Results

Overview

We started the development of our VitroZymes in the laboratory by testing the sensitivity of the riboswitch (BBa_K4868000) in our live-cell system, which has been undocumented so far. This experiment laid the foundation for the utilization of the selection system encoded in our insertion piece called RGFr (BBa_K4868001). Further testing was required to confirm the induction of the respective dehalogenases in IPTG-rich media. Additionally, the enzymatic activity of two of the original enzymes, DeHa 4 (BBa_K4868998) and DeHa 5 (BBa_K4868999) was investigated using F-NMR. Lastly, we planned to compare the enzyme kinetics of the original dehalogenases, DeHa 1, 2, 4, and 5 to the most fluorescent dehalogenase variants from a mutant library created with error-prone PCR. Unfortunately, this last experiment was not performed due to hardware delivery issues.

Testing the inducibility of the fluoride sensitive riboswitch of RGFr

Before it is possible to use the riboswitch efficiently it is necessary to test if it works and how much fluoride is required to measure a difference in the sfGFP expression.

Our plasmid consists of RGFr where there is a constitutive active promotor followed by the riboswitch and a superfolder Green fluorescent protein (sfGFP). The riboswitch controls the translation of sfGFP and is first activated when fluoride ions are present in the cell1. This means that the greater the amount of fluoride present in the cell, the more sfGFP will be translated. Hence, when our Dehalogenases degrade PFOA the fluoride concentration will increase, resulting in greater amounts of expressed sfGFP. Therefore, more efficient variants of the DeHas will generate a higher amount of fluoride which will lead to an increased expression of GFP, which enables efficient selection of our optimized dehalogenase enzymes.

In the article the riboswitch has been tested in a cell-free system, and it was found that concentrations as low as 0.1 mM1 affect the activity of the riboswitch. See Figure 1 on the Design page for a more graphic illustration.

However, live-cell systems, such as the one used for the selection of the most efficient enzymes, differ in how they work, since there are different pathways and proteins in a bacterium that cannot be replicated in a cell-free system. Fluorescence microscopy was done for qualitative evaluation of the fluoride induced sfGFP, for NEB T7 (NEB #C2566I) containing our plasmid which is grown in varying fluoride concentrations (Fig. 1).

The NEB T7 (NEB #C2566I) strain showed a gradual increase in GFP intensity in increased NaF concentrations. Low levels of sfGFP could also be observed in the presence of 0 mM NaF, indicating that the riboswitch has leakage. sfGFP distribution throughout the population seemed to be not entirely uniform.

Fmicroscope-NEBT7-fig
Fig. 1:Fluorescence microscopy images of NEB T7 cells (NEB #C2566I) carrying the RGFr containing plasmid with fluoride inducible sfGFP expression at 0, 47, 84, and 112 mM NaF. The image shows the increase in sfGFP expression in NEB cells was increased with the NaF concentration.

Quantitative assessment of the riboswitch induction via FACS and plate reader experiments

FACS and plate reader were used to assess the fluoride inducibility of the riboswitch controlled sfGFP in the NEB T7 E. Coli strain (NEB #C2566I).

In the FACS and plate reader experiment the results show that the riboswitch needs around 50 mM NaF in most of the plasmids to register a change in sfGFP expression (Fig 2). This indicates that the riboswitch sensitivity is low, which could cause problems when testing for PFOA degradation. This is because PFOA has a solubility of 1.2 mM in water which would result in a fluoride concentration of 16.8 mM PFOA which might be too low to observe a fluorescence response. As a solution for this we grew them on agar plates covered with a PFOA layer.

F-plate-fig F-FACS-fig
Fig. 2: Left: Plate reader experiment, which measured fluorescence with various gene insertions (DeHa1, 2, 4 and 5) into the insertion site of RGFr. NEB WT shows the fluorescence of the wild type NEB T7 (NEB #C2566I) strain without any plasmid. sfGFP fluorescence is normalized to the optical density. Right: Flow-cytometry measurements of the same cultures. The 99 percentile (top 1% fluorescent cells) of sfGFP fluorescence (FITC-A) is shown for each strain.

GFP Fluorescence Distribution of Cells Containing the Original Dehalogenases

It was necessary to access the effectivity of the original dehalogenases and therefore an experiment was set up using FACS and violin plots (Fig. 3). This experiment determined the distribution of fluorescence from sfGFP in samples with the different DeHAs, a sample containing RFP at the insertion site. It can be seen from the figure below that the riboswitch is inducible with fluoride in all the plasmids harboring DeHas (Fig. 4-7). Notably, it can be observed that there are often fluorescent and non-fluorescent populations within the same sample. No fluorescence induction could be observed in wild type (NEB #C2566I) strain (Fig. 8). Notably a big fluorescence population is observable at 0 mM NaF. This indicates a leakage of the riboswitch.

fig3-violin
Fig. 3: Violin plot of GFP fluorescence from NEB T7 (NEB #C2566I) containing RGFr. It can be seen that the higher the concentration of NaF the more fluorescence can be detected for RGFr (BBa_K4868001).
fig4-violin
Fig. 4:Violin plot of GFP fluorescence from NEB T7 (NEB #C2566I) containing RGFr with DeHa 1 (BBa_K4868996).
fig5-violin
Fig. 5: Violin plot of GFP fluorescence from NEB T7 (NEB #C2566I) containing RFGr with DeHa 2.
fig6-violin
Fig. 6: Violin plot of GFP fluorescence from NEB T7 (NEB #C2566I) containing RFGr with DeHa 4.
fig7-violin
Fig. 7: Violin plot of GFP fluorescence from NEB T7 (NEB #C2566I) containing RGFr and DeHa 5.
fig8-violin-control
Fig. 8: Violin plot of GFP fluorescence from NEB T7 cells wildtype (NEB #C2566I). Roughly the same level of fluorescence can be seen in all the samples at all the tested concentrations of NaF.

Selecting for efficient error-prone DeHas via FACS

To obtain a variety of different DeHas (DeHa1, 2, 4, and 5 respectively) an error-prone PCR was performed. Golden Gate was performed to insert the error-prone DeHas into the RGFr carrying plasmid, resulting in a diverse mix of plasmids with slightly different versions of the respective DeHa. Subsequently this DeHa mutant libraries were transformed into NEB T7 (NEB #C2566I) and plated out on LB-media plates with an antibiotic for selection and a PFOA layer (88 µL/cm2 of evaporated 1.2 mM PFOA solution). Therefore, colonies with more efficient DeHa versions would result in higher fluorescent cells.

The plates were incubated overnight at 37°C for optimal growth of the bacteria and were then moved to 25°C, since it was estimated that the enzymes have a higher efficiency at that temperature.

The selection of our highly efficient mutated DeHas was performed using FACS. The samples were prepared by resuspending and pooling all colonies of a transformation plate, mixing them thoroughly. This was performed for each DeHa (1, 2, 4, and 5) respectively. The single top 0.1% fluorescent cells were sorted at 2000 events/second into 96-well plates. 40 variants from all DeHas that showed subsequently the highest fluorescence on the 96-well plate in the plate reader were sent for sequencing.

Furthermore, the fluorescence profile of the original DeHa populations (also grown on PFOA plates) were measured and compared to the fluorescent profile of the mutant populations (Fig. 9) The recorded data were used to compare the fluorescence distribution of each original DeHas compared to the corresponding mutated dehalogenase pool (Fig. 9).

fig9-violin
Fig. 9: Violin plots that show the distribution of fluorescence of cells containing each original dehalogenase compared to its mutated counterparts. A is DeHa 1, B is DeHa 2, C is DeHa 4 and 5 is DeHa 5. By comparing the original enzymes (yellow plot) with the mutated ones (blue plot), it can be seen that the mutated DeHa 2 and 5 fluorescence are more than the original variant, For DeHa 4 and 1 there is no significant difference, and DeHa 1 have gotten worse.

When looking at Figure 9, a clear difference can be observed between the original DeHa and the mutated DeHa pool of DeHa 2 and DeHa 5. Based on the plots of DeHa 2 and DeHa 5 it indicates that the original DeHa 2 and DeHa 5 are not very efficient at degrading PFOA compared to the mutated counterparts, under the same conditions. However, DeHa 1 and DeHa 4 show approximately the same level of sfGFP as their mutated counterparts. Considering that there were thousands of variants generated per DeHa to begin with and only a small fraction of the pooled variants were analyzed, there is a certain likelihood that no beneficail mutations were present in this pool of mutants.

Evaluating performance of selected DeHa versions

The selected 40 variants were tested in a cell system, where each mutation variant, and the original DeHas, were plated out on agar plates with a concentrated PFOA layer in the same manner as for the plates for the FACS experiment. Fluorescence of the colonies was monitored with exposure to blue light in the GelDoc (BIO-RAD 10000125685) to determine which of the plates had the most fluorescent colonies. For further details on how this experiment was conducted, please visit our Experimental Design page. A side-by-side comparison in fluorescence of the mutated variants and original DeHas can be seen in figure 10.

fig10-F-bargraph
Fig. 10: Fluorescence comparison of the mutated DeHa variants and original DeHas. Original DeHas are marked in red, DeHa 1 is marked in blue, DeHa 2 is marked in green, DeHa 4 is marked in purple and Deha 5 is marked in yellow. The black error bars indicate the variation in fluorescence measured between the triplicates of each variant.

The pattern that can be seen in Figure 10 matches the previously obtained FACS data seen in Figure 9. Both the mutated DeHa 2 and DeHa 5 variants show an overall increase in fluorescence compared to the original DeHa 2 and DeHa 5, whereas the fluorescence of the mutated DeHa 1 and DeHa 4 variants does not show an overall increase in fluorescence. Based on Figure 10 it can be concluded that most of the DeHa 2 and DeHa 5 mutant variants express an increased amount of sfGFP compared to the original DeHa 2 and DeHa 5. Since the expression of sfGFP comes from an increased amount of fluoride, which activates the riboswitch, it can be assumed that the increase in fluorescence is due to higher efficiency of the mutated DeHas. Whether the enzymes show the same increase in efficiency in vitro making them applicable still needs further investigation.

Insertion and sequencing of the mutant DeHa variants

Sequencing of the selected mutant variants confirmed the insertion of 7 DeHa 1 variants, 9 DeHa 2 variants, 7 DeHa 4 variants, and 7 DeHa 5 variants. When taking a closer look at the translations of the mutant variants it could be seen that a lot of the mutant variants had mutations that led to one or several stop-codons in the amino acid sequence. The amino acid sequences of the amino acids were aligned to see if there were any locations where several variants have mutations to determine sites of the enzymes that most likely are essential for the activity of the enzyme. Figure 11 shows an alignment of the DeHa 1 variants, where the stop codon in DeHa1_var1 and the regions with mutations are highlighted.

fig11-alignmentDeHa1
Fig. 11: Alignment of DeHa1 variant amino acid sequences compared to the original DeHa1. The blue colors indicate the identity percentage, and the red highlights mark the stop codon and positions where multiple variants have mutations.

The same alignment was done for the DeHa 2 variants (Figure 12). There were however no DeHa 2 variants that didn’t have a stop codon. Furthermore, DeHa2_var3 and DeHa2_var4 had frameshift mutations, which is why they were excluded from the alignment. The frameshift mutations might however still have had a gain-of-function impact on the enzyme structure, since the fluorescence experiment showed high fluorescence in DeHa2_var4. All of the stop codons and positions where multiple variants have mutations are highlighted in Fig. 12

fig12-alignmentDeHa2
Fig. 12: Alignment of DeHa2 variant amino acid sequences compared to the original DeHa 2. The blue colors indicate the identity percentage, and the red highlights mark the stop codons and positions where multiple variants have mutations.

In the alignment of DeHa 4 and its variants (Figure 13), multiple variants had frameshift mutations and those were excluded from the alignment. Since there also was not seen an increase in fluorescence of the mutated DeHa 4 variants compared to the original DeHa 4 the other DeHas were prioritized for further analysis.

fig13-alignmentDeHa4
Fig. 13: Alignment of DeHa4 variant amino acid sequences compared to the original DeHa 4. The blue colors indicate the identity percentage, and the red highlight marks the stop codon.

Two of the DeHa 5 mutant variants had frameshift mutations which is why they were excluded from the alignment (figure 14). Furthermore, the overall mutation rate was lower compared to the other DeHas. The DeHa 5 mutant variants did however show the biggest increase in fluorescence compared to the original DeHa 5.

fig14-alignmentDeHa5
Fig. 14: Alignment of DeHa5 variant amino acid sequences compared to the original DeHa5. The blue colors indicate the identity percentage.

Though the sequencing results show many variants containing one or more stop codons it does not necessarily mean that the mutated variants DeHas will not be fully transcribed or functional as enzymes. For example, it has been reported that bacteria (such as E. coli) have varying UGA readthrough, which means that bacteria can in certain situations transcribe past a stop codon(2).

Other studies supported varying efficiencies for E. coli to terminate transcription. Furthermore, another study suggested that the base pair following the stop codon dictates whether transcription halts or proceeds3. Other studies suggest that the stop codon is nothing more than an indication of where to terminate the signal and there are many other factors needed for transcription termination4.

There is also a possibility that the enzymes could be transcribed until the stop codon, and folded, while maintaining their enzymatic pocket and thereby still be functional enzymes. Therefore, the stop codons are not to be seen as determining factors for the possible further investigations of the mutated variants enzymes. Additionally, the sequencing results revealed the cells showing the most fluorescent when running them in FACS, had in fact mutations forming stop codons in their sequence.

Additionally, docking was used to simulate the binding of PFOA to the mutated DeHa variants. See more on this link.

Protein purification

For testing the DeHas in a cell-free system they had to be purified. First, there had to be proven DeHa expression. This was done by purifying them using a His-tag purification kit (Cube Biotech 80101) and then running the purified proteins on an SDS page (Fig. 15). DeHa 1 and DeHa 5 were run on a gel together and DeHa 2 and 4 on another gel, as seen on figures 15. As expected, the bands for Deha 4 and Deha 5 are visible at ~34 kDa and ~25 kDa respectively. Conversely, for Deha 1 and DeHa 2, there were no clear bands at the expected size of 26 kDa. The expected protein sizes can be seen in table 1. Likely the enzymes were nonetheless expressed, but the purification protocol remains to be optimized for these enzymes, as they neigher could be found in the flow-through or wash fractions of their respective purification (data not published).

If there was more time at disposal, it would have been possible to optimize the purification process using different temperatures and different elution solutions for instance.

Table 1) Overview of molecular weight of RFP and dehalogenases 1, 2, 4, and 5, both with and without a 6x His-tag.
Mol. weight (kDa) Mol. weight w/ 6x His-tag (kDa)
RFP 25.30 -
DeHa 1 25.69 26.53
DeHa 2 25.51 26.35
DeHa 4 33.6 34.44
DeHa 5 24.14 24.98
fig15-deha-1-5-2-gel fig15-deha-2-2-4-gel
Fig. 15: 2 SDS-page results from purified samples of original DeHa 1, 5 and 2 (left) and DeHa 2 and 4 (right). For size reference, the protein ladder PageRuler™ Plus Prestained Protein Ladder, 10 to 250 kDa (Cat. No. 26619). Each dehalogenase was run in duplicates of 5, corresponding to each elution step of the protein purification. This was done to investigate which elution step yielded the most protein. On the left image, no bands show up for DeHa 1, nor for DeHa 2. DeHa 5 yields protein in each elution step. On the right image, DeHa 2 shows several bands, but not around 25 kDa, as expected. Duplicates 3-5 of DeHa 4 shows strong bands in the range between 25 kDa and 35 kDa.

F-NMR

Fluorine-19 nuclear magnetic resonance spectroscopy (19F-NMR) was used to prove the enzymes did break down PFOA. For this purpose, one sample was taken after 80 hours, 68 hours, and 44 hours. 19F-NMR-spectra were obtained. The spectra can be seen below, in figure 16:

fig16-FNMRspectra
Fig. 16: 97F-NMR spectra for solutions with PFOA and DeHa 4 and DeHa 5 after 80h, 68hr, and 44 hr. The spectra show no difference for the chemical shift or integral height. This indicates the enzymes did not work.

Unfortunately, there were no differences in the spectra when they were compared to the samples with PFOA. This indicated that the enzymes did not work on PFOA and therefore no C-F bonds were broken. This could have been due to the low concentration of enzyme (5 µg/mL), or the enzyme unfolded before they were able to interact with PFOA. One explanation could be that the enzymes have a degradation rate bigger than 80 hours. Additionally, temperature optimization could be investigated. The protocol for the used kit for protein purification could likewise be adjusted.

Model

next-page-button

Supplementary results

Adaptive Laboratory Evolution

If you are interested in the Whole Genome Sequencing results for the ALE1 and ALE2 strains, please contact: sofielmoebes@gmail.com

Prior to the plasmid assembly it was hypothesized that it would benefit the project if the bacteria could tolerate an increased amount of fluoride, as fluoride ions are released during the enzymatic degradation of PFOA. So, the first idea of the project was that the bacteria could be integrated into a closed system that could work as a live PFAS degradation filter. The aim was to enable the bacteria to tolerate a NaF tolerance of 200 mM.

Therefore, a side project was initiated aimed to create an E. coli with increased fluoride tolerance. However, the project and the corresponding results did not meet the criteria, and the side project was discontinued.

We aimed to increase this tolerance to 200 mM NaF using adaptive laboratory evolution (ALE). In the first ALE experiment, the bacteria were cultivated in a 96-well plate containing LB media with different NaF concentrations. Over a 4-week period, we inoculated a new 96-well plate containing a NaF dilution row with an aliquot of the well from the day before. We will refer from here on to the strain obtained through this ALE as ALE1. We attempted another ALE-experiment in which we grew the bacteria on LB-agar plates containing NaF. Each day, we streak-plated colonies onto fresh plates a higher NaF concentration than the day before. We refer to the mutant strain obtained from this experiment as ALE2 from here onwards. For an overview of the experimental setup, please refer to figure 9S.

ale-experiment-set-up
Fig. 9S: Figure showing the experimental setup for ALE1 and ALE2. ALE1 was created using liguid culture while ALE2 was created using solid culture. Created with biorender.com

To conclude the experiment, the growth of both ALE1 and ALE2 on different concentrations of NaF, ranging from 0 mM to 400 mM, was compared to that of the original Rosetta (Fig. 10S). The three different strains were plated out in a 1:10 serial dilution from a factor 1 to a factor of 10-7, on LB agar plates containing 0 mM, 50 mM, 100 mM, 150 mM, 200 mM, 250 mM, 300 mM, 350 mM, and 400 mM NaF. Overall, there are no remarkable differences in bacterial growth between the strains on plates with low (0 mM and 50 mM) NaF. On plates which contain 200 mM NaF a distinct difference in growth could be observed between the two strains. ALE1 does not show any clear difference in growth to the original Rosetta strain, with both only showing growth down to a dilution of 10-2 at a concentration of 200 mM. ALE2, conversely consistently grew a clear lawn down to a dilution of 10-3 at a concentration of 200 mM.

10s-left 10s-right
Fig. 10S:Final comparison experiment of growth of ALE1 & Rosetta (left), and ALE2 & Rosetta (right). Both ALE1 and Rosetta shows growth down to a culture dilution of 10-2 at a concentration of 200 mM. ALE2, conversely consistently grew a clear lawn down to a concentration of 10-3 at a concentration of 200 mM.

At concentrations above 350mM NaF, none of the cells grew.

At conclusion the ALE strains were not suitable for the purpose of this specific study. They, however, exhibit significant potential in studies that rely on the tolerance of E. coli to fluoride.

From the results above, it can be concluded that ALE2 does have increased fluoride tolerance, compared to the original Rosetta strain. Unfortunately, fluorescence microscopy images show that ALE1 and ALE2 are unsuitable for selection of optimized dehalogenases. Please see figure 11S for comparison of fluorescence of ALE1, NEB and original Rosetta.

F-microscope
Fig. 11S: Fluorescence microscopy images of ALE1 (with increased NaF-tolerance), NEB C3040I and original Rosetta, at 0, 50, 100 and 150mM NaF. The image shows that the GFP-expression of both NEB C3040I and Rosetta cells increases with the NaF-concentration, whereas the GFP-expression in ALE1 does not. A morphology change in NEB C3040I cells, especially, can be observed at 150mM NaF, indicating cellular stress.

  1. Thavarajah, W., Silverman, A. D., Verosloff, M. S., Kelley-Loughnane, N., Jewett, M. C., & Lucks, J. B. (2020). Point-of-Use Detection of Environmental Fluoride via a Cell-Free Riboswitch-Based Biosensor. ACS Synth Biol, 9(1), 10-18. https://doi.org/10.1021/acssynbio.9b00347
  2. Fan, Y., Evans, C. R., Barber, K. W., Banerjee, K., Weiss, K. J., Margolin, W., Igoshin, O. A., Rinehart, J., & Ling, J. (2017). Heterogeneity of Stop Codon Readthrough in Single Bacterial Cells and Implications for Population Fitness. Mol Cell, 67(5), 826-836.e825. https://doi.org/10.1016/j.molcel.2017.07.010
  3. Poole, E. S., Brown, C. M., & Tate, W. P. (1995). The identity of the base following the stop codon determines the efficiency of in vivo translational termination in Escherichia coli. Embo j, 14(1), 151-158. https://doi.org/10.1002/j.1460-2075.1995.tb06985.x
  4. Tate, W. P., Poole, E. S., Dalphin, M. E., Major, L. L., Crawford, D. J., & Mannering, S. A. (1996). The translational stop signal: codon with a context, or extended factor recognition element? Biochimie, 78(11-12), 945-952. https://doi.org/10.1016/s0300-9084(97)86716-8