Part:BBa_K5033003

OncoBiotica: mFadA[B]_GSLinker_CDA[Gamma]

This page is currently under construction!

If you are interested in an overview of the parts designed by the iGEM Team Aachen 2024, visit our Parts page.

This part, developed by iGEM Aachen 2024, is our 'Gamma' mutant of the basic part BBa_K5033000. Within the codA cytosine deaminase (CDA) domain, it contains a R91T (the arginine on position 91 has been exchanged with threonine) as well as a D314A mutation (the aspartic acid on position 314 has been exchanged with alanine). In the context of our fusionprotein they are R129T and D352A mutations. This mutant has been selected for research in the lab with the help of literature research and modeling.
Its main use is to see if we can improve the caltalytic activity of our fusionprotein.
It is an optimized version of BBa_K5033000 which served as the foundation for exploring the concept of microbiota-directed cancer therapy. This part encodes a fusion protein designed to combine two functionalities. Binding specific bacteria and having an optimized enzymatic function. This part is to be cloned into a vector based on an inducable expression system. iGEM Aachen 2024 used a pET21b(+) vector.
iGEM Aachen 2024 successfully demonstrated that the enzymatic activity is approximately doubled in comparison to the CDA-wild type fusionprotein. In addition this variant is expected to have a much higher relative affinity to 5-fluorocytosine over cytosine, making it the most promising candidate for a novel microbiota-directed cancer therapy. This is one of five mutants analyzed by iGEM Aachen 2024 in addition to the CDA-wild type fusionprotein.
See the other four variants: 'Alpha' (BBa_K5033001), 'Beta' (BBa_K5033002), 'Epsilon' (BBa_K5033004) and 'Theta' (BBa_K5033005).

Figure 1: Schematic view of the fusion protein's coding sequence.

Part Composition

The first protein domain is derived from the part BBa_K4990002 but has been codon optimized for expression in E. coli. It is the mFadA B-domain, found in various Fusobacterium strains. This part has already been well described by the iGEM23_CPU-CHINA team. This domain should be able to bind to FadA pili on Fusobacterium nucleatum and its former subspecies Fusobacterium nucleatum, F. polymorphum, F. vincentii, F. animalis via self assembly.
To further investigate the binding domain's functionality, iGEM Aachen 2024 created the basic part BBa_K5033006. This variant replaces the enzyme in our fusion protein with eGFP as a reporter protein.
The second functional protein domain is linked to the mFadA B-domain by a synthetic flexible linker consisting of Glycin and Serine in alternating order. This linker is eleven amino acids long.

This second functional domain of the fusionprotein is a mutant of the codA cytosine deaminase (CDA) that is native to E. coli. This mutant contains the R91T and D314A mutations of the enzyme. This Enzyme converts cytosine to uracil in its host organism but it is also able to convert 5-fluorocytosine (5-FC) to 5-fluorouracil (5-FU).[1]

In this case the Enzyme can be used for an enzyme directed prodrug therapy. To be precise, 5-FC is the non-toxic substrate and 5-FU is the active chemotherapeutic agent.
The fusionprotein encoded by this part also contains a downstream hexa-histidine tag for protein purification.

Protein Modeling

To find interesting mutations that shall be investigated in the lab, our team used a research based modeling approach.
Before transformation of this biological part (cloned into the pET21b(+) plasmid backbone), the structure of the expected fusionprotein was modeled.

Selection of the Mutant

'Gamma' (R91T/D31A) was predicted using computer assisted recombination (CompassR) to have a high folding stability (ΔΔG_fold = -4.88 kcal/mol) making it an interest candidate for our lab group. The D314A substitiution was shown to improve the enzyme efficiency realtive to the wild type enzyme [2]. Additionally, the R91T substitution improves the relative enzyme specificity to our substrate 5-FC, by reducing the affinity to cytosine significantly [3]. Thus, it is reasonable to assume that the combined mutations could potentially outperform the individual ones in terms of enzyme specificity and efficiency.

Biochemical Properties

The fundamental biochemical properties like molecular mass and extinction coefficient are important for a lot of synBio work done with proteins. To see an overview of these properties, have a look at figure 2.

Figure 2: Biochemical properties of mFadA[B]_GSLinker_CDA[Gamma] as described by Benchling.

Protein Structure Prediction

Figure 3: Tertiary structure of mFadA[B]_GSLinker_CDA[Gamma] as predicted by AlphaFold2. From left to right: mFadA[B] as an alpha helix (blue), the flexible linker, the enzyme-mutant and the freely accessible His-Tag (orange).

The tertiary structure has been predicted using AlphaFold2 by DeepMind. In this case it is especially important, that the binding domain and the His-Tag are freely available.

Modeling of Substrate and Active Site Interaction

Why RoseTTAFold All-Atom?

Proteins rarely act alone. Although substantial progress in the prediction of protein structures has been made, modeling of proteins and their ligands still remains challenging. The development of RoseTTAFold All-Atom (RFAA) aims to tackle this issue by building a neural network that is trained to accurately model general biomolecules containing a wide range of nonprotein components. In contrast to other tools that only include sequence based modeling, RFAA incorporates a graphical representation that models non-protein molecules at the atomic level, capturing their chemical bonds and interactions. In combination with the training data set that also includes ligand-bound protein structures from the Protein Data Bank (pdb), it allows RFAA to predict protein structures, ions and non-protein ligands. Interestingly, during our project, DeepMind released a new AlphaFold version (v3) that includes selected ions and ligands. However, an earlier release would not have been advantageous for us, as 5-FC and cytosine are not among the selected ligands that AlphaFold3 includes. Nevertheless, this shows that the improvements made this year mark a significant step forward, paving the way for more refined and accurate modeling of proteins and ligands in the future.

We observed a good overall structural alignment of the wild type enzymes' crystal structure [4] to the RoseTTAFold All-Atom model. Upon closer inspection of the active site, we noticed small differences in torsion angles of the side chains which naturally led to slight differences in bond lengths between amino acids and the ligand. However, these differences are inherent to the modeling process and do not reflect significant deviations. Therefore, they do not compromise the reliability of our approach for predicting structural changes in the mutants. This allowed us to apply the approach to the generated mutants by CompassR.

Modeling Results

Figure 4: Model of interactions of the amino acids in the active site of the fusionprotein with 5-fluorocytosine. As predicted using RoseTTAFold All-Atom.

Looking at the model generated by RFAA, it can be observed that the structure of the active center of the mutant with 5-FC appears to be similar of the D314A ('Alpha' (BBa_K5033001)) and F186W/D314A ('Beta' (BBa_K5033002)) mutants. The methyl group of A314 is at similar distance to the fluorine atom (3.772 Å). Additionally, T91 is not present in the direct vicinity of the active center but close to it on an alpha-helix. Arginine bears a positively charged side chain that is substituted with a polar uncharged side chain of threonine. Thus, the substitution could have minor secondary effects on the structure of the active center that are not visible through our modeling approach. In contrast to the WT-structure, the model of R91T/D314A complexed with cytosine exhibits ignificant deviations. Most notably, the orientation of cytosine is substantially altered compared to the WT. This reorientation leads to different interactions between the amino acids in the active center and the ligand. The cytosine appears to be rotaded by approximately 100°. This results in the planar sructure of the pyrimidine ring to be no longer arranged parallely to W319 but facing towards it, causing a sterical hindrance. Furthermore, D313 is positioned further away from where the reaction takes place (the oxygen of cytosine) than in the WT . Additionally, the iron atom is displaced between D313 and the susbtrate. Moreover, it seems that cytosine is shifted towards H63, H61 and H214. This results in additional hydrogen bonds to the substrate. Generally, the active center appears to be very crowded with new bonds that were not present in the WT. All these effects could lead to a reduced efficiency and specificity towards cytosine. The R91T mutation alone exhibits specificity and catalytic efficiency towards cytosine and 5-FC comparable to that of the WT enzyme [6]. When R91T is combined with D314A, the resulting active site structure with 5-FC resembles that of the D314A single mutant. However, the structure with cytosine shows significant deviations to the WT and D314A. This suggests that the R91T/D314A double mutant may exhibit reduced specificity for cytosine, making it an interesting candidate for further experimental validation in the lab.

Cloning of the Plasmid

To build the plasmid containing the gene for our Gamma variant. We used the plasmid we already had for our WT-Fusionprotein (BBa_K5033000; pET21b(+)_mFad[A]_GSLinker_CDA[WT]). The gene sequence for this part contains a BamHI restriction site between the linker and the enzyme. The backbone contains a XhoI restriction site at the end of the gene insert.
After modeling of the Gamma variant we ordered the gene fragment, encoding this variant. We made sure to include the correct restriction sites.

The backbone was prepared using the BamHI and XhoI restriction enzymes. After digestion, the cut backbone was cleaned up using an agarose gel and a gel extraction kit. The same was done for the Insert.

After gel cleanup the cut backbone and insert were ligated using the T4 Ligase.

To enhance the efficiency of the plasmid transformation into E. coli BL21 (DE3) the plasmid was first propagated via transformation in E. coli DH5α.

The propagated pET21b(+)_mFadA[B]_GSLinker_CDA[Gamma] plasmid could then be purified with a plasmid miniprep kit and used for transformation into the production organism E. coli BL21 (DE3).

Producing the Fusionprotein

After successful transformation of the pET21b(+)_mFadA[B]_GSLinker_CDA[Gamma] plasmid into the production organism E. coli BL21 (DE3) the protein could be expressed and purified. The pET21b(+) backbone has a lac operon (including the lacI repressor), which can be induced with IPTG (IUPAC: Propan-2-yl 1-thio-β-D-galactopyranoside).

Expression and Purification of the Fusionprotein

The fusionprotein was expressed by adding IPTG to the medium to a final concentration of 1mM.
The His-tagged protein was purified using a Protino Ni-IDA 2000 packed column by Macherey & Nagel®.

Figure 5: SDS pages showing the proteins in the elution fractions. The number corresponds to the imidazole concentration (in mM) in the elution buffer respectively. Example: E10 is an elution buffer with 10mM imidazole. There are two gel pictures combined because the fractions were put on seperate gels (the gels ran simultaniously in the same electrophoresis chamber). It was made sure that there are ladders on both gel parts.

The fusionptrotein is expected to have a molecular weight of 52.02kDA (cf. Fig. 2). This corresponds to the big bands visible on the gel.
This gel shows that the E10 fraction still has a lot of impurities. The E50 fraction was desalted and stored in 50mM TRIS buffer, to use for the kinetic assays.

Kinetic Assays

If you are interested in the methods used, take a look at our Experiments page.

UV/Vis

An assay using a spectrometer was devised. This measures the peak of the absorbance spectrum, which shifts to lower wavelengths as the ratio of 5-FU to 5-FC increases. We established a linear relationship, which we calibrated using standards. More details can be found in the UV-Vis Peak Shift Assay protocol on our Experiments page.
Note that this method does not give a high resolution beyond 5% 5-FU/5-FC intervals.

High-Performance Liquid Chromatographie (HPLC)

We used Reverse Phase High Performance Liquid Chromatography for quantitative Analysis of 5-fluorocytosine and 5-fluorouracil in mutual solution. The results seen below were all measured with the same method. (see Experiments page)
Standards at between 10 µM and 500 µM were made to translate the peak area into compound concentration. After measuring, the chromatograms were evaluated with “OpenChrom” by Lablicate. For this, a baseline subtraction filter was applied, after this the standard first derivitave peak detector and trapezoid peak integrator were run. We identified para-aminobenzoicacid as a potential internal standard, but no problems which would necessitate the use of an internal standard arose.

Figure 6: Example HPLC Chromatogram obtained from measurement using 400 µM 5-FC and 200 µM 5-FU; the first peak being the 5-FC, the second 5-FU.

¹⁹F-NMR

To get insights into the enzyme's kinetic, NMR (nuclear magnetic resonance) experiments, especially ¹⁹F-NMR, were performed for the wild type fusionprotein and the gamma mutant.

Reaction monitoring of the Gamma mutant

Figure 7 shows exemplarly the third ¹⁹F-NMR spectrum of the Gamma mutants's examination which was taken 25 min after the reaction's start.

Figure 7: Third ¹⁹F-NMR spectrum of the Gamma mutants's examination.

The singlet signals at δ = -167.98 ppm and δ = -169.14 ppm can be assigned to 5-fluorocytosine and 5-fluorouracil, respectively. The compound causing the signal at δ = -119.77 ppm is not identified as mentioned above. It can be stated that the signals have the same chemical shift δ as the signals in the examination of the wild type. Here, the integrals for 5-FC and 5-FU are I_5-FC = 358.9E5 and I_5-FU = 84.22E5 but if they are compared to the wild type's integrals the magnitude of the mutant's integrals differs from the wild type measurement. This is not a sign for lower amounts in the reaction mixture because the integrals can just compared within one measurement. Also in this case, the evolution of the integral values are according to our expectations as mentioned above.

The spectra for the Gamma mutant can be evaluated efficiently when they are stacked (cf. Fig. 8).

Figure 8: Stacked ¹⁹F-NMR spectra of the Gamma mutants's examination. ¹⁹F-NMR spectrum of the Gamma mutants's examination.

To obtain the integrals, the range was set for 5-FC as δ = [-167.901,-168.101] and for 5-FU as δ = [-169.102,-169.300]. The blue curve which represents the integral of 5-FC's signal is constantly decreasing from 366.6E5 at the beginning to 69.34E5 at the end which is according to the expectation because the substrate is consumed by the enzyme and therefore the amount of it in the reaction mixture decreases. The shape of this curve has the same shape as the substrate's curve for the wild type so the same justification can be applied here. As recognized in the wild typ's measurement before, there are some outliers (e.g. for the spectrum measured at t = 1036 min) which can be explained in the same way as for the wild type.

Fig. 9 shows the evolution of the integrals of the Gamma mutant' expressed enzyme.

Figure 9: Evolution of the integrals of the gamma mutants's examination.

The blue curve represents the increasing product amount in the reaction mixture. Here, the integral value increases from 80.85E5 at the beginning to 319.7E5 at the end which is also according to the literature. The reason for the curve's behavior are stated above.

All in all, the examination of the Gamma mutant was successful since the integrals which are indicators for the substrate's and product's amount in the reaction mixture are decreasing for 5-FC and increasing for 5-FU.

Conversion rate of the reactions using the wild type and the Gamma mutant enzyme

An important quantity for reactions is the so called conversion rate X which describes how much of substrate is converted at a given time. To obtain this quantity, the concentrations for the substrate and product c_i for each time have to be computed. To get the concentrations, the amount of substance n_i need to be divided by the volume of the reaction mixture. In the next step, the conversion rate can be calculated using the following equation where c_total = 20 mM.

Enzyme Kinetics

Similar to the wild type enzyme, the mutant cytosine deaminase followed Michaelis-Menten kinetics. However, due to the limited solubility of 5-fluorocytosine (5-FC), achieving substrate concentrations sufficient for determining the mutants' maximum velocity (V_max) was difficult, leading to variability in kinetic measurements. This made the direct determination of K_m and V_max unreliable. To assess the mutants' performance, three key metrics were employed: (1) the time required to reach half-maximal product formation t_1/2RP, (2) the initial reaction rate (initial velocity) v₀, and (3) the total product formed over time as a measure of relative efficiency RFA.

For each mutant and method (HPLC, NMR, and UV/Vis Spectroscopy), relative substrate and product concentrations were plotted against time, and these graphs formed the basis of the analysis. The time points at which half-maximal product formation was reached were graphically determined, and substrate and product concentrations for each variant are plotted against time in figure 10. As with the wild type, the relative efficiency of the mutant was calculated by determining the area under the graph within a predefined time interval, assuming linearity between data points. This area represents the total product formed over time and provides a valuable comparison of efficiency of the mutant.

Figure 10: Representation of the relative amount of substance n_rel for the gamma mutant against the time t obtained from HPLC measurements.

Figure 11: Representation of the relative amount of substance n_rel for the gamma mutant obtained from NMR measurement against the time t.

The initial velocities were determined by linear approximation of relative product formation during the first 80 seconds for both the HPLC and UV/Vis methods. For NMR, where significantly lower enzyme concentrations were used, the reaction times were longer, and the initial velocities were calculated over an extended time interval. In Figure 12, the first points of the graphs for the wild type and Gamma variants are plotted against time, with the slope of the regression line representing the relative velocity for each condition. Each mutant's performance was directly compared to the wild type enzyme under these conditions.

Figure 12: Representation of the starting reaction velocity by plotting the relative amount of substance n_rel against time t containing the first 30000 s.

The relative activity of each mutant was calculated as the mean of the two key parameters: the time to half-maximal product formation and the initial velocity, with the standard deviation providing an estimate of error. This combined measure of relative activity, allowed for a direct comparison of each variant's performance (cf. fig. 13).

Figure 13: Summary of the obtained RPA, t_(1/2)RP and v₀ values from NMR spectroscopy.

However, in NMR, Gamma showed only 1.2 times the efficiency compared to the wild type, likely due to the suboptimal conditions of very low enzyme concentrations and extended reaction times. This discrepancy (to the measurement with HPLC) highlights the challenges in comparing performance across different methods, as illustrated in Figure 14, where the lower relative efficiency of Gamma is evident in the NMR assays.

Figure 14: Comparison of the relative enyzme performance for the wildtype and gamma variant; data obtained from NMR and HPLC.

To further characterize the Gamma variant, we directly compared its performance to the Alpha variant. This comparison ensured that the additional R91T mutation in Gamma did not significantly impair its catalytic activity, while potentially enhancing the relative substrate affinity to 5-FC drastically.

This detailed approach allowed us to quantify the relative activity and efficiency of each mutant, highlighting the differences in their performance compared to the wild type.

Conclusion

Our mutant variants have shown significantly increased catalytic activity compared to the wild type enzyme, a key achievement in optimizing the fusion protein for therapeutic use. When integrating our modeling results, the Gamma variant emerges as particularly promising. Not only does it exhibit catalytic activity comparable to that of the Alpha mutant, but the modeling also suggests that Gamma has a lower affinity for cytosine, potentially offering improved selectivity for 5-FC. The selective affinity for 5-FC over cytosine is critical in ensuring the enzyme's therapeutic specificity, and while current evidence for this enhanced selectivity is based on Dry Lab page predictions, it provides a strong basis for further investigation.

Looking ahead, our goal is to replicate these promising results and further characterize the behavior of the mutant enzymes. Specifically, we aim to investigate how the mutations affect both catalytic activity and substrate selectivity under various conditions. In future experiments, we will test the affinity of these mutants for cytosine, focusing on competitive substrate assays that will examine how cytosine and 5-FC compete for enzymatic binding. This is particularly important because cytosine is naturally present in the human body, and understanding this competition is crucial for optimizing the enzyme’s application in vivo. Since the used analyzing techniques are not sufficient to absolutely determine the product's structure, further analytics should be done for complete structural elucidation. For this reason we got into contact with the mass spectrometry unit of the IOC (institute of Organic Chemistry) to plan those for the future.

Finally, the modular nature of our approach offers flexibility in swapping out the enzyme for different enzyme/prodrug pairs, depending on therapeutic needs.

References

[1] Aučynaitė, A., Rutkienė, R., Tauraitė, D., Meškys, R., Urbonavičius, J., 2018. Discovery of Bacterial Deaminases That Convert 5-Fluoroisocytosine Into 5-Fluorouracil. Frontiers in Microbiology 9. https://doi.org/10.3389/fmicb.2018.02375
[2] Sheri D. Mahan, Greg C. Ireton, Catherine Knoeber, Barry L. Stoddard, Margaret E. Black, Random mutagenesis and selection of Escherichia coli cytosine deaminase for cancer gene therapy, Protein Engineering, Design and Selection, Volume 17, Issue 8, August 2004, Pages 625–633, https://doi.org/10.1093/protein/gzh074
[3] Kohila, V., Jaiswalb, A., Ghosh, S., 2012. Med. Chem. Commun., 2012,3, 1316-1322 https://doi.org/10.1039/C2MD20209C
[4] PDB Entry - 1RA0. https://doi.org/10.2210/pdb1RA0/pdb

Sequence and Features

Assembly Compatibility:

• 10
  COMPATIBLE WITH RFC[10]
• 12
  COMPATIBLE WITH RFC[12]
• 21
  INCOMPATIBLE WITH RFC[21]

  Illegal BamHI site found at 115
  Illegal XhoI site found at 1396
  
• 23
  COMPATIBLE WITH RFC[23]
• 25
  INCOMPATIBLE WITH RFC[25]

  Illegal NgoMIV site found at 1254
  Illegal NgoMIV site found at 1341
  
• 1000
  COMPATIBLE WITH RFC[1000]