We hope that EcN can complete the metabolism of erythritol, so in order to explore whether EcN can make use of erythritol and the best conditions of its subsequent reaction and metabolism. We need to evaluate the transport ability of ABC sugar transporter to erythritol, and how sldH and L-RI can carry out multi-enzyme cascade to realize the reaction and achieve better results. So as to explore whether EcN can normally transport erythritol and its subsequent reaction and metabolism. At the same time, we verified the validity of the gene sequence of eryEFG system by modeling.
We use ChimeraX and pymol software to map the protein structure, and we will use Pymol to align the structure of two protein. For the designed proteins with linker and his tag tags or those that have not been submitted in pdb database, we use AlphaFold 2 to model the structure and get the corresponding pdb files. For the docking modeling of many proteins and small molecules, we mainly use CB-Dock developed by Professor Cao Yang from the college of Life Sciences, Sichuan University.
CB-Dock is an online molecular docking tool with friendly interface. Its characteristic is that it can predict the binding site of a given protein, that is, a new pocket detection method based on curvature is used to calculate the center and size of the pocket. And the docking is performed by the popular docking program Autodock Vina.
Fig.1 EryD (with erythritolthritol)
Fig.2 CB-dock results of EryE, EryF, EryG.
Fig.3 Structure of EryE, EryF, EryG with erythritol
| Gene | Vina score with erythritol |
|---|---|
| eryD | -4.6 |
| eryE | -4.1 |
| eryF | -4.1 |
| eryG | -4.7 |
Table1 CB-dock scores of EryD, EryEFG with erythritol. "Vina score" means the score of the complex obtained by molecular docking of receptor and ligand with corresponding pocket parameters by Vina program. The lower the value, the higher the affinity of receptor and ligand.
Based on an erythritol transport mechanism found in previous research, we found the related transport system eryEFG according to the loci position. In order to verify whether eryEFG is responsible for erythritol transport, we have made a model prediction using CB-dock2. The lower the CB-dock2 score, the stronger the binding ability of the protein to the ligand (erythritolthritol). According to the theory that structure determines function, combined with our previous modeling of erythritol binding protein EryD (Fig. 1). By comparing the docking results of the eryEFG transporters with EryD, we found that EryE, EryF and EryG have binding effects with erythritol (Fig. 2, Fig. 3, Table 1), indicating a great potential as transport channel proteins. Among the three proteins, EryG had the strongest binding effect on erythritol, even surpassing EryD.
| Gene | Homologous protein | blast total score | blast Query Cover | CB-dock2 Score with erythritol |
|---|---|---|---|---|
| eryE | / | / | / | -4.1 |
| mglA | eryE | 350 | 93% | -4.6 |
| rbsA | eryE | 348 | 96% | -4.4 |
| lsrA | eryE | 285 | 95% | -4.3 |
| alsA | eryE | 279 | 94% | -4.8 |
| yphE | eryE | 271 | 93% | -4.3 |
| livF | eryE | 195 | 80% | -4.3 |
| eryF | / | / | / | -4.1 |
| yphD | eryF | 128 | 86% | -3.9 |
| mglC | eryF | 113 | 80% | -4.2 |
| alsC | eryF | 100 | 85% | -4.1 |
| lsrC | eryF | 90.5 | 84% | -3.9 |
| eryG | / | / | / | -4.7 |
| rbsB | eryG | 95.1 | 86% | -4.4 |
Table 2 Homologous proteins list, blast scores and CB-dock scores.
Fig. 4 CB-dock results of homologous proteins of EryE, EryF, EryG
Fig. 5 Structure of several homologous proteins with erythritol
According to our results in wet lab, we found that without the specific erythritolthritol transport proteins erythritolE, erythritolF and erythritolG, the perythritolU-carrying strains could use the erythritolthritol as a carbon source. We hypothesized that other sugar transporters in E. coli might also function as erythritolthritol transporters. So we used protein blast to find the homologous proteins of erythritolE, erythritolF and erythritolG (Table2). We used CB-dock2 to simulate their binding efficiency with erythritolthritol (Fig. 4, Fig. 5) and found that almost all of the homologous protein could transport the erythritolthritol, even many proteins have a higher affinity for erythritolthritol than their homologous proteins (Table 2). However, because the modeling results of CB-dock can only show its binding ability, the real transport efficiency of various homologous proteins for erythritolthritol needs to be further verified and characterized by wet experiments.
Fig.6 Multi-enzyme cascade modeling overview
OConsidering that the transition from erythritol to L-erythrulose to L-erythrulose is a continuous process, while L-RI is localized in the cytoplasm, SLDH is localized in the cell membrane. Here, we introduced a surface display that anchors the protein L-RI outside the membrane. (We designed to add a linker to L-RI and SldB so that L-RI is indirectly connected to SldB, thus anchoring L-RI on the outer side of the cell membrane, which has the advantage of anchoring the enzyme on the membrane to improve the reaction efficiency); on the other hand, due to the close spatial distance, the L-erythrulose generated from the SldB reaction can be rapidly transferred to L-RI to participate in the reaction to generate L-erythrulose, thus forming a multi-enzyme cascade reaction.
Idea: Our first idea was to connect the linker to the C- or N-terminal of the L-RI. For the sldH, the his tag should preferably be behind the sldH as required by the experimental group, so that the sldH composite structure was built. However, we found it difficult to model the sldH composite structure with software such as ChimeraX, so we considered putting his tag on the L-RI for modeling.
For the L-RI-linked ones, we designed four types of linkages, linker-his tag-L-RI, linker-L-RI-his tag, L-RI-his tag-linker and his tag- L-RI -linker. by Alphfold2 modeling and pymol-align analysis, the The ones with better fit to L-RI were screened, while the optimal combination was screened by CB-DOCK2 small molecule docking for the four results.
| CB-dock2 | his tag-L-RI-linker | linker-his tag-L-RI | linker-L-RI- his tag | L-RI-his tag-linker |
|---|---|---|---|---|
| Score | -4.6 | -4.0 | -4.1 | -3.7 |
Table 3, CB-dock2 score for 4 types of L-RI connections.
| Alphfold2 | his tag-L-RI-linker | linker-L-RI- his tag | linker- his tag-L-RI | L-RI- his tag-linker |
|---|---|---|---|---|
| Model_1 | 70.6 | 69.9 | 66.9 | 75.3 |
| Model_2 | 79.5 | 79.1 | 79.8 | 79.7 |
Table 4 Alphfold2 score for 4 types of L-RI connections
To determine whether the L-RI with the addition of linker still had the same reactive site as the original L-RI, we compared the L-erythrulose docking effect of L-RI-linker with that of L-RI. The result was that the reaction site of L-RI did not change after the addition of linker, and L-erythrulose still appeared at the original reaction site.
Fig.7 L-RI-linker(with erythritol)
Fig.8 Alignment result of L-RI(yellow) and L-RI-linker(pink)
Idea : For the sldH composite structure, we considered modeling sldH separately when the mechanism of action was not clear. However, it turned out that SldB was better modeled and SldA was poorly modeled; meanwhile, we found that sldH does not act separately and needs to be involved in the reaction together, so we decided not to model SldA and SldB separately and shifted to the prediction of the composite structure.
We first tried the complex protein structure prediction of SldA and SldB, trying to construct a complex protein with the same structure as SLDH. However, we gave up the complex protein structure prediction of SldA and SldB because both alphafold2 and SWISS-MODEL modeling results were not satisfactory: the structure of the linker peptide sequence did not correspond to SLDH, and the spatial structure of SldA and SldB did not correspond to SLDH.
Idea: Since we found that sldH needed to participate in the reaction together, we could not use the original DOCK method to judge. So we used the method of pymol comparison to determine that the structure of SldB with his tag did not change after the model.
In the case of the failure of the composite protein structure prediction, we chose another route: linker and his tag were performed separately for SldA and SldB. We attached linker and his tag to the N-terminal end of SldB and subsequently performed alphafold2 structure prediction, and obtained the desired prediction and his tag-linker-SldB. Subsequently, we performed pymol-align analysis of the obtained his tag-linker-SldB with SldB and found that the fit was better, and the SldB structural domain was not changed, which could indicate that its reaction structure was not changed.
Fig.9 Chromatogram of protein structure prediction results of SldB
After the successful prediction of the SldB linker and his tag structure, we also wanted to implement this operation in the SldA protein, i.e., adding the linker and his tag to the C-terminus of SldA. we found that the run scores of the SldA sequence, except for the linker and his tag sequences, were similar to our previously run SldA sequence, and we inferred that the addition of the linker and his tag will not change the reaction structure of SldA.
Fig.10 his tag - linker – SldB
Finally, synthesizing the protein modeling work, we chose to add his tag to the C-terminus of SldA, i.e., SldB-SldA- his tag. we conveyed the results and the modeling process based on the modeling to the experimental group, and finally succeeded in obtaining better results based on the modeling idea, and successfully achieved the effect of modeling to guide the experiment.
1. Hiromi Yoshida . et al. X-ray structure of a novel L-ribose isomerase acting on a non-natural sugar L-ribose as its ideal substrate. The FEBS journal 281 , 3150-3164 (2014).
2. Hoshino . et al. Membrane-bound D-sorbitol dehydrogenase of Gluconobacter suboxydans IFO 3255--enzymatic and genetic characterization. Biochimica et biophysica acta 1647 , 278-288(2003).
3. J F Sperry and D C Robertson. Erythritol catabolism by Brucella abortus. Journal of bacteriology 121 , 619-630(1975).
4. Liu Y , Grimm M , Dai W T , et al. CB-Dock: a web server for cavity detection-guided protein–ligand blind docking[J]. Acta Pharmacologica Sinica.2020, 41(1):7.
5. Jerome . et al. Eberhardt AutoDock Vina 1.2.0: New Docking Methods, Expanded Force Field, and Python Bindings. Journal of Chemical Information and Modeling 2021 61 (8), 3891-3898(2021).