Since the beginnings of synthetic biology, there have always been questions to be raised when genetically modifying organisms: Do they pose a danger to humans, animals or nature?
Therefore, in almost all nations in the world the introduction of genetically modified organisms (GMOs) into the environment is strictly regulated (for more information have a look at our talk with Dr. Ulrich Ehlers)[1]. There are good reasons for this: Genetically modified, non-native organisms can have completely unforeseen effects when entering a novel ecosystem. As an example, an inserted antibiotic resistance might be passed on to other bacteria, leading to multiple resistances. This raises the important question: How do we prevent these organisms or their recombinant genetic material from entering the environment?
Most importantly, when working with GMOs, strict attention must be paid to the so-called biocontainment. In general, GMOs are not supposed to leave the designated facility. In the laboratory, disinfection or autoclaving of the equipment are the most common methods to efficiently neutralize all remaining microorganisms. Working inside of special rooms like cleanrooms or at a sterile bench can not only improve the quality and purity of a product, but can further reduce the risk of a GMO escaping the facility.
In contrast to the physical and chemical methods previously mentioned, there is also the option for a biological containment method, called kill switch. A kill switch is a genetic circuit used for selective elimination of a GMO[2]. The GMO is supposed to stay in its defined area. When it leaves the defined area, a genetic switch is triggered, leading to the death of the microorganism.
Variants and Precursors
There is a need for biocontainment systems that couple environmental sensing with genetic switch-based control of cell viability. These systems are essential to prevent the escape of GMOs into the environment and thus the responsible application of biotechnology in direct contact to our environment. The genetic switches, generally referred to as kill switches, need to be reprogrammable, in order to change for example their environmental inputs, regulatory architecture and killing mechanism. In the scientific literature, only a few publications deal with kill switches. The best known publications are called "Deadman" and "Passcode", published by the group of James Collins[2]. Both kill switches were designed for and constructed in Escherichia coli. The Deadman kill switch uses a transcription-based, monostable toggle switch system to ensure fast and robust killing of target cells. In the absence of an inducer not found in nature, e.g. anhydrotetracycline. In contrast, the Passcode kill switch utilizes hybrid LacI-GalR family transcription factors to implement complex environmental requirements for cell survival. Nevertheless, there are further different possibilities for the implementation of genetic kill switch systems. A popular and frequently used approach is the utilization of toxin-antitoxin systems:
This system, as the name suggests, consists of two components: The toxin and the corresponding antitoxin. The two genes that encode these components may be present on a plasmid. If the cells are not under selection pressure, the plasmid will be lost in the next generation and they will die, because the gene that encodes the antitoxin is not present. Thus, the antitoxin is not further produced, remaining antitoxins are degraded and the toxic proteins remains active and kills the new cell[3].
A well-researched toxin-antitoxin system is the mazEF system in E. coli. But mazEF -like modules are also found in the genome of many other bacteria including pathogens[3]. The natural function of the mazEF system is to prevent the spread of a bacterial phage infections. Thereby MazF is a sequence-specific RNA endoribonuclease and thus a toxin that initiates a programmed cell death pathway. Its counterpart, the antitoxin MazE binds to MazF and renders it inactive, but also also prone for enzymatic degradation (see fig. 1)[4].
Figure 1: Mechanism of the mazEF system[4]. The toxin MazF and the antitoxin MazE are produced within the cells. Binding of MazE to MazF inhibits toxic activity of MazF. Consequently, in a balanced system the cell is viable, but is killed upon excess of MazF.
Various iGEM teams have built on that the mazEF and implemented their own ideas[5, 6, 7]. iGEM Wageningen 2019, for example, has used the mazEF system together with the KaiABC system for circadian oscillations in gene expression[5]. One common problem with toxin-antitoxin systems is the basal expression of the toxin gene due to promoter leakiness and consequently unwanted cell death.
A second example for the use of the mazEF system is provided by iGEM Munich 2012. They used the mazEF system in combination with sigma factor G of Bacillus subtilis to kill sporulating cells. However, the project struggled with too strong expression of the toxin and a consequent imbalance of toxin and antitoxin[6].
Another popular approach is engineering auxotrophy: One way to overcome the difficulties of the toxin-antitoxin system is to knock out an essential gene. Essential genes are genes that encode cellular components that are vital for the life of the cell. If essential genes are knocked out, the cell can only survive if it can compensate for the function of the missing component, e.g. synthesis of an essential metabolite, from other sources, e.g. uptake of supplemented metabolites from outside of the cell. Consequently, if bacteria escape into the environment, they cannot compensate for their auxotrophy due to the absence of supplemented essential molecules[8].
iGEM NCKU Tainan 2015 used synthetic auxotrophy to switch off the dapA gene encoding 4-hydroxy-tetrahydrodipicolinate synthase. This enzyme is a part of the diaminopimelate pathway and therefore plays an important role in cell wall synthesis[9]. If dapA is knocked out, diaminopimelate (DAP) cannot be synthesized and the cells can only grow by taking up external supply of DAP. If the cell leaves the controlled conditions into the environment, it will not survive[7].
However, engineering synthetic auxotrophy was not feasible for our project, as these components cannot be easily added in the wastewater treatment plant as the components would be flushed away, before they could enter the cells. Therefore, large amounts of essential molecules would have to be used, which is not cost-effective and not practically feasible. Furthermore, the auxotrophic molecules would then float in the water, and microorganisms could take them up after leaving the biofilm. Consequently, there would be no guarantee that the microorganisms would actually die after leaving the biofilm. That is why we thought of something else: Our approach is based on the idea to use intercellular communication with an of an inducible gene switch to control an essential gene. The expression of the essential gene is genetically modified in a way that it can be controlled by a promoter induced by high cellular density. Within the biofilm, the molecules necessary for the transcription of the essential gene are present, the protein is synthesized and the cell is viable. If the cell leaves the biofilm, the necessary molecules inducing expression of the essential gene are missing and the cell consequently cannot synthesize the essential components and thus cannot survive.
Evolution of our Kill Switch Design
We developed a concept for a kill switch model to hinder the wash-out of living organisms from the biofilm. Thereby our concept went through five phases. In this section, we present our reasoning, show which decisions were based on which essential parameters and how the design of our kill switch has evolved over the course of our project. The final design or our "phase 5" in the evolution of our kill switch design can be seen in the section Final Kill Switch: Design and Function.
Phase 1
Our initial idea (fig. 2) for the basic design of a kill switch consisted of joining two strains of B. subtilis in a communal biofilm. In each strain, an essential gene is knocked out in the genome and the lack of the gene is bridged by the expression of the respective gene from a plasmid. Essential gene expression is controlled by orthogonal promoter systems which are induced upon binding of small molecules produced by the respective communal partner strain. Consequently, the essential genes are only expressed, if cells of the two strains are in close proximity to each other, which allows the transfer of inducing small molecules in the biofilm. Only if both inducing molecules are bound, the expression of the essential gene is induced. Cells leaving the biofilm leave an environment with inducer concentrations enabling essential gene expression, which will eventually lead to cell death. The essential gene we utilized first was dnaA since it encodes a protein that plays an important role in initiating DNA replication in B. subtilis. During phase 1 of our design creation process we switched to rpsB, because of several diverse factors. Read about our thought process in the associated section: Which Essential Gene do we Knock Out?
Figure 2: Phase 1 design of our kill switch. Two joined strains of bacteria communicate through two different inducer molecules. Furthermore, an essential gene is knocked out on the genome of each strand. If both molecules are present in one of the cells, the alternative essential gene, that is located on the plasmid that the strains were transformed with beforehand, is expressed. If only one inducer is available, the essential gene is not expressed and the cells die.
Phase 2
After some research, we learned about the process of quorum sensing, which is already functional in various bacterial species[10]. Quorum sensing is a fitting process for the function we want to create: First, the individuals of the biofilm are dependent on the community of cells. Strains that use quorum sensing produce autoinducers to regulate gene transcriptions. The amount of these autoinducers secreted by the cells is proportional to the cell density in the medium, because a higher cell density represents more cells that are capable of producing and secreting the inducer substance. If the concentration reaches a certain threshold concentration, the inducer is sensed by surrounding cells and this step initiates a phosphorylation cascade inside the respective cells. The cascade terminates with the activation of a regulator protein which finally alters the expression rate of the specific genes under control of quorum sensing (Background - Quorum Sensing).
The big difference to our initial idea in phase 1 is that quorum sensing only utilizes one universal autoinducer, namely ComX[11] of the comQXPA system(Background - comQXPA system). Quorum sensing is a well understood, wide spread and already applied process which is why we decided on using it[12]. There are other crucial benefits of quorum sensing: It is used to regulate certain genes transcription and e.g. start their expression in high cell density situations. Because of these factors, we do not have to implement a system as explained in our initial idea in phase 1, which especially saves us time and additional work. Since quorum sensing is a system that can be naturally found in B. subtilis[12], we hypothesize that this basic system itself will not fail. But because this is just a hypothesis build up by our team, read here more about the critical aspects and challenges. Another challenge is that when working with plasmids in the gram-positive and naturally competent soil bacterium B. subtilis, genetic instability and the loss of plasmids over time are present issues that can take place[13]. The less external DNA we have to transform into the bacteria, the better for the final result.
Due to the many advantages of the quorum sensing system, we created a new design of our cassette. In our design of phase 2,
the quorum sensing promoter PdegQ[14] is utilized to control the expression of the essential gene rpsB.
Naturally, the promoter is activated by the phosphorylated ComA regulator protein ComA-P and plays an important role in exoprotease synthesis[15]. Consequently, in low cell densities and thus low ComX inducer concentration, ComA is not phosphorylated and expression activation of the essential gene under control of PdegQ is not possible. Only in high cell densities and corresponding high ComX levels, the regulator protein ComA is phosphorylated to form ComA-P and thus gene expression of our essential gene is activated.
Figure 3: Phase 2 design of our kill switch. While the cell is in the biofilm, comX concentration is high enough for comx-P to bind on the degQ promoter, gene expression of the essential gene can take place. If a single cell detaches from the biofilm, no more ComX is available. Gene expression cannot take place, the cell dies.
At this time, another important question arose: What about the first phase of biofilm growth?
If the cell density is still low, the concentration of the autoinducer ComX is likewise also low. Consequently, the signal transduction is low or absent and we likely do not receive sufficient expression of the essential gene in our cells. As a result, the cells are likely not viable on their own in this first stage of growth.
Phase 3
We thought to solve this problem by adding a second copy of the essential gene to the previously designed plasmid. In contrast to the first copy, the expression of the second copy of rpsB is controlled by an inducible promoter. In practice, we would add an inducer to activate expression of the essential gene during the growth phase, which most likely has to take place outside the actual area of application anyway. The demands on a suitable inducer and the corresponding inducible promoter were:
1) The inducer should be inexpensive and should not remain in the cells indefinitely.
2) The promoter should only activate the transcription once the inducer is added.
Vice versa, in absence of the inducer the transcription level, referred to as basal expression, shall be as low as possible. Promoters with low basal expression are also called tight promoters, which are commonly desirable for synthetic genetic circuits like our kill switch.
We chose PSalTTC[16], as it acts as an orthogonal promoter with very little crosstalk to the other promoter, which we integrated into our cassette later on during the project (Phase 4).
Figure 4: Phase 3 design of our kill switch. A second copy of our essential gene rpsB is located on the same plasmid as the cassette from phase 1 of our design evolution. This rpsB gene is under control of an inducible promoter to survive the growth phase of the biofilm. In this phase, the cell density is still low and consequently, the concentration of the autoinducer ComX is also low. Without this extra copy of rpsB with the respective inducible promoter, the cells would not be viable.
An alternative for the inducible promoter to ensure that the cells survive the growth phase would have been the direct addition of the autoinducer ComX from outside. However, we decided against this idea, because buying this substance would not be affordable for wastewater treatment plants (WWTPs) and producing ComX ourselves would be very costly and time consuming, due to the eleven coupling steps and the necessity to establish a solid-phase synthesis scheme[17]. It would generally be possible and easier to manufacture the inducer ComX directly in cells.
Eventually, it is difficult to evaluate, when the cell density is high enough for the cells to self-regulate the expression of the essential gene without an added inducer. High cell density corresponds to high ComX concentration, but the estimation of the threshold concentration of ComX for the activation of the quorum sensing cascade and the corresponding promoter PdegQ is a challenge, when the experiments are only planned in silico and hardly any experimental data for this topic is available. If possible, we would like to generate data for these topic ourselves in the future with our kill switch model.
Phase 4
As we continued to work on the design, a larger problem occurred: Not only is our recombinant plasmid highly homologous to the essential gene in the genome itself, but the plasmid contains internal homology through the dual presence of the essential gene on it, each under the control of a different promoter (the inducible promoter PSalTTC for the growth phase and PdegQ for the final kill switch application).
This is problematic for us for the following reason:
While thinking about the later implementation of the kill switch, we became aware of the advantages of genome integration, especially since we work with B. subtilis. In the process of a homologous recombination, many different sections could recombine with each other, if multiple homologies are present, and thus we are less likely to achieve the desired integration of our cassette into the B. subtilis genome. Based on this information, we wanted to find a solution in the following design development process.
The method of homologous recombination also brought us into contact with the topic of recombinases. These enzymes can be used to remove selection markers such as antibiotic resistances from the integrated segment in the genome[18]. The markers are no longer needed after the integration and only lead to unnecessary metabolic burden. But there is more to know about recombinases: Their abilities are dependent on the existence of lateral flanks marking recombination sites called recombinase target sites (RTS). Genome sites labeled by RTSs can be manipulated by recombinases in several ways[18]. Furthermore, the function of the recombinase is highly depended on the relative orientation of these RTS flanking the area of interest on the genome. If the RTSs are arranged in a head-to-tail fashion, the area within the RTSs is excised (called “resolution”) and both RTSs are fused. If the RTSs are arranged in head-to-head orientation, like in our final kill switch design, the sequence within the RTSs is quasi flipped (called “inversion”).
The following figure shows our design for the
phase 4
in detail.
Figure 5: Phase 4 design of our kill switch. Schematic representation of the designed kill switch cassette of phase 4. RpsB expression is under the control of a upstream DNA region (red) flanked by the two recombinase sites (lox 66 and 71, grey arrows) aligned head-to-head. The sense strand encodes the inducible promoter PSalTTC and the antisense strand encodes the inducible promoter PdegQ. The addition of cuminic acid induces the expression of the Cre recombinase. As a result, the DNA region is inversed and the inducible PdegQ starts expressing rpsB.
Eventually, when we discussed our concept of
phase 4
with our advisors of the Kabisch lab, we realized that minor changes are still necessary to enable a straightforward genomic integration of our kill switch design in B. subtilis. Those changes and the complete final design are explained in
Final Kill Switch: Design and Function.
Final Kill Switch: Design and Function
Our concept is challenged by bringing together cell growth from low to high cell densities and a kill switch design which is activated by high cell density. Although the design appears contrary on first view, we aim to enable biofilm growth under controllable conditions during which the lethal function of the kill switch cannot be activated.
The bacteria are then genetically altered by a quorum sensing signal upon crossing a threshold in cell density, what leads ultimately to the setup of the kill switch. This is accomplished by a novel cassette design:
It consists of the chosen essential gene, in our case encoding the ribosomal protein RpsB (Which Essential Gene do we Knock Out?)
which is under control of an engineered promoter region. This promoter region encodes two promoters. One promoter is encoded on the sense strand, the second one is encoded on the antisense strand. Upon passing a cell density threshold, the promoters are ready to be flipped.
An inducer from the outside activates a recombinase which flips the promoter region and thus exchanges the promoter on the sense strand by the one previously encoded on the antisense strand.
By this, the essential gene can be controlled by to different promoters that can be interchanged through recombinase activity. As mentioned above, the twist with this design approach is that the promoter region controlling the expression of rpsB can by inverted by the recombinase.
We named the system “recombinase induced promoter exchange (RIPEX)”.
The RIPEX system and the engineered promoter region consist of two promoters: For the growth phase the constitutive promoter Pveg controls the essential gene rpsB. As already mentioned above
Evolution of our Kill Switch Design
, the growth of the bacteria can be controlled by this constitutive promoter.
The second promoter PdegQ naturally controls expression of the gene degQ, with the help of quorum sensing. In quorum sensing, ComA is phorsphorylated
(Background - Quorum Sensing).
by the phosphorylated regulatory protein ComA-P. PdegQ is located on the antisense strand in contrast to the constitutive promoter Pveg which is located on the sense strand inside of the engineered promoter region. It is intended to be used by the bacteria as soon as the cell density has exceeded a certain threshold value and the inversion performed by a recombinase is induced by Xylose.
In conclusion, this leads to an irreversible and unidirectional inversion of the promoter region. After the inversion, the essential gene is exclusively and irreversibly under the control of the quorum sensing promoter PdegQ. Consequently, the constitutive promoter Pveg, who is located on the antisense strand after inversion of the combined promoter region, should not lead to expression of the essential gene rpsB.
As stated before, there is another inducible promoter PXylA inside the cassette, controlling the expression of the recombinase. Through induction with Xylose, it can be used to specifically control, when the rotation of the combined promoter region and the switch from the constitutive to the quorum sensing promoter should take place. Or in other words: when the RIPEX system should be activated.
This is accomplished by using the Cre recombinase (Cre/lox-system) and two mutated recombination sites designed by the 2007 Paris iGEM Team lox66 (upstream of the desired region, BBa_I718016) and lox71 (downstream, BBa_I718017). After inversion, the sequences are altered and the recombinase does not recognize them as sites, which indicate a possible inversion. In conclusion, this leads to an irreversible and unidirectional inversion of the promoter region.
Through this novel technique , the problem of homology mentioned before
Evolution of our Kill Switch Design
is overcome and the early growth phase can be survived by the bacteria in a controllable way (before the cell density and therefore the concentration of the autoinducer ComX is high enough for the quorum sensing promoter to function).
After the inversion through the Cre/lox-system, the cells are exclusively depended on the high cell density and the community of the biofilm, because the cells receive only within the biofilm necessary for the expression of rpsB.
When an individual cell leaves the biofilm, the concentration of ComX in its surrounding environment is lower than inside of the biofilm. The essential gene is not expressed as well as inside the living community and in the expected case the cell dies upon leaving the biofilm.
Which Essential Gene do we Knock Out?
Since our kill switch is based on linking the availability of an essential gene (product) to the presence of a biofilm, this essential gene first had to be knocked out. In our first approach we chose the gene dnaA . Since it encodes a protein that plays an important role in initiating DNA replication in prokaryotes[19], its absence leads to cell death. However, at the end of the first phase of our kill switch design we noticed that the choice of this gene caused a problem: dnaA is located in an operon containing another gene[19, 20]. In the case of a knockout by destroying the promoter, the entire operon would no longer be transcribed. This could have led to unpredictable complications. In addition, DnaA is only essential for cell division, so it is possible that the microorganisms do not die but go into dormancy instead.
For this reason, we searched for an easier accessible gene - and found rpsB. Its gene product is the ribosomal protein S2 in the 30S subunit of prokaryotic ribosomes. As one of the biggest ribosomal proteins in the small subunit[21], RpsB is essential for the functionality of the ribosomes and thus for translation. When a bacterium leaves the biofilm, RpsB will no longer be produced due to our killswitch. This leads to the bacterium being unable to keep up metabolic function, resulting in cell death. Another advantage is that rpsB is the only gene under control of the corresponding operon[22], therefore being much easier to control.
Integration of our Cassette
The implementation of additional, often heterologous, genes in organisms has always been a key step in Synthetic Biology. Several ways to insert new genes in organisms exist. Usually for prokaryotes, a plasmid with the genes of interest is introduced into the bacterium. With a corresponding origin of replication (ori), a fitting promoter region controlling the expression of the gene of interest, and an antibiotic resistance for selection, such a plasmid can often be very effective. Yet, several bacteria cannot robustly carry plasmids. In these cases, the genomic integration of the gene of interest is recommended.
Although B. subtilis can carry plasmids, its highly efficient homologous recombination machinery[23]
allows the integration of our expression cassette in the genome. The cell normally uses this mechanism to repair DNA double-strand breaks, by using a homologous DNA fragment as a blueprint. This machinery also works without the generation of double-strand breaks. Therefore, genes of interest can be integrated directly into the genome of B. subtilis. Only a plasmid with double-stranded DNA, containing our genetic cassette, an antibiotic resistance for selection and homologous sequences to the genome, are required for insertion (see fig. 6)[24]. Theoretically a homology of ~70 base pairs is sufficient, however, we decided to use around 500 base pairs of homologies for both flanks after talking to our advisors from the Kabisch Lab.
We decided to use this technique to integrate our kill switch cassette into the genome of B. subtilis. This additionally prevents our cassette from being lost during replication, which is a major problem when using plasmids, as a constant selection pressure has to be maintained.
Figure 6: Homologous Recombination
Critical Aspects
There are many aspects to consider when designing and impelementing that have to be payed attention to in the design and implementation of a complex kill switch like ours.
First of all, the genetic code might not be stable over time.
Mutations, recombination and other cellular mechanisms may alter the genetic code.
All genetic circuits are susceptible to mutation.
A single point mutation could deactivate an entire circuit, rendering it inactive.
However, there have been improvements in genetic stability, especially in kill switches[25].
Another problem depends on the Cre recombinase and its recombination sites.
If the recombination process loses its irreversibility, the whole kill switch could fail, as the constitutive promoter might get switched back in front of the essential gene.
That way, the essential gene is not under the control of the quorum sensing dependent promoter anymore, and the B. subtilis will not die when leaving the biofilm.
The possibility of this problem depends heavily on a mutation in the cre gene and the lox 66 and lox 71 recombination-site before first inversion[26].
The safety of this kill switch also depends on whether the bacteria die after leaving the biofilm or only enter an inactive phase. We planned experiments to answer this question but because of the COVID-19 pandemic we were not able to execute them.
The fact that rpsB is an essential ribosomal protein leads to the presumption that the bacteria will die, as most likely its entire metabolic functions will collapse[27].
Another potential downfall could be the comQXPA pathway itself,
since it is also conditionally required in sporulation processes[28].
The question is whether the kill switch leads to infertile endospores. Due to the restrictions in 2020, we could not answer this question with experiments.
Future Testing
Our concept of a kill switch is solely theoretical, because we were not able to work on the kill switch in the wetlab due to the COVID-19 pandemic. For the practical continuation of this project, it is essential that we can test the functionality of the kill switch. A critical parameter of the kill switch to characterize is for example the threshold of cell density at which the kill switch system is activated.
Most importantly, we have to ensure that our kill switch does what it is supposed to do: Killing our genetically modified Bacillus subtilis cells upon leaving the biofilm environment. We expect that the cells will die, if the essential ribosomal gene rpsB is not expressed, because many bacterial antibiotics inhibit the protein biosynthesis or ribosomal proteins. Thereby, some antibiotics are able to kill cells by blocking the activity of these important factors. [29, 30, 31] Nevertheless, we cannot be fully sure whether knocking out the essential gene kills the cells or only inhibits their growth. Therefore, we have to test in the laboratory, if the cells are viable without the expression of the essential gene. This can be done via live/dead staining. In our planned assay, cells with damaged membrane, which are considered dead or dying, show a red staining, while cells with intact membrane are stained green upon incubation with SYTO 9/propidium iodide stain.
Further, we have to show experimentally, at which cell density the cells produce enough of the autoinducer ComX to be viable with the quorum sensing promoter PdeqQ solely controlling the expression of the essential rpsB gene. For this purpose, the cells are grown in separate test flasks up to a defined cell density. In the next step the expression of the recombinase is induced to initiate the irreversible inversion of the promoter region. Thereby the constitutive promoter in front of the gene is replaced by the ComX-dependent promoter. We hypothesize that if the cell density is not high enough, the induced switch of the promoter region would lead to the death of the cells. If it is the case, the cell density is high enough after the inducer has been set down for the growth phase, the cells should be able to live without a without an additional inducer. Cell viability and consequently the expression of the essential gene can be determined by live/dead staining, as already described above.
