Bacteria Engineered to Drop One Amino Acid From Life's Core Alphabet
Every living cell we've ever studied runs on the same 20 amino acids. Researchers just built bacteria that run core cellular machinery on 19 — and the cells survived.
Explanation
Proteins — the molecular machines that do almost everything inside a cell — are built from chains of 20 chemical building blocks called amino acids. This set has been universal across all known life for billions of years. Until now.
A team used AI-guided protein design to systematically eliminate one amino acid from the essential machinery of bacterial cells. The bacteria were re-engineered so that the proteins which would normally require that amino acid were redesigned to work without it. The cells remained viable — meaning life's "alphabet" isn't as locked-in as biology has always implied.
Why does this matter today? Because a cell that runs on a reduced amino acid alphabet is, by definition, chemically isolated from normal biology. Viruses that hijack standard cellular machinery would struggle to replicate inside it. Proteins produced by such a cell could be made "invisible" to natural biological systems — useful for therapeutics that dodge immune responses, or for biocontainment of synthetic organisms so they can't swap genetic material with the wild.
The AI angle is key: brute-force redesign of this scale wasn't feasible before large protein-language models could predict how substitutions would ripple through protein structure and function. This is one of the first real-world demonstrations of AI rewriting the rules of biochemistry at a systems level, not just tweaking a single enzyme.
The immediate caveat: "key machinery" is doing some work in the headline. These aren't fully 19-amino-acid organisms — the reassignment is targeted, not genome-wide. The gap between a proof-of-concept reduction and a fully orthogonal synthetic cell is still large. Watch whether the same approach scales to multiple amino acid removals, or whether each deletion compounds the fitness cost exponentially.
The canonical 20-amino-acid genetic code has resisted natural variation — with minor exceptions like selenocysteine and pyrrolysine in narrow lineages — because the translational apparatus is deeply co-evolved and highly pleiotropic. Reassigning even one codon globally tends to be catastrophic. Prior synthetic biology work (notably the Chin lab's sense-codon reassignment and the Romesberg/Scripps unnatural base pair lines) expanded the alphabet outward; this work compresses it inward, which is a meaningfully different engineering challenge.
The AI-guided redesign component is the technical crux. Protein language models (likely ESM-class or AlphaFold-derivative scoring) were used to identify substitution paths that preserve fold and function across the subset of proteome components targeted — essentially navigating a high-dimensional fitness landscape that would be intractable by rational design or directed evolution alone at this scale.
The biocontainment implications are the most near-term consequential angle. A chassis organism whose core machinery is orthogonal to natural amino acid availability creates a metabolic firewall: escape into the environment means starvation for the synthetic cell, not proliferation. This is a more elegant solution than auxotrophy-based containment (which can be bypassed by environmental complementation) because the dependency is architectural, not metabolic.
Open questions worth tracking: (1) What is the fitness cost, and how does it scale with additional deletions? (2) Which amino acid was removed — the choice matters enormously for downstream applications, particularly if it's a high-frequency residue like leucine vs. a low-frequency one like tryptophan. (3) Can the same AI pipeline handle simultaneous multi-residue reassignment, or does each step require independent model retraining? (4) Does reduced-alphabet translation introduce novel error modes — misfolding, aggregation — under stress conditions?
The result is incremental in scope but significant in principle: it empirically falsifies the assumption that the 20-AA code is a hard biological floor, and it hands synthetic biologists a new axis of chassis orthogonality to exploit.
Reality meter
Why this score?
Trust Layer Score basis
A detailed evidence breakdown is being added. For now, the score basis is the source list below and the reality meter above.
- 1 source on file
- Avg trust 95/100
- Trust 95/100
Time horizon
Community read
Glossary
- selenocysteine
- A rare 21st amino acid incorporated into proteins in certain organisms through a specialized genetic code mechanism, distinct from the standard 20 amino acids.
- pyrrolysine
- A rare 22nd amino acid found in some archaea and bacteria, incorporated through a modified genetic code system separate from the canonical 20-amino-acid code.
- pleiotropic
- Describing a genetic or molecular system where a single component affects multiple different traits or functions, making changes to it have widespread consequences.
- protein language models
- AI systems trained on large datasets of protein sequences to predict how amino acid changes affect protein structure and function, such as ESM or AlphaFold-based models.
- biocontainment
- Engineering strategies that prevent genetically modified organisms from surviving or reproducing outside controlled laboratory environments.
- auxotrophy
- A condition where an organism cannot synthesize a specific essential nutrient and must obtain it from its environment to survive.
- chassis organism
- A genetically engineered microorganism used as a foundational platform for synthetic biology, designed to be modified for specific applications.
What's your read?
Your read shapes future topic weighting.
Your vote feeds topic weights, community direction and future prioritisation. Open community direction
Sources
Optional Submit a prediction Optional: add your prediction on the core question if you like.
Prediction
Will a fully viable bacterial strain with two or more amino acids removed from its core machinery be demonstrated within the next three years?