Improved prime editors enable pathogenic allele correction and cancer modelling in adult mice

Abstract

Prime editors (PEs) mediate genome modification without utilizing double-stranded DNA breaks or exogenous donor DNA as a template. PEs facilitate nucleotide substitutions or local insertions or deletions within the genome based on the template sequence encoded within the prime editing guide RNA (pegRNA). However, the efficacy of prime editing in adult mice has not been established. Here we report an NLS-optimized SpCas9-based prime editor that improves genome editing efficiency in both fluorescent reporter cells and at endogenous loci in cultured cell lines. Using this genome modification system, we could also seed tumor formation through somatic cell editing in the adult mouse. Finally, we successfully utilize dual adeno-associated virus (AAVs) for the delivery of a split-intein prime editor and demonstrate that this system enables the correction of a pathogenic mutation in the mouse liver. Our findings further establish the broad potential of this genome editing technology for the directed installation of sequence modifications in vivo, with important implications for disease modeling and correction.

Fig. 1. Improved NLS composition enhances prime editing efficiency.

a Schematic representation of the original prime editor (PE2) and optimized prime editor (PE2*) carrying additional NLS sequences at the N-terminus and C-terminus. The M-MLV reverse transcriptase in PE2 was fused to a SaCas9 nickase (SaPE2*) or a SaCas9^KKH nickase (Sa^KKHPE2*) with the same NLS composition to develop orthogonal prime editors. BP-SV40 NLS = bipartite SV40 NLS; vBP-SV40 NLS = variant BP-SV40 NLS. b Diagram of the A•T-to-G•C transition required to convert a stop codon to GLN to restore function to a mCherry reporter in HEK293T cells (top). Frequencies of targeted A•T-to-G•C transition by different prime editors (PE2, PE2*, and Sa^KKHPE2*) were quantified by flow cytometry (bottom). c Diagram of the deletion reporter in HEK293T cells containing a broken GFP with 47-bp insertion, P2A, and out-of-frame mCherry (top). Targeted, precise deletion of 47 bp will restore GFP expression, whereas indels that create a particular reading frame alteration produce mCherry expression. Frequencies of precise deletion (GFP+) and indel (mCherry+) introduced by different prime editors (PE2, PE2*, and Sa^KKHPE2*) were quantified by flow cytometry (bottom). d Diagram of the insertion reporter in HEK293 cells containing a broken GFP with 39-bp insertion, T2A, and mCherry (top). Targeted, precise insertion of 18 bp that substitutes for a disrupting sequence can restore GFP expression, whereas indels that create a particular reading frame alteration produce mCherry expression. Frequencies of targeted 18-bp replacement and indel generation by different prime editors (PE2, PE2*, and Sa^KKHPE2*) were quantified by flow cytometry (bottom). All expression vectors were delivered by transient transfection. The presence of sgRNAs to promote nicking of the complementary strand is indicated in each figure legend. Results were obtained from six independent experiments and presented as mean ± SD. **P < 0.01, ***P < 0.001, ****P < 0.0001 by one-way ANOVA with Tukey’s multiple comparisons test between each PE2 and PE2* using the same nicking sgRNA.

Fig. 2. Improved PE2* increases editing efficiency at endogenous loci.

a Comparison of editing efficiency for nucleotide substitution, targeted 3-bp deletion, and 6-bp insertion with PE2 and PE2* at EMX1 locus in HEK293T cells. Indels broadly indicate mutations to the endogenous sequence that do not result in the desired sequence alteration. b–c Editing efficiency for nucleotide substitution, targeted 3-bp deletion, and 6-bp insertion with SaPE2* (b) and Sa^KKHPE2* (c) at EMX1 locus in HEK293T cells. d Sequence of CCR5 locus and pegRNA used for the 32 bp deletion. Two mutations in red were included to demonstrate that sequence collapse was not a function of nuclease-induced microhomology-mediated deletion and to reduce re-cutting of deletion allele. The bottom panel shows the alignment of pegRNA with the CCR5 sense strand. e Comparison of efficiency for generating a targeted 32-bp deletion with PE2, PE2*, and Sa^KKHPE2* within CCR5 in HeLa cells. All expression vectors were delivered by transient transfection. The presence of sgRNAs to promote nicking of the complementary strand is indicated in each figure panel. Results were obtained from three independent experiments and presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by one-way ANOVA with Tukey’s multiple comparisons test. ns, not significant.

Fig. 3. Improved PE2* increases the correction efficiency of a pathogenic mutation in vivo.

a Installation (via G•C-to-A•T) of the pathogenic SERPINA1 E342K mutation in HEK293T cells using PE2, PE2*, and Sa^KKHPE2*. Editing efficiencies reflect sequencing reads which contain the desired edit. The presence of sgRNAs to promote nicking of the complementary strand is indicated on the x-axis. Results were obtained from three independent experiments and are presented as mean ± SD. b pegRNA used for correction (via A•T-to-G•C) of the E342K mutation includes a spacer sequence, a sgRNA scaffold, an RT template including edited bases (red), and a primer-binding site (PBS). A PAM mutation (AGG to AAG) was introduced to reduce re-cutting of the locus that results in a synonymous codon change. c Evaluating PE expression and subcellular distribution in mouse liver. FVB mice were injected with PE2 or PE2* expression plasmids containing a 3xHA-tag. IHC was performed with an HA-tag antibody. Scale bars: 100 µm (×20 lens). d Average percentage of HA-tag signal from the nucleus. Each dot is the average calculated signal intensity within the nucleus relative to the whole cell from all positive cells in a microscopic image. Numbers are mean ± sem (n = 20 total images from 3 mice). e Schematic overview of correction strategy of the SERPINA1 E342K mutation in PiZ transgenic mouse model of AATD. Prime editor, pegRNA, and nicking sgRNA plasmid were delivered by hydrodynamic tail-vein injection. f Comparison of the efficiency of K342E correction and indels in mouse livers in PE2 or PE2* treatment groups. Precise editing is defined as the fraction of sequencing reads with both A to G prime editing and synonymous PAM modification. Results were obtained from three mice and presented as mean ± SD. **P < 0.01, ***P < 0.001, ****P < 0.0001 by one-way ANOVA with Tukey’s multiple comparisons test.

Fig. 4. Generating mouse cancer models using improved PE2*.

a pegRNA used for installation (via C•G-to-T•A) of the oncogenic S45F in Ctnnb1 in mouse liver. b Schematic overview of the somatic cell editing strategy to drive tumor formation. Prime editor (PE2 or PE2*), pegRNA for Ctnnb1 S45F, and nicking sgRNA plasmids were delivered by hydrodynamic tail-vein injection along with the MYC transposon and transposase plasmids. c Representative images of tumor burden in mouse liver with PE2 or PE2*. d Tumor numbers in the livers of mice 25 days after injection with PE2 or PE2*. The Control group was pegRNA only. Results were obtained from 4 mice and presented as mean ± SD. e Sanger sequencing from normal liver and representative tumors. The dashed box denotes C to T editing in tumors. *P < 0.05 by one-way ANOVA with Tukey’s multiple comparisons test. f Schematic of Ctnnb1 S45 deletion strategy using PE2* (S45del). pegRNA used for 3 bp deletion (TCC) is shown. g PE2* treatment leads to oncogenic activation of Ctnnb1. Prime editor (PE2*), pegRNA (Ctnnb1 S45del or SERPINA1), and nicking sgRNA plasmids were delivered by hydrodynamic tail-vein injection along with the MYC transposon and transposase plasmids. Mice treated with the pegCtnnb1 S45del (n = 4) displayed a large number of liver tumors whereas mice treated with pegSERPINA1 as a control displayed no noticeable oncogenic lesions. beta-Catenin (CTNNB1) IHC staining was performed. Scale bars: 100 µm (×20 lens). h Prime editing efficiency and indels determined by targeted deep sequencing in control liver and representative tumors. Results were obtained from 3 tumors in each group and presented as mean ± SD.

Fig. 5. Systemic injection of dual AAV8 split-intein prime editor achieves pathogenic mutation correction in PiZ mice.

a Schematic of split-intein dual AAV prime editor. Full-length primer editor (original PE2) was reconstituted from two PE2 fragments employing the Npu DNAE split intein. C, carboxy-terminal; N, amino-terminal. b Schematic of the in vivo experiments. Dual AAV8 split-intein prime editor (2 × 10¹¹ vg total) was delivered to six-week-old PiZ mice by tail-vein injection. Livers were harvested at 2 (n = 2), 4 (n = 3), and 10 (n = 3) weeks after injection and the genomic DNA was isolated for sequencing. c Prime editing efficiency of K342E correction and indels determined by targeted deep sequencing in mouse livers of dual AAV-treated mice. Precise editing is defined as the fraction of sequencing reads with both A to G prime editing and synonymous PAM modification. Results were obtained from two (2 weeks) or three mice (6 and 10 weeks) and presented as mean ± SD. **P < 0.01, ***P < 0.001 by one-way ANOVA with Tukey’s multiple comparisons test. d Composition of edited alleles at SERPINA1 by UDiTaS analysis. The circle plot shows the fraction of edits that are precise (intended base conversion), small indels (<50 bp) or substitution, deletions between pegRNA and nicking sgRNA sites (<100 bp), large deletions (>100 bp), and AAV fragment insertion. Numbers are average of 3 mice in 10 weeks treated cohort. e The statistically significant large deletion sequences detected by UDiTaS in the 10 weeks treated cohort are displayed as bars spanning the sequence that is deleted (a representative liver of n = 3 mice). Positions of the pegRNA and nicking sgRNA are indicated by dotted lines and the approximate positions of the locus-specific UDiTaS primers are indicated by arrows below the bar chart. The deletion size and number of UMIs associated with each deletion are indicated to the right of each bar. Statistical significance was calculated as a Benjamini–Hochberg adjusted p-value with a cut-off of 0.05.