Our project was to use a biosensor to characterize the production of a compound of interest, avoiding the use of expensive and time consuming technologies like HPLC or mass spectrometry.
We wanted to produce a library of mutants from one of the five D-psicose 3-epimerases (Dpe): two from Agrobacterium tumefaciens (UniProt ID Q7CVR7 and A9CH28), one from Clostrium cellulolyticum (UniProt ID B8I944), one from Flavonifractor plautii (UniProt ID G9YVF8) and one from Pseudomonas cichorii (UniProt ID O50580). To better understand the molecular mechanisms of these enzymes, we first performed an alignment of these sequences and found a good identity score, aligning the metal binding site, the substrate binding site (DTFH motif) and the catalytic residues, all well conserved in this family of enzymes. In their crystal form, the epimerases form tetramers or dimers and each sub-unit contains a catalytic site.
Three of the D-psicose 3-epimerases had well resolved crystallographic 3D structures in the Protein Data Bank (PDB) available with different substrates [1,2,3], and the literature describing their structure and catalytic mechanisms (Table 1).
|Substrate / Organism||Agrobacterium tumefaciens||Clostridium cellulolyticum||Pseudomonas cichorii|
For the D-psicose 3-epimerases in the table 1, we compared the crystal structures with and without substrates (fructose, psicose, sorbose and tagatose) and found RMSD values of less than one angstrom, suggesting there are minimum conformational changes across species in the structure of the enzymes while binding the substrate or the product.
On this basis, we proceeded to perform docking simulations of fructose against three receptors in table 1 to characterize the different enzyme-substrate complexes, and determine which enzyme would be the best one to realize the biocatalysis of D-fructose in D-psicose.
Material and Methods
Docking simulations were performed with the software AutoDock4.2.6 .
1) We prepared the receptor and the ligand as required by the software. Since selenomethionine was not supported by the program, we converted it into methionine using Chimera, another molecular modeling software . We incorporated manganese atoms into the simulations as unbonded electrically neutral atoms. We did not include any water molecules present in the crystal structures, including those in the catalytic site.
2) We used the defaults parameters for the Lamarckian genetic algorithm that generates ligand poses in the catalytic pocket of the enzyme.
3) We analyzed visually the (fifty) poses resulting from each docking with the PyMOL visualization software.
All programs used are under educational licence or free software.
We first performed “re-docking” to ensure that our search parameters allowed us to reproduce the crystallographic structure of the substrate and obtained good results for 2HK1 (RSMD value inferior to 0.4A), but not for 3VNK (RMSD value superior to A1) for the rigid ligands in the bioactive conformation, suggesting that the presence of water in the binding site could have an important contribution to the positioning of the ligand.
So we performed a docking simulation with water incorporated in the receptor structure and were able to obtain poses much closer to the crystallographic structure (RMSDs inferior to 0.4A). When docking a flexible ligand (D-fructose or D-psicose) with all five torsions free, neither the explicit incorporation of water nor the explicit hydrated docking approach  improved the results of the docking.
We therefore performed simulations without solvent molecules using AutoDock Tools following the procedure described there  and obtained the results presented in Table 2.
|Ranks||Binding affinity (kCal/mol)||RMSD||Binding affinity (kCal/mol)||RMSD||Binding affinity (kCal/mol)||RMSD|
Analyzing the results of flexible ligand docking, we searched for poses that would reproduce the position of the Keto function and of the hydroxyl group of carbon 3 (where the epimerization occurs) of the crystallographic substrate’s pose. Based on these criteria, we found good results (poses) regardless of RMSD values for both receptors, as represented below (Figures 2, 3 and 4).
The software AutoDock4  allowed us to find a pose close for the crystallographic substrate structure, but a degeneration in the RMSD values regarding the predicted binding affinities did not allow us to compare the results for different structures.
The structure of D-fructose which is a linear chain of carbon atoms with side chain oxygens atoms, it is hard to discriminate its binding site because of its symmetry and because of intrinsic functions of the docking software that detect hydrogen bonds between atoms without noting the presence of hydrogen.
The software AutoDock4  was able to predict the right interaction site (even on larger grid).
A further implementation of the model could include a calculation of the partial charge of the Mn ion by Quantum Mechanics computations. Subsequent molecular dynamics simulations would allow us to determine minimum energy structures, as well as the free energy landscape, leading to improvement in the accuracy of the docking prediction and discrimination between different receptors and mutants.
-  Kim K, Kim HJ, Oh DK, Cha SS, Rhee S. Crystal structure of D-psicose 3-epimerase from Agrobacterium tumefaciens and its complex with true substrate D-fructose: a pivotal role of metal in catalysis, an active site for the non-phosphorylated substrate, and its conformational changes. J Mol Biol (2006) 361, 920-931.
-  Chan HC, Zhu Y, Hu Y, Ko TP, Huang CH, Ren F, Chen CC, Ma Y, Guo RT, Sun Y. Crystal structures of D-psicose 3-epimerase from Clostridium cellulolyticum H10 and its complex with ketohexose sugars. Protein Cell (2012) 3, 123-131.
-  Yoshida H, Yoshihara A, Ishii T, Izumori K, Kamitori S. X-ray structures of the Pseudomonas cichorii D-tagatose 3-epimerase mutant form C66S recognizing deoxy sugars as substrates. Appl Microbiol Biotechnol (2016) 100, 10403-10415.
-  Morris GM, Huey R, Lindstrom W, Sanner MF, Belew RK, Goodsell DS, Olson AJ. AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility. J Comput Chem (2009) 30, 2785-2791.
-  Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, Ferrin TE. UCSF Chimera--a visualization system for exploratory research and analysis. J Comput Chem (2004) 25, 1605-1612.
-  Forli S, Olson AJ. A force field with discrete displaceable waters and desolvation entropy for hydrated ligand docking. J Med Chem (2012) 55, 623-638.
-  Forli S, Huey R, Pique ME, Sanner MF, Goodsell DS, Olson AJ. Computational protein-ligand docking and virtual drug screening with the AutoDock suite. Nat Protoc (2016) 11, 905-919.
Dr. R. Charbel Maroun1
- 1 Laboratory of Structure and Activity of Normal and Pathological Biomolecules (SABNP), INSERM, Univ. Evry, University Paris-Saclay