DRY LAB

1. Introduction

Our investigation for an improved strain of the MHETase begins with the paper published by Cui et al. in which they used the GRAPE strategy to get a better performing variant of PETase, a PET degrading enzyme. Their approach relied on finding out and accumulating the stabilising mutants using ddG as a parameter. Then we found a paper published by Lu et al. in which they used a neural network machine learning model to predict the mutations which can improve the enzyme’s activity and accumulating the proving mutants from Cui et al.’s studies created their version of the PETase, FastPETase which is till date the best performing PETase.

2. PETase

Polyethylene terephthalate (PET) is a largely utilized synthetic polymer that is produced by the decondensation process of terephthalic acid (TPA) and ethylene glycol (EG). It is popularly consumed due to its versatility, reusability, durability and cost-efficiency.(1) However, accumulation of large amounts of PET wastes in the surroundings has caused severe environmental issues.
Ideonella sakaiensis 201-F6 bacteria produced the unique PET-hydrolyzing enzyme known as PETase.(2) In comparison to previous published enzymes, PETase has a considerably greater enzymatic activity and specificity for PET at room temperature (30–40 °C). (3) Consequently, it is a potentially effective biocatalyst for the development of biological PET degradation.. Additionally, the use of developed pelB allowed for up to 1.7-fold greater PETase secretion. PETase hydrolyzes the PET model component, bis (2-hydroxyethyl) terephthalic acid (BHET), PET powder, and PET film more effectively as a result of better PETase secretion. (4)

3. FAST-PETase

Lu et al. used machine learning to design enzyme protein structures and created a new PET hydrolase with high PET degradation capability. (5). They enhanced PET hydrolase performance by synthetically balancing the link between PET hydrolase stability and activity using deep learning neural networks (Fig. 1a). The local chemical microenvironment of amino acids was discovered by the three-dimensional (3D) self-supervised convolutional neural network (CNN), which was then utilised to extract optimised characteristics from the microstructure.

Then discrete probability distributions were added to these optimised structural characteristics. matched to the normal amino acid structure by categorization for predicting enzyme protein structure. The four mutations (S121E, T140D, R224Q, and N233K) with the best catalytic activity and thermal stability were chosen for improvement from a ranking of the predicted enzyme structures.

ThermoPETase's variation with N233K and R224Q added on top of S121E demonstrated the highest performance and was given the designation FAST-PETase for future testing. FAST-PETase's superior adaptability was demonstrated by its ability to degrade more PET and release more PET monomers within 24 hours when compared to other PET hydrolases. (6)

4. MHETase

Mono(2-hydroxyethyl) terephthalate (MHET) is broken down by the enzyme MHETase into terephthalic acid and ethylene glycol. It is a crucial enzyme in the enzymatic breakdown of PET. Despite its degradation efficiency, our project aims to amplify the outcome of its combined effect on PET alongside FAST-PETase. To achieve this amplification, we discovered two approaches that computationally aid in analyzing the evolution of this enzyme post mutation.

Two available approaches:

1.   Machine-learning aided engineering of hydrolases: Utilization of MutCompute24 (https://mutcompute.com), which uses a three-dimensional (3D) self-supervised convolutional neural network (CNN) to find stabilising mutations. This algorithm can accurately predict positions within a protein in which wild-type (WT) amino acids are not optimised for their local environments by training on over 19,000 sequence-balanced protein structures from the Protein Data Bank (PDB). This algorithm learns the local chemical microenvironments of amino acids. It is essentially an in silico thorough mutagenesis scan to use MutCompute to get a discrete probability distribution for the structural fit of each of the 20 canonical amino acids at each site in MHETase. Then, predictions were sorted according to their expected probability (fold change of fit). Numerous anticipated mutations can be produced by employing a stepwise combination technique. Further characterization is possible for variants with enhanced thermostability (measured by protein melting temperature, Tm) and catalytic activity (measured by esterase activity and plastic degradation rates). (5)

2.   The GRAPE strategy: Enhancing the thermostability of enzymes is done computationally using the GRAPE (greedy accumulated approach for protein engineering) method. (7) In order to maximise the exploration of epistatic effects in terms of additivity and/or synergism across sets of mutations while minimising experimental effort, the suggested GRAPE technique combines the benefits of greedy and clustering algorithms.(8)

Methodology

We decided to combine both the approaches. First we got the mutations from the machine learning model i.e. MutCompute and then we ran a ddG value check on the mutants. Variants having the least ddG values among the predicted ones were chosen for further studies with a goal of having a stabilizing mutation. The proving variants were then docked with the ligand MHET (4-[(2-Hydroxyethoxy)carbonyl]benzoate, Pubchem CID: 22062452) using AutoDock Vina. After docking, we performed Molecular Dynamic Simulations to assess the dynamics of the protein-ligand complex.

1. Predictions from MutCompute (ML model)
   The model is available on https://mutcompute.com/ and the results were obtained as a csv file. 6QZ4 was used as an input for the model. It contains the wild type probability and the predicted probability of a residue at specific positions.

   

2. ddG Values
   We used PyRosetta to mutate the structure and calculate the ddG values and exported the results into a csv format. ddG is a measurement of the energy shift between the folded and unfolded forms of a protein (G folding), as well as the shift that occurs when a point mutation is present.

   

   Taking -1.0 kJ/mol as a threshold, we created pair mutations and recorded ddG values for the same. We used -1.0 kJ/mol as a threshold in order to pick single-point mutations that would cause the least amount of energy change, providing a more stable mutation.

   

3. Docking
   We perform molecular docking in order to predict the binding mode of MHETase with MHET polymer. Molecular docking was performed using AutoDock Vina. 6QGA was used for docking and MHET (4-[(2-Hydroxyethoxy)carbonyl]benzoate, Pubchem CID: 22062452) was used as the ligand.
   
   Docking Scores

   

   The conformation was found to be correct when compared to how the ligand binds to the enzyme in the real world.

   1. RMSD
      
      
      1. For Protein
      
      
      
      
      
      
      
      2. For protein-ligand complex
      
      
      
      
      
      
      
   2. RMSF
   
   
   
   
   
   
   

4. MMPBSA
   MMPBSA analysis to find out the free energy of the protein-ligand complex using GB protocol.

   1. WT- -19.53 kcal/mol
   
   
   
   
   
   
   
   2. M433E- -18.88 kcal/mol
   
   
   
   
   
   
   
   3. S491A- -21.60 kcal/mol
   
   
   
   
   
   
   
   4. S196A, T434S- -20.14 kcal/mol

   The above experimental data thus shows that our mutants are stable and efficient compared to the wild type MHETase.

References

1. Sinha, Vijay & Patel, Mayank & Patel, Jigar. (2010). Pet Waste Management by Chemical Recycling: A Review. Journal of Polymers and the Environment. 18. 8-25. 10.1007/s10924-008-0106-7.

2. Yoshida S, Hiraga K, Takehana T, Taniguchi I, Yamaji H, Maeda Y, Toyohara K, Miyamoto K, Kimura Y, Oda K. A bacterium that degrades and assimilates poly(ethylene terephthalate). Science. 2016 Mar 11;351(6278):1196-9. doi: 10.1126/science.aad6359. PMID: 26965627.

3. Taniguchi, Ikuo & Yoshida, Shosuke & Hiraga, Kazumi & Miyamoto, Kenji & Kimura, Yoshiharu & Oda, Kohei. (2019). Biodegradation of PET: Current Status and Application Aspects. ACS Catalysis. 9. 4089-4105. 10.1021/acscatal.8b05171.

4. Shi L, Liu H, Gao S, Weng Y, Zhu L. Enhanced Extracellular Production of IsPETase in Escherichia coli via Engineering of the pelB Signal Peptide. J Agric Food Chem. 2021 Feb 24;69(7):2245-2252. doi: 10.1021/acs.jafc.0c07469. Epub 2021 Feb 12. PMID: 33576230.

5. Lu, H., Diaz, D.J., Czarnecki, N.J. et al. Machine learning-aided engineering of hydrolases for PET depolymerization. Nature 604, 662–667 (2022). https://doi.org/10.1038/s41586-022-04599-z

6. Shijie Yu, Qinghai Li, Yanguo Zhang, Hui Zhou, New possibility for PET plastic recycling by a tailored hydrolytic enzyme, Green Energy & Environment, 2023, ISSN 2468-0257, https://doi.org/10.1016/j.gee.2023.02.00

7. Jinyuan Sun, Yinglu Cui, Bian Wu, Chapter Ten - GRAPE, a greedy accumulated strategy for computational protein engineering, Editor(s): Gert Weber, Uwe T. Bornscheuer, Ren Wei, Methods in Enzymology, Academic Press, Volume 648, 2021, Pages 207-230, ISSN 0076-6879, ISBN 9780128220122, https://doi.org/10.1016/bs.mie.2020.12.026.

8. Cui, Ying-Lu & Chen, Yanchun & Liu, Xinyue & Dong, Saijun & Tian, Yu’e & Yuxin, Qiao & Mitra, Ruchira & Han, Jing & Li, Chunli & Han, Xu & Liu, Weidong & Chen, Quan & Wei, Wangqing & Wang, Xin & Du, Wenbin & Tang, Shuangyan & Xiang, Hua & Liu, Haiyan & Liang, Yong & Wu, Bian. (2021). Computational Redesign of a PETase for Plastic Biodegradation under Ambient Condition by the GRAPE Strategy. ACS Catalysis. 11. 1340-1350. 10.1021/acscatal.0c05126.