Researchers at the University of California San Diego have shown that a bacterial enzyme central to gene expression can accurately read a genetic alphabet of eight letters rather than the four used by every organism on Earth. The work was published in Nature Communications on Sept. 2 and was led by Dong Wang of the UC San Diego Skaggs School of Pharmacy and Pharmaceutical Sciences. It combined transcription assays with four cryo-electron microscopy structures at resolutions between 2.42 and 2.75 angstroms to show how Escherichia coli RNA polymerase handles synthetic base pairs.

The alphabet in question is called Hachimoji, part of what its designers term the Artificially Expanded Genetic Information System. As the UC San Diego announcement explains, it adds two engineered pairs, P with Z and B with S, to the familiar A, T, G, and C. The authors write that their results establish the feasibility of an eight-letter genetic alphabet for transcription.

What the study did not do is equally important, and the paper says so directly. This was purified enzyme in buffer with short synthetic DNA and RNA scaffolds, not a living organism. Translating these expanded codons into protein has not been achieved. Readers should hold both halves of that at once.

Making a Base Pair Is Not the Same as Getting a Cell to Read It

Synthetic base pairs are not new. Chemists have been building them since the late 1980s, and the eight-letter Hachimoji system was described in Science in 2019. Earlier work showed these pairs are well accepted by Taq DNA polymerase during replication and by T7 RNA polymerase, a simple single-subunit enzyme borrowed from a virus.

The harder question was whether the multi-subunit machinery that actually runs gene expression in cellular life would tolerate them. That machinery has quality-control checkpoints. Its trigger loop, a flexible protein element, folds shut when a correctly paired nucleotide is loaded, which is part of how cells avoid transcription errors. The same group had already resolved their earlier structures of a six-letter DNA alphabet containing the B and S pair in 2023, leaving the P and Z pair uncharacterised for cellular polymerases.

The new assays and images answered it. Both P with Z and its modified version adopted canonical Watson-Crick geometry in the active site and induced trigger loop folding, with key protein residues in nearly the same positions seen with natural pairs. Single-turnover measurements found incorporation rates only about two-fold lower than the natural G to C pairing under matched conditions, and chase experiments showed the enzyme continued elongating normally past a synthetic pair, with no pause at the following position.

A Nitro Group, a Mispairing Problem and a Fix Called Z*

The study is unusually candid about a flaw. The Z base carries a nitro group that lowers the acidity of the nucleobase to a pKa of roughly 7.8, which means that at the pH inside a cell it readily loses a proton and takes a form that mimics cytosine well enough to pair with guanine. The team observed exactly that: misincorporation of G opposite Z stood out above every other non-cognate combination, and the paper notes that the same behaviour has been reported during DNA replication, indicating a chemical vulnerability rather than a quirk of this enzyme.

Their fix was chemical. Swapping the nitro group for a carboxamide raises the pKa above 10, which largely prevents the deprotonation. That analogue, written Z*, sharply reduced the mispairing in their assays and produced no other major misincorporation on the templates tested.

The tradeoff is that Z* is slower. Comparing the two datasets, the team found that the nitro group itself was stabilising a water-mediated interaction with the enzyme's bridge helix that helps hold the shape needed for the trigger loop to close. Without it, the complex sat in an open, less catalytically advanced state far more often.

The authors do not oversell the fix. They describe Z* as an initial step toward a fully optimised pair and say further refinement will be required. They also note that their single-nucleotide assays force a mismatch by offering only one nucleotide at a time, so the error rates they measured are likely overestimates of what would happen in a complete system where the correct nucleotide competes.

Purified Enzymes in Buffer, Not Organisms in Soil

The framing that matters for an environment and science audience is the distinction between a test tube and a living system. This is laboratory structural biology, peer-reviewed and published, using E. coli RNA polymerase purified from expression plasmids, short oligonucleotides ordered from a commercial supplier, and synthetic nucleotides manufactured by a Florida company. Nothing was released, nothing replicated in an organism, and no environmental exposure was tested or modelled.

One fact from the methods section is worth noting. The four synthetic letters do not occur in nature. Their building blocks had to be chemically synthesised and supplied to the reaction, in this case phosphoramidites and ribotriphosphates purchased from Firebird Biomolecular Sciences of Alachua, Florida. Any organism engineered to depend on those letters would depend on a manufactured chemical feedstock, a point this paper records without examining its implications.

Readers should also be cautious about the species gap. Whether an E. coli enzyme predicts how human or plant transcription machinery would respond to these letters is not something this work addresses.

Reagents, Disclosure and What Comes Next

The record here is public and checkable. Cryo-EM data were collected at the Stanford-SLAC Cryo-EM Center, which is supported by the National Institute of General Medical Sciences, and the authors declared no competing interests. Steven Benner, a co-corresponding author and one of the original architects of the Hachimoji system, is based with two co-authors at the Foundation for Applied Molecular Evolution in Alachua, Florida, the same town as the company that supplied the synthetic nucleotides. All four structures and maps have been deposited in the Protein Data Bank and the Electron Microscopy Data Bank, and the full paper is openly available, so other groups can check the work.

A companion paper from the same group appeared in the Proceedings of the National Academy of Sciences on Aug. 12, reporting that RNA polymerase can also recognise a different synthetic pair that lacks the hydrogen bonds normally considered essential to holding DNA base pairs together.

The authors set out two directions. One leads toward messenger RNA and eventually protein, where translation of the full eight-letter system remains unachieved, although ribosomal incorporation of a non-standard amino acid was demonstrated with the B and S pair back in 1992. The other leads to functional non-coding RNAs, where an expanded chemical alphabet could produce molecules that bind targets more tightly than four-letter RNA can. A six-letter DNA aptamer built with P and Z has already been used to deliver doxorubicin selectively to liver cancer cells, which is the precedent the authors cite.

The reasonable takeaway is narrow and genuinely interesting. A well-understood cellular enzyme turns out to read engineered genetic letters using the same machinery it uses for natural ones, which removes one obstacle among several standing between synthetic base pairs and functioning inside a cell. It does not mean organisms running on eight letters exist, are imminent, or pose any current environmental question. The next milestones to watch are a demonstration in living bacteria and a further-refined version of the Z base, neither of which has been reported.

What Readers Want to Know

What did the researchers actually demonstrate? That E. coli RNA polymerase can accurately read and transcribe a DNA template containing eight letters, four natural and four synthetic, and that it recognises the synthetic pairs using the same structural mechanisms it uses for natural ones.

Was this done in a living organism? No. The experiments used purified enzyme and short synthetic DNA and RNA scaffolds in solution. No organism was engineered, and nothing was released.

Can these eight letters make proteins? Not yet. The authors state that translation of Hachimoji codons into protein remains a future goal. Ribosomal incorporation of a non-standard amino acid has been shown previously for one synthetic pair, not for the full eight-letter system.

What was the main problem the study found? The Z base tends to mispair with guanine because it loses a proton at cellular pH. The team's modified version, Z*, reduced that error substantially but incorporates more slowly, and they describe it as a first step rather than a solution.

Is there an environmental or biosafety concern right now? No current one. The synthetic letters do not occur in nature and must be manufactured and supplied, but this study did not test containment, environmental release, or ecological effects, and no engineered organism using them exists.

Where can the underlying data be checked? All four cryo-EM maps and their atomic coordinates were deposited in the Electron Microscopy Data Bank and the Protein Data Bank, and the paper is open access with source data provided.

What should readers watch for next? A demonstration of the eight-letter system inside living bacteria, and a further-improved version of the Z base. Neither has been reported.