📄 bioengineering

Megascale recombination generates millions of catalytically competent PETases

This study demonstrates that megascale recombination of ten bacterial PETase sequences can generate millions of diverse, catalytically competent chimeric enzymes, proving that large-scale sequence recombination is a scalable and effective strategy for protein engineering compatible with modern high-throughput screening.

Original authors: Heinzelman, P., Nisonoff, H., Busia, A., Listgarten, J., Romero, P. A.

Published 2026-09-22

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Heinzelman, P., Nisonoff, H., Busia, A., Listgarten, J., Romero, P. A.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Proteins are the microscopic machines that power life, folding into intricate shapes to perform specific tasks like breaking down food or building cells. For decades, scientists trying to improve these machines have relied on a method of trial and error that involves making tiny, random changes to a protein's code and hoping to find a better version. This approach works well for small adjustments, but it struggles when the goal is to create something entirely new or to explore the vast landscape of all possible protein shapes. A more powerful strategy involves recombination, a process where scientists cut and paste large chunks of DNA from different, naturally occurring proteins to create hybrid versions. While this method can generate massive diversity, a lingering question has been whether these Frankenstein-like hybrids could still function. If you mix too many parts from different sources, does the machine fall apart, or can it still run?

A team of researchers at Duke University and the University of California, Berkeley, set out to answer this by building a library of enzyme hybrids on a scale never before attempted. They focused on a specific type of protein called PETase, which naturally eats plastic, and combined genetic fragments from ten different bacterial versions of this enzyme. By stitching together ten distinct blocks of DNA from these ten parents, they created a theoretical library containing ten billion unique combinations. The central challenge was not just making these billions of variants, but finding the few that still worked. To do this, the scientists used yeast cells as tiny factories to display the proteins on their surfaces. They then introduced a special chemical probe that only sticks to the active center of a working enzyme. If the probe attached, it meant the hybrid protein had successfully folded into a shape with a functional working part, even though it was made of pieces from ten different ancestors.

The results were striking. Despite the fact that each hybrid enzyme contained dozens of changes compared to its closest natural relative, about two percent of the billions of variants in the library still showed signs of being functional. This small percentage translates to an estimated one hundred million working enzymes hidden within the massive library. The researchers then used a sorting machine to separate the working yeast from the non-working ones, enriching the population of active enzymes from two percent to over sixty percent in a single round. This proved that the enzymes did not need to be nearly identical to their parents to function; they could be radically different and still retain their ability to work.

To confirm that these findings were not an artifact of the yeast display system, the team took the most promising hybrids and produced them in bacteria, a different type of cell, to see if they could be purified and tested in a standard lab setting. They screened fifty of the top candidates using a liquid test that measures how well the enzyme breaks down a plastic-like substance. Several of the hybrid enzymes performed as well as, or even better than, the original natural enzyme. The researchers then purified three of the best performers to measure their activity precisely. These purified enzymes, which were mosaics of ten different parents, retained their ability to break down the substrate effectively. The study showed that the working enzymes were not all clones of a single perfect design; instead, they represented a wide variety of different genetic combinations, suggesting that nature has many different ways to solve the same problem.

This work demonstrates that scientists can now generate and screen libraries of proteins that are large enough to match the capabilities of modern high-throughput screening technology. By showing that massive recombination does not destroy function, the study opens the door to exploring vast regions of protein space that were previously too risky to investigate. It suggests that the path to discovering new enzymes or improving existing ones does not require starting from scratch or making tiny, incremental changes. Instead, it is possible to mix and match large pieces of genetic code from diverse sources and still find working solutions. This approach provides a scalable way to generate the massive amounts of data needed to train computer models, potentially leading to the discovery of enzymes that can tackle difficult environmental challenges or perform new chemical reactions that nature has not yet invented.

Technical Summary: Megascale Recombination Generates Millions of Catalytically Competent PETases

Problem Statement
Protein engineering is transitioning toward data-centric disciplines reliant on machine learning and high-throughput screening (HTS) technologies capable of evaluating millions to billions of variants. However, a critical bottleneck exists: most experimental libraries sample only narrow, local regions of sequence space via point mutagenesis, failing to match the scale of modern screening capacities. While recombination-based strategies offer the potential to generate substantially larger sequence changes while preserving structural compatibility, it remains unclear whether highly recombined enzymes can retain catalytic competence when library sizes approach the millions-to-billions regime. Previous recombination libraries have generally remained far below the scale accessible to current HTS technologies, leaving a gap in understanding whether functional diversity can be maintained at megascale.

Methodology
To address this, the authors constructed a "megascale" recombination library using ten bacterial PETases (polyethylene terephthalate hydrolases) as parental sequences.

Library Design: The catalytic domain of Ideonella sakaiensis PETase was partitioned into ten contiguous sequence blocks. These blocks were designed using structure-aware analysis to minimize disruption of intramolecular contacts while maximizing sequence diversity. The ten parental enzymes were selected to span a broad range of sequence diversity (56–78% pairwise identity) while retaining the canonical $\alpha/\beta$ -hydrolase fold.
Assembly Strategy: A modular Golden Gate assembly strategy was employed. Ten parent-derived DNA fragments for each of the ten block positions were synthesized. These were assembled in two stages: first, two "half-libraries" (blocks 1–5 and 6–10) were constructed, each with a theoretical diversity of $10^5$ . These halves were subsequently joined to create a full library with a theoretical diversity of $10^{10}$ unique chimeric sequences.
Screening Platform: The library was displayed on the surface of Saccharomyces cerevisiae. Functional screening utilized a fluorophosphonate (FP) probe that covalently labels reactive active-site serines, a hallmark of serine hydrolase competence. This allowed for the parallel measurement of surface expression and catalytic competence via flow cytometry (FACS).
Downstream Validation: Variants enriched by FACS were transferred to E. coli for soluble expression. Activity was assessed using a soluble chromogenic surrogate substrate (p-nitrophenyl butyrate) in both crude lysates and purified enzyme preparations. Deep sequencing (Oxford Nanopore) was performed on both unsorted and sorted populations to analyze recombination patterns and sequence diversity.

Key Results

High Frequency of Functional Variants: Despite the library containing variants with dozens of amino acid substitutions relative to their closest parents, approximately 2% of the sampled yeast-displayed variants exhibited FP reactivity. Given the theoretical diversity, this suggests the library contains an estimated $10^8$ catalytically competent enzymes.
Successful Enrichment: A single round of FACS sorting enriched the population of FP-reactive variants from ~2% to >60% of the total yeast population.
Sequence Diversity and Recombination Patterns: Deep sequencing revealed that the functional population remained highly diverse, with a median pairwise Hamming distance of 51. No single parental sequence dominated the enriched population; instead, catalytic competence emerged from multiple distinct combinations of parental fragments. Enrichment patterns were distributed across the entire enzyme structure rather than localized to a single region.
Soluble Activity: Secondary screening of 50 enriched variants in E. coli lysates confirmed that multiple chimeras retained substantial esterase activity on p-nitrophenyl butyrate. Three top-performing variants (PETase-R1, R2, R3) were purified and showed enzyme-normalized activity comparable to or greater than the wild-type I. sakaiensis PETase.
Library Quality: Sequencing validation confirmed that >99% of analyzed constructs were full-length chimeras without insertions, deletions, or residual restriction sites.

Significance and Claims
The paper demonstrates that large-scale recombination can generate millions of diverse, catalytically competent enzymes that are compatible with modern high-throughput screening workflows. The authors claim this work establishes recombination-based diversity generation as a scalable strategy for protein engineering, bridging the gap between library design and screening capacity.

Key implications highlighted by the authors include:

Scalability: Recombination can be effectively scaled into the million-to-billion variant regime while maintaining substantial fractions of functional sequences.
Data Generation: The approach provides a foundation for generating large-scale sequence-function datasets that are broadly distributed across sequence space, which is critical for training machine learning models to predict long-range epistatic interactions.
Practicality: The method requires relatively modest experimental infrastructure (standard Golden Gate assembly and synthesized gene blocks) and is extensible to other enzyme families.

The authors remain modest regarding the limitations, noting that the FP probe measures catalytic-serine competence rather than direct PET hydrolysis, and that p-nitrophenyl butyrate is a surrogate substrate. Furthermore, only a small fraction of the theoretical $10^{10}$ space was experimentally sampled. Nevertheless, the recovery of numerous active, soluble chimeras suggests that extensive functional diversity exists within the broader recombination landscape.

Technical Summary: Megascale Recombination Generates Millions of Catalytically Competent PETases

More like this