Tyrosinases are a ubiquitous group of copper-containing metalloenzymes that hydroxylate and oxidize phenolic molecules. In our project, tyrosinase was anchored to the cell membrane to catalyze BPA using the method of surface display. Therefore, whether tyrosinase could maintain its original function on the cell membrane became a question to be explored. The existing verification methods can only continuously debug the interprotein linker, and then test the enzyme activity through experiments. In order to avoid this tedious work, we used molecular simulation to predict the structure of the fusion-expressed tyrosinase protein, and compared with the original structure, the results showed that the fusion-expressed tyrosinase protein had a stable structure and catalytic activity. This conclusion is consistent with the experimental data.
Molecular simulation refers to quantitatively predicting some structural information, molecular properties and chemical reactions that are difficult to be determined by experimental methods with the help of computers, according to theoretical chemistry and laws of mechanics. We use alphafold2 to predict the 3D structure of proteins, AlphaFold2 is a deep learning algorithm with attention mechanisms, which uses several techniques like evolutionarily related sequences, multiple sequence alignment, amino acid residue pairs, etc.[1]
The original structure
The tyrosinase protein we used was a protein with an optimized sequence, and its 3D structure could not be found on PDB. Through sequence homology analysis, we found the 5M6B protein with only a few amino acids sequence differences. Thus, we use the 5M6B as the original structure of tyrosinase.
Figure1.5M6B information.
Next, we compared the structure of the tyrosinase protein predicted by alphafold2 with that of 5M6B. The results show that the 3D structures of the two proteins are very similar, and the RMSD is only 0.213, so we believe that the prediction result of alphafold2 is credible.
Figure2. Three-dimensional structure align.
To predict the 3D structure of fusion protein:
We input the amino acid sequence of INP_linker_tyrosinase into alphafold2 for structure prediction and set the iteration five times. PLDDT (Per-confidence scores) images of prediction results are shown in figure 3, with blue areas representing the convergence of prediction results and red areas representing the fluctuation of prediction results. alphafold2 achieved convergence in predicting the structure of both INP and tyrosinase, and there was great uncertainty in predicting the structure of the linker. Since the linker is a peptide chain and there is no obvious secondary structure, which leads to uncertainty in structure prediction, which is in line with biological principles.
Figure3. Per-residue confidence scores of INP_linker_tyrosinase.
In addition to PLDDT, PAE (Domain position confidence) is also a common evaluation indicator. Predicted PAE gives a distance error for every pair of residues It gives AlphaFold's estimate of position error at residue x when the predicted and true structures are aligned on residue y. Values range from 0 - 35 Angstroms. It is usually shown as a heatmap image with residue numbers running along vertical and horizontal axes and colour at each pixel indicating the PAE value for the corresponding pair of residues. If the relative position of two domains is confidently predicted then the PAE values will be low (less than 5A) for pairs of residues with one residue in each domain. The blue region representing the two proteins and linker can be clearly seen on the PAE image, which supports the accuracy of the prediction results.
Figure4. Domain position confidence of INP_linker_tyrosinase.
Three-dimensional structure comparison:
By comparing the predicted fusion protein structure with the structure of 5M6B, it was found that tyrosinase and 5M6B had obvious overlap, and the RMSD was 0.663. This result indicated that the structure and function of tyrosinase could not be changed by anchoring to the cell membrane with INP fusion expression.
Figure5. Results before and after the alignment.
Reference：
[1] Jumper J, Evans R, Pritzel A, et al. Highly accurate protein structure prediction with AlphaFold[J]. Nature, 2021, 596(7873): 583-589
