The problem

Base editing is an innovative technology for efficiently generating base conversions without inducing double-strand DNA breaks or requiring DNA donor templates1. Current base editors — namely, adenine base editors (ABEs), cytosine base editors (CBEs) and C•G-to-G•C base editors (CGBEs) — can install A-to-G, C-to-T and C-to-G conversions, respectively. However, unlike cytosine deamination, which can induce base transitions and transversions as uracil can be removed from DNA by uracil DNA N-glycosylase, triggering the base excision repair (BER) pathway, adenine deamination only generates A-to-G conversions as no known enzymes can efficiently remove hypoxanthine from DNA to activate BER in cells2,3. As adenosine transversions could potentially correct a quarter of pathogenic single-nucleotide variations (SNVs), 17% of which require A-to-C conversions, efficient adenine transversion editors (and especially A-to-C editors) are needed for use in basic research and in potential clinical applications (Fig. 1a).

Fig. 1: Development of adenine transversion editors for specific A-to-C editing.
figure 1

a, An overview of base conversions induced by canonical base editors is shown (left; unachievable adenine transversions are labeled ×; GBEs are glycosylase base editors), with a distribution of base mutations needed to reverse human pathogenic SNVs in the ClinVar database (accessed 24 May 2021) (right). b, A conceptual diagram of the adenine base transversion editors constructed to mediate A-to-C or A-to-T edits in DNA is shown (left), alongside the potential molecular mechanism underscoring the repair of intracellular adenine mutations. In normal conditions, adenine transitions (A-to-G) are achieved through adenine deamination to generate inosine (I), which is read as guanosine by DNA polymerase. Once inosines are excised by DNA glycosylase, the BER pathway is initiated to induce the incorporation of cytosines or thymines (right). PAM, protospacer-adjacent motif. © 2023, Chen, L. et al.

Full size image

The solution

As adenosine deamination results in inosine, which is read as guanine by DNA polymerase, ABEs catalyze A-to-G transitions with almost 100% purity. To generate adenosine transversions, an apurinic or apyrimidinic site (that is, a site without a purine or pyrimidine base) must be induced by DNA glycosylases that initiate BER at inosines; however, such DNA glycosylases had been unknown. We speculated that removing hypoxanthine from the inosine that is created by ABE-mediated adenosine deamination would stimulate adenine transversions through DNA repair (Fig. 1b). To identify a DNA glycosylase that could remove hypoxanthine, we focused on alkyladenine DNA glycosylases (AAGs) and endonuclease V, enzymes that can remove hypoxanthine from DNA in vitro.

By screening eight AAGs from distinct species and endonuclease V from Escherichia coli, we found that the direct fusion of mouse AAG (mAAG) with nickase Cas9 (nCas9) and adenosine deaminase TadA-8e (generating AXBE, where X represents any nucleotide) could induce a programable A-to-Y (where Y represents C or T) transversion with up to 46% efficiency in different cell types. Through testing 93 adenines in various sequence contexts at 26 target sites, we found that AXBE prefers to induce adenine transversion at a YA*R motif (where R represents A or G and * indicates the targetable A). Structure-guided evolution of mAAG determined that a mAAG variant containing R165E and Y179F (mAAG-EF) had an enhanced ability to increase A-to-Y editing events, resulting in cytosines as the major products. As this AXBEv2 variant (which includes mAAG-EF) generates a high level of A-to-C editing but simultaneously induces A-to-G bystander mutations, we further engineered AXBEs to have minimal bystander edits. Embedding TadA-8e and mAAG-EF into nCas9 and introducing a N108Q mutation in TadA-8e4 yielded A-to-C base editors (ACBEs) (including ACBE-Q) that can specifically induce high levels of A-to-C conversions with minimal bystander and off-target editing effects. We also found that ACBEs, especially ACBE-Q, offer a more convenient and efficient option than prime editors (editors that can introduce all 12 possible base conversions, small insertions and deletions) for precisely editing A•T to C•G in multiple applications.

The implications

As AXBEs generate product diversity (up to 436 codons and 115 amino-acid conversions) that cannot be achieved by canonical base editors, AXBEs and AYBEs (which are similar to AXBEs and were constructed by fusing engineered human AAG to nCas9 and TadA-8e in an independent study)5 could prove useful for mutagenesis applications such as genetic screening, molecular evolution and lineage tracing. Furthermore, as ACBEs showed high A-to-C efficiency with reduced bystander edits, they are better suited to installing A-to-C edits with minimal optimization than prime editors that yield higher product purities; thus, ACBEs and prime editors offer complementary strengths. As ACBEs can efficiently excise hypoxanthine from DNA, they also provide a unique tool for studying mechanisms of DNA damage repair. We also found that ACBE-Q can have up to a 100% A-to-C conversion frequency in mouse embryos. As we also show that TadA-8e and mAAG can be fused with different Cas9 variants, generating ACBEs based on other CRISPR–Cas systems could broaden their targeting range and potentially correct C•G-to-A•T pathogenic SNVs, the second most common category of human pathogenic SNVs.

In short, the development of AXBEs and ACBEs expands the capabilities of base editing for basic research and potentially for therapeutic applications. Next we will focus on increasing the efficiency and product purity of adenine transversion editors.

Liang Chen & Dali Li

East China Normal University, Shanghai, China.

Expert opinion

“Chen et al. developed a class of base editors by fusing an adenine base editor with mouse alkyladenine DNA glycosylase (mAAG) to enable A-to-C/T conversion. Overall, this study is straightforward and convincing, and the new adenine transversion base editors have a considerable competitive advantage, which is useful in this fast-moving field.” Xingxu Huang, ShanghaiTech University, Shanghai, China.

Behind the paper

Base editing technology is very promising for clinical applications, but no base editors were able to induce the adenosine transversions that could correct about 25% of known pathogenic SNVs. Although previous studies failed to develop adenine transversion editors by fusing potential DNA glycosylases to other base editing components, we investigated all known enzymes across distinct species with the potential to remove hypoxanthine from DNA to create apurinic or apyrimidinic sites. Moreover, we used more than one target in the initial screen of the functional enzymes. Fortunately, we found three glycosylases that, when fused with nCas9 and TadA-8e, can stimulate adenine transversions, and we used mouse AAG to generate the initial AXBE. Based on this prototype of adenine transversion editor, structure-guided mutagenesis of mAAG enabled us to increase AXBE activity. Further engineering of TadA-8e and of where it is embedded within the nCas9 domain helped us optimize our A-to-C base editors. L.C. & D.L.

From the editor

“Chen et al. developed adenine transversion editors for precise A•T-to-C•G base editing. They fused an adenine base editor with mAAG, enabling A-to-C/T conversion. The refined ACBE-Q editor exhibited a narrower window and higher purity. ACBE-Q corrected pathogenic SNVs and generated a mouse disease model.” Editorial Team, Nature Biotechnology.