Evolution in a bottle, but with the pause button
In the long human effort to reconcile industrial civilization with the natural world, a team at the National University of Singapore has quietly expanded what is possible. By enlisting a modified virus as an agent of controlled chaos, their LySE platform can now evolve entire gene clusters in bacteria — not just single genes — opening a path toward microbes that digest plastic, capture carbon, and produce medicines with a precision that nature alone could never achieve on a useful timescale. It is a reminder that some of our most powerful tools for repairing the world may come not from overriding biology, but from learning to steer it.
- Millions of tonnes of plastic accumulate in landfills and oceans each year, and the bacteria capable of breaking them down are too inefficient to matter at scale — until now.
- Previous gene-evolution tools hit a hard ceiling at roughly 8,000 DNA letters, leaving the large, coordinated gene clusters that govern complex cellular work effectively out of reach.
- LySE weaponizes a bacteriophage's own replication errors — mutating DNA at 160,000 times the normal rate — while using that same sloppiness to weaken the virus and keep it from spreading out of control.
- In just five cycles, E. coli evolved with LySE grew 50 percent better on ethylene glycol, a plastic-derived chemical, with every beneficial mutation traceable and transferable to other bacterial species.
- The platform now points toward pharmaceuticals, pollutant breakdown, and AI-designed metabolic pathways that have never existed in nature — optimized by evolution before they ever enter a living cell.
Every year, millions of tonnes of plastic pile up in landfills and drift into oceans. Scientists have long imagined bacteria as a solution — microbes engineered to break plastic into useful chemicals — but the challenge is not swapping out a single gene. It means retuning entire clusters of genes, dozens of instructions that must work in concert, like refitting every station on a factory floor at once.
Researchers at the National University of Singapore have built a tool for exactly that problem. They call it LySE — Lytic Selection and Evolution — and it uses a modified bacteriophage, a virus that naturally infects bacteria, to generate and test thousands of genetic variations across DNA sequences up to 40,000 letters long. That is five times the length previous phage-based methods could handle, large enough to cover most of the gene clusters that govern important cellular processes.
The mechanism is elegant in its controlled recklessness. The engineered virus carries the target genes and copies them with a deliberately error-prone enzyme — one that mutates DNA at 160,000 times the normal bacterial rate. Those same errors damage the virus itself, preventing uncontrolled spread. By adjusting the ratio of virus to bacteria, researchers can toggle between a mutation phase, generating new variants, and a selection phase, testing which variants actually work better.
Assistant Professor Julius Fredens, who led the work at NUS, called it a best-of-both-worlds approach: the speed of continuous evolution without the "cheater" mutations that typically undermine it. In a proof-of-concept, the team evolved a five-gene pathway letting E. coli feed on ethylene glycol, a chemical found in PET plastics. After five rounds, the best strain grew 50.9 percent more efficiently on ethylene glycol alone. Crucially, all changes stayed confined to the target cluster, making the optimized pathway easy to transfer into other bacterial species — a requirement for any industrial application.
The team envisions LySE reaching well beyond plastic degradation, into pharmaceutical synthesis, environmental pollutant breakdown, and carbon capture. Their most ambitious plan is to take metabolic pathways designed entirely by artificial intelligence — routes that have never existed in nature — and use LySE to evolve them into forms that actually function inside living cells. That convergence of computational design and rapid biological evolution, they suggest, is where the platform's deepest potential lies.
Every year, millions of tonnes of plastic end up in landfills and oceans. Scientists have long imagined a solution: engineer bacteria to break plastic down into useful chemicals. But teaching a microbe to do this efficiently is not like flipping a switch. It requires retuning not one gene, but entire clusters of them—dozens of genetic instructions that must work together in perfect coordination, like retooling every station on a factory floor instead of replacing a single machine.
Researchers at the National University of Singapore have developed a tool that could make this kind of large-scale genetic engineering practical. They call it LySE—Lytic Selection and Evolution—and it works by harnessing a modified virus to rapidly generate and test thousands of small genetic variations across long stretches of DNA. Where previous methods could only handle about 8,000 DNA letters, LySE can work with sequences up to 40,000 letters long, large enough to encompass most of the gene clusters needed for important cellular processes.
The system works by exploiting a bacteriophage, a virus that naturally infects bacteria. The researchers engineered the virus to carry a small ring of DNA containing the genes they want to improve. When the phage replicates, it copies these genes using a deliberately error-prone enzyme—one that makes mutations at a rate 160,000 times higher than a bacterium's normal DNA copying system. This intentional sloppiness is the key insight: because the copying enzyme makes so many mistakes, it also damages the virus's own DNA, weakening the phage and preventing it from spreading uncontrollably. By adjusting the ratio of virus to bacteria, researchers can toggle between a mutation phase, where new genetic variations are generated and packed into fresh virus particles, and a selection phase, where those mutated genes are inserted into normal bacteria and tested for improved function.
Assistant Professor Julius Fredens, who led the work at the NUS Yong Loo Lin School of Medicine, described the approach as a "best-of-both-worlds" system. Traditional directed evolution methods are slow but highly controlled; continuous evolution methods are fast but prone to generating "cheater" mutations that trick the test without actually improving the target genes. LySE sidesteps both problems by maintaining precise control while operating at speed.
In a proof-of-concept experiment, the team used LySE to improve a five-gene pathway that allows E. coli bacteria to feed on ethylene glycol, a chemical used in manufacturing PET plastics. After just five rounds of evolution, with progressively less glucose available to force the bacteria to rely on ethylene glycol, the best-performing strain produced 50.9 percent more biomass using ethylene glycol as its sole food source. Sequencing revealed that LySE had modified both the regulatory switches that control gene expression and the protein-coding genes themselves, and each beneficial mutation was confirmed by reintroducing it individually into fresh bacterial hosts.
What makes this result significant is not just the improvement itself, but how the improvement was achieved. Because all the genetic changes remained confined to the target gene cluster, the researchers could easily transfer the optimized pathway into entirely different bacterial species—a critical requirement for any technology intended to work at industrial scale. Without LySE, bacteria attempting to digest plastic tend to mutate their entire genome in search of solutions, accumulating unwanted changes throughout their DNA. LySE constrains evolution to the genes that matter.
The platform's ability to handle gene clusters five times larger than previous phage-based methods opens doors to applications far beyond plastic degradation. The team envisions using LySE to optimize biosynthetic pathways for pharmaceuticals, engineer microbes that break down environmental pollutants, and evolve entirely synthetic metabolic routes for carbon capture. Because the workflow requires only standard laboratory equipment and the simple mixing of virus lysates with cell cultures, the technology is accessible to labs without specialized expertise in phage biology.
Looking forward, Fredens and his team have filed a patent for LySE and are planning to tackle an even more ambitious challenge: taking metabolic pathways designed by artificial intelligence—routes that have never existed in nature—and using LySE to optimize them so they actually function efficiently inside living cells. That convergence of computational design and rapid biological evolution, they suggest, is where the real potential lies.
Citas Notables
Our goal was to create a best-of-both-worlds system: a tool that rapidly evolves large biological pathways while still letting us hit the pause button to control the process and prevent unwanted genetic errors.— Assistant Professor Julius Fredens, NUS Department of Biochemistry
With LySE, we can take AI-designed enzymes and metabolic pathways and rapidly optimise them to work in practice. That is where massive potential lies.— Assistant Professor Julius Fredens
La Conversación del Hearth Otra perspectiva de la historia
Why does it matter that LySE can handle 40,000 DNA letters instead of 8,000?
Because most of the gene clusters that do useful work in cells are bigger than 8,000 letters. A single gene might be 1,000 letters. A pathway that does something meaningful—like breaking down plastic—needs five or ten genes working together. You can't evolve what you can't fit into your tool.
The engineered enzyme is deliberately sloppy. That seems backwards. Doesn't precision matter in biology?
It does, but not the way you'd think. Yes, the sloppiness creates mutations. But it also damages the virus itself, which is the whole point. It keeps the virus weak enough that you stay in control. You're trading precision in copying for precision in the overall system.
So the bacteria don't actually learn to eat plastic better. The researchers are just selecting for the ones that already can.
That's the heart of it. You're not teaching anything. You're creating variation—thousands of slightly different versions of the same genes—and then keeping the ones that work best. Evolution in a bottle, but with the pause button.
Why is it important that the improved genes can be moved into different bacteria?
Because you don't want to deploy the exact strain you evolved in the lab. You want to take what you learned—the optimized genes—and put them into bacteria that are already safe, or already suited to the environment where you'll use them. If the improvements are tangled up with mutations elsewhere in the genome, you can't do that cleanly.
What's the real bottleneck this solves?
Time and control. You could evolve bacteria the old way, but it takes months. You could use continuous evolution and get results in days, but you lose the ability to steer the process and you get cheaters. LySE gives you speed without losing your hands on the wheel.
What happens next?
They want to take pathways that AI has designed—metabolic routes that have never existed in nature—and use LySE to make them actually work in living cells. That's the frontier. Design something on a computer, then use evolution to debug it in reality.