Nearly every organism on the planet uses the same genetic code, corresponding to the same 20 amino acids and giving rise to the seemingly endless protein diversity seen in nature. But proteins can be built from more than just those canonical amino acids.
Researchers in George Church’s lab at Harvard University have developed automated genetic transfer RNA (tRNA) expansion (AGENTEX), which can be used to design entirely custom genetic codes and incorporate nearly any amino acids of a researcher’s choosing into proteins (Nature 2026, DOI: 10.1038/s41586-026-10949-y). The scientists say that could greatly expand the capabilities of protein engineering and synthetic biology.
Our genetic code contains a lot of redundancy: each amino acid is encoded by multiple codons, each made up of three nucleotides. For decades, scientists have attempted to reassign some of those degenerate codons to instead code for noncanonical amino acids, expanding the genetic code.
Although researchers have been able to incorporate hundreds of noncanonical amino acids into proteins, Church says, they’ve been able to reassign only a few codons at a time. With AGENTEX, researchers can create new genetic codes of up to 34 amino acids of their choosing.
The Church lab was able to unlock custom genetic codes thanks to a newly discovered understanding of tRNA biology.
At the 3′ end of all tRNAs is a CCA sequence. That final A nucleotide is the site where an amino acid is attached, and it was previously thought that aminoacyl-tRNA synthetases (aaRSs) would attach an amino acid only to tRNAs with a 3′ CCA sequence.
“I was solidly betting that CCA was the most conserved sequence in the entire planet,” Church says. “You know, it’s present on every tRNA of every organism, every organelle, and I thought, that’s it. It’s carved in stone. Nothing’s going to break it.”
But Felix Radford, a chemical biologist in Church’s group, found that the Escherichia coli aaRSs could charge other 3′ sequences too.
Radford made tRNAs with the middle C in the 3′ sequence swapped to a different nucleotide and put them into an E. coli cell lysate. All alternative tRNA sequences were charged with an amino acid, albeit to varying degrees—3′ CGA sequences worked the best.
Even though the team discovered that CCA isn’t required for aminoacylation, the team found that the sequence is still required for loading a tRNA into the ribosome. Base pairing between the tRNA and ribosomal RNA occurs within that CCA sequence. But ribosomal RNA can be modified to base pair to a non-CCA sequence instead. Radford says this is “good for developing this technology, because you essentially have two ribosomes in parallel, one accepting CCA tRNAs and the other accepting CGA tRNAs.”
Elisabeth Gardiner, chief scientific officer of Tevard Biosciences, which works on suppressor tRNA gene therapies, says via email, “This paper is an early and tantalizing step toward engineering organisms with expanded chemistry and yet another reminder that the ‘universal’ genetic code might be more of a strong suggestion than a law of biology.” Gardiner was not involved in the Church lab’s research.
Church says his lab is working to test custom genetic codes in vivo and to use artificial intelligence to help design new proteins, though he says there isn’t much training data yet for proteins that incorporate noncanonical amino acids.
In the meantime, Radford says, he’s interested in “making biopolymers that incorporate as many monomers as we can, which can have applications to materials, to therapeutics, to new catalysts, new chemistries.”
Max Barnhart is an assistant editor and life sciences reporter at C&EN.
