Roadmap of Fisherly
Our goal is to develop a biosensor that detects bioamine levels in fish and produces colourimetric outputs to indicate whether or not the fish is safe to consume. In order to achieve that, we need to design a modulated bioamine sensor that has a three-stage processor comprising a sensing, processing, and output actuating module.
All constructs in “Build” were cloned using traditional cloning method in E. coli DH5α, please refer to the circuit experiments page for detailed protocols.
As we wanted to create a biosensor that can detect total bioamine levels, we searched for an enzyme that can convert these external signals into intracellular transcriptional targets that our chassis could recognise - this was when we came across rat diamine oxidase[1]. Under the activity of diamine oxidase, bioamines that indicate spoilage such as histamine, cadaverine, and putrescine, are oxidised to generate H2O2. Thus, we decided to proceed with an in-silico test to determine how this activity of oxidase affects the final outputs.
To obtain a preliminary understanding of how the rate of H2O2 conversion from diamine by rDAO influences the final colourimetric outputs, we built an in-silico model that reflects the enzymatic activities of rDAO, which is constructed in MATLAB using Ordinary Differential Equations (ODEs). A sensitivity analysis was also performed to confirm the significance of correlation between rDAO’s conversion rate and the biosensor’s final outputs. Refer to the Modelling page for details.
The result of our mathematical modelling can be seen below. Figure 2 is obtained using Matlab for our circuit with different values of k1, which is the reaction rate constant for the conversion of H2O2, while Figure 3 is the sensitivity analysis performed using Sobol Analysis.
By using Figure 2 and Figure 3, we found that the k1 rate constant is an important parameter in our circuit. In other words, the concentration of rDAO is significant to the output of our circuit. From Dry Lab, k1 is related to [rDAO] with the equation k1 = k[rDAO]. Thus, increasing [rDAO] increases k1, which subsequently increases RFP output of our system.
Thus, we decided to do further research with the idea of increasing rDAO expression.
Just like hDAO, rDAO contains a disulfide bond, and we realised that this is the problem. In prokaryotes, the formation of disulfide bonds in the cytoplasm is unfavourable due to the bacterial cytoplasmic reducing environment. Instead, the correct formation of disulfide bonds occurs more favourably in the periplasm, which has a more oxidising environment. The formation of disulfide bonds in the periplasm is done using systems: DsbA-DsbB and DsbC-DsbD.
The DsbA protein oxidises SH moiety of cysteine, forming the disulfide bond, while DsbB recycles the reduced DsbA back to its oxidised active form. However, DsbA prefers to form disulfide bonds in a vectorial manner, which may lead to an incorrect disulfide bond formation, and thus incorrect protein folding. This is where the DsbC-DsbD system comes in. This system fixes the incorrect disulfide bond by isomerizing the incorrect disulfide bonds.
With this, we hypothesised that the low expression of rDAO was due to the unfavourable folding of disulfide bonds in the cytoplasm.
We then performed research to see whether the introduction of a translocation tag to rDAO has been done before. We found that Rosini and her team had connected the pelB translocation tag to rDAO and expressed the rDAO in their Origami2(DE3) E. coli strain[4]. Seeing their success, we decided to tag pelB (BBa_J32015) to rDAO and transformed the composite part into DH5α.
We decided to characterise the activity of rDAO in catalysing the reaction of bioamine to produce hydrogen peroxide. Ninhydrin was used for this experiment as it reacts with bioamine to produce Ruhemann’s purple that can be measured at 570 nm. More importantly, the absorbance measured at 570 nm correlates with the concentration of bioamine. Taking advantage of this property, we designed an experiment to measure the decrease of absorbance over time, which corresponds to the decrease in diamine concentration and reflects the functionality of rDAO.
A statistically significant decrease (p < 0.01) in absorbance can be seen with PelB-rDAO while a less significant decrease (p < 0.05) can be seen with only rDAO construct. On the other hand, there is a significant increase in the absorbance for the negative control. This increase occurs due to the production of protein in our DH5α bacteria. Therefore, this proves that the rDAO enzyme in our construct can reduce the diamine concentration in our bacteria. Knowing that the function of rDAO is to catalyze the conversion of bioamine to H2O2, our result indirectly proves that H2O2 is produced by the rDAO enzyme. Moreover, the construct with pelB-rDAO caused a larger decrease in the percentage change of absorbance compared to the rDAO construct, demonstrating a higher enzymatic activity of PelB-rDAO compared to the latter.
PelB-rDAO demonstrates a significantly higher functionality when compared to rDAO. We concluded that our improvement of the existing part BBa_K3684002 (rDAO) by placing BBa_J32015 (Pel-B leader sequence) at its N-terminal to direct the protein to the periplasmic space of E. coli for the correct formation of disulfide bonds is successful.
The existing part BBa_K3684002 (rDAO) has been improved by placing BBa_J32015 (pelB leader sequence) at its N-terminal to direct the newly synthesised protein to the periplasm of E. coli for the correct formation of disulfide bonds - allowing for an increase in functional expression of rDAO.
As explained by dry lab, increasing the functional expression of rDAO increases the value of k1. In other words, based on Figure 2, given a high level of bioamines in the fish sample, a higher value of k1 will allow users to see a brighter red output on our sensor with lower detection time.
Our aim for designing the Processing and Output Actuating modules was to modulate the H2O2 input and display clear colourimetric outputs to inform the users about the quality of fish. We were inspired by E. coli’s oxidative stress sensing mechanism and decided to adopt the OxyR transcription factor and two OxyR inducible promoters, OxySp and KatGp, into our genetic circuit[2][3].
The H2O2 generated from rDAO oxidises the constitutively-expressed transcription factor oxyR. At low bioamine concentrations, this would activate the oxyS promoter and turn on the transcription of our green output; at high bioamines concentrations, it would activate the katG promoter, producing a red output.
Whole cell Experiments Constructs:
1. Comparison between Two-plasmid and whole-plasmid design
The combination of Pcon-OxyR and PoxyS-GFP into the same plasmid offers convenience for controlling the factual presence ratio of these two parts. However, as mentioned by PeroxiHub[5], this whole-plasmid design would cause potential resource competition for surrounding proteins such as RNA Pol, which in turn will lead to the decreased expression of desired genetic products. Hence, to implement OxyR in a cell-free environment, we intended to compare the performances of two-plasmid and whole-plasmid designs as below:
G1: Pcon-OxyR_pSB1C3 (26 µg/mL); PoxyS-GFP_pSB1C3 (26 µg/mL);
G2: Pcon-OxyR-PoxyS-GFP_pSB1C3 (26 µg/mL).
2. Characterization of oxyS and katG promoter
Since OxyR is a vital hub for the signal response, it’s needed in a sufficient amount in a cell-free system. To meet this condition, we came up with two solutions. Firstly, one could transform the OxySp protein into the BL21 strain, which serves as the cell lysate source and harvest the cells during the log phase. Alternatively, one can perform a pre-incubation step after CFS assembly to enable the pre-expression of OxySp. In this experiment, we tried the second method out and incorporated a one-hour preincubation procedure by incubating the plasmids below Pcon-OxyR-PoxyS-GFP_pSB1C3 (31µg/mL) / Pcon-OxyR-PkatG-RFP_pSB1C3 (31 µg/mL) with cell extracts and mastermix (with amino acids) to pre-express the OxyR proteins, which according to PeroxiHub would also enable the dissipation of hydrogen peroxide, hence reducing the background signal in the original system.
Whole-Cell experiments
Here, we aim to characterise the concentration of H2O2 required to activate OxyS and KatG promoters. Combined with a constitutively expressed oxyR gene, these peroxide-inducible promoters are designed upstream of GFP and RFP respectively. Subsequent fluorescence of these reporter proteins was measured individually to distinguish the activation threshold of both promoters. Due to the sensitivity of fluorescence measurements, the testing of OxyS and KatG promoters was done simultaneously to minimise the external influence on the results.
Cell-free Experiments
1. Comparison between Two-plasmid and whole-plasmid design
Here, we incorperated a one-hour preincubation procedure by incubating the plasmids below Group 1: Pcon-OxyR_pSB1C3 (26 µg/mL)/ Group 2: Pcon-OxyR-PoxyS-GFP_pSB1C3 (26 µg/mL) with cell extracts and mastermix (with amino acids) to pre-express the OxyR proteins and enable the dissipation of hydrogen peroxide in the original system[5]. All OxyR-involved cell-free assemblies onwards included this pre-incubation step.
After preincubation, we applied 0 mM, 0.75 mM and 1.5 mM H2O2 to respective samples with two replicates, also for the negative controls that do not contain plasmids. For more detailed cell assembly formula, please refer to the table below. Then, we incubated these reactions under 37 degrees for 7.5 Hours and subjected them to fluorescence reading (Ex. 480nm; Em: 530nm).
From Figure 18, we observed that the maximum fold change of OxySp-OxyR is higher than that of KatGp-OxyR. This data persuades us that amplification is required to produce an intense red colour that signals the user of ‘danger’.
Upon observing our Wet Lab data, we consulted Mr. Willem Landman, who is an expert in implementing corporate food safety systems in Hong Kong, about this potential problem. He explained that quality control staff in small-scale fish businesses often require a lot of training from the company in order to be trusted to perform such important tasks; as such, it should be our priority to ensure that our biosensor is foolproof and can generate an easy-to-see output that even an inexperienced staff would understand. His suggestion made amplifying the fluorescence output one of our top priorities at this time of our project.
For cell-free experiments, as indicated by Figure 14, both groups lack a significant increase as the application of H2O2 concentration. This might be due to the insufficient incubation time since the completion of cell-free reactions by time could from two hours to 16 hours as an account of cell extract differences, mastermix formula and so forth. Thus, we switched to a longer incubation time of 16 hours for our following optimization experiments, which is the longest incubation time recommended by Figure 19[6] below which concluded the typical incubation time of commercial kits very elaborately. For simplicity and quality control, whole-plasmid design was preferentially conserved in our CFS workflow.
After increasing the incubation time to 16 hours, there was an interpretable increase in relative fluorescence outputs of experimental samples which contain the Pcon-OxyR-PoxyS-GFP_pSB1C3 circuit construct. According to preliminary data (only one replicate was involved), 20 µM H2O2 had already saturated the production of GFP driven by oxySp promoter (Figure 16). This is less favorable due to hypersensitivity could presumably generate false positive results. Moreover, the threshold relative fluorescence on the plateau is still relatively low compared to the common signal-to-noise ratios (2-50) presented by other CFS biosensor research[5][7]. The same H2O2 induction experiment was also performed on Pcon-OxyR-PkatG-RFP_pSB1C3 simultaneously, while no interpretable activation curve was obtained (Figure 17).
In line with the live cell experiments' results, oxySp shows high sensitivity towards oxidative stress change. Thus, instead of choosing oxySp as an early-spoilage indicator, a more comprehensive design in our genetic circuit is required to provide more accurate outputs triggered by distinct bioamine concentrations.
As shown in Figure 18, our wet lab results indicate that both GFP and RFP will be produced at higher concentrations of H2O2. This means that when the fish is spoiled, instead of a red output, a yellow colour will display, which is not what we desire.
Our Dry Lab modelling using Matlab and Python shows very similar results; red outputs are not observed at high concentrations of bioamines since the green outputs are also being continuously produced, ultimately resulting in a yellow output (Figure 20).
In our discussion with Mr. Martin Dijk from Seafood Friday, a local business that directly ships top-quality seafood from Holland to its distributing centre in Hong Kong, we learned that one of the biggest problems in quality control for seafood products is the ambiguity in determining its quality. Since most of the inspection for fish businesses in Hong Kong is performed manually, it is of utmost importance to make sure our biosensor generates easy-to-understand signals. Thus, the previously mentioned colour-mixing issue is another potential problem that we cannot simply neglect; we must find a way to develop a system where the green output would diminish as the red output increases when high bioamine levels are detected.
Having a 2 colourimetric output system has posed problems in regard to colour mixing. As a result, we decided to degrade the green output when the red output is expressed. Constitutively expressed modified GFP is stable as there is an inhibitory sequence (77 amino acids from mRFP) at its C terminal, shielding the LVA tag; this way, the endogenous E. coli proteases ClpXP and ClpAP cannot recognise and degrade the GFP.
At low bioamine concentrations
At high bioamine concentrations
The aim of this experiment is to characterise the interaction between the expression TEV protease and the subsequent cleavage of the inhibitory sequence. Cultures were grown at different concentrations of IPTG for 5 h. The fluorescence tracked by a plate reader.
As shown in figure 25, at the last two concentration points, a decrease in fluorescence of the pTac - TEV - Pc - GFP - LVA - CS - IS construct can be observed. Which suggests that the expression of TEV can remove the shielding effect of the inhibitory sequence, thereby inducing the degradation of GFP. Moreover, the construct pTac - TEV - Pc - GFP - LVA - CS remains at a low fluorescence level, indicating that the lack of an inhibitory sequence causes the construct to be continuously degraded. These 2 observations indicate that our constructs function well.
At low concentrations of IPTG, the fluorescence level of the test construct (pTac - TEV - Pc - GFP - LVA - CS - IS) is lower than that of the positive control (pTac - TEV - Pc - GFP).
This might be due to several reasons:
1. The inhibitory sequence did not fully shield GFP from degradation
The shielding effect of the IS is dependent on the length of the amino acid sequence and its configuration. To increase shielding effect, more amino acids can be added to the IS or a completely different IS from other proteins can be used. A comparative assay with a similar setup can be conducted to test the functionality of different IS.
2. The addition of an inhibitory sequence interfered with the folding of GFP
As the IS is sourced from 77 amino acids from the C terminal of mRFP[9] and ligated directly to the back of the pTac - TEV - Pc - GFP - LVA - CS, the secondary and tertiary protein folding of the IS may cause steric interference/ collision to GFP, decreasing its fluorescent output. Therefore, we suggest that a longer spacer sequence can be inserted between the LVA tag and IS to aid in proper protein folding.
The degradation rescue system is an effective tool for amplifying the dynamic range of output. As TEV proteases have catalytic activity, a single molecule of protease can act upon many molecules of the substrate as opposed to a transcription factor that can bind to only one site at a time.
RFP is tagged with a potyvirus SSRA C-terminal degradation tag LVA, which can be recognised and degraded by proteases ClpXP and ClpAP, which are endogenous in E. coli cytoplasm[8].
At low bioamine concentrations
At high bioamine concentrations
Our goal here is to characterise the interaction between the expression TEV protease and the subsequent cleavage of the LVA degradation tag. The cultures were grown at different concentrations of IPTG for 5 hours and the fluorescence was tracked by a plate reader.
After introducing the amplification and degradation rescue systems, we can clearly see an improvement in the colourmetric outputs of our biosensor through modelling. As seen in the graphs below, different levels of bioamines would lead to different gradients from green to red.
When exposed to a low bioamines level (200 ppm), our sensor will generate either a green or yellow output, indicating that the fish is not spoiled and is safe to consume. When the bioamine level increases to around 600 ppm, the sensor would display a orange color instead. Finally, if the sensor is exposed to recommended upper limit for bioamine concentration (1000 ppm), it would generate a red output, indicating the user that the fish is no longer safe to consume and should be discarded.
[1] Elena Rosini, Serena Nossa, Mattia Valentino, Paola D’Arrigo, Stéphane Marinesco, Loredano Pollegioni, Expression of rat diamine oxidase in Escherichia coli, Journal of Molecular Catalysis B: Enzymatic, Volume 82, 2012, Pages 115-120.
[2] Rubens, J., Selvaggio, G., & Lu, T. (2016). Synthetic mixed-signal computation in living cells. Nature Communications, 7(1). https://doi.org/10.1038/ncomms11658.
[3] Åslund, F., Zheng, M., Beckwith, J. & Storz, G. Regulation of the OxyR transcription factor by hydrogen peroxide and the cellular thiol—disulfide status. Proc. Natl Acad. Sci. USA 96, 6161–6165 (1999).
[4] Fernandez-Rodriguez, J., & Voigt, C. A. (2016). Post-translational control of genetic circuits using Potyvirus proteases. Nucleic acids research, 44(13), 6493–6502. https://doi.org/10.1093/nar/gkw537
[5]Soudier, Paul et al. “PeroxiHUB: A Modular Cell-Free Biosensing Platform Using H2O2 as Signal Integrator.” ACS Synthetic Biology 8 (2022): 2578–2588. Crossref. Web. 12 Oct. 2022.
[6] Garenne, D., Haines, M. C., Romantseva, E. F., Freemont, P., Strychalski, E. A., & Noireaux, V. (2021, July 15). Cell-free gene expression. Nature News. Retrieved October 8, 2022, from https://www.nature.com/articles/s43586-021-00046-x#Sec41
[7] Pardee, K., Green, A. A., Ferrante, T., Cameron, D. E., DaleyKeyser, A., Yin, P., & Collins, J. J. (2014, November 6). Paper-based synthetic Gene Networks. Cell. Retrieved October 8, 2022, from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4243060/#SD1
[8] Fernandez-Rodriguez, J., & Voigt, C. A. (2016). Post-translational control of genetic circuits using Potyvirus proteases. Nucleic acids research, 44(13), 6493–6502. https://doi.org/10.1093/nar/gkw537
[9] Jungbluth, M., Renicke, C. & Taxis, C. Targeted protein depletion in Saccharomyces cerevisiae by activation of a bidirectional degron. BMC Syst Biol 4, 176 (2010). https://doi.org/10.1186/1752-0509-4-176