Team:UPF Barcelona/Engineering
In this page we will introduce the Engineering Design Cycle that we followed in order to create our Intein Mediated T3 (IMT3) Biosensor. We made our first version IMT3 sensor, then we made another design round that led us to the IMT3 sensor that we present with Hormonic. However, a third engineering process to enhance it to a better version was already being planned. Here you can see the journey we went through to generate the BBa_K3484000.
Engineering success
The enginnering cycle
Synthetic biology is a science discipline that differs a lot from the others. Usually, you would use the scientific method to obtain results in the majority of life sciences, centering your work on making observations and performing experiments. However, synthetic biology goes beyond that, it is all about constructing something new from parts (biological parts if you may). That is why, sometimes the engineering design process fits better with these projects, since a product has to be built, a machine that performs a task, where it is more important what it does rather than how, yet still has to be tested. Funnily enough, that's why iGEM stands for genetically engineered machines.
T3 biosensor development
Research: How can we sense T3?
We went through the literature trying to find any previous T3 sensor in bacteria. We found the sensor used by researchers in the University of Princeton, where they used the ΔI-SMTR mini intein coupled to the TS enzyme creating an OD sensor [1]. To see more of our biosensor background go to our T3 biosensor design page.
Imagine: How can our bacteria communicate with a computer?
Since the main characteristic of inteins is their ability to reconstruct virtually any split protein, we thought that, if split correctly, we could use a Fluorescent Protein. Using light as a reporter we can quantitatively and continuously know the amount of hormone. That's why we thought of splitting eGFP, since we saw in papers that the ΔI-SMTR mini intein without the THR-β could put together the two parts of the eGFP [2].
Design: How can our bacteria express our protein the best way?
We want to couple our ΔI-SMTR mini intein, which consist of the THR-β flanked by the two intein splicing domains. To this we added the eGFP split between residues 70 and 71 to keep the necessary cysteine after the second splicing domain. To see more in detail the construction of this sensing protein, we recommend you visit our T3 biosensor design page.
We had our idea for chimeric protein, however, in the same study where they showed the reconstruction of the eGFP, they also mentioned the propensity of forming inclusion bodies. This normally happens when bacteria are transformed to express at a high rate big proteins, the machinery of protein folding probably falls short. So in our design we had to include a solubility tag, and we found 4 possible candidates [3], based on size (Tab.1).
| Solubility Tag | Length (aa) |
| FH8 | 69 |
| SUMO | 100 |
| GB1 | 56 |
| MBP | 396 |
We wanted a small tag since we did not want to enlarge our protein, so even though Maltose Binding Protein (MBP) showed good solubility, we considered it too big. From the other 3 options, GB1 was the smallest, however, neither GB1 nor SUMO solubility ability was good compared to Fh8. The MBP and Fh8 are ranked among the best solubility enhancer tags [4]. So we placed the latter in the N-terminal of the first half of eGFP.
Another tag that we wanted to add was the Flagx3 tag in the C-Terminal of the second half of eGFP, for immunodetection purposes. Although any tag would have worked for these detection, experts working within the Transational Synthetic Biology Lab, suggested us to work with this one due to its great performance in past experiments.
For the non-coding sequences, we tried to use the most standard parts already used in iGEM. We used the BBa_K880005 which contains a strong constitutive promoter (BBa_J23100) and a strong RBS (BBa_B0034) as well, perfect for recombinant protein expression. For terminators, we used the common double terminator BBa_B0014. These parts were recommended to us by Dr. Carlos Toscano, who was an advisor for our University team in 2018 and used these parts.
Build: How can we build it?
After the design phase of our Biosensor and taking into account the difficult access to the lab for confinement, we made the decision of synthesizing it directly as a clonal plasmids. We ordered to IDT the IMT3_eGFP_pUC-Kan sequence with a high copy number vector.
Test: Do we get the results we expected?
We transformed our plasmid to E. Coli BL21 cells and performed tests to check how the sensor was working. We performed a Western blot to check the expression of our protein, and we saw that splicing was occurring when induced with T3 and that the solubility tag avoided major protein aggregates. We then tested it quantitatively by placing our sensing cells with different T3 concentrations ranging from 10uM to 80 uM, and unfortunately we obtained some random behaviour. To see the final results visit our Results page.
Learn: What we did not know from our design?
From the results of the initial plate readers, we saw that a lot of times higher values of T3 in the media did not correspond to higher fluorescence. Moreover, different plate reader results showed different behaviours between the ranges 20 uM to 80 uM. This made us think that maybe we were exploring the wrong range of concentrations of our sensor to report correctly. If the sensor needed more concentration to report correctly, it meant that it needed concentrations higher than 80uM. However, we do not think this was the case since T3 is insoluble at concentrations higher than 80uM. So, having in mind the low levels of T3 the Thyroid hormone receptors normally sense in the body (around 1 pM), the sensor was probably saturated, raising that random behaviour. We had to explore concentrations lower than 20uM.
Improve: Is our reporter the best one?
We also realised that maybe eGFP wasn't the best GFP we could use. Fluorescent levels were very low, and having in mind that we had to detect them from our own fluorescent sensor, we needed more signal. Looking for other reporters, we found a version of the GFP called superfolder GFP (sfGFP). This variation had two main characteristics: it was faster and brighter.
Design: How can we introduce the new reporter?
Once we decided to use sfGFP as our new reporter, we could also vary the split point. We looked up in the literature splitting points and created two versions. The first one was keeping the same split region as the eGFP because the residue sequence is very similar. The 70/71 split point from the eGFP corresponded to the residues 69/70 for the sfGFP, in which residue 70, just after the C-terminal of the intein, had to be changed to a Cysteine.
The second version came from the 2014 Heidelberg iGEM team, where they used trans splicing inteins (two different proteins) to join two halves of sfGFP. The split point was between the residues 64 and 65, and once again, residue 65 had to be changed to a cysteine.
Build: How can we build it?
Once again, we decided to synthesize it again the same way as the IMT3_eGFP_pUC-Kan, same promoter, rbs, tags and terminator. Hence we ordered the two second versions of our sensor: IMT3_sfGFP70/71_pUC-Amp and IMT3_sfGFP64/65_pUC-Amp, changing the antibiotic resistance to avoid any possible cross contamination, since we would work with both strains at the same time a lot.
However, IDT warned us that the sequence for IMT3_sfGFP64/65_pUC-Amp “included a partial or full coding region from Brucella melitensis” and could not be sent to us from the United States to Spain. So at the end, the IMT3_sfGFP70/71_pUC-Amp became the only IMT3_sfGFP_pUC-Amp sequence.
Test: Does it perform better?
This time, knowing that we were probably saturating our sensor, we chose to test our construct with T3 concentrations ranging from 1pM to 100uM. We tested both constructs: IMT3_eGFP_pUC-Kan and IMT3_sfGFP_pUC-Amp and results showed a better performance of the sfGFP version, with less randomness, brighter and more precise. To see the final sfGFP sensor results visit our results page.
Learn: Why do these improvements work?
As we expected, since sfGFP has more brightness, the fluorescence values related to each T3 concentration escalated, so the difference between values also escalated. Also, thinking that the IMT3 sensor was saturating at values between 20uM and 80uM, increasing the testing range to a logarithmic scale from 1pM to 100uM (10-12 to 10-4 M), allowed us to appreciate the sensing capacities of our sensor and allowed us to fit it better to the model.
Improve: In what other ways can we modify our sensor?
So far we have achieved a sensor able to detect T3 concentrations down to 10-7 M and that has a dynamic range that goes as far as 10-4 M. However, the IMT3 biosensor can be further improved. Decreasing even more the detection limit is something that should be done if we would like to directly detect T3 concentrations similar to the ones found in the blood and interstitial liquid (around 1 pM). Another improvement would be to widen the dynamic range, making the sensor more sensible, or even changing the reporter for other applications.
Design: How can we make new versions?
Low Copy Vectors
One approach that can be done to reduce the random noise from the sensor, thus making it more precise, is to place our IMT3 sensor in a low copy vector.
At the moment IMT3_sfGFP_pUC-Amp is in a high copy pUC vector, meaning that for each cell there are between 500 and 700 plasmids per cell. These 200 plasmid variations can cause noise. We think that cloning our IMT3_sfGFP sensor to a low copy number vector, such as the pSC101 [5], with a compy number of ~5 could reduce the noise, thus decreasing the lower detection limit. However, we might lose some intensity of the signal. We did this assembly via Golden Gate assembly using BsaI Type IIS restriction enzyme. but we couldn't characterise its performance due to time limitations. See Golden Gate Assembly in the Protocol page.
Double LBD
Talking with our instructors, they told us that, since our sensor had a ligand and receptor dynamic, increasing the number of receptors could modify the sensibility or the ability of the sensor to detect lower T3 concentrations [6]. In our case, the receptors are the Ligand Binding Domains of the THR-β. We thought that by placing one next to the other (Fig.1) in between the Intein Splicing domains we could tune the minimum detectable concentration.
To implement that we thought of doing it via Golden Gate assembly. We could use PCR amplification to get the LBD of the THR-β and insert it next to the other one, however, sine this fragment is around 780 kb, having such long equal sequences can easily trigger homologous recombination mechanisms, making it very difficult to have this sequence repeated. That is why we ordered to Twist the LBD THR-β codon optimized to differ from the sequence we had in the sensor yet produce the same amino acids. This fragment will be placed next to the other receptor without creating homologous recombination.
Directed evolution
We realised that based on literature our modification options were limited. Not much has been published about mutant variants of the THR-β besides the ones related with diseased, which often are inhibitory mutations. We know which residues directly interact with T3 when binding at a molecular level [7], but we don't know what changes to do in order to improve its performance. Actually we don't have to know what changes make to the protein, natural selection (in our case our selection) can lead to better sensor variants.
Here we are stepping in the field of Directed Evolution, a fairly recent academic field yet has been present since the origins of humankind. It is based on creating diversity and applying a proper selective pressure to navigate towards a better sequence. This has been already done in the past selecting the trees that produced more fruits or the cows that produced more milk. But now, with better technologies, we can select, for example, brighter bacteria. Using the Mutagenesis plasmid 6 (MP6) [8] which randomly inserts mutations, we designed an experimental pipeline to be able to generate variability in our IMT3_sfGFP biosensor, select them by fluorescence via cell sorting and repeat it to make different evolution cycles (Fig.2).
References
[1] Gierach, I., Li, J., Wu, W. Y., Grover, G. J., & Wood, D. W. (2012). Bacterial biosensors for screening isoform-selective ligands for human thyroid receptors α-1 and β-1. FEBS open bio, 2, 247–253. https://doi.org/10.1016/j.fob.2012.08.002
[2] Gangopadhyay, J. P., Jiang, S., & Paulus, H. (2003). An in Vitro Screening System for Protein Splicing Inhibitors Based on Green Fluorescent Protein as an Indicator. Analytical Chemistry, 75(10), 2456–2462. DOI:10.1021/ac020756b
[3] Costa, S., Almeida, A., Castro, A., Domingues, L. (2014). Fusion tags for protein solubility, purification, and immunogenicity in Escherichia coli: the novel Fh8 system. Front. Microbiol., 5, 63, https://doi.org/10.3389/fmicb.2014.00063
[4] Costa, S. J., Almeida, A., Castro, A., Domingues, L., Besir, H. (2013). The novel Fh8 and H fusion partners for soluble protein expression in Escherichia coli: a comparison with the traditional gene fusion technology. Appl Microbiol Biotechnol, 97, 6779–6791. DOI 10.1007/s00253-012-4559-1
[5] Plasmids 101: Origin of Replication. Retrieved July 25, 2020, from [https://blog.addgene.org/plasmid-101-origin-of-replication]
[6] Gonzalez-Flo, E., Alaball, M. E., Macia, J. (2020). Two-Component Biosensors: Unveiling the Mechanisms of Predictable Tunability. ACS Synthetic Biology, 9(6), 1328-1335. DOI: 10.1021/acssynbio.0c00010
[7] UniProtKB - P10828 (THB_HUMAN). Retrieved October 22, 2020, from [https://www.uniprot.org/uniprot/P10828]
[8] Badran, A., Liu, D. (2015). Development of potent in vivo mutagenesis plasmids with broad mutational spectra. Nat Commun, 6, 8425. https://doi.org/10.1038/ncomms9425
FOLLOW US
