Journal of Proteomics & Bioinformatics
Open Access

Flavobacterium johnsoniae Tyrosine Ammonia Lyase (Fjtal) In-Silico Structure Prediction and Molecular Docking to L-Tyrosine, P-Coumaric Acid (pHCA) and Caffeic Acid

Seyyed Amirrez Mousavi Majda^{* *}Correspondence: Seyyed Amirrez Mousavi Majda, Department of Biotechnology, College of Science, University of Tehran, Tehran, Iran, Email:

Author info »

Introduction

In the recent methods that have been developed for the sustainable production of phenylpropanoids in recombinant E. coli, FjTAL has been proposed as one of the potential enzymes to convert L-tyrosine to p-coumaric acid (Figure 1) [1-3].

Figure 1: The artificial biosynthetic pathway developed by Rodrigues et al. (2015)2 and Haslinger and Prather (2020)1 and the structure of related metabolites at pH=7.4, TAL: Tyrosine Ammonia Lyase, Ex: FjTAL, RgTAL, SeSam8, RsTAL, HaTAL. *Many different enzymes are utilized for the next step of the artificial pathway including CYP199A2, 4-coumarate 3-hydroxylase, hydroxyphenylacetate 3-hydroxylase.

Although it has been demonstrated that Flavobacterium johnsoniae Tyrosine Ammonia Lyase (FjTAL), has satisfactory affinity for L-tyrosine (Km~6.7 μM), little has been reported on the binding affinity of other structurally-related metabolites to FjTAL [4,5]. It has been hypothesized by Sariaslani (2007) [6] and Haslinger and Prather (2020) that p-coumaric acid, the product of elimination reaction catalyzed by FjTAL, might inhibit the activity of the enzyme, exerting a kind of negative feedback response. The goal of this study is to computationally predict the 3D structure of FjTAL, paving the way to help other researchers to reasonably engineer the protein for the future studies. It is worth mentioning that since not even a single experimental data on the structure of the enzyme in Protein Data Bank (PDB) was available, one had to only rely on computational methods in order to determine the structure of the enzyme.

The in-silico experiment design

The general strategy devised in this study is depicted in Figure 2. To predict the 3D structure of FjTAL, the sequence of the enzyme had to be retrieved. The codon-optimized DNA sequence of FjTAL was retrieved from the supplementary file of Haslinger and Prather (2020) [1], and translated it into amino acid alphabet by using the Biostrings package [7]. It was also observed that a record had already existed in NCBI with the exact sequence (SCV44818.1 unnamed protein product).With the FjTAL protein sequence available, four webservers were employed in order to predict tertiary structure of FjTAL. Two of these webservers, SWISS-MODEL [8-12] and Phyre212, implement homology modelling algorithms while the other two, AlphaFold2 [13-15] and RoseTTAFold [16], predict the structures without the aid of template (ab-initio prediction). FjTAL protein sequence was fed into each of these four servers. Then, the quality of these structures was validated using the SAVES of UCLA [17-19] and the ones that did not meet the standards were eliminated. Eventually, the top five pdb files which were found to be superior among all, were stored as pdb files for the next steps. After that, the secondary structures of the predicted models were extracted from the pdb files and secondary structure prediction methods that were only dependent on the input sequence rather than tertiary structure (Figure 2).

Figure 2: The workflow carried out in this study.

Materials and Methods

TIn this study quite a few computational tools were employed to help us fulfill the computational experiment. In brief, the workflow started with employing Biostring package to retrieve the sequence of FjTAL. Then the tertiary structure prediction was done by using Phyre2, SWISS-MODEL as well as ColabFold and Robetta. SAVES of UCLA was helpful in ranking the predicted structures. The molecular docking task (see section 3.3) was executed by CB-Dock and the results were subsequently analyzed by ChimeraX, Mol* and BIOVIA. The complete list of the software and webservers is given in Table 1.

Table 1: The list of computational tools used throughout this survey. Mol* is used to determine the residues that constituted the binding pocket of FjTAL in each of these complexes.

Results and Discussion

FjTAL structure prediction and quality assesment

As mentioned before, four webservers were employed to predict the structure of FjTAL. Phyre2, SWISS-MODEL, ColabFold and Robetta generate one, two, five and five structures respectively. It is worth noting that ColabFold initially generates five unrelaxed structures and five AMBER-relaxed structures which we used only the latter for the purposes of this study. SWISS-MODEL also predicted that FjTAL is likely to be either a homotetrameric or a homodimeric protein. However, most of the closely-related homologs of FjTAL were proven to be homotetrameric and thus, we have concluded this may also be a valid assumption for FjTAL as well.

All thirteen pdb files were passed to SAVES webserver to assess the quality of the predicted structures. The ideal structures had to pass Verify3D test and achieve a score<>95 in ERRAT program (top five structures were selected) (Table 2).

Table 2: The results of quality assessment using SAVES. FjTAL_Final_Phyre2.pdb is the file generated on Phyre2 server. FjTAL_X_SWISS.pdb files are the output of SWISS-MODEL. FjTAL_X_Rob.pdb files are the tertiary structures predicted by Robetta/RoseTTAFold. FjTAL_X_AF2 files are the output of ColabFold/AlphaFold2.

As shown in Table 2, ab-initio algorithms significantly outperformed Phyre2 and SWISS-MODEL. The superiority of AlphaFold2 and RoseTTAFold over classical methods had also been proposed in other studies such as CAMEO [20-22] as well. The top five structures were selected and stored.

FjTAL secondary and tertiary structure

The top models from previous sections (yellow highlights in Table 2) were passed to Stride22 to extract the secondary structures from pdb files.

We also tried to predict the secondary structures using JPred4 [23] and PROTEUS2 [24] webservers which require the protein sequence as their only input (Table 3).

Table 3: The results of secondary structure prediction and extraction.

In Table 3 the prediction results by JPred4, PROTEUS2 and the extraction results by Stride are tabulated. As shown in Table, while, to some extents, the results differ from one model to another, one can state that FjTAL is mostly consisted of alpha helices with very few, short length, beta strands.

Note that JPred4 and PROTEUS2 do not generate any data on the number of 310 helices that may exist in the structures and also notice that the number of beta strands differ significantly in each record, mainly because FjTAL seems to have very short (length=<4 amino acid) beta strands. This has led to emergence of presumably false positive results in the case of beta strands especially in the prediction-based results (Figures 3 and 4).

Figure 3: The secondary structures retrieved from FjTAL_1_AF2 as one of the top candidate structures. H: Alpha helix, G: 3-10 helix, I: PI-helix, E: Extended conformation (beta strand), B: Isolated bridge, T: Turn, C: Coil (none of the above).

Figure 4: Secondary structures predicted by PROTEUS2. (C: coil, E: beta strand, H: alpha helix).

The structure of FjTAL_1_AF2 is shown in Figure 5 as an example [25,26]. We also passed the pdb files to BIOVIA Discovery Studio in order to calculate RMSD, as a measure of similarity among the structures. Note that there exists a significant difference (ΔRMSD>2 Angstrom) in the structures predicted by Robetta/ RoseTTAFold and ColabFold/AlphaFold2. Their difference mainly stems from the coils and turns rather than the alpha helices. Without the availability of an experimentally-verified 3D structure for the enzyme, such workflow was the only logical way one could turn into to find out more about the structure of FjTAL and its relevance to its function (Table 4 and Figure 6).

Figure 5: Ribbon visualization of FjTAL_1_AF2.pdb as one of the candidate structures for FjTAL.

Table 4: RMSD values for each pair of models. We have arbitrarily taken FjTAL_1_AF2 as the reference model.

Figure 6: The superimposed structure of FjTAL.

FjTAL molecular docking to L-tyrosine, P-coumaric acid and caffeic acid

With the availability of the predicted structure of FjTAL, a molecular docking task was executed. The structures of ligands i.e., L-tyrosine, p-coumaric acid and caffeic acid were retrieved from ChemSpider and their structures were subsequently adjusted [27-30] to pH=7.4. Then, an MM2 minimization task was performed to retrieve the best energetically-favored representation of the ligands. The five receptor pdb files were docked with three potential ligands, L-tyrosine, p-coumaric acid and caffeic acid via the vina 31-based docking server, CB-Dock [31-33]. CB-Dock reports five cavities and their corresponding vina score for each pair of ligand and receptor. The top results, all of which possessing a vina score=<-5.5, were selected for further investigations.

We used mol*[34] to determine the residues that constituted the binding pocket of FjTAL in each of these complexes. Interestingly two distinct binding sites, named after their first residues as VQAY and VIY cavities, consistently appeared in most of the structures (Figures 7-9).

Figure 7: The predicted binding site of pH-adjusted, MM2 minimized, caffeic acid and FJTAL_1_AF2.pdb with vina score = -5.8. The blue residues depict the VQAY pocket, which is comprised of less evolutionary conserved residues.

Figure 8: The predicted binding site of pH-adjusted, MM2 minimized, caffeic acid and FjTAL_1_AF2.pdb with vina score = -5.9. The red residues depict the VIY pocket, consisted of many evolutionary conserved residues.

Figure 9: The molecular docking top results and their corresponding binding cavities. VQAY cavity is in blue and VIY cavity is in red. Other colors represent rare and presumably false binding sites.

Some of the residues that constitute the VIY cavity, are known to form the prosthetic group, 3,5-dihydro-5-methylidene-4H- imidazol-4-one (MIO), which has been proved to act as an electrophilic species in the deamination reaction catalyzed by Phenylalanine Ammonia Lyase (PAL) [35]. VIY cavity homologs also appeared in the experimentally-verified structure of Rhodobacter sphaeroides Tyrosine Ammonia Lyase, RsTAL in complex with caffeate and coumarate [36]. The VQAY cavity exhibits very few evolutionary-conserved residues, suggesting that it is either completely an artifact or it could be another binding site to the ligands as well as the VIY cavity. Site directed mutagenesis studies can help us further illustrate the importance of VQAY cavity. The molecular docking results demonstrate that pHCA and caffeic acid bind to FjTAL in the same pocket as L-tyrosine, supporting the idea that these two may competitively inhibit the activity of FjTAL, and further result in a negative feedback response [6].

Conclusion

In this study, the 3D structure of Flavobacterium johnsoniae Tyrosine Ammonia Lyase (FjTAL) has been predicted by employing the state of art computational tools. The candidate structures were filtered and subsequently docked with three potential ligands, L-tyrosine, p-coumaric acid and caffeic acid. The results indicated that two potential binding cavities may exist on the surface of FjTAL, one of which exhibited great sequence identity with the experimentally-verified docking site of RsTAL. The other detected binding site may be a false positive result or a second binding site for the ligands. Eventually, site directed mutagenesis studies and X-ray crystallography data would be extremely beneficial to help us understand the structure of FjTAL comprehensively.

References

1. Haslinger K, Prather KL. Heterologous caffeic acid biosynthesis in Escherichia coli is affected by choice of tyrosine ammonia lyase and redox partners for bacterial Cytochrome P450. Microb Cell Fact. 2020;19(1):1-2.

   [Cross ref] [Google Scholar][PubMed].

2. Rodrigues JL, Araújo RG, Prather KL, Kluskens LD, Rodrigues LR. Heterologous production of caffeic acid from tyrosine in Escherichia coli. Enzyme Microb Technol. 2015;71:36-44.

   [Cross ref] [Google Scholar][PubMed] .

3. Cousins KR. ChemDraw Ultra 9.0. CambridgeSoft, 100 CambridgePark Drive, Cambridge, MA 02140. www. cambridgesoft. com. See Web site for pricing options. J Am Chem Soc. 2005;127: 4115–4116.

   [Cross ref] [Google Scholar].

4. Jendresen CB, Stahlhut SG, Li M, Gaspar P, Siedler S, Förster J, et al. Highly active and specific tyrosine ammonia-lyases from diverse origins enable enhanced production of aromatic compounds in bacteria and Saccharomyces cerevisiae. Appl Environ Microbiol. 2015;81(13):4458-76.

   [Cross ref] [Google Scholar][PubMed].

5. Xue Z, McCluskey M, Cantera K, Ben-Bassat A, Sariaslani FS, Huang L. Improved production of p-hydroxycinnamic acid from tyrosine using a novel thermostable phenylalanine/tyrosine ammonia lyase enzyme. Enzyme Microb Technol. 2007;42(1):58-64.

   [Cross ref] [Google Scholar].

6. Sariaslani FS. Development of a combined biological and chemical process for production of industrial aromatics from renewable resources. Annu Rev Microbiol. 2007;61:51-69.

   [Cross ref] [Google Scholar][PubMed].

7. Pagès H, Aboyoun P, Gentleman R, DebRoy S. Biostrings: Efficient Manipulation of Biological 424 Strings. Bioconductor version: Release (3.10).

   [Google Scholar].

8. Benkert P, Biasini M, Schwede T. Toward the estimation of the absolute quality of individual protein structure models. Bioinformatics. 2011;27(3):343-50.

   [Cross ref] [Google Scholar][PubMed].

9. Mariani V, Biasini M, Barbato A, Schwede T. lDDT: A local superposition-free score for comparing protein structures and models using distance difference tests. Bioinformatics. 2013;29(21):2722-8.

   [Cross ref] [Google Scholar][PubMed].

10. Waterhouse A, Bertoni M, Bienert S, Studer G, Tauriello G, Gumienny R, et al. SWISS-MODEL: homology modelling of protein structures and complexes. Nucleic Acids Res. 2018 ;46(W1):W296-303.

    [Cross ref] [Google Scholar][PubMed].

11. Bienert S, Waterhouse A, de Beer TA, Tauriello G, Studer G, Bordoli L, et al. The SWISS-MODEL Repository—new features and functionality. Nucleic Acids Res. 2017;45(D1):D313-9.

    [Cross ref] [Google Scholar][PubMed].

12. Kelley LA, Mezulis S, Yates CM, Wass MN, Sternberg MJ. The Phyre2 web portal for protein modeling, prediction and analysis. Nat Protoc.2015;10(6):845-58.

    [Cross ref] [Google Scholar][PubMed].

13. Mirdita M, Schütze K, Moriwaki Y, Heo L, Ovchinnikov S, Steinegger M. ColabFold-Making protein folding accessible to all.bioRxiv.2021

    [Cross ref] [Google Scholar].

14. Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583-9.

    [Cross ref] [Google Scholar][PubMed].

15. Case DA, Aktulga HM, Belfon K, Ben-Shalom I, Brozell SR, Cerutti DS, et al. Amber 2021. University of California, San Francisco; 2021.

    [Google Scholar].

16. Baek M, DiMaio F, Anishchenko I, Dauparas J, Ovchinnikov S, Lee GR, et al. Accurate prediction of protein structures and interactions using a three-track neural network. Science. 2021;373(6557):871-6.

    [Cross ref] [Google Scholar][PubMed].

17. Lüthy R, Bowie JU, Eisenberg D. Assessment of protein models with three-dimensional profiles. Nature. 1992;356(6364):83-5.

    [Cross ref] [Google Scholar][PubMed].

18. Bowie JU, Lüthy R, Eisenberg D. A method to identify protein sequences that fold into a known three-dimensional structure. Science. 1991;253(5016):164-70.

    [Cross ref] [Google Scholar][PubMed].

19. Colovos C, Yeates TO. Verification of protein structures: Patterns of nonbonded atomic interactions. Protein Sci. 1993;2(9):1511-9.

    [Cross ref] [Google Scholar][PubMed].

20. Haas J, Gumienny R, Barbato A, Ackermann F, Tauriello G, Bertoni M, et al. Introducing “best single template” models as reference baseline for the Continuous Automated Model Evaluation (CAMEO). Proteins Struct Funct Bioinforma. 2019 ;87(12):1378-87.

    [Cross ref] [Google Scholar][PubMed].

21. Robin X, Haas J, Gumienny R, Smolinski A, Tauriello G, Schwede T. Continuous Automated Model EvaluatiOn (CAMEO)—Perspectives on the future of fully automated evaluation of structure prediction methods. Proteins Struct Funct Bioinforma.2021;89(12):1977-86.

    [Cross ref] [Google Scholar][PubMed].

22. Frishman D, Argos P. Knowledge‐based protein secondary structure assignment. Proteins Struct. Funct. Bioinforma. 1995;23(4):566-79.

    [Cross ref] [Google Scholar][PubMed].

23. Drozdetskiy A, Cole C, Procter J, Barton GJ. JPred4: A protein secondary structure prediction server. Nucleic Acids Res.  2015;43(W1):W389-94.

    [Cross ref] [Google Scholar][PubMed].

24. Montgomerie S, Cruz JA, Shrivastava S, Arndt D, Berjanskii M, Wishart DS. PROTEUS2: A web server for comprehensive protein structure prediction and structure-based annotation. Nucleic Acids Res.  2008;36(suppl_2):W202-9.

    [Cross ref] [Google Scholar][PubMed].

25. Pettersen EF, Goddard TD, Huang CC, Meng EC, Couch GS, Croll TI, et al. UCSF ChimeraX: Structure visualization for researchers, educators, and developers. Protein Sci. 2021;30(1):70-82.

    [Cross ref] [Google Scholar][PubMed].

26. Goddard TD, Huang CC, Meng EC, Pettersen EF, Couch GS, Morris JH, et al. UCSF ChimeraX: Meeting modern challenges in visualization and analysis. Protein Sci. 2018;27(1):14-25.

    [Cross ref] [Google Scholar][PubMed].

27. O'Boyle NM, Banck M, James CA, Morley C, Vandermeersch T, Hutchison GR. Open Babel: An open chemical toolbox. J Cheminform. 2011;3(1):1-4.

    [Cross ref] [Google Scholar][PubMed].

28. O'Boyle NM, Morley C, Hutchison GR. Pybel: A Python wrapper for the OpenBabel cheminformatics toolkit. Chem Cent J. 2008;2(1):1-7.

    [Cross ref] [Google Scholar][PubMed].

29. Bray SA, Lucas X, Kumar A, Grüning BA. The ChemicalToolbox: Reproducible, user-friendly cheminformatics analysis on the Galaxy platform. J Cheminform. 2020;12(1):1-7.

    [Cross ref] [Google Scholar][PubMed].

30. Afgan E, Baker D, Batut B, Van Den Beek M, Bouvier D, Čech M, et al . The Galaxy platform for accessible, reproducible and collaborative biomedical analyses: 2018 update. Nucleic Acids Res. 2018;46(W1):W537-44.

    [Cross ref] [Google Scholar][PubMed].

31. Trott O, Olson AJ. AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem. 2010;31(2):455-61.

    [Cross ref] [Google Scholar][PubMed].

32. Liu Y, Grimm M, Dai WT, Hou MC, Xiao ZX, Cao Y. CB-Dock: A web server for cavity detection-guided protein–ligand blind docking. Acta Pharmacol Sin. 2020;41(1):138-44.

    [Cross ref] [Google Scholar][PubMed].

33. Cao Y, Li L. Improved protein–ligand binding affinity prediction by using a curvature-dependent surface-area model. Bioinformatics. 2014;30(12):1674-80.

    [Cross ref] [Google Scholar][PubMed].

34. Sehnal D, Bittrich S, Deshpande M, Svobodová R, Berka K, Bazgier V, et al. Mol* Viewer: Modern web app for 3D visualization and analysis of large biomolecular structures. Nucleic Acids Res. 2021;49(W1):W431-7.

    [Cross ref] [Google Scholar][PubMed].

35. Calabrese JC, Jordan DB, Boodhoo A, Sariaslani S, Vannelli T. Crystal structure of phenylalanine ammonia lyase: multiple helix dipoles implicated in catalysis. Biochemistry. 2004;43(36):11403-16.

    [Cross ref] [Google Scholar][PubMed].

36. Louie GV, Bowman ME, Moffitt MC, Baiga TJ, Moore BS, Noel JP. Structural determinants and modulation of substrate specificity in phenylalanine-tyrosine ammonia-lyases. Chem Biol. 2006 ;13(12):1327-38.

    [Cross ref] [Google Scholar][PubMed].