Team:Bielefeld-CeBiTec/Bacteria

Abstract

We reached our goals in establishing a complex, partially synthetic signaling cascade in E. coli to detect the TNT surrogate BTCA. We showed that an in-silico designed receptor can bind its ligand BTCA with a higher specificity than the natural receptor it is based on. This is a proof of concept for our project and shows that it is possible to design de novo receptors by computational means and thus engineer organisms to detect certain chemicals and act as a biosensor with our approach. In addition, we have carried out several toxicity tests on chemical weapons substitutes and degradation products.
Nici

Wiki Tour

  • After testing our system in bacteria, we tested it in a real plant. It looks very much like me: Tobacco

Bacterial Chemotaxis – The root of P.L.A.N.T.

Two-component systems are found in eukaryotes and prokaryotes and allow sensing of environmental conditions in a stimulus-response mechanism. Since two-component mechanism can be found in plants and bacteria and is well conserved throughout various species, it is feasible to engineer chimeric two component systems. Our signaling cascade in plants is based on a two-component system which is involved in chemotaxis of Escherichia coli (E. coli). Chemotaxis is a complex mechanism that allows cells to sense and move towards attractants or move away from repellents. The following passage focuses on the chemotaxis of E. coli, since it has been proven as the organism we successfully established the signaling cascade in.
The first step of the complex signaling during the chemotaxis is the binding of the ligand, for example a carbon source such as ribose or glucose, to a specific receptor located in the periplasm, the periplasmic binding protein (PBP). PBPs are homologous proteins, that consist of two globular domains and a connecting hinge region. Upon binding of a ligand, a drastic conformational change is induced. The so-called “Venus-flytrap”-mechanism enables embedding of the ligand between the globular domains and allows the binding of the periplasmic binding protein by membrane-bound receptors that are specific for each PBP [1]. During our work, we mainly focused on the ribose binding protein (RBP). In the case of RBP, as well as the galactose/glucose binding protein (GBP), the membrane bound receptor involved in the signaling cascade is the Trg receptor. This receptor type officiates as a homodimer and consists of a periplasmic ligand-binding domain, a helical transmembrane segment and a helical cytoplasmic region [2]. The cytoplasmic part of the receptor serves as a framework for a super-molecular receptor-kinase signaling complex. In this complex the histidine kinase CheA is bound to the receptor. The binding is facilitated by a conformational change upon PBP binding to Trg [3]. The activity of the kinase on the other hand is downregulated upon ligand binding by Trg, which in turn inhibits the activity of the two response regulators CheB and CheY [3]. CheB is a methylase that acts antagonistic to the methylation enzyme CheR. CheR methylates Trg and makes it less responsive to the binding of the ligand, while CheB reverts this reaction, therefore increasing sensitivity to a ligand. Phosphorylated CheY docks to the flagellar motor protein and induces a tumbling state [3]. The inhibition of the activity of both CheB and CheY by the binding of the ligand results in a stable forward swimming state along an increasing attractant gradient.

Figure 1: The bacterial signaling cascade.

The bacterial signaling cascade

Our signaling cascade is depicted in figure 1. The two-component system of the chemotaxis was engineered to allow recognition of a ligand of interest as a stimulus and result in a visible output. In our case, the stimulus were chemical weapon degradation products and a green fluorescent protein (GFP) is used as response indicator. The engineered signaling cascade is based on a computationally evolved RBP, that binds chemical weapon degradation products. Baumgartner et al. first integrated the downstream chemoreceptor Trg to the osmosensor EnvZ in 1994 [4]. The resulting Trz fusion protein was composed of the periplasmic and the transmembrane domain of the chemotaxis receptor Trg which was linked via the HAMP-domain with the cytoplasmic part of the histidine kinase EnvZ [4]. Both parts of the Trz1 protein are derived from E. coli and combine the ability of recognizing ligand-bound RBPs of Trg with capability to phosphorylate the transcription factor OmpR upon contact with the RBPs. The resulting signaling cascade can detect the RBP-bound ligands. As a consequence of detection, expression of genes that are controlled by the OmpC promoter, which is activated by phosphorylated OmpR is induced [5, 9]. In the past it has been attempted to alter the binding pockets of PBPs to allow detecting of other ligands, and thus turn E.coli into a biosensor for various small molecules, while maintaining the same signaling cascade and reporter output downstream [6, 7]. This approach has been successfully transferred from bacteria onto plants by the group of Medford in 2011 [8]. In our project, we aimed at re-designing this signaling cascade in plants, to make it suitable for the detection of chemical weapons and their precursors in a natural environment.
We chose to first test the signaling cascade in E. coli as a proof of concept before integrating our novel system within Nicotiana benthamiana, as a time efficient way to establish the various modified PBPs connected with the signaling cascade and to assess in more detail their specificity and affinity to the chemicals.

Plasmid structure of the bacterial signaling cascade

The signaling cascade was designed on the pSB1C3 (Figure 2). We chose to make the signaling cascade inducible to be able to compare the output of the cascade before and after expression. This also reduces stress during early growth phases. We chose the L-arabinose inducible pBAD promotor system, consisting of AraC, the regulator of the arabinose operon and the AraBAD promotor, which is regulated by AraC activity. After induction with arabinose, the promotor has a low to medium activity, reducing stress and formation of inclusion bodies. The AraBAD promotor controls the expression of the synthetic membrane-bound Trz receptor and the ssTNT receptor [7]. The sequence of the TNT receptor was modified to be cut out by HindIII to simplify exchange with our computationally designed receptors. However, the TNT receptor was missing a periplasmic localization peptide and was therefore not exported into the periplasmic space after expression, thus could not activate the signaling cascade. The OmpR-inducible OmpC promotor regulates the expression of the GFP, that marks the ending of the signaling cascade. OmpR occurs natively in E. coli and is therefore not required to be expressed by separately [3, 9]. Since OmpC can also be induced by osmotic stress, we had to assess whether the addition of arabinose and ligands influenced the expression of GFP, due to the changes in osmolarity.

Figure 2: Plasmid structure of the bacterial signaling cascade.

Preliminary work

We constructed the bacterial signaling cascade encoded by a one plasmid system and the single parts were ordered as gene synthesis products. Complementary overhang regions to the pSB1C3 plasmid were added via touchdown PCR. The signaling cascade was cloned into pSB1C3 vector by Gibson assembly and transformed into E. coli DH5α. Positive colonies were verified by colony-touchdown PCR and grown in liquid culture overnight at 37°C. The plasmids were isolated and again verified by touchdown PCR. Since the TNT receptor is not translocated to the periplasmic space, it was tested whether the endogenous RBP/GBP expression is sufficient for activating the signaling cascade. A preliminary growth experiment was conducted with eight positive clones. The clones were inoculated with an optical density (OD) of 0.1 in 15 mL LB media in Erlenmeyer flasks and grown at 37°C at 180 rpm in an incubator. After 2 hours 0.2% (0.013 M) arabinose was added. After additional 4 hours, ribose (20 mM) was added. Liquid cultures were subsequently grown overnight. The liquid cultures were then centrifuged. For six out of 8 cultures, a greenish color in the pellets was observed. One strain showed a pellet with most intense green color, hence, it was used for subsequent experiments.
A second growth experiment was carried out under the same conditions as described above, with the exception that other ligands were supplemented. Since E. coli favors glucose over ribose as a carbon source, we concluded that the GBP may have a higher affinity to Trz and, therefore, may be the more effective inductor of the signaling cascade. Glucose was added at a concentration of 20 mM and 200 mM to evaluate whether a higher concentration leads to more GFP expression. Additionally, we tested the influence of the changes in osmolarity by adding different amounts (50%/25%/12.5% w/v) of polyethylene glycol 4000 (PEG4000) without prior induction. PEG4000 is a chemically inert polymer, that is not metabolized by E. coli and is therefore well suited as osmolyte. A control run was carried out with only arabinose to rule out, that arabinose is sufficient to activate the signaling cascade. Further experiments were performed with the ligand and without arabinose induction, to rule out the osmotic pressure from the ligand addition as a contributing factor. The conditions are shown in table 1.

The liquid cultures were pelleted by centrifugation. Several pellets appeared greenish while the negative control lacked a greenish color (Figure 3). The gradation was the following from greenest to palest:

  1. Arabinose + Ribose
  2. Arabinose + Glucose (20 mM) / Arabinose + Glucose (200 mM)
  3. PEG4000 (50% w/v) / PEG4000 (25% w/v)


The other pellets, except for the negative control also displayed a greenish color but were not differentiable by eye.

To verify, that the green color was due to increased GFP expression, we used confocal laser scanning microscopy (CLSM) (Figure 4). The pellets were dissolved in 1 mL PBS pH 7.2, and random spots were examined. The resulting histograms were plotted with their fluorescence intensities against their frequencies (Figure 5).

It appears that induction by arabinose + ribose leads to the highest GFP expression. The induction by osmotic stress does not seem to differ drastically from the base fluorescence of the negative control, therefore indicating, that in fact the signaling cascade is expressed and contributes to the fluorescence. The results also suggested that ribose is more suitable for following experiments since it seems to lead to higher induction of the signaling cascade. However, the results only show a trend and are not statistically relevant. Furthermore, the overall low fluorescence activity may stem from the low expression of RBP, the starting point of the signaling cascade. Therefore, the next step was the overexpression of RBP to increase the initial stimulus, that consequently leads to a higher output signal.

Table 1: The combinations of additives for the second growth experiment.
Figure 3: The pellets of arabinose + ribose supplemented E. colis (left) and the negative control (right).
Figure 4: Confocal laser scanning microscopy of different additive combinations. A positive control with GFP under the control of a T7 promotor was added.
Figure 5: The histograms of the microscopy images plotted with their fluorescence intensities against their frequencies. The positive control from before was excluded.

Overexpression of RBP

Removal of the TNT-receptor by HindIII digestion was not successful. We decided to co-transform the pSB1C3 plasmid containing the signaling cascade with the pJOE containing a RBP, that was constructed for our library experiments . Furthermore, the fact that TNT-receptor remained in the signaling cascade did not pose as a problem, as it is not exported into the periplasmic space and does not influence the intracellular signaling. In theory, the plasmids should be incompatible because they belong to the same origin of replication incompatibility group. However, both plasmids are high copy plasmids, making it less likely that one plasmid is displaced. Additionally they contain different antibiotic resistance genes, providing a necessity for keeping both plasmids. Thus, in our case, using both plasmid presumably leads solely to a reduced copy number within the cell. The plasmids were co-transformed into E. coli BL21(DE3), as this expression strain contains a lac-operon promotor controlled T7 polymerase gene, that allows expression of the T7 promotor controlled RBP of pJOE vector. After heat shock transformation and subsequent over-night growth on the LB-agar plates supplemented with ampicillin and chloramphenicol, biological triplicates were picked, grown in 15 mL LB-media at 37°C over-night in Erlenmeyer flasks. 10 mL of the liquid cultures were used for plasmid isolation. The parts of the bacterial signaling cascade and the RBP were confirmed by touchdown-PCR. Plasmid isolation was chosen over colony PCR because RBP naturally occurs in BL21(DE3) and would inherently lead to false positive results.

The remaining 5 mL of each culture were used for a preliminary growth experiment to assess which culture has the highest fluorescence after induction with arabinose. The colonies were again inoculated at an OD of 0.1 in LB-medium with antibiotics. After 2h, arabinose was added, after 4h IPTG (1 mM) was added and after 6h ribose was added. The cultures were grown over-night and centrifuged, where the pellet of the three replicates appeared in the same greenish color. The strain with the greenest pellet was chosen for the following experiments.

The next growth experiment was conducted to quantify whether the additional overexpression of RBP leads to higher fluorescence. Addition of arabinose, IPTG and ribose to the media was performed as for the preliminary experiment. One strain without RBP overexpression and one strain with RBP overexpression were tested with the same combinations of additives, shown in the table 2. No biological replicates were used in this experiment. The liquid cultures were diluted to OD 2 and 300 µL of each culture were transferred thrice to a 96 well plate. GFP fluorescence was measured in the Tecan infinite M200 microplate reader (from now on called Tecan reader). Each well was measured in 9 technical replicates at overlapping but different spots. The mean of the technical replicates was calculated automatically by the manufacturer’s software. The means and the standard deviations are shown in figure 6. The fluorescence of the background of the LB-medium was subtracted.

In general, we demonstrated in our experiment, that the GFP expression increases with RBP overexpression, showing that the endogenous RBP expression is a bottleneck in the signaling cascade. Induction with arabinose and subsequent addition of ribose leads to the highest fluorescence. The ratio of the cultures induced and supplemented with ribose relative to the negative control increased greatly in RBP overexpressing strains. This indicates that the overexpression does in fact increase the stimulus of the signaling cascade and further validates, that the signaling cascade is working. IPTG does not show a positive effect on the signaling cascade. Possible explanations to this observation include that the lac-promotor is leaky to uninduced expression or that IPTG induction leads to formation of inclusion bodies. The pellets from other cultures, with exception of the negative control, also demonstrated a greenish color, however not differentiable by eye.

Table 2: The combinations of additives for the growth experiment for overexpression of RBP.
Figure 6: The results of the RBP overexpression. The different conditions are shown on the x-axis, the relative fluorescence on the y-axis. Cultures with the signaling cascade without overexpression of RBP (left) and cultures with the signaling cascade and additional overexpression of RBP (right) are shown. Error bars indicate standard deviation of the 9 measurements of each well.

Overexpression of computationally designed receptors

After the successful implementation of the experiments towards establishing the signaling cascade with RBP overexpression, we decided to overexpress our computationally designed diisopropyl methylphosphonate (DIMP) and 1,3,5-benzenetricarboxylic acid (BTCA) receptors for the purpose of testing them in-vivo. We used vectors pRSETB-DIMP-receptor/BTCA-receptor for expression of the receptors for in-vitro testing, were used for a co-transformation with the signaling cascade in BL21(DE3). The above-mentioned plasmids belong to the same ori incompatibility group, but are sustained by the cell by the same mechanisms as for the pJOE-RBP construct used in co-transformation. After heat shock transformation, colonies were picked and cultivated in 15 mL LB-medium supplemented with ampicillin and chloramphenicol, and grown under the same conditions as before. After plasmid isolation, the correct incorporation of the parts of the signaling cascade and the respective receptors were verified by PCR. A growth experiment was conducted. Cells were grown in 1 mL LB supplied with antibiotics in a 96-deep well plates at 37°C on a rotary shaker at 500 rpm. Additives were added in the same manner as before, combinations of additives are shown in the table below. Three biological replicates were tested. Fluorescence was measured in 96-well plates in the Tecan-reader. For each well, nine measurements were conducted, and cells were diluted to an OD of 1. The fluorescence of the LB-medium was subtracted.

All three replicates show the same trends in fluorescence levels. The sole addition of either arabinose or BTCA in various concentrations led to fluorescence levels similar or below the fluorescence levels of the negative control (Figure 7). This finding indicates no significant activation of the signaling cascade. The saturation concentration of BTCA in ethanol was experimentally determined to be 0.142 M. It was diluted accordingly. Addition of arabinose and BTCA led to significantly higher fluorescence levels, with the culture exposed to the highest concentration demonstrated the highest fluorescence. Addition of 1% ethanol to the cultures grown on arabinose, also increased fluorescence in comparison to the negative control. This can be explained by the osmotic pressure from ethanol, which activates the osmotic stress responsive promotor. However, the addition of the ligand only did not result in higher fluorescence and the arabinose supplementation is not sufficient for inducing osmotic stress, it is likely, that the signaling cascade is the main contributor to the higher fluorescence levels. This provides evidence for the receptor binding the ligand. This conclusion is in line with the results of the in-vitro testing and is a proof of concept for the computational receptor design of the receptor binding specifically to the BTCA. To provide further evidence, yet another growth experiment was done to compare the abilities of the BTCA-receptor and RBP to activate the signaling cascade. The DIMP-receptor showed no affinity to DIMP nor ribose (data not shown). Possible ways to obtain a functional DIMP-receptor are computational re-design, or mutation of RBP by Darwin assembly. Our attempt toward construction thereof can be found in description of our library experiments .

Table 3: The combinations of additives for the growth experiment for overexpression of the BTCA-receptor/DIMP-receptor.
Figure 7: The results of the BTCA-receptor overexpression of replicate 2. The different conditions are shown on the x-axis, the relative fluorescence on the y-axis. Error bars indicate standard deviation of the 9 measurements of each well.

Endogenous receptor vs. computationally designed receptor

For the testing of efficiency of ligand binding of the RBP compared to the computationally designed BTCA-receptor, we conducted a cultivation experiment under the same conditions as the last experiments. The combinations of additives are shown in the table 4. Biological triplicates of the BTCA-receptor overexpressing strain 2 and of a RBP overexpressing strain were used. Fluorescence was measured in 96-well plates in the Tecan-reader. Each well was measured nine times, cells were diluted to an OD of 1. The fluorescence of the background of LB-medium was subtracted.

The ethanol, and negative controls do not differ significantly, confirming that ethanol has no effect on the fluorescence, and that in fact BTCA activating the signaling cascade is the main contributor to the increased fluorescence (Figure 8). The RBP overexpressing culture supplemented with ribose shows no significantly higher fluorescence than the negative control. One possible reason for this is, that no arabinose was added to the culture media. Further in-depth experiment to elucidate the exact reason were not concluded yet due to time constraints, however the most important comparison between cultures/constructs with overexpressed RBP and the BTCA-receptor with addition of BTCA are currently undergoing. The mean fluorescence intensity measured is approximately twice as high for the BTCA-receptor binding BTCA as for RBP binding BTCA. A one-sided T-test for unpaired data showed a significant higher fluorescence for the BTCA-receptor binding BTCA (p = 0.041, n = 3). This shows that the BTCA-receptor has a higher affinity to BTCA than RBP. It is a proof of concept for the computational receptor design and our project.

Table 4: The combinations of additives for the growth experiment for overexpression of the BTCA-receptor/RBP.
Figure 8: The results of the BTCA-receptor overexpression vs. the RBP overexpression. The different conditions are shown on the x-axis, the relative fluorescence on the y-axis. Error bars indicate standard deviation of the three biological replicates.

3D deconvolution widefield fluorescent microscopy and super resolution 3D structured illumination microscopy

Lastly, we performed microscopy experiments to detect the GFP induced by the signaling cascade on a single cell level. For this purpose, we had access to a Deltavision OMX V4, a cutting-edge life cell imaging microscope, which uses Fourier transformation to gain a resolution higher than the diffraction limit of visible light. We did 3D deconvolution Widefield Fluorescent Microscopy and super resolution 3D-structured illumination microscopy. The BTCA-receptor culture supplied with BTCA and the respective negative control were used in this experiment.
Figure 9 shows the 3D deconvolution widefield fluorescent microscopy, the differential interference contrast channel and the eGFP channel and the merged overlay of both channels in various magnifications. Evidently, the BTCA supplemented BTCA-receptor culture displayed a GFP fluorescence while the negative control showed none.
The results of the super resolution 3D-structured illumination microscopy can be seen in figure 10. The differential interference contrast channel, the eGFP channel, and the merged overlay of both channels in different magnifications are shown. The super resolution 3D-structured illumination microscopy confirms once again that we obtain a green fluorescence after arabinose induction and addition of BTCA to the culture, while the negative control showed no fluorescence.

Figure 9: 3D deconvolution widefield fluorescent microscopy of BL21(DE3) with the signaling cascade overexpressing the BTCA-receptor with either arabinose and BTCA added during growth (BTCA-receptor + BTCA) or nothing added (negative control). The differential interference contrast (DIC) channel, eGFP channel and the merged overlay are shown. Different magnifications of the same bacteria are shown.
Figure 10: Super resolution 3D-structured illumination microscopy of BL21(DE3) with the signaling cascade overexpressing the BTCA-receptor arabinose and BTCA added during growth (BTCA-receptor + BTCA). The differential interference contrast (DIC) channel, eGFP channel and the merged overlay are shown. Different magnifications of the same bacteria are shown.

Test of chemical toxicity on bacteria

Degradation products of chemical weapons can be less dangerous than the weapons themselves. However, for many of them, there are still toxic effects known. For some others, there is not sufficient data available. As we use organisms for their detection, it is crucial that they withstand possible toxic effects. We tested this not only for plants but also for bacteria as our second test system. Another goal was to test which concentrations of chemical weapon degradation products E. coli tolerates and also allows protein expression, as this is necessary for the signaling cascade to work.
E. coli DH5𝛼 expressing mRFP in pSB1C3 was cultivated in a BioLector bioreactor (m2p-labs). Liquid chemical weapon degradation products (DIMP, DEMP, TDG) were diluted in LB media. For BTCA, a saturated solution in ethanol (0.14 M) was prepared, which then was diluted in LB media. MPA was directly solved in LB media (0.56 M) which served as the first dilution level. For each substance and dilution level, triplicates were cultivated. OD600 and fluorescence intensity for dilutions from 10⁻¹ to 10⁻⁷ (for MPA 10⁻⁰ to 10⁻⁶) were measured during cultivation using the Biolector and FLowerPlates.


DEMP and MPA

In figure 11 it can be seen, that significant mRFP expression starts at dilutions of 10⁻³ (slightly at 10⁻²) for DEMP and for MPA at dilutions of 10⁻¹. For DEMP this is supported by measured values (figure 12). The measurements are not shown for the other chemicals, as they only support observations already seen on images of the well plates. Positive and negative controls show the expected results. The 10⁻⁰ dilution of MPA shows a color that is distinct from both, controls and other dilutions. This can either mean that there was some growth or that MPA influences the color of the LB media. The 10⁻¹ dilution of MPA seems to show a color that is slightly darker than positive control and other dilutions. This could mean that there is more growth in that concentration than in the other samples. One explanation is that the bacteria may be able to metabolize MPA and therefore grow to a higher optical density. This should be taken into account for the detection of MPA, as the MPA concentration would decrease over time.


DIMP and TDG

In figure 13 fluorescence can be clearly seen for DIMP starting at a dilution of 10⁻² and for TDG starting at 10⁻³, at least for two of three replicates. All fluorescent samples appear to have similar intensity compared to the positive control.


BTCA and EtOH

BTCA is almost not soluble in water so that it was solved in EtOH. However, this could affect the outcome. To determine whether observed effects are caused by BTCA or by EtOH, EtOH was measured as a comparison. In figure 14 fluorescence can be clearly seen for BTCA as well as for EtOH starting at a dilution of 10⁻². Positive and negative controls show the expected results. The 10⁻² samples seem to be slightly brighter than the other wells. This may indicate that the bacteria are able to metabolize EtOH and possibly also BTCA. This should be considered when using bacteria to detect these chemicals, as concentrations would change. Another explanation could be that EtOH fluoresces itself. The 10⁻¹ wells of BTCA seem to differ in color. This again could be an effect caused by BTCA, as it seems to be the case for MPA.


Summary

E. coli is able to grow in MPA starting from the 10⁻¹ dilutions, in EtOH, BTCA, and DIMP starting from 10⁻² dilutions and for TDG and DEMP starting at 10⁻³. These dilutions are therefore the highest that can be detected by an E. coli based detection system. It should be taken into account that EtOH, BTCA, and MPA may be metabolized by E. coli so that concentrations in the media would change during cultivation.

Figure 11: Photo of well plate with fluorescent E. coli taken on an UV light table. The first column contains positive (P) and negative (N) controls, the other columns the dilutions of chemicals DEMP and MPA.
Figure 12: Exemplary plot of mean values of measured fluorescence for DEMP.
Figure 13: Photo of well plate with fluorescent E. coli taken on an UV light table. The first column contains positive (P) and negative (N) controls, the other columns the dilutions of chemicals DIMP and TDG.
Figure 14: Photo of well plate with fluorescent E. coli taken on an UV light table. The first column contains positive (P) and negative (N) controls, the other columns the dilutions of chemicals BTCA and EtOH.

Conclusion

We reached our goals in establishing a complex, partially synthetic signaling cascade in E. coli to detect the TNT surrogate BTCA. We showed that one of our computationally designed receptors can bind its ligand with a higher specificity than the receptor it is based on. This is a proof of concept for our project and shows that it is possible to design de novo receptors and thus engineer organisms to detect certain chemicals and act as a biosensor with our approach. The signaling cascade poses as a powerful and diverse tool that could be designed by de novo and constructed by the means of synthetic biology and be applied in a wide range of applications ranging from biotechnology to biomedicine.

References

  1. Borrok, M. J., Zhu, Y., Forest, K. T. & Kiessling, L. L. Structure-based design of a periplasmic binding protein antagonist that prevents domain closure. ACS chemical biology 4, 447–456 (2009).
  2. Antommattei, F. M., Munzner, J. B. & Weis, R. M. Ligand-specific activation of Escherichia coli chemoreceptor transmethylation. Journal of bacteriology 186, 7556-7563 (2004)
  3. Falke, J. J., Bass, R. B., Butler, S. L., Chervitz, S. A. & Danielson, M. A. The two-component signaling pathway of bacterial chemotaxis: a molecular view of signal transduction by receptors, kinases, and adaptation enzymes. Annual review of cell and developmental biology 13, 457–512 (1997).
  4. Baumgartner, J. W. et al. Transmembrane signalling by a hybrid protein: communication from the domain of chemoreceptor Trg that recognizes sugar-binding proteins to the kinase/phosphatase domain of osmosensor EnvZ. Journal of bacteriology 176, 1157–1163 (1994).
  5. Tavares, D. Changing the ligand-binding specificity of E. coli periplasmic binding protein RbsB by rational design and screening. University of Lausanne 2020.
  6. Tavares, D. et al. Computational redesign of the Escherichia coli ribose-binding protein ligand binding pocket for 1,3-cyclohexanediol and cyclohexanol. Scientific reports 9, 16940 (2019).
  7. Antunes, M. S. et al. Programmable ligand detection system in plants through a synthetic signal transduction pathway. PloS one 6, e16292 (2011).
  8. Looger, L. L., Dwyer, M. A., Smith, J. J. & Hellinga, H. W. Computational design of receptor and sensor proteins with novel functions. Nature 423, 185–190 (2003).
  9. https://www.uniprot.org/uniprot/P0AA16.