LA JOLLA, CALIFORNIA
Four billion years. Four letters.
A, T, C and G. Adenine, thymine, cytosine, guanine. Every gene in every organism that has ever lived on this planet — the bacteria in your gut, the oak outside your window, the tiger, the tulip, the coronavirus, you — is written in that four-character alphabet and nothing else. It is the single most conserved fact in all of biology. Life experimented with almost everything else: cell walls, wings, photosynthesis, consciousness. It never once changed its alphabet.
This week, researchers at the University of California San Diego published something that quietly punctures the assumption that it couldn't.
Using biochemistry and a cryo-electron microscope capable of resolving structures finer than the width of a single atom, they caught one of the most fundamental enzymes in all of biology — RNA polymerase, the machine that reads DNA and writes RNA, the very first step of gene expression in every living thing — reading an eight-letter genetic alphabet.
Not struggling with it. Not tolerating it. Reading it, in the words of the study, "in the same manner as natural base pairs."
The enzyme, in other words, did not appear to notice that it was reading a language that has never existed on Earth.
THE FOUR LETTERS THAT AREN'T FROM HERE
The alphabet in question has a name, and the name is charming: hachimoji. It comes from the Japanese hachi, eight, and moji, letter. Eight letters.
Hachimoji DNA was introduced to the world in the journal Science in February 2019 by Steven A. Benner of the Foundation for Applied Molecular Evolution, along with a large collaboration. Benner and colleagues designed and synthesised four entirely artificial chemical bases — designated P, Z, B and S — and slotted them alongside nature's A, T, C and G.
The synthetic letters pair up just as the natural ones do. A pairs with T; C pairs with G; P pairs with Z; and B pairs with S. Four pairs instead of two. Eight letters instead of four.
Crucially, hachimoji DNA was not just a chemical curiosity that could sit in a tube looking pretty. Benner's team showed it met the structural requirements for a genetic system capable of supporting Darwinian evolution: a polyelectrolyte backbone, predictable thermodynamic stability, stereoregular building blocks. It could be transcribed into hachimoji RNA. That RNA folded into a functioning fluorescent aptamer — a molecule that does a job.
But between "we can make this chemistry work in a tube" and "a living cell can use this" lies a very large gap. And the bridge across that gap is RNA polymerase.
WHY THE ENZYME IS THE WHOLE BALLGAME
Here is why this specific enzyme matters so much, and why the new work is a bigger deal than a headline about "eight-letter DNA" might suggest.
Storing information is easy. Reading it is hard.
Cells do not do anything with DNA directly. To make use of a gene, a cell must first transcribe it — copy the DNA sequence into a strand of RNA, which then goes on to direct the building of proteins. That transcription job is done, in bacteria and in you, by a large multi-subunit machine called RNA polymerase. It is the gatekeeper. If RNA polymerase cannot read your synthetic letters, then those letters are inert. Beautiful, expensive, inert.
So the question the San Diego team set out to answer was blunt: can the cell's own native machinery — not a specially engineered enzyme, not a designer polymerase from a directed-evolution campaign, but the ordinary workhorse polymerase of Escherichia coli — actually transcribe a genetic alphabet with double the letters?
"Because cellular transcription is executed by multi-subunit RNA polymerases, the compatibility of unnatural base pairs with this machinery is essential for engineering expanded genetic systems," the authors write in the paper's opening.
The answer, published September 2 in Nature Communications under the title "Structural basis of transcription of the hachimoji eight-letter alphabet by E. coli RNA polymerase," is yes.
THE PROBLEM WITH Z — AND THE FIX
The work was not frictionless, and the way the team handled the friction is the most interesting part of the story for anyone who cares how science actually gets done.
When they ran the transcription experiments, they found a fidelity problem. The polymerase was sometimes pairing the synthetic base Z with the natural base G — a misincorporation error. In a genetic system, that is not a rounding error; that is a typo in the source code, and typos in genetic source code compound.
So the team did what synthetic chemists do. They redesigned the letter.
They synthesised a higher-fidelity version of Z, which they call Z*, in which the nitro group at the C5 position is replaced with a carboxamide group. One functional group swap on one ring position — and the misincorporation problem substantially improves.
That detail deserves a beat. This is the difference between studying a genetic alphabet and engineering one. Nature's four letters are the product of four billion years of selection and are stuck as they are. Hachimoji's letters are chemistry, and chemistry can be edited on a bench in a week. When the P and Z pair misbehaved, the researchers didn't work around it — they built a better Z.
WHAT THE PICTURES SHOWED
The structural work is the heart of the paper, and it is genuinely beautiful.
Using cryo-electron microscopy — flash-freezing molecules and imaging them with electron beams, a technique that has revolutionised structural biology over the past decade — the team determined four separate structures of RNA polymerase caught in the act of incorporating either dZ:PTP or dP:Z*TP. The resolutions range from 2.42 to 2.75 ångströms.
For non-specialists: an ångström is one ten-billionth of a metre. A carbon atom is about 1.5 ångströms across. These are pictures at a scale where you can see individual chemical bonds forming.
And what those pictures show is the punchline. RNA polymerase recognises the synthetic base pairs using the same biochemical and structural signals it uses for the natural ones. The enzyme's active site closes around the artificial pair the way it closes around a normal one. The geometry checks out. The machinery does not care.
Combined with the group's earlier work on the S:B pair, the authors conclude that "E. coli RNA polymerase is able to efficiently recognize these unnatural base pairs in the same manner as natural base pairs," and that the results "establish the feasibility of an eight-letter genetic alphabet for transcription."
The study was led by Dong Wang, PhD, professor at the UC San Diego Skaggs School of Pharmacy and Pharmaceutical Sciences, with first author Qingrong Li. The collaboration included Hyo-Joong Kim, Yan Liu, Juntaek Oh, Peini Hou, Shuichi Hoshika, Grigore Pintilie, Sriram Aiyer, Jenny Chong, structural biologist Dmitry Lyumkis — and Steven Benner himself, the chemist who invented the letters seven years ago now watching a cell's own enzyme read them.
THE SECOND PAPER, AND THE STRANGER RESULT
Buried in the announcement is a companion finding that is arguably weirder than the headline one.
In a separate study published August 12 in the Proceedings of the National Academy of Sciences, the same group — led again by Wang, with first author Qingrong Li and collaborators including Michiko Kimoto and Ichiro Hirao — reported that RNA polymerase can also recognise a completely different class of synthetic base pair: a hydrophobic unnatural base pair that has no hydrogen bonds holding it together at all.
Take a moment with that. The textbook explanation of how DNA works, the one on every classroom poster, is hydrogen bonding: A pairs with T because two hydrogen bonds form between them; C pairs with G because three do. That specificity is supposed to be the whole basis of genetic fidelity.
The PNAS paper — titled "Hydrophobic unnatural base pair promotes trigger loop closure and catalysis in cellular RNA polymerase independent of hydrogen bonding" — reports that the enzyme's trigger loop closes and catalysis proceeds even without them. The polymerase is apparently reading shape and fit as much as it is reading hydrogen bonds.
Two papers, three weeks apart, from one lab, both pointing at the same uncomfortable and exhilarating conclusion: the machinery of life is considerably more accommodating than four billion years of monoculture would suggest.
EXPERT PERSPECTIVE: WHY THIS IS A FOUNDATION, NOT A PRODUCT
It is worth being precise about what has and has not been demonstrated, because "scientists double the genetic alphabet" is the kind of headline that runs ahead of the science.
What has been shown: purified E. coli RNA polymerase, working on templates containing synthetic base pairs, transcribes them efficiently and accurately, and the structural basis for that accuracy is now understood at near-atomic resolution.
What has not been shown: a living organism running its full metabolism on an eight-letter genome. That remains a much harder problem involving DNA replication fidelity across generations, the supply of synthetic nucleotide building blocks inside a cell, repair enzymes that have never seen these letters, and a ribosome that would need to translate the resulting RNA into protein.
The UC San Diego announcement is careful about this. It frames the work as providing "a molecular foundation for future technologies that use expanded genetic codes" — a foundation, not a finished building.
But foundations are what the field has been missing. Expanded genetic alphabets have existed as chemistry since 2019. What has been absent is proof that biology's own machinery will play along, and a structural explanation of why. That is what these two papers supply.
WHAT THIS MEANS FOR YOU
The nearest-term applications are not science-fictional. They are diagnostic and therapeutic, and one of them already exists.
Earlier research has used expanded genetic alphabets to build synthetic DNA molecules — aptamers — capable of recognising liver cancer cells. Aptamers are short nucleic acid strands that fold into three-dimensional shapes and stick tightly and specifically to a target, functioning rather like antibodies but far smaller, cheaper to manufacture, and easier to modify.
The catch with conventional aptamers is that they are built from only four chemical letters, which sharply limits the shapes and chemical surfaces they can form. Add four more letters and the accessible design space expands enormously. More shapes means more targets, tighter binding, better discrimination between a cancer cell and a healthy one.
Practical consequences, if this line of work matures: blood tests that detect specific cancers earlier and with fewer false positives. Targeted therapeutics that bind their target and nothing else, sparing patients the collateral damage that makes so much of oncology brutal. Biosensors for pathogens or contaminants in water and food supplies.
There is a second category that sounds stranger but may arrive sooner: information storage. DNA is an extraordinarily dense and durable storage medium — you can, in principle, put an entire data centre's contents in a test tube and read it back millennia later. Doubling the alphabet increases the information density per base and enables more sophisticated barcoding, tagging, and error-correction schemes. Hachimoji DNA was flagged for exactly these applications back in 2019: bar-coding, combinatorial tagging, retrievable information storage, self-assembling nanostructures.
And then there is the third category: engineered biological systems that produce compounds not found in nature. Enzymes are proteins, proteins are built from instructions in nucleic acids, and the range of what cells can be made to manufacture is bounded by what can be encoded. A larger alphabet is, over the long run, a larger vocabulary for that manufacturing.
There is also a genuine biosafety argument in favour of this approach that deserves airing. An organism dependent on synthetic nucleotides that do not exist in the wild cannot survive outside a lab that supplies them. Expanded-alphabet systems come with a built-in containment mechanism: cut off the supply of P, Z, B and S, and the organism simply stops. That is a more robust safeguard than most of the containment strategies used in synthetic biology today.
THE BIGGER QUESTION NOBODY CAN ANSWER YET
Underneath all the practical talk sits a question that has bothered origin-of-life researchers for decades.
Why four?
Was the four-letter alphabet selected for — genuinely optimal, the best possible trade-off between information density and copying fidelity, arrived at by relentless competition? Or was it a frozen accident, the first workable system that happened to lock in before anything else got a chance, after which the cost of switching became prohibitive?
Every result like this week's nudges the answer toward "accident." If the ordinary polymerase of an ordinary gut bacterium can read eight letters at near-native efficiency, without being redesigned, using the same structural logic it applies to the natural four — then four was not a hard constraint. It was a starting condition.
Which raises the possibility that life elsewhere, if it exists, may not be written in our alphabet at all. Astrobiologists searching for biosignatures have generally looked for chemistry resembling ours. Hachimoji is a standing reminder that the space of workable genetic systems may be considerably larger than the one sample we have.
For now, the practical achievement stands on its own. A bacterial enzyme that spent billions of years reading a four-character language was handed one with eight characters, and read it fluently — while a microscope in La Jolla watched, one ten-billionth of a metre at a time.
Life has been running on four letters since before there were continents. This week, in a lab overlooking the Pacific, it turned out to have been an option all along, not a rule.
