Results

Chimeric receptor development

---

Introduction

In our initial experiments we aimed to assemble our chimeric receptor which will be able to sense and bind to the target Anatoxin-a. In order to achieve this, as first step we amplified by PCR using self-designed primers the DNA sequence of the chemoreceptor Pctd sensing domain, the EnvZ signaling domain of the EnvZ/OmpR two-component system, and the vector backbone which harbors the reporter protein GFP. Afterwards, all amplified DNA fragments were assembled together via the Gibson assembly technique and the new plasmid was transformed into our final E. coli strain. Lastly, the E. coli biosensor was tested in the presence of several compounds to assess its sensitivity and reliability.

Results

DNA fragments amplification (PCRs)

Fig.2 shows the agarose gel of the different amplified DNA fragments by PCR, which were assembled afterwards. All fragments were successfully amplified as the bands seem to have the expected size.

In lanes 2 and 3 are found two different fragments of the EnvZ response regulator. The larger one in lane 2, with a size of 842bp, contains the HAMP domain which has been described as essential in the signaling process of the two-component system. The other lacks the HAMP domain and has a smaller size of 708bp.

The HAMP domain connects the periplasmic ligand-binding domain (LBD) of the chemoreceptor Pctd to the cytoplasmic signaling domain, thus it plays an important role in the transmission of sensory signals. Indeed, the packing stability of the HAMP domain has been described to be highly critical to signaling and any displacement of its structural integrity could lead to a non-functional sensory protein. Additionally, extensive mutational analysis has shown that a single point mutation in the HAMP domain can result in the modification of the receptor ligand specificity as well as the constitutive state, therefore the integrity of HAMP becomes crucial.

In our approach, since we aim to generate a hybrid receptor from Pctd and EnvZ, and both of them have their own HAMP domain, we have generated different DNA fragments which might contain or not the HAMP domain that will be fusioned together and assessed to obtain the chimeric receptor with the highest output.

Lanes 4 and 5 show the sensor domain of the chemoreceptor Pctd. Like EnvZ fragments, the pctd gene was amplified twice, with and without the HAMP domain, with sizes of 1365bp and 1195bp, respectively.

Lastly, the backbone pkg116 is shown in lane 6 with an expected size of 4228 bp. It will be used as our reporter plasmid in which the GFP fluorescence will serve as signal to indicate the presence of the toxin. The different inserts of the chimeric receptor will be assembled with this vector.

Plasmid assembly (Gibson + sequencing & gel electrophoresis)

The Gibson assembly method was performed to generate our construct which harbors the chimeric receptor. Following this purpose, 2 different receptors were assembled and transformed directly into chemically competent DH5a cells: PHE (pctd-HAMP + EnvZ + pkg116), in which the HAMP domain comes from the pctd gene, and PEH (pctd + EnvZ-HAMP + pkg116), where the HAMP domain is natively from the EnvZ gene.

Both receptors will be assessed to figure out which has the higher output and therefore we could determine which HAMP domain should be used, either from the receptor or the response regulator, in future attempts to generate new hybrid receptors.

Transformation into E. coli strains (+ colony PCR)

DH5a cells

The first bacterial strain used for transformation of our assembled plasmid was E-coli DH5a cells. This strain was used due to their high transformation efficacy, which allowed us to obtain a great number of transformed colonies to perform a colony PCR.

Figures 3 and 4 show the electrophoresis gel of a colony PCR of the DH5a cells transformed with the assembled PHE and PEH, respectively. Interestingly, the assembly of PEH (Fig 4) seemed to be more efficient than PHE (Fig. 3) as the proportion of DNA bands – each coming from one different DH5a colony (Fig.5) – with the expected length of 2.2kb is higher.

Furthermore, the PHE and PEH plasmids were also isolated through by miniprep and sent for sequencing to check and verify the correct orientation of the inserts. The results (not shown) confirmed that the inserts were assembled in the correct orientation.

VS1007

Once it was ensured that the plasmid was assembled and transformed successfully into DH5a cells, we proceeded to transform both plasmids in our final E. coli strain VS1007 which carries the reporter plasmid harbouring GFP. Like in DH5a cells, we also performed a colony PCR as well as sequencing to verify the integrity of our plasmids.

In Figure 6 it is shown the gel electrophoresis of colony PCR and it can be seen in lanes 6 and 7 the positive colonies carrying the PEH plasmid and PHE plasmid, respectively.

Conclusion & future prospects

We successfully developed our biosensor which is based on the E. coli strain VS1007, that carries the reporter plasmid harboring GFP, by first amplifying and assembling the DNA sequences of our chimeric receptor via Gibson assembly, followed by transformation of the receptor into this bacterial strain.

Owing to the essential role of the HAMP domain in the signaling pathway triggered upon ligand binding, as well as the fact both proteins of our hybrid receptor – Pctd and EnvZ – have their own HAMP domain, we generated two versions of our chimeric receptor, PEH and PHE, that will be assessed in following experiments to obtain the one with higher output.

Strikingly, sequencing results revealed the presence of a P203S mutation in the chimeric receptor, and the same mutation was also found in the original Pctd template DNA. This amino acid is close to the binding pocket of Pctd, thus it may affect negatively the affinity of the receptor to its ligand.

Under this circumstance, we proceeded to assess the consequences of this unexpected mutation in the specificity and reliability of our biosensor. Furthermore, if the results of our experiments testing the biosensor are not good, we have planned to remove the mutation by using self-designed primers to PCR out this mutation near the binding site from the beginning, thereby obtaining a new version of our biosensor.

Drylab results

---

In order to produce a chimeric receptor which Anatoxin-a can bind to, we tried to predict the binding affinity of Antoxin-a to PctD in silicio. As first step we predicted the binding site correlating with Anatoxin-a. This was done with SSNet (1).

SSnet

Molecule	p-Value
Acetylcholine	0.02057565189898014
Choline	0.054201576858758926
Betaine	0.04926067963242531
Anatoxin-a	0.15658815205097198

  

It has previously been shown that ACh, Choline and Betaine all bind the same binding sitei. In our modelling with SSnet Anatoxin-a seems has a lower possibility to bind there compared to the natural ligands(highest p-value). Still SSnet was predicting the binding site of Anatoxin-a as the same as the one from ACh.

In Figure 7 you can see the positions of Pctd where SSnet predicted some binding possibility. In red are the positions where Anatoxin-a is most likely to bind, those are the same positions as for ACh and the other tested molecules. Because the protein is molecular dynamic, the two bidning sites at each of the same chains have not the same quality for binding. That leads to the next step: Predict the binding of Anatoxin-a. 

In the second step, we then predicted the binding affinity of Anatoxin-a and other ligands using SeeSar (2)

.

SeeSar

In Figure 8 is the binding prediction with the highest affinity predicted by SeeSar for Acetylcholine. As you can see, Acetylcholine is sitting deep inside of the binding site (blue-light molecule). Figure 9 is the highest affinity prediction for Anatoxin-a around the binding site, there were other predictions at the intacellular site of PctD. Compared to Acetycholine, Betaine and Choline, Anatoxin-a doesn’t really bind inside of the binding site, sitting way outside of it. It seems that Anatoxin-a is too big to enter it. 

Betaine and Choline are binding at the same location as Acetylcholine. These are the results of SeeSar with the highest binding affinity.

GROMACS

Since the results from both tools suggested binding of Anatoxin-a to PctD is considerably weaker compared to the natural ligands, we questioned wether predictions of Anatoxin-a binding would be more promising when accounting for dynamic conformations of the PctD receptor. We collaborated with iGEM Team iGEM Team Patras to carry out molecular dynamics simulation using Gromacs (3).

  In the Video of the Molecular Dynamic Simulation Anatoxin-a is flipping between the intracellular site (Figure 10)that was also predicted by SeeSar and outside of the real binding site (Figure 11), also like SeeSar predicted.

Even with Molecular Dynamic Simulation, there seems to not be enough space inside of the binding site for Anatoxin-a, even GROMACS has difficulties finding the correct binding sites. Next step would be to try to define the binding site of the protein before the simulation. But when you focus on those binding sites, you can see that they are not really flexible, what makes the chance for the big Anatoxin-a molecule to enter it way smaller. 

Conclusion

After the in silicio analyisis, it is not likely that Anatoxin-a binds to PctD. Initially we wanted to do some mutational analysis, to improve binding of Anatoxin-a to PctD but Prof Thiel explained us that prediction of the effect of mutations on protein structures in silicio is currently too unreliable yet. This is because accurate protein structure prediction algorithms mostly rely on finding similar second structures in homologous proteins. Since the amount of known structures of homologous proteins is considerably lower when you introduce point mutations, the algorithms cannot reliable predict the effect of the point mutation.

Therefore, screening of mutated versions of PctD needs to be carried out in wetlab.

PctD-LBD experiments

---

Introduction

PctD is a scarcely characterized chemoreceptor, found in Pseudomonas aeruginosa. It has been found that it binds to choline and acetylcholine. (4) Anatoxin-a mimics acetylcholine and binds to the nicotinic Acetylcholine receptors (5). To our knowledge, Anatoxin-a has not yet been shown to bind PctD. Since we wanted to use it for our biosensor, we wanted to test the binding of Anatoxin-a to PctD.

Therefore, we planned to express the PctD ligand binding domain (PctD-LBD) in BL21(DE3)pLys E.coli, purify it and use isothermal titration calorimetry (ITC) to test the binding of Anatoxin-a to the protein. This is important to establish, because the functionality of our biosensor is dependent on detection of Anatoxin-a by the PctD ligand binding domain.

We used a plasmid, that was kindly gifted to us by the lab of Victor Sourjik, that contained the His-tagged PctD-LBD to transform BL21(DE3)pLys E.coli cells and overexpress the protein (4).

To purify the protein we used a HisTrap column and the Äkta-FPLC system. We used SDS gel electrophoresis to analyze the expression and purification.

Next, we performed ITC with the purified PctD-LBD and choline as a ligand. This was meant to be a positive control for our ITC experiment with Anatoxin-a as a ligand. Unfortunately, due to delivery issues, Anatoxin-a was not available for us to perform experiments with.

Results

Protein purification

Purification with HisTrap column yielded two peaks at approximately 70 mM imidazol and 90 to 110 mM imidazol. The peaks are indicated with black arrows in Figure 12. To confirm that one of the peaks contains the purified PctD-LBD, samples of selected fractions of both peaks were analyzed with an SDS-PAGE alongside samples of the expression process (Fig.13).

The PctD ligand binding domain with a His-tag has a molecular weight of 39,5 kDa. Bands of the correct size can be seen in the SDS gel in the lanes “L” and “P”, that contain the whole cell lysate and the pellet of the unsoluble fraction, respectively. Bands of the same size can be seen in the lanes of the elution fractions 54 and 57. As expected, all of those bands are thicker than the rest of the visible bands, because the PctD ligand binding domain was overexpressed in our cells. This confirms, that Peak 2 contains the PctD-LBD protein.

The lanes of fractions E54 and E57 show further bands, suggesting that our target protein still contains some contaminations. But the bands are thin and faint, so we considered the fractions pure enough to work with the proteins further.

The fractions of Peak 2 were pooled and a dialysis against ITC buffer was performed. The protein solution was then pooled with the protein we obtained from another successful protein purification (following the same protocol).

The protein concentration was measured with a microvolume spectrophotometer and yielded a concentration of 2.5 mg/ml.

Isothermal titration calorimetrie

For the ITC measurement, we prepared our protein (PctD-LBD) in a concentration of 35µM and choline in a concentration of 490µM in ITC buffer. These concentrations were chosen so that the protein concentration of the protein is between 10-100*Kd (Kd = dissociation constant of ligand-protein interaction), and the final concentration of ligand is 2-times the protein concentration. The dissociation constant (Kd = 2.6 µM) was known from literature (4).

During the ITC measurement, 40 µl of ligand was injected in 2µl doses in intervals of 2,5 min. Temperature changes for each injection were recorded.

If binding of ligand occurs, temperature spikes for each injection are expected, with decreasing peak area over time. This is because the binding sites of the receptors get saturated over time.

Our raw data is shown in Figure 14. The rate of heat release DP in µcal/s is plotte against the time. After integration the peak area is obtained. This corresponds to the enthalpy change dH. When plotting the enthalpy change against the molar ratio, a saturation curve is expected. This is visualized in Figure 15.

You can see that there is no continuous decrease in peak area. The first two peaks have a higher peak area and therefore a higher enthalpy change, the rest of the peaks are of similar, smaller size. Therefore, potting the enthalpy against the molar ratio (Fig. 15) does not show an optimal saturation curve, because the slope consists of only 1 data point.

The data we got could be explained by the fact that the ligand binding is not sufficiently exothermic, so that the signal intensity is too low to be observed. But it has already been shown by others that binding of Choline to PctD-LBD can be detected and quantified by ITC.

Another possible explanation for this is that our effective protein concentration (concentration of functional protein that can bind the ligand) was much lower than the total protein concentration in the solution. This might have resulted in the ligand binding site being saturated very fast, therefore explaining the raw data we got.

Additionally, the analysis software corresponding to the ITC system, was not able to fit a curve to our data and output information about the dissociation constant and other variables, because there was too little heat overall.

Instead, we decided to fit a curve to the data by hand (Fig. 16). We set the dissociation constant from the literature (4) and the concentrations of our ligand and protein as constants, allowing us to vary the number of binding sites of our protein (N(sites)). The detailed fitting parameters can also be found in Figure 16. The curve that fitted our data had a value of N(sites) = 0.01 per protein. This corresponds to only 1% of present protein having a binding site for Choline, supporting our hypothesis that only part of our protein is functional.

Conclusion

To summarize this part of our project, we successfully expressed and purified a protein with the expected molecular weight of around 39.5 kDA, indicating that it is PctD-LBD.

The results of the ITC measurement did not allow us to quantify the binding of Choline to PctD-LBD. But the data we obtained allows the assumption that there is binding of choline to a small subset of protein that is functional. Whether this is indeed the case has to be examined in further experiments.

Future prospects

As it has already been shown in other publications that binding of Acetylcholine and Choline can be verified by ITC, the next step will be to confirm functionality of our purified protein e.g. through Blue Native PAGE or CD spectroscopy. Based on the results, the ITC can be repeated with adjusted concentrations of protein and ligand. Also, the purification protocol of PctD-LBD can be revised if the yield of functional protein is very low.

We expresssed the PctD-LBD in BL21(DE3) pLysS cells. In this strain, T7 RNA polymerase under control of lac UV5 promotor can be induced by IPTG. The T7 polymerase then transcribes target genes under control of the T7 promotor (6). Additionally, the strain contains the pLysS plasmid which encodes T7 lysozyme, reducing the expression of T7 RNA polymerase and thus decreasing expression of the target gene in absence of IPTG (7).

In the paper characterizing the PctD chemoreceptor, PctD-LBD expression was carried out in the BL21-AI strain, which has been engineered to reduce background expression by a different method. In this strain, the T7 RNA polymerase is expressed under the strictly controlled pBAD promotor which is activated by arabinose (4) (8).

The change of expression system compared to the literature might have contributed to misfolding of the protein, rendering it inactive. Expressing PctD-LBD in Bl21-AI cells could be carried out in an attempt to produce more functional protein.

After we verify the expressed protein is functional and binding of Choline and Acetylcholine can be shown by ITC, the last step will be to repeat the experiment with Anatoxin-a to answer the question whether Anatoxin-a also binds PctD.

Testing of Anatoxin-a Biosensor

---

Introduction

As the description of the cloning and assembly of our biosensor shows, we created 2 potential candidates for our biosensor: VS1007 PHE and VS1007 PEH. Both candidates are based on the strain VS996 which lacks all chemoreceptors as well as EnvZ. VS1007 additionally contains the reporter plasmid pAM107 PompC-GFP expressing GFP under the PompC promotor. The biosensor candidates carry a second plasmid which contains the hybrid receptor - either the version PctD-HAMP-EnvZ (PHE) or the version PctD-EnvZ-HAMP (PEH). The hybrid receptor is under control of the PnahG promotor making its expression inducible with salicylic acid or acetylsalicylic acid.

To assure that our chimeric receptors and the signaling pathway leading to expression of GFP work as intended, we carried out the next experiments with the natural ligands choline and acetylcholine. Our plan was to test the biosensor with Anatoxin-a after the initial testing with choline and acetylcholine would have confirmed the functionality of the sensor.

First, we performed a growth experiment to assess the fitness cost of our plasmids. For this, we grew the constructed E. coli strains (containing 0, 1 or 2 plasmids) over a period of several hours. We measured the OD600 every hour to compare their growth curves.

Second, to get a first impression of the fluorescence of our potential biosensor, we imaged the different strains, incubated with no ligand, choline and acetylcholine with an epifluorescence microscope.

Finally, we chose the most promising candidate for our biosensor to measure the response to different concentrations of choline and acetylcholine over an extended period of time.

Results

Growth experiment

The 2 biosensor candidates (VS1007 PHE, VS1007 PEH), as well as the reporter strain (VS1007) and the untransformed VS996 strain were grown in 10ml LB medium in 15 ml polypropylene tubes at 37°C and 200 rpm for 7h. The figure below (Fig. 17) shows the growth curves for each of the strains.

For each timepoint, 3 samples of each strain were measured. The growth curves are constructed from the mean values of these 3 measurements. The standard deviation is given for each timepoint as well.

Fluorescence Microscopy

As we use GFP as a reporter, we expect to see a change in fluorescence of our E. coli cells upon binding of a ligand to the PctD ligand binding domain of our PctD-EnvZ chimeric receptor. Therefore, we used fluorescence microscopy to check whether our biosensor shows the intended behavior. choline and acetylcholine were used as ligands as they are natural substrates of PctD and should activate EnvZ signalling if the hybrid receptor is functional.

We used overnight cultures of our biosensor candidates and our reporter strains and induced the Pnahg rpomoter with acetylsalicylic acid. Approximately 1h before the imaging, the ligands (200µM) were added. We then used a fluorescence microscope to image each of the strains with and without ligand. The images are collected in the Figures 18-20.

The reporter strain VS1007 shows highly visible fluorescence under all 3 conditions. This indicates that the reporter plasmid promotor is leaky, resulting in GFP expression without stimulus.

VS1007, transformed with the plasmid containing PHE shows no fluorescence except for the fluorescence in E.coli. Adding of the ligand shows no visible effect.

VS1007, transformed with the plasmid containing PHE, shows visible fluorescence under all 3 conditions. Because the VS1007 reporter strain shows leakage, some or all fluorescence from the VS1007 PEH strain stems from the reporter plasmid. To see differences in fluorescence of the 2 strains or under different conditions, the fluorescence output needs to be quantified.

VS1007 PEH is the strain that seemed most promising for our biosensor, so we wanted to further test its function. We set up an experiment to quantitatively measure the fluorescence of the strain over time in response to different concentrations of ligand.

Plate reader experiment

To quantify the fluorescence of our bacteria, we set up an experiment that allowed us to measure green fluorescence and absorbance at 600nm in a plate reader at the same time. We used the strain VS996 as controls and compared VS1007 and VS1007 PEH with no ligand and ligand in the following concentrations: 10 µM, 100 µM and 1000 µM. As in previous experiments, acetylcholine and choline were used as substrates.

We inoculated M9 medium for the different test conditions (15 in total) with overnight cultures of the three different strains (VS996, VS1007, VS1007 PEH) to a starting OD600 of 0.2. We induced the expression of the hybrid receptor with 4μM acetylsalicylic acid and added the ligands for the designated samples. Cells were incubated for 4h after induction and samples were taken in triplicates for measurement with the plate reader at 0h, 2h, 4h.

Analysis of absorption at 600 nm (Fig.21 ) shows that absorbance at 600 nm after 4 h is decreased in samples containing acetylcholine and choline in both VS1007 and VS1007 PEH. However, this is at least partly due to differences at the starting absorbance at 600 nm indicating growth is not strongly affected by the compounds. Absorbance at 600 nm is 2 to 3 times higher for VS996 compared to samples of VS1007 and VS1007 PEH which can again be attributed to a difference in the starting cell density.

Interestingly, for VS1007 PEH cells absorbance at 600 nm stayed almost level over the course of incubation indicating slower growth compared to the VS1007 cells. This effect does not depend on the addition of acetylcholine or cholin. Possibly, the induction of the hybrid receptor with acetylsalicylic acid placed a burden on the cells and impaired their growth. In the previous growth experiments, this effect could not be observed because the expression of the hybrid receptor had not been induced.

The figures below show the fluorescence normalized by the cell number (fluorescence/absorbance at 600 nm) over time.

First, we compared the background fluorescence of the different strains without added ligand (Fig. 22).

It is evident that VS966 shows very little fluorescence, probably only the autofluorescence of E. coli cells. Compared to that, the VS1007 reporter strain shows an approximately 3-fold increase in fluorescence. This is expected due to some leakage of the ompC reporter and consequential synthesis of GFP.

Interestingly, the VS1007 PEH strain shows an approx. 6-fold increase in fluorescence compared to the reporter strain. This indicates that our chimeric receptor activates GFP expression without stimulation by a ligand. This possibility is consistent with the literature. There are cases of chimeric chemoreceptors that activate expression of their reporter gene while inactive, and suppress it upon activation and vice versa (9).

The Figures 23 to 26 summarize the normalized fluorescence of the VS1007 reporter strain when incubated with different concentrations of acetylcholine and choline. This strain does not have the plasmid containing the PEH gene, therefore no change of fluorescence upon adding any ligand is expected.

A paired Student´s T-test was performed for the data points of t = 4h for all pairings. The p-values p can be found in the section “p-values”. Significant results (p < 0.05) are indicated by brackets and asterisks in the following figures.

As expected, for addition of choline no significant changes are observed. When adding acetylcholine, the difference between 0µM and 100µM and between 100 µM and 1000 µM are considered significant. But for the first pair, increase of acetylcholine results in increased normalized fluorescence, while the second pair shows the opposite effect. In line with our expectations, this indicates that we do not observe a real effect of acetylcholine here.

Next, the normalized fluorescence of the VS1007 PEH strain when incubated with different concentrations of acetylcholine and choline was compared. Based on the results for the samples with no added ligand, we expect to see a reduction in fluorescence over time. This reduction over time is expected to be more pronounced at increasing ligand concentrations. This is because based on our results we hypothesize that our chimeric chemoreceptor suppresses GFP expression upon ligand binding.

For most test conditions, the relative fluorescence decreases gradually over time. Only for the test condition with 100 µM choline, the fluorescent is increased after 2h of induction and then decreases below the initial level after 4h. Also, an increase of normalized fluorescence (t = 4h) between 0µM and 1000µM, as well as between 10µM and 1000 µM acetylcholine is observed. Paired t-test revealed that this result can be classified as significant. Adding to the high variance across the samples and the non-uniform trend between samples, this suggests that the observed effect is due to statistical fluctuations instead of a biological effect.

For addition of choline, all pairings but 2 showed a significant difference in normalized fluorescence after 4 hours of incubation (Fig. 26). The higher the choline concentration, the higher the normalized fluorescence. Again, high variances across samples can be observed. Taken together with the results for acetylcholine, this could indicate a biological effect. But opposed to our hypothesis, higher acetylcholine or choline concentrations resulted in higher normalized fluorescence.

Conclusion

We transformed the VS1007 reporter strain with plasmids that contain 2 different variants of a PctD-EnvZ chimeric chemoreceptor. We were able to show that our new strains are not impaired in their growth.

We showed fluorescence in the VS1007 and VS1007 PEH strains under the fluorescence microscope, but we could not show any effect of choline or acetylcholine on our biosensor candidate. Therefore, we sought to quantify the effect of different concentrations of choline and acetylcholine on VS1007 PEH by plate reader experiments.

In those experiments we showed that VS1007 had low background fluorescence. However, for VS1007 PEH we observed high fluorescence indicating active EnvZ signalling in the absence of ligands. Assuming a functional hybrid receptor, we would expect a ligand concentration dependent decrease in fluorescence over time. However, both ligands show an increase fluorescence for choline and acetylcholine compared to the samples where no ligand was added.

As both choline and acetylcholine show a trend to increase fluorescence with increasing concentration after 4h of incubation, an effect of those substrate on our biosensor can be suspected.

Future prospects

To further test the function of our biosensor candidate VS1007 PEH, the next step will be to repeat the plate reader experiment described above over an extended period of time with more biological replicates to validate the results. If an effect of choline and acetylcholine is verified, the experiments then could be repeated with Anatoxin-a.

If the biosensor shows low affinity to Anatoxin-a, or weak binding is shown in ITC experiments, the next step will be to increase the affinity of PctD by directed evolution. Then the fluorescence response of the new biosensor strain to binding of Anatoxin-a needs to be quantified, similar to the experiments described above.

After a functional Anatoxin-a biosensor is obtained, a distributable form of the biosensor can be generated. To be usefull for testing facilities, a testing kit for water samples would be the best option. To further optimize this testing kit, the fluorescence reporter could be changed to RFP, so the signal can easily be distinguished from the autofluorescence of chlorophyl. Another improvement would be to broaden the range of cyanotoxins the biosensor can detect.