Compact zinc finger base editors that edit mitochondrial or nuclear DNA in vitro and in vivo

Abstract

DddA-derived cytosine base editors (DdCBEs) use programmable DNA-binding TALE repeat arrays, rather than CRISPR proteins, a split double-stranded DNA cytidine deaminase (DddA), and a uracil glycosylase inhibitor to mediate C•G-to-T•A editing in nuclear and organelle DNA. Here we report the development of zinc finger DdCBEs (ZF-DdCBEs) and the improvement of their editing performance through engineering their architectures, defining improved ZF scaffolds, and installing DddA activity-enhancing mutations. We engineer variants with improved DNA specificity by integrating four strategies to reduce off-target editing. We use optimized ZF-DdCBEs to install or correct disease-associated mutations in mitochondria and in the nucleus. Leveraging their small size, we use a single AAV9 to deliver into heart, liver, and skeletal muscle in post-natal mice ZF-DdCBEs that efficiently install disease-associated mutations. While off-target editing of ZF-DdCBEs is likely too high for therapeutic applications, these findings demonstrate a compact, all-protein base editing research tool for precise editing of organelle or nuclear DNA without double-strand DNA breaks.

Fig. 1. Optimizing ZF-DdCBEs increases base editing efficiency in mitochondria.

a Architectures of optimized ZF-DdCBEs showing progression from v1 to v8. The components are a mitochondrial targeting signal, FLAG tag, nuclear export signal(s), ZF array with either canonical ZF scaffold (dark grey) or optimized ZF scaffold (light grey), Gly/Ser-rich flexible linker, split DddA deaminase (with or without activity-enhancing mutations and specificity-enhancing mutations) and UGI. b A v8 ZF-DdCBE pair with canonical C-terminal architecture. The ZF-DdCBE pair shown is 9-ND51+R13-ND51. c, d Mitochondrial DNA base editing efficiencies of HEK293T cells treated with (c) six optimized ZF-DdCBE pairs used to establish architectural improvements or (d) seven additional optimized ZF-DdCBE pairs. e A v8 ZF-DdCBE pair with N-terminal architecture. The ZF-DdCBE pair shown is LT51-Mt-tk+RB38-Mt-tk. For (b, e) ZF binding sites are underlined and the cytosine with the highest editing efficiency is colored in blue. For (c–d) values and errors reflect the mean ± s.d. of n = 3 independent biological replicates. The editing efficiencies shown are for the most efficiently edited C•G within the spacing region. Source data are provided as a Source Data file.

Fig. 2. Optimized ZF-DdCBEs display increased base editing efficiency over ZFDs in mitochondria.

a, b Comparison of mitochondrial DNA base editing efficiencies of HEK293T cells treated with either ZFD or optimized ZF-DdCBE pairs at genomic target sites chosen by (a) Lim et al., or (b) by this study. For (a, b) values and errors reflect the mean ± s.d. of n = 3 independent biological replicates. The editing efficiencies shown are for the most efficiently edited C•G within the spacing region. Source data are provided as a Source Data file.

Fig. 3. High-specificity ZF-DdCBE variants reduce mitochondrial off-target editing.

a Mitochondrial DNA base editing efficiencies within amplicon ND4 of HEK293T cells treated with ND4-DdCBE. b Mitochondrial DNA base editing efficiencies within amplicon ATP8 of HEK293T cells treated with v7 ZF-DdCBE pair R8-3i-ATP8+4-3i-ATP8. c Off-target editing efficiencies within mitochondrial off-target amplicon ND5.1 of HEK293T cells treated with ND4-DdCBE, v7 ZF-DdCBE pair R8-3i-ATP8+4-3i-ATP8, or individual components of the v7 ZF-DdCBE architecture. d–f On-target and average off-target editing efficiencies within amplicon ATP8 of HEK293T cells treated with canonical v7 ZF-DdCBE pair R8-3i-ATP8+4-3i-ATP8 (colored in red) or variants containing (d) DddA^N and DddA^C truncations, (e) Ala point mutations within DddA^C, or (f) combinations of DddA^N and DddA^C truncations, point mutations within DddA^C, and/or fused catalytically inactivated DddA^N. High-specificity variants HS1 to HS5 are colored in blue. For (a, b) and (d–f) values reflect the mean of n = 3 independent biological replicates. For (c), values and errors reflect the mean ± s.d. of n = 3 independent biological replicates. For (d–f) the editing efficiencies shown are for the most efficiently edited C•G within the spacing region. Source data are provided as a Source Data file.

Fig. 4. ZF-DdCBEs install pathogenic mutations in cultured cells in vitro.

a The m.8340G>A mutation in human MT-TK disrupts the T-arm of mt-tRNA^Lys. b Mitochondrial DNA base editing efficiencies of HEK293T cells treated with an optimized ZF-DdCBE pair designed to install m.8340G>A. c The m.7743G>A mutation in mouse Mt-tk disrupts the T-arm of mt-tRNA^Lys. d Mitochondrial DNA base editing efficiencies of C2C12 cells treated with an optimized ZF-DdCBE pair designed to install m.7743G>A. e Mitochondrial DNA base editing efficiencies of C2C12 cells treated with an optimized ZF-DdCBE pair designed to install m.3177G>A. For (b, d and e), values and errors reflect the mean ± s.d. of n = 3 independent biological replicates. For each site the DNA spacing region, split DddA orientation, ZF array lengths, and ZF-targeted DNA strands (LT left top, LB left bottom, RB right bottom) are shown, and the cytosine with the highest editing efficiency is colored in blue. Source data are provided as a Source Data file.

Fig. 5. ZF-DdCBEs enable base editing of nuclear DNA.

a Nuclear DNA base editing efficiencies of HEK293T cells treated with five 3ZF+3ZF nuclear-targeted ZF-DdCBE pairs, or ZF-DdCBE variants with extended ZF arrays. ZF-DdCBE pairs were designed to install edits within or nearby nuclear genes COL5A1, DCAF8L2, EMILIN2, EMILIN2, and TRAM1L1, respectively. The editing efficiencies shown are for the most efficiently edited C•G within the spacing region. b Nuclear DNA base editing efficiencies of HEK293T-HBB cells treated with an optimized ZF-DdCBE pair designed to correct the HBB −28A>G mutation. The DNA spacing region, split DddA orientation, ZF array lengths, and ZF-targeted DNA strands (LT left top, RB right bottom) are shown, and the pathogenic cytosine is colored in blue. For (a and b), values and errors reflect the mean ± s.d. of n = 3 independent biological replicates. Source data are provided as a Source Data file.

Fig. 6. In vivo base editing of pathogenic sites in mtDNA.

a Mitochondrial DNA base editing efficiencies installing m.7743G>A of tissue samples from mice treated with buffer, dAAV-Mt-tk, or AAV-Mt-tk. b Mitochondrial DNA base editing efficiencies of tissue samples from AAV-Mt-tk-treated mice. c Off-target editing efficiencies within representative mitochondrial off-target amplicon OT8 of tissue samples from mice treated with buffer, dAAV-Mt-tk, or AAV-Mt-tk. d Mitochondrial DNA base editing efficiencies installing m.3177G>A of tissue samples from mice treated with buffer or AAV-Nd1. e Mitochondrial DNA base editing efficiencies of tissue samples from AAV-Nd1-treated mice. f Off-target editing efficiencies within representative mitochondrial off-target amplicon OT7 of tissue samples from mice treated with buffer, or AAV-Nd1. For (a and b), values and errors reflect the mean ± s.d. of n = 4, 4 and 7 for mice treated with buffer, AAV-Mt-tk, or dAAV-Mt-tk, respectively. For (c), values reflect the mean of n = 4, 4 and 7 for mice treated with buffer, AAV-Mt-tk, or dAAV-Mt-tk, respectively. For (d and e), values and errors reflect the mean ± s.d. of n = 4 and 7 for mice treated with buffer or AAV-Nd1, respectively. For (f), values reflect the mean of n = 4 and 7 for mice treated with buffer or AAV-Nd1, respectively. Source data are provided as a Source Data file.