Our goal was to design a self-regulated system to optimize bioproduction and ensure plasmid stability. Therefore, we designed a three component system containing a riboswitch, a toxin-antitoxin system and a gene of interest for production (bluB). Each part is highly regulated and yet they interact with each other to form an autoregulatory circuit that regulates the survival of the engineered cells. To achieve our goal, we went through several engineering cycles. The first step of the engineering cycle is designing the project. Followed by the second step, the construction of the system. Third, testing and measuring the results and finally learning from our data to finetune the system until the desired results are reached.
CELLECT is composed of three individual parts and while testing each of them, we went through each of the engineering cycles multiple times:
- dealing with our chosen cloning method
- the expression of the toxin, MazF, in E. coli
- expression of AdoCbl and
- the design of the riboswitch
Our system contains multiple separate parts. To test their individual functionality and to have the possibility to change specific sequences as ribosome binding sites (RBS) or promoters, we designed multiple plasmids for testing each part individually. Therefore, it was necessary to choose a cloning strategy which allowed such modularity to simplify our cloning process.
After some deliberation, we decided on the Golden Gate method. One big advantage of this method is that we could assemble single and multiple DNA fragments even in the presence of redundant elements. Since no one from our team has had any previous experience in cloning, we wanted to chose a Golden Gate enzyme which, according to research, has a high efficiency. The enzyme we decided on using as a restriction enzyme is BsmBI. Following, we designed overhangs with the corresponding sites to each fragment.
To build our plasmids various fragments were synthesized: Antitoxin, bluB and the riboswitch combined with the toxin. Some of the fragments as the TetR- and Amp promoter, mTurquoise, mVenus and the rrnB Terminator, we could amplify out of plasmids we got from our PI. After running PCRs to ensure that the restriction sites of BsmBI and the right overhangs were inserted, we cloned our fragments into the pGGAselect backbone.
The DNA construction was amplified and analyzed by agarose gel electrophoresis. It was successful for two of our plasmids containing bluB and the toxin. However, we faced multiple problems using this approach. The efficiency of our assemblies was very low, which led us to screen dozens of colonies on our plates in search of one successful clone.
We learned that we could adjust the concentration of chloramphenicol on our plates to minimize the amounts of clones that do not carry our target plasmid. We further contacted researchers with experience in Golden Gate cloning to get advice. We learned that the enzyme we had chosen to use for digesting our fragments (BsmBI) seems to be less common than other restriction enzymes used in Golden Gate. In addition, a post-doc (Dr. Morgane Batzenschlager) from a lab next door at the university advised us to use a backbone other than pGGAselect, e.g. EC47732 pL1V-F1-47732 with blue/white selection after addition of X-Gal, AmpR-BsaI or Bpil cloning to facilitate screening of our clones. Furthermore, we learned that incubating the reaction overnight and at lower temperatures and with more cycles could be beneficial. BsmBI works optimal at 55°C but temperatures over 45°C inhibit activity of the T4 Ligase. Even if we lowered the temperature to 42°C, as recommended by NEB, where the T4 Ligase should not be affected, it seemed that BsmBI cut less efficient and therefore this approach did not improve our results. Additionally, we figured out that we could avoid contamination of the original backbone plasmid by doing a DPN1 digest overnight.
However, even by implementing everything we learned, we still didn’t get good results from our cloning strategy. We then cut all our fragments individually and loaded them onto a gel to find out if there were any problems already with the restriction digest by BsmBI.
By analyzing our agarose gels, we figured out that digesting our terminator did not work efficiently. It is possible that the random bases we decided on as the overhangs and also as buffer between the recognition and restriction site caused methylation that made it impossible to cut them. Since we had already successfully cloned the riboswitch with Gibson Assembly and our supervisors had the most experience with this cloning method, we decided to switch completely to Gibson Assembly.
By changing our cloning method to Gibson it was necessary to design new primers for each fragment. We paid attention to ensure that the primers had sufficiently long binding sites and that the overhangs did not have a melting temperature under 50°C.
Using the plasmids we already cloned via Golden Gate as a template, we could shorten their sequence via PCR or include fragments in them.
After amplifying each plasmid we did an DPN1 digest overnight instead of 15 min. to ensure that we obtained pure amplified DNA fragments.
We learned that we achieved results more efficiently by switching to Gibson Assembly. With this method we cloned most of our plasmids.
To test the function of our riboswitch and to have a cost-effective detection method besides the ethanolamine-medium, we decided on designing a biosensor. It was possible to test whether our riboswitch is functional by measuring sfGFP fluorescence. Even though we tested three different riboswitches, this engineering cycle is specific for the sensor containing the E. coli K12 riboswitch.
After a long research period, we decided on using the naturally occuring riboswitch from E. coli K12, which can detect B12 and differentiate between Cobinamide and Cobalamin. Our riboswitch negatively affects the expression genes downstream when its ligand adenosylcobalamin (AdoCbl) binds. Therefore, we designed our biosensor so that the riboswitch is located upstream of the lacI repressor gene, which in turn regulates the expression of a fluorescent protein, such as the superfolder GFP (sfGFP). When AdoCbl is present, the riboswitch represses LacI expression, the trc promoter in front of sfGFP is no longer repressed and sfGFP is produced.
In other words, our construct reverses the inhibition of the riboswitch, resulting in a positive readout through increased fluorescence at elevated AdoCbl concentrations. The first step was to find a method to test whether the biosensor only expresses sfGFP in the presence of the target compound AdoCbl.
The riboswitch was inserted between the lacI promoter and lacI. In order to be able to upload the biosensor as a composite part to the registry of iGEM, a silent mutation would have been necessary between the trc promoter and the sfGFP gene to remove an EcoRI restriction site. However, since the reporter gene is variable in this sensor and can be exchanged with any reporter of choice, we decided to keep on working without the silent mutation. Therefore, sensor piG_K12BSa consists of the lacI promoter, the riboswitch, the lacI repressor gene and the trc promoter.
In the experiments we performed to test the biosensor we treated cultures with various concentrations of AdoCbl and measured the fluorescence. We observed a high background fluorescence signal even without AdoCbl supplementation, which implied an insufficient repressor production. Moreover, the total fluorescence signal of the sensor, induced with IPTG, is overall lower than the one of a comparable plasmid containing sfGFP induced with IPTG. This implied that too much repressor was present. Also the differences in fluorescence with different amounts of AdoCbl added to the medium were minor (see more on Sensing B12).
A similar problem was also observed for a toxin experiment. There, insufficient toxin expression was supposedly caused by the RBS which is part of the riboswitch. From literature research we learned that this specific RBS usually controls the expression of a transporter protein that isn't required in high amounts, so we concluded the RBS being rather weak. A weak RBS would explain the high fluorescence when the sensor is uninduced. To incorporate the riboswitch to the previously mentioned plasmid we had removed the original RBS of lacI. If the RBS of the riboswitch is significantly weaker than the original one, that would explain low repressor production.
We designed new gibson primers with which we could add a supposedly stronger RBS, taken out of the TetR promoter of plasmid piG_23, directly between riboswitch and lacI, hoping to not disturb the conformation of the riboswitch.
We amplified the backbone via PCR, digested the amplified DNA with DpnI overnight, and inserted the stronger RBS of piG_23 via Gibson Assembly.
We repeated the whole experiment with the recloned version, piG_K12BSb and noticed that compared to the previous construct, containing the old RBS, the fluorescence overall decreased.
We deduced that the high fluorescence in the uninduced state, also not including any AdoCbl, actually was a consequence of the weak RBS provided by the riboswitch. However, this did not explain the minor differences in fluorescence we saw when AdoCbl was added to the medium, nor the overall low fluorescence when we induced the sensor with IPTG.
IPTG is supposed to bind any present repressor, therefore the reduced fluorescence in presence of IPTG suggested either a lower sfGFP production, or an excessive constant repressor expression.
This new insight, certainly unlocked a solution for the issue regarding the toxin that has not been working when the riboswitch was located upstream of the gene. However, it did not improve the function of the sensor past lowering the background fluorescence.
It is still puzzling that the sensor is unable to reach fluorescence as high as the IPTG induced pIG23_04 plasmid. Because if the only difference had been the weak RBS, an insufficient repressor expression of the lacI gene would cause the sensor to always show fluorescence as brightly as the induced original plasmid. As this is obviously not the case, an overproportional repressor production would be the explanation. Thus, even when AdoCbl is present, which is supposed to activate the riboswitch and stop repressor production of LacI, a decreased fluorescence is to be expected.
To find a solution, we firstly need to understand the different ways in which the riboswitch and IPTG inhibit the repressor LacI. While IPTG binds all present repressor, changing its conformation and therefore stopping it from binding and repressing the trc promoter, the riboswitch doesn´t inhibit already present LacI. The riboswitch changes its conformation in the presence of AdoCbl, inhibiting transcription and translation of lacI. Simplified, IPTG removes any produced repressor, while the riboswitch only stops production of more repressor. Earlier produced LacI is still present in the system and capable of repressing the trc promoter and consequential sfGFP production. In theory, it could also be possible that the repressor is still produced because not all copies of the riboswitch are always inhibited, or because repressor production starts before AdoCbl can bind to the riboswitch. In any case, some LacI would always be present until it is degraded. This could be the explanation for the minor differences in fluorescence with different AdoCbl concentrations. As a result, a new solution for improving the detection ability of the sensor would be to change the experimental set up.
A way of testing the repressor production would be measuring sensor activity at multiple time points over an extended period of time. So far measurements were taken only 12 hours after AdoCbl was added to the medium. So theoretically, this timespan could either be too much or too little time to accurately depict the function of the sensor.
If it's too long, too much repressor could have been expressed through leakiness or because all riboswitch can not be activated at all times.
If it is too short, previously produced repressor may not have been degraded yet, and even if at a certain point all riboswitches on all plasmids ware active and able to suppress LacI production, already produced LacI would still be present and able to suppress sfGFP production.
So far we were unable to test the sensor in this new experimental set up.
In order to evaluate whether LB medium contains significant amounts of B12 (which would consequently exclude its use in further experiments) and to prove that our E. coli cells actually synthesize AdoCbl, we searched for a method to detect B12, more specifically, AdoCbl.
The first B12 detection method we chose to test was ELISA, using a Vitamin B12 ELISA kit (DEB12E01) from Demeditec. It was a generally well-documented detection method, quicker and easier compared to LC-MS, and we could perform it in our own lab. Important considerations that influenced our choice. However, finding an ELISA designed for B12 detection in cell lysates specifically remained a challenge. The Demeditec kit we tested was designed for cyanocobalamin detection in vitaminized foods, as the company contact person made us aware. At the same time, they also shared with us a publication documenting the same ELISA kit being used for B12 detection in cell lysates of Mycolicibacterium smegmatis. Therefore, we finally decided to test the applicability of the kit with E. coli cell lysates and LB medium.
First, we used the ELISA kit to test B12 concentration in LB medium. Additionally, we tested two other media used for E. coli and cyanobacteria cultivation (both supposedly containing no B12), and also cell lysates with different amounts of AdoCbl added (as a validation of our detection method). It was important to test the kit’s responsiveness to varying amounts of AdoCbl in cell lysates (and its usability with cell lysates in general) since that would determine its usability for our project. Lastly, we tested the kit’s specificity with samples containing cobinamide, the precursor we would add exogenously to the medium for B12 production.
A high B12 content was detected in the LB medium and therefore we decided to change our cultivation medium for B12 production experiments to M9 (minimal) medium. However, we also had to change the detection method for further experiments since the kit appeared to have a high cross-reactivity with cobinamide (which would always be present in the production cultures). Next, we realized that converting all cobalamin forms in cell lysates to cyanocobalamin (the B12 form the specific ELISA kit is made for) is not possible for routine measurements in our lab, and if we converted the photo-labile forms (adenosylcobalamin and methylcobalamin that we expect to produce) to hydroxocobalamin, the kit wouldn't allow for precise detection of the latter. Hence, a different B12 detection method was needed.
As an alternative we have chosen ethanolamine medium as an easy and cheap detection method. The principle behind the ethanolamine medium is that it only allows bacterial growth when AdoCbl is present. E. coli produce ethanolamine ammonia lyase (EAL), an enzyme that breaks down ethanolamine into ammonia and acetaldehyde. Furthermore, EAL requires AdoCbl as a cofactor. The fact that E. coli can utilize ethanolamine as a nitrogen source only in the presence of AdoCbl, the ethanolamine medium offers a convincing detection method.
After some unsuccessful attempts at making the actual ethanolamine medium, we finally found a protocol which worked.
The aim of our first experiment was to test whether E. coli is able to grow in the medium with ethanolamine as the sole nitrogen source, both with and without AdoCbl supplementation. We did a growth curve comparison for LB medium and Ethanolamine medium. We furthermore tried to use the differences in growth of bacteria supplemented with different amounts of AdoCbl to create a calibration curve.
We learned that with this method it is not possible to determine precise AdoCbl concentrations, only qualitatively. However as previously published, it is an efficient and easy method for qualitatively measuring if B12 was present. However, we still had the wish to determine the exact concentration of B12 and other precursors in order to prove actual B12 production. We learned that mass spectrometry can be used to determine more precisely how much AdoCbl is present in the cells and whether it really is AdoCbl or another substance. We ended up coming in contact with a research group at the University Medical Center of Freiburg who was willing to do these measurements for us. We are incredibly grateful for their support!
For the functionality of our system a potent toxin is essential. We performed experiments for toxin expression, attempting to assess the lethality of MazF. Unfortunately, the first results implied that our chosen toxin was unexpectedly non lethal to the cells. Subsequently, we searched for the cause of inadequate toxicity.
To test toxin expression, we designed a plasmid containing the toxin gene mazF downstream of the riboswitch, including the associated RBS. We chose to test the toxin together with the riboswitch to directly determine the expression level of the toxin that would eventually be present in our self-regulating system. We added a 6xHis-Tag to the toxin to perform western blot experiments to visualize toxin expression.
We took the toxin sequence, as well as the riboswitch sequence, from a publication and synthesized it as a contiguous sequence. The fragment was then amplified by PCR, cloned into pGGAselect via Gibson assembly and finally transformed into E. coli MG1655.
To test the toxicity of expressed mazF, we measured optical density and determined colony forming units (CFUs) for different doxycycline (DOX) inducer concentrations. We expected to see reduced growth (decrease in optical density), as well as decreasing CFUs. Unfortunately, we did not see a decrease in survival with higher concentration of inducer. Even at a DOX concentration of 1000 ng/mL, at which the toxin should have been sufficiently expressed, no major changes were observed compared to the negative control. However, western blot results showed toxin expression after DOX induction.
We found that our toxin was expressed with an upstream riboswitch, but was not inducing cell death. We had become aware that we might get into trouble with the toxin already early on in the project. We introduced our idea at the biology Faculty Day of the university on June 30th, 2023 (see more on integrated human practise). It is an event for all those interested in biology, and where the various projects of the university's research groups are presented. The question arose to what extent we prevent the cells from mutating the toxin, or the associated promoter, over time. Rationally, bacteria in our cultures which no longer express a functional MazF would gain an advantage, and would therfore prevail. Since our toxin-antitoxin genes do not overlap, as they do in the native system, the cells are not forced to keep the functional mazF gene.
We redesigned the system, so that the bacteria gain a survival advantage when no mutattions occur in the promoter regulating toxin expression. We had the idea to implement a second antibiotic resistance e.g. kanamycin behind the TetR promoter and between the bluB gene and riboswitch and toxin. If the bacteria mutated the TetR promoter, they also lose the ability to express the kanamycin resistance and will not be able to survive the selection pressure(see more on design page).
We redesigned the system, so that the bacteria have a survival advantage when they do not mutate the promoter regulating toxin expression. We had the idea to implement a second antibiotic resistance e.g. Kanamycin behind the TetR promoter and between bluB gene and riboswitch and toxin. If the bacteria mutate the TetR promoter they also lose the expression of Kanamycin and will not be able to survive the selection pressure (see more on the design page).
We also decided to redesign the protocol of our experiment. For the next experiments we collected single colonies from the CFUs and sent them for sequencing to gain more information about the possible mutation of the toxin.
After several cloning cycles we did not manage to clone the kanamycin plasmid, so we decided to focus on cloning and testing the individual parts. We screened several generations of bacteria, which had been induced in our experiments to express the toxin, by sequencing. Primers were used which specifically allow sequencing of the toxin.
According to the sequencing results of the toxin, we had no mutations inside the relevant gene which could have affected the function of the toxin. Apparently, the concerns of toxin mutation do not seem to have any impact on our system in the period we have tested. However, despite this positive finding the problem remained, that the toxin didn’t cause cell death upon induction.
After long troubleshooting, we looked for possible answers as to why our toxin experiments were unsuccessful. We hypothesized that the RBS of the riboswitch might be too weak to produce enough toxin to induce cell death or inhibit cell growth. After extensive literature research,this conclusion was further supported.
As a backbone, we used a plasmid that we received from our supervisor. This already contained the TetR promoter and the rrnB terminator. We amplified our toxin from the previous plasmid and inserted it into the backbone via Gibson Assembly. The final plasmid (piG_23) was transformed into E. coli MG1655.
For simplicity, the experiment with piG_23 and pGGAselect as a control, could be carried out in LB medium due to the lack of a riboswitch. Our results showed a decrease in growth and CFUs following induction with 50 ng/mL DOX.
This was a first indication of increased toxin expression and consequently increased toxin activity in the cells. We suspect that the RBS of the riboswitch was indeed the cause of the previously insufficient toxin activity.
To make sure that our assumptions were correct, we cloned new toxin constructs that contain the different RBS. Thus, a plasmid (piG_23b) was designed with the weaker RBS originating from the riboswitch . In this case, we would expect the toxicity to decrease and the levels to match those of the other toxin plasmid (with riboswitch).
New primers were designed to exchange the respective RBS via Gibson Assembly and to reintroduce them into the backbone. The cloned plasmids were then transformed into E. coli MG1655. We added a 6xHis-Tag to the toxin to perform western blot experiments to visualize toxin expression.
Cells containing piG_23b, as well as control cells containing pGGAselect, showed a steady increase in growth over time in the non-induced and induced state with 100 ng/mL DOX(see more on toxin/antitoxin results page). This result indicates no difference in growth, while significant growth inhibition can be observed for piG_23 cells induced with 100 ng/mL DOX.
We learned that by replacing the RBS of the riboswitch with a stronger RBS, the expression of downstream genes is altered.
Next we would be interested to clone the TetR RBS into the plasmid consistin of TetR/Riboswitch/Toxin/Terminator in addition to the already existing Riboswitch RBS. This plasmid thus contains two RBS in order to guarantee the functionality of the riboswitch. In an experiment with this plasmid, we expect an increased toxicity, similar to piG_23. Unfortunately this was out of the capabilities of our iGEM project.
-  Cai Y, Xia M, Dong H, Qian Y, Zhang TC, Zhu B, et al. Engineering a vitamin B12 high-throughput screening system by riboswitch sensor in Sinorhizobium meliloti. 2018 May 11;18(1).
-  Li GW, Burkhardt D, Gross C, Weissman Jonathan S. Quantifying Absolute Protein Synthesis Rates Reveals Principles Underlying Allocation of Cellular Resources. Cell. 2014 Apr;157(3):624–35
- Li GW, Burkhardt D, Gross C, Weissman Jonathan S. Quantifying Absolute Protein Synthesis Rates Reveals Principles Underlying Allocation of Cellular Resources. Cell. 2014 Apr;157(3):624–35 
-  CHANG GW, CHANG JT. Evidence for the B12-dependent enzyme ethanolamine deaminase in Salmonella. Nature. 1975 Mar;254(5496):150–1.
-  Using a Riboswitch Sensor to Examine Coenzyme B12 Metabolism and Transport in E. coli. Chemistry & Biology [Internet]. 2010 Jul 30 [cited 2021 Apr 5];17(7):756–65. Available from: https://www.sciencedirect.com/science/article/pii/S1074552110002437?via%3Dihub
-  Escherichia coli str. K-12 substr. MG1655, complete genome. NCBI Nucleotide [Internet]. 2022 Mar 9 [cited 2023 Oct 9]; Available from: https://www.ncbi.nlm.nih.gov/nuccore/NC_000913.3?report=fasta&from=2910756&to=2911091&strand=true
-  Lemos JAC, Brown TA, Abranches J, Burne RA. Characteristics ofStreptococcus mutansstrains lacking the MazEF and RelBE toxinâantitoxin modules. FEMS Microbiology Letters. 2005 Dec;253(2):251–7.
-  Li GW, Burkhardt D, Gross C, Weissman Jonathan S. Quantifying Absolute Protein Synthesis Rates Reveals Principles Underlying Allocation of Cellular Resources. Cell. 2014 Apr;157(3):624–35.