For roughly a decade, molecular biologists have lived with a strange asymmetry. The tools to read the genome treat all four DNA letters equally. A sequencing machine does not care whether it is counting an adenine, a cytosine, a guanine, or a thymine. But the tools to rewrite the genome have been stubbornly constrained. You could change one base at a time, or two. You could never change all three mutagenic letters in a single reaction.
That asymmetry has now been broken. A team led by Liang Chen at the Lingang Laboratory and East China Normal University, with equal-contribution first authors Mengjia Hong, Changming Luan, Meng Yuan, Hao Huang, and Xinyuan Guo, reported July 27, 2026 in Nature Communications the development of smACG triple base editors. The name stands for simultaneous mutagenesis of A, C, and G. For the first time, researchers can convert adenine, cytosine, and guanine within the same allele in a single experimental step.
The leap from editing one or two bases to editing all three is not incremental. It changes the experimental landscape in a way that efficiency numbers alone do not capture.
The difference between reading and rewriting
Base editing emerged in 2016 as a gentler alternative to CRISPR-Cas9. Rather than breaking both strands of DNA and letting the cell repair the damage, base editors chemically convert one nucleotide letter into another without a double-strand break. Adenine base editors change A to G (read as A to G in the genome, functionally an A-to-I deamination interpreted as G). Cytosine base editors change C to T. These are single-letter changes, the molecular equivalent of a pencil eraser on a single character.
Dual base editors followed, allowing simultaneous A-to-G and C-to-T editing. But G, the third mutagenic base, remained inaccessible. A guanine base editor was developed in 2020, but it could not be combined with the other two in the same reaction. The toolkit had A, C, and T covered. Guanine was the missing operator.
The problem is chemical. Cytosine deamination is straightforward: removing an amine group from cytosine produces uracil, which the cell reads as thymine. Adenine deamination works similarly. But guanine requires a different strategy. The Chen group solved this by engineering a fusion of an adenine deaminase, a cytosine deaminase, and an alkyladenine DNA glycosylase variant within the Cas9 scaffold. The glycosylase excises guanine from the DNA backbone, creating an abasic site. The cell’s base excision repair pathway fills that gap, preferentially inserting the desired replacement base.
The result is a single protein complex that can, in one pass, turn A into G, C into T, and G into any of the other three bases.
smACGmax and the efficiency question
The team tested multiple architectural variants. The best performing, designated smACGmax, achieved up to 41 percent triple-base conversion efficiency across a panel of target sites with diverse sequence contexts. That means in nearly half of all edited alleles, all three targeted bases were simultaneously converted.
Single-base conversion efficiencies for smACGmax reached as high as 82 percent for A-to-G, 77 percent for C-to-T, and 55 percent for G-to-anything. The G editing efficiency varied depending on the replacement base, with G-to-A being the most efficient and G-to-T the least.
A persistent concern with any engineered deaminase is off-target editing, unintended changes elsewhere in the genome or in RNA transcripts. The team assessed RNA off-target effects using transcriptome-wide sequencing and found that smACGmax showed substantially lower RNA off-target activity compared to previously reported dual-base editors. This suggests that the engineered enzyme architecture, despite packing three catalytic activities, is more specific than earlier constructs with fewer activities.
From letters to landscapes
What matters about triple base editing is not the percentage improvement over dual editors. What matters is that the mutation landscape accessible to a single experiment has expanded from a handful of possible changes to virtually all possible point mutations at a given target site. A dual base editor targeting a five-base window can produce 32 combinations. A triple base editor can produce 243. This is a shift from sparse sampling to near-exhaustive interrogation of sequence space.
The Chen group demonstrated the practical power of this expanded landscape with two proof-of-concept experiments.
Finding diphtheria toxin resistance in one shot
The first application targeted the HBEGF gene, which encodes the receptor that diphtheria toxin uses to enter human cells. Mutations in HBEGF can confer resistance to the toxin, but identifying those mutations has traditionally required laborious selection experiments or directed evolution.
The team applied smACGmax to a window within the HBEGF coding sequence, generating a library of edited cells carrying diverse combinations of base changes. Sequencing showed that the editor achieved 94 percent coverage of the target region, meaning nearly every base in the window was successfully edited across the cell population. When the team challenged the edited cells with diphtheria toxin, resistant clones emerged readily. Sequencing those clones revealed the specific mutation combinations that confer resistance.
The experiment demonstrates a general principle: triple base editing can be used as a one-shot functional screening platform. Rather than introducing mutations one at a time and testing each variant individually, researchers can generate a high-diversity library of genetic variants in a single reaction and then apply a selective pressure to fish out the functional hits.
Dissecting the spliceosome one base at a time
The second application targeted SF3B1, a splicing factor frequently mutated in cancers including myelodysplastic syndrome, chronic lymphocytic leukemia, and uveal melanoma. Mutations in SF3B1 alter the splicing of hundreds of downstream genes, but the relationship between specific SF3B1 variants and their splicing consequences has been difficult to dissect because naturally occurring tumors carry many co-occurring mutations.
The team used smACGmax to generate a panel of SF3B1 variants spanning single, double, and triple base changes within a hotspot region. By comparing the splicing patterns produced by each variant, they were able to characterize how specific amino acid substitutions alter splicing specificity. Some triple mutants produced splicing changes that could not be predicted from the effects of the individual single mutants, highlighting the importance of studying combinatorial mutations rather than assuming additivity.
This kind of experiment was essentially impossible before triple base editing. Generating the same variant panel with traditional methods would have required dozens of independent mutagenesis reactions and cloning steps. smACGmax produced the entire panel in a single experiment.
The landscape argument
The broader significance of triple base editing lies in the change in experimental reach it enables. Knowing what the genome says is not the same as knowing what it can say. Every position in the genome is, in principle, mutable. The space of possible human genomes is astronomically larger than the number of humans who have ever lived. The question for functional genomics is whether we can systematically explore the mutational landscape around specific genes to understand what changes matter for disease, drug resistance, and biological function.
Triple base editors do not answer that question, but they remove a significant barrier to asking it. With smACG, an experimenter can target a region of interest and generate essentially all possible point mutations in a single reaction. The limiting factor shifts from the tool to the throughput of the functional assay.
The technology is not without limitations. The 41 percent triple-base efficiency means that in most edited cells, not all three targeted bases are converted. The G editing efficiency, at 55 percent for the most favorable replacement, is the weakest link, and the editing window constrained by the Cas9 structure covers roughly 5 to 7 bases.
But the principle is established. Base editing has moved from single letters to dual letters to the complete mutagenic alphabet. The asymmetry between reading and rewriting the genome has been narrowed, and the kind of questions molecular biologists can ask has changed.
The Chen group has filed a patent application (CNIPA No. 202411508492.5) covering the technology. The work was supported by the National Key R&D Program of China, the National Natural Science Foundation of China, the Shanghai Municipal Commission, and the Lingang Laboratory. The paper is published open access under a CC BY-NC-ND 4.0 license.
The genome has always been a four-letter language. Now, for the first time, experimenters have an alphabet that matches the full grammar.