# 🥇 Modeling ## Introduction ~|| We have performed various types of simulations to improve the optogenetic regulation system and optimize the vanillin biosynthesis in the *Arthrospira platensis*. We used the programs for protein visualization and molecular dynamics: Molecular Operating Environment (MOE), PyMOL, Visual Molecular Dynamics (VMD), GROMACS, and servers for protein structure prediction: SWISS-MODEL [1], trRosetta [2], pyDockWEB [3], ClusPro 2.0 [4]. In our optogenetic system gene expression occurs when the 760 nm light is active and is suppressed by dark matter, red light (640 nm) or blue light (450 nm). |  || After literature search, we have decided to create a system based on the Redchuk's research works [5]. Its key element is a bacterial phytochrome, which was previously proposed by use for regulation of gene expression by the Vilnius-Lithuania team in 2019. Redchuk's variants of optogenetic systems are adapted to work in eukaryotic cells, which needs the use of NLS. We have applied these mechanisms to build an optogenetic system, adapting it to work in prokaryotic cells. The optogenetic system consists of two photoreceptors, which are combined into one chimeric protein and regulate transcription in two ways: under the influence of blue light, the optimized BcLOV4 domain, which plays the role of a repressor, is attached to the lipid membrane by means of positive and hydrophobic amino acids of the α-helix due to conformational changes, and under the influence of red light, the activation of the BphP1 switch occurs. BphP1 needs biliverdin IXα tetrapyrrole as cofactor that has the largest electron-conjugated system. Upon 760 nm light stimulation, BphP1 passes from the inactive dimerized state (Pr) to the active monomeric state (Pfr). The binding site becomes available for BphP1's contacting compound, Q-PAS1, which was previously in a dimerized and DNA-bound state. The Gal4 transcription factor is attached to it via a peptide bond, acting as a transcription inhibitor and recognizing the sequence 5’-CGG-N11-CCG-3' in the promoter region. BphP1 dissociates Gal4 from regulatory DNA sites, which leads to the expression of vanillin biosynthesis genes. It is noteworthy that BcLOV4 in this system is needed as an additional "fuse" against accidental transcription. The action of BcLOV4 protein is based on LOV photosensory signaling which relies on a flavin chromophore bound within a Per–Arnt–Sim (PAS)-type sensory domain. BcLOV4 dynamically associates with anionic plasma membrane phospholipids by a blue light-switched electrostatic interaction. This reversible association is rapidly triggered by blue light and ceases within seconds when illumination is switched off. The truncated variant of BcLOV4, in which the regulator of G-protein signaling (RGS) domain was removed (∆1-240), was chosen for the analysis because the RGS domain inhibits the membrane interaction [6].  In summary, bacteriophytochrome BphP1 and its contacting compound Q-PAS1 from *Rhodopseudomonas palustris*, transcription factor Gal4 from *Saccharomyces cerevisiae*, and photoreceptor BcLOV4 from *Botrytis cinerea* are the main components of our optogenetic system. The following modeling tasks were completed: 1. Monomers and dimers structures of BphP1-BcLOV4, Gal4-Q-PAS1 chimeric proteins were obtained and studied. 2. We tried to optimize chimeric proteins obtained by homology analysis and energy minimization in the MOE program. The molecular dynamics analysis of truncated and full-length fusion proteins was carried out using the GROMACS program. 3. The binding of BphP1-BcLOV4 to Gal4-Q-PAS1 was analyzed using protein-protein docking performed on the ClusPro 2.0 and pyDockWEB servers. 4. We studied interaction of Gal4 (with a truncated and full-length amino acid sequences) with the promoter region of DNA. 5. The structural features of the BcLOV4 protein, allowing it to bind to the lipid membrane of *E. coli* and *Arthrospira platensis* IPPAS B-256, were analyzed. ## Preparation and primary analysis of fusion proteins ~|| First, we downloaded the BphP1, Gal4, Q-PAS1 structures from the Protein Data Bank (PDB) and used the PyMOL molecular imaging software to gain insights on their structures. However, there was no BcLOV4 crystal structure in the PDB. The most reliable way to obtain crystal structure is to build models using *ab initio* method or homology analysis. In the first case, determining of the secondary and tertiary protein structures is performed using physicochemical parameters and neural network algorithms. One of the drawbacks of this strategy is that it cannot accurately account for the orientation of amino acid side chains (see Critical Assessment of Protein Structure Prediction for more information). The *ab initio* simulation tools are trRosetta, I-TASSER. We have applied these programs to predict the BcLOV4 three-dimensional structure from its amino acid sequence obtained from the UniProt database. We took into account the modeling parameters and three-dimensional structures of BcLOV4 protein in both programs to obtain reliable and accurate data for further analysis.| ! [English])||  To analyze the stability of the chimeric proteins, we constructed their information structures based on protein sequences. The information structure principle is based on dividing the sequence into blocks of 5 amino acids and determining the frequency of each block in a non-homologous database. Each amino acid is assigned the sum of the occurrence frequencies of the five-term blocks in which it is included. Using the nonlinear smoothing procedure, we obtained the graphs of the smoothing functions φ(i, a), where *i* is the position of the amino acid, *a* is the width at half-height [7]. The y-axis indicates the amino acid positions, and the x-axis indicates the width at half-height. We found out that individual domain structures do not violate the functional integrity of the BphP1-BcLOV4, Gal4-Q-PAS1 fusion proteins.  and Gal4-Q-PAS1 chimeric protein, Gal4, Q-PAS1 domains (right)) There are templates with three-dimensional structures similar to target protein in the PDB. If an empirically determined three-dimensional structure is available for a quite similar protein (50% or better sequence identity would be good), software can be used to arrange the backbone of target sequence identically to this template. On the basis of a sequence alignment between the target protein and the template structure, a three-dimensional model for the target protein is generated. Such homology modeling shows good results when there are many similar templates. Otherwise, there is high probability of error. The structure prediction and optimization of the BphP1-BcLOV4, Gal4-Q-PAS1 chimeric proteins by homology modeling were performed using the SWISS-MODEL and MOE programs. We compared the chimeric proteins conformations with the crystallized structures of the individual domains, on the basis of which the best conformations were selected. During homology analysis, we obtained truncated structures of chimeric proteins. In the initial conformations of molecular modeling systems, molecules can have overlapping atoms or groups of atoms. As a result, the initial energies of the system can be very high, which can lead to large errors at the initial stage of the MD algorithms operation. To avoid this, the energy of the system is usually minimized before the start of MD calculations. One of the main energy minimization methods applied in molecular modeling is the method of steepest descent. We applied this method in the MOE program to test the stability of the conformation of amino acids. After energy minimization in MOE, these proteins were analyzed in further analysis. We used the PyMOL molecular imaging software to analyze the results. PyMOL allows us to generate scene-based movies of our resulting models and create better pictures due to its rendering tools. Below are BphP1-BcLOV4 and Gal4-Q-PAS1 monomers with full-length amino acid sequences (left) and shortened amino acid sequences (right). --- ~|| ! [English]) | ! [English]) || || ! [English]) | ! [English]) || --- Protein-protein docking predicts the orientation and binding of multiple proteins. There are many different programs that can be used for protein-protein docking, such as MOE and SwissDock. However, to get more accurate results, protein-protein docking was performed using the ClusPro 2.0 and pyDockWEB servers. The work of these programs is based on calculating the potential energy of each new conformation, then the structure with the lowest energy is selected as the most probable candidate. Although BphP1-BcLOV4 dimers with a truncated amino acid sequence are structurally different from BphP1-BcLOV4 dimers with a full-length amino acid sequence, in both cases the formed complexes prevent the binding of the Q-PAS1 domain. As seen, Gal4-Q-PAS1 are dimers formed by two chains with two ligand binding sites. Below are BphP1-BcLOV4 and Gal4-Q-PAS1 dimers with full-length amino acid sequences (left) and shortened amino acid sequences (right). --- ~|| ! [English]) | ! [English]) || || ! [English]) | ! [English]) || --- The binding sites of BphP1 and Q-PAS1 have been determined with crystallography and are therefore known. Because of this, we did not have to perform blind docking of fusion proteins. However, some other problems were caused by absence of crystal structure of BcLOV4, its large size and multiple torsion angles. That is, it is difficult to perform accurate protein-protein docking with bigger proteins, because of the amount of molecule orientations, so many runs with different parameters and software were needed to achieve the correct visual orientation of proteins. Although we analyzed the chimeric proteins structures computationally, in the end we did not run the experiments in the lab. We have analyzed the protein-protein interaction sites and prepared files for a more detailed molecular dynamics calculation in the GROMACS program. ## Molecular dynamics in GROMACS One of our main disciplines of simulation is molecular dynamics. We used a great variety of different programs and tools to achieve the best possible results for the simulations. Mainly, we applied GROMACS for simulation of our optogenetic systems and MD trajectory analysis. To visualize simulations, we used VMD, because this program allows us to create quality images and videos of our models and MD trajectories. To test our optogenetic system, we primarily have analyzed the Gal4-Q-PAS1 and BphP1-BcLOV4 monomers. At the same time, it was examined in the GROMACS and MOE programs to what extent these structures deviate from those obtained by homology modeling on SWISS-MODEL server. An important parameter in molecular dynamics is the root-mean-square deviation of atomic positions, RMSD for short. It is the measure of the average distance between the atoms (usually the backbone atoms) of superimposed proteins. Once we were sure that the obtained structures deviate insignificantly when calculating the molecular dynamics in water in the OPLS-AA/L force field (RMSD < 0.5 nm for 100 ps), and the monomers remain stable, we proceeded to construct dimers. Below are molecular dynamics simulations and RMSD for BphP1-BcLOV4 chimeric protein with full-length amino acid sequence (first row), Gal4-Q-PAS1 chimeric protein with full-length amino acid sequence (second row), BphP1-BcLOV4 chimeric protein with shortened amino acid sequence (third row) and Gal4-Q-PAS1 chimeric protein with shortened amino acid sequence (fourth row). --- ~|| ![https://static.igem.org/mediawiki/2021/4/41/T--LMSU--Modeling-otherBphP1mono-gif.gif] | ![https://static.igem.org/mediawiki/2021/0/04/T--LMSU--Modeling-RMSD-otherBphP1mono.png] || --- ~|| ![https://static.igem.org/mediawiki/2021/archive/3/35/20211018101545%21T--LMSU--Modeling_otherGal4mono-gif.gif] | ![https://static.igem.org/mediawiki/2021/1/12/T--LMSU--Modeling-RMSD-otherGal4mono.png] || --- ~|| ![https://static.igem.org/mediawiki/2021/archive/e/eb/20211018101620%21T--LMSU--Modeling-ourBphP1mono-gif.gif] | ![https://static.igem.org/mediawiki/2021/7/70/T--LMSU--Modeling-RMSD-ourBphP1mono.png] || --- ~|| ![https://static.igem.org/mediawiki/2021/archive/e/e9/20211018101411%21T--LMSU--Modeling-ourGal4mono-gif.gif] | ![https://static.igem.org/mediawiki/2021/5/59/T--LMSU--Modeling-RMSD-ourGal4mono.png] || --- The molecular dynamics of dimers was also calculated in the OPLS-AA/L force field. At this stage, we faced a number of difficulties: the BphP1-BcLOV4 dimer turned out to be too large, as a result of which the calculation of the energy of the system was impossible. It was decided to remove BcLOV4 domain, since it did not participate in the formation of the dimeric complex. Besides RMSD, important factor for molecular dynamics is the potential energy, which determines the forces on the atoms. The result is reliable if Epot is negative, and of the order of 10^6-10^7 for chimeric proteins in water, depending on the system size. During the energy minimization phase, the system maximum force should not exceed 1000 kJ\*mol-1\*nm-1. Calculations of molecular dynamics and interaction kinetics demonstrated stability of BphP1 complex (RMSD < 0.5 nm for 100 ps, Epot = -1.6825\*e+7 kJ\*mol-1, Etot = -1.411\*e+7 kJ\*mol-1). At the same time, the BphP1-BcLOV4 lowest total score, calculated using PyDockWEB, for this model (S) is -33.784 kJ\*mol-1. Total score is calculated based on electrostatics, desolvation energy and limited van der Waals contribution. Similar parameters were achieved for Gal4-Q-PAS1 dimer with full-length amino acid sequence: RMSD < 0.5 nm for 100 ps, Epot = -5.07\*e+6 kJ\*mol-1, Etot = -4.25\*e+6 kJ\*mol-1, S = -124.942 kJ\*mol-1. It was possible to carry out all calculations without removing parts of the dimer. Below are molecular dynamics parameters for BphP1 dimer with full-length amino acid sequnce (first graphs) and Gal4-Q-PAS1 dimer with full-length amino acid sequnce (second graphs). --- ~|| ![https://static.igem.org/mediawiki/2021/e/e1/T--LMSU--Modeling-otherBphP1dim-gif.gif] | ![https://static.igem.org/mediawiki/2021/2/22/T--LMSU--Modeling-RMSD-otherBphP1dim.png] || || ![https://static.igem.org/mediawiki/2021/3/3f/T--LMSU--Modeling-En-tot-otherBphP1dim.png] | ![https://static.igem.org/mediawiki/2021/d/db/T--LMSU--Modeling-En-pot-otherBphP1dim.png] || --- ~|| ![https://static.igem.org/mediawiki/2021/4/40/T--LMSU--Modeling-otherGal4dim-gif.gif] | ![https://static.igem.org/mediawiki/2021/5/53/T--LMSU--Modeling-RMSD-otherGal4dim.png] || || ![https://static.igem.org/mediawiki/2021/0/02/T--LMSU--Modeling-En-tot-otherGal4dim.png] | ![https://static.igem.org/mediawiki/2021/b/b7/T--LMSU--Modeling-En-pot-otherGal4dim.png] || --- Epot = -1.507\*e+7 kJ\*mol-1, Etot = -1.2625\*e+7 kJ\*mol-1, RMSD < 0.5 nm for 100 ps for BphP1-BcLOV4 dimer with shortened amino acid sequence and Epot = -5.886\*e+6 kJ\*mol-1, Etot = -4.937\*e+6 kJ\*mol-1, RMSD < 0.5 nm for 100 ps for Gal4-Q-PAS1 dimer with shortened amino acid sequence. This indicates that the removal of a part of the amino acid sequence may change the state of the system (the potential energy value decreased). The PyDockWEB total score for the BphP1-BcLOV4 dimer with a shortened amino acid sequence is -22.024 kJ\*mol-1, and for the Gal4-Q-PAS1 dimer with a shortened amino acid sequence is -45.299 kJ\*mol-1, which is less favorable than for full-length dimers (-33.784 kJ\*mol-1 and -124.942 kJ\*mol-1, respectively). Below are molecular dynamics parameters for BphP1-BcLOV4 dimer with shortened amino acid sequnce (first graphs) and Gal4-Q-PAS1 dimer with shortened amino acid sequnce (second graphs). --- ~|| ![https://static.igem.org/mediawiki/2021/9/9a/T--LMSU--Modeling-ourBphP1dim-gif.gif] | ![https://static.igem.org/mediawiki/2021/f/f1/T--LMSU--Modeling-RMSD-ourBphP1dim.png] || || ![https://static.igem.org/mediawiki/2021/c/c2/T--LMSU--Modeling-En-tot-ourBphP1dim.png] | ![https://static.igem.org/mediawiki/2021/2/22/T--LMSU--Modeling-En-pot-ourBphP1dim.png] || --- ~|| ![https://static.igem.org/mediawiki/2021/1/15/T--LMSU--Modeling-Gal4-Q-PAS1dim-gif.gif] | ![https://static.igem.org/mediawiki/2021/1/1c/T--LMSU--Modeling-RMSD-Gal4-Q-PAS1dim.png] || || ![https://static.igem.org/mediawiki/2021/b/b1/T--LMSU--Modeling-En-tot-Gal4-Q-PAS1dim.png] | ![https://static.igem.org/mediawiki/2021/2/25/T--LMSU--Modeling-En-pot-Gal4-Q-PAS1dim.png] || --- Then we proved the stability of the BphP1-BcLOV4 and Gal4-Q-PAS1 structures binding upon 760 nm light stimulation. We compared these structures with the shortened variants of BphP1-BcLOV4 and Gal4-Q-PAS1 and turned out that both complexes were stable (RMSD < 0.5 nm for 100 ps). However, the complex of chimeric proteins with a full-length amino acid sequence had a lower potential energy but higher PyDockWEB total score (Epot = -8.11\*e+6 kJ\*mol-1 and S = -35.312 kJ\*mol-1, respectively) than the complex of shortened proteins (Epot = -7.185\*e+6 kJ\*mol-1 and S = -36.828 kJ\*mol-1, respectively). As expected, the Gal4-Q-PAS1 orientation relative to BphP1-BcLOV4 in two variants prevented the BphP1-BcLOV4 and Gal4-Q-PAS1 dimers formation. Below are three-dimensional structures of BphP1-BcLOV4 and Gal4-Q-PAS1 complex with full-length amino acid sequnces (first graphs) and shortened amino acid sequnces (second graphs). --- ! [English]) ~|| ![https://static.igem.org/mediawiki/2021/1/18/T--LMSU--Modeling_otherBphP1-Gal4-gif.gif] | ![https://static.igem.org/mediawiki/2021/d/d3/T--LMSU--Modeling_RMSD-otherBphP1-Gal4.png] || || ![https://static.igem.org/mediawiki/2021/e/e0/T--LMSU--Modeling_En_tot-otherBphP1-Gal4.png] | ![https://static.igem.org/mediawiki/2021/6/6f/T--LMSU--Modeling_En_pot-otherBphP1-Gal4.png] || --- ! [English]) ~|| ![https://static.igem.org/mediawiki/2021/6/69/T--LMSU--Modeling_ourBphP1-Gal4-gif.gif] | ![https://static.igem.org/mediawiki/2021/f/f7/T--LMSU--Modeling_RMSD-ourBphP1-Gal4.png] || || ![https://static.igem.org/mediawiki/2021/3/3f/T--LMSU--Modeling_En_tot-ourBphP1-Gal4.png] | ![https://static.igem.org/mediawiki/2021/9/93/T--LMSU--Modeling_En_pot-ourBphP1-Gal4.png] || --- We also performed a number of simulations of the Gal4 HTH-domains binding to 5’-CGG-N11-CCG-3' promoter region of DNA and obtained the following molecular dynamics values: Epot = -1.946\*e+6 kJ\*mol-1 for full-length one and Epot = -1.828\*e+6 kJ\*mol-1 for shortened one (RMSD < 0.5 nm for 200 and 400 ps, respectively). It turned out that the energy of the BphP1-BcLOV4 complex with Gal4-Q-PAS1 (Epot = -8.11\*e+6 kJ\*mol-1 and Epot = -7.185\*e+6 kJ\*mol-1 for proteins with full-length and shortened amino acid sequences, respectively) is lower than the energy of the DNA-bound Gal4-Q-PAS1 (Epot = -1.946\*e+6 kJ\*mol-1 and Epot = -1.828\*e+6 kJ\*mol-1 for proteins with full-length and shortened amino acid sequences, respectively) in the AMBER99SB force field, which means that BphP1 can dissociate Gal4 from regulatory DNA sites and system works correctly. Below are DNA-bound Gal4 protein with full-length amino acid sequnce (first graphs) and shortened amino acid sequnce (second graphs). --- ! [English]) ~|| ![https://static.igem.org/mediawiki/2021/f/f6/T--LMSU--Modeling-otherGal4-DNA-gif.gif] | ![https://static.igem.org/mediawiki/2021/3/36/T--LMSU--Modeling-RMSD-otherGal4-DNA.png] || || ![https://static.igem.org/mediawiki/2021/d/d3/T--LMSU--Modeling-En-tot-otherGal4-DNA.png] | ![https://static.igem.org/mediawiki/2021/4/41/T--LMSU--Modeling-En-pot-otherGal4-DNA.png] || --- ! [English]) ~|| ![https://static.igem.org/mediawiki/2021/d/d2/T--LMSU--Modeling-ourGal4-DNA-gif.gif] | ![https://static.igem.org/mediawiki/2021/4/45/T--LMSU--Modeling-RMSD-ourGal4-DNA.png] || || ![https://static.igem.org/mediawiki/2021/c/c4/T--LMSU--Modeling-En-tot-ourGal4-DNA.png] | ![https://static.igem.org/mediawiki/2021/c/c3/T--LMSU--Modeling-En-pot-ourGal4-DNA.png] || --- Using a set of CHARMM force fields, we simulated the binding of the BcLOV4 alpha helix to the lipid membranes of *E. coli* (Epot = -2.3\*e+6 kJ\*mol-1) and *Arthrospira platensis* IPPAS B-256 (Epot = -2.1\*e+6 kJ\*mol-1). *E. coli* accumulates two major membrane phospholipids: phosphatidylethanolamine, phosphatidylglycerol, while the *Spirulina* membrane is consisted of glycolipids: monogalactosyl diacylglycerolipids, digalactosyl diacylglycerolipids, and sulfoquinovosyl diacylglycerolipids. ---  --- Since there are no models of these membranes in the CHARMM-GUI database, we decided to create dipalmitoyl phosphatidylethanolamine (C16:0/16:0) membrane model for *E. coli* (first graphs below) and dipalmitoyl glycerol (C16:0/16:0) membrane model for *Spirulina* in CHARMM-GUI (second graphs below) [8]. --- ![https://static.igem.org/mediawiki/2021/a/ac/T--LMSU--Modeling-BcLOV4-membr-E-coli-gif.gif] ~|| ![https://static.igem.org/mediawiki/2021/2/27/T--LMSU--Modeling-En-tot-BcLOV4-membr-E-coli.png] | ![https://static.igem.org/mediawiki/2021/0/05/T--LMSU--Modeling-En-pot-BcLOV4-membr-E-coli.png] || || ![https://static.igem.org/mediawiki/2021/4/45/T--LMSU--Modeling-RMSD-BcLOV4-membr-E-coli.png] | ![https://static.igem.org/mediawiki/2021/9/99/T--LMSU--Modeling-Temp-BcLOV4-membr-E-coli.png] || --- ![https://static.igem.org/mediawiki/2021/a/ae/T--LMSU--Modeling-BcLOV4-memb-SP-gif.gif] ~|| ![https://static.igem.org/mediawiki/2021/2/2f/T--LMSU--Modeling-En-tot-BcLOV4-memb-SP.png] | ![https://static.igem.org/mediawiki/2021/0/02/T--LMSU--Modeling-En-pot-BcLOV4-memb-SP.png] || || ![https://static.igem.org/mediawiki/2021/e/ef/T--LMSU--Modeling-RMSD-BcLOV4-memb-SP.png] | ![https://static.igem.org/mediawiki/2021/5/50/T--LMSU--Modeling-Temp-BcLOV4-memb-SP.png] || --- The results demonstrate that membrane localization of RGS-truncated BcLOV4 protein is mediated by a polybasic amphipathic helix after the LOV domain (RMSD < 0.5 nm for 100 ps). The reversible electrostatic interaction dependent on the anionic content of the membrane without preference for a specific group. Although the potential energy and RMSD mainly used for proving that the simulation was physically valid, we verified the stability of the optogenetic system components, and also determined which variant of the optogenetic system (full-length or shortened) can be more effective. Our findings confirm that this optogenetic system can be implemented and works more efficiently with full-length chimeric proteins at an average temperature of 300-304 K. ## References [1] https://swissmodel.expasy.org/ [2] https://yanglab.nankai.edu.cn/trRosetta/ [3] Jimenez-Garcia B., Pons C. and Fernandez-Recio J. "pyDockWEB: a web server for rigid-body protein-protein docking using electrostatics and desolvation scoring". Bioinformatics (2013) 29(13):1698-1699. [4] https://cluspro.org/ [5] T. A. Redchuk, E. S. Omelina, K. G. Chernov, and V. V. Verkhusha, ‘Near-infrared optogenetic pair for protein regulation and spectral multiplexing’, Nat. Chem. Biol., vol. 13, no. 6, pp. 633–639, Jun. 2017, doi: 10.1038/nchembio.2343. [6] S. T. Glantz, E. E. Berlew, Z. Jaber, B. S. Schuster, K. H. Gardner, and B. Y. Chow, ‘Directly light-regulated binding of RGS-LOV photoreceptors to anionic membrane phospholipids’, Proc. Natl. Acad. Sci. U. S. A., vol. 115, no. 33, pp. E7720–E7727, Aug. 2018, doi: 10.1073/pnas.1802832115. [7] A. Nekrasov, A. Anashkina, and A. Zinchenko, A New Paradigm of Protein Structural Organization. 2013. [8] https://www.charmm-gui.org/
Team:LMSU/Model
Project
Lab
People
Parts
Big Picture
Judging