Throughout the project, we performed multiple engineering cycles on both the tech and lab front, with obstacles constantly needing to be addressed. How does the fluorescence of our reporters look visually? Could it be detected by our software? What would protein production look like in our host cells? Does our current FRET/quenching approach work in vitro? These were just some of the problems we came across and wanted to overcome. In doing so, this led to different iterations of our Fungalescence synthetic construct.


Choosing Fluorescent Protein Pairs

The first challenge we came across was what fluorescent proteins would be viable for use in our Fungalescence sensor. To facilitate this decision, we spent time determining the best expression platform for distinguishing between fluorescent proteins. We also worked on deciding the optimal fluorescent proteins for interaction with one another via FRET interaction. For ease and efficiency, we decided to transform various fluorescent proteins containing plasmids into E. coli DH5a to create more copies of our DNA.These plasmids were also transformed into E. coli Rosetta-Gami (DE3) and induced on arabinose-containing plates to promote protein expression while under the control of pBAD. The colonies formed on plates would be used for an initial analysis as we expected to observe coloured colonies. Ultimately, we secured eight pBAD plasmids containing eight different fluorescent proteins, which we simply transformed into DH5a. After transformation, we found that mScarlet, mNeonGreen, sfGFP, and mClover3 were suitable for our construct due to their visual colouring from the protein expression levels and transformation efficiency. We examined emission and absorption spectra to further narrow down potential candidates. Ultimately, we chose mScarlet and mNeonGreen as they allowed for a suitable excitation with green light and the ability for FRET between the pair. However, this engineering cycle left us with some new questions to consider. Protein expression using arabinose induction was not at sufficient levels for us to feel confident to incorporate it into our design. If we were able to obtain higher protein expression, it would allow us to see a better signal that could be detected by our drone software.

Characterization of Fluorescent Proteins

With mScarlet and mNeonGreen as the chosen candidates, we wanted to create a new expression platform to achieve higher protein production and allow us to better characterize our fluorescent proteins. Fortunately, our secondary primary investigator was able to donate a pET-22b(+) plasmid under the control of a T7 promoter. This allows for IPTG induction, which is much stronger than arabinose induction. Therefore, we decided to put our mScarlet and mNeonGreen coding sequences under the expression of the T7 promoter to hopefully see more protein expression. We PCR amplified both genes from their pBAD vectors and ligated them into the pET backbone using Gibson Assembly. After transformation into E. coli Rosetta GAMI and induced on plates containing IPTG, we did not see a significant difference in terms of saturation of colour in comparison to the pBAD constructs. Nevertheless, we were also able to successfully excite both red and green colonies using green light and were able to detect their fluorescence. We originally planned to have our construct fluoresce green in the presence of Fusarium graminearum and fluoresce red in the absence of our fungal pathogen due to FRET. However, after characterizing the fluorescence, we were unsatisfied with the nature of our reporter signals. After discussions with our engineering team, they explained that the green fluorescence was hard to detect and the colour did not contrast well with the shades of browns and greens that are observed within crop fields. Fortunately, mScarlet’s red fluorescence provided an ideal reporter signal as the red colour has significant contrast with the colours of the field environment environment.

After the results we observed in the lab, we chose to use mScarlet as the reporter signal. FRET interactions rely on the transmission of energy between donor and acceptor molecules. In the case of fluorescent proteins, donors emit light energy which excites neighboring acceptors, causing subsequent fluorescent emissions. mScarlet fluoresces between ~550-650 nm (orange-red), severely limits our options for FRET acceptors since very few fluorescent-type proteins can be excited by this range of light. After reviewing the literature and various databases, we were able to identify a chromoprotein, ShadowR, that can quench mScarlet through a FRET interaction that would serve our purposes well. This is what inspired us on our next engineering cycle in the project involving the assembly, transformation, and expression of our Cameleon-Q-Sensor in E. coli Rosetta-Gami B(DE3) and Shuffle-T7.

Figure 1. A. Transformation plate of pET-22b(+) with Cameleon-Q-Sensor construct in Rosetta-Gami. Plated on LB + 100mg/mL Ampicillin agar. We speculate that the purple colour of the colonies is due to ShadowR quenching mScarlet. B, C, & D. Streak plating from transformation of pJUMP with URS+ShadowR construct in SHuffle-T7. Temperature growth characterization from left to right: 25°C, 30°C, and 37°C.

Creation and Purification of Antigen Construct

Simultaneously, as we were characterizing our fluorescent proteins, we got to work trying to express a wild-type antigen of F. graminearum to act as a potential target to determine a nanobody for binding. For this experiment, we thought to express our protein in E. coli Rosetta-Gami (DE3) to allow for its purification with an associated HisTag. We utilized our pET plasmid with the T7 promoter as we could induce protein expression with IPTG and had access to an empty version of the vector. To reduce the complexity of our antigen, Endo-1,4-beta xylanase enzyme (UniProt accession I1RII8), we searched for the exposed alpha-helical motifs on the enzyme and put it into our helix backbone that had N-terminal and C-terminal helix initiators to control the folding of the construct. As well, this contained a His-tag to make purification easier. However, when ordering the construct from IDT, there was a miscommunication on the plasmid the insert would arrive in, resulting in the loss of a terminator for proper transcription due to a different backbone being present. Therefore we could not directly transform our vector into our competent cells. The restriction sites we were originally planning to use for cloning were removed, so we could not remedy the problem with a simple restriction digest. To solve this, we developed a PCR strategy where we linearized our pET backbone and generated overhangs compatible with our PCR’d insert and combined them together with Gibson cloning.

To test protein expression and our purification methods, we first set out to test two different purification methods. The resulting samples would be subjected to a Bradford assay and run on a polyacrylamide gel to determine if protein expression worked. The first method we used was buffer-based lysis, which allows for good quantities if the protein is hydrophobic. The second was using the French Press, which works well provided the protein is hydrophilic. After performing the lysis and a Bradford assay, we found that the French press was a much more effective method compared to the buffer lysis. As such, we decided to use the French press in any further protein purifications. However, the numbers we obtained from 100 mL of culture were still not satisfactory, thus we increased our amount of culture to 1L for future applications. While visualizing the gel, we initially used a TGX stain-free method, however, we were unable to visualize any bands. We then transitioned to a more traditional Coomassie blue method. Unfortunately, the band that was developed as a result was undefined and faint. The subsequent decision was made to utilize silver staining as it is more sensitive. Using this method more significantly defined bands were visible. As such, we were able to create parameters for future protein purifications that served us well in subsequent experimentation.

Figure 2. DS-PAGE of purified antigen construct using buffer lysis and French press, visualized using TGX-Stain Free.
Figure 3. SDS-PAGE of purified antigen construct using buffer lysis and French press, visualized using Coomassie blue staining.
Figure 4. SDS-PAGE of purified antigen construct using French press, visualized using silver staining.

Following BioBrick Standards

As we proceeded through characterization of our fluorescent proteins and the design of our sensor construct, our generated parts did not adhere to iGEM Biobrick standards. This is because we used cloning strategies that were immediately convenient for the progress of our project and did not require us to order any specific parts at the time. As such, we decided to make these parts compatible for submission to the parts registry and under the expression of a different promoter. After using the T7 promoter contained within pET-22b(+) for the majority of our engineering process, we found that while the promoter yielded good results, optimizing the amount of IPTG to add to induce higher protein expression without becoming toxic to the cell was difficult, as we experienced when we purified our antigen from large volumes of culture. Furthermore, inducible promoters, such as the T7 promoter, are advantageous when the goal is to strategically regulate protein expression, whereas constitutive promoters are useful when the goal is to produce proteins continuously without requiring an inducer. With this in mind, we decided to redesign our expression system and put our parts under the expression of a constitutive promoter to express more proteins than with the IPTG inducible T7 promoter. This was also more cost effective as we no longer had to purchase/use IPTG to produce protein. We developed a PCR strategy that amplified the biobricks parts that we ordered, while adding the required prefixes and suffixes to adhere to iGEM submission standards.

It was decided that the pJUMP 28-1A plasmid, provided in this year's distribution kit, would act as the entry vector for all of our basic and composite parts. A few features of this plasmid proved quite useful, namely, the sfGFP insert and strong flanking terminators. The first allowed us to perform “green-white” screening of successful ligation reactions. Due to pJUMPs terminators, it was no longer necessary to incorporate our own terminator parts, which sped up the assembly process.

Using restriction cloning, we digested our biobricks parts and the pJUMP entry vector and utilized T4 DNA Ligase to insert the biobricks parts into the backbone. We were able to combine and characterize mScarlet and ShadowR under the control of the J23101-B0034 (BBa_K4755001) part as well as we were able to combine the Cameleon linker and ShadowR. First, we attempted transforming the constructs into E. coli Rosetta-Gami B(DE3) which yielded no colonies. We postulate this was due to the shared Kanamycin resistance of Rosetta-Gami and our plasmid. Next we attempted to transform in E. coli SHuffle-T7 which yielded excellent results, resulting in bright colonies in terms of saturation of colour in comparison to our previous constructs. Another distinction between the previous pET-22b(+) constructs and the new pJUMP biobricks constructs was the observation of distinct colour in liquid cultures, indicative of increased protein production. Unfortunately, the competition was coming to an end as these cloning were being completed and we were unable to combine the URS-mScarlet and Chameleon-ShadowR biobricks constructs to create the entire fungalescence sensory construct, but this would be the natural direction for the completion of this stage of the engineering cycle.

Figre 5. Initial transformation of DH5ɑ with the pJUMP+URS+mScarlet ligation reaction.Three white colonies were identified. Plated on LB+ 50mg/mL Kanamycin agar.
Figre 6. Re-streaked red colony from transformation of E. coli SHuffle-T7 with Colony 3 extracted plasmid. Red colour of colonies indicative of mScarlet production and success of cloning. Plated on LB+ 50mg/mL Kanamycin agar.
Figre 7. 50mL LB liquid + 50 µL 50mg/mL Kanamycin media inoculated from re-streaked colony 3 SHuffle-T7 transformed with pJUMP+URS+mScarlet construct (Figure 6.). Red colour of liquid indicative of mScarlet production and success of cloning.
Figre 8. Initial transformation of DH5ɑ with the pJUMP+URS+ShadowR ligation reaction. Four white colonies were identified. Plated on LB+ 50mg/mL Kanamycin agar.
Figre 9. Re-streaked white colony from transformation of E. coli SHuffle-T7 with Colony 2 extracted plasmid. Gray/purple colour of colonies indicative of ShadowR production and success of cloning. Plated on LB + 50mg/mL Kanamycin agar.
Figre 10. 5mL LB liquid + 5µL 50mg/mL Kanamycin media inoculated with re-streaked colony 2 SHuffle-T7 transformed with pJUMP+URS+ShadowR construct (Figure 9.). Tubes were placed on a UV transilluminator to illuminate culture. Gray colour of liquid indicative of ShadowR production and success of cloning.
Figre 11. Initial transformation of DH5ɑ with the pJUMP+Cameleon+ShadowR ligation reaction. Ten purple colonies were chosen for colony PCR to verify inserts. Plated on LB+ 50mg/mL Kanamycin agar.



On the tech team side, most of our work had involved trial and error. When it came to designing for 3D printing, we made edits and tested them using our university resources. This approach worked well, as we encountered no issues with the trial-and-error method. In fact, the trial-and-error process proved effective because it provided us with direct results and a better understanding of the areas we needed to improve. This approach paralleled our work in AI development as well.

Upon receiving data (sample images) from the lab team, we embarked on optimizing the image masking techniques to meet our project's requirements. This optimization occurred several times throughout the project: once for the color we needed to detect, once for improvements based on that color, once for determining the appropriate picture-taking distance, and one final optimization for the ultimate quality checks.

Another area where we experimented with new techniques was during our camera selection phase. We initially attempted to work with a multispectral camera; however, we encountered difficulties connecting it to our Raspberry Pi, which made us appreciate the simplicity we would have had if we had opted to use a regular RGB GoPro-like camera.