AI alters E. coli ribosomes

Columbia University researchers have reported an E. coli strain whose ribosome runs without isoleucine in a set of engineered ribosomal proteins, using generative artificial intelligence to design replacements that preserved function. The work appears in *Science* in a paper titled “Toward life with a 19–amino acid alphabet through generative artificial intelligence design.” The authors said no known free-living organism uses fewer than the standard 20 canonical amino acids, making the experiment a direct test of whether cells can operate with a reduced protein alphabet. (science.org) The team’s engineered strain, called Ec19, remained genomically stable for more than 450 generations in laboratory passage, according to the paper. (wanglab.c2b2.columbia.edu) ### Which amino acid did the researchers try to remove, and why that one? The *Science* paper identifies isoleucine as the amino acid targeted for elimination from the redesigned proteins. The authors wrote that evolutionary sequence analysis and protein modeling suggested isoleucine was among the least-conserved amino acids and could often be substituted by chemically similar residues such as valine. The researchers said that possibility made isoleucine a candidate for testing whether part of the cell’s protein machinery could be rebuilt with a smaller alphabet. The paper frames that as a way to probe biochemical constraints on early evolution and to explore simpler synthetic cells. (science.org) ### What part of E. coli did they redesign first? The team focused on the ribosome, the molecular machine that makes proteins, and specifically on all 52 essential E. coli ribosomal proteins. The paper says those proteins form one of the most conserved and indispensable complexes in the cell, making the ribosome a stringent test case. (science.org) Science’s research summary says the redesign replaced all 382 isoleucine residues across the ribosomal protein set. The authors then combined 21 isoleucine-free ribosomal subunits into a single genomic locus in E. coli to create the Ec19 strain that supported cellular growth. (science.org) ### How did artificial intelligence enter the experiment? The paper says simple one-for-one swaps of isoleucine with valine or leucine worked poorly at first. In 39 essential or highly expressed proteins tested, only about 43% of those straightforward variants remained functional in vivo, according to the *Science* summary. (science.org) The researchers then used sequence-based language models — ESM2 and MSA Transformer — alongside structure-based models ProteinMPNN and AlphaFold2 to propose alternative isoleucine-free sequences. The authors said those models generated de novo variants intended to preserve protein structure and function while removing the targeted amino acid. (science.org) ### Did the altered bacteria survive and keep the changes? Ec19 supported cellular growth after the 21 engineered ribosomal subunits were installed, the paper says. The *Science* summary reports that the redesign-and-replacement process maintained relative cellular fitness at more than 90% of wild type during the ribosomal protein work. (science.org) The engineered strain then remained genomically stable for more than 450 generations of continuous passage in the laboratory, with no isoleucine reversion mutations identified by whole-genome sequencing, according to the paper. That result is one of the clearest indicators that the redesigned ribosomal system did not immediately revert to the standard amino-acid set. (science.org) ### Does this mean the whole bacterium now uses only 19 amino acids? The paper does not say the entire E. coli proteome has been rewritten to eliminate isoleucine. Wang Lab describes Ec19 as a “major advancement” produced by removing all isoleucines in the E. coli ribosome, and separately says the broader project aims to make life with a 19-amino-acid alphabet. (science.org) That makes the reported bacterium a partial but concrete step: the ribosome itself was redesigned, while the larger project remains the construction of a strain using 19 canonical amino acids across the organism. The authors present the ribosome-first approach as an experimental platform for that next stage. (science.org) ### Where was the work published, and what comes next? *Science* published the paper in late April or early May 2026 under DOI 10.1126/science.aeb5171, and Wang Lab posted a notice on May 1 saying the lab had published work on making life with a 19-amino-acid alphabet. Science Magazine’s social-media post helped circulate the findings again on May 14. (wanglab.c2b2.columbia.edu) Wang Lab says the Ec19 project is aimed at “make life with an 19-amino acid alphabet” and identifies generative AI protein design, gene synthesis and genome engineering as the tools for the next steps. The named participants on the paper include Liyuan Liu, Charlotte Rochereau, Simon Kozlov, Guillaume Urtecho, Sergey Ovchinnikov and Harris H. (wanglab.c2b2.columbia.edu) Wang. (wanglab.c2b2.columbia.edu)