E. coli redesigned to drop isoleucine

Bacteria are built from proteins, and proteins are built from 20 standard amino acids. That list has looked basically untouchable. You can swap codons, add exotic amino acids, or reassign parts of the genetic code — but taking one of the canonical 20 away from essential cell machinery is a much harder trick. A team from Columbia and Harvard has now pushed that line by redesigning the E. coli ribosome so this core machine can run without isoleucine in its protein parts. ### What exactly changed? The news is narrower — and more interesting — than “scientists deleted isoleucine from life.” They did not build a whole bacterium that never uses isoleucine anywhere. They focused on the ribosome, the cell’s protein-making machine, because it is both essential and brutally complex. In E. coli, that meant redesigning ribosomal proteins so the ribosome’s protein components no longer needed isoleucine residues to fold and function. (science.org) ### Why start with the ribosome? Because the ribosome is the worst reasonable place to try this. It is ancient, universal, and packed with constraints. If you can remove one canonical amino acid from proteins in a machine this central, you have a serious proof that the protein alphabet is more flexible than biology textbooks make it seem. That is why outside commentary framed this as a first step toward “19-amino-acid” life rather than a one-off protein engineering stunt. (science.org) ### Why is isoleucine hard to remove? Isoleucine is hydrophobic — it likes to hide inside proteins and help pack their cores. In principle you can swap it for leucine or valine, which are chemically similar. But turns out that naive substitutions break a lot of proteins. In the team’s first pass across 39 essential or highly expressed E. coli proteins, only about 43% of simple isoleucine-to-leucine or isoleucine-to-valine variants still worked in vivo. That failure rate is the whole reason AI design entered the story. (science.org) ### So what did the AI actually do? It was not “AI made a new organism” in the sci-fi sense. The researchers used evolutionary and structure-guided generative models to propose smarter residue substitutions, then screened those designs experimentally. The models helped search a huge design space — basically, not just “replace this amino acid,” but “replace it with what, and in what local sequence context, so the whole protein still behaves?” The paper specifically describes combining sequence-generation methods with structure-aware evaluation to redesign a ΔI ribosome. (science.org) ### How far did they get? Far enough to matter. The team reports redesigning 21 ribosomal proteins at their native genomic locus and replacing all 382 isoleucine residues in that system. That does not mean the whole proteome is now isoleucine-free. But it does mean one of the cell’s most indispensable assemblies can be rebuilt around a reduced alphabet. That is a real threshold crossing. ### Does this make the bacteria weirdly fragile? (science.org) Probably, at least for now. These kinds of deeply engineered cells usually pay a fitness cost. The point was not to beat wild-type E. coli in a growth contest. The point was to show that essential biology can survive more radical redesign than expected. That opens the door to iterative cleanup later — better growth, broader amino-acid reduction, or tighter coupling to recoded genomes. ### Why do synthetic biologists care? Because reducing the amino-acid alphabet could make engineered organisms more isolated from natural biology. If your proteins, ribosomes, or code dependencies drift far enough from the wild type, horizontal gene transfer gets less useful and viruses may have a harder time hijacking the cell. That logic already underpins other recoded-organism work. This result adds a new lever: redesigning the proteins themselves, not just the codons that encode them. (nature.com) ### What is the bottom line? The big idea is not “life now uses 19 amino acids.” Not yet. The big idea is that AI-guided protein design just helped rewrite an essential cellular machine at genome scale. That makes the canonical amino-acid set look less like a law of nature and more like a very old design choice. (science.org) (nature.com)