Team:Exeter/Project
Introduction
Bio-safety is an important issue in synthetic biology. Concerns surrounding synthetic organisms escaping into the environment have prompted the development of novel methods of bio-containment. Many iGEM projects that require an organism to be released from the lab use kill switches to address concerns about the effect of genetically modified organisms (GMOs) on the environment. Unfortunately, kill switches - inducible genetic devices that cause cell death - are poorly categorised in the standard registry of genetic parts. There is a distinct lack of quantitative data which prevents them being used with confidence.
Our initial concerns surrounding the use of kill switches as bio-safety devices were centred
around their efficiency, robustness over time and ability to prevent transfer of synthetic DNA into the
wild population. We contacted Dr Markus Gershater, chief scientific officer at Synthace Ltd, to ask him
what the application of kill switches might be in an industrial setting and what evidence would be satisfactory
for their use. Dr Gershater gave the view that kill switches would not be as effective or economical as the
physical and chemical bio-containment methods that Synthace currently employ. One of his concerns was that
any leakiness in a system would provide a strong selection pressure against cells with fully functional kill
switches. In order for Dr Gershater to be satisfied, the kill switch would need to be tested across a wide
range of environmental conditions and still remain effective. He would also need to see high efficiency levels,
as in the context of large culture vessel, even a low proportion of cell survival would result in a large
population lacking an effective kill switch.
There’s no need for kill switches in the industry area that we operate. Even in the case where we might think there was a need for extra containment, the idea of putting in multiple systems to a production strain is concerning
-Dr Markus Gershater, Synthace Ltd
We also contacted Dr Tom Ellis of Imperial College London and
asked his opinion on kill switch reliability. He gave the view that combining multiple mechanisms could
greatly reduce organism escape rates after kill switch induction. Combining several kill switches was an
approach we had discussed during the development phase of our project. Dr Gershater advised that the
different systems would need to be truly orthogonal. For example, in an industrial setting two different
kill switches that both rely on protein production could potentially be circumvented by the over expression
of a useful enzyme that is being commercially produced.
There is of course no such thing as completely (infinitely) efficient, so instead it is best to aim for the most efficient possible system - e.g. chance of escape less than 1 in 1020. This is not unrealistic because multiple mechanisms can be combined that each have less than 1 in a billion escape rates, which gets you into this territory.
-Dr Tom Ellis. Imperial College London
We spoke to Professor Richard Titball leader of the Microbes and Disease research group at Exeter University. We asked him about potential applications of kill switches. He talked about how physical containment methods traditionally used in microbiology may have limitations when applied to vaccines, an area of his research, as they are administered to the population. He thought that if triggered by a specific environmental condition, a kill switch could be an elegant bio-containment solution as it is a system that can be finely tuned. In practice however he was skeptical that kill switches could be made reliable. Interestingly when we discussed the public perception of synthetic DNA and its potential release into the environment, Prof Titball believed that it was an issue that shouldn't be discussed by the scientific community alone, but that the public should be involved in the risk/benefit assessment of the use of genetically modified organisms. This prompted us to find ways to engage the public in order to better their understanding of synthetic biology and include them in the debate. You can see our interview with Prof Titball on the Exeter iGEM 2016 youtube channel.
This is not a zero risk activity, on the other hand there are massive benefits that might be realised from GMOs. The only way I can see of moving us forward is to educate the public a bit more about the issues.
-Prof Richard Titball. University of Exeter
After talking to individuals from industry and academia about the strengths and limitations of kill switches, it was clear that there was not a consensus on their efficacy. We decided to investigate the effectiveness of different types of kill switch, to quantify their robustness after several generations and investigate the possibility for horizontal gene transfer (HGT). We aimed to test if multiple kill switches in a system would reduce failure rate and if integration into the genome would increase stability.
We have developed three types of kill switch to cover a broad range of strategies that may be employed in kill switch design:
- A metabolic kill switch that uses the production of reactive oxygen species to kill the cell
- An enzymatic kill switch that uses the production of lysozyme (Gallus gallus)
- And two systems that target DNA:
- A CRISPR based kill switch based on the work of Caliando and Voigt (2015) that targets specific sites for degradation
- A broad spectrum system that non-specifically degrades DNA mediated by expression of DNase 1
Continuous culture
Before starting the project we spoke to Prof. Robert Beardmore EPSRC Leadership Fellow in the Mathematical
Biosciences at Exeter University. Much of his research has been into antibiotic resistance. We discussed how high
selection pressure is applied by prolonged use of antibiotics and how kill switches may be analogous to this.
It is clear that cells which develop a mutation that inactivates the kill switch would be strongly selected for.
It was estimated that functional loss of the kill switch would occur in a short amount of time as a result,
and if this was the case, could have strong implications for kill switch longevity.
To test this we decided
to use a ministat (Fig. 4) to perform a continuous culture. The ministat was developed in the Dunham Lab at the
University of Washington (Miller et al, 2013). Each ministat chamber is fed from its own media container (Fig. 1)
via a peristaltic pump (Fig. 2), with the culture volume set by the height of the effluent needle in the chamber (Fig. 3). Air is
bubbled through flasks of water to hydrate it and then used to agitate the culture. Chambers were inoculated with
freshly transformed E. coli BL21 (DE3) and samples taken to test if the kill switches were still viable.
By simulating in miniature how a kill switch might behave in an industrial setting, the ministat provides a proof
of concept for how a kill switch might be maintained in larger chemostats during a continuous culture. A protocol
for running experiments in the ministat can be found here
Fig. 1 Media container used to feed a single ministat chamber.
Fig. 2 Peristaltic pump
Fig. 3 Ministat chambers in heatblock and 1 litre Duran bottle used to collect effluent
Fig. 4 Ministat running a preliminary
experiment to calibrate parameters such as dilution rate and temperature of the heat block. 50 ml burettes used
here to accurately measure effluent levels. 1 litre Duran bottles were used for effluent collection in the main
experiment due to greater volumes of effluent.
Our own growth curve was performed to determine the maximum specific growth rate of E. coli BL21 (DE3) in our lab, but could not be conducted for a sufficient length of time to be accurate. A maximum specific growth rate value of 1.730 was used (Cox, 2004). The ministat must be run at a dilution rate less than maximum specific growth rate, this prevents the culture being washed out of the growth chambers. The dilution rate of the culture was calculated by measuring flow rate at a setting of 7.5 rpm on the peristaltic pump. For practical reasons the pump could not be run faster than this due to the amount of media needed. The dilution rate was set at 0.2 which produced cultures that grew at an average optical density (OD) of 3.47 for KillerRed samples, 3.64 for KillerOrange samples and 3.17 for lysozyme samples. OD was measured daily with a Bug Lab OD scanner. When the same sample was measured in a tecan infinite 200 pro plate reader the Bug Lab showed readings approximately three times higher. The difference between the samples was consistent regardless of the method used to measure OD and OD measurements remained stable throughout the experiment.
Method and Results
Metabolic kill switch
KillerRed (Fig. 5) and KillerOrange are homologues of Green fluorescence protein (GFP) which, when irradiated with green and blue light respectively, generate reactive oxygen species (ROS). KillerRed has been shown to effectively kill cells when exposed to green light (540–580 nm) and is much less effective under blue light (460–490 nm) (Bulina et al, 2006). KillerOrange effectively kills cells when exposed to 450-495nm (Sarkisyan 2015), the range that KillerRed does not. There is a β-barrel present in both these proteins. A water-filled channel that is in contact with a chromophore area and located at the cap of the said β-barrel is thought to confer these proteins with their phototoxic capabilities (Pletnev S et al, 2009).
Fig. 5 Structure of KillerRed showing the open channel which allows contact between the chromophore and the solvent.
The mechanism by which ROS kill cells isn’t completely understood. However ROS have been shown to have various detrimental roles in cells such as the oxidation of thiols, ascorbate and proteins containing (Fe-S)4 clusters as well as reducing various transition metals (Farr and Kogama, 1991). Prolonged exposure and or high levels of ROS triggers apoptosis like mechanisms (Held, 2015).
Our metabolic kill switches build on previous iGEM projects which have used the expression of highly phototoxic
fluorescent proteins to kill the cells by exposing the culture to light. In 2013, the iGEM team from Carnegie Mellon developed a phage delivery system of the KillerRed gene, which was then expressed in the infected bacteria, killing it on exposure to light. Carnegie Mellon 2014 continued characterisation of KillerRed and its monomeric form Supernova by analysing their photobleaching characteristics. Neither team tested the longevity of the kill switch or provided details on the light intensity that the cultures were exposed to. We aim to quantify the length of time for which the kill switch remains viable and provide absolute values for the intensity of our light source.
Firstly we improved KillerRed, an existing registry part, by codon optimising it for E. coli and
improved its characterisation by exposing cultures expressing the protein to previously untested light intensity. We designed and submitted KillerOrange as a new part and tested it in the same way. We compared the phototoxicity of KillerRed and KillerOrange to the commonly used Red fluorescence protein (RFP). This was to determine how phototoxic KillerRed and KillerOrange are in relation to RFP and if prolonged excitation of a commonly used fluorescent protein is deleterious to bacterial growth. Once we had established the efficiency of KillerRed and KillerOrange, ministat chambers were
inoculated with samples of each to determine the robustness of the kill switches over time.
Method
The following samples were tested for phototoxicity by exposing them to 1.2 mW/cm2 of white light from a 4x8 LED array (Fig. 6) for 6 hrs. Samples were then spread plated and colony forming units (CFUs) were counted. All parts were carried on the pSB1C3 plasmid and transformed into E. coli BL21 (DE3). Samples that were induced were done so with Isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.2 nM.
Henceforth samples will be refered to as:
- KillerOrange induced
- KillerOrange not induced
- KillerRed induced
- KillerRed not induced
- RFP
- Control: BL21 (DE3)
See protocol for detailed method.
KillerRed is excited by green/yellow light (540-580 nm) and KillerOrange by blue light (460-490 nm). We constructed a box around the LED array to prevent ambient light entering. Access to the inside of the box was gained through an opening cut in the front. With help from Ryan Edginton, we used a portable spectrometer (Ocean Optics USB2000+VIS-NIR-ES, connected to a CC3 cosine corrector with a 3.9 mm collection diameter attached to a 0.55 mm diameter optical fibre) to measure light spectra and absolute intensity in the visible range.
Fig. 6 Light box and thermocouple
Fig. 7 A graph showing the relative intensities of the spectra for our light source
Metabolic Kill Switch: Results
Characterisation experiment
The following graphs show the average percentage viable cells for induced and uninduced samples after 6 hrs of exposure to 1.2 mW/cm2 of white light and an average temperature of 38.63 °C. CFU count for the control condition was treated as 100 % and viable cells calculated as a proportion of that value. CFUs were not counted above 300. Error bars represent the standard error of the mean. Little to no difference in CFU count is shown between the control (BL21 (DE3)) and kill switch samples when they are kept in the dark. There is a significant difference in the number of CFUs when the kill switch samples are exposed to light.
Fig. 8. Percentage viable cells after 6 hrs in the light box. BL21 (DE3) transformed
with pSB1C3 containing an RFP maker is compared to a control with no plasmid. 10 ml falcon tubes containing
4.5 ml of sample were covered in tin foil before being placed in the light box.
Fig. 9. Percentage viable cells after 6 hrs in the light box. BL21 (DE3) transformed
with pSB1C3 containing an RFP maker is compared to a control with no plasmid. 10 ml falcon tubes containing
4.5 ml of sample were placed label down in the light box to allow maximum exposure to the light.
Fig. 10. Percentage viable cells after 6 hrs in the light box. BL21 (DE3) transformed
with KillerRed (BBa_K1914003) is compared to a control with no plasmid. 10 ml falcon tubes containing 4.5 ml
of sample were covered in tin foil before being placed in the light box.
Fig. 11. Percentage viable cells after 6 hrs in the light box. BL21 (DE3) transformed
with KillerRed (BBa_K1914003) is compared to a control with no plasmid. 10 ml falcon tubes containing 4.5 ml
of sample were placed label down in the light box to allow maximum exposure to the light.
Fig. 12. Percentage viable cells after 6 hrs in the light box. BL21 (DE3) transformed
with KillerOrange (BBa_K1914001) is compared to a control with no plasmid. 10 ml falcon tubes containing 4.5
ml of sample were covered in tin foil before being placed in the light box.
Fig. 13.Percentage viable cells after 6 hrs in the light box. BL21 (DE3)
transformed with KillerOrange (BBa_K1914001) is compared to a control with no plasmid. 10 ml falcon tubes
containing 4.5 ml of sample were placed label down in the light box to allow maximum exposure to the light.
Ministat experiment
All samples from the ministat were tested using the KillerRed, KillerOrange protocol found here. Glycerol stocks were made from the samples taken at each time interval, testing was done using these glycerol stocks.
The following graphs show the average number of colonies of samples taken at 0 h, 24 h, 120 h and 168 h of continuous culture and then tested in the light box. Values were averaged across three biological repeats. Colonies were not counted above 300 and so this is the maximum value given. All samples were induced to a final concentration of 0.2 nM IPTG. All samples were diluted 1000 times in a final volume of 4.5 ml liquid broth (LB). Error bars represent the standard error of the mean.
Fig. 14. Comparison of CFUs formed by KillerRed exposed to light and kept in the dark for each sample taken from the ministat. The efficiency of the kill switch decreases over time as shown by the increasing number of CFUs.
Fig. 15. Comparison of CFUs formed by KillerOrange exposed to light and kept in the dark. The efficiency of the kill switch decreases over time as shown by the increasing number of CFUs. The effect is not as obvious in KillerOrange compared to KillerRed as the starting efficiency of KillerOrange is lower.
Enzymatic kill switch
Lysozymes are a group of enzymes that are an important part of the immune response against bacteria (Myrnes et al, 2013). They are defined as 1,4-fl-N-acetylmuramidases that cleave the glycosidic bonds between carbon 1 of N-acetylmuramic acid and carbon 4 of N-acetylglucosamine in the peptidoglycan that makes up a bacterial cell wall (Jollès and Jollès, 1984). Lysozymes are commonly used in mass spectrometry for protein mass calibration and are also effective lysing agents against Gram-positive and to lesser extent, Gram-negative bacteria (Sigma aaldrich, 2016). UNICAMP-Brazil 2009 iGEM team used the Lysozyme Gallus gallus part BBa_K284001 previously and other lysis mechanisms have been used as kill switches by TU-Delft 2013, Newcastle 2010, Imperial College London 2011 and METU-Ankara 2011. As others team have used lysis mechanisms in their kill switches we thought Lysozyme (Gallus gallus) would be a suitable candidate to test the effectiveness of lysis as a kill switch mechanism and investigate the potential for HGT if lysis is successful. We added an OmpA signal peptide to Lysozyme C which targets it to the periplasm, this was to ensure that the enzyme would be translocated to the cell wall where it would be most effective. Our model predicted the complete degradation of the cell wall to be within the first generation of E. coli (details of the model can be found here). This gave us confidence that lysozyme would be effective.
Method
To show the activity of lysozyme, a molecular probes EnzCheck lysozyme assay kit (Thermo fisher scientific) was used. The coding sequence (CDS) contains an OmpA signal peptide targeting it to the perisplasm therfore lysozyme will only be detectable if the cells have lysed. The kit uses a substrate containing Micrococcus lysodeikticus cell walls labelled with fluorescein to such as degree that fluorescence is quenched. The presence of lysozyme causes a sharp increase in fluorescence by easing the quenching. The increase in fluorescence is proportional to lysozyme activity in the sample. The fluorescence assay was used to measure the activity of the freshly transformed kill switch and that of the cultures grown in the ministat. CFUs were also used as a measure of efficiency by comparing the number of colonies to a control. 5 ml ovenights of E. coli BL21 (DE3) transformed with pSB1C3 lysozyme were used to inoculate 250 ml Erlenmeyer flasks containing 50 ml of LB laced with 35 µg/ml chloramphenicol. Once an OD of 0.23 was reached IPTG was added to a final concentration of 0.2 nM. Protein production was allowed to proceed for 2 hrs. The sample was serially diluted (10-2,10-3,10-4). 200 µl of each dilution factor was spread plated and incubated at 37 °C overnight. CFUs were then compared to a control treated in the same way.
The potential for horizontal gene tranfser was tested using the lysozyme (Gallus gallus) provided in the EnzCheck lysosyme assay kit. Cells were lysed, the enzyme inactivated and then transformation of the resulting lysate performed. For a detailed protocol see HGT protocol
Enzymatic Kill switch: Results
Characterisation
No difference in CFUs was observed between the control and the samples producing lysozyme. The results of the EnzCheck lysozyme assay were inconclusive.
Horizontal Gene Transfer
The HGT experiment showed that DNA present in lysate can be successfully transformed into a different E. coli strain with an average of 4 colonies per transformation (stdev=3.38). The BL21 (DE3) competent cells all gained the antibiotic resistance and RFP marker from the plasmid present in the lysed DH5α. 2 colonies from each plate were cultured over night and showed a fluorescence value concordant with that of the original culture. The starting cultures of DH5α had an average starting OD of 1.11 and fluorescence value of 258 arbitrary units (A.U) before lysis. The BL21 (DE3) cultures transformed with the lysate had an average OD of 0.75 and average fluorescence of 306 A.U. None of the spread plated lysate produced any colonies, showing that all cells were killed in the lysis reaction.
DNase
DNase I is a nonspecific deoxyribonuclease originally extracted from bovine pancreatic tissue. It degrades both double-stranded and single-stranded DNA resulting in the release of di-, tri- and oligonucleotide products with 5´ -phosphorylated and 3´-hydroxylated ends (Vanecko, 1961). DNase I has also been shown to work on chromatin and DNA:RNA hybrids (Kunitz, 1950). DNase I degrades these target polymer molecules through the hydrolytic cleavage of phosphodiester linkages in their backbone (Suck, 1986).
For a kill switch to be effective as a bio-containment device, the release of synthetic DNA must be mitigated. We aimed to do this in our project using the expression of DNase I. DNase I is commonly used in a laboratory setting to degrade unwanted DNA. It was shown by Worrall and Connolly (1990) that expression of DNase I is possible in E. coli as long as it is under the control of a promoter with a strong off state. We constructed a part with DNase I under the control of the T7 promoter. Unfortunately no transformations were successful and all colonies produced contained empty plasmid backbone. Worral and Connolly reported that a promoter which is less tightly regulated (pKK223-3) would result in transformation failure. As was shown in our metabolic kill switch, the T7 promoter we used to control expression of the CDS is very leaky. This is likely the reason why transformations were unsuccessful as immediately after transformation, production of DNase I would commence killing all the cells. If this is the case future work on a system that uses DNase I as a kill switch but under much tighter control, may prove very effective.
Discussion
Metabolic Kill Switch: KillerRed and KillerOrange.
We have shown that KillerRed and KillerOrange can effectively kill cells (Fig. 11 & 13) under much lower light intensity
than is used in the literature (Sarkisyan et al, 2015). On investigation into the kind of light source that was needed to produce
the 1 W/cm2 of previous experiments (Bulina et al, 2005), it became clear that 1 W/cm2
was impractically bright and would generate large amounts of heat which would kill E. coli. We decided to use a much less powerful LED array that produces 1.2 mW/cm2 at the wavelengths most effective for KillerRed and KillerOrange (Fig. 7) and expose our samples to light for a greater length of time. We showed that this was still effective with an average
survival rate in the + IPTG condition of 2.2% for KillerRed (Fig. 11) and 12.7 % for KillerOrange (Fig. 13). A wider range of exposure times
and light intensities would greatly improve the characterisation of these parts, unfortunately time limitations prevented
us from testing this.
The leakiness of the T7 promoter used in our kill switches was quantified by comparing protein production in the + IPTG condition and - IPTG condition. A one tail t-test assuming equal variance was performed for the mean fluorescence values of the cultures tested in the light box. Fluorescence was used as a measure of protein production. No statistically significant difference was found between the + IPTG condition and – IPTG condition (a significance value of < 0.05 was used. p-value for KillerRed 0.18, p-value for KillerOrange 0.16). CFU counts
for + IPTG conditions were within the standard error of – IPTG (Fig. 11 & 13). For KillerRed the induced kill switch appears to be more
effective whereas the uninduced switch is more effective in KillerOrange. The leakiness of the T7 promoter has likely
lead to near equal expression in both conditions, possibly exacerbated by the length of time that the cultures were left
to grow in order for the protein to fully mature.
The literature showed that cells had been kept in a cold room at 4
°C for 24 hrs before exposing the samples to light (Sarkisyan et al, 2015), the reason given for this was to "increase the fraction of mature protein". We tested the validity of this as cultures were incubated at 37 °C 220 rpm overnight not 4 °C and the phototoxicity of KillerRed and KillerOrange was still evident. The light box itself had a negative effect on E.
coli growth. In our experiment each sample was first diluted to 10-3,10-4 and 10-5 before exposure
to light. The dark condition for the control formed a lawn of bacteria on the agar plate regardless of the starting dilution factor, however in the light condition, the 10-3 dilution produced the same amount of colonies as the dark but in greater dilutions the number of colonies decreased. This decrease was not significant enough to have affected the results but it should be noted that exposure to 1.2 mW/cm2 for 6 hrs slows the growth rate of E. coli BL21 DE3.
The continuous culture of KillerRed showed a 15 fold increase in the percentage of viable cells after 168 hrs. A similar pattern is shown for KillerOrange but with around a two fold increase. Both KillerRed and KillerOrange show greater numbers of colonies forming over time (Fig. 14 & 15). This number approaches the amount produced in the dark condition by 168 hrs. The average fluorescence reading for 0 hr KillerRed samples was 506.3 A.U (recorded at an average OD of of 0.745). After 168 hrs the average fluorescence reading was 436 A.U (at an average OD of 0.96). It seems unlikely due to the readings being similar that a mutation has occurred in the kill switch itself. As fluorescence is proportional to the amount of ROS being produced, up regulation of native E. coli enzymes that mitigate the effects of ROS may be the cause of the increase in cell survival. Future transcriptome analysis could provide interesting data on the mechanism of this change, this was unfortunately beyond the scope of this project. This shows that there may be many ways for bacteria to circumvent the effects of a kill switch given the high selection pressure they pose.
Enzymatic Kill Switch: Lysozyme
The Enzcheck fluorescence assay used to determine lysozyme activity produced values that were not consistent with the CFU count of the plated sample. The HGT experiment we conducted showed that 50 µl of 500 U/ml lysozyme C normally used in the EnzCheck assay to produce a standard curve would effectively kill all the cells in a 50 µl sample of DH5α. The assay showed that 20x diluted sample produced near 500 U/ml activity readings yet this culture would still produce a lawn of bacteria when 200 µl was spread plated. It is noted that the standard curve was of poor quality. The samples of lysozyme were assayed in the same way after continuous culture and did show a decrease in lysozyme activity over time, however the original readings that were used as a comparison have an error of sufficient size that this is not conclusive. The CFU count for lysozyme showed no difference from the control. Lysozyme added to a sample extra-cellularly was shown to lyse all the cells in our HGT experiment, even though Gram-negative bacteria are partially protected from its action due to their outer membrane (Callewaert, 2008). Yet lysozyme produced intra-cellularly and targeted to the periplasm was not effective. There may have been issues with translocation of the protein to the target area, however this seems unlikely due to the effectiveness of its extracellular action. Another explanation may be that lysozyme as a kill switch mechanism is inherently ineffective, as any cells that are resistant will proliferate and a population that are not affected by its production very quickly develops. Regardless of the ability of lysozyme to kill cells effectively, we have shown that HGT is a concern with using this form of kill switch; antibiotic resistance markers commonly used in synthetic biology can be transferred to a different strain of E. coli and in principle any wild type organisms outside the lab. Therefore kill switches that use lysis as a mechanism for cell death should not be considered bio-containment.
