MPNN-guided redesign of PET hydrolases with enhanced catalytic activity below the PET glass transition temperature

This study utilizes ProteinMPNN and LigandMPNN to redesign the PET hydrolase PHL7, successfully generating variants (notably D5) that achieve high depolymerization activity at 50°C with improved expression and stability, thereby enabling efficient plastic recycling and potential virgin PET resynthesis under mild, energy-efficient conditions.

The Gist

The Big Picture: The Plastic Problem and the "Lazy" Cleaner

Imagine the world is drowning in plastic water bottles. We need a way to turn that trash back into new, clean plastic without melting it down in a hot, energy-guzzling furnace. Nature has a solution: enzymes. Think of enzymes as tiny, biological scissors that can cut plastic chains apart.

One specific enzyme, called PHL7, is a superstar. It's really good at cutting plastic, but it has a major flaw: it's a "heat freak." It only works well when the plastic is almost melting (around 70°C or 158°F).

• The Problem: Heating plastic to that temperature takes a lot of energy and money. Also, PHL7 is hard to make in the lab (low yield) and it falls apart quickly if it gets too hot.
• The Goal: The scientists wanted to redesign PHL7 so it could work like a champion at a much cooler temperature (50°C), making the recycling process cheaper and greener.

The Strategy: Using AI as a "Genetic Architect"

Instead of randomly guessing which parts of the enzyme to change (which is like trying to fix a watch by throwing random gears at it), the team used Artificial Intelligence.

They used two AI tools, ProteinMPNN and LigandMPNN. Think of these as highly advanced "Genetic Architects."

• They looked at the blueprint of the original enzyme.
• They asked the AI: "If we keep the scissors (the active site) exactly the same, but change the handle and the body of the tool, can we make it stronger, easier to build, and able to work in cooler water?"

The AI generated 36 new designs. It was like asking a chef to create 36 new versions of a recipe, keeping the secret spice the same but changing the cooking method.

The Results: The "Goldilocks" Variants

The team built these 36 new enzymes in the lab. Here is what happened:

1. The Yield Boom: Most of the new designs were much easier to produce. While the original enzyme was like a shy plant that only grew a few leaves, some of the new designs grew like weeds, producing 120 times more enzyme in the same amount of time. This means the cost to make the enzyme dropped dramatically.
2. The Temperature Shift: This is the magic part. The original enzyme (PHL7) was a "sprinter" that ran fast but only in hot conditions. The new designs, specifically D5 and D11, were "marathon runners."
   • At high heat (70°C), the new designs actually did worse than the original because they were a bit more fragile (less stable).
   • But, at a cooler 50°C, the new designs were superstars. They broke down plastic just as well as the original did at 70°C.

The Analogy: Imagine the original enzyme is a race car that needs a super-hot engine to go fast, but the engine overheats and breaks. The new designs are like a hybrid car. They don't need the super-hot engine; they run perfectly on a cooler, more efficient setting, getting you to the finish line just as fast without burning out.

Why Did It Work? The "Loose Joints" Theory

The scientists wanted to know why the new enzymes worked better in the cold. They ran computer simulations (Molecular Dynamics) to watch the enzymes move.

They found that the new designs had more flexible "joints" (loops) near the cutting site.

• The Original: Was like a stiff, rigid robot. It needed a lot of heat to get its stiff joints moving so it could grab the plastic.
• The New Design (D5/D11): Was like a flexible dancer. Because the "joints" were looser, it could wiggle and grab the plastic easily even in cooler water.

However, there is a trade-off. Being flexible means being less stable. If you heat the new enzyme up too much, it gets too wiggly and falls apart. But since the goal was to work at cooler temperatures, this "instability" was actually a feature, not a bug!

The Bonus: A Better Way to Recycle

When these enzymes cut plastic, they produce two main things:

1. TPA: The basic building block.
2. MHET: A half-finished building block.

The new enzymes (D5) produced more MHET than the original. Why does this matter?

• Old Way: You have to break MHET down into TPA, then rebuild it. It's like taking a Lego house apart to the individual bricks and rebuilding it.
• New Way: You can use MHET directly to rebuild the plastic. It's like taking a Lego house apart into big chunks and snapping them back together. This saves energy and is a more efficient "circular economy."

The Bottom Line

This paper shows that by using AI to redesign enzymes, we can create "super-cleaning" tools that:

1. Are cheaper to make (high yield).
2. Work at lower temperatures (saving energy).
3. Produce better building blocks for recycling.

It's a huge step toward a future where we can recycle plastic bottles into new bottles efficiently, without needing massive, energy-hungry factories. The scientists didn't just find a better enzyme; they found a smarter way to design life itself to solve our trash problems.

Technical Summary

1. Problem Statement

The enzymatic depolymerization of Polyethylene Terephthalate (PET) is a promising route for sustainable plastic circularity. However, industrial viability is hindered by the limitations of naturally occurring PET hydrolases (PETases). Specifically, the enzyme Polyester Hydrolase Leipzig 7 (PHL7), while effective at high temperatures (~70°C), suffers from:

• Poor recombinant expression: Low protein yields in E. coli.
• Thermal instability: Rapid inactivation at temperatures above 60°C, despite a melting temperature ($T_m$) near 79°C.
• Temperature constraints: It requires temperatures near the PET glass transition temperature ($T_g \approx 70^\circ\text{C}$) for efficient activity, which is energy-intensive.
• Activity-Stability Trade-off: Natural enzymes often lose activity when engineered for stability, or vice versa.

The goal was to redesign PHL7 to improve protein expression, maintain or enhance catalytic activity, and enable efficient PET degradation at milder temperatures (below 60°C) to reduce energy costs and enable closed-loop recycling.

2. Methodology

The authors employed a deep learning-guided inverse folding strategy combined with evolutionary information to redesign the PHL7 sequence.

• Computational Design:

  • Tools: Used ProteinMPNN and LigandMPNN for fixed-backbone sequence redesign.
  • Input Structure: Crystallographic structure of TPA-bound PHL7 (PDB ID: 8BRB).
  • Constraints:
    • Active Site: Residues within 6 Å of the bound TPA ligand were fixed to preserve catalytic geometry.
    • Evolutionary Conservation: 16 Multiple Sequence Alignments (MSAs) were generated using hhblits against UniRef30. Residues with high conservation (30%, 50%, 70% cutoffs) were identified and fixed to maintain structural integrity.
  • Generation: Generated 1,176 sequences (588 per model).
  • Filtering: Structures were predicted using ESMFold and filtered by confidence metrics (pLDDT > 85, C$\alpha$-RMSD < 1.5 Å) and energy minimization (RosettaFastRelax, total energy < -200 kcal/mol).
  • Selection: Sequences were clustered using Levenshtein distance, resulting in 36 representative variants for experimental testing.

• Experimental Screening:

  • Expression: Codon-optimized genes were expressed in E. coli (BL21(DE3)) with a CyDisCo system to facilitate disulfide bond formation.
  • High-Throughput Purification: Micro-scale (2 mL) purification followed by yield quantification.
  • Characterization:
    • Structure: Circular Dichroism (CD) for secondary structure.
    • Stability: Differential Scanning Fluorimetry (DSF) for melting temperature ($T_m$).
    • Activity: Screening against amorphous PET films and Polycaprolactone (PCL) nanoparticles at 50–80°C.
    • Kinetics: HPLC quantification of degradation products (TPA, MHET, BHET) and kinetic parameter calculation ($k_{cat}$, $K_M$) using inverted Michaelis-Menten equations.
  • Mechanistic Analysis: Molecular Dynamics (MD) simulations (200 ns trajectories at 50, 60, and 70°C) to analyze active site flexibility and conformational dynamics.

3. Key Contributions

• Development of High-Yield Variants: Successfully redesigned PHL7 to achieve significantly higher protein expression yields in E. coli compared to the wild type.
• Shift in Optimal Temperature: Identified variants that exhibit superior catalytic performance at 50°C, a temperature where native PHL7 is largely inactive or unstable over long durations.
• Mechanistic Insight: Demonstrated that reduced thermal stability (lower $T_m$) can be strategically traded for increased local flexibility in the active site, enabling efficient catalysis at lower temperatures.
• Product Profile Control: Engineered variants that favor the production of mono-(2-hydroxyethyl) terephthalate (MHET) over terephthalic acid (TPA) at lower temperatures, facilitating more energy-efficient direct repolymerization routes.

4. Key Results

• Expression Yields:

  • 31 out of 36 designed variants were successfully purified.
  • Most variants showed significantly higher yields than wild-type PHL7 (average 5 mg/L vs. 275 mg/L for designs).
  • Two top variants, D5 and D11, showed 7-fold and 13-fold yield increases, respectively, upon scaling to 1 L cultures.

• Thermal Stability vs. Activity:

  • Stability: All active variants (D5, D11) exhibited lower thermal stability than wild-type PHL7 ($T_m$ dropped from ~82.6°C to ~72.8°C for D5 and ~70.7°C for D11).
  • Activity: Despite lower $T_m$, D5 and D11 showed enhanced PET depolymerization rates at 50°C and 60°C.
  • Performance: At 50°C, D5 achieved the same product yield as wild-type PHL7 at 70°C after 24 hours.
  • Kinetics: While wild-type PHL7 had faster initial turnover rates at 50°C, it rapidly inactivated. D5 and D11 maintained activity longer, resulting in higher total degradation over 24 hours.

• Product Profile:

  • At 50°C, D5 and D11 produced a higher ratio of MHET to TPA compared to wild-type PHL7. MHET is a preferred intermediate for direct repolymerization into virgin PET, offering a more efficient circular economy pathway.

• Mechanistic Findings (MD Simulations):

  • MD simulations revealed that D5 and D11 possess enhanced local flexibility in key active site loops (W-loop, D-loop, H-loop) at 50°C compared to the rigid wild-type.
  • This increased flexibility facilitates substrate binding and catalysis at lower temperatures but leads to partial denaturation at higher temperatures (explaining the activity loss at 70°C).
  • Structural analysis showed a reduction in complex salt-bridge networks in the designs, correlating with the observed decrease in $T_m$.

5. Significance

• Energy Efficiency: By enabling efficient PET degradation at 50°C, this work reduces the energy input required for industrial biorecycling, avoiding the need to heat PET to its glass transition temperature.
• Circular Economy: The shift toward MHET production at mild temperatures supports emerging MHET-initiated repolymerization strategies, which are more energy-efficient than traditional TPA/EG condensation.
• Design Paradigm: The study validates a "stability-for-activity" trade-off strategy. It demonstrates that deep learning tools (ProteinMPNN/LigandMPNN) can successfully navigate sequence space to create enzymes optimized for specific operational conditions (mild temperatures) rather than just maximum thermostability.
• Scalability: The high expression yields of the redesigned enzymes address a major bottleneck in the economic viability of enzymatic plastic recycling.

In conclusion, this work provides a blueprint for engineering biocatalysts that balance expression, stability, and activity to enable sustainable, low-energy plastic recycling.