A Bacterial Enzyme Just Read an Eight-Letter DNA Alphabet
-
Click here to listen to this article - Share via
- E. coli RNA polymerase transcribed DNA templates built from an eight-letter DNA alphabet, adding the synthetic pairs P:Z and B:S to natural A:T and G:C, according to the study published Sept. 2, 2026 in Nature Communications.
- Incorporation of the synthetic P:Z pair ran only about twofold slower than the natural G:C pair under the same conditions, the study authors reported.
- The most common error was the synthetic letter Z pairing with natural guanine, which happens because Z loses a proton at the pH inside cells, the authors wrote.
- A redesigned version of Z, called Z*, swaps a nitro group for a carboxamide and raises the base’s pKa above 10, and the authors reported it substantially reduced that mispairing.
- Four cryo-electron microscopy structures at 2.42 to 2.75 angstrom resolution show the synthetic pairs sitting in the enzyme’s active site in standard Watson-Crick geometry, the same shape natural pairs take.
Table of Contents
- What an Eight-Letter DNA Alphabet Is
- What the Structures Show
- What This Could Be Used For
- Research Gap
- Closing Thoughts
What an Eight-Letter DNA Alphabet Is
Every living thing on Earth stores genetic information in four chemical letters: adenine, thymine, cytosine and guanine, written A, T, C and G. They pair in fixed couples, A with T and C with G, and that pairing is how genetic information gets copied and read. An eight-letter DNA alphabet keeps those four and adds four more.
Chemists have spent years adding letters to that set. One result is hachimoji DNA, an eight-letter system reported in Science in 2019 by a team including Steven A. Benner, PhD, of the Foundation for Applied Molecular Evolution [2].
Hachimoji, Japanese for eight letters, keeps the natural four and adds two synthetic pairs: P with Z, and B with S. The synthetic letters rearrange which parts of the molecule donate and accept hydrogen bonds while keeping the same overall shape as a natural pair. Because the synthetic pairs hold that shape, enzymes that evolved on natural DNA can act on them [2].
Why a Natural Enzyme Matters Here
Getting a cell to read synthetic DNA is a separate problem from building it in a tube.
Cells read DNA using RNA polymerase, a multi-part enzyme that copies a DNA sequence into RNA. Bacteria, plants and humans all use a multi-subunit version of it. Earlier hachimoji work relied on T7 RNA polymerase, a single-subunit enzyme from a virus that infects bacteria, which is simpler and more permissive.
The new study used E. coli RNA polymerase, from Escherichia coli, the multi-subunit kind cellular life actually runs on. The team, led by Dong Wang, PhD, a professor at the UC San Diego Skaggs School of Pharmacy and Pharmaceutical Sciences, reported that the enzyme transcribed templates containing all four synthetic letters without being engineered to do so.
The enzyme incorporated the P:Z pair only about twofold slower than the natural G:C pair under the same conditions [1].
Wang’s group had already shown the B:S pair works in this enzyme in 2023 [3]. The P:Z pair was the untested half, and the two pairs had never been shown to work side by side in a multi-subunit polymerase.
The authors also reported that the enzyme extended efficiently past a synthetic pair, generating long RNA transcripts.
What the Structures Show
The team froze the enzyme mid-reaction and imaged it with cryo-electron microscopy, a method that reconstructs a molecule’s structure from images of thousands of frozen copies. Four structures came out at resolutions between 2.42 and 2.75 angstroms, sharp enough to place individual water molecules [1].
In every structure, the synthetic pair sat in the active site in standard Watson-Crick geometry, the paired shape natural DNA letters take.
Two of the enzyme’s moving parts behaved normally as well. In the trigger-loop-closed structures, the trigger loop, a segment that folds shut over a correctly matched incoming building block, folded, and the bridge helix bent the way it does with natural DNA [1].
“P:Z and P:Z* behave as native-like base pairs during the early steps of substrate loading,” the authors wrote.
The enzyme passes the synthetic letters through the same proofreading checkpoints it applies to natural DNA.
The Error That Had to Be Fixed
Across 32 combinations of template letter and incoming building block, one error stood out: the synthetic letter Z pairing with natural guanine.
The cause is chemical. Z carries a nitro group that pulls electrons away and lowers the molecule’s pKa to roughly 7.8, close to the pH inside a cell. At that pH a meaningful fraction of Z loses a proton, and the deprotonated form takes on a shape that reads like cytosine, which guanine pairs with normally [1].
The same misfire had already been documented in DNA replication, and the authors described the deprotonation mechanism behind it as universal and enzyme-independent rather than specific to any one polymerase.
Their fix was to change the chemistry. Z* replaces the nitro group with a carboxamide, raising the pKa above 10 so the base stays protonated at neutral pH [1].
In the assays, guanine misincorporation opposite Z* dropped substantially, and stray products seen with Z against other templates did not appear with Z*.
The Tradeoff Behind the Fix
The correction slowed the enzyme, and the structures show why.
In the Z-containing complexes, 76.8% of imaged particles had the trigger loop closed, the conformation that precedes catalysis. In the Z* complexes, that flipped: 80.6% sat open. Kinetic measurements matched, with the original Z scaffold supporting faster incorporation than Z* [1].
The structures locate the reason at a single water molecule. In the Z complex, a water sits 2.62 angstroms from the nitro group’s nitrogen and hydrogen-bonds to two points on the bridge helix, holding the helix in its bent position and helping the trigger loop close [1].
Z* has no nitro group, so that water bridge does not form.
The chemical group causing the errors is also the one helping the enzyme close on the incoming letter. The authors described Z* as an initial step toward a better P:Z pair and said further work on the Z scaffold will be needed [1].
What This Could Be Used For
The nearest application is RNA molecules with chemistry that natural RNA does not have.
Aptamers are short nucleic acid strands that fold into shapes that bind a specific target, and some are being developed as drugs and diagnostic reagents. A larger letter set gives designers more chemical variety to fold with, and the authors named aptamer design as the application their results support [1].
Protein coding is not yet possible. Making protein from hachimoji instructions requires the translation machinery, ribosomes and transfer RNAs, to handle expanded codons, and the authors marked translation as a future goal.
Nothing here is a clinical tool, and no part of this work has been tested in people. Readers should treat it as laboratory research rather than anything bearing on current treatment.
Research Gap
Every result in the paper comes from purified components in a tube, not from living cells. The enzyme, the DNA templates and the building blocks were assembled in defined reactions.
The authors also flagged a limit in how the error rates were measured. Single-nucleotide assays offer the enzyme one building block at a time with no competition from the correct one, so measured misincorporation is likely an overestimate of what would happen in a complete system where all eight compete at once [1].
A cell adds problems this setup does not test: supplying synthetic building blocks, keeping them stable, and dealing with repair systems that treat unfamiliar chemistry as damage. Those lie outside the experiments reported here, and the authors present the work as a mechanistic foundation rather than a demonstration in living cells.
Closing Thoughts
The structural case is now reasonably complete: a natural multi-subunit RNA polymerase recognizes all four hachimoji synthetic letters through the same checkpoints it applies to natural DNA. Two things remain unsettled. The current fix for the Z:G error trades accuracy against speed, and nothing yet shows whether this survives the move from purified reactions into a living cell.
References
[1] Li, Q., Kim, H.-J., Liu, Y., Oh, J., Hou, P., Hoshika, S., Pintilie, G., Aiyer, S., Chong, J., Lyumkis, D., Benner, S. A., & Wang, D. (2026). Structural basis of transcription of the hachimoji eight-letter alphabet by E. coli RNA polymerase. Nature Communications, 17(1), 9060. https://doi.org/10.1038/s41467-026-76668-0
[2] Hoshika, S., Leal, N. A., Kim, M.-J., Kim, M.-S., Karalkar, N. B., Kim, H.-J., Bates, A. M., Watkins, N. E., SantaLucia, H. A., Meyer, A. J., DasGupta, S., Piccirilli, J. A., Ellington, A. D., SantaLucia, J., Georgiadis, M. M., & Benner, S. A. (2019). Hachimoji DNA and RNA: A genetic system with eight building blocks. Science, 363(6429), 884–887. https://doi.org/10.1126/science.aat0971
[3] Oh, J., Shan, Z., Hoshika, S., Xu, J., Chong, J., Benner, S. A., Lyumkis, D., & Wang, D. (2023). A unified Watson-Crick geometry drives transcription of six-letter expanded DNA alphabets by E. coli RNA polymerase. Nature Communications, 14(1), 8219. https://doi.org/10.1038/s41467-023-43735-9
