Abstract
In this project, we used approaches such as molecular modeling and molecular dynamic simulations to promote the project’s progress and reduced the error cost. During the primary stages, we modeled and optimized the fused protein; in the latter stages, we conducted analysis on the scaffold where no immunofluorescence is detected, therefore upgrading the scaffold, allowing the project to develop smoothly. We used SWISS-MODEL, ITASSER, trRosetta, Procheck, Verify3D, PyMOL, NAMD, VMD, and Snap Gene in this progress.
Software introduction and their principles
I-TASSER (Iterative Threading ASSEmbly Refinement) is a hierarchical approach to protein structure prediction and structure-based function annotation. It first identifies structural templates from the PDB by multiple threading approach LOMETS (Local Meta-Threading Server, version 3) is a next-generation meta-server approach to template-based protein structure prediction and structure-based function annotation)with full-length atomic models constructed by iterative template-based fragment assembly simulations. Function insights of the target are then derived by re-threading the 3D models through protein function database BioLiP. The server is in active development with the goal to provide the most accurate protein structure and function predictions using state-of-the-art algorithms. The server is only for non-commercial use.
The trRosetta is an algorithm for fast and accurate protein structure prediction. It builds the protein structure based on direct energy minimizations with a restrained Rosetta. The restraints include inter-residue distance and orientation distributions, predicted by a deep neural network. Homologous templates are included in the network prediction to improve the accuracy further. In benchmark tests on CASP13 and CAMEO derived sets, the trRosetta outperforms all previously described methods
Model construction, evaluation, and optimization of PETase-spy tag fusion protein
We created the model of PETase linked with spy tag to evaluate and optimize it. Different apps/online server to model PETase-spy tags have led to various results.
To select the best model, we compared different models and rated them through VERIFY3D. The result showed that 83.69% of the residues have averaged 3D-1D score >= 0.2, suggesting our model a good quality one (The model here was constructed using trRosetta).
Model construction, evaluation, and optimization of MHETase-snoop tag fusion protein
Similarly, we created the model of MHETase linked with Snooptag to evaluate and optimize it. Different platforms gave different results when modeling MHET-Snoop tag. We compared the ratings of the different models using various methods, such as VERIFY3D. The result showed that 87.23% of the residues have averaged 3D-1D score >= 0.2, suggesting our model a good quality one (The model here was constructed using trRosetta).
Model construction of CBM-SC-SC-SNC-SC-V5-7813 protein scaffold
In results, we discovered that while building the scaffold CBM-SC-SC-SNC-SC-V5-7813, no immunofluorescence was detected. We hypothesized that the V5 tag is obstructed, which hampered the binding with the antibody, so no immunofluorescence can be detected. We used I-TASSER to model the protein and found that the V5 tag is embedded, which fits our hypothesis.
The prediction results are shown below, we selected the one with the highest rating.
Although the selected model was listed as a highest rating model by ITASSER, it was identified as a fail model by VERIFY3D. The VERIFY3D result showed that only 46.67% of the residues have averaged 3D-1D score >= 0.2.
As a result, we optimized the structure using MD simulation to improve the R group structure and achieve the best energy status of the protein. The modelled structure was solvated in at least 10 Å in each direction and NaCl at a concentration 0.15 mol/l with a neutralization option added. 1 nanosecond (ns) water and ion equilibrations, 1000 steps of energy minimization and gradual heating up to 300 K were performed. Main 20 ns trajectories were calculated using Langevin molecular dynamics simulation in 300 K with 2 fs time step in atmospheric pressure and long-range electrostatic interactions calculated using particle-mesh Ewald summation. After optimization, the Root Mean Square Deviation (RMSD) of the model was around 3.5 Å, indicating the optimized model a good quality one (Fig. 7). We then resubmitted the optimized structure to VERIFY3D, and the evaluation result showed that 91.69% of the residues have averaged 3D-1D score >= 0.2.
Model construction of CBM-V5-SC-SC-SNC-SC-7813 protein scaffold
To explore whether the V5-tag could be exposed on the surface when placed in front of the catcher, we used I-TASSER to construct a scaffold model of CBM-V5-SC-SC-SNC-SC-7813. As shown in the figure 8, when the v5-tag is placed in front of the catcher, it can be completely exposed on the surface.
At the same time, we predicted the structure of the scaffolds that were proved to be successful in immunofluorescence from previous experiments. The modeling results are as follows. The red area is v5-tag. It can be found that the predicted results are consistent with the experiment, and v5-tag is exposed on the surface.
Although we can detect the immunofluorescence of the strain with V5 tag near the N-terminus or in the middle, there’s still no PETase activity was determined. We changed our strategy that the catchers were replace by its corresponding tags. The scaffold SP-CBM-V5-ST-ST-SNT-ST-7813 was expressed in the yeast, and the model of CBM-V5-ST-ST-SNT-ST-7813 was constructed. It can be found that the predicted result is good (Fig. 11). At the same time, the PETase was fused with spycatcher, and MHETase was fused with snoopcatcher.
A new FAST-PETase was recently reported with excellent ability to biodegrade PET plastic. The FAST-PETase is a mutated PETase from Ideonella sakaiensis. To compare the catalytic performance of such enzyme with that of our enzymes, we thus constructed related fusion enzyme FAST-PETase-spycatcher and displayed on cell surface. Their modeled structures were obtained through trRosetta and ITASSER online servers. VERIFY3D was used to evaluate these constructed models. Evaluation results showed that our models constructed by ITASSER were convincing.
Conclusion
Establishing protein models not only helps us to analyze the problems that arise during the experiment, but also guides the experimental design by combining the model, which reduces our workload and ensures the experiment to be carried out smoothly. After spending most of our early days analyzing the reasons for the lack of immunofluorescence and inactivity of the surface display system with the catcher as a scaffold, building a structural model of the protein disclosed that the problem might be protein folding in the molecular level. Eventually, this conclusion prompted us to Try a catcher and tag swap strategy to solve the problem and constructed a Candida tropicalis recombinant strain that can degrade PET efficiently.
References
Kiefer, Florian, et al. "The SWISS-MODEL Repository and associated resources." Nucleic acids research. 37.suppl_1 (2009): D387-D392.
Yang, Jianyi, et al. "The I-TASSER Suite: protein structure and function prediction." Nature methods 12.1 (2015): 7-8.
Humphrey, William, Andrew Dalke, and Klaus Schulten. "VMD: visual molecular dynamics." Journal of molecular graphics 14.1 (1996): 33-38.
Eisenberg, D., R. Lüthy, and J. U. Bowie. "VERIFY3D: assessment of protein models with three-dimensional profiles. InMethods in enzymology 1997 Jan 1 (Vol. 277, pp. 396-404)."
Lu, Hongyuan, et al. "Machine learning-aided engineering of hydrolases for PET depolymerization." Nature 604.7907 (2022): 662-667.