Programmable C:G to G:C genome editing with CRISPR-Cas9-directed base excision repair proteins

Abstract

Many genetic diseases are caused by single-nucleotide polymorphisms. Base editors can correct these mutations at single-nucleotide resolution, but until recently, only allowed for transition edits, addressing four out of twelve possible DNA base substitutions. Here, we develop a class of C:G to G:C Base Editors to create single-base genomic transversions in human cells. Our C:G to G:C Base Editors consist of a nickase-Cas9 fused to a cytidine deaminase and base excision repair proteins. Characterization of >30 base editor candidates reveal that they predominantly perform C:G to G:C editing (up to 90% purity), with rAPOBEC-nCas9-rXRCC1 being the most efficient (mean 15.4% and up to 37% without selection). C:G to G:C Base Editors target cytidine in WCW, ACC or GCT sequence contexts and within a precise three-nucleotide window of the target protospacer. We further target genes linked to dyslipidemia, hypertrophic cardiomyopathy, and deafness, showing the therapeutic potential of these base editors in interrogating and correcting human genetic diseases.

Fig. 1. Initial screen of CGBE candidates for C:G to G:C editing.

a CBEs like BE3 and BE4 predominantly convert C:G to T:A while CGBE aims to predominantly convert C:G to G:C. b CGBE candidates were designed in three orientations – ACX, AXC, and XAC, where X denotes the fused BER protein. c Seven candidates were selected for their high C:G to G:C editing at both HEK2 and HEK3. The lower editing at HEK3 is likely due to a disfavored motif (refer to data in Fig. 2a). Targeted C’s are in red. PAMs are underlined. *p < 0.05; **p < 0.01; ***p < 0.001 using one-way ANOVA (Dunn- Šidák) of C:G to G:C editing against ‘Untreated’. Exact p values are available in Source Data. Each dot represents editing of an individual biological replicate; bars represent mean values; error bars represent SEM of three biologically independent replicates.

Fig. 2. Sequence context and editing window for two selected CGBEs.

a Evaluation of C:G to G:C editing at each NCN DNA motif. 16 different gRNAs were designed to target the genomic region around the HEK2 site (HEK2-1 to HEK2-16 in Supplementary Table 2), chosen such that the gRNA-to-target combinations together cover all NCN motif contexts and that genomic distance among gRNAs is minimized (14/16 gRNAs, including the initial HEK2-1 gRNA:target, reside within a 1.8 kb region, while the other 2 gRNAs target within 10 kb). Each dot represents C:G to G:C editing of an individual biological replicate; bars represent mean values of two biologically independent replicates. b DNA WebLogo created with target motifs in which C:G at position 6 was edited to G:C (n = 2 biologically independent replicates; error bars are Bayesian 95% confidence intervals). c Editing window of CGBEs using gRNAs with alternating 5′-W-C-3′ motifs. Targeted C’s are in red. PAMs are underlined. *p < 0.05; **p < 0.01 using one-way ANOVA (Dunn-Šidák) of C:G to G:C editing against ‘BE3’. Exact p values are available in Source Data. Each dot represents C:G to G:C editing of an individual biological replicate; bars represent mean values; error bars represent SEM of three biologically independent replicates.

Fig. 3. CGBE induces efficient C:G to G:C editing as the predominant product.

a Removal of UGI from BE3 increases C:G to G:C editing (blue); fusion of rXRCC1 further increases C:G to G:C editing. The major byproduct is C:G to T:A editing (orange). b Mean C:G to G:C editing/C:G to T:A editing ratio. For both plots, p values were obtained via Mann–Whitney tests between the indicated editors. Each dot represents editing of an individual biological replicate. gRNAs chosen have C’s within WCW, ACC, or GCT motifs. Black lines represent mean values of 43 (BE3), 29 (BE3 (no UGI)), or 46 (ACX, rPB(8KD) and ACX, rXRCC1) biologically independent replicates.