A split cytosine deaminase architecture enables robust inducible base editing.

Directed base substitution with base editing technology enables efficient and programmable conversion of C:G or A:T base pairs to T:A or G:C in the genome. Although this technology has shown great potentials in a variety of basic research, off-target editing is among one of the biggest challenges toward its way to clinical application. Base editing tools, especially the tools converting C to T, caused unpredictable off-target editing throughout the genome, which raise the concern that long-term application of these tools would induce genomic instability or even tumorigenesis. To overcome this challenge, we designed an inducible base editing tool that was active only in the presence of a clinically safe chemical, rapamycin. In the guidance of structural information, we designed four split-human APOBEC3A (A3A) -BE3 base editors in which these A3A deaminase enzymes were split at sites that were opposite to the protein-nucleotide interface. We showed that by inducible deaminase reconstruction with a rapamycin responsible interaction system (FRB and FKBP); three out of four split-A3A-derived base editors showed robust inducible base editing. However, in the absence of rapamycin, their editing ability was dramatically inhibited. Among these split editors, splicing at Aa85 of A3A generated the most efficient inducible editing. In addition, compared to the full-length base editor, the splitting did not obviously alter the editing window and motif preference, but slightly increased the product purity. We also expanded this strategy to another frequently used cytosine deaminase, rat APOBEC1 (rA1), and observed a similar induction response. In summary, these results demonstrated the concept that splitting deaminases is a practicable method for timely controlling of base editing tools.