At a glance:
- A new tool allows researchers to design and produce synthetic proteins faster, more safely, and at larger scale than previously possible.
- Each protein can contain up to 34 custom amino acids rather than the naturally occurring 20.
- The work promises to expand the development of novel protein-based medicines and other materials.
- The technology arises from a discovery that overturns decades of understanding of how transfer RNAs (tRNAs) help turn the genetic code into proteins.
Scientists have been hard at work for more than 20 years to instruct biological systems, such as E. coli cells, to produce proteins they don’t naturally do. Success could mean faster, cheaper, innovative solutions for medicine, agriculture, materials science, environmental remediation, and other industries.
The journey so far has been arduous. It seemed that the best — and possibly only — way to achieve the goal was to reprogram an organism’s DNA, but that has proven fiendishly difficult and time-consuming and produces feeble results.
The lab of geneticist George Church at Harvard Medical School and the Wyss Institute for Biologically Inspired Engineering at Harvard University has now devised a simpler, faster, safer, and larger-scale method to make new proteins without requiring genome recoding or even any organisms.
Their tool, called AGENTEX and reported Aug. 26 in Nature, allows researchers to design proteins using up to 34 amino acids rather than the naturally occurring 20; custom engineer two types of building blocks (tRNAs and ribosomes) necessary to make those proteins; and insert the building blocks into a standard lab concoction that contains cell components but no actual cells. There, the components churn out the new proteins without interfering with natural protein-making machinery.
“AGENTEX enables researchers to generate entirely new genetic codes on demand in test tubes and use them at scale to build proteins far beyond what nature has evolved,” said first author Felix Radford, HMS research fellow in genetics in the Church Lab. “This is more rapid and safe than existing methods, as it does not rely on handling living cells or altering their genomes.”
Along the way, the team discovered something unexpected about one of the most fundamental processes of life on Earth: how DNA gets translated into proteins.
“It’s been really amazing to be working in this environment because not only are we developing new, innovative technologies but we’re also discovering new science at the same time,” Radford said.
A suddenly expanded amino-acid alphabet
All natural life is made of proteins, and proteins are built from combinations of just 20 amino acids. Each amino acid is coded by a triplet of DNA or RNA bases called a codon. But there are about three times as many codons as amino acids. For instance, UCU, UCC, UCA, and UCG all code for serine. This redundancy suggests that many codons could be reprogrammed to make something else.
After nine years of effort, the Church Lab demonstrated in 2013 that this could be done, freeing up one codon in E. coli bacteria so the microbes could make whatever new amino acid scientists wanted. It took the lab another 10 years to free up a second codon for E. coli to make other new amino acids. Scientists could then engineer cells to make proteins containing up to 22 different amino acids.
Progress around the world has remained slow and fraught with challenges, including the fact that codon reprogramming hampers other normal cell functions and that genetically engineered organisms need to be kept safely “walled off” from natural life.
AGENTEX takes a leap forward by making 34 codons customizable, allowing researchers to make proteins with up to 34 different amino acids. And it does so in test tubes without touching an organism’s natural DNA.
“It took us a decade per new amino acid added to the code, so this remarkably opens the door to 34 at once and with almost none of the usual collateral damage to the genome,” said senior author Church, the Robert Winthrop Professor of Genetics in the Blavatnik Institute at HMS and founding core faculty and lead of synthetic biology at the Wyss Institute.
“The breakthrough turns protein engineering into something closer to a molecular design and discovery platform,” added Radford. “Thousands of unique molecules can be built, tested, and evolved in parallel, without the years of genome rewriting in living cells that was previously required to add each new amino acid.”
A crucial discovery about tRNAs
Underpinning the work was a discovery that changes decades of understanding about transfer RNAs (tRNAs). These are the molecules that add amino acids one by one into a chain to build a protein, making them crucial for both natural and engineered protein synthesis. There are tRNAs that correspond to each codon in the genome; for example, one matches with UCU and attaches a serine to the chain.
Every tRNA has the genetic sequence CCA on its tail end. Dogma has held that any other sequence in that spot will flag the tRNA as defective. Enzymes won’t give the tRNA its amino acid cargo, and ribosomes — the factories in which proteins are made — won’t allow the tRNA in to deliver it.
To their surprise, and contrary to previous evidence, Radford and colleagues discovered that the first part isn’t true. The enzymes do allow some tRNAs with alternative tail-end sequences to receive amino acids. A tool the team built as part of AGENTEX, dubbed tSCAN, identified nonstandard sequences that work best, including CGA.
“We’ve shown that we can alter one of the most fundamental portions of one of the most fundamental systems found in nature, the protein-synthesis system that has existed pretty much unchanged for billions of years across all organisms, and it’s functional,” Radford said. “The CCA end is much more flexible than people assumed.”
New science, new tool
Revealing that scientists can make functional tRNAs that don’t end with CCA is important because the non-CCA tRNAs won’t be allowed into natural ribosomes when added to cells or cell-component soup. Instead, scientists can engineer ribosomes that only work with tRNAs that have a specific alternative sequence, such as CGA. Those alternative tRNAs can be assigned to carry whatever amino acid researchers want, natural or new.
Scientists have chased after this separate, side-by-side protein processing for years.
AGENTEX — short for automated genetic tRNA expansion — provides a workflow for doing all of these steps from start to finish. It includes software Church’s team wrote, available for free, that can be run on an open-source robot sold by the company Opentrons.
To begin, researchers can use AGENTEX to design and produce tRNAs with different non-CCA end sequences and any of 34 amino acids. AGENTEX can batch-test them in cell-component soups known as lysate solutions and report which end-sequence variations have the most success getting tRNAs’ amino acids properly attached.
In the next stage, AGENTEX can take the most successful tRNAs, design and produce matching ribosomes, and add everything to more lysate solutions. There, the components churn out the brand-new proteins.
If scientists are after something more than protein synthesis, they can also use AGENTEX to run quick tRNA screens before moving into more complex models. It’s not yet clear what else may need to be done for the system to work optimally in cells or organisms, Radford said.
The team envisions integrating artificial intelligence into AGENTEX to further optimize protein design.
An array of applications
Among many potential applications, the results hold promise for developing new therapeutics to combat disease.
A bonus of the work is that tSCAN offers a new way to study tRNAs in general, including mutations known to cause diseases such as diabetes and hearing loss, the authors said. That could likewise lead to better understanding and improvement of human health.
“Trying to understand life at the molecular level and using that knowledge to develop new technologies like AGENTEX is really amazing because you can make a positive impact on the world,” Radford said. “You can use the power of evolution to transform the development of therapeutics, materials, food, pretty much anything.”
Authorship, funding, disclosures
Additional authors are Nayan Sapers, Hana M. Burgess, Lucy Ort, and Bogdan Budnik. Burgess is now at Weill Cornell Medicine Graduate School of Medical Sciences.
This work was funded by the National Science Foundation (grant 2123243), the Department of Energy (grant DE-FG02-02ER63445), and an HMS Dean’s Innovation Award for the Use of Artificial Intelligence in Education, Research, and Administration.
Radford and Church are listed as inventors of a provisional patent application filed through Harvard University (PCT 017868). Church is a founder of companies with related financial interests: GRO Biosciences, enEvolv (Ginkgo Bioworks), and Pearl Bio. Other financial interests of Church are listed here. All other authors declare no competing interests.
Article Title
Automated Prototyping of Genetic Codes
Article Publication Date
26-Aug-2026
