Difference between revisions of "Team:XMU-China/Model"

To achieve the goal to detect the Glyphosate, we designed a system which is composed of EcAKR4-1, GRHPR and iNAP. We convert glyphosate into AMPA and glyoxylate through a Glyphosate oxidase (GOX) from Echinochloa Colona named EcAKR4-1. Then glyoxylate can be transferred into glycolic acid with the function of a Human Glyoxylate Reductase named GRHPR, which will simultaneously convert a molecule of NADPH to NADP+.

We use iNAP, a fused protein of T-Tex and cpYFP to detect the amount of NADPH in the system. When NADPH exists, the conformational changes of the fusion protein make it become fluorescent activity. Therefore, the amount of glyphosate can be indirectly quantified by determining the amount of NADPH consumed in the process from fluorescence intensity.

We modeled to describe the dynamics of the detection system and giving a full version of working curve which can be used in the degradation systems.
The whole reactions in the detection system are as following:

And those reactions can be described by the following ODE equations:

The reaction between iNAP and NADPH is quite fast so we using the rapid equilibrium assumption. After fitting with experimental data, we found our ODE equations can describe all the dynamics in Fig.1.

Taking the fluorescence in 。。。。。。And the working curve can be simulation and obtained in Fig.2.

Modeling of PhnJ was aimed in search of optimized enzymatic protein to achieve idealized degradative function of PhnJ after being transformed into E. coli. The following processes and methods led us to explore the optimal solution.

2.1 An Top-down strategy : in the holoenzyme view

Before we going deeply into the mechanism of PhnJ, we firstly try to understand the holoenzyme—PhnGHIJK, where the PhnJ is assembled to after translation.

PhnGHIJ, the C-P lyase core complex, contains four protein subunits, in which PhnJ is the core enzyme, catalyzing the lysis of C-P bond in glyphosate and methylphosphonate as a C-P lyase. PhnI is a nucleosidase to deglycosylate ATP and GTP into ribose 5-triphosphate. PhnGHL support the reaction of C-P lyase pathway in the place of PhnI to transfer the phosphonate to the ATP position in ribose C-1` position to displace adenine and generate the reaction intermediate ribose 5`-triphosphonate alkyl phosphonate. PhnM release the pyrophosphate and PhnJ cleaves the C-P bond and PhnP and PhnN convert it to PRPP [2-18]. PhnK associates with PhnGHIJ as the fifth protein, resembles ABC cassette proteins with unclear stoichiometry [1].

“The C-P lyase core complex resembles the letter 'H' with rounded arms that are twisted approximately 45° in and out of the plane with respect to each other.” [3] PhnG and PhnJH compose the two arms of “H”, while two PhnI touch and stretch over the two arms in the middle, as the center of C-P lyase complex.

PhnI is comprised with four-stranded, antiparallel beta-sheet next to a four-helix bundle with two helical extension in both C- and N- termini to grasp and tether PhnJ by extensive interaction. Furthermore, PhnJ connects and associates to PhnH through packing of two conserved alpha-helices in both proteins. The interaction of PhnHJ heterodimer is similar to the subunits structure of PhnH homodimer.

Just as described above, the collaboration between PhnG, PhnH, PhnI, PhnJ, PhnK is quite important so we suppose that increasing the binding affinity between the four proteins may contribute to the performance of Glyphosate degradation.

For the PhnJ (exogenous), our work mainly focuses on getting the structure by homologous modeling.

The gene sequence of PhnJ in Enterobacter cloacea K7 is reported to be highly conserved and high identity (99%) to the phnJ sequence in Enterobacter cloacae strain EcWSU1[4]. Therefore, we used codon-optimization to the preference of E. coli and submited the gene sequence in the website of SWISS-MODEL (https://swissmodel.expasy.org/), filtrated and obtained the optimal homologous protein model of PhnJ (exogenous).

In order to achieve a more precise and accurate result of PhnJ protein model, we implemented MOE to build another homologous model with different algorithm.

Plus, MOE offers geometry analysis of protein model, which brings much convenience to our model assessment and evaluation.

Initially, we assessed the SWISS-MODEL result of PhnJ in Molprobity (http://molprobity.biochem.duke.edu/) , with the function of “Analyze geometry without all-atom contacts”. It was reported that the model possesses numbers of clashes, outlier rotamers, and CaBlam outliers.

Thus, we built another homologous PhnJ model in software MOE and achieve geometry analysis within the software. Moreover, with the basic understanding of the C-P lyase complex in the aspect of structure and protein interaction, we designed several mutations in PhnJ and transformed them into the fittest configuration of each mutated amino acid. Basically, the interface of PhnJ to PhnH and PhnI was the essence of mutation. For instance, the hydrophilic lysine in the PhnJH interface was mutated into hydrophobic phenylalanine to achieve a strengthened hydrophobic interaction and diminish the surface water inside. So we try to choose the following three mutation, including R21M, which mutate the 21th amino acid from R to M, P45Q and T16S & R40Y which have two sites mutated.

The mutation is Homologous modeling by Swiss-Model, and the mutated site can be seen in Fig?-Fig.?.

Protein-Protein docking is carried out by the ClusPro server and the binding energy can be seen in the table.1.

2.2 An Bottom-up strategy : in the PhnJ view

Only obtaining the higher binding affinity can’t guarantee that the mutated protein will have better performance to degradation the glyphosate. We should go deeper into the reaction mechanism of PhnJ to check whether the mutated action leads to a higher speed of reaction or just inactive the enzyme.

The reaction mechanism of PhnJ can divide into the following six-step.

Step 1. Initial formation of the 5`deoxyadenosyl radical. The mechanism is assumed to be identical to other Radical SAM type proteins.

Step 2. The adenosyl radical abstracts a hydrogen atom from the catalytic glycine residue. The products of this activation step represent the ground state of this enzyme.

Step 3. The glycyl radical abstracts a hydrogen atom from the catalytic cysteine residue.

Step 4. The cysteinyl residue attacks the phosphate group, forming a pentavalent intermediate.

Step 5. The pentavalent intermediate collapses, eliminating a methyl radical with concomitant abstraction of a hydrogen atom from the glycine residue.

Step 6. One of the hydroxyl groups of the ribose intermediate initiates a nucleophilic attack on the covalently bound phosphate to regenerate the cysteine residue and final product. The enzyme is now in a state that it can perform another round of catalysis.

In the all six-step reaction mechanism, some super-distance hydrogen atom transforms are worth to study, like Step2 and Step3. The previous research shows that the hydrogen atom transforms which carried by the Hydrogen bond network, some proton channel or protein re-folding. In other views for PhnJ itself, the super-distance hydrogen atom transforms just happened between the binding site and catalytic site which means that it’s possible that the substrate is transported inside or outside the protein. But for the most common situation in such a complex system, those mechanisms mentioned above are likely to work together to finish hydrogen atom transforms.

So here we have following three parts to evaluate the state of hydrogen atom transforms in mutation which is operated in three pars in reaction mechanism including Step2, Step3 and key reaction process in Step5.

2.2.1 Finding the adenosyl radical transforms path in Step2，an general method combining Monte Carlo sample and reinforcement learning

As we can see in reaction mechanism Step2，we can find that first super-distance hydrogen atom transforms in PhnJ’s reaction coordinates. After Initial formation of the 5`deoxyadenosyl radical, The adenosyl radical abstracts a hydrogen atom from the catalytic glycine residue. But the distance between adenosyl radical and Gly32 is too far to finish the direct transform without the moving of adenosyl radical. In order to finding the possible hydrogen atom transforms path in Step2, we propose an algorithm call F2ARTP (Finding the adenosyl radical transforms path) to calculate the path.

The FHATP algorithm can be split into two parts: the Monte Carlo sampling which finally gives a score function to evaluate the possible docking position of the substrate using the Gibson binding energy and, reinforcement learning which will find the path that satisfies the minimization of total reaction energy and moving distance. The workflow of FHATP can be seen in Fig.?,

The Monte Carlo Sampling parts is used to collect the sample which can be used to integrate the scoring functions which is used to evaluate the possible docking position of the substrate and work as the return function in Deep Q-Learning. It can be split into following parts:

1. Pre-processing the protein structure for molecular dynamics. Adding missing hydrogen atoms and convert non-standard residues to their standard equivalents to make the PDB structure suitable for performing the molecular dynamics.

2. Using the classical molecular dynamics to collect the possible flexible structural sample which can be used for molecular docking.

3. Pick a reasonable number of frames from the results of MD and using as Accepter to perform molecular docking to find all the possible binding sites for substrate around the protein and their Gibson binding energy.

4. All step here using the hypothesis that the protein is the flexible part while the substrate is the rigid part because we need to identify the position of the substrate using the only one coordinates in cartesian coordinates, which is a convenience for us to integrating the result of molecular docking. In the step, we should find a mapping between the possible docking position and its binding energy. We can choose the table function or neural network.

The two key methods: molecular dynamics and molecular docking is described as following.

Molecular dynamics (MD) is a computer simulation method for studying the physical movements of atoms and molecules. The atoms and molecules are allowing to interact for a fixed period of time, giving a view of the dynamic evolution of the system. In the most common version, the trajectories of atoms and molecules are determined by numerically solving Newton's equations of motion for a system of interacting particles, whose forces between the particles and their potential energies are often calculated using interatomic potentials or molecular mechanics force fields.

To a system which consists of $N$