Team:UPF Barcelona/Results


As a proof of concept for our device, which regulates a correct dosing depending on a measured hormonal level, we designed a biological circuit capable of producing and sensing lactone in a controlled way. This circuit is formed by two kinds of cells, ones responsible for producing lactone, and ones responsible for sensing the presence of lactone and producing super-folder GFP (sfGFP) in response. With this circuit, we can simulate and understand how the thyroid axis works and how our device would work.



The design of the biosensor was made only with the bibliographic information available. Not being able to access the laboratory made us fine tune the “in silico” sequence design as much as possible to cover all of our bases and expect as little experimental surprises as possible. This would allow us to minimize the troubleshooting time as much as possible to seize the restricted laboratory access time. This is the reason why we already added to the initial construct the Fh8 solubility tag, the Flag tag, a strong pair of Promoter-RBS and the BioBrick assembly prefix/suffix, among others. However, since our biosensor had never been done before in the literature, we surely had to test our silico design in vitro. All the protocols we used are available as downloadable files at the Experiments page.

Sequence Checking

The two final constructs were synthesized and provided by IDT. Despite how much we trust iGEM’s sponsors, we had to check the sequences of both Intein Mediated T3 (IMT3) biosensors. To do this, we separately transformed E. Coli DH5α with our two plasmids: IMT3_eGFP_pUC-Kan (enhanced GFP as a reporter version) and IMT3_sfGFP_pUC-Amp (super folder GFP as a reporter version).

Colony PCR followed by sequencing, both with the commonly used VF2 and VR primers, confirmed the sequence. The sequence-confirmed plasmids were extracted and transformed into E. Coli BL21 (DE3) via electroporation, and Colony PCR was performed once again.

Fig 1. Colony PCR of IMT3_eGFP_pUC-Kan confirmed a weight around 3,5 Kb. First well was a positive control, the next 7 were different colonies.
Fig 2. Colony PCR of IMT3_sfGFP_pUC-Amp confirmed a weight around 3,4 Kb. Wells 1, 2 and 3 correspond to colonies 1, 2 and 3 respectively.

Biosensor Expression

To evaluate whether our IMT3 sensing protein was being expressed as well as its splicing abilities in a qualitative way, we performed Western blots to detect our target proteins.

Fig 3. A) Western blot from BL21 with IMT3_eGFP_pUC-Kan where the whole construct (85 kDa) and after splicing joint eGFP (37kDa) can be seen. B.1) Positive control: mammalian lysiate from a colleague with the same FLAGx3 tag. B.2) Negative control, BL21 without plasmid incubated overnight with 100uM of T3. C) Western blots from the whole construct. C.1 is a positive control. C.2 and C.3 samples were treated with the boiling method. C.4 and C.5 were treated with Urea incubation before loading the gel.

The Wester Blot showed that the full protein was being produced (Fig.3A), corresponding to the 85kDa signal. Splicing activity can be seen from the presence of the 37kDa protein, corresponding to the two halves of the GFP and both tags, increasingly shown in the samples of induced cells with T3, suggesting that the splicing is indeed coupled with presence with T3.

We could also observe that there was no significant difference between the samples treated with Urea and without, suggesting that the Fh8 solubility tag was avoiding major aggregate formation (Fig.3C).

Biosensor Performance

In order to characterize the functionality of both Intein Mediated T3 (IMT3) biosensors, in parallel to its expression study, Long plate reader Experiments were set up to sense the amount of fluorescence produced by our biosensor with different T3 concentrations.

BL21 (DE3) containing IMT3_eGFP_pUC-Kan, IMT3_sfGFP_pUC-Amp and no plasmid were grown overnight in LB. The next day one part cells were inoculated in 30 fresh LB media and grown for 3 to 4 hours. They were induced with different concentrations of T3 ranging from 1pM to 100uM plus controls (Tab.2). Fluorescence intensity and OD600 were analysed every 30 minutes over 24 hours with a plate reader (Tab.1).

Table 2. Combinatios of T3 concentrations and sensor types. IMT3_eGFP_pUC-Kan (Orange), IMT3_sfGFP_pUC-Amp (green), Bl21 no plasmid (yellow) and Blank, LB+Amp (white).

Sensor fluorescence was obtained by subtracting the blank LB fluorescence and normalizing to the OD (Fig.4-5). All the values were obtained from the mean of the 3 wells with the same conditions. Then it was analysed at steady state (time to infinity) to generate the transfer functions (Fig.6-7).

Equation 1. Normalization equations used to generate the results.
Fig 4. Temporal behaviour of fluorescence under different T3 concentration with IMT3_eGFP_pUC-Kan over 24 hours. The values are the mean of the 3 same concentration wells.
Fig 5. Temporal behaviour of fluorescence under different T3 concentration with IMT3_sfGFP_pUC-Amp over 24 hours. The values are the mean of the 3 same concentration wells.
Fig 6. Transfer function of IMT3_eGFP_pUC-Kan correlating the steady fluorescence with the amount of T3. Experimental data (red dots) and fitted model (blue line).
Fig 7. Transfer function of IMT3_sfGFP_pUC-Amp correlating the steady fluorescence with the amount of T3. Experimental data (red dots) and fitted model (blue line.

IMT3_sfGFP_pUC-Amp fluorescence over the time showed an increase proportional to the concentration of T3 in the media (Fig.5). IMT3_sfGFP_pUC-Amp also had higher fluorescent values (between 700 and 1800 RFU) (fig.5) compared with IMT3_eGFP_pUC-Kan (between 400 and 900 RFU) (fig.4).

Overall, the IMT3_eGFP_pUC-Kan experimental results were not concluding enough to see its sensing range.

However, with the IMT3_sfGFP_pUC-Amp, we have successfully developed a biosensor able to detect changes in T3 concentration and report them with a fluorescent signal proportional to the ligand concentration. We could observe that the sensor response had a dynamic range of: 10-7 to 10-4 M of T3, as can be seen above (Fig.7).


We have already seen the characteristics of the lactone circuit and its analogy to the thyroid axis in the Proof of Concept page . Here you can find the results obtained from the different tests and characterizations that were performed using this circuit.

Characterization of the lactone biosensor

To characterize the functioning of the lactone biosensor cell, different concentrations of lactone going from 100 pM to 100 µM were used. The response to these concentrations over time was monitored using a Plate-Reader analysis on GFP fluorescence emission (Fig.8). From this data we could fit the resulting sfGFP concentration to our lactone biosensor model, showing a great correlation between the experimental and the theoretical results.

Figure 8. GFP fluorescence emission with different AHL concentrations with respect to time.

In the graph below (Fig.9) we can see these fitting results. The graph represents fluorescence, in Relative Fluorescence Units (RFU) normalized with respect to the optical density, relative to the different lactone concentrations. The shape of this function follows a Hill function, showing a dynamic range that goes from 10-8 to 10-5 M. This means that our biosensor would follow a sinusoidal dynamics.

Figure 9. Final GFP fluorescence emission with respect to different AHL concentrations.

Characterization of the lactone system: Combining the producer and the sensor cells

As for the genetic circuit, we characterized its behaviour under different concentrations of both arabinose and glucose. In order to model these dynamics two transfer functions can be derived: one representing the producer cell and the other representing the sensor cell. These models can be found in our Modeling page. The experimental results (Fig.10) showed a great similarity with the theoretical results previously obtained.

Figure 10. Experimental and modelling results of the genetic circuit.

As expected, the presence of sfGFP is dramatically reduced as the concentration of the inhibitor (glucose) increases or as the concentration of the activator (arabinose) decreases. The graphs in Fig.10 can be simplified into the table below (Fig.11).

Figure 11. Truth table for the behaviour of genetic circuit with respect to glucose and arabinose.


Future work: improve our biosensor

Alongside the good results, it is important to keep in mind that this is not the final state of our IMT3 sensor. From the first version of the sensor IMT3_eGFP, where the results were not as promising, we managed to improve it to the second version (IMT3_sfGFP), which had more stable results improving the accuracy and the precision from our previous version. We already started the third engineering cycle, planning on how to improve our sensor to its next version to further enhance its characteristics.

We would like to study the effect of having two Ligand Binding Domains in tandem instead of only one. Increasing the number of receptors could modify the sensibility or the ability of the sensor to detect lower T3 concentrations [1]. A Low copy number vector was also considered, to reduce the randomness of the high copy plasmid, where sometimes from cell to cell the number of plasmids can vary up to 200 copies. Changing the GFP split point or introducing mutations to the intein or the residues that interact with the ligand are key targets that are meant to be explored.

Specially this last point, mutagenesis, can be extrapolated to Directed Evolution principles, which opens a realm of protein optimization, in which our IMT3 sensor could excel. All these proposed modifications are further explained in our Engineering Success Page.

Extending our biosensor to other applications

In Hormonic, the IMT3 biosensor was built with the objective of finding a good way to continually report to a computational unit the levels of T3 to potentially treat patients with hypothyroidism, but along the way, we envisioned plenty of other possible applications.

With cell free system technologies being cheaper every year, IMT3 biosensors could be introduced to the diagnosis field, by the creation of quick tests, where the reporter protein, instead of being a fluorescent protein, could be a chromoprotein or any other depending on the application. Specially in these times (2020), the adaptability of intein mediated biosensors could help develop new quick diagnosis systems for COVID-19, considering the interaction of Human hACE2 protein with the Spike protein of the virus, finding the binding domains and studying the pairing with the intein domains.

The use of Inteins as sensors has been studied for over two decades. However, this is the first time it has been shown that the two intein domains can be coupled in the same chimeric protein with a ligand binding domain to report fluorescence. Only in the human body there are several proteins that bind to other molecules or proteins that have been studied. This opens the potential for the creation of other Intein Mediated Biosensors, that, not only will detect different molecules, but also could be reported as fluorescence and coupled to computational devices.

Moreover, with the proven capability of our PID to restore the feedback, these types of sensors adapted to sense other molecules allow the capacity of restoring other hormonal disorders, such as diabetes, menopause, imbalances in sexual hormones, or even osteoporosis.

All in all, achieving a functional IMT3 biosensor means much more than reporting T3 concentration. It means that the post-translational protein splicing phenomena is a very strong and versatile tool.

We would also like to acknowledge that this IMT3 biosensor will be used as educational material as an example of a protein biosensor, which will be explained at a high school summer camp organized by “La Fundacio La Pedrera” in Barcelona. For more information visit our Science communication page.


[1] Gonzalez-Flo, E., Alaball, M. E., Macía, J. Two-Component Biosensors: Unveiling the Mechanisms of Predictable Tunability. ACS Synthetic Biology, 2020 9 (6), 1328-1335 DOI: 10.1021/acssynbio.0c00010