Nuclear and mitochondrial DNA editing in human cells with zinc finger deaminases

Abstract

Base editing in nuclear DNA and mitochondrial DNA (mtDNA) is broadly useful for biomedical research, medicine, and biotechnology. Here, we present a base editing platform, termed zinc finger deaminases (ZFDs), composed of custom-designed zinc-finger DNA-binding proteins, the split interbacterial toxin deaminase DddA_tox, and a uracil glycosylase inhibitor (UGI), which catalyze targeted C-to-T base conversions without inducing unwanted small insertions and deletions (indels) in human cells. We assemble plasmids encoding ZFDs using publicly available zinc finger resources to achieve base editing at frequencies of up to 60% in nuclear DNA and 30% in mtDNA. Because ZFDs, unlike CRISPR-derived base editors, do not cleave DNA to yield single- or double-strand breaks, no unwanted indels caused by error-prone non-homologous end joining are produced at target sites. Furthermore, recombinant ZFD proteins, expressed in and purified from E. coli, penetrate cultured human cells spontaneously to induce targeted base conversions, demonstrating the proof-of-principle of gene-free gene therapy.

Fig. 1. Development of ZFDs.

a ZFD (zinc finger deaminase) architecture. Split-DddA_tox halves are fused to the C terminus of ZFPs (zinc finger proteins) (C type). b Optimization of the ZFD platform using pTarget libraries. pTarget plasmids contain a spacer region that ranges in size from 1–24 bp (shown in red) and ZFP DNA binding sites (shown in green). ZFD constructs contain AA (amino acid) linkers of different lengths (shown in yellow and orange) and different DddA_tox split sites and orientations (shown in blue). c, d ZFD activities were measured at on-target sites in the pTarget library to examine the effect of the variables described in (b). ZFD pairs with linkers of the same (c) or different (d) lengths in the left and right ZFD were tested. Base editing frequencies were measured by targeted deep sequencing of the relevant region of pTarget plasmids. Data are shown as means from n = 2 biologically independent samples. Source data are provided as a Source Data file.

Fig. 2. Cytosine base editing by ZFDs at endogenous target sites.

a Architecture of nuclear DNA-targeting ZFDs. Split-DddA_tox halves are fused to the C terminus (C type) or N terminus (N-type) of ZFPs. ZFD pairs were designed in CC or NC configurations, which are composed of a C-type left ZFD and a C-type right ZFD or an N-type left ZFD and a C-type right ZFD, respectively. b Base editing frequencies induced by ZFDs at endogenous target sites in HEK 293 T cells. All statistical analysis for comparing with untreated samples was conducted using unpaired Student’s t-test (two-tailed) in GraphPad Prism 8. Statistical significance as compared with untreated samples was denoted with *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, n.s. (not significant) P > 0.05. Data are shown as means with standard error of the mean (s.e.m.) from n = 3 biologically independent samples. c–f ZFD-induced base editing efficiencies at each base position within the spacer at the NUMBL (c), INPP5D-2 (d), TRAC-CC (e), and TRAC-NC (f) target sites in HEK 293 T cells. Data are shown as means ± s.e.m. from n = 3 biologically independent samples. g ZFD-induced base editing frequencies in K562 cells following electroporation or direct delivery of ZFDs or ZFD-encoding plasmids. ZFD proteins with one or four NLSs were tested, and equimolar amounts of left and right ZFDs were used. Electroporation was performed using an Amaxa 4D-Nucleofector. For direct ZFD delivery, K562 cells were incubated with a cell medium containing left and right ZFD proteins. Cells were treated either once (1x) or twice (2x) in the same way. Data are shown as means from n = 2 biologically independent samples. Source data are provided as a Source Data file.

Fig. 3. mtDNA editing with mitoZFDs.

a Base editing frequencies in mtDNA induced by mitoZFDs and a TALE-DdCBE in HEK 293 T cells. All statistical analysis for comparing with untreated samples was conducted using unpaired Student’s t-test (two-tailed) in GraphPad Prism 8. Statistical significance as compared with untreated samples was denoted with *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, n.s. (not significant) P > 0.05. Data are shown as means from n = 2 biologically independent samples. b–g mitoZFD-induced base editing efficiencies at each base position within the spacer at the ND2 (b), ND4L (c), COX2 (d), ND5-2 (e), and ND1 (f) target sites, and TALE-DdCBE-induced base editing efficiencies at the ND1 (g) target site, in HEK 293 T cells. Data are shown as means from n = 2 biologically independent samples. h Comparison of DNA changes and amino acid changes in the ND1 gene introduced by mitoZFD and TALE-DdCBE. The reference sequence (Ref.) is at the top. In the alleles, the red letters indicate changes in the amino acid sequence. (* indicates a stop codon.) The frequency of sequencing reads (%) for each mutant allele was measured by targeted deep sequencing. The spacer regions for the ZFD pair and the TALE-DdCBE pair are indicated with blue dashed lines. Source data are provided as a Source Data file.

Fig. 4. Activity of mitoZFDs, TALE-based DdCBEs, and ZFD/DdCBE hybrid pairs.

a DNA sequences of the binding regions of the mitoZFD and TALE-DdCBE pairs. Sites recognized by the TALE-DdCBEs are highlighted in green and for the mitoZFDs in blue. The upper sequence represents the mtDNA heavy strand and the lower sequence represents the mtDNA light strand. b Frequencies of cytosines edited by ZFDs, TALE-DdCBEs, and ZFD/DdCBE hybrid pairs. All statistical analysis for comparing with untreated samples was conducted using unpaired Student’s t-test (two-tailed) in GraphPad Prism 8. Statistical significance as compared with untreated samples was denoted with *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001, n.s. (not significant) P > 0.05. Data are shown as means from n = 2 biologically independent samples. c Heat maps of base editing activities at each base position. The red box indicates the spacer region for each construct. The blue arrows indicate the position in the mtDNA. Source data are provided as a Source Data file.

Fig. 5. Improving the mitochondrial genome-wide target specificity of mitoZFDs.

a QQ mitoZFD variants contain R(-5)Q mutations in each zinc finger in the ZFD to remove non-specific DNA contacts. F1–F4; Finger 1–4. (If no R was present at position-5 of the zinc finger framework, a nearby K or R was converted to Q.) b Whole-mtDNA sequencing of mitoZFD-treated cells. Editing frequencies at on-target and off-target sites are indicated by red and black dots, respectively. Error bars are shown as standard error of the mean (s.e.m.) for n = 2 biologically independent samples. All C/G-to-T/A base changes present at frequencies >1% are presented. c–e Editing efficiencies and specificities depend on the dose of ZFD-encoding mRNA delivered. Data are shown as means from n = 2 biologically independent samples. c The average C/G-to-T/A editing frequency for all C/Gs in the mitochondrial genome. d The number of edited C/G sites with base editing frequencies >1%. e The specificity ratio was calculated by dividing (average editing frequency at on-target Cs) by (average editing frequency at off-target Cs). Source data are provided as a Source Data file.