To verify the applicability of the prime-editing system in the mammalian genome, we used the tdTomato-expressing reporter system at the AAVS1 locus in HEK293T cells. We designed the stop codon by inserting a T nucleotide into the tdTomato sequence using prime editor 3 (PE3) with a primer binding site (PBS) length of 8 nt and a reverse transcriptase (RT) template length of 17 nt (PBS8-RT17). Moreover, prime-editing guide RNA (pegRNA) was designed to remove the protospacer adjacent motif (PAM) sequence on the non-target strand to inhibit editing on the edited strand. We observed that tdTomato-negative cells represent 32% by flow cytometric analysis (Additional file 1: Figure S1), suggesting that the PE enables the desired target mutagenesis in our system.
Next, to apply the PE to mouse model generation, we tried to design targets involving eight transversion mutations or the insertion or deletion of one or more nucleotides. We first induced stop codons in two mouse genes, Igf2 and Adamts20. Igf2 can induce a dwarf phenotype caused by a mutation in the Igf2 allele inherited from the father [11]. We inserted TA nucleotides into exon4 of the Igf2 gene to generate a stop codon for loss of function. Besides, the nucleotide of the PAM sequence was changed from NGG (where “N” is any nucleotide base) to NCG to prevent continuous editing on the edited strand. Adamts20 is a gene involved in the development of melanocytes. The premature stop codon at the E584 site of the Adamts20 locus is linked to a typical white-belt phenotype [12]. We induced the conversion of nucleotides from CG to TT at exon12 of the Adamts20 locus, resulting in a premature stop codon (E584*) and PAM modification (NGG to NAG) (Additional file 1: Figure S2a and S2b). To induce mutagenesis in the mouse targets, we used a PE3 system consisting of PE (nCas9 fused with engineered M-MLV RT), pegRNA, and nicking sgRNA (nsgRNA); the nsgRNA enhances the editing efficiency by promoting DNA repair activity through the cleavage of the non-edited strand [7].
First, we tested the editing efficiency of the pegRNAs with varying PBS lengths (8–14 nt) and RT template lengths (10–18 nt) to optimize the prime-editing at the Igf2 and Adamts20 sites. We excluded the length of PBS with thymine at the 3′-end (which could be part of the transcription termination signal) and the length of RT with cytosine at the 5′-end (which can interfere with the pegRNA structure) [7]. We transfected three plasmids encoding PE, pegRNA, and nsgRNA into NIH/3T3 cells via electroporation and harvested the cells after 72 h for targeted deep-sequencing. However, the editing efficiency with PE3 in the NIH/3T3 cells was less than 3% on the Igf2 and Adamts20 targets (Additional file 1: Figure S2c and 2d). These results suggest that the prime editor requires improvement for use in mouse model generation.
In one attempt to improve the editing efficiency of PE, we employed dsgRNAs based on the idea of proxy-CRISPR [13] instead of catalytically dead endonuclease to unwind the chromatin structure of the target sites. A dsgRNA is a 14- or 15-nt guide RNA that shows inactivated catalysis yet binds to the target site guiding Cas endonuclease [14]. Thus, we hypothesize that PE may play two roles: one is prime-editing at the target site with a pegRNA and the other is modulating the chromatin neighboring the target site with a dsgRNA. We designed proximal dsgRNAs adjacent to the Igf2 and Adamts20 target sites in the range of 7–62 nucleotide positions from the spacer of pegRNA. We applied proximal dsgRNAs to various pegRNA lengths at the Igf2 and Adamts20 sites to identify the editing efficiency, and interestingly, that of PE3 using proximal dsgRNA was improved in most groups. We chose PBS9-RT14 pegRNA and PBS11-RT13 pegRNA with the highest efficiencies for the Igf2 and Adamts20 targets, respectively (Additional file 1: Figure S2c and S2d).
Next, to select a dsgRNA with high editing efficiency, we designed and tested additional proximal dsgRNAs in Igf2, Adamts20, Casp1 (4-bp deletion), Hoxd13 (G to T conversion), Angpt1 (CGG to TGA conversion), and Ksr2 (TGAT insertion). The PE3 with a proximal dsgRNA was delivered into NIH/3T3 and C2C12 cells via electroporation with plasmids encoding PE, pegRNA, nsgRNA, and each proximal dsgRNA. Targeted deep-sequencing data reveals that proximal dsgRNA selectively improved the editing efficiency in most of the targets compared to PE3 (Additional file 1: Figure S3). Overall, the editing efficiency with dsgRNA depended on the proximal dsgRNA position, and so a screening process for an optimal dsgRNA is required for each target and cell type to induce effective targeted mutagenesis.
We also engineered the PE using chromatin-modulating peptides (CMPs), high-mobility group nucleosome binding domain 1 (HN1), and histone H1 central globular domain (H1G) [15] to increase the editing efficiency. CMP-PE-V1 consists of HN1 at the N-terminus and H1G at the C-terminus of nCas9. CMP-PE-V2 consists of HN1 at the N-terminus of nCas9 and H1G at the C-terminus of engineered M-MLV RT (Fig. 1a). We delivered engineered CMP-PE3-V1 (CMP-PE-V1 with pegRNA/nsgRNA) or CMP-PE3-V2 (CMP-PE-V2 with pegRNA/nsgRNA) into two mouse cell lines to compare the editing efficiency with PE3 (Additional file 1: Figure S4). We observed that CMP-PE3-V1 was significantly more efficient in all target sites compared to PE3. In particular, the editing efficiency of CMP-PE3-V1 was up to 2.55-fold higher in the Igf2 target and up to 3.92-fold higher in the Adamts20 target than by PE3, respectively, in NIH/3T3 (Additional file 1: Figure S4). These results suggest that engineered PEs using CMP HN1 and H1G can improve editing efficiency.
CMP-PE3-V1 and dsgRNA (CMP-PE3-V1 + dsgRNA) were delivered to mouse cells to test for any synergistic effects; CMP-PE3-V1+dsgRNA achieved improvements of up to 4.20-fold, 5.11-fold, and 3.56-fold prime-editing efficiency at the Igf2, Adamts20, and Hoxd13 target sites, respectively, compared to PE3. However, the synergistic effect differed depending on the cell line and target (Fig. 1b–g). These results suggest that although CMP-PE3 could efficiently enhance the prime-editing in all of the targets and cell types, the efficiency of PE3+dsgRNA or CMP-PE3+dsgRNA varied depending on the target site and cell type.
Next, we carried out targeted mutagenesis in mouse embryos via microinjection using advanced prime-editing systems. We selected the Igf2 target site that showed relatively low undesired mutations among the designed mouse targets. CMP-PE3-V1 and CMP-PE3-V1+dsgRNA appeared to have significantly higher editing efficiency at the Igf2 target in mouse embryos. Notably, in the case of CMP-PE3-V1+dsgRNA, the desired mutations at the target site of the Igf2 were observed in 21 out of 22 embryos (95%), suggesting that CMP-PE3-V1+dsgRNA has a much higher prime-editing efficiency than PE3 and PE3+dsgRNA (Fig. 1h and Additional file 1: Table S1). We further tested using CMP-PE-V1 and a proximal dsgRNA to verify the prime-editing efficiency at Adamts20, Hoxd13, Angpt1, Ksr2, and Ar in mouse embryos. Targeted deep-sequencing data revealed that CMP-PE-V1 and a proximal dsgRNA improved prime-editing efficiency in the Adamts20, Hoxd13, and Angpt1 targets but not the Ksr2 and Ar targets compared to PE3 (Fig. 1i–m and Additional file 1: Table S1). Taken together, our results suggest that improved prime-editing method using a proximal dsgRNA and chromatin-modulating peptides in mouse embryos can be efficient.
To determine whether CMP-PE-V1 and dsgRNA can improve chromatin accessibility by unraveling the chromatin structure of the target site, we performed a DNaseI digestion assay and qPCR to verify the chromatin status of each target site. In the NIH/3T3 cell line, Igf2, Adamts20, and Hoxd13 appeared as relatively closed-chromatin structures, and in C2C12, all targets except Igf2 were open chromatin (Fig. 1n and Additional file 1: Figure S5). This result suggests that even when the target sequences are the same for the two cell lines, their chromatin structure states are different. Furthermore, we analyzed whether CMP-PE-V1 or dsgRNA could alter the chromatin status at the Igf2 target site (a representative target with a closed-chromatin structure). From the results, we identified that unlike PE3, the closed-chromatin structure was gradually opened by CMP-PE-V1, dsgRNA, or CMP-PE-V1+dsgRNA (Fig. 1o). These results are direct evidence that the use of CMP-PE-V1 or dsgRNA can improve prime-editing efficiency by unraveling chromatin structures and improving chromatin accessibility.
Next, we carried out targeted mutagenesis of Igf2 in mouse embryos via microinjection using PBS9-RT14 and dsgRNA +7 with a relatively low frequency of undesired mutations. Injected mouse embryos were transplanted into surrogate mothers, and pups were obtained with G to C transversion and TA insertion at the Igf2 locus at prime-editing frequencies of up to 47% (2 out of 10 pups) (Fig. 2a–c). Moreover, we identified that 7 out of 9 pups from an F1 littermate of the Igf2 mutant mouse had germline transmission (Additional file 1: Figure S6).
To assess the off-target effects of the prime editor, we used Cas-OFFinder [16] to identify potential off-target sites by pegRNA and nsgRNA of the Igf2 target with up to three nucleotide mismatches in the mouse genome. Off-target mutations were not detected at the potential off-target sites compared to the wild-type (Additional file 1: Figure S7). Moreover, we carried out whole-genome sequencing to identify off-target effects in the Igf2 mutant mouse and found a single off-target site of nsgRNA. However, we determined that this site was a false positive through Sanger sequencing with genomic DNA (Fig. 2d, e). To identify the phenotype of the Igf2 mutant mice, the Igf2 p+/m− male (F1) was mated with a wild-type female. The Igf2 p−/m+ mouse carrying the mutation for the Igf2 gene inherited from the paternal allele exhibited the dwarf phenotype, which is consistent with the desired mutant genotype (Fig. 2f, g). These results suggest that the newly improved prime-editing tools can be effectively applied to mouse model generation.