Overview

In the first round, we prioritized mutation sites by combining sequence homology and structure-agnostic screening. Specifically, BLAST and Multiple Sequence Alignment (MSA) were used to identify non-conserved positions as permissive targets for substitution. We then executed Alanine-Scanning Mutagenesis (ASM) to assess residue essentiality and filter out positions with strong functional penalties. Sixteen positions passed these criteria and were advanced to Site-Saturation Mutagenesis (SSM) to enumerate single-amino-acid variants; in parallel, we incorporated literature-reported beneficial substitutions to construct three candidate mutants rationally. In the second round, we augmented the design with MutCompute, a convolutional neural network predictor of stability effects. These variants constitute our final design set for downstream biochemical and functional evaluation.

Figure 1: Full Drylab Mutagenesis Workflow

TfCut2 5ZOA

Thermobifida fusca cutinase (TfCut2) is a well-characterized enzyme frequently cited in the literature for its ability to degrade a wide range of plastic substrates, particularly polyethylene terephthalate (PET) and polybutylene adipate terephthalate (PBAT). To identify the most suitable enzyme for our system, we adopted a docking-affinity-based screening approach. Among the enzymes tested, TfCut2 consistently showed strong binding performance in AutoDock Vina simulations with PET- and PBAT-related ligands.

However, we recognized that PET and PBAT are long-chain polymers, and docking only their monomeric units would not accurately represent real enzyme-substrate interactions. To better capture the structural complexity of these plastics, we expanded our docking strategy. Inspired by the Beijing United 2022 iGEM team, we tested a series of more representative ligand models, including PET dimers and trimers.

Based on past literature comparing various plastic-degrading enzymes, TfCut2-KW3—a thermostable variant—became a top candidate due to its favorable and consistent docking affinities.

However, further analysis revealed concerns about its structural reliability. The AlphaFold-predicted structure indicated a high level of uncertainty in the conformation of KW3’s C-terminal tail, suggesting it may be highly flexible or disordered. Notably, many predicted ligand-binding interactions were concentrated near this ambiguous region, casting doubt on the validity of the docking results. To resolve this issue, we opted for TfCut2 (PDB ID: 5ZOA), a closely related variant that lacks the uncertain tail. This truncated form retains the core catalytic domain while offering greater structural stability and interpretability for downstream modeling and engineering.

TfCut2 structure — Figure 2. Tfcut2-KW3 Alphafold predicted structure

Rational Design Approach

Finding Conserved Residues: BLAST + MSA

The Basic Local Alignment Search Tool (BLAST) finds regions of local similarity between sequences. The program compares nucleotide or protein sequences to sequence databases and calculates the statistical significance of matches. We BLASTed Thermobifida fusca cutinase (TfCut2; PDB ID: 5ZOA) against the NCBI protein database to identify homologous proteins. This allowed us to find commonly conserved residues in the Thermobifida fusca family.

Multiple Sequence Alignment (MSA) allows for the comparison of multiple biological sequences to detect conserved and variable regions. We performed MSA on TfCut2 across 13 plastic-degrading species, including PETases, LCC, and fungi, as shown in Figure 2. The aligned sequences provided significant insights into the vital residues for the functioning of PET-degrading enzymes, as shown in Figure 3.

Figure 3: The 13 plastic-degrading species

Figure 4: MSA of 13 plastic-degrading species

Alanine-Scanning Mutagenesis

Alanine Scanning Mutagenesis (ASM) is a systematic approach in which all non-conserved residues are replaced with alanine, one at a time. This technique allows us to assess the contribution of individual residues to the protein’s function. Alanine is chosen because its small, non-reactive methyl side chain minimizes steric and electronic disturbances, making it less likely to distort the overall protein structure. Furthermore, it preserves the backbone structure, meaning any observed change in activity can be attributed to the absence of the original side chain rather than a structural collapse.

The outcomes of ASM can be broadly categorized into four groups, which can be understood through an analogy. Imagine you wanted to improve the performance of a car. One conservative strategy would be to change the tires, perhaps switching to a set with slightly better grip. The car might handle corners a little better, but the fundamental machine remains unchanged. These are like residues that, when replaced with alanine, have little effect on binding affinity—safe, incremental modifications that fine-tune without major risk.

On the other hand, making changes to the engine or transmission represents a much riskier but potentially game-changing modification. Such changes could dramatically increase the car’s speed or power output, but they could also cause breakdowns or poor performance. Similarly, mutating residues that are critical for substrate binding may greatly affect enzyme function. These are residues that, when replaced with alanine, cause a substantial drop in binding affinity, often because they are near or directly involved in the binding site. These positions are critical for function, and altering them carries both high risks and the potential for significant performance improvements.

Plastic	Category	Residues
PET	Tires	173, 188, 193, 231, 180, 183, 82, 65
PET	Engine	63, 77, 197, 196, 250, 83, 184, 251
PBAT	Tires	173, 188, 193, 231, 180, 183, 82, 65
PBAT	Engine	63, 77, 197, 196, 250, 83, 184, 251

SSM: Computational Simulation

Site Saturation Mutagenesis (SSM) is a technique used to systematically explore the effects of all possible amino acid substitutions at a single residue. Continuing with the car analogy, SSM is like testing all possible types of tires, then repeating this for the engine and all the other parts deemed interesting during ASM. This exhaustive approach allows us to evaluate which specific substitution at a critical position can optimize the enzyme’s function.

In search of mutants with the potential to degrade multiple plastics, we selected eight key residues—split evenly between PET and PBAT interactions—for SSM. At each position, we generated 20 point mutations for each of the possible amino acids. This comprehensive mutagenesis library enabled us to thoroughly screen the functional impact of over 400 substitutions.

We then filtered these mutants based on their predicted binding affinities and additional rational engineering criteria, such as thermal stability and structural considerations, to identify variants with enhanced substrate interactions and improved catalytic potential.

Figure 5: A graph of all our SSM point mutations by their affinity to PET and PBAT

While PBAT is not quite relevant to our goal of cotton-PET blend degradation, we haven’t narrowed our project into textiles when we performed SSM and finalized our first 3 variants. At the time, we had wanted to lean into Tfcut’s ability to degrade multiple plastics like PET and PBAT. As such, we attempted to balance our mutants and evaluate them on a combination of PET and PBAT affinity. This is why promising mutants such as A65M and T63D were left out.

In the screening process for our final 3 mutants, A65H stood out as the mutation with the highest product (PET*PBAT). It was also promising because of its proximity to the active site (residue Y60). Another promising candidate was H77R, which also stood out due to it having the second-highest binding affinity when you took into account both PET and PBAT docking scores. After being chosen as the highest performing mutations in terms of binding affinity, A65H and H77R were later combined with other novel mutations as well as known mutations from literature to create 3 of our mutants.

Machine Learning Approach

MutCompute is a computational tool that uses artificial intelligence to predict the stability of an amino acid at a specific residue. It is similar to having an engineer scan every part of a building to recommend the strongest possible materials for each location. Instead of testing mutations randomly, MutCompute uses a convolutional neural network trained on massive protein datasets. This means it does not just consider each residue in isolation – it also analyzes how each residue interacts with its neighbors in the folded protein structure.

While our 3 rationally engineered mutants focused on improving enzyme function and binding affinity, MutCompute complements them by focusing on enhancing structural stability.

By predicting mutations that may outperform the folding and thermal stability of TfCut2, MutCompute allows us to design enzyme variants that can withstand more extreme conditions during degradation, such as higher pH levels or temperatures. These harsher conditions can act like a pretreatment, helping to soften or partially break down PET’s crystalline structure, making it easier for the enzyme to access and degrade the plastic efficiently.

The integration of function-focused mutations from the first round with stability-focused mutations from MutCompute ensures that our engineered enzymes are optimized: not only are the enzymes diversified, but they are also highly active and structurally robust under degradation conditions.

Part of MutCompute’s output is something called avg_log_ratio, which stands for average log₂ fold change. This value tells us how much more stable the predicted amino acid is compared to the wild type at that position. A higher positive avg_log_ratio means the predicted residue is significantly better for stability. If it is zero, they are equally stable.

log₂ fold change = log₂(P_predictedAA / P_wtAA)

Figure 6: MutCompute output for every residue in TfCut2

For example, at position 77, MutCompute predicted that Tyrosine (Y) is far more stable than the wild type Histidine (H), leading to a high avg_log_ratio and strong potential improvement.

Figure 7: MutCompute output for residue 77 in TfCut2

However, MutCompute does not consider whether a residue is functionally conserved. Conserved residues are like load-bearing beams in a building – mutating them risks collapse. Unconserved residues are more like decorative panels that can be swapped with minimal impact.

This leads to different strategies. We can choose the aggressive path, mutating residues with the highest avg_log_ratio even if they are conserved, or the conservative path, mutating only non-conserved residues to reduce risk.

In summary, MutCompute gave us an AI-guided blueprint to engineer TfCut2 systematically. It allowed us to design mutations based not only on evolutionary data but also on predicted structural stability, potentially creating enzyme variants with improved thermostability and degradation performance for plastic recycling.

Design Rationale

Six Mutants

Rationally engineered	Deep-Learning Guided
G62A, P193H, S197D	D12S
H77R, D204C, E253C	T234L
A65H, L90A, I213S	Full MutCompute

Constructing Our Novel Mutants

Variant 1: G62A/P193H/S197D

Figure 8: Catalytic triad formed by P193H, S194, and S197D in variant 1

The mutants G62A, P193H, and S197D were selected based on their synergistic potential. Based on our research, we predicted that including a catalytic triad would increase enzyme activity. Mutated residues H193 and D197, together with wild-type residue S194, would form a catalytic triad, which would theoretically enhance catalytic efficiency. G62A is a known mutation from the literature that is located inside the PET-binding groove of TfCut2. It has been shown to decrease MHET inhibition and improve activity on PET film 3.4 fold. We theorized that the combination of a catalytic triad and a well-established mutation would be able to form an effective mutant.

Variant 2: H77R/D204C/E253C

The mutations H77R, D204C, and E253C were chosen for their promising functional effects. H77R exhibited one of the highest overall binding affinities for both PET and PBAT. Meanwhile, the paired mutations D204C and E253C are known to form a disulfide bridge, which enhances ionic bonding and contributes to improved thermal stability of the enzyme. Together, these mutations were targeted to boost both substrate binding and enzyme robustness.

Variant 3: A65H/L90A/I213S

Figure 9: PET docking positions in variant 3 with A65H mutation

The mutations A65H, L90A, and I213S were selected based on their potential to enhance enzyme activity. After some analysis on the different PET docking positions of our mutations, we found that the inclusion of the A65H mutation actually created more docking positions for PET. In the wild type TfCut2. The PET only had one position to dock in. This is significant because MHET is also able to dock close by as shown in figure 7. This means that there is a high likelihood that MHET will actually strongly inhibit the PET degradation. After the inclusion of A65H, the PET is able to dock in many more positions, as shown in Figure 8. This reduces the effects of MHET inhibition for the variant.

Furthermore, both L90A and I213S are established mutations derived from comparisons to LCC, a powerful PET-degrading enzyme found in compost. LCC serves as a model enzyme for guiding mutation strategies in TfCut2 due to its exceptional plastic-degrading capabilities. These naturally compatible mutations were incorporated to leverage LCC’s proven functional insights.

MutCompute Variants

This mutation changes Aspartic Acid to Serine at position 12 and has the highest avg_log_ratio among unconserved residues. While its stability improvement may be smaller than T234L, it carries far less risk of disrupting activity. We chose D12S as the safest MutCompute-based mutation to enhance stability without compromising TfCut2’s function.

MutCompute identified Threonine to Leucine at position 234 as the most stabilizing mutation across all 261 residues. However, this site is conserved, meaning there is a potential trade-off between improved stability and enzyme function. We included this mutant to test whether MutCompute’s top-predicted stability mutation can retain its function despite conservation risk.

Variant 6: Full Mutcompute Variant

Figure 12: Full MutCompute variant with all predicted stabilizing mutations except those at conserved residues

In this variant, we implemented all MutCompute-predicted stabilizing mutations except those at conserved residues. This design tests whether broad structural stability improvements can cumulatively enhance thermostability and degradation efficiency while maintaining function

Results

PET Film

Figure 13. Calculated mass of released TPA (dark gray bars) and the corresponding equivalent mass of degraded PET (light gray bars) in miligrams (mg). These values were calculated from the A260 measurements in (A) using the TPA standard curve from (B). Data are presented as the mean ± standard error (SEM) from three independent replicates.

After multiple rounds of pure PET film degradation, it is quite evident that variant 3 leads the charge with the highest activity in terms of PET weight loss as well as actual UV absorbance at 260 nm. It was followed closely by variant 4, and with the other mutants moderately ahead of the wild-type tfcut. On the other hand, it is equally clear how ineffective variant 6 is.

Part of this result was certainly surprising. Because G62A was such a well-documented mutation, we had thought that it would have a much more evident effect. While it did edge out the wild-type, it certainly wasn’t the 3.4-fold result we were expecting, which hadn’t taken into account the catalytic triad. This leads us to believe that perhaps the addition of a catalytic triad may have done more harm than good.

If we instead looked at the degradation results when we were still in the optimization phase, we would see that variants 4 and 5 are actually able to almost hold their own and outperform variant 3 in some cases. Compared to the wild-type, it definitely suggests that many of the variants are much more stable and are able to function at much lower pH values and higher temperatures than the wild-type. While variant 4 may not be the strongest variant under optimized conditions, the D12S mutation appears to have a significant impact on the thermostability of the variant, enabling it to function optimally at 70°C.

Although the T234L mutation also appears to confer higher thermostability, the tradeoff of mutating a conserved residue is also quite evident. While it definitely performs better at higher temperatures, its overall catalytic efficiency is subpar because the mutation messed with the active site. This is super interesting because it reveals one of the weaknesses of a tool like MutCompute. Mutcompute’s strength is its ability to recognize patterns across millions of proteins. However, this same trait means that it is not able to adapt to the desired function of a specific protein. Using T234L as an example, MutCompute was able to determine that threonine would not be the most stable amino acid at the 234th position. Yet, although threonine is not the most stable amino acid, it seems to play a big role in the catalytic efficiency of the enzyme, something a deep convolutional neural network like MutCompute would be unable to identify. In other words, deep learning tools like MutCompute are sort of like a jack of all trades but master of none.

What's curious is that even though we didn’t mutate any conserved residues for variant 6, all the small changes that slightly affected the active site compound led to a completely ineffective enzyme when it came to PET film degradation.

Pure PET Textile

Moving from pure PET films to pure PET textiles presents new problems. Because of the way polyester is produced, the PET is in a state of much higher crystallinity. This is where we can gather many interesting insights into the different tradeoffs for the different variants. For most of the variants, untreated textiles led to the complete nullification of enzyme activity.

When pretreated, looking at variant 1, something about the catalytic triad has definitely compromised the catalytic efficiency of the variant, completely overshadowing whatever gains G62A added.

Similar to PET film degradation, variant 2 slightly outperforms the wild type.

For our star player variant 3, we hypothesize that while the A65H mutation adds more possible docking configurations, which reduces MHET inhibition, it doesn’t necessarily increase the pet affinity of each new active site. Instead, it creates slightly weaker docking positions. When we are dealing with low crystallinity PET film, the reduced MHET inhibition is well worth the slightly compromised binding affinity. But with the pretreated PET textiles, 2 factors come into play that make this tradeoff a lot worse. Because the activity across all the mutants is significantly lower than the activity on the pretreated textiles, it also means that there isn't as much MHET being produced. Because all the mutants have inherently less MHET inhibition (because there isn’t that much MHET), the benefits of the A65H are greatly diminished. However, from the figure above, Variant 3 is still outperforming the variants pretty significantly. But crystalline PET is simply a lot harder to degrade, the worse binding affinity that results from the A65H mutation just doesn’t cut it anymore (pun intended).

Once again, when we look at variants 4 and 5, we can see how mutating a conserved residue consistently harms Variant 5. Compared with the wild type 5ZOA, it's clear that Variant 4 performs better while Variant 5 performs worse. While Variant 5 is supposed to have a higher overall stability, Variant 4 can consistently outperform it because it can stabilize the enzyme without compromising the active site. The relationship between Variant 4 and Variant 5 really drives home why you are not supposed to touch the conserved residues (even when guided by an omniscient AI overlord).

Variant 6 still lies flat.

Conclusion

We created six novel variants of TfCut2, with a view to (a) increasing the catalytic activity, (b) decreasing MHET inhibition, or (c) reinforcing the structural and thermostability of the enzyme.

Rationally engineered	Deep-Learning Guided
G62A, P193H, S197D	D12S
H77R, D204C, E253C	T234L
A65H, L90A, I213S	Full Mutcompute

One significant caveat for these data is that there’s no clear one-to-one correspondence with the PET film degradation. Because degradation of crystallized and amorphous PET is so different, its optimal conditions may vary significantly. We have intentionally omitted data from cotton-PET blend degradation in this section because the nature of cotton-PET blend degradation introduces more variables that need to be controlled, e.g., textile weaving pattern and leftover cellulase buffer. More information can be found on the results page. Moving forward, after Wiki Freeze, we hope to be able to gather more comprehensive data from both pure PET and cotton-PET blend degradations. With this data, we hope to continue to be able to better understand the abilities of our mutants and the reasons behind them.

References

Beijing_United. (2022). Model. iGEM. https://2022.igem.wiki/beijing-united/model
Anand, P., Nagarajan, D., Mukherjee, S., & Chandra, N. (2014). ABS–Scan: In silico alanine scanning mutagenesis for binding site residues in protein–ligand complex. F1000Research, 3, 214. https://doi.org/10.12688/f1000research.5165.2
Barth, M., Oeser, T., Wei, R., Then, J., Schmidt, J., & Zimmermann, W. (2015). Effect of hydrolysis products on the enzymatic degradation of polyethylene terephthalate nanoparticles by a polyester hydrolase from Thermobifida fusca. Biochemical Engineering Journal, 93, 222–228. https://doi.org/10.1016/j.bej.2014.10.012
Barth, M., Wei, R., Oeser, T., Then, J., Schmidt, J., Wohlgemuth, F., & Zimmermann, W. (2015). Enzymatic hydrolysis of polyethylene terephthalate films in an ultrafiltration membrane reactor. Journal of Membrane Science, 494, 182–187. https://doi.org/10.1016/j.memsci.2015.07.030
Deves Sabari V L, Gokulnath Rajmohan, B, R. S., Srivaishnavi S, Kishore Nagasubramanian, Senthil Kumar G, & Venkatachalam, P. (2025). Improving the binding affinity of plastic degrading cutinase with polyethylene terephthalate (PET) and polyurethane (PU); An In-silico study. Heliyon, 11(2), e41640–e41640. https://doi.org/10.1016/j.heliyon.2025.e41640
Enyoh, C. E., Maduka, T. O., Duru, C. E., Osigwe, S. C., Ikpa, C. B. C., & Wang, Q. (2022). In sillico binding affinity studies of microbial enzymatic degradation of plastics. Journal of Hazardous Materials Advances, 6, 100076. https://doi.org/10.1016/j.hazadv.2022.100076
Li, Q., Zheng, Y., Su, T., Wang, Q., Liang, Q., Zhang, Z., Qi, Q., & Tian, J. (2022). Computational design of a cutinase for plastic biodegradation by mining molecular dynamics simulations trajectories. Computational and Structural Biotechnology Journal, 20, 459–470. https://doi.org/10.1016/j.csbj.2021.12.042
Meng, S., Li, Z., Zhang, P., Contreras, F., Ji, Y., & Ulrich Schwaneberg. (2023). Deep learning guided enzyme engineering of Thermobifida fusca cutinase for increased PET depolymerization. Chinese Journal of Catalysis, 50, 229–238. https://doi.org/10.1016/s1872-2067(23)64470-5
Mrigwani, A., Pitaliya, M., Kaur, H., Kasilingam, B., Thakur, B., & Guptasarma, P. (2022). Rational mutagenesis of Thermobifida fusca cutinase to modulate the enzymatic degradation of polyethylene terephthalate. Biotechnology and Bioengineering. https://doi.org/10.1002/bit.28305
Silva, C., Da, S., Silva, N., Matamá, T., Araújo, R., Martins, M., Chen, S., Chen, J., Wu, J., Casal, M., & Cavaco-Paulo, A. (2011). Engineered Thermobifida fusca cutinase with increased activity on polyester substrates. Biotechnology Journal, 6(10), 1230–1239. https://doi.org/10.1002/biot.201000391
Wei, R., Oeser, T., Schmidt, J., Meier, R., Barth, M., Then, J., & Zimmermann, W. (2016). Engineered bacterial polyester hydrolases efficiently degrade polyethylene terephthalate due to relieved product inhibition. Biotechnology and Bioengineering, 113(8), 1658–1665. https://doi.org/10.1002/bit.25941

Introduction

Our mathematical models were designed to complement the process of engineering TfCut2 mutant variants for PET degradation. We guided wet lab design by identifying optimal conditions for degradation and quantified the effects of mutagenesis on catalytic efficiency by interpreting experimental results.

Mathematical Modelling Objectives:
► Utilize 3D Response Surface Methodology (RSM) from wet lab pNPB assays and PET film degradation assays to determine optimal temperature and pH conditions.

► Apply Enzyme Kinetics Modeling to fit wet-lab experimental data, evaluate enzyme mutant performances, and verify mutant functional design.

3D Response Surface Methodology

Response surface methodology (RSM) is an experimental design used to model the influence of multiple factors and determine optimal conditions with few experimental trials. 3D response surface plots are generated to visualize how well the model fits. Mathematically, RSM models the relationship between a response variable w and multiple independent variables x₁, x₂, ..., x_n through regression analysis. RSM has the ability to evaluate the influence of multiple factors with few experimental trials.

Polynomial regression models can be fitted to experimental data, allowing us to obtain approximations for the response surface. For two factors, the model takes the form:

The two factor RSM model can be written in a matrix form:

R-studio computes the coefficients by using ordinary least squares, which minimizes the sum of squared residuals:

Determining the Optima

Once the algorithm determines the coefficients, the optima can be found using the gradient. By taking the partial derivatives and solving for critical points, we are able to obtain the candidates for optimal temperature and pH.

Then, by using the 2-variable determinant test, the algorithm is able to compute the optimal temperature and pH. If there are at least 3 independent variables, the Hessian is required to obtain the optima.

In our enzyme-substrate experiments, RSM was used to optimize temperature and pH in order to obtain maximized TfCut2 activity. Using RStudio, we constructed 3D RSM for both pNPB assays and PET film assays.

x₁ = temperature factor
x₂ = pH factor

3D RSM: pNPB assays

We simulated TfCut2 enzymatic activity using absorbance value (y) across a range of temperature x₁ and pH x₂ values.

Each dry lab member produced their own R code to analyze the same experimental dataset, which allowed us to validate results and confirm a precise interval. The predicted optima from some dry lab member’s model were compiled, and we calculated the average values to obtain a final prediction.

Dry Lab Predicted Average pH and Temperature — Figure 1: Dry Lab’s Predicted Average pH and Temperature for pNPB assays

Figure 2: Dry Lab’s 3D RSM Plot for pNPB assays

Average optimal pH: 7.216 ± 0.010 SD
Average optimal temperature: 52.663°C ± 0.064 SD
Predicted Maximum OD 405: 0.2246

The resulting 3D surface shows an optimal point represented by the red star, indicating the optimal conditions pH 7.2 and 52°C, suggesting that TfCut2 performs best under slightly basic pH and moderate temperatures.

3D RSM: PET Film assays

We simulated TfCut2 enzymatic activity using absorbance value A260 (y) across a range of temperature x₁ and pH x₂ values. We computed the average of the optimum across the variants to determine the conditions.

A260 = f(Temperature, pH)

RSM Plot Variant 3 — Figure 3: Dry Lab’s 3D RSM Plot for PET Film Assays (Variant 3, 4, All Enzymes)

RSM Plot Variant 4 — Figure 3: Dry Lab’s 3D RSM Plot for PET Film Assays (Variant 3, 4, All Enzymes)

Average optimal pH: ~ 7.5
Average optimal temperature: ~ 51°C

The resulting 3D surfaces show the global optimal conditions pH 7.5 and 51°C. Variants 3 and 4 exhibited the most catalytic activity under the uniform conditions.

Michaelis-Menten Kinetics

TfCut2 first forms an enzyme-substrate complex with PET. Kinetic parameters k_a and k_d describe the forward adsorption and backwards desorption. Once formed, the TfCut2-PET complex is hydrolyzed into products. (Barth et al., 2014)

Figure 4: Enzyme-substrate binding mechanism

The total enzyme concentration e₀ is the sum of the free enzyme [E], active enzyme bound to PET [ES], and inactive enzymes bound to products [EP].

Figure 5: Enzyme concentration mechanism

When TfCut2 binds to intermediate products like MHET (Figure 5), an inactive enzyme-product complex EP is formed, resulting in product inhibition. This interaction is described by inhibition binding constants β₁ and β₂. (Barth et al., 2014)

Figure 6: Enzyme concentration equation

When TfCut2 binds to intermediate products like MHET (Figure 6), an inactive enzyme-product complex EP is formed, resulting in product inhibition. This interaction is described by inhibition binding constants β₁ and β₂. (Barth et al., 2014)

MM Equation

The enzymatic activity of TfCut2 on PET can be approximated by the classic Michaelis–Menten model, which assumes that:

a single substrate is converted into a single product
reaction must be irreversible
enzymes exhibit no allosteric behavior
no end product inhibition

Figure 7: Michaelis-Menten Equation

Initially, v rapidly rises when substrate [S] rises. When the enzyme is saturated, v plateaus to maximum rate V_max. The constant K_m indicates a specific [S] at which the reaction reaches half of V_max, serving as a measure for binding affinity. The two parameters enable valuable comparisons across the different variants.

The process of obtaining wetlab PET film results was supported by the combined application of the Michaelis-Menten kinetic theory and Response Surface Methodology.

Figure 8: Classic Michaelis-Menten Graph

Efficiency Parameters

The quantitative efficiency of TfCut2 degrading PET can be characterized by two fundamental parameters: K, the equilibrium binding constant, and catalytic efficiency. The product formation rate, or the degradation rate, also serves as an essential efficiency comparison.

K is determined by the adsorption and desorption parameters, and it indicates binding strength and stability during hydrolysis. The catalytic efficiency k_cat/K_m is equivalent to the ratio of the turnover number to the Michaelis constant. The turnover number describes how many substrate molecules one enzyme molecule can convert into product per second.

pNPB Kinetics Experiment

pNPB Assay

We used pNPB (p-nitrophenyl butyrate) as a model substrate to mimic the ester bond hydrolysis in PET degradation. Unlike PET, which is insoluble and exhibits crystalline structure, pNPB is a small soluble monomer that releases pNP (p-nitrophenol) upon hydrolysis.

While pNPB does not replicate complexity of PET, it nevertheless provides a basis to confirm verification of enzymatic activity and differentiating catalytic performance between the wild type and engineered variants. pNPB kinetics offers a framework to quantify enzymatic performance, generating insights that contextualize PET assays.

Figure 9: Insoluble, crystalline structure of PET (left), soluble structure of pNPB (right)

In short, this specific process serves as a preliminary assay to confirm the catalytic functions of our recombinantly expressed proteins. Ultimately we aim to distinguish the functional differences among our carefully designed variants.

Experimental Setup

► We prepared 8 different pNP powder concentrations (0 to 70 μM) and measured the OD405 (A405) readings for each concentration using an Elisa Reader
► Plotting the 8 points, we used linear regression fitting and obtained a standard curve of A405 vs pNP concentration
► Simultaneously, using optimal conditions determined by RSM, we completed pNPB substrate controlled hydrolysis with 5 dilutions (1:1, 1:5, 1:10, 1:15, 1:20) and obtained A405 data in a 60 minute interval (12 absorbances each, 5 min intervals)
► Using the standard curve, we converted the A405 data into concentrations then hydrolysis rates
► We obtained kinetic curves, which allowed us to determine parameters V_max and K_m

Materials

Figure 10: 4-Nitrophenyl butyrate (pNPB), 4-Nitrophenol, 99% (pNP), TfCut2 Variants, Buffer (HEPES+CaCl₂), Elisa Reader

Kinetics: Variant Functional Designs

Variant 1

Variant 1 was engineered to form a catalytic triad and reduce inhibition. The catalytic triad was formed by S194 (WT), H193, and D197 (mutations). In PET degradation, we would expect Variant 1’s hydrolysis rates to outperform the wild-type’s as MHET builds up. The mutation G62A aimed to reduce MHET inhibition, a known constriction that substantially decreases overall hydrolysis rates. However, MHET is not a product of pNPB and the inhibition reduction results cannot be applied to PET experiments.

Variant 2

Variant 2 was engineered to enhance substrate binding and structural stability. Mutations H774, D204C, E253C were introduced to create a disulfide bridge, with the goal of improving thermal stability and reinforcing the enzyme structure. We would expect significantly lower K_m values in our experiments than the wild type, as the more rigid and stable enzyme should maintain its function at more extreme temperatures when the wild type begins to denature. At higher temperatures this variant should retain a higher V_max value.

Variant 3

Variant 3 was engineered to reduce MHET inhibition, increase catalytic activity, and mimic LCC enzyme’s high performance. The A65H mutation was intended to act similarly to the G62A mutation in Variant 1, so we expect to see Variant 3’s hydrolysis rates to outperform the wild-type’s as MHET builds up. I213S was tested in past studies, and shown to improve turnover rates. More significantly, the L90A and I213S combination was chosen because it is found in the highly active LCC enzyme, giving our mutant an extremely high performance.

Variant 4, 5, 6

Variants 4, 5, and 6 were designed by using a convolutional neural network Mutcompute (https://mutcompute.com), in an attempt to enhance stability and ultimately improve degradation rates.

Standard Curve Analysis

Figure 11: Standard curve fit of pNP concentration and corresponding A405

The relationship between absorbance and pNP concentration was determined using linear regression:

OD₄₀₅ = 0.0051 [pNP] + 0.0296

[pNP] = (OD₄₀₅ − 0.0296) / 0.0051

Conditions: 35°C and pH 7.5

The OD₄₀₅ absorbance levels were measured at continuous intervals over a 30-minute period. A total of 5 absorbance readings were obtained at equal time intervals, then converted into concentrations using the standard curve equation. This enabled the generation of a fitted model for kinetic analysis.

pNP Results

Figure 12: pNP Formation, Absorbance at 405 nm (OD405) vs. Time (mins)

Under the conditions of 35°C and pH 7.5, we obtained results for five different starting dilutions across all variants.

► At conditions 35°C and pH 7.5, we obtained experimental data across five starting dilutions. Variant 1 mirrored the wild type with no steep activity changes, while Variant 2 showed the fastest initial rate but plateaued early on. Variant 3 behaved similarly to the wild type and Variant 1. Variant 4 showed minimal activity until after 50 minutes when it rapidly increased toward 60 minutes. Variant 5 displayed moderate activity throughout, while Variant 6 remained catalytically dead.

► No variants performed significantly better than the wild type. This does not mean our engineered enzymes aren’t better, rather, the pNPB assay may not fully reflect catalytic improvements because it is only a model substrate. Variant 3 later proved the most effective in actual PET degradation, confirming its superior catalytic performance in the intended functional design.

Inhibition Modelling

Product inhibition happens when an enzyme binds to an intermediate product or final monomer, forming inactive enzyme-product complexes. In order to evaluate and compare how different TfCut2 variants are affected by inhibition, the inhibition parameter i can be calculated; i is inversely proportional to inhibition strength. This factor plays a crucial role in describing enzyme mutant abilities associated with product inhibition, which is not explicitly revealed from hydrolysis rates.

Figure 13: Inhibition Equation

Since MHET accumulates and has a larger inhibition binding constant relative to other products, it is the dominant inhibitor. Other intermediate and end products such as EG and TPA bind weakly to TfCut2, resulting in minimal inhibition.

Mutagenesis: Inhibition Reduction

The G62A mutation helps TfCut2 work better by mitigating MHET inhibition. Replacing glycine with alanine at position 62 of TfCut2’s amino acid code tightens the active site. As a result, MHET cannot linger, allowing enzymes to continue hydrolyzing PET more efficiently without being blocked by its own byproduct. Variant 1 was specifically designed with the mutation, and the A65H mutation in Variant 3 was designed to mirror the effect.

14: Inhibition Illustration (Wild Type vs G62A Mutation)

Modified Michaelis–Menten Equation

As product concentrations increase, inhibition significantly lowers the hydrolysis rate. The method of approximating the degradation rates and parameters must be modified in order to more accurately describe the enzyme performances. Since the original Michaelis-Menten equation does not model enzyme-product inhibition, some modifications must be made in order to incorporate additional parameters.

14: Modified Michaelis–Menten Equation

Additionally, the enzyme-product complex can be expressed where binding and dissociation of the products are accounted for. This makes it possible to simulate dynamic changes in [EP] over time.

15: Modified Michaelis–Menten Equation with Inhibition

Differential Equations Modelling

In addition to parameter fitting, we used differential equations to describe the changing behavior of enzyme–substrate interactions over time. By modelling the rate of change in concentrations of substrates, intermediates, and products, we are able to understand how different mutations impact specific steps in the degradation process of PET textiles.

First Order Rate Equations

► The first equation describes the initial step of PET hydrolysis. The concentration of PET decreases exponentially over time, and the rate of the reaction is proportional to the amount of PET remaining. The k₁ constant represents the rate constant for the conversion of PET into BHET.

► The second equation models the formation and degradation of BHET. BHET is generated from PET degradation and is simultaneously converted into MHET. This balance means that BHET concentration typically rises initially, then reaches a peak, and finally decreases as it is consumed in the following reaction step (Fig 4).

► The final equation represents the formation of MHET from BHET and its conversion into TPA and EG. Like BHET, MHET behaves as an intermediate, increasing as BHET is hydrolyzed and decreasing as it is further degraded.

Langmuir Hinshelwood Model: TfCut2 Adsorption and PET Surface Reaction

► This equation describes how the concentration of the enzyme–substrate complex [ES] changes over time. The first term represents the formation of the complex, and the next two terms represent its breakdown by either dissociation back into free enzyme and substrate or by catalytic conversion of substrate into product.

► This model captures the balance between formation and loss of the enzyme–substrate complex. (Salvestrini, 2017)

References

Barth, M., Oeser, T., Wei, R., Then, J., Schmidt, J., & Zimmermann, W. (2015). Effect of hydrolysis products on the enzymatic degradation of polyethylene terephthalate nanoparticles by a polyester hydrolase from Thermobifida fusca. Biochemical Engineering Journal, 93, 222–228. https://doi.org/10.1016/j.bej.2014.10.012"
Burns, R. A., El-Sayed, M. Y., & Roberts, M. F. (1982). Kinetic model for surface-active enzymes based on the Langmuir adsorption isotherm: phospholipase C (Bacillus cereus) activity toward dimyristoyl phosphatidylcholine/detergent micelles. Proceedings of the National Academy of Sciences, 79(16), 4902–4906. https://doi.org/10.1073/pnas.79.16.4902"
Chen, H.-Y., & Chen, C. (2025). A Study of the Response Surface Methodology Model with Regression Analysis in Three Fields of Engineering. Applied System Innovation, 8(4), 99–99. https://doi.org/10.3390/asi8040099
Hårdin, H. M., Antonios Zagaris, Klaas Krab, & Westerhoff, H. V. (2009). Simplified yet highly accurate enzyme kinetics for cases of low substrate concentrations. FEBS Journal, 276(19), 5491–5506. https://doi.org/10.1111/j.1742-4658.2009.07233
Pirillo, V., Loredano Pollegioni, & Molla, G. (2021). Analytical methods for the investigation of enzyme-catalyzed degradation of polyethylene terephthalate. FEBS Journal, 288(16), 4730–4745. https://doi.org/10.1111/febs.15850
Salvestrini, S. (2017). Analysis of the Langmuir rate equation in its differential and integrated form for adsorption processes and a comparison with the pseudo first and pseudo second order models. Reaction Kinetics Mechanisms and Catalysis, 123(2), 455–472. https://doi.org/10.1007/s11144-017-1295-7
Wei, R., Oeser, T., Schmidt, J., Meier, R., Barth, M., Then, J., & Zimmermann, W. (2016). Engineered bacterial polyester hydrolases efficiently degrade polyethylene terephthalate due to relieved product inhibition. Biotechnology and Bioengineering, 113(8), 1658–1665. https://doi.org/10.1002/bit.25941

TfCut2 Mutant

Our TfCut2 Mutant Library combines literature-reviewed mutations with novel variants we developed through mutagenesis and computational approaches such as MSA, ASM, SSM, molecular docking, and MutCompute. By gathering both established and newly designed mutants, we aim to provide a resource that advances enzymatic plastic degradation research, supports sustainable solutions for the textile industry, and inspires future iGEM teams to explore TfCut2. Given TfCut2's well-documented reliability and potential to degrade plastics like PET and PBAT, our variants were specifically designed to enhance thermostability, reduce product inhibition, and increase weight loss efficiency. Through this contribution, we hope to strengthen the community's toolkit and open new possibilities for engineering next-generation polyester hydrolases.

Conservation between all plastic degrading species

MSA comparison with 1. PETase species

Takes into account top 50 similar species from BLAST results

Unconserved (tested in literature)