Comparative evaluation of in vivo editing efficiencies for three representative TadA‐derived cytosine base editors in zebrafish. a) Schematic of the mRNA construct for four cytosine base editors. bpNLS: bipartite nuclear localization, apobec1, eTadA*, TadA‐CDa and CBE‐1.14: various adenine deaminase, XTEN: a 32aa flexible linker, nSpCas9: SpCas9 nickase, GGS linker: GGSSGGS amino acid, P2A: Porcine teschovirus‐1 2A, UGI: Uracil glycosylase inhibitor. b) The C‐to‐T editing efficiency of AncBE4max, TadCBEmax, TadCBEa, CBE‐1.14 was examined at 10 endogenous genomic loci. The heatmap represents the average editing percentage derived from three independent experiments. c) Evaluation of the efficiency and targeting window of all four CBEs based on 10 sites in Figure 1b. Each data point reflects the mean editing activity at a specific site (PAM located at positions 21–23). Data from three independent experiments were analyzed. d) Assessment of the mean editing efficiency of four cytosine base editors using the plot based on the data in Figure 1b. Mean editing efficiency at each site is represented by individual data points, with the central dotted line indicating the overall mean. Two‐tailed paired t‐test was performed: not significant (ns) P ≥0.05, * P < 0.05, ** P < 0.01, and *** P < 0.001. e) The indel efficiency comparison among four cytosine base editors targeting ten different loci. Values are presented as mean value ± standard deviation (SD), n = 3 biological replicates. Data are expressed as mean ± SD. f) Assessment of the mean indel efficiency of four cytosine base editors using the plot based on the data in Figure 1e. Mean indel frequency at each site is represented by individual data points, with the central dotted line indicating the overall mean. Two‐tailed paired t‐tests were performed: not significant (ns) P ≥0.05, * P < 0.05, ** P < 0.01, and *** P < 0.001.

Efficient cytosine base editing mediated by zTadCBE in zebrafish. a) Schematic overview of the engineering strategy for zTadCBE. Starting from TadCBEmax (low indel), modifications were introduced sequentially: TadCBEa (increased efficiency), TadCBEa‐2xUGI (reduced indels while maintaining efficiency), and final zTadCBE, which incorporates additional mutations (V82S & Q154R) in the deaminase domain (eTadA*) for enhanced activity and editing precision. Spheres of different colors indicate deaminase variants with distinct mutations, nSpCas9 (D10A): SpCas9 nickase, UGI: Uracil glycosylase inhibitor. b) Heatmaps showing C‐to‐T base editing efficiencies among TadCBEmax, TadCBEa, TadCBEa‐2XUGI, and zTadCBE across ten target loci. Base position within the gRNA is denoted numerically, and values are reported as mean ± standard deviation (SD), with n = 3 biological replicates. Statistical analysis was conducted using a two‐tailed paired t‐test, not significant (ns) P ≥0.05, * P < 0.05, ** P < 0.01, and *** P < 0.001. c) Analysis of mean editing efficiency for TadCBEmax, TadCBEa, TadCBEa‐2XUGI, and zTadCBE based on data in Figure 2b. Mean editing efficiency per site is shown by individual data points, with the central dotted line representing the overall mean. Two‐tailed paired t‐tests were performed: not significant (ns) P ≥0.05, * P < 0.05, ** P < 0.01, and *** P < 0.001. d) Analysis of mean indel frequency for TadCBEmax, TadCBEa, TadCBEa‐2XUGI, and zTadCBE based on data in Figure 2b. Mean editing efficiency per site is shown by individual data points, with the central dotted line representing the overall mean. Two‐tailed paired t‐tests were performed: not significant (ns) P ≥0.05, * P < 0.05, ** P < 0.01, and *** P < 0.001. e) Heatmap illustrating the average C‐to‐T editing efficiency of zTadCBE across 23 target sites. Editing efficiency is shown on a color scale, where blue represents 100% efficiency and white represents 0% efficiency. f) Base editing efficiencies of the zTadCBE systems at the target C in different sequence contexts. Each data point reflects the mean editing activity at a specific site. Statistical analysis was conducted using a two‐tailed paired t‐test, not significant (ns) P ≥0.05, * P < 0.05, ** P < 0.01, and *** P < 0.001.

Broadening the Targeting Range and Editing Windows by zTadCBE variants. a) Comparison of average C‐to‐T editing efficiencies between zTadCBE‐SpRY and CBE4max‐SpRY using twelve gRNAs targeting NNN PAMs. zTadCBE‐SpRY shows consistently higher editing efficiencies at multiple non‐NGG PAM sites, demonstrating improved PAM flexibility and activity. The position of the edited base within each gRNA is indicated numerically. Data are presented as mean values ± standard deviation (SD), calculated from three biological replicates. b) Analysis of mean editing efficiency for zTadCBE‐SpRY and CBE4max‐SpRY based on data in Figure 3a. Mean editing efficiency per site is shown by individual data points, with the central dotted line representing the overall mean. Two‐tailed paired t‐tests were performed, with P values annotated at the top of the violin plot. c) Schematics showing constructs of zTadCBE‐ex1 designed to shift the editing window of cytosine base editing.Heatmap illustrating the average C‐to‐T editing efficiency of zTadCBE‐ex1 across 16 target sites. Editing efficiency is shown on a color scale, where blue represents 80% efficiency and white represents 0% efficiency. d) Comparison of editing efficiency and activity windows between zTadCBE and zTadCBE‐ex1. Each data point indicates the mean editing efficiency at a given target site. The editing windows, defined from the 5′ to 3′ end of the protospacer, are highlighted in pink (positions 4–8) for zTadCBE and in blue (positions 4–13) for zTadCBE‐ex1. Data were derived from three biologically independent replicates. e) Schematic representations of zTadCBE‐ex2 constructs. A heatmap depicting the average C‐to‐T editing efficiency of zTadCBE‐ex2 across 19 target sites is shown, with a color gradient indicating efficiency levels, where blue denotes 70% efficiency and white indicates 0% efficiency. f) Comparison of editing efficiency and activity windows between zTadCBE and zTadCBE‐ex2. Each data point indicates the mean editing efficiency at a given target site. The editing windows, defined from the 5′ to 3′ end of the protospacer, are highlighted in pink (positions 4–8) for zTadCBE and in yellow‐brown (positions 5–16) for zTadCBE‐ex2. Data were derived from three biologically independent replicates.

Off‐target analysis of zTadCBE in zebrafish. a) On‐target, product purity, and gRNA‐dependent off‐target analysis of zTadCBE induced C‐to‐T editing at p53, kcnq3‐T2, and spata5l1‐T2 sites using NGS. The top three high‐scoring off‐target sites (PAMs are underlined) are shown. Mismatched bases are indicated in lowercase. The PAM sequences are underlined in red. b) Schematic of Cas9‐independent deamination of cytosines within dSaCas9‐induced R‐loops by SpCas9 zTadCBE. c) Bar graphs depicting the efficiency of on‐target editing for different conditions, with heights representing the percentage of successful edits and colors indicating different targets, cdh23‐R2032W, kcnq3‐T2, and med12‐g2. Data are presented as mean values ± standard deviation (SD), calculated from three biological replicates. d) Bar graphs showing the editing efficiencies of three R‐loop regions. e) Transcriptome analysis of edited cytosine nucleotides in zebrafish embryos. Embryos were injected with zTadCBE+ spata5l1‐T2, zTadCBE + kcnq3‐T2 or zTadCBE mRNA only. Data from two independent replicates are shown. f) RNA C‐to‐U conversion frequencies in injected zebrafish embryos. The numbers of C‐to‐U RNA edits are indicated on the plots. Data from two independent replicates are shown.

Disease modeling using zTadCBE editors. a) Schematic representation of CRISPR target site in the cdh23 (R2032W) showing the genomic structure with exons (black bars) and introns in zebrafish. The targeted sequence is displayed with the PAM underlined and bold. The targeted nucleotide is highlighted in red. Next‐generation sequencing analysis of F0 zebrafish embryos injected with zTadCBE‐SpRY mRNA and gRNA targeting the cdh23 (R2032W) locus. The predominant sequencing products with the highest frequencies are shown. The red asterisk indicates the desired edited product. b) Phenotypes of 5 dpf cdh23 (R2032W) F0 embryos induced by zTadCBE‐SpRY. Compared to the control, F0 cdh23 (R2032W) embryos exhibited a significant reduction in functional neuromast hair cells (labeled with YO‐PRO‐1 (green)). FM1‐43FX uptake was markedly reduced in hair cells of cdh23 (R2032W) F0 embryos, indicating impaired mechanotransduction. OTOF immunostaining showed no appreciable difference between WT and mutant groups, suggesting preserved hair cell identity despite functional deficits. c) Schematic diagrams of med12 (Q432*) in zebrafish. The targeted sequence is displayed with the PAM underlined and bold. The targeted nucleotide is highlighted in red, and a black bar indicates the related coding frame. Next‐generation sequencing analysis of F0 zebrafish embryos injected with zTadCBE mRNA and gRNA targeting the med12 (Q4322*) locus. The predominant sequencing products with the highest frequencies are shown. The red asterisk indicates the desired edited product. d) Phenotypes of med12 (Q432*) F0 embryos induced by zTadCBE. Compared to the control, med12 (Q432*) F0 embryos exhibited microcephaly, microphthalmia, and pericardial edema (indicated by the red arrowhead) phenotypes. MyHC and phalloidin staining revealed that 3‐dpf med12 (Q432*) F0 embryos exhibit abnormal heart chamber morphogenesis, implicating heart looping defects. Alcian blue staining further demonstrated craniofacial abnormalities in med12 (Q432*) F0 embryos at 4 dpf. From the lateral view, the magenta line shows the angle and extent of Meckel's cartilage and palatoquadrate. In the ventral view, med12(Q432*) F0 embryos present a more expanded angle for ceratohyal, misaligned and shortened palatoquadrate, and kinked Meckel's cartilage. Overall, the craniofacial morphology of med12(Q432*) F0 embryos exhibits severe deformation. (ch: ceratohyal, pq: palatoquadrate, m: Meckel's cartilage). e) Schematic diagrams of etfa (Q266*) in zebrafish. The targeted sequence is displayed with the PAM underlined and bold. The targeted nucleotide is highlighted in red, the related coding frame is indicated by a black bar. Next‐generation sequencing analysis of F0 zebrafish embryos injected with zTadCBE mRNA and gRNA targeting the etfa (Q266*) locus. The predominant sequencing products with the highest frequencies are shown. The red asterisk indicates the desired edited product. f) Schematic diagrams of myhz1.1 (R1398C) in zebrafish. The targeted sequence is displayed with the PAM underlined and bold. The targeted nucleotide is highlighted in red, and the related coding frame is indicated by a black bar. Next‐generation sequencing analysis of F0 zebrafish embryos injected with zTadCBE mRNA and gRNA targeting the myhz1.1 (R1398C) locus. The predominant sequencing products with the highest frequencies are shown. The red asterisk indicates the desired edited product. g) Phenotypes of etfa (Q266*) F0 embryos induced by zTadCBE. Compared to the control, etfa (Q266*) F0 embryos exhibited global lipid accumulation when labelled with Oil Red O. In contrast, the deposition of lipid is restricted to the liver in the control. Fluorescent imaging further showed ectopic lipid deposition in vascular structures (indicated by black arrowhead). h) Phenotypes of 5 dpf myhz1.1 (R1398C) F0 embryos induced by zTadCBE. Phalloidin staining of F‐actin revealed a disorganized myotome structure across somites in myhz1.1 myhz1.1 (R1398C) F0 embryos, with defects most pronounced in the anterior trunk region compared to the control.