Improving the DNA specificity and applicability of base editing through protein engineering and protein delivery

Abstract

We recently developed base editing, a genome-editing approach that enables the programmable conversion of one base pair into another without double-stranded DNA cleavage, excess stochastic insertions and deletions, or dependence on homology-directed repair. The application of base editing is limited by off-target activity and reliance on intracellular DNA delivery. Here we describe two advances that address these limitations. First, we greatly reduce off-target base editing by installing mutations into our third-generation base editor (BE3) to generate a high-fidelity base editor (HF-BE3). Next, we purify and deliver BE3 and HF-BE3 as ribonucleoprotein (RNP) complexes into mammalian cells, establishing DNA-free base editing. RNP delivery of BE3 confers higher specificity even than plasmid transfection of HF-BE3, while maintaining comparable on-target editing levels. Finally, we apply these advances to deliver BE3 RNPs into both zebrafish embryos and the inner ear of live mice to achieve specific, DNA-free base editing in vivo.

Figure 1. Engineering and in vitro characterization of a high-fidelity base editor (HF-BE3).

(a) Schematic representation of HF-BE3. Point mutations introduced into BE3 to generate HF-BE3 are shown in green. The representation used PDB structures 4UN3 (Cas9), 4ROV (cytidine deaminase) and 1UGI (uracil DNA glycosylase inhibitor). (b) In vitro deamination of synthetic substrates containing ‘TC' repeat protospacers. Values and error bars reflect the mean and range of two independent replicates performed on different days.

Figure 2. Activity of a high-fidelity base editor (HF-BE3) in human cells.

(a–c) On- and off-target editing associated with plasmid transfection of BE3 and HF-BE3 was assayed using high-throughput sequencing of genomic DNA from HEK293T cells treated with sgRNAs targeting non-repetitive genomic loci EMX1 (a), FANCF (b) and HEK293 site 3 (c). On- and off-target loci associated with each sgRNA are separated by a vertical line. (d) On- and off-target editing associated with the highly repetitive sgRNA targeting VEGFA site 2. Values and error bars reflect mean±s.d. of three independent biological replicates performed on different days. For a–c, stars indicate significant editing based on a comparison between the treated sample and an untreated control. *P≤0.05, **P≤0.01 and ***P≤0.001 (Student's two-tailed t-test). For d, asterisks are not shown since all treated samples displayed significant editing relative to the control. Individual P values are listed in Supplementary Table 1.

Figure 3. Protein delivery of base editors into cultured human cells.

(a–d) On- and off-target editing associated with RNP delivery of base editors complexed with sgRNAs targeting EMX1 (a), FANCF (b), HEK293 site 3 (c) and VEGFA site 2 (d). Off-target base editing was undetectable at all of the sequenced loci for non-repetitive sgRNAs. Values and error bars reflect mean±s.d. of three independent biological replicates performed on different days. Stars indicate significant editing based on a comparison between the treated sample and an untreated control. *P≤0.05, **P≤0.01 and ***P≤0.001 (Student's two-tailed t-test).

Figure 4. Effect of dosage of BE3 protein or plasmid on the efficiency of on-target and off-target base editing in cultured human cells.

(a) On-target editing efficiency at each of the four genomic loci was averaged across all edited cytosines in the activity window for each sgRNA. Values and error bars reflect mean±s.e.m. of three independent biological replicates performed on different days. (b,c) On- and off-target editing at the EMX1 site arising from BE3 plasmid titration (b) or BE3 protein titration (c) in HEK293T cells. Values and error bars reflect mean±s.d. of three independent biological replicates performed on different days.

Figure 5. DNA-free in vivobase editing in zebrafish embryos and in the inner ear of live mice using RNP delivery of BE3.

(a) On-target genome editing in zebrafish harvested 4 days after injection of BE3 complexed with indicated sgRNA. Values and error bars reflect mean±s.d. of three injected and three control zebrafish. Controls were injected with BE3 complexed with an unrelated sgRNA. (b) Schematic showing in vivo injection of BE3:sgRNA complexes encapsulated into cationic lipid nanoparticles. (c) Base editing of cytosine residues in the base editor window at the VEGFA site 2 genomic locus. (d) On-target editing at each cytosine in the base-editing window of the VEGFA site 2 target locus. (c,d) Values and error bars reflect mean±s.e.m. of three mice injected with sgRNA targeting VEGFA site 2, three uninjected mice and one mouse injected with unrelated sgRNA.