Multi-objective Engineering of Trimethylamine Monooxygenase for Improved Thermostability and Cofactor Use

This study employs a multi-objective engineering strategy to enhance the thermostability and cofactor flexibility of trimethylamine monooxygenase, successfully identifying variants with improved heat-resistant NADPH activity while revealing the persistent challenge of maintaining NADH function under thermal stress.

The Gist

The Big Problem: The "Smelly Fish" and the "Expensive Battery"

Imagine you have a factory that turns leftover fish parts into delicious, healthy food. But there's a catch: the process creates a terrible, fishy smell (caused by a chemical called Trimethylamine or TMA). Nobody wants to buy smelly food.

To fix this, scientists use a special biological machine called an enzyme (specifically, a Flavin-Containing Monooxygenase). Think of this enzyme as a smell-eating vacuum cleaner. It sucks up the smelly TMA and turns it into a harmless, odorless substance.

However, this vacuum cleaner has two major problems that make it too expensive for big factories:

1. It breaks down in the heat: Industrial factories are hot. This enzyme is like a chocolate bar left in the sun; it melts and stops working when things get too warm.
2. It needs a pricey battery: To work, the enzyme needs a fuel source called NADPH. This is like a premium, expensive brand of battery. There is a cheaper, more common battery called NADH, but unfortunately, this specific enzyme refuses to use it. It's like having a car that only runs on premium gasoline, even though regular gas is everywhere and much cheaper.

The Goal: The "Super-Vacuum"

The scientists wanted to create a Super-Vacuum. They wanted an enzyme that:

• Can survive the heat of a factory (Thermostability).
• Can run on the cheap, common battery (NADH compatibility).

The Journey: Three Attempts to Fix the Machine

The team tried three different strategies to build this Super-Vacuum.

Attempt 1: The "Tinkerer" Approach (Trying to force the cheap battery in)

First, they looked at the "battery slot" of the enzyme. They knew the enzyme used to work with both batteries, but a previous version (made by another team to make it heat-resistant) had lost the ability to use the cheap one.

• The Analogy: Imagine the enzyme is a lock, and the NADPH battery is a key. The heat-resistant version of the lock had been modified so tightly that the cheap key (NADH) no longer fit. The scientists tried to file down the lock to make the cheap key fit again.
• The Result: They managed to get the cheap key to fit a little bit, but in doing so, they broke the lock for the expensive key. The machine became unstable or stopped working entirely. It was a trade-off: you couldn't have both.

Attempt 2: The "Conservative" Approach (Tightening the screws)

Next, they went back to the original, non-heat-resistant enzyme. They tried to make it stronger by only making very small, safe changes—like tightening the screws on a machine without changing the gears.

• The Analogy: They tried to reinforce the engine of a car by only using parts that were already in the car's manual.
• The Result: The engine got slightly stronger, but not strong enough to survive the factory heat. It was too cautious.

Attempt 3: The "AI Detective" Approach (The Winning Strategy)

Finally, they used a powerful combination of tools: Genetic Algorithms (computer programs that evolve solutions like nature does), Evolutionary History (looking at how similar enzymes evolved over millions of years), and AI Language Models (AI that understands the "grammar" of protein sequences).

• The Analogy: Instead of just guessing, they hired a team of detectives.
  • One detective looked at the blueprints (Physics/Rosetta) to see if the structure was solid.
  • Another detective looked at the family history (Evolution/Potts models) to see what changes nature had successfully made in the past.
  • A third detective used AI to predict which changes would make the machine work best.
• The Result: This team found a "Goldilocks" solution. They created a new version of the enzyme (called BSC029) that was incredibly tough. When they heated it up, it didn't melt; it kept working!
  • The Catch: While it was great at using the expensive battery (NADPH) after heating, it still struggled to use the cheap battery (NADH). Only one specific version (BSC025) could use the cheap battery, and even then, only before it got hot.

The "Why": The Secret of the Flipped Battery

Why was it so hard to make the enzyme use the cheap battery? The scientists used super-computers to watch the enzyme in action (like a high-speed movie).

They discovered that the enzyme has a dance floor where the battery enters.

• The Expensive Battery (NADPH): It has a special "handle" (a phosphate group) that grabs onto a specific spot on the dance floor. This forces the battery to stand in the perfect position to do its job.
• The Cheap Battery (NADH): It lacks that handle. Without the handle, it's like a dancer without a partner to guide them. It spins around and often ends up standing upside down (flipped). When it's upside down, it can't do the work.

When the scientists made the enzyme stronger (more heat-resistant), they accidentally made the dance floor even more rigid. This made it even harder for the "handle-less" cheap battery to find the right spot. It got stuck in the wrong position more often.

The Conclusion: A Step Forward, Not a Finish Line

This paper teaches us that fixing a machine is rarely about fixing just one part.

• Stability and Function are linked: You can't just make an enzyme stronger without changing how it moves and interacts with its fuel.
• The "Flip" is the problem: The cheap battery keeps flipping over because it lacks the anchor that the expensive one has.

The Takeaway:
The scientists successfully built a heat-proof enzyme that is better than anything before it. However, getting it to run on the cheap battery while it's hot is still a massive challenge. It's like they built a car that can drive in a blizzard, but it still only runs on premium gas.

What's next?
The paper suggests that maybe we shouldn't just try to fix the car (the enzyme). Maybe we should also invent a new kind of "premium gas" that is cheap and stable, or find a way to give the cheap battery a "handle" so it can dance correctly. It will take a mix of engineering the enzyme and engineering the fuel to solve this completely.

Technical Summary

1. Problem Statement

The industrial processing of marine by-products generates trimethylamine (TMA), a compound responsible for undesirable fishy odors. While Flavin-containing Monooxygenases (FMOs), specifically from Methylophaga aminisulfidivorans (mFMO), can oxidize TMA into odorless trimethylamine N-oxide (TMAO), their industrial application is hindered by two major constraints:

1. Thermostability: Native mFMO lacks sufficient thermal stability for industrial processes.
2. Cofactor Dependency: Native mFMO and its previously engineered thermostable variant (mFMO_20, designed via the PROSS algorithm) rely exclusively on NADPH, which is expensive and chemically unstable. The industrial goal is to switch cofactor preference to the cheaper, more stable NADH.

Previous attempts to engineer mFMO_20 for NADH usage failed; while mFMO_20 retained NADPH activity, its NADH activity dropped to near zero. The study addresses the challenge of simultaneously improving thermostability and restoring NADH compatibility, a classic multi-objective optimization problem where stability and activity often trade off against each other.

2. Methodology

The authors employed a multi-stage computational and experimental workflow combining physics-based simulations, evolutionary metrics, and machine learning:

• Computational Analysis of Binding Modes:

  • Used PELE (Protein Energy Landscape Exploration) Monte Carlo simulations to analyze NADH and NADPH binding in both wild-type (WT) mFMO and the thermostable mFMO_20.
  • Defined two binding orientations: NTA (productive, nicotinamide near FAD) and ADE (non-productive, adenine near FAD).
  • Calculated free-energy surfaces and binding energy gaps to understand why mFMO_20 lost NADH activity.

• Multi-Objective Optimization (MOO) Strategies:

  • Strategy 1 (Targeted Restoration): Applied a Genetic Algorithm (GA) to the mFMO_20 scaffold to bias NADH binding toward the productive NTA orientation. Objectives included minimizing total Rosetta energy (stability) and maximizing the energy gap between productive (NTA) and non-productive (ADE) states.
  • Strategy 2 (Conservative Stability): Applied a PROSS-like workflow to the wild-type scaffold. This restricted mutations to highly conserved residues to preserve intrinsic cofactor plasticity while seeking stability gains.
  • Strategy 3 (Hybrid Multi-Signal Optimization): Expanded the search space using a less restricted library. This approach integrated Rosetta energies (physics), Potts model energies (evolutionary co-variation), and ESM language model scores (statistical fitness from large protein datasets) to guide the GA.

• Experimental Validation:

  • Selected variants were expressed in E. coli, purified, and subjected to heat treatment (45°C for 30 mins).
  • Enzyme activity was measured for both NADPH and NADH using TMA as a substrate, monitoring absorbance at 340 nm.

• Mechanistic Validation:

  • Performed microsecond-scale Molecular Dynamics (MD) simulations (10 replicas of 1 µs each) to quantify the "reactive-frame occupancy" (fraction of time the hydride transfer distance < 2.7 Å).

3. Key Contributions & Findings

A. Mechanistic Insight into Cofactor Switching

• Binding Orientation Shift: Simulations revealed that while WT mFMO stabilizes NADH in the productive NTA orientation, the thermostable mFMO_20 variant shifts NADH preference to the non-productive ADE orientation.
• Loss of Anchoring: The loss of the NADPH 2'-phosphate group (present in NADPH but absent in NADH) removes a critical anchoring interaction. In the rigidified mFMO_20 scaffold, this leads to NADH adopting a flipped, non-catalytic conformation.
• Dynamic Accessibility: MD simulations showed that catalytic activity correlates with the frequency of accessing the reactive geometry (hydride transfer distance < 2.7 Å), not just the static binding energy. mFMO_20 rarely sampled reactive geometries for either cofactor, explaining its low activity.

B. Optimization Results

• Strategy 1 (mFMO_20 Restoration): Attempts to force NADH binding in mFMO_20 via MOO yielded variants with measurable NADH activity, but they completely lost NADPH activity, confirming the difficulty of reprogramming the cofactor pocket without disrupting the catalytic cycle.
• Strategy 2 (Conservative WT): Introducing conserved hydrophobic mutations into WT mFMO preserved NADH activity but resulted in only modest thermostability gains, failing to match the stability of mFMO_20.
• Strategy 3 (Hybrid Multi-Signal): This was the most successful approach. By combining Rosetta, Potts, and ESM scores, the authors identified variants that balanced stability and activity.
  • BSC029: Retained 64% of WT NADPH activity after heat treatment (compared to 20% for mFMO_20), showing superior thermal resistance.
  • BSC025: Uniquely retained detectable NADH activity after heat treatment, a feat not achieved by other variants or mFMO_20.

C. Algorithmic Behavior

• The study observed that the most effective variants emerged in the early generations (10–20) of the genetic algorithm. Later generations accumulated mutations that offered diminishing returns or reduced fitness, highlighting the importance of early Pareto-front selection.
• Integrating evolutionary (Potts) and statistical (ESM) signals proved superior to Rosetta energy alone, as Rosetta scores did not correlate well with experimental thermostability in this context.

4. Significance

• Proof of Concept for Multi-Objective Engineering: The paper demonstrates that balancing stability and cofactor specificity requires a multi-perspective approach. Relying solely on physics-based scoring (Rosetta) or single-site conservation is insufficient; integrating evolutionary constraints (Potts) and language model priors (ESM) is crucial for navigating the complex fitness landscape.
• Mechanistic Understanding of Trade-offs: It elucidates that cofactor specificity is not just a static structural feature but a dynamic property governed by the interplay between scaffold rigidity, cofactor orientation, and the accessibility of reactive states.
• Industrial Relevance: While a perfect "NADH-compatible, highly thermostable" variant was not fully realized (only one variant retained trace NADH activity post-heat), the study provides a robust framework for future engineering. It suggests that future success may require combining enzyme engineering with cofactor analog engineering (e.g., using chemically stabilized NADP+ analogs) to overcome the inherent structural constraints of the enzyme.

In conclusion, this work advances the field of enzyme design by showing that successful multi-objective optimization requires a hybrid computational strategy that respects both biophysical laws and evolutionary history, while acknowledging the intrinsic dynamical constraints of cofactor binding.