Harnessing DSB repair to promote efficient homology-dependent and -independent prime editing

Abstract

Prime editing recently emerged as a next-generation approach for precise genome editing. Here we exploit DNA double-strand break (DSB) repair to develop two strategies that install precise genomic insertions using an SpCas9 nuclease-based prime editor (PEn). We first demonstrate that PEn coupled to a regular prime editing guide RNA (pegRNA) efficiently promotes short genomic insertions through a homology-dependent DSB repair mechanism. While PEn editing leads to increased levels of by-products, it can rescue pegRNAs that perform poorly with a nickase-based prime editor. We also present a small molecule approach that yields increased product purity of PEn editing. Next, we develop a homology-independent PEn editing strategy, which installs genomic insertions at DSBs through the non-homologous end joining pathway (NHEJ). Lastly, we show that PEn-mediated insertions at DSBs prevent Cas9-induced large chromosomal deletions and provide evidence that continuous Cas9-mediated cutting is one of the mechanisms by which Cas9-induced large deletions arise. Altogether, this work expands the current prime editing toolbox by leveraging distinct DNA repair mechanisms including NHEJ, which represents the primary pathway of DSB repair in mammalian cells.

Fig. 1. SpCas9 nuclease-based prime editing.

a NGS analysis of PEn or PE2-mediated targeted DNA insertions of indicated sizes using 10 different pegRNAs targeting endogenous loci in HEK293T cells. Plots show mean ± SD of n = 3 biologically independent replicates. “prime edits – all” and “prime edits – precise” categories are superimposed. P-values were determined using Student’s paired t test (two-tailed) *P <  0.05, **P <  0.01, ***P <  0.001. Calculated P values: HEK3 = 0.0050, DPM2 = 0.3075, AAVS1 = 0.2569, EGFR = 0.0732, TRAC = 0.0433, PCSK9 = 0.0028, FANCF = 0.0733, TRBC = 0.0021, PDCD1 = 0.0055, CTLA4 = 0.0003. b Representative alignment and allele frequencies of AAVS1 locus edited with PEn and the indicated RT template in HEK293T cells. For each category, the top 10 variants are shown with a minimum frequency of 0.1%. c Model of homology-dependent and NHEJ modes of PEn-mediated insertions at DSBs. Source data for Fig. 1a are provided as a Source Data file.

Fig. 2. NHEJ mediates imprecise PEn editing.

a Selected data from Fig. 1a with additional DNA-PK inhibitor treatments of PEn samples. Plots represent PEn or PE2 editing using 8 different pegRNAs targeting endogenous loci in HEK293T. PEn edited cells were additionally pre-treated with DNA-PK inhibitor (PEn+i) or DMSO (PEn). Plots show mean ± SD of n = 3 biologically independent replicates. “prime edits – all” and “prime edits – precise” categories are superimposed. Histograms below the bar plots represent percentages and size distribution of PEn prime edited alleles for each target site with or without DNA-PK inhibition. P-values were determined using Student’s paired t test (two-tailed) *P <  0.05, **P <  0.01, ***P <  0.001. Calculated P values: PEn vs. PEn+i (DPM2 = 0.00001, AAVS1 = 0.0273, TRAC = 0.0536, PCSK9 = 0.0168, TRBC = 0.1189, CTLA4 = 0.1029, PDCD1 = 0.0091, EGFR = 0.0282), PEn+i vs. PE2 (DPM2 = 0.3229, AAVS1 = 0.0222, TRAC = 0.0467, PCSK9 = 0.0149, TRBC = 0.0121, CTLA4 = 0.0020, PDCD1 = 0.0202, EGFR = 0.0342). b Representative alignment and allele frequencies of AAVS1 locus edited with PEn and the indicated RT template in HEK293T cells treated with DNA-PK inhibitor. For each category, the top 10 variants are shown with a minimum frequency of 0.1%. c NGS analysis of PEn editing outcomes of AAVS1 locus in wild-type and POLQ-/- and HEK293T cells with or without DNA-PK inhibitor treatment. The plot shows mean ± SD of n = 3 biologically independent replicates. Source data for Fig. 2a, c are provided as a Source Data file.

Fig. 3. PEn editing through NHEJ.

a Model of PRimed INSertions (PRINS) – an NHEJ-mediated mode of PEn editing using springRNA. b NGS analysis of PRINS-mediated editing at AAVS1 in HEK293T cells with or without DNA-PK inhibitor. c Representative alignment and allele frequencies of AAVS1 locus edited with PRINS and the indicated RT template in HEK293T cells. For each category, the top 10 variants are shown with a minimum frequency of 0.1%. d NGS analysis of PRINS-mediated editing at indicated loci using a panel of springRNAs in HeLa cells. e NGS analysis of PRINS-mediated editing at indicated loci using a panel of springRNAs in HCT116 cells. f NGS analysis of PRINS-mediated 1 nt insertions at AAVS1 using springRNAs with four different RT-templates in HEK293T cells. All plots show mean ± SD of n = 3 biologically independent replicates. “prime edits – all” and “prime edits – precise” categories are superimposed. Source data for Fig. 3b, d–f are provided as a Source Data file.

Fig. 4. Off-target analysis of PEn editing.

NGS analysis of editing outcomes at three on-target and eleven off-target sites with indicated editors and peg/springRNAs. Editing levels are shown as percentages of modified reads in each sample. The values represent the average of n = 3 biologically independent replicates. Mismatches to the on-target gRNA sequence are highlighted in red, the PBS region is highlighted in blue. Source data for Fig. 5 are provided as a Source Data file.

Fig. 5. Large on-target deletion induction by Cas9, PE2, or PEn editing.

a Diphtheria toxin (DT) selection-driven assay to detect on-target large deletions induced by different genome editing systems. b Relative rates of surviving colonies after DT selection of cells edited by indicated genome editors. Data normalized to Cas9. The plot shows the mean of n = 2 biologically independent replicates. c Alignment of long HBEGF reads from samples targeted with Cas9, PE2, or PEn and harvested before DT selection. Red lines denote Cas9 cleavage site. Scalebar 5000 bp. Panels on the right show window of 100 bp around the cleavage site. d Relative comparison of the rates of surviving colonies after DT selection of cells edited with PEn and either springRNA with a random insert of PAM-reconstituting springRNA. Data normalized to PEn + springRNA. The plot shows the mean of n = 2 biologically independent replicates. e Alignment of long HBEGF reads from samples targeted with PEn and either springRNA with a random non-PAM insert of PAM-reconstituting springRNA (PAMins) harvested before DT selection. The y-axis is set from minimal to maximal read depth for each sample. Source data for Fig. 5b, d are provided as a Source Data file.