representation of the circuit is displayed in Figure 2. This leads to one of the following situations:
- Inside biofilter
- Outside biofilter
In high methane concentrations, a high formaldehyde concentration will lead to the repression of toxin production and the transcription of antitoxin, so the cells are unaffected and remain alive.
In case E. coli escapes the biofilter, methane and thereby formaldehyde concentrations will drop. Toxin production is thus activated and antitoxin production repressed, leading to cell death.
To prevent cell death when the methane concentration temporarily drops, we have replaced the LacI promoter in front of the hok gene by a hybrid promoter that responds to input signals of both LacI and LuxR . This second input signal is part of the second proximity-dependent kill switch (see below). Only when LacI is low and LuxR is high, is the repression on Plux/lac lifted and the kill switch activated.
Proximity-based kill switch
As the methane-dependent kill switch is only incorporated in E. coli, a second safety mechanism is needed that kills P. putida if it escapes the biofilter. Therefore, we make use of a second characteristic of the biofilter: the cell density of the bacteria inside the biofilter is higher than outside. To ‘sense’ the cell density, quorum sensing is used, which is the ability of bacteria to regulate gene expression depending on the cell density. This leads to a second proximity-dependent kill switch, which makes the survival of both bacteria dependent on the high cell density of the biofilter. As described above, it is also coupled to the methane-dependent kill switch, so toxin production in E. coli is inhibited by a high methane concentration as well as a high cell density.
In our biofilter, P. putida will be producing the signalling molecule N-acyl homoserine lactone (AHL). This molecule easily diffuses out of the cell . Thus, when the cell density is high inside the biofilter, the AHL concentration is high as well. In our circuit, AHL is repressing toxin production and activating antitoxin production, which will make the cell survive inside the biofilter. Outside the biofilter, the opposite is true. The low cell density yields a low AHL concentration. This, in turn, leads to an increase in toxin production and antitoxin repression, killing any escaper cell. Figure 8 provides a schematic overview of the proximity-based safety circuit in P. putida. We built a similar system in E. coli, without the ability to produce AHL, making it dependent on P. putida for AHL, adding an extra layer of safety. More details can be found on the proximity-dependent kill switch page.
This safety circuit is coupled to the methane-dependent kill switch by means of a hybrid promoter. This promoter is activated by free LuxR and downregulated by LacI. Outside the biofilter the AHL concentration is low, giving a high free LuxR concentration and the methane concentration is low, giving a low LacI concentration. In E. coli toxin production is put under the control of this promoter, killing escaper cells.
In coculture conditions, to complement the safety circuits, the establishment of a cross-feeding community of E. coli and P. putida ensures the co-dependency of the two bacteria. This additional measure was designed to ensure the containment of the engineered bacteria also in case of mutations of the safety circuits previously described. The co-dependency between E. coli and P. putida was based on amino acids exchange and carbon-source dependency. Specifically, Tyrosine (Tyr) and Tryptophan (Trp) were found to be good candidates for the establishment of a cross-feeding community of the two bacteria. In this scenario, E. coli was knocked out for Trp and was engineered to overproduce Tyr, while P. putida was designed to be knocked out for Tyr while overproducing Trp (Figure 12). The implementation of this strategy showed promising results in vivo. More information can be found here.
In addition, P. putida was engineered to rely solely on the acetate secreted by E. coli for its carbon source (Figure 12), conferring an additional layer of safety to the system. The carbon-source dependency between E. coli and P. putida was characterised and it was shown not only to confer an additional layer of safety but to play an important role in the maintenance of the ratio between the bacterial strains during coculture.
Altogether, in the coculture condition, the double auxotrophy and carbon-source dependency, together with the previously described safety circuits, ensure the containment of the genetically engineered bacteria inside the biofilter.
- Thisted, Thomas, and Kenn Gerdes. 1992. “Mechanism of Post-Segregational Killing of Plasmid Rl by the Hok / Sok System Sok Antisense RNA Regulates Hok Gene Expression Indirectly Through the Overlapping Mok Gene” Journal of molecular biology, 223 (1): 41–54.
- Chen, Y. et al., (2018). Tuning the dynamic range of bacterial promoters regulated by ligand-inducible transcription factors. Nature communications, 9(1), 1-8.
- N. Saeidi, M. Arshath, M. W. Chang, and C. L. Poh, “Characterization of a quorum sensing device for synthetic biology design: Experimental and modeling validation,” Chem. Eng. Sci., vol. 103, pp. 91–99, 2013, doi: 10.1016/j.ces.2012.12.016.