Every discarded plastic bottle carries a small contradiction. The qualities that made it useful—light weight, strength, resistance to water, and the ability to survive months on a shelf—also help it persist when its brief working life is over. A bottle might hold a drink for an afternoon, yet the material remains a challenge for waste systems long afterward. Recycling can extend its usefulness, but turning used plastic into high-quality new plastic is rarely as simple as putting it into the correct bin.
Now scientists are trying an unusually precise way to approach that problem. Instead of finding stronger machinery to tear plastic apart, they are redesigning proteins that can loosen the chemical bonds holding one common kind of plastic together. A study published in Nature Communications on November 3, 2026, reports experiments with 603 designed versions of plastic-degrading enzymes. More than 400 retained the ability to break down polyethylene terephthalate, better known as PET, and many worked better than the natural enzymes from which they were derived.
The surprising part is not merely that an enzyme can act on plastic. Researchers have known that for years. It is that large sets of changes to a protein's molecular recipe can sometimes improve its performance without destroying its delicate three-dimensional shape. The new work combines information preserved by evolution with deliberate chemical engineering and repeated laboratory tests. It offers a fresh way to search for useful enzymes across a landscape of possibilities too enormous to explore by trial and error alone.
This is not a discovery that makes litter disappear, and it does not mean a plastic bottle left in a park will suddenly biodegrade. The achievement is narrower, and potentially more useful: a method for developing better molecular tools for controlled recycling processes. To see why that matters, we first have to understand what makes PET so difficult to take apart.
Background: the plastic with a very long afterlife
PET is a polyester. Its molecules are long chains assembled from smaller chemical building blocks, linked through bonds known as ester bonds. It appears in beverage bottles, food packaging, and polyester textiles. The same general chemistry can give PET different physical characteristics depending on how the material is manufactured, stretched, cooled, and processed. That difference matters enormously when a recycling system tries to recover it.
Most people know the basic promise of recycling: collect a bottle, process it, and make something new. In mechanical recycling, facilities sort, wash, shred, and remelt suitable plastic into reusable material. This approach is important and, for well-separated PET streams, often practical. But dyes, additives, food residues, other polymers, and repeated heat treatment can complicate the process. Some products are easier to turn into new bottles than others, and not every form of PET can circulate indefinitely without additional treatment or quality controls.
Chemical recycling takes a different route. Rather than preserving the long polymer chains, it aims to break them into smaller molecules that can be purified and used again. Enzymatic recycling is one possible version of this idea. It uses protein catalysts to cut particular chemical connections, ideally recovering valuable components rather than merely producing smaller pieces of waste.
The scale of the challenge explains the urgency. In its Global Plastics Outlook, the Organisation for Economic Co-operation and Development estimated that the world generated about 353 million metric tonnes of plastic waste in 2019. After accounting for losses, just 9 percent was ultimately recycled. That figure refers to plastic waste overall, not specifically to PET or to the performance of an individual country's bottle collection program. It is a sobering indicator of how far current systems remain from a genuinely circular material economy.
Even a highly effective enzyme would not replace collection, sorting, or efforts to reduce unnecessary plastic use. It could, however, create another route for some materials that are difficult to recover at high quality through conventional methods. That possibility has encouraged scientists to treat the plastic molecule itself as a design challenge.
What was discovered: hundreds of working molecular redesigns
The November 2026 study asked a deceptively simple question: how boldly can researchers redesign a PET-cutting enzyme before it stops functioning? A typical enzyme is a chain of amino acids folded into a precise structure. Change a few amino acids, and the protein might work faster, tolerate heat better, or interact differently with a surface. Change too many, and it might misfold, become unstable, or lose activity altogether.
The researchers started from eight different parent PET hydrolases—enzymes that can hydrolyze, or chemically split, bonds in PET. They then used an evolution-informed computational approach, called EVdesign, to propose new amino-acid sequences. Rather than simply changing one position at a time, the approach searches for combinations of changes that resemble the relationships found among related proteins in nature. The study reports 603 experimentally tested designs, with more than 400 variants carrying roughly 12 to 68 substitutions that could still depolymerize PET.
Those numbers require a careful reading. A working enzyme in a controlled assay is not automatically a practical recycling product. Nor does a higher activity measurement mean that an entire bottle was converted into pristine raw materials. The result shows that scientists can generate a remarkably large and diverse collection of functional candidates. It provides options for future optimization, not a finished industrial solution.
One enzyme stood out as particularly welcoming to redesign: TfCut2, originally associated with the bacterium Thermobifida fusca. In the paper's TfCut2-derived group, 399 of 513 designs remained soluble and active, despite extensive sequence changes. The researchers found that this family gave them a productive starting point for further rounds of engineering. Other parent enzymes behaved differently, demonstrating that there is no universal protein template that responds equally well to the same computational strategy.
The team also found that a handful of carefully selected changes could stabilize an enzyme sufficiently to support more ambitious redesigns. Subsequent rounds produced variants that performed strongly across multiple conditions. In 15 of 16 experimental combinations, at least one designed enzyme matched or exceeded the 48-hour product formation of a well-established benchmark, LCC-ICCG, under the particular test conditions used. That is a meaningful laboratory comparison, but it should not be mistaken for proof of superior performance in a commercial reactor.
The broader discovery is a principle about engineering: evolution-guided exploration and deliberate structural improvements can work together. Instead of choosing between nature's accumulated information and a chemist's targeted intervention, researchers can use both to expand the range of proteins worth testing.
How it works: turning an impossibly large search into a manageable one
Imagine a protein as a very long sentence written using an alphabet of 20 amino acids. The order of its letters helps determine how that sentence folds into a three-dimensional object. Some positions are relatively flexible; others are crucial. Two positions far apart in the written sequence may touch after folding, so changing one can alter what works at another. A random edit is therefore difficult to judge in isolation.
Evolution provides clues. Related organisms have accumulated different protein sequences over immense periods. Where certain amino acids consistently appear together, their relationship may reflect constraints that help preserve the protein's structure or function. EVdesign builds statistical models from alignments of homologous proteins—proteins that share evolutionary ancestry—to estimate which new sequences might remain viable.
That does not mean a computer understands every atom's behavior or predicts the perfect enzyme from first principles. It means a model narrows the search toward combinations that appear compatible with patterns observed in existing proteins. The researchers then use the model to sample new sequences while controlling how similar those sequences are to the original enzyme and to one another. Diversity becomes something they can adjust deliberately.
After a promising design is chosen, it still has to enter the physical world. A DNA sequence encoding the new protein must be synthesized or assembled. Cells can then produce the enzyme, which is purified and tested. Laboratory instruments measure characteristics such as thermal stability and the release of PET breakdown products under defined conditions. A prediction becomes evidence only after the protein is made and examined.
For this research, a semi-automated experimental pipeline made it feasible to examine hundreds of candidates. The team measured performance against different PET substrates, including more ordered crystalline material and less ordered amorphous forms, while changing temperatures and acidity. Those variations matter because a recycling process must work on real physical substances, not only on a convenient ideal sample.
Consider two hypothetical enzyme candidates. One may cut relatively soft PET efficiently at a moderate temperature but lose its shape when the process gets hotter. Another may survive high temperatures yet attach poorly to the plastic surface. Neither is simply “better” in every situation. By testing many sequences across multiple conditions, scientists can identify trade-offs and decide what kind of enzyme a particular process actually needs.
The study's importance lies partly in replacing a single winner-takes-all search with an expandable library. A library of active proteins provides material for comparison, further experimentation, and even future models that learn from measured successes and failures.
The hidden difficulty: an enzyme has to reach the bond
A simplified illustration of plastic recycling can make the process look effortless. Picture an enzyme landing on a chain and snipping it neatly into pieces. But solid PET is not an open string floating freely in water. It is a dense material, often containing tightly packed regions where molecular chains are ordered into crystals. The enzyme generally has to work at accessible surfaces, and its success depends on whether the necessary chemical bonds can be reached.
This helps explain why PET films, powders, and bottles do not necessarily behave alike. A finely divided sample exposes a different amount of surface than a thick object. Heating can alter chain mobility and help enzymes work, but the temperatures that improve access may also damage fragile proteins. Some feedstocks need washing, grinding, or other pretreatment before their molecules become accessible enough for efficient enzymatic processing.
Crystallinity is especially important. In amorphous regions, chains are less tightly ordered. In crystalline regions, their arrangement can make attack more difficult. A study highlighted by Spain's CSIC in May 2026 examined molecular obstacles to the enzymatic breakdown of crystalline PET, illustrating why laboratory progress on one type of substrate cannot be generalized to every form of packaging or textile.
Acidity presents another complication. When PET is split, acidic reaction products can accumulate and alter the surrounding pH. Enzymes often operate best within particular pH ranges. Maintaining those ranges may require chemicals and processing equipment. A promising enzyme that performs better under more acidic conditions could potentially reduce part of that burden, but only if the advantage survives realistic reactor testing.
These physical and chemical hurdles are why the 2026 team examined multiple substrate types and operating conditions. The test results should be read as a map of strengths and weaknesses, not a single speed record. An enzyme that wins under one condition might fall behind when the temperature, pH, or PET structure changes.
A clever two-part strategy: stabilize first, explore again
The new paper offers a revealing example of cooperation between two styles of molecular engineering. The first style looks at natural evolutionary patterns to identify plausible combinations of mutations. The second deliberately introduces changes with a known or expected physical benefit. Neither strategy is infallible, but their strengths can complement one another.
In TfCut2, researchers introduced four mutations analogous to a set previously used to improve another PET-degrading enzyme. These are sometimes identified by the shorthand ICCG. The changes markedly improved the heat tolerance of the TfCut2-based proteins tested. In the study's validation experiments, adding these substitutions also improved measured activity, though the magnitude varied across the redesigned proteins and test conditions.
Why is that so interesting? A stable protein can provide room for further experimentation. If an enzyme begins close to its thermal or structural limit, additional mutations may push it into failure. Strengthen its foundation, and some otherwise risky changes become possible. The researchers then applied evolution-informed design again, including to already redesigned sequences, generating successive generations of new variants.
The result was not a smooth climb in which every generation became better in every respect. Some modifications reduced heat tolerance or performance. The best variants emerged as uncommon standouts from a larger population. In fact, the study found that designs with greater sequence diversity tended to include more failures but also created opportunities for unusually strong performers. That is a trade-off familiar from exploration in many fields: staying close to what already works improves the odds of modest success, while venturing farther occasionally reveals a more valuable surprise.
A particularly striking finding concerned two high-performing descendants called TfCut2_EV1_G2 and TfCut2_EV1_G3. They carried dozens of substitutions relative to the original TfCut2, yet showed strong results in different experimental settings. Other heavily changed variants retained nearly the same overall protein backbone structure when examined using X-ray crystallography. The evidence suggests that a protein can sometimes tolerate substantial revisions while keeping its essential fold.
It is tempting to describe this as evolution moving at computer speed. A more accurate description is guided protein engineering: computational models propose sequences, and laboratory measurements determine which ones actually work. Nature supplied the clues; the researchers supplied the objectives; experiments supplied the reality check.
Why it matters: recycling that rebuilds, rather than only reshapes
A PET bottle is valuable not because its original shape is special, but because its constituent molecules can become useful materials again. The appeal of depolymerization is the possibility of returning a used polymer to chemical building blocks that can be purified and repolymerized. In principle, this offers a route to high-quality recycled PET even when an incoming stream is unsuitable for straightforward mechanical reuse.
PET's principal building blocks include terephthalic acid and ethylene glycol. Breaking down the polymer can produce these components, sometimes through intermediate compounds that require additional processing. Recovering and purifying them would allow manufacturers to make PET with properties suited to demanding applications, subject to appropriate quality and safety standards. The ultimate ambition is not simply to grind a discarded bottle into tiny pieces. It is to preserve the usefulness of its underlying chemistry.
There are several reasons that might matter. Polyester fabrics often contain dyes, finishes, and blends that make recycling difficult. Some PET packaging contains multiple layers or is contaminated by other materials. If enzymes and associated processing steps can handle selected challenging streams economically, they might recover value that would otherwise be lost. But the key phrase is “if they can.” Mixed textiles, coatings, pigments, and other polymers introduce problems that the new study did not solve.
An enzyme is also a catalyst, meaning it can facilitate chemical reactions without being consumed in the same way as a stoichiometric ingredient. In a working process it may nevertheless lose activity, become trapped on surfaces, or require replacement. Manufacturers must produce enzymes in sufficient quantities, maintain suitable reaction conditions, and recover products efficiently. The environmental and financial consequences depend on the entire system, not just a protein's activity in a test tube.
Mechanical recycling will remain important where it already works well. It usually avoids the extra step of breaking the polymer all the way back into smaller chemicals and then rebuilding it. Enzymatic depolymerization should therefore be evaluated as a potential complement, not automatically a superior replacement. The right method depends on feedstock quality, energy demand, chemical use, transport distances, product specifications, and local infrastructure.
The deeper benefit of improved enzymes may be flexibility. Different candidates could be matched to different conditions: some for lower temperatures, others for higher temperatures, and still others for particular PET surfaces or pH ranges. A diverse enzyme library offers more ways to adapt processing to actual waste rather than imagining every waste stream behaves identically.
The evidence: what the experiments show, and what they do not
The strongest evidence in the new study comes from the combination of extensive design, protein production, and direct biochemical testing. The researchers did not merely publish promising computer predictions. They produced hundreds of proteins, measured stability, and tested PET depolymerization under multiple laboratory conditions. In some cases, they also examined protein structures to understand whether extensive mutations changed the overall fold.
The comparison with the benchmark LCC-ICCG is encouraging but needs context. The researchers compared product formation over 48 hours in defined, relatively small-scale assays. In 15 of 16 tested conditions, at least one designed enzyme matched or exceeded the benchmark's result. That statement refers to at least one variant per condition, not every candidate. It also does not say a single enzyme was the overall winner in all conditions. A commercial recycler needs reliable, sustained performance at much greater scale and under changing feedstock conditions.
Another limitation is that high relative improvement can coexist with low absolute output. If the starting enzyme barely works under a strongly acidic condition, multiplying that performance several times may still leave a process that is impractical. The authors explicitly caution that absolute activities at the lowest pH remained low even when the relative gains looked impressive. This distinction matters whenever scientific headlines highlight large fold changes.
The study's endpoint measurements also leave unanswered questions about exactly why each variant performs as it does. A greater amount of product after 48 hours could reflect faster chemical turnover, longer enzyme survival, more effective binding to plastic, or a combination of those effects. Time-resolved experiments will be needed to separate them. Researchers will also need to determine whether the enzymes maintain their advantages in concentrated industrial mixtures and with inconsistent real-world waste.
Finally, the paper does not establish that enzyme-based recycling has become economically competitive everywhere or that its environmental footprint is necessarily lower than other recovery pathways. Such claims require specific life-cycle and cost analyses using realistic equipment, inputs, outputs, and local energy conditions. A laboratory breakthrough is an important step, not permission to skip the rest of the evidence.
A decade of progress behind one new paper
The 2026 experiment did not begin from nothing. In 2016, a team led by Shosuke Yoshida described Ideonella sakaiensis, a bacterium found in material collected near a PET recycling site in Japan. The researchers identified enzymes associated with the breakdown of PET and subsequent conversion of an intermediate compound. That finding helped make biological PET degradation a major area of investigation, though the original work involved specific experimental materials and should not be read as proof that ordinary bottles disappear rapidly in the environment.
Scientists subsequently tried many ways to improve PET hydrolases. Some explored naturally occurring enzymes in microbes and compost environments. Others changed individual amino acids to improve heat tolerance or interactions with the polymer. In 2020, a study led by Vincent Tournier in Nature reported an engineered PET depolymerase and demonstrated the potential for recovering monomers that could be used to make new PET. That work helped move the discussion from an intriguing biological observation toward process development.
More recently, the field has begun combining laboratory robotics, molecular modeling, evolutionary statistics, and machine-learning-assisted searches. In August 2026, a study in Advanced Science described using deep learning to identify a different PET-degrading enzyme and testing its activity under physiological conditions. That work addressed another specific research question; it did not prove that introducing enzymes into human bodies is a safe or practical way to remove microplastics.
The November paper adds a distinctive contribution to this history. Rather than reporting only one new champion enzyme, it demonstrates how to produce a broad family of modified proteins and iterate on successful candidates. This may be valuable precisely because no single enzyme will perfectly suit every possible waste stream or recycling plant.
Why the word “plastic-eating” is misleading
The phrase “plastic-eating enzyme” is memorable, but it can blur several different processes. An enzyme is not an animal consuming an object for energy. It catalyzes a chemical reaction. A microbe may use the reaction products in its metabolism, but a purified enzyme itself does not reproduce, make choices, or hunt for litter. Nor can an enzyme necessarily recognize every substance casually called plastic.
PET is only one member of a much larger family of materials. Polyethylene bags, polypropylene containers, polystyrene foam, and many other common plastics have different molecular backbones and chemical properties. An enzyme that cleaves ester bonds in PET cannot simply be assumed to break the carbon-carbon backbone of polyethylene or polypropylene. A news story about PET should not be rewritten as an announcement that scientists have solved “all plastic pollution.”
It is equally important to distinguish breakdown from disappearance. Breaking a polymer into shorter chains or smaller fragments is not the same as fully recovering its monomers, mineralizing it, or removing all potential environmental impacts. A successful recycling process must account for where products, residues, chemicals, and energy go. If a treatment merely turns large objects into dispersed fragments, it has not necessarily delivered an environmental benefit.
The intended setting for these enzymes is also crucial. Controlled industrial equipment can optimize temperature, pH, mixing, and feedstock preparation. A roadside ditch, riverbank, or beach offers none of that control. Even if one enzyme works under certain laboratory conditions, sunlight, dilution, temperature changes, contamination, and inaccessible polymer surfaces can prevent comparable results outdoors. The new research supports better recycling technology, not an excuse to litter.
What happens next: from a promising enzyme to a credible process
The next challenge is to test the most promising variants under conditions that resemble real recycling operations. Researchers need to study the time course of reactions, the total fraction of PET recovered as useful chemicals, and how activity changes as the reaction mixture evolves. They also need to understand which molecular properties explain differences among close relatives. A variant that produces more material after two days might not be the best choice for a reactor designed around a much shorter processing cycle.
Feedstock tests will be especially revealing. Recycled bottles arrive with variable colors, labels, cap materials, residues, and histories of heating or stretching. Polyester textiles add dyes and blended fibers. Before enzyme treatment, facilities may need to remove contaminants, reduce particle size, or adjust the material's crystallinity. Each preparation step adds complexity and expense. A fair performance comparison has to measure that entire chain of operations.
Enzyme manufacturing is another important hurdle. A protein that performs brilliantly but is difficult to produce or purify at scale may be less useful than a slightly slower candidate that can be made reliably and cheaply. The cost of maintaining reaction conditions, handling chemicals, separating monomers, and managing wastewater also belongs in the calculation. Better laboratory activity can create opportunities for cost savings, but actual savings must be measured, not assumed.
Future research may explore mixtures of complementary enzymes, targets beyond PET, and design models trained on the new experimental data. The large collection of working and nonworking variants could help researchers learn more about how protein sequences, stability, and surface interactions relate. That kind of information may eventually benefit other industrial enzymes, although the present study directly tests PET-degrading proteins rather than proving broad applicability across unrelated chemistry.
Independent replication and transparent comparisons will be essential. Strong results should survive testing in different laboratories, on different batches of waste, and at scales larger than the original assays. Environmental assessments should compare realistic alternatives, including reducing waste and using mechanical recycling where appropriate. A technology cannot be called sustainable simply because its most visible component is biological.
Why this discovery is bigger than one bottle
There is an appealing lesson hidden in the experiment. Scientists did not find a magical protein waiting to be switched on. They created a disciplined way to explore how much a complex molecule can be changed while preserving a useful task. The answer turned out to be more flexible than a cautious one-mutation-at-a-time strategy might suggest. Hundreds of redesigned enzymes still worked, and some performed exceptionally well in their measured settings.
That flexibility matters beyond the immediate recycling application because chemistry is full of trade-offs. A catalyst might be active but fragile, selective but slow, or stable only within a narrow range of conditions. Learning how to move through protein sequence space—using evolutionary evidence, targeted modifications, and repeated experiments—gives researchers more ways to navigate those trade-offs. It is not the same as predicting every outcome. It is a better way to decide which possibilities are worth testing next.
For the public, the responsible conclusion is simpler. The familiar bottle remains a material that requires thoughtful design, collection, and management. Enzymatic recycling is promising because it may help recover more of that material's chemical value. The study published on November 3, 2026, brings that goal closer in the laboratory by showing how to build a diverse toolkit of PET-degrading enzymes. But whether those tools become useful at scale depends on work that is still ahead.
The most important change may be in how scientists frame the problem. Instead of asking whether nature already contains one perfect enzyme for recycling PET, they are asking which combinations of natural knowledge and deliberate design can produce a whole range of workable solutions. A used bottle does not become new simply because a protein can cut it. Yet the possibility of carefully taking its molecules apart—and giving those molecules another productive life—is becoming a more testable, more sophisticated idea.
Sources
- Nature Communications (November 3, 2026), “Iterative computational and rational design generates diverse PET hydrolases for plastic depolymerization”: https://www.nature.com/articles/s41467-026-78471-3
- Organisation for Economic Co-operation and Development, Global Plastics Outlook (2022): https://www.oecd.org/en/publications/global-plastics-outlook_de747aef-en.html
- Yoshida et al., Science (2016), “A bacterium that degrades and assimilates poly(ethylene terephthalate)”: https://pubmed.ncbi.nlm.nih.gov/26965627/
- Tournier et al., Nature (2020), “An engineered PET depolymerase to break down and recycle plastic bottles”: https://www.nature.com/articles/s41586-020-2149-4
- Institut de Ciències del Mar (ICM-CSIC), “Molecular hurdles hindering the total recycling of plastic bottles unveiled” (May 26, 2026): https://icm.csic.es/en/news/molecular-hurdles-hindering-total-recycling-plastic-bottles-unveiled
- Wang et al., Advanced Science (August 18, 2026), “Deep Learning-Driven Discovery and Engineering of an Efficient PETase for Depolymerization and Detoxification of PET Microplastics Under Physiological Conditions”: https://advanced.onlinelibrary.wiley.com/doi/10.1002/advs.77163
Body word count: 4,113 words, excluding the title, section headings, category label, photo captions, and sources.
The conversation
Post a Comment