ENGINEERING SUCCESS
"Science can amuse and fascinate us all,
but it is engineering that changes the world."
- Isaac Asimov
"Science can amuse and fascinate us all,
but it is engineering that changes the world."
- Isaac Asimov
Throughout its course, the AflaxOFF project has experienced a lot of ups and downs. Through ongoing engagement with stakeholders, in-depth literature reviews, and measurements taken in our own laboratory, we have been constantly redefining our project, both in its technical and practical dimensions, in order to find the best answer to the problem we are addressing.
Our Engineering Success lies in the search for the most suitable chassis for the implementation of our solution. This involved the challenge of shifting a molecular machinery initially designed to be functional in one organism, to a whole new one. We believe this pivotage to be a proof of our project's flexibility and resilience.
Early in the project, once our initial idea was determined, we defined the requirements for our engineered organism. We needed to construct an organism which was capable of detecting AFB1 and produce sequestering molecules upon such recognition.
Firstly, due to the strong biomedical orientation of our project, we thought about Lactococcus lactis, due to its probiotic nature. However, through dialog with advisors and early investigations on the available literature, we noticed a lack of well-characterized tools for engineering sensing platforms in this gram positive bacterium.
General architecture of the EMeRALD platform: scFvs are used as ligand binding domains (LBDs)
Therefore, we decided to perform our proof of concept in Escherichia coli due to the greater amount of knowledge supporting its engineering and its fast growth and genetic manipulability. In order to render this bacterium responsive to AFB1, we found a promising option in the sensing-response platform EMeRALD. This system is based on the modification of the ligand binding domain of E. coli’s CadC transmembrane transcriptional activator, rendering a receptor activated through induced dimerization by a ligand of interest.
To ensure maximal capture of aflatoxin B1 by our bacteria upon toxin detection, we came up with the idea of regulating the system through positive feedback. We designed a construct where receptor expression is maintained by its own regulated promoter, pCadBA. This way, only a few receptors are produced due to promoter leakage when AFB1 is absent. Once the toxin is detected, the expression is triggered, significantly increasing the number of receptors on the membrane and enhancing aflatoxin B1 adsorption.
With our design in mind, we ordered certain parts to start testing them individually in the laboratory. We addressed the building of our genetic constructs with modularity in mind and concluded that synthesizing the entire receptors, which differed only in its ligand binding domains, would be a waste of resources and loss of maneuverability, especially for future design optimizations. Therefore we decided to split the receptors into domains, with CadC and the artificial transmembrane region Leu(16) on one side, and the scFvs as LBDs on the other.
Proposed extended-MoClo for assembly of EMeRALD-based receptors: a novel fusion site is designed based on the sequence of CadC to generate new receptors in a straight-forward way
The whole transcriptional units were intended to be assembled through Golden Gate, following MoClo standard and adding a novel fusion site for the scarless and in-frame fusion of both domains of the CDS. This was achieved by taking the four first nucleotides from the constant part, i.e., the signaling C-terminal domain of the receptor, as the fusion site, which were also appended to each of the LBDs.
Prior to this assembly, we pre-cloned all of our synthesized parts into SEVA181 plasmid thanks to flanking EcoRI and PstI restriction sites. By using an RFP expressing device as dropout, we were able to select the positive transformants. However, during this stage of our lab work, we faced certain difficulties with minipreps, which could affect our downstream assembly efficiency. Therefore, we decided to suspend our wetlab tests until a solution was found.
In parallel to our wet lab activities, we worked computationally to implement and investigate the model developed by Lu & Wang (2017) on receptor heterodimerization in order to gain insights into the behaviour of our EMeRALD-based system.
During our first trial stage, we aimed to evaluate whether the expected functionality of our parts had been accomplished, and gather data on their behaviour for further optimization and increase of complexity in future cycles.
Our tests involved addressing receptor localization, CadC signaling capacity and efficiency of AFB1 binding, by means of coupled-protein fluorescence and AFB1 determination through ELISA, both in resuspended cell pellets and supernatant.
Despite our pause of laboratory activity, we still managed to extract information on the system’s expected behaviour through our in silico model. This was achieved by performing real-time and steady-state simulations of the system, addressing the relationship between receptor signal and key variables such as ligand and receptor concentrations.
Simulation of a ligand-induced heterodimerization equilibrium (concentration-time); the curves represent the evolution of the different complexes over time: P1, unbound subunit 1; P2, unbound subunit 2; P1L, ligand-subunit 1 monomeric complex; LP2, ligand-subunit 2 monomeric complex; P1LP2, ligand-heterodimeric complex
A first simulation was run with set initial concentrations of both of the receptor subunits and ligand. As expected, the amount of dimer-ligand associations increases gradually over time. However, certain subunits remain inactive in the form of monomer-ligand complexes.
Steady-state concentrations of the different complexes of an heterodimerization equilibrium as a function of total ligand concentration: P1, unbound subunit 1; P2, unbound subunit 2; P1L, ligand-subunit 1 monomeric complex; LP2, ligand-subunit 2 monomeric complex; P1LP2, ligand-heterodimeric complex
When the steady-state concentrations of the different species were assessed with varying concentration of ligand in the medium, we noticed a peculiar dynamic. In contrast to ligand-monomer equilibriums, where active complex concentration increases with [L] until saturation, the concentration of the dimeric complex was found to experience a maximum value at [L]max, with a following decay due to the accumulation of monomer-ligand associations. In addition, we found this value [L]max to be dependent on total receptor concentration, with a positive relation between both.
Finally, when the analysis performed above was repeated with two concentrations of receptor differing by a 10-fold, a substantial difference was observed in both the concentrations of ligand at which signal experiences an increase, the slope of such increment, and the ratio of receptor activation at [L]max, with all of them being greater at higher levels of expression of the receptor.
Based on the results derived from our model, we could reach the following conclusions:
In view of the analysis retrieved from our calculations, we decided to search for alternatives to AFB1 adsorption different from the own receptors. While reviewing the literature with such objective, conflicting information was found about the interaction of AFB1 with Escherichia coli.
We read some articles indicating that AFB1 was adsorbed onto the bacterium’s cell wall, while others supported that it caused serious damage to this structure (Chen et al., 2020). These findings raised concerns for our approach: if AFB1 cannot penetrate the cell wall, then it cannot be sensed by the receptors; and what is more: if AFB1 alters the cell wall’s structure, then the bacterium’s viability will be impaired.
The accumulation of uncertainties regarding the use of Escherichia coli as our model organism led us to an unavoidable decision: to ensure our project's viability, there was a need to migrate our system to a novel chassis.
After our decision to turn to a new organism different from Escherichia coli, we came back to our initial idea of engineering the probiotic Lactococcus lactis, which harbors a different surface structure with a thicker cell wall. We attempted to find in this species, orthologs of CadC and other one-component systems from gram positive bacteria, such as Bacillus subtilis.
Orthologs are genes from different species which share a common ancestor and retain a similar function. This concept was originally conceived by Fitch for phylogenetic inference. Nowadays, orthologs are used to study and predict gene function, being fundamental for gene annotation and the advancement of functional genomics. This approach is of great relevance given that the vast majority of genes in sequenced genomes will never be addressed experimentally.
Table with orthologs inferred with OrthoFinder for Escherichia coli and Lactococcus lactis proteomes
To find potential orthologs, we went beyond literature and applied OrthoFinder. This software is an innovative method for inferring orthologs of protein-coding genes, outperforming other methods in accuracy and speed, with improvements ranging from 8% to 33%. Unlike other approaches that may omit short genes and incorrectly group long genes, OrthoFinder normalizes alignment results based on sequence length, allowing it to infer orthologs from incomplete data with high precision.
Unfortunately, we were unable to identify any suitable candidates to engineer a system in Lactococcus lactis using OrthoFinder.
In this context of instability, the idea of shifting to the eukaryotic host Saccharomyces cerevisiae was proposed. Several evidence supported the idea of S. cerevisiae, including the ability of AFB1 to reach the cell’s membrane (Su et al., 2023), the deep research around this organism and the existence of a probiotic strain named Boulardii.
However, this 180-degree change, transitioning from a prokaryotic to an eukaryotic system which we had not previously worked with, involved migrating our EMeRALD based system to yeast, keeping up with the same principles of functionality and modularity.
General architecture of the GEMS platform: scFvs are used as ligand binding domains (LBDs) and the split ubiqutin system is exploited for signaling: LBDs, ligand binding domains; EpoR, erythropoietin receptor; Cub, C-terminal ubiquitin fragment; NubG, N-terminal ubiquitin fragment with I13G mutation; sTF, synthetic transcription factor
After extensive research, we found a great solution in a system proposed by Su et al. (2023) consisting of a synthetic receptor based on the merging of the GEMS-platform (Scheller et al., 2018) and the split-ubiquitin system (Sinder et al., 2010). Such platform harbors an scaffold derived from EpoR and ligand binding domains which drive the dimerization and reconstitution of ubiquitin, mediating receptor activation through the release of an attached transcription factor.
In order to solve the potential hurdles of our previous design, the functions of detection and adsorption were separated into individual molecules, with the later being now performed by a single-domain antibody (sdAb) expressed in the cell’s surface upon receptor signaling. In addition, to maintain the positive feedback, the transcription factor is also produced in its free form upon receptor activation. Both responses, as well as a fluorescent protein as a reporter, were facilitated thanks to the addition of IGG6 into our set of parts, which would allow us to support polycistronic expression of all the proteins without dramatically increasing the complexity of our construction.
Using a similar methodology as with E. coli, an assembly strategy was proposed based on an extended-MoClo with two additional fusion sites, determined by the core protein domain of the EpoR scaffold, and permitting the construction of several different receptors by exchange of, this time, both the LBDs and the intracellular signaling domains.
Growth and transformation media for Saccharomyces cerevisiae: 40% filtered glucosa (top-left); 1M lithium acetate (top-center); YP, Yeast-extract Peptone (top-right); selective SC-ura agar plate (bottom-left)
Prior to implementing our parts to the new host S. cerevisiae, much work had to be done to learn how to grow and engineer this yeast. The first step to this purpose was preparing the media required for its growth, including selective media without uracil for selection of colonies harboring a gene complementing uracil auxotrophy, as well as the reagents necessary for yeast transformation. In addition and as importantly, we required a shuttle vector that would allow for replication and selection both in E. coli, where we would be performing cloning and assemblies, and S. cerevisiae, the subject of our assays.
For this purpose, we began transforming plasmids from the Open Yeast Collection (OYC) provided in the iGEM Kit Plates, with the intention of generating both an episomal and integrative plasmid through Golden Gate Assembly.
Despite the verification of most of the parts required for the assembly of our plasmid, we had struggles with a few. Time pressure led us to purchase another vector, namely YEplac195, to fulfill this gap and be able to continue towards our engineering goal.
Due to the time invested in our project until this point, we decided to order each of our designed transcriptional units as a whole and not split, in order to accelerate our work.
Based on the protocols gathered for yeast transformation, we managed to successfully incorporate our YEplac195 plasmid into S. cerevisiae. Regarding the testing of our devices, although still ongoing, they are intended to provide us information on the system’s functionality:
Saccharomyces cerevisiae GM01 transformed with plasmid YEplac195, confering complementation for uracil auxotrophy; SC-ura agar plate
Our receptors harboring split-mCerulean as intracellular domains are designed to provide insights on the dimerization capacity of our scFvs, which will require fluorescence measurements to monitor the formation of ligand-dimer complexes. The same kind of assay is intended to be performed to test our part IGG6, in the context of a tricistron expressing the fluorescent proteins yEmRFP, yECFP and yeGFP. For this purpose, the Fluorescence Measurement Calibrant Kit will be used to assess efficiency of each position of the polycistronic gene. To assess the adsorption capacity of our yeast, thanks to our Human Practices team labour, we count with ELISA kits provided by ProGnosis to determine the concentration of AFB1 retired from the medium by our modified organism.
After assays, we expect to retrieve essential conclusions from our results in order to guide our following round of design, which will involve bringing together all the interacting parts with the required optimizations, or redesign a certain part because of its dysfunctionality.
In short, the course of our project has been marked by two great stages of development, in which an iterative approach was applied through the DBTL Cycle in order to assess the viability of implementation of our molecular solution to different organisms:
This iterating approach led to the construction of two analogous batches of parts, of great utility for the construction of sensing-response platforms, both in Escherichia coli and Saccharomyces cerevisiae.
Chen, Y., Yang, Y., Wang, Y., Peng, Y., Nie, J., Gao, G., & Zhi, J. (2020). Development of an Escherichia coli-based electrochemical biosensor for mycotoxin toxicity detection. Bioelectrochemistry (Amsterdam, Netherlands), 133, 107453. https://doi.org/10.1016/j.bioelechem.2019.107453
Scheller, L., Strittmatter, T., Fuchs, D., Bojar, D., & Fussenegger, M. (2018). Generalized extracellular molecule sensor platform for programming cellular behavior. Nature chemical biology, 14(7), 723–729. https://doi.org/10.1038/s41589-018-0046-z
Snider, J., Kittanakom, S., Curak, J., & Stagljar, I. (2010). Split-ubiquitin based membrane yeast two-hybrid (MYTH) system: a powerful tool for identifying protein-protein interactions. Journal of visualized experiments : JoVE, (36), 1698. https://doi.org/10.3791/1698
