A natural enzyme reads an eight-letter genetic alphabet

Researchers at the University of California San Diego reported on 2 September 2026 in Nature Communications that an unmodified bacterial enzyme reads a synthetic genetic alphabet of eight letters as accurately as it reads the natural four. The work, from Dong Wang’s group, used cryo-electron microscopy to capture E. coli RNA polymerase in the act of recognising and incorporating the added synthetic base pairs.

The alphabet is Hachimoji, which doubles A, T, G and C with four laboratory-made bases. The question that mattered was never whether such letters can be synthesised — that has been done — but whether the machinery inside a living cell will transcribe them without stumbling. Structural evidence that it does removes an assumed obstacle rather than announcing a new capability.

A companion study from the same group, published in PNAS in August 2026, went further still: the polymerase also handled synthetic pairs held together without hydrogen bonds at all, which is the interaction textbooks treat as the basis of base pairing.

Nothing here has been put into an organism. What it establishes is that an expanded genetic code would not need a purpose-built transcription system to be read — the existing one appears to cope.