Efficient generation of mouse models with the prime editing system

Most genetic diseases in humans are caused by single-nucleotide mutations. Although genome editing with either the CRISPR-based cytosine base editor (CBE) 1 or the adenine base editor (ABE) 2 holds great promise for gene correction of C-to-T and A-to-G base substitutions in some genetic diseases 3,4 , both editors are useless for correction of other variants such as base transversion, small insertions and deletions (indels).

PE3 system, to further improve editing efficiency, an additional sgRNA is used to induce nick on the non-edited strand to trigger the endogenous mismatch repair pathway 5 .
In comparison with base editors, PE induces base institutions in more extended regions with fewer bystander mutations 5 . With its unique versatility and accuracy, this technology broadens the scope of genome editing and opens a new avenue for targeted mutagenesis and gene correction in many organisms. However, the efficiency of PE was reported only in five different cell types 5 ; it has not been investigated in animals. Here, we demonstrate that PE can be employed to generate mutant mice with singlenucleotide substitutions.
We first validated the editing versatility of PEs in human HEK293T cells at eight loci (Supplementary Table S1), including two loci (RUNX1 and RNF2) that were reported by Anzalone et al 5 . PE3 was selected for gene targeting validation, due to its higher editing efficiency compared with PE2 5 . Sanger sequencing revealed that PE3 induced significant base conversions at six (RUNX1, RNF2, EMX1, VEGFA, SRD5A3, and KCNA1) out of eight targeted sites ( Supplementary Fig. S1a, b).
PE3 was then used to induce point mutations in the X-linked androgen receptor (Ar) gene and the homeobox protein Hox-D13 (Hoxd13) gene in mouse neuro-2a (N2a) cells 6,7 . Both targeted mutations in mice are homologous to human variants associated with clinical diseases in Clin-Var 8,9 . pegRNAs and nick-editing sgRNAs targeting these two genes were designed (Supplementary Table S2). We designed pegRNAs starting with a primer binding site (PBS) length of 13 nt and an RT template length of~13-15 nt. Nicks were positioned 3′ of the edit~40-60 bp from the pegRNA-induced nick. Sanger sequencing revealed that PE3 efficiently (~8-40%) mediated base transversions at three target sites of Hoxd13 and Ar (pegHoxd13-1 for G to C, pegHoxd13-2 for G to T, pegAr-2 for G to T) ( Fig. 1a; © The Author(s) 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
WT PAM d c 10    Supplementary Fig. S2a-d). Next, to optimize the performance of pegRNAs, we systematically evaluated pegRNAs with RT template lengths of 10-20 nt and a PBS length of 13 nt for pegHoxd13-1, pegHoxd13-2, and pegAr-2 in N2a cells. Targeted deep sequencing revealed that the editing efficiency of pegRNAs with RT template lengths ranging from 14 to 17 nt were stably efficient in Hoxd13 and Ar, respectively ( Fig. 1c; Supplementary Fig. S3a). Then, using the efficient RT template lengths of 14-17 nt, we systematically evaluated pegRNAs in the context of PBS lengths between 10 and 17 nt. Targeted deep sequencing revealed that pegRNAs with a PBS length of 16, 13, and 12 nt for pegHoxd13-1, pegHoxd13-2, and pegAr-2, respectively, had the highest efficiencies; these pegRNAs were used for our in vivo study (Fig. 1b, d; Supplementary Fig. S3b). Anzalone et al. reported that the PE3b system can be applied to the edit lying within a second protospacer to induce fewer indels 5 . Accordingly, we designed nick sgRNA for Hoxd13 based on these parameters and found that the editing efficiency of PE3b was significantly lower than that mediated by PE3 in Hoxd13 ( Supplementary Fig. S4), we then selected PE3 for the further analysis.
To edit mouse embryos with PE3, pCMV-PE2encoding mRNA and corresponding nick sgRNA were co-injected with different pegRNAs into one-cell embryos. For Hoxd13, nucleotide conversions were observed in 8 out of 18 (44%) and 12 out of 16 (75%) blastocysts, respectively, with mutation frequencies ranging from 1.1 to 18.5% (Fig. 1e). We observed a low frequency of indels in injected embryos ( Supplementary  Fig. S5a). Interestingly, we observed not only desirable transversion at the target sites, but also a mixture of nucleotide conversions at these sites ( Supplementary  Fig. S6), implying a relative low fidelity of the PE3 system in murine embryos that the PE is in need of improved fidelity. Moreover, we observed much lower editing efficiency in embryos for the Ar gene (Supplementary Fig. S7), suggesting that screening of pegRNAs in vitro is essential before conducting in vivo studies.
We next delivered mouse embryos that were injected with pCMV-PE2-encoding mRNA and pegRNAs for Hoxd13 into surrogate mothers. Targeted deep sequencing indicated that eight mutant mice out of 30 (Hoxd13-1, editing efficiency of G to C > 1%) and two mutant mice out of 19 (Hoxd13-2, editing efficiency of G to T > 1%) carried mutations at the targeted loci (Fig. 1f, g). Furthermore, the editing efficiency in mutant mice is equivalent to that in prime-edited plants 10 . Together with indel frequency data in embryos and mutant mice (Supplementary Fig. S5), we revealed a low level of indels in the PE system in vivo. As expected, targeted deep sequencing revealed that the editing efficiency varied in the examined nine tissues (heart, liver, spleen, lung, kidney, brain, testis, intestine, and toe) of the two mutant mice (#1-5 and #2-14) ( Supplementary Fig. S8), indicating that the somatic mosaicism and allelic complexity were induced by zygotic injection. As observed in mouse embryos, we found existence of varied frequencies of undesirable base changes at the two target sites (Fig. 1g).
To comprehensively investigate off-target sites throughout the genome, we conducted whole-genome sequencing for the Hoxd13 mutant mice #1-26 (G-to-C) and #2-1 (G-to-T). A total of 3,745,168 and 3,840,538 SNPs were identified for #1-26 and #2-1, respectively ( Supplementary Fig. S10). After filtering out naturally-occurring variants in the dbSNPs database and in two wild-type mice, we picked out the variants with G-to-H (H indicates A/C/T) and C-to-H substitutions, and examined whether the remaining SNPs located at the putative off-target sites. Of 3805 predicted offtarget sites, no base substitutions were uniquely found in #1-26 and #2-1 ( Supplementary Fig. S10). These results demonstrate that the PE3-mediated base conversion is highly specific in vivo.
In sum, we present here the first report of using PEs to generate targeted base conversion mutations in animals. We validate the prime editing system in human cells and demonstrate the versatility of PEs in mice in vivo, though there is a higher frequency of unwanted mutations at target loci. These data support the clinical potential of PEs in correcting a broad range of mutations for genetic diseases.