Eight Letters Instead of Four: A Bacterial Enzyme Reads Synthetic DNA

All life on Earth writes its genetic information with the same four chemical letters: A, T, C and G. For decades, synthetic biologists have asked whether that alphabet could be enlarged without jamming the cellular machinery that reads it. The hachimoji system, developed in Steven Benner’s laboratory, adds two artificial pairs of letters, P:Z and B:S, designed so that they fit into the double helix with the same shape as natural pairs. The open question was whether the enzymes that transcribe genes inside cells would accept them.

A study published on September 2, 2026 in Nature Communications by Qingrong Li, Dong Wang and colleagues shows that they do. Using the RNA polymerase of Escherichia coli — the enzyme that copies DNA into RNA — the authors found that it reads all eight letters and picks the right partner every time, without mixing up the two artificial pairs. The synthetic letters were copied almost as fast as natural ones, at roughly half the speed of a standard G:C pair, and the enzyme then carried on transcribing normally. Four cryo-electron microscopy images, resolved at close to atomic detail (between 2.42 and 2.75 ångström), show why: inside the enzyme’s active site, the artificial pairs look and behave like ordinary ones.

The most interesting result concerns accuracy. One artificial letter, Z, carries a chemical group that makes it lose a proton at the pH of a living cell, and in that altered form it resembles C closely enough to pair by mistake with G. Replacing that group with another one almost eliminates the error, but also slows the enzyme down, because the same group holds in place a single water molecule that helps the enzyme close around the incoming letter. Accuracy and speed, in other words, depend on the same small piece of chemistry.

Some caution is due. The work was done in a test tube, on purified enzymes and short DNA fragments rather than in living bacteria, and the assays were designed to provoke errors, so the true error rate is probably lower. Furthermore, turning eight-letter RNA into protein has not been achieved yet. Even so, six-letter DNA molecules have already been built to carry drugs to liver cancer cells, and extending the same chemistry to RNA could produce new sensors, diagnostic tools and therapeutic oligonucleotides. The next question to answer is whether a living cell tolerates the expanded alphabet.