Testing the Functionality of piG_10b for B12 Bioproduction
In this experiment we tested the functionality of piG_10b, the second prototype of CELLECT, in a B12 production setup and compared it with piG_10a, piG_01b and pGGAselect as a negative control. We investigated cell growth, B12 yield and expression of bluB and mazF.
E. coli MG1655[piG_10b], E. coli MG1655[piG_10a], E. coli MG1655[piG_01b] and E. coli MG1655[pGGAselect] cells from a glycerol stock were used to inoculate respective 100 mL pre-cultures in M9 medium.
These pre-cultures then grew to sufficient optical densities to inoculate 50 mL cultures to an OD600 of 0.3. Substrates and inducer were added at OD600~0.6. We tested induced and non-induced setups with cobinamide (Cbi) and hydroxocobalamin (OHCbl) for piG_10b, piG_10a, piG_01b and pGGAselect. pGGAselect is the empty backbone to the other 3 plasmids.
The cultures grew at 37°C, shaking at 200 rpm. Optical density was measured every 6 hours during the first 24 hours and once more after 48 hours. Samples for Western Blot were taken after 0, 12, 24 and 48 hours. LC-MS samples were taken 0, 24 and 48 hours after induction and substrate supplementation.
For detection of B12 concentrations with Liquid Chromatography - Mass Spectrometry (LC-MS), we took 1 mL of cell culture, centrifuged it, discarded the supernatant and stored the cell pellet at -80°C. These samples were then purified and exposed to light deliberately before sending for external LC-MS analysis, blinded to prevent bias.
The experiment was repeated a total of two times.
The setups of E. coli MG1655[piG_10b] induced with doxycycline (DOX) showed no growth, similarly to the culture of E. coli MG1655[piGGAselect] treated with Cbi and DOX (Figure 3).
All other cultures grew relatively similar.
Western Blot shows bluB expression only for induced piG_10a and piG_01b (Figure 4). No signal for BluB was observed in any of the induced E. coli MG1655[piG_10b] setups.
We did see expression of mazF for induced piG_10b, while none was visible for induced piG_10a. Additionally, the setup of induced E. coli MG1655[pGGAselect] with Cbi displayed a signal for MazF as well.
We can see B12 production in E. coli MG1655[piG_10a] and E. coli MG1655[pGGAselect], induced with DOX and supplemented with Cbi respectively. Induced E. coli MG1655[piG_10a] with Cbi show OHCbl contents about 50% compared to E. coli MG1655[piG_01b] under the same conditions.
Expression of bluB and conversely B12 production does not work in E. coli MG1655[piG_10b]. We observed mazF expression and growth inhibition of induced cells cultures.
The results from this experiment imply that neither riboswitch nor bluB gene are working in this construct. This is surprising because bacteria carrying piG_10a do express BluB and produce B12 in the very same experiment. We modified piG_10a by exchanging the ribosome binding site (RBS) upstream of mazF for a stronger RBS, creating piG_10b. More about the cloning procedures is listed on our Cloning Plasmids page.
As the results from this experiment appeared suspicious to us, we had a look at the sequencing results. The quality of the peaks that would determine single nucleotides was rather low, indicating that there might have been a subculture in the glycerol stock we took the cells from. Unfortunately, we did not manage to repeat this experiment with newly cloned bacteria in time to validate the findings from this experiment.
Testing Inducible Kanamycin Resistance
To test if the Kanamycin resistance on piG_11 is inducible by DOX and can be further used for the improved design of our system.
E. coli MG1655 containing the piG_11 where incubated in LB at 37°C and 200 RPM for 1 hour. Afterwards they were split into four different setups.
The setups were 2x LB without anything added, LB with 100 ng/mL DOX and LB with 100 ng/mL DOX and 200 nM AdoCbl added.
They were further incubated in their respective setup at 37°C and 200 RPM for 1 hour. Then plated out on LB plates with different Antibiotics added. Antibiotics added to the LB plates were: 3x 50 mg/mL kanamycin and 34 µg/mL chloramphenicol and one plate with just 34 µg/mL chloramphenicol .
We see a lot of bacteria growth for all four plates (Figure 7). But the most colonies can be seen on plate and 4, while individual colonies can be seen on plate 2 and 3.
The results are inconclusive, because we see the most colonies for the non induced plates.
We would expect a lot of colonies for setup 4) (Figure 7.4) since piG_11 has CHL resistance as its main antibiotic on the plasmid.
For setup 1) we would expect no colonies since the kanamycin resistance is uninduced and they are grown on kanamycin. This is however not what we observe leading us to conclude that the leaky promoter is leaky enough to support growth on kanamycin (Figure 7.1).
Interesting is that for the setups induced with DOX we observe a lower colony number than for the non induced setups (Figure 7.2 & 7.3). The main reason for this could be the induced toxin even though our early toxin results, comparable on our Toxin-Antitoxin Results page, have shown that the RBS of the ribsowitch should not support cell death.
This is further supported by the setup with AdoCbl added (Figure 7.3) as we would expect to have more colonies there if the toxin would indead be the cause for the lower colony number. Since AdoCbl should bind the ribowitch leading to lower toxin amounts.
Conclusivly can be said that the kanamycin resistance is not inducable since the leakyness causes the cells to survive without inducing the resistance. Therefore we dit not go forward with this approach. It remains to be seen however if the kanamycin would meet its purpose of protecting the toxin promoter from mutations.
The fact that the cells are kanamycin resistant without induction could even be seen as an advantage since no inducer is requiered and the purpose of protecting the toxin promoter would still be given.
Outlook for CELLECT
Since we designed CELLECT, our self-regulated-system, to ensure genetic stability and improve bioproduction, the optimal next step would be to integrate it directly into the bacterial genome to eliminate the need of antibiotic resistance. Especially, as large-scale bioproduction does not allow for supplementation of antibiotics due to economical and environmental reasons . This was also pointed out by the industrial biotechnology company BRAIN Biotech during our meeting. More about our meeting with BRAIN Biotech is accessible on our Integrated Human-Practices page. However, implementing the system into the bacterial genome would go beyond the scope of our project, so we decided on using plasmids carrying our system as proof of principle.
Unfortunately, due to the shortage of time and the long cloning process, it was not possible to achieve all of our goals. For some parts, we have not yet had the chance to perform the necessary experiments to provide us with the desired answers. So here are some improvements we will implement or tests if we choose to move on with CELLECT, exceeding the iGEM competition.
Improving Toxin/Antitoxin SystemOne of the next steps would be to gain more information about the Toxin/Antitoxin interactions by using the fluorescence resonance energy transfer (FRET) method. FRET is a distance-dependent physical process by which energy is transferred from an excited donor fluorophore (mTurquoise) to another acceptor fluorophore (mVenus), when they are in close proximity (10-100Å) to each other . By coupling mTurquoise to the toxin mazF and fusing the antitoxin mazE to mVenus, we are interested in detecting the formation and the lifespan of the unstable toxin-antitoxin complex.
As toxin expression is regulated through the same inducible promoter as the bluB gene, the expression of both proteins is coupled to each other. Therefore the required amount of antitoxin to buffer toxin production after addition of inducer is a crucial question to be answered.
It is essential to determine the lethal toxin concentration in the cell in order to adjust the production of antitoxin accordingly with the help of a constitutive promoter. Ensuring the survival of the cell at the onset of production, preventing cells to die because toxin production is faster than B12 synthesis, while also ensuring cell death in case of unproductiveness.
At the beginning of the project, we looked into several different toxin-antitoxin systems. After intense research and review of the current literature, we decided on MazE/F as most suitable for the moment. It is well studied and has already been implemented as a part of the iGEM registry.
By collecting experimental data on this TAS we want to solidify our choice as the best possibility for CELLECT. If our experimental data where to suggest MazE/F is not suitable we wouldn't hesitate to switch the TAS.
According to a publication, growth inhibitory effects of MazF seem to be higher during the lag phase of a culture compared to mid-log phase cultures of E. coli . Therefore, we are interested in the effect of DOX induction at OD600=0.1 instead of OD600 = 0.5/6 on MazF toxicity.
Eliminating Mutation RisksThe most susceptible sites for mutations in our CELLECT plasmid are the toxin gene mazF and the TetR promoter in front of bluB and mazF. As the cells are exposed to a selection pressure favoring random mutations that stop production of the toxin, new subpopulations with mutations of these plasmid parts could emerge.
With sequencing results repeated over a time span of around four days, we were able to show that the toxin gene did not mutate which could have affected the function of the toxin. All details regarding this experiment are obtainable on our Toxin-Antitoxin Results page.
Even if the concerns of toxin mutation do not seem to have any impact on our system in the period we have tested, we are interested in finding a solution to minimize the probability of mutations that could affect our toxin long term. Implementing CELLECT in a bioreactor would require solid stability for long periods of time and numerous cell generations.
Therefore, we would introduce a second selection marker. As an example, placing a kanamycin resistance gene in front of the riboswitch followed by the toxin under the same promoter. This would be our CELLECT system two. To see more about CELLECT system two, have a look at our Design page.
In theory, if the TetR promoter in front of the riboswitch and toxin gene mutates and therefore transcription of this part is stopped, the kanamycin resistance is also not expressed and the cells are vulnerable to antibiotic selection.
However, the promoter we used for the CELLECT system appeared to be leaky. Therefore during initial experiments with the kanamycin resistance, we experienced that even without induction via DOX cells survived in media supplemented with kanamycin. The solution would be to replace the leaky TetR promoter with a tighter inducible promoter for CELLECT system two. Optionally, a higher kanamycin concentration could be used to counteract the leaky expression, as induction still increases expression of the resistance gene.
Despite having these options to improve CELLECT in mind, time was figuratively slipping through our fingers. Therefore we decided to focus on our system without the kanamycin resistance firstly.
A second idea to stabilize the mutation-sensitive areas is an overlap in the toxin and antitoxin gene. Creating artificial overlaps and frameshifts is a technique developed specifically for improved gene stability . For an improved CELLECT system, the toxin and antitoxin genes could be arranged with an overlap including a shifted frame, causing mutations in the toxin gene to affect the antitoxin gene as well. As the toxin is more stable than the antitoxin, such a mutation would be deadly for the cell. Based on this, we thought of adding a second antitoxin gene in a frame shift to the toxin gene, protecting the toxin from mutations. This method, however, requires a lot of know-how and optimisation and was therefore not feasible to pursue during iGEM.
Improving ProductionDue to the fact that hardly any data exist on whether BluB differs in its activity in the monomer and in the dimer, we decided to model this using biological structural analysis and ultimately molecular dynamics simulations. All results and data generated with this model of molecular dynamics is provided on our Modeling page.
To do this, we first used AlphaFold , a tool for protein structure prediction based on an amino acid sequence. BluB as a monomer and as a fusion dimer were created from our plasmid sequence of piG_01. We showed that our artificial fusion dimer is stable in our simulation of over 30 ns and forms the enzymatic pocket in a similar way as published crystallographic BluB data . The BluB monomer however showed imperfect protein folding because the c-terminal tails of BluB stabilize the substrate Flavin mononucleotide.
We therefore suggest that a fusion BluB could help improve AdoCbl production yields because enzymatic BluB is bound to its complex partner, increasing the likelihood of the dimer forming and overall stability. We plan on cloning this fusion protein and are thrilled to see if we can achieve experimental approval of this assumption by reaching higher B12 yields.
Improving RiboswitchFor the implemented riboswitch, we were sadly unsuccessful in constructing and cloning a reliable and precise AdoCbl sensor. Although different AdoCbl concentrations only lead to minor, yet expected changes in fluorescence, supplementing Cbi does not show any predictable effect on fluorescence, thereby suggesting that the riboswitch does fold accordingly. You can find more details regarding the riboswitch on our Sensing B12 Results page.
However, at this point, we are unfortunately still unsure whether the riboswitch we have chosen folds as intended, when another gene is expressed upstream.
The limitation for the need of an already existing riboswitch could possibly be overcome by the use of a synthetic riboswitch made on the basis of an aptamer . However, this is a complex process and requires extensive testing which was outside of the capabilities of our iGEM project.
Implementing Missing ControlsDue to time pressure we currently do not have a negative control for our system. This would be a mutated, non-functional bluB. Here, no vitamin B12 is produced and the cells should therefore die due to lacking inhibition of the toxin gene via the riboswitch. Such a plasmid (tetR_mut_bluB_riboK12_mazF) is already designed, but so far has not been cloned successfully.
Alternative ApplicationsA fascinating next step for the development of CELLECT would be exploring alternative applications aside from bioproduction. With its universal design, the autoregulatory system is not limited to bioproduction only. It can also be used for 1) degradation by making substance-degrading microbes dependent on the associated intermediate and for 2) biocontainment: in this case, the used organisms survive just as long as the intermediate is produced from the degraded product.
-  Wang ZK, Gong J, Qin J, Li H, Lü ZM, Shi J, et al. Improving the intensity of integrated expression for microbial production. ACS Synthetic Biology [Internet]. 2021 Nov 5;10(11):2796–807. Available from: https://doi.org/10.1021/acssynbio.1c00334
-  Sekar RB, Periasamy A. Fluorescence resonance energy transfer (FRET) microscopy imaging of live cell protein localizations. J Cell Biol. 2003 Mar 3;160(5):629-33. doi: 10.1083/jcb.200210140. PMID: 12615908; PMCID: PMC2173363. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2173363/
-  Potnis AA, Raghavan PS, Shelke A, Nikam TD, Rajaram H. Comparative analysis of MazEF and HicAB toxin–antitoxin systems of the cyanobacterium,Anabaena sp. PCC7120. Visweswariah S, editor. FEMS Microbiology Letters. 2016 Dec 8;364(1):fnw279.
-  Decrulle AL, Frénoy A, Meiller-Legrand TA, Bernheim A, Lotton C, Gutierrez A, et al. Engineering gene overlaps to sustain genetic constructs in vivo. PLoS computational biology [Internet]. 2021 Oct 1 [cited 2023 Oct 9];17(10):e1009475. Available from: https://pubmed.ncbi.nlm.nih.gov/34624014/
-  Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021 Jul 15;596(7873):583–9.
-  PDB ID: 2ISJ
-  artificial riboswitch Findeiß S, Etzel M, Will S, Mörl M, Stadler P. Design of Artificial Riboswitches as Biosensors. Sensors [Internet]. 2017 Aug 30;17(9):1990. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5621056/