Fig. 1: Genetic circuits for (a) HicA-HicB and (b) RES-Xre systems in MG1655.
To complete the growth switch, we co-transformed both toxin-GFP plasmid and antitoxin plasmid into MG1655 knock-out HicA-HicB strain (for HicA-HicB toxin-antitoxin system) and native MG1655 strain (for RES-Xre toxin-antitoxin system).Fig. 2: Growth curve showing growth arrest in cells induced with (a) HicA toxin gene and (b) RES toxin gene expression after the addition of IPTG at 1h. (c) The extent of growth arrest by HicA toxin was found to be significant as shown by the difference in maximum OD600 and initial OD600 between HicA-mediated growth-arrested cells and normal-growing control cells. (d) Similarly, the extent of growth arrest by RES toxin was also significant.
However, there was an unexpected resumption of growth in cells containing RES-Xre plasmids at approximately 10 hours, which was not observed in cells containing HicA-HicB plasmids. Previously, Skjerning et al. performed RES characterization for 6h, with no resumption of growth observed (Skjerning, Senissar, Winther, Gerdes, & Brodersen, 2019). We wondered if this auto-resumption of growth was simply not captured in the aforementioned work due to the shorter characterization duration. Hence, we proceeded to characterize the exact plasmid used in their work for 20h and saw a similar auto-resumption in growth despite the absence of Xre expression (Fig. 3). This confirmed that there is auto-resumption for the RES system. As the RES toxin functions to degrade cellular NAD+, it may be due to the existence of an inbuilt feedback system in the cells which helped to cope with the declining level of NAD+, thereby resulting in an auto-resumption of growth. Due to time constraints, we were unable to further validate this phenomenon. Nevertheless, quantifying intracellular NAD+ at different time points of the characterization can be done to validate this hypothesis for future work.Fig. 3: Growth curve of E. coli cells transformed with Skierning et al. RES plasmid, showing a resumption in growth, albeit much later (14 hour).
Fig. 4: Growth curve of growth-arrested cells displaying resumption in growth after the induction of (a) HicB antitoxin and (b) Xre antitoxin gene expression with 0.0133M arabinose at 3.5h. (c) The extent of growth resumption by HicB antitoxin was found to be significant as shown by the difference in maximum OD600 and initial OD600 between growth-arrested cells and HicB-mediated growth-resumed cells. (d) Similarly, the extent of growth resumption mediated by Xre antitoxin was also significant.
To prove that the toxins trigger growth arrest without killing the cells, we performed a Colony Forming Unit (CFU) assay before and after the induction of antitoxin expression in growth-arrested cells. The cells cultured in tubes had their growth arrested for 10 days prior to the CFU assay. We hypothesized that if the HicA toxin is bacteriostatic, expression of the HicB antitoxin should restore cell growth and allow cell division to take place. Hence, the number of colonies formed after the induction of antitoxin expression should be greater than that in growth-arrested cells without antitoxin expression. Consistent with our hypothesis, the results show that across all three dilutions, there were more colonies formed when growth-arrested cells were induced with antitoxin expression (Fig. 5). Taken together, the results demonstrated that the growth-arrested cells were able to remain viable even after 10 days of growth arrest, retaining the ability to grow and divide after antitoxin expression was induced.Fig. 5: Growth-arrested cells maintained with 1mM IPTG for 10 days demonstrated the ability to restore growth and divide after the induction of HicB antitoxin expression as shown by (a) CFU plating and (b) CFU quantification at a 106 dilution factor.
Fig. 6: (a) Growth curve of growth-arrested cells induced with HicA toxin expression at 1h showing a reduced (b) total GFP and (c) average GFP production rate, demonstrating that the inhibition of cell growth inhibited its rate of protein production. (d) The level of protein production (maximum - initial) in HicA-mediated growth-arrested cells were found to be significantly lower compared to normal-growing control cells.
With that being shown, we speculate that the induction of HicB antitoxin gene expression would likewise trigger our growth switch to restore GFP production rate. Indeed, cells showing growth resumption displayed a higher protein production rate in comparison with growth-arrested cells that were uninduced with antitoxin expression (Fig. 7b). Similarly, statistical analysis was performed using nine experimental samples and the level of protein production was found to be significantly higher in growth-resumed cells compared to growth-arrested cells (Fig. 7d). This demonstrated that our growth switch has the ability to restore protein production in previously growth-arrested cells, via the induction of antitoxin expression.Fig. 7: (a) Growth curve of growth-arrested cells induced with HicB antitoxin expression at 3.5h showing an increase in (b) total GFP and (c) average GFP production rate, demonstrating that the resumption of cell growth restored their protein production rate. (d) The level of protein production (maximum - initial) in HicB-mediated growth-resumed cells were found to be significantly higher compared to growth-arrested cells.
Taken together, we are excited to show that our growth switch has the ability to control cell growth and regulate protein production, providing scientists with a tool to manipulate cells for fundamental and application-based studies.Fig. 8: Growth curve of cells treated with fixed concentration of IPTG and varying concentrations of arabinose at 1h and 3.5h respectively to generate experimental data for model construction.
Additionally, our growth switch should ideally be (1) achieving a maximum growth arrest upon induction of toxin expression, (2) showing a faster growth resumption upon the induction of antitoxin expression and (3) using the lowest possible inducer concentration to achieve (1) and (2). To find out the optimal inducer concentrations that could fulfil requirements (1) to (3), the modelling team came up with a surface plot (in the sensitivity analysis, growth switch section) which was useful at informing the wet lab team on the inducer concentrations they should be using or avoiding. By coupling a model prediction to our growth switch, we are able to provide users with predictions of growth profiles when parameters such as inducer concentration and induction time are varied. Check out our software tool to find out more!Fig. 9: Genetic circuit of HicA-HicB toxin-antitoxin modules and LuxCDABE operon in MG1655.
We show that on all three days, triggering of growth resumption and protein production resulted in growth-arrested cells producing a higher level of luminescence compared to normal-growing control cells (Fig. 10 & 11). Growth-arrested cells possess a lower starting OD600 value compared to normal-growing control cells due to the active inhibition of growth in the former (Fig. 10a - 10c).Fig. 10: HicA-HicB-mediated growth switch generating a prolonged functional lifespan in luminescence-producing cells. (a), (d), (g) Growth curves showing growth-arrested cells with a lower starting OD600 indicative of growth inhibition across different days. (b), (e), (h) Total luminescence curves consistently showing a higher level of luminescence production upon in growth-arrested cells induced with HicA toxin expression, compared to normal-growing control cells without toxin expression. (c), (f), (i) Averaged luminescence curves showing a similar trend - growth-arrested cells producing higher level of luminescence per OD600 than normal-growing control cells.
Fig. 11: Photo of normal-growing control cells and luminescent growth-arrested and revived cells taken on day 5.
Taken together, this demonstration suggests the potential of our growth switch in implementing a more sustainable and controllable production of bioluminescence. As part of our human practice, we interviewed Ms Teresa van Dongen, a bioluminescent lamp designer, and learnt that the inability to turn the bioluminescent lamp 'OFF' is one of its shortcomings. With our growth switch, we addressed this problem by allowing bioluminescence production to be turned 'ON' and 'OFF' at will, providing existing bioluminescence lighting with an enhanced function of acting much like an actual torchlight.Fig. 12: Genetic circuit of HicA-HicB toxin antitoxin module incorporated into a biosensor system for AHL detection.
Fig. 13: Pictorial diagram of cell-based AHL biosensor genetic circuitry. Growth-arrested cells were maintained with IPTG every three days to keep the cells dormant, reducing their rate of LasR production. Upon the presence of AHL, the basal level of LasR proteins produced would then activate LasI promoter to ‘wake’ the cells and produce reporter protein, RFP. On the other hand, normal-growing cells would be growing and constitutively producing LasR until AHL is detected to produce RFP.
Here, we show that on day 11 the addition of AHL induced growth-arrested cells to produce a higher level of RFP compared to normal-growing control cells (Fig. 14b - 14c). Growth-arrested cells similarly possess a lower starting OD600 value compared to normal-growing control cells due to active growth inhibition in the former (Fig. 14a).Fig. 14: HicA-HicB-mediated growth switch displaying a prolonged functional lifespan in cell-based AHL biosensor cells on day 11. (a) Growth curve of growth-arrested cells shows a lower starting OD600 compared to normal-growing control cells, which was indicative of growth inhibition up till day 11. (b) Total fluorescence curve shows growth-arrested cells producing more RFP protein than normal-growing cells, which was clearly supported by (c) average fluorescence curve whereby each cell produced on average a two-fold higher level of RFP protein.
As part of our human practice, we spoke to Dr Karen Polizzi from Imperial College and learnt that cell-based biosensors do not necessarily have to express high amounts of protein for detection. Hence, we speculate that the amount of RFP produced by growth-arrested cells (RFU= 6,000 - 10,000) on day 11 to be sufficient in biosensor applications, as long as this value exceeds the minimum detection threshold required for the application. While further work has to be done to optimize this system, taken together, we foresee the usefulness of our growth switch to be extremely relevant for in situ functions such as environmental and therapeutic monitoring. We thus propose E.co LIVE as a novel method to prolong the functional lifespan of engineered cells at 37°C, keeping them viable and functional in situ and removing reliance on a cold chain.Fig. 15: (a) Total ATP level between growth-arrested cells and normal-growing cells in bioluminescence demonstration. Statistical analysis performed using four experimental samples revealed a significant difference in the level of ATP between both growth-arrested cells and normal-growing cells, suggesting that the inhibition of translation results in reduced ATP usage. (b) Similarly, total ATP level was found to be higher in growth-arrested cells compared to normal-growing cells for AHL biosensing demonstration. Statistical analysis was not performed due to insufficient experimental samples.
Fig. 16: Comparison of promoter activity over 8h. Growth curve of cells with varying spacer length (a) in the presence of blue light and (b) in the dark was plotted. Looking at the total RFP graph (c) in the presence of blue light and (d) in the dark, as well as the average RFP graph (e) in the presence of blue light and (f) in the dark, we clearly see that the blue-light repressible promoter with 30bp spacer consistently produced higher level of protein despite having a relatively higher OD600 than the other two samples, evidently shown by the average fluorescence curve.
Fig. 17: Genetic circuit showing HicA toxin gene placed under IPTG-inducible promoter and HicB antitoxin gene placed under blue-light repressible promoter containing a 30bp spacer. Meanwhile, the EL222 gene is constitutively expressed.
Here, we are proud to show that our growth switch coupled to the improved blue-light repressible promoter retained its desired functionality in response to the presence of blue light. Using a custom-made blue light LED illuminator, we could clearly see that cells pre-treated with IPTG grown under blue light produced visibly lesser GFP than cells treated with IPTG but grown in the dark (control cell 1) and also untreated cells grown under blue light (control cell 2). This was consistent with our hypothesis (Fig. 18).Fig. 18: Microplate showing a visual representation of GFP production under a blue light illuminator for (a) control cells shone with blue light only, (b) test cells induced with IPTG and shone with blue light, (c) control cells induced with IPTG but kept in the dark. All three conditions were performed for 24h before visualization.
To visualize for the recovery of GFP production, we left the plate to shake in the dark for 24h. Formerly growth-arrested cells were subsequently observed to produce comparable amounts of GFP when compared to both controls, indicating a recovery in protein production (Fig. 19).Fig. 19: Microplate showing visual representation of GFP production under a blue light illuminator for (a) control cells exposed to blue light only, (b) test cells induced with IPTG and exposed to blue light, (c) control cells induced with IPTG but kept in the dark. All three samples were kept in the dark for 24h before visualization.
To further validate that the growth switch is capable of controlling growth and regulating protein production in response to blue light, we proceeded to characterize the cells using a microplate reader. After incubating the cells that were pre-treated with IPTG under blue light for the whole night, we measured cell growth and level of fluorescence production by the cells the next day. As expected, a plateau was observed at the beginning of the measurement, indicating that cell growth was indeed arrested from overnight treatment. Furthermore, the starting OD600 of cells treated with IPTG and exposed to blue light was relatively lower than the light-only control cells, which further supports the hypothesis that growth was arrested in the former cells. Over time, it appeared that absence of blue light in the microplate reader triggered the expression of antitoxin. resulting in growth resumption as shown in Fig. 20a.Fig. 20: Effect of blue-light controlled growth switch on (a) cell growth (b) total protein production and (c) average protein production when growth-arrested cells were kept in dark and measured for 12h.
From the experiments conducted above, a positive trend between cell growth and protein production under the control of blue light was observed, indicating that the presence of blue light and IPTG induced growth arrest. As a result, a lower level of GFP was produced at the beginning of the measurement (Fig. 20b). Over time, we see that the level of GFP produced increased as the absence of blue light triggered the production of antitoxin, thereby ‘waking’ the cells up to produce more GFP. The validity of these results are tempered by the variance observed in the readings, necessitating further improvements in experimental designs despite the overall clear trend. Taken together, we show that our growth switch developed to control cell growth and regulate protein production is functional even when it is partially controlled by a blue-light inducible system. As such, we envision the adoption of other inducible systems (e.g temperature inducible system) to also result in a functional growth switch with the ability to prolong cellular functional lifespan. Through this demonstration, we were able to showcase the flexibility and modularity of our growth switch, making it a broad and useful toolbox for scientists.Fig. 21: Genetic circuit of SgrS gene placed under the control of tetracycline-inducible promoter.
Fig. 22: (a) Growth curve of cells induced with different aTc concentrations showing a negative correlation between aTc concentration and growth rate of cells. (b) Looking at different aTc concentrations used during the characterization, higher aTc concentrations reflected a lower level of glucose consumption as well as a smaller increase in growth, compared to lower aTc concentrations.
Taken together, these results support our hypothesis - that controlling cellular glucose uptake rate can effectively control cellular growth rate and produce different growth trends.Fig. 23: SgrS-mediated growth knob regulating protein production. (a) Growth curve of cells co-transformed with SgrS plasmid and inducible RFP plasmid showing a negative correlation between aTc concentration and growth rate of cells. (b) Total RFP graph revealed a spectrum of protein production profiles - increasing aTc concentrations showing a decreasing protein production rate and generating a gentler protein production profile as hypothesized. (c) Average RFP produced by cells treated with different aTc concentrations was plotted to support the analysis.
Although a simple reporter gene was used to demonstrate the functionality of our growth knob, we envision this growth knob to be useful at fine-tuning growth and protein production in applications requiring different protein production profiles. For this growth knob, we have also constructed a model based on the experimental data generated by our wet lab team. To check out the different simulations we have run with this model click here!Fig. 24: Genetic circuits showing each plasmid containing an opposing pair of toxin and antitoxin to generate selection pressure for plasmid retention. The top plasmid contains Axe antitoxin from Txe-Axe toxin-antitoxin system, Hok toxin from Hok-Sok toxin-antitoxin system, RFP gene and chloramphenicol resistance gene. On the other hand, the bottom plasmid contains Txe toxin from Txe-Axe toxin-antitoxin system, Sok antitoxin from Hok-Sok toxin-antitoxin system, GFP gene and kanamycin resistance gene.
After obtaining the Gibson products for both plasmids, we transformed them into 10-beta cells for cloning and sent the extracted plasmids for sequencing. Interestingly, we discovered that one of the toxin pairs, more specifically Hok, contains a nonsense mutation encoding for a stop codon within the gene itself (Fig. 25). This renders the Hok-Sok toxin-antitoxin pair non-functional as the Hok gene has likely lost their ability to enforce selection pressure on the cells.Fig. 25: Sequencing results showing that the deletion and point mutation changed tyrosine (TAC) into a stop codon (UAA) at amino acid position 35.
Nevertheless, we continued to transform both plasmids into MG1655 and plated the cells on either chloramphenicol or kanamycin plates to test their functionality. For cells plated on the chloramphenicol plate, we observed most colonies to be red and non-fluorescing (Fig. 26a). This was expected because despite having a mutated Hok toxin gene, the chloramphenicol resistance gene and RFP gene present in the same plasmid are still functional, allowing the cells to grow and form red colonies. The reason why we expected fewer orange colonies on chloramphenicol plate, was because the presence of mutated Hok gene would have rendered the other plasmid containing the ‘antidote’ Sok and GFP unnecessary since it is no longer required to neutralize the Hok toxin, which is unable to impose a killing pressure on the cells. On the other hand, we observed that most colonies were orange in color when cultured on a kanamycin plate (Fig. 26b). This was expected because sequencing results revealed the Txe toxin gene to be intact. As such, Txe toxins produced have to be neutralized by its cognate antitoxin Axe, which was transcribed and translated from the other plasmid. Hence, for cells to survive on a kanamycin plate, both plasmids have to be taken up by the cells, thus causing both GFP and RFP to be concurrently produced and resulting in an orange color formed. Altogether, this demonstrates that despite having only one pair of functional toxin-antitoxin (Txe-Axe) present in separate plasmid, almost all of the transformed cells could still retain both plasmids as shown in Fig. 26b. More importantly, the results provide preliminary insight into the feasibility of our conceptual dual plasmid retention system.Fig. 26: MG1655 transformed with both plasmids plated on (a) chloramphenicol and on (b) kanamycin, visualized with a blue light illuminator. Under the blue light illuminator with a wavelength ranging between 380nm to 500nm, GFP with the excitation wavelength of 485nm would be able to fluoresce while RFP with the excitation wavelength of 528nm would not be able to fluoresce. However, as the emission wavelength of GFP (528nm) was coincidentally the same as the excitation wavelength of RFP, cells containing both GFP and RFP would fluoresce to produce an orange fluorescence as shown in (b).