Polyethylene terephthalate (PET) is a kind of synthetic polyester widely used in packaging and textiles. It is composed of terephthalic acid (TPA) and ethylene glycol (EG) through ester bond. Its low cost, portability, durability and gas barrier ability are widely used in beverage bottles, product packaging and textile industries.
Despite its many advantages, the production and disposal of PET have significant environmental impacts.
Although PET has high recycling value, the global recycling rate remains low. Many PET wastes fail to enter the recycling system, ending up in landfills or being incinerated, resulting in resource wastage and environmental pollution. The recycling process, involving cleaning, sorting, and reprocessing, requires significant energy and water resources, and recycled PET often does not match the quality of virgin material. Traditional chemical degradation methods, such as high-temperature pyrolysis or solvent treatment, are typically energy-intensive and produce harmful by-products, which do not align with sustainable development goals.
To address the environmental issues of PET, scientists have started researching biodegradation methods, especially the use of enzymes to degrade PET.
1) Discovery of Enzymes: Since the discovery of a PET hydrolase from Thermobifida fusca in 2005. In 2016, Japanese researchers isolated an enzyme called PETase from a bacterium named Ideonella sakaiensis. This enzyme can break down PET.
2) Improvement of Enzymes: Subsequent studies have shown that genetic engineering can enhance the activity of PETase, accelerating the degradation of PET. Our laboratory has made significant progress in the design and directed evolution of PETase.
3) Industrial Application: Currently, some companies are exploring the use of PET-degrading enzymes in industrial-scale PET recycling processes to improve recycling efficiency and quality.
Although the PET degrading enzyme has shown excellent performance in the laboratory environment, we still need to face many challenges in the process of promoting its industrial application. The core issues are concentrated in the following aspects:
1) High Production Costs: Firstly, the high production cost is a big obstacle to the wide application of the PET degrading enzyme. The current production process leads to the high cost of the enzyme, and it is difficult to realize large-scale commercial application. According to the existing literature, even if the genetically engineered PETase is expressed in Escherichia coli, its yield can only reach the level of hundreds of milligrams per liter of culture medium, which is far from the requirements of industrial large-scale application.
2) Low Production Efficiency: Secondly, low production efficiency is another problem to be solved urgently. The existing enzyme expression system is inefficient, which limits the possibility of large-scale production. For example, the expression of MHETase in natural reservoir is low, which affects the efficiency of the whole degradation process.
In order to deal with the above problems, researchers adopted diversified strategies to improve the yield and activity of enzymes. At present, the main strategies to improve enzyme production include:
However, due to factors such as the difficulty of transforming chassis cells and the limitations of expression regulation technology, these transformation strategies have not yet reached the standard of industrial production. Therefore, continuous efforts are needed to further optimize these strategies in order to realize efficient industrial production of enzymes!
Our model is constructed based on two aspects: promoters and signal peptides.
1) Promoters: Our model is a generator-predictor framework. Based on the randomly mutated promoter dataset provided by the experimental group, we used a CNN-LSTM model (Convolutional Neural Network and Long Short-Term Memory Network) to predict promoter expression strength based on sequence data and created a promoter strength predictor. Then, we applied a genetic algorithm to generate mutated promoters and used the previously developed predictor as an evaluation tool to optimize the promoters. Later, we improved the generator by switching from the genetic algorithm to a more efficient and interpretable binomial statistical model.
2) Signal Peptides:Similarly, our signal peptide model follows a generator-predictor structure. Using a large dataset of signal peptides obtained from a library, we trained a predictor using a random forest model and screened a small subset of signal peptides to provide to the wet lab group. After receiving the signal peptide data returned by the wet lab, we fine-tuned the predictor with this small dataset. In addition, we employed a Hidden Markov Model (HMM) as the generator to produce mutants of the best-performing signal peptides from the previous wet lab experiments, which were then evaluated using the predictor.
We successfully constructed a highly efficient expression strain, with expression efficiency improved by x% compared to the original system. The efficient ICCG expression system we developed increased ICCG yield by xx% compared to the initial conditions reported in the literature.