According to the needs of the experiment, we need to construct a predictor to determine whether a peptide sequence can bind to PE. After the preliminary literature search, there are 174 pieces of related data collected. Due to the small amount of data, we first performed machine learning training; then we performed special processing and tried to introduce neural networks.
Firstly, the original data is used to train the SVM model; then, the data is increased by 1:10 to get nearly two thousand pieces of data. Then, the trained SVM is used to sift the augmented data according to specific criteria to get a new data set; to dig deeper into the sequence information, the sequence is transformed into a structural graph, the features are extracted using graph neural network (GNN) and location specific scoring matrix, and the attention module is introduced to perform further learning to get a predictor.
The above dataset is divided into training and test sets, experiments are conducted using the built model and corresponding experimental metrics such as accuracy, recall and F1 score are measured.
Finally, the test results are further analysed and the model parameters are adjusted and optimised.
The initial step of constructing a plasmid is finding an appropriate plasmid backbone. We chose pAB1, a broad host range plasmid for IPTG-inducible expression of EGFP, as our basic plasmid backbone for expressing protein in P. aeruginosa. After ordering the plasmid, we found that RBS sequence will be cut off if changes the GFP sequencing to our gene of interest. And if we want to release our engineered bacteria to the soil, a constitutive expression promoter is needed. Replacing the promoter is an urgent need.
We selected 4 promoters that were indicated in the paper that could be expressed in P. aeruginosa, pRPO, Pbla, PoprL and pS.Then linked the promoter sequence into the sequence by enzyme cleavage of NsiI and XbaI on pAB1.
The pAB1 plasmids with new promoter were transferred into the DH5α strain, and EGFP was successfully expressed.
When synthesizing short promoter fragments, we used the annealing extension method of two complementary single strand DNA for rapid synthesis. We selected the cleavage sites of NsiI and XbaI for enzyme digestion and ligation operation, replacing the lac operator and ptrc by our aiming promoters. Initially, we tried to use pRPO as a constitutive promoter, but the pRPO promoter could drive the expression of gfp gene in E.coli DH5α strain, but could not drive the expression of the gene in P. aeruginosa PAO1. Therefore, we later found pS, PoprL and Pbla promoters for testing, and found that gfp driven by all three of these promoters could be successfully expressed in PAO1 strain. Given that the pS promoter strain expressed the strongest and most stable fluorescence intensity, we ended up using the pS promoter as our constitutive promoter.
After the microorganisms approaching to the microplastics, they can begin the biodeterioration step. Therefore, we intend to construct a fusion protein of PEBP and PE degrading enzymes to enhance the biodeterioration ability of our engineered bacteria. Different kinds of PE degrading enszymes play a role in different situations, so it is crucial to select a suitable PE degrading enzyme, and PEase from wax worm saliva is a suitable candidate. We couple PEBP to PEase. Signal6.0 was used to predict a signal peptide of Esterase-A from P. aeruginosa and bind to the N-terminal of PEBP. The membrane binding domain is predicted using software and is attached to the C end of the PEase. In this way, we have designed a whole PE binding and deteriorate module.
We ordered the complete PE adsorption-depolymerization module sequence from a company and cloned it into the pAB1 vector using restriction enzymes. Colony PCR and sequencing confirmed the successful construction of our vector. We introduced an enzyme cutting site between PEBP and PEase to facilitate the replacement of different PEBP sequences. Additionally, we had the company synthesize a sequence with PEBP fused to EGFP to assess the ability of PEBP-expressing bacteria to bind to microplastics.
SDS page was used for gene expression detection, and the correct bands of pAB1-PEBP-GFP were obtained, indicating that the protein was indeed correctly expressed in the bacteria. However, we didn't obtain correct band of pAB1-PEBP-PEase plasmid, so for this plasmid we tested it by modeling method.
The PEBP-GFP protein was purified and extracted. From the experimental results, we can clearly see that the fluorescence intensity of PEBP-GFP co-incubated microplastics is obviously higher than GFP co-incubated microplastics and blank control group, indicating that PEBP does play a role in adsorption.
For PEase, we tested its effect through moduling. We conducted structure predicting of PEBP-PEase. The predicting module clearly displayed a membrane spanning domain and an enzyme domain, which we exactly expected our bacteria to produce. After that molecular docking was conducted to study the interactions between the PEase enzyme and both large and small alkane molecules. The result indicating that PEase is able to deteriorate PE plastic.
In this cycle we successfully constructed pAB1-PEBP-PEase and pAB1-PEBP-GFP plasmids, the effects of PEBP were verified by co-incubation, and PEase was verified by modeling methods.
Microplastics are depolymerized into smaller plastic molecules, which then need to be assimilated by microorganisms and further depolymerized within the cells. In the assimilation step, short-chain alkanes can pass through the cell membrane on their own, while medium-short-chain alkanes can enter the cell via AlkB2. For the subsequent mineralization step, we introduced a CYP enzyme with the CYPY96F mutation to efficiently hydroxylate free short-chain alkanes. Medium-long-chain alkanes can be hydroxylated during their entry through AlkB2. To improve the hydroxylation efficiency of medium-long-chain alkanes, we coupled the Rd45 auxiliary factor to AlkB2 to assist with electron transfer. Given that ethanol dehydrogenation is a crucial step in alkane chain mineralization, we also coupled AdhA ethanol dehydrogenase, which can dehydrogenate long-chain alcohols, to AlkB2 to facilitate downstream reactions.
We had the company synthesize fragments of the AlkB2-Rd45-AdhA fusion protein and the CYP-VHb fragment, adding enzyme cutting sites at both ends. We ultimately constructed the entire vector using restriction cloning. We obtained the correct colony PCR results and sequencing results, indicating that we successfully constructed the vector.
SDS page was used for gene expression detection, and the correct bands were obtained, indicating that the protein was indeed correctly expressed in the bacteria.
For Degradation module, we tested its effect through co-incubation. We added 0.1g microplastics into 45ml IS medium, and added 1ml bacterial solution with OD600 of 0.65 for co-culture. After co-incubation for 10 days, Fourier Transform Infrared Spectrometer was used to detect microplastics. It can be found that the peak of carbon-carbon chain in the microplastics of the experimental group gradually decreased with the increase of days, and the oxygen-containing groups gradually increased, and the change was more obvious than that of the control group, indicating that Degradation module can play the role of biodeterioration.
In this module, we initially wanted to construct a whole plasmid capable of expressing both AlkB2-Rd45-AdhA and CYPY96F as well as VHb proteins. To construct the we needed to attach two long fragments to a vector. At first, we used the enzyme binding method to make the connection, but we tried for a long time without success. It was realized that the method of binding two long f ragments by enzyme digestion was inefficient, and it was difficult to make the vector express three proteins at the same time. We changed the way we connected to homologous recombination and connected one fragment at a time, and we got the engineering success. The effects of AlkB2-Rd45-AdhA and CYPY96F were verified respectively.