For four months we have worked closely together with the BOKU-Vienna iGEM team. Our common objective was ‘Producing a compound using engineered phages as a delivery vector’. We have been in contact via Whatsapp, email and Zoom. After our discussions they changed their strategy and used the T7 phage instead of the originally considered lambda phage. Thanks to their suggestions we also made changes to our project, coming up with methods to select for our engineered phages. When the BOKU-Vienna iGEM team was starting experiments to engineer phages we advised them on phage engineering and troubleshooting. After an interesting discussion about safety we concluded that safety regarding engineered phages is challenging to incorporate, and we reached out to some experts to come up with ideas. We continued our Partnership with some comparative experiments.
During the same time we also worked in synergy with the UT-Austin iGEM team. We both had a shared goal: ‘Maximising heterologous protein production with bacteriophages’. This Partnership was drylab based. The UT-Austin iGEM team created a model to show the effect phage insertions/deletions/replacements have on toxin production. We created a model describing host-phage interactions on a population level scale. To gain more knowledge they used their Pinetree model to run simulations to compare our 3 different GFP replacement locations. Using their simulations we were able to identify that the best site of insertion from our choices was site 4.3. We were then able to use this data to model the phage-host infection and toxin production dynamics in the locust gut (see model 1). The main conclusion we could draw from the results is that late phage genes are most likely associated with high expression.
We agreed to simulate their bacterium/bacteriophage population dynamics with our model. With our model (phage-host infection dynamics in water), we demonstrated the importance of phage infection probability in systems, like theirs, where the cell concentration is low. Based on this we were able to suggest a strategy to improve the design of their detection.
We have really enjoyed working together with such nice teams. It was fruitful to share what we were doing and to get feedback on our experiments, models and project in general. Thank you BOKU-Vienna iGEM team and UT-Austin iGEM team!
Our first meeting took place on the 12th of June via Zoom (Figure 1). During this introductory meeting, we informed each other about our projects.
iGEM TU Delft:
We explained to them our project idea which consists of engineering bacteriophages that contain molecules toxic to the locust. In particular, we explained them our phage engineering strategy, including why we chose T7 bacteriophage as a phage model, as well as why we wanted to use Bacteriophage Recombineering of Electroporated DNA (BRED) as phage engineering strategy to insert a GFP reporter in three positions of the phage genome.
iGEM BOKU - Vienna:
iGEM BOKU-Vienna wanted to help treat septic shock in patients. Their project includes the production of gelsoline with lambda phages to tackle lipopolysaccharides (LPS) endotoxins. LPS are released when gram negative bacteria are killed. This release leads to negative side effects in patients and can even cause septic shock. Using bacteriophages, they want to express proteins that bind to the LPS so that the LPS becomes harmless. For more details about their project you can visit their Wiki.
After sharing our projects, we found out that we had a common objective: ‘Producing a compound using engineered phages as a delivery vector’ . Thus we decided to keep in touch and help each other throughout the iGEM experience.
In July we exchanged information about our project via Zoom meeting (Figure 2) as well as through emails to keep each other updated on the project's progress.
iGEM BOKU-Vienna told us they were using the phiC31Integrase Vector System for BRED. Instead of replacing a gene they were inserting a whole cassette, including promoter and terminator using Golden Gate cloning. They also informed us that our exchange of information made them change their strategy and use the T7 phage instead of the originally considered lambda phage. With this, they aimed to increase the comparability of the outcome of our approaches and eventually exchange protocols later on.
Selection and screening of engineered phages
iGEM BOKU-Vienna asked us how we were selecting our engineered phages. We explained to them that to screen engineered phages, we planned to check fluorescence emitted by our reporter GFP once inserted in the targeted positions of the phage genome. We also explained that we wanted to do PCRs with flanking primers to check for the presence of GFP in the phage genome.
They suggested that we might need an accurate way to verify if the insertion occurred at the right place within the genome. Up until now, we only had primers flanking the GFP gene and we could not determine the location of insertion. After this discussion, and several talks with our wet lab team on how to check recombination, we decided to design primers that hybridise with GFP and are flanked with the insertion site.
In addition, iGEM BOKU-Vienna asked us how we would select for the engineered phages. Thanks to their heads up, we came up with a CRISPR-Cas9 method to select for engineered phages. Later in the project, it turned out that we needed this method since our BRED efficiency was lower than expected.
We concluded that safety was a concern for both of our projects, thus we exchanged different safety ideas. They were replacing the gp10 gene of the T7 phage with their genes of interest. By supplying the missing g10 protein on a plasmid in the host bacteria they prevent uncontrollable spreading, thus ensuring safety. Unfortunately, for our project we cannot engineer the bacteria, as we want to use the bacteria already present in the locust gut. Together we discussed the possibility of kill-switches to stop phage spreading, and quorum sensing to regulate infection of phages.
These sparring sessions helped us think about ideas and to critically analyse if we could apply them to our project. These talks instigated discussions within our own team to see if we could use kill-switches or quorum sensing for our project. Due to talks with iGEM BOKU-Vienna about safety of our projects, we concluded that safety regarding engineered phages is challenging to incorporate. After talking with them about the safety of our projects, we reached out to some experts. With all these discussions, our team concluded that even though safety is a challenging task, it is highly essential. Then, we got inspired to create an extensive safety page for our project (see Safety).
During the rest of July we exchanged more project details and worked out on how we could help each other. We thought about possible dry lab and wet lab collaborations.
iGEM BOKU-Vienna changed to also use the lambda RED recombinase system as phiC31 integrase had some problems. As our projects became very similar we saw opportunities for a wet lab partnership.
Phage characterisation experiments
A noteworthy difference in our projects is that we are replacing non-essential genes with GFP whereas iGEM BOKU-Vienna will replace the g10 gene. In addition, we are using a ‘ready made’ T7 phage and they are expressing linear DNA encoding for the T7 phage. Because of these differences, we decided to perform comparative experiments. The goal of these comparative experiments was to see if our wildtype phages had similar characteristics and to see what the influence of the different gene replacements were. We suggested looking at plaque count and size to compare diffusion speed and virulence of the phages. iGEM BOKU-Vienna suggested we could also compare the dose of infection of both phages to determine the infectivity of the phages.
In the beginning of August we both expressed that we would like to meet in person if possible. We were looking for opportunities to meet up in real life. Unfortunately, at the end of the month we concluded that a meet up would be difficult as travel restrictions due to COVID19 increased.
We exchanged protocols to perform the comparative experiments. We wrote a protocol to compare plaque size and they wrote a protocol to compare infection doses. As both teams, up until this point, did not have any engineered phages we decided to do the experiments only with the wildtype phages. We planned to do these experiments in October.
Phage engineering advice
Towards the end of September, the BOKU-Vienna iGEM team reached out for help as they were starting with the phage engineering experiments. We gave them advice regarding phage engineering and troubleshooting:
- Before starting experiments determine the right initial phage concentration to produce enough plaques, but not so many that they overlap.
- Use primers for the regions flanking the gene, and primers for the inserted gene.
- If not able to obtain the engineered phages by picking plaques, combine all plaques and do a PCR.
- To isolate the engineered phages use a CRISPR-Cas9 system targeting the gene to replace.
The experiments were performed using the wild-type phages of each team (Figure 3). Both teams conducted experiments determining the phage titer needed to clear a liquid culture of susceptible bacteria (infection dose assay) to determine the infectivity of the phages. In addition, the diameter of the plaque size was measured over time to compare the diffusion speed and virulence of the phages.
Results plaque assay
Plaque size experimental results from both teams showed a linear increase in plaque size over time (Figure 4). The average growth rate of the plaques from the phages of the BOKU-Vienna team is 0.48±0.02 mm/h, and from the TU Delft team 0.37±0.01 mm/h. It should be noted that the experiment was only done by both teams once, the experiment should be repeated to draw substantiated conclusions.
Results infection dose
The results of the infection dose assay from the Vienna team showed that the minimal phage dose needed to clear a liquid culture of susceptible bacteria is 1.28 phages per bacteria. Unexpectedly, the results of the Delft team did not show a clear trend between the bacterial culture with any of the tested phage concentrations (Figure 5).
Conclusion phage characterisation experiments
It was interesting to observe that although we were using the same phage, we collected very different results for both plaque assay and infection dose experiments. Both experiments indicate that the phages used by the Vienna team have a higher rate of infection. This is unexpected since they used the same phage.
From these experiments we concluded that our phages, although they have a similar sequence, don’t give the same results. This could be because:
- Our protocols for phage characterisation are not accurate enough.
- There is too much room for human error in our protocols.
The comparative experiments did not give us the expected results. The experiments should be repeated to find out what causes this difference. Ideally we would have done different protocols and compared those results but due to time we were not able to do so.
In conclusion, the partnership with the BOKU-Vienna iGEM taught us a lot. Our interesting discussion about safety regarding engineered phages caused us to reach out to experts. In addition, we improved our methods to select for engineered bacteriophages. We changed our primers and came up with a CRISPR-Cas9 selection method. For the phage engineering experiments we gave them advice regarding phage engineering and troubleshooting. We also wrote protocols for phage characterisation experiments and conducted preliminary experiments.
We were very happy to have worked together with Vienna during iGEM. It was fruitful to share what we were doing and to get feedback on our method and project. We learned a lot from the shared discussions and used this to improve our project. Thank you iGEM BOKU-Vienna!
From July we have partnered with the iGEM team from the University of Texas at Austin. After watching our video ‘Synbio: infinite possibilities’ that we created as collaboration with multiple teams, the iGEM UT Austin team realized that we were also working with bacteriophages and reached out to us. The collaboration has been extremely useful to both teams. We have had many discussions and ran simulations of our models for each other. This exchange of knowledge helped us develop and gain insight into our project further than we could have ourselves. We have incorporated the results from their model into our future perspective for project design.
The UT Austin 2020 iGEM team reached out to us via email late July as they had found out we were also doing research on bacteriophages. They suggested scheduling a meeting to discuss both projects and to see if there would be a possibility to help each other out.
Initially, we exchanged information via email about our projects and concluded that there might be ways to work together. This is why we scheduled our first Zoom meeting in August.
Discussing our project
During our first Zoom meeting with the Austin team (Figure 6) we discussed our projects and goals.
iGEM TU Delft:
We explained the goal of our project: to tackle the locust crisis using bacteriophage. We intended to prove our concept of engineering a phage producing a homologous toxin in the wet lab. The engineering involved the replacement of a non-essential gene of T7 phages with GFP. Gene expression of phages occurs in three stages: early, middle and late, we hypothesised that by replacing early or middle genes toxin expression would maximise toxin production. Thus, we chose to replace 0.6A (early gene), 1.1 (early gene) or 4.3 (middle gene) in the lab to identify the best site of replacement. In the drylab we wanted to model phage infection and toxin production to simulate how our biopesticide, PHOCUS, would work. To this end, we intended to use a model describing the temporal evolution of the susceptible cell, infected cell, phage, and toxin concentrations in an ideally-mixed system. As the locust gut is most likely not an ideally-mixed system, we additionally adapted a 2D model describing also the temporal evolution of susceptible cell, infected cell, phage and toxin concentrations to investigate how the spatial organisation of cells influences toxin production.
iGEM UT Austin:
The goal of their project is to engineer a reporter phage for detection of bacteria in water. They want to use T7 bacteriophage as a biosensor to detect E.coli concentrations in water supplies by expressing GFP. This technology would be applicable to fields where fast and effective water testing is needed. Their work is completely in silico and uses a stochastic gene expression simulator, Pinetree, that simulates protein output. Using these simulations, the team wants to study what would be the most efficient way, in terms of gene editing, to maximise reporter production and minimise the time it takes for the reporter to reach a detectable concentration. For more details about their project you can visit their Wiki.
During this meeting, we found out that we had a shared objective: ‘Maximising heterologous protein production with bacteriophages’. Since the Austin team was working on a dry lab project this year and we were involved in modelling as well, we decided to turn our interaction into a dry lab partnership. This partnership seemed very useful and complementary from the start, as we modelled completely different aspects of our shared goal. While Austin was modelling phage induced gene expression on a single cell level, we were modelling host-phage infection on a population level scale. Therefore, we figured that the results of both models could nicely add to the design of both of our projects.
As explaining models via Zoom is rather difficult, we agreed to send each other more detailed information about the models via email and include some papers.
iGEM UT-Austin’s contribution
Austin uses a Stochastic gene simulator called Pinetree. With this model they can simulate the effect of potential deletions, insertions or replacements of toxins or genes on the final gene expression. They also constructed lysis and burst size calculators that take in the output from the simulation. They offered to run simulations of our mutated T7 sequences to estimate lysis time and burst size and to create visuals of protein output of the mutated T7 phage. With this data, we could improve the design of our engineered T7 phage. Moreover, their model could give valuable insights in what sequence properties, such as promoter/terminator type, promoter/terminator position and RNAse cleavage site, would be important to take into account in the design of any phage.
Our Contribution to UT-Austin
We based our model on the model by Beretta and Kuang (1998), describing temporal evolution of susceptible bacteria (S), infected bacteria (I) and the amount of phages (P). We modified our model to include an equation describing the amount of heterologous protein produced over time upon lysis of infected cells. We offered to model their phages in our system so they could understand the interaction between host and phage and GFP production on a larger scale.
We could run simulations for each other as we both had different models, iGEM UT-Austin could model the effect of phage insertions/deletions/replacements on toxin production and we had a model describing host-phage interactions on a population level scale. Our model could help them understand which initial bacterial and phage concentration results in what GFP concentration and which parameters are relevant to design a quick detection method. Their model could help us understand where we should ideally insert our toxin to ensure maximum production and what effect different insertions have on the toxin production.
Creating concrete plans
After understanding their model, we scheduled a new meeting to ask more detailed questions and to make concrete plans on how to work together.
Our Contribution to UT-Austin
During our second Zoom meeting (Figure 7) we agreed to simulate their bacterium/bacteriophage population dynamics with our model. This could give them two insights:
- We can determine how much time is needed to produce enough GFP for detection. This would give them an indication for the speed of their system.
- We can determine how many initial phages would be required to lyse the cells quickly in their system.
For this we needed their parameters for:
- The concentration of cells in their system
- The lysis time and burst size of their phage according to their simulations
After the meeting via email, Austin indicated that they had not decided yet on the final mutated genome, which would influence production of GFP. It would not help them much if we modelled the production of GFP without knowing the final genome ,as the amount of GFP produced per cell depends on the final mutated genome. However, they could use our model to predict the number of lysed cells from which they can deduce the total amount of GFP produced, knowing the amount of GFP produced per cell and assuming no GFP degradation. Their model suggested that there is a tradeoff between lysis time and burst size, so they also wanted to know whether it would be best to prioritise lysis time or burst size. They wondered if we could predict the number of cells lysed given the burst size and lysis time. We answered several questions they had about the model and parameters, and agreed to also investigate this new research question. We agreed to send an accumulative graph showing the total number of lysed cells after a certain time.
iGEM UT-Austin’s contribution
Austin would generate graphs for the average amount of heterologous protein produced at lysis for different positions on the viral genome and the distribution of lysis time and burst size for regular and codon optimised genes. They would use their Pinetree model to run simulations to compare our 3 different GFP replacement locations. This could give us different insights:
- Which gene replacement position (0.6A, 1.1 or 4.3) yields the highest GFP/toxin expression.
- The effect of insertion place or codon optimisation on the lysis time and viral particle production of the phage (phage fitness).
From us they needed:
- Our insertion sites (or actually which genes we replaced)
- The annotated gene sequence Genbank format of our GFP/our toxin
iGEM UT-Austin’s contribution - preliminary results
During further email contact, Austin explained that the toxin output is expected to be approximately equivalent to the output of the native gene at that site. Therefore, it is wise to choose native genes with high protein expression levels as replacement sites, for example those coding for capsid proteins. With the simulations they had previously done for their own experiments with their model, they could already conclude that our proposed replacement sites are not very effective at producing a large quantity of toxin (Figure 8). For instance, 0.6A is an early gene with a low protein output because it is transcribed by E. coli polymerase. We learned that in the T7 phage, early genes do not have a strong promoter and that upstream terminators can hinder expression of later genes downstream. They could already identify that site 4.3 would probably be the best out of the three , producing an average of 1596 proteins per cell lysed. Therefore we were able to use this data to get preliminary results for our own model. They suggested that we later compare this to the output of our experiments with the engineered phages.
On the 30th of September we scheduled a third Zoom meeting (Figure 9). During this meeting we finalised the discussions on what we would need from each other and when we would send each other the results.
Running the simulations
Our results for UT-Austin
Nearing the end of the competition, our models were running smoothly, enabling us to perform simulations for Austin with the data they sent us (Table 1). We tweaked our model and set up parameters based on the aquatic environment they expected in their design.
|Genome||Lysis Time||Burst Size|
For each genome we ran the simulation with E.coli densities (1,10,88,126,1000 CFU/100ml) and MOI (1,10,100) (code). In total, we produced 45 sets of graphs for them, each showed the change in populations of susceptible cells, infected cells, phages and lysed cells for the conditions given. Figure 10 is an example of such a graph for the gfp-h-h-gp10 genome, 1000 CFU/100mL and a MOI of 100.
These results showed us that the used cell densities are too low for the phages to successfully propagate within the time frame of 2 hours. This can be attributed to the fact that the rate at which susceptible cells are infected in the model (ε) is too low for these low cell densities. This parameter is very important for the system as it shows the rate at which 1 bacteria infects 1 cell in 1 ml of water. Therefore, when the concentration of phage and bacteria is high, the overall rate of infection is also higher. This would imply that, in diluted conditions such as water, the probability of infection is very low. As expected, when we additionally simulated higher concentrations (such as 106-107 CFU/100ml), the results were faster phage propagation and more lysed cells. Therefore we first suggested increasing the initial phage concentration or using methods to increase the rate of infection, such as sample mixing. Even with this solution, since the number of cells lysed is low, the concentration of GFP will probably be too low for an accurate reading, so instead we suggested that they concentrate bacteria by using a filter before testing. This would result in a much higher likelihood of infection of the bacteria by the phages and consequent GFP production for a reliable readout. For more detailed information see the partnership page of UT Austin.
iGEM UT-Austin’s results for us
Our primary goal of UT Austin’s model was to get an insight into how the 3 different replacement sites we chose (0.6A, 1.1 and 4.3) would influence the production of the toxin (Cry7Ca1; see Design). However, we took this opportunity to also investigate the effect of codon optimisation on the final expression. Furthermore, we decided to let Austin simulate the expression of eGFP at the three replacement sites. This simulation could then be nicely compared to the wetlab experiments we performed. Therefore, we sent them the following sequences:
- Cry7Ca1, the full sequence, codon optimised
- Cry7Ca1, the full sequence, native protein
- Cry7Ca1, shorter sequence (higher toxicity, see Design), codon optimised
- our eGFP, which we have engineered in T7.
Unfortunately due to some technical simulation error, the replacement of the 0.6A gene was not simulated and only results for the 1.1 and 4.3 sites were obtained. We used their data in our model to predict how much heterologous protein would be produced. Based on their model it was better to use codon optimisation as it yielded more proteins per cell for the full protein (Table 2). While the difference is very small for 1 cell, it might make a large difference on a larger scale, suggesting all genes must be codon optimised for the host cell.
|Protein||Production at site 1.1||Production at site 4.3|
|Full native Cry7Ca1 sequence||886||1551|
|Full codon-optimized Cry7Ca1 sequence||945||1616|
|Short codon-optimized Cry7Ca1 sequence||770||1552|
We had planned a set of experiments to identify the best site of gene replacement but this was not possible due to the time constraints in the lab set by the pandemic. Nevertheless, the data from Austin helped us identify that 4.3 would probably be the best site from our choices (Table 2). Using their simulation we were also able to identify that site 10A could be even better as 40,983 native proteins are produced per cell lysed are produced, compared to 1552 produced in 4.3 (Figure 11). This site encodes for a capsid protein which is a late gene. To maximise protein production by other phages, it would seem best to insert at similar sites in their genomes. For unannotated phages, we could identify genes for capsid proteins in using an existing phage annotation software like Phaster . However, each phage has a different genome size limitation and insertion could potentially lead to a non-functional phage. The preliminary results based on software must be investigated using experimentation to reject this possibility.
Although their Pinetree simulations were very useful, these simulations were specific for E.coli and T7. This could be a limitation when implementing other phages and cells. For instance, the final design of our biopesticide will probably consist of a cocktail of phages and not specifically the T7 phage. Ideally, we would want to run similar Pinetree simulations, but then for different phage and bacteria combinations. Unfortunately, this is yet only possible for very well characterised bacteria and phages, like T7 and E.coli, as the simulation strongly depends on experimental data for transcription and translation rates and on the completeness of the phage annotation.
The partnership with UT Austin provided us with insights in which genomic traits of a phage are important to account for when engineering a phage to maximise protein expression. The main conclusion we could draw from the results is that late phage genes are most likely associated with high expression. Therefore, a valid engineering strategy could be to identify the site of a robust late gene, like a gene encoding for the capsid protein, and insert our toxin in close vicinity of this gene.
From our simulations we were able to show Austin the importance of the rate constant at which susceptible cells are infected. This is a representation of the phage infection probability in systems where bacteria concentrations are diluted. To ensure phage propagation in such systems we recommended using high phage doses and methods, like well-mixing, to increase the phage infection probability. Furthermore, as the number of (lysed) cells is rather low, we would suggest a concentration step to obtain a reliable readout.
Our collaboration with the UT Austin team was successful. The partnership resulted in interesting insights for us regarding the best placement of genes in the T7 phage genome. We believe that our model helped them gain an understanding on the amount of GFP produced by their selected phages. To that end, we were able to suggest simple ways to improve their design.
- Arndt, D., Grant, J., Marcu, A., Sajed, T., Pon, A., Liang, Y., Wishart, D.S. (2016) PHASTER: a better, faster version of the PHAST phage search tool. Nucleic Acids Res., 2016 May 3.