I promise we didn't practice evolving any humans!
Throughout our education, all members of our team have noticed that much of modern synthetic biology is often functionally paywalled. This financial hurdle can arise from the need for specialist equipment, skills and expertise, proprietary science, or safety guidelines. We are privileged to go to a university that helps mitigate these problems for its students, with access to experts, cutting edge laboratories and academic literature. However, not everyone shares the same access and opportunities as we do. Our project was inspired by how open science focuses on bridging this gap, and we focused on three main goals:
Bioreactors are fundamental for many experiments but entry-level bioreactors are often prohibitively expensive. Open-source hardware and software are now sophisticated enough to emulate many basic functionalities of commercial research bioreactors, but at a fraction of the cost. For our specialist technology we developed an affordable, customisable turbidostat bioreactor using off-the-shelf open source components and software. This was designed to complement our synthetic biology goal: a genetic circuit for a cell-based, in vivo form of directed evolution that dispenses with reliance on phages and all the complications that phages can present in lab settings. Finally, we collated a series of genetic manipulation tools into a single bioinformatics toolbox with an accessible, intuitive interface.
The goal of our bioreactor was to maximise functionality, ease of construction and customizability whilst minimising cost and required expertise. The choice to build the main vessel from perspex over glass was due to perspex’s physical and chemical resilience, and its compatibility with regular hand tools. Opting for plastic over glass for the main vessel will lead to perspex degradation over time, meaning the bioreactor is not suitable for long-term continuous culturing and the main vessel will eventually need replacement. However, perspex is 100% recyclable. But using plastic over glass means we must rely solely on chemical sterilisation with Virkon – a lengthy process.
This same ethos informed our decision to use Raspberry Pi and Python as the fundamentals for our electronics and programming. Both are hugely open-source, with massive online communities and resources that allow people from different skill sets and backgrounds to educate themselves and connect with each other.
Cell-based directed evolution was an attempt to move the technique away from phage-based methods, along with the extra variables and potential contamination risks that phages inherently introduce into paradigms. Moreover, removing reliance on phages allows for directed evolution to be performed more safely and in a wider variety of labs.
We consider the bioreactor to be a success. The entire bioreactor cost under £200 (including wastage) and is fully autonomous. The MicroPython code is versatile and largely based on existing code available online. The turbidostat is capable of automatically adjusting inflow in response to changes in turbidity. It can be built mostly using off-the-shelf components and hand tools, substituting a coping saw, hand saw, hand drill and sandpaper/file for laser cutters, band saws, pillar drills and belt-sanders respectively. It is certainly the first table-top toroid turbidostat bioreactor in iGEM and quite possibly the first ever table-top toroid turbidostat of this budget level.
We adjusted our goals from a fully functional genetic circuit for guided evolution, to testing of individual genetic components to provide a proof-of-concept for future scientists to test and build upon. This was partly due to logistical constraints stemming from factors such as COVID-19 and Brexit, which was partially what led to our Distribution Kit arriving too late. However we are aware that our project proved to be more ambitious than we had originally anticipated. We are nevertheless confident in the theoretical validity and soundness of our genetic components as they have all been tested before in literature (even if we’re still assembling our constructs).
Directed evolution is an inherently powerful tool capable of being used towards works of either great benefit or harm. Theoretically, our cell-based technique is a versatile tool that has borderline limitless applications for all forms of directed evolution. A simple but very realistic possibility of misuse would be using it to alter genes of pathogens to make them more resistant to conventional antibiotics, in order to create a new superbug.
In order to mitigate both intentional and inadvertent harm coming to anyone, we followed lab protocol and safety guidelines to the letter. We kept track of all our samples of V. natriegens, E. coli and all other genetic material, storing them safely and securely to guard against their misplacement or abuse. Furthermore, during outreach events at all age and education levels we continually advocated for responsible, ethical and sustainable applications of synthetic biology.
As we are competing in the foundational advance category, our success could have wide-reaching implications for a range of communities if future work is based upon ours. Our bioreactor also has theoretical and practical uses outside of a foundational context as well, and hopefully will help bring responsible synthetic biology into a more attainable category for people around the world.
However, our work may still raise objections from some. Our use of V. natriegens was due to its rapid doubling time and the obvious applications it has for directed evolution. However, it is a relatively new chassis organism with typically proprietary media and comparatively little academic research. Certain communities may feel that synthetic biology is akin to “playing God”, and could take exception to our goal of making synthetic biology and genetic engineering more attainable for more people.
From an early stage, our team was interested in building modular plasmid systems for directed evolution in both prokaryotes (Escherichia coli) and eukaryotes (Saccharomyces cerevisiae). Previous talks with various academics and throughout some of our studies, we came to understand that PACE (Phage Assisted Continuous Evolution) typically requires a dedicated laboratory due to the high risk of contamination that phages present. We hence saw how profound a new, cell-based form of directed evolution could be and decided to take this on as one of our overarching goals. Our team was also enamoured with the prospect of taking on quite a difficult proof-of-concept application (evolving a promiscuous enzyme to produce just one product, lactose, instead of its other many side-products). Alongside this we saw the value in creating a bioreactor to accelerate plasmid engineering via cell-based in vivo directed evolution, as a successful bioreactor would rapidly automate and accelerate this phase.
Our first meeting was with Joe Price, the CEO of a directed evolution startup called Evolutor. He advised that our original proof-of-concept in food technology (producing lactose) is an inadvisable place to begin due to a great deal of red tape in place to prevent commercialisation of GMO products with sequences that differ substantially from their natural counterparts. Joe instead suggested that we targeted the biochemical or biomaterial markets instead. We saw how much more applicable a fully developed method of directed evolution could be for optimising experimental or existing molecules with applications in these industries instead.
Joe then gave us a large list of companies to look into as potential customers for our evolved proteins. He also put us in touch with Sara Holland, an expert in both synbio and its regulation within its legal system. Sara henceforth became a valuable point of contact for subsequent regulatory queries, and has greatly informed further decisions and how we view our project in industry and societal contexts. This reinforced our decision to stay away from food technology applications, and in keeping other industries in mind.
After a period of research and collective introversion, we organised a meeting with Tuck Seng Wong. Tuck has devoted much of his professional career towards progressing directed evolution, and is in many ways a potential future end user if our project one day proves successful. He immediately saw the value of a cell-based directed evolution method, and was excited by the prospect. When we expressed our interest in performing this in both E. coli and S. cerevisiae he cautioned heavily against it, highlighting how ambitious our project goals actually were and that we would likely have more success focusing on something more specific. Hence, we decided to focus solely on bacterial evolution.
Further to focusing our project on bacteria, Tuck suggested Vibrio natriegens as an even faster-growing chassis organism than E. coli. With theoretical doubling times as low as 7 minutes, and being highly compatible with many existing protocols already used for E. coli, we saw the potential value V. natriegens had to our project and quickly resolved to base our work upon it. To simplify our proof of concept, Tuck suggested we stay away from something as complex as lactose and instead focus on restoring antibiotic resistance. Tuck also mentioned that, out of the existing targeted mutation systems, MutaT7 with the base-editor fused to a T7 RNA polymerase was one of the best tested. In the context of PACE, Tuck revealed that despite widespread use it results in mutations of all phage DNA, meaning it cannot be used to target a single gene. Hence we integrated both of these suggestions into our final MutaT7 Test Cassette, and were vindicated in our decision to perform cell-based directed evolution.
Regarding our bioreactor Tuck said that nothing fancy was really needed. As a real-world customer and end user performing cutting edge research, all that was required for him was a bioreactor with controllable stirring, pumping, temperature and OD measurement would be sufficient. He was intrigued by the toroid design, something he had not come across before. As a result of our talks with Tuck, our project became focused on MutaT7 evolution in V. natriegens coupled with a simple turbidostat bioreactor to accelerate growth and evolution phases.
To gain more insight into bioreactor design, we organised a meeting with Krys Bangert, Technical Team Lead of Bioreactor Design at the University of Sheffield. We had the original idea of having a laser for our OD measurement as we are likely to have issues with self-shading if in the future we wanted to culture photosynthetic organisms. We hence settled on a regular LED-phototransistor pairing, which in hindsight actually adds to our open-source bioreactor goals. Krys also highlighted some common issues with stirring, such as cell lysis from crushing or turbulent eddies that are similar in size to cells (especially if cells have weak walls). Overall our magnetic stirring system was likely to be fine so long as it’s gentle, and we could check this by looking at cells under the microscope to verify that there weren’t an undue amount of cells lysing. Krys also suggested using OD800 over OD600 wavelengths for our OD measurement system, as it could help minimise background radiation interference if we use perspex as construction material. Following talks with Krys we had a much better understanding of bioreactor design, and were confident in beginning constructing our own.
During the Münster Meetup we met René Inckemann, instructor for the 2018 Marburg Team and now Vice Head of Steering of the German Synthetic Biology Association. Through a happy coincidence, it turned out that René’s team had worked with V. natriegens in their submission and had achieved doubling times as low as 7 minutes! Discussing with and gleaning insights from René led to us incorporating some of his advice in our media optimisation experiments, with the goal of making them more applicable to real scientists who work with V. natriegens such as himself. In other words, we should keep in mind the likely culture conditions that end users were likely to provide for their cultures and not focus solely on achieving the lowest possible doubling time, as the fastest doubling time may not be compatible with feasible lab practices for many users.
Another person we met during the Münster Meetup was Jan Kalkowski, the GASB Education Officer and a Münster Meetup Judge. We were already aware that our plasmid system poses a bit of a problem when you evolve a successful gene and want to find its sequence. Our original system had the gene of interest (GOI) on a high copy number plasmid, so even if a cell shows an increase in fitness, it’s not clear which of the potentially hundreds of copies of that gene are responsible for that improvement. Jan pointed this out to us and suggested that we change our approach so the gene being evolved is only present in one copy in the cell. Hence, our short-term solution was to move the GOI and selection system to a low copy number plasmid. This should make it possible to get 1-5 sequences and run confirmation experiments to see which sequence is responsible for the increased fitness. Our longer-term solution is to eventually integrate our GOI into the genome. This would solve Jan’s problem of having several copies of the mutated gene in most cases, but when a cell is dividing as fast as V. natriegens, there may actually exist several partially replicated copies of the genome in any given cell at any given time. In the future then, we should test if slowing growth by lowering the temperature or something similar would result in fewer duplicates of our GOI.
Not long after the Münster Meetup, Tuck Seng Wong pointed us in the direction of Michael Magaraci, who was the Team Advisor for iGEM UPenn as well as a Research Scientist at Protein Evolution. Michael had spent his whole PhD working on directed evolution, and proved to be invaluable in highlighting some of the most common pain-points in the process as well as how well rEvolver might address them. He shared how long his cycles of error-prone PCR, transformation, screening, and reisolation took — each round took about a week of full-time effort. This meant that his project, which required 10 rounds of evolution, fully consumed his time for 10 weeks! Moreover, the intensely manual nature of much of the work increased the likelihood of making mistakes, potentially losing a week’s worth of work. By making continuous evolution a viable alternative could have saved weeks of effort. On top of this, Michael highlighted how he found that growing the cells was often painstaking due to how long it took, and how often one would need to come in and refresh the cultures. Our bioreactor can dispense with manual effort almost entirely and help to maintain a much more controlled environment, and adopting V. natriegens into his work could speed up the whole process.
We also highlighted to Michael how we originally wanted to tie our continuous selection to growth rate using a set of growth slowing genes which could be up or down regulated in response to a biosensor signal within the cell. We believed these might be convenient to use since they don’t require any special media or culture conditions, Michael mentioned that evolution always seems to find the easiest way around a roadblock. He was concerned that our microbes would just evolve to disrupt the growth slowing gene. While our targeted mutations of the ORF make it more likely to improve our gene than break the growth slowers, we also decided to take an alternative approach by using growth enhancing genes like sorbitol dehydrogenase, which could provide cells with a vital second carbon source when growing on minimal media. It’s harder to evolve these gains in function than it is to generate a loss of function in a slowing gene.
However, one problem we haven’t managed to fully address is Michael’s emphasis on the importance of selecting for exactly what you are looking for. His PhD project was creating an infrared fluorescent protein for imaging deep tissue (eukaryotic cells). The issue was that because the directed evolution process was carried out entirely in E. coli, the product was an excellently functioning protein in prokaryotes but exhibited little to no fluorescence at all once expressed in eukaryotic cells. In the future, expanding the rEvolver toolkit to other model organisms would greatly help in solving this issue, but as per Tuck’s advice this is not something we are currently addressing.
In September we met virtually with Douglas Kell, Systems and Computational Biologist, and CEO of the Biotechnology and Biological Sciences Research Council. Douglas was the first to really flag up the potential trouble with transforming three plasmids simultaneously into our cells. In response to this feedback, we started looking at genome editing instead of CRISPRi knockdowns so we don’t need to keep our third plasmid in the cells with the other two simultaneously. Douglas was again dubious about the growth slowing genes — worried that, like Michael Magaraci, the cells would just evolve to disrupt our growth slowing genes instead of improving the GOI. Moreover, Douglas was similarly critical of an antibiotic-based method as cells can evolve general antibiotic efflux pumps (instead of the biosensor upregulating a synthetic resistance gene).
On a more positive note, Douglas thought that our preliminary plans for modelling the evolution system were good and suggested that we compare our simulated results to nature using a phylogenetic approach. This led us to comparing our generated diversity to that existing in nature as seen in our NmLgtB BLAST data. We were also worried that our system would be quite limited compared to error-prone PCR, but Douglas mentioned that error-prone PCR is actually also quite likely to get stuck in local minima. Even if our system is a bit limited in the diversity it can generate, other existing systems like error-prone PCR are very popular in the field and we are confident that our system could find a place as well.
When we mentioned our media optimisation work, Douglas was quite enthused and suggested that we focus on this so that our V. natriegens cells are as nice to work with as possible! We took this advice and followed up our media screening experiment with a second round of optimisation. Finally, Douglas, like Michael, mentioned that it would be super valuable to have a version of rEvolver that works in eukaryotes and even pointed us towards the fastest doubling time eukaryote, Kluyveromyces marxianus. After this, we decided to narrow our future plans from working on just another eukaryote (like S. cerevisiae) to this new, faster growing organism!
Our final meeting was with Daniel Stukenberg, on René Inckemann’s advice. Daniel was Marburg’s 2018 Team Leader and V. natriegens expert. In all of his work with Vibrio, Daniel recently published a new system called NT-CRISPR for genome editing of V. natriegens. We integrated this with our advice from Douglas Kell regarding three plasmids being difficult to manage, and resolved to use NT-CRISPR to knock our targeted DNA repair genes out of the genome instead of knocking them down. We also brought up that we were working with the Vmax Express strain of V. natriegens, and Daniel mentioned that not many actually use this strain. To make our media optimisation results and codon optimisation for V. natriegens more relevant to the scientists that are already working with this version, we should use the wild-type strain and with a -Dns gene knockout. Alongside all this advice and insight, Daniel has been generous enough to send us the plasmids for NT-CRISPR and the wild-type strain of V. natriegens which we will most likely be using for next year’s project.
When supervising the Marburg team, Daniel had actually planned a quite similar directed evolution project in V. natriegens but this was cancelled due to COVID. He said that it would have been good to have some evidence showing that evolution is actually sped up by using V. natriegens. This is what inspired our modelling of gene fixation rate and its dependence on growth rate. Daniel also mentioned that some even more targeted mutation could be valuable, citing how they were looking at using EvolvR originally. This has led us to consider HiSCRIBE based mutagenesis in our future plans (allowing for the targeted mutation of arbitrary small DNA regions).
Lau K, Thon V, Yu H, Ding L, Chen Y, Muthana MM, Wong D, Huang R, Chen X. Highly efficient chemoenzymatic synthesis of beta1-4-linked galactosides with promiscuous bacterial beta1-4-galactosyltransferases. Chem Commun (Camb). 2010 Sep 7;46(33):6066-8. doi: 10.1039/c0cc01381a. Epub 2010 Jul 12. PMID: 20625591; PMCID: PMC3114949.
Rocha, S.N., Abrahão-Neto, J. & Gombert, A.K. Physiological diversity within the kluyveromyces marxianus species. Antonie van Leeuwenhoek 100, 619–630 (2011). https://doi.org/10.1007/s10482-011-9617-7
Stukenberg, D., Hoff, J., Faber, A. et al. NT-CRISPR, combining natural transformation and CRISPR-Cas9 counterselection for markerless and scarless genome editing in Vibrio natriegens. Commun Biol 5, 265 (2022). https://doi.org/10.1038/s42003-022-03150-0
Halperin, S.O., Tou, C.J., Wong, E.B. et al. CRISPR-guided DNA polymerases enable diversification of all nucleotides in a tunable window. Nature 560, 248–252 (2018). https://doi.org/10.1038/s41586-018-0384-8