Systematic quantification of HDR and NHEJ reveals effects of locus, nuclease, and cell type on genome-editing

Abstract

Precise genome-editing relies on the repair of sequence-specific nuclease-induced DNA nicking or double-strand breaks (DSBs) by homology-directed repair (HDR). However, nonhomologous end-joining (NHEJ), an error-prone repair, acts concurrently, reducing the rate of high-fidelity edits. The identification of genome-editing conditions that favor HDR over NHEJ has been hindered by the lack of a simple method to measure HDR and NHEJ directly and simultaneously at endogenous loci. To overcome this challenge, we developed a novel, rapid, digital PCR-based assay that can simultaneously detect one HDR or NHEJ event out of 1,000 copies of the genome. Using this assay, we systematically monitored genome-editing outcomes of CRISPR-associated protein 9 (Cas9), Cas9 nickases, catalytically dead Cas9 fused to FokI, and transcription activator-like effector nuclease at three disease-associated endogenous gene loci in HEK293T cells, HeLa cells, and human induced pluripotent stem cells. Although it is widely thought that NHEJ generally occurs more often than HDR, we found that more HDR than NHEJ was induced under multiple conditions. Surprisingly, the HDR/NHEJ ratios were highly dependent on gene locus, nuclease platform, and cell type. The new assay system, and our findings based on it, will enable mechanistic studies of genome-editing and help improve genome-editing technology.

Figure 1. Design and validation of the assay to simultaneously detect HDR and NHEJ at an endogenous gene locus.

(a) The WT allele will be FAM+ and HEX+ because the reference (FAM) and NHEJ (HEX) probes bind to it. With HDR, the HDR probe can bind to the HDR allele, making it FAM++ (higher-amplitude fluorescence than FAM+). With NHEJ, the NHEJ probe loses its binding site, so the NHEJ allele will be FAM+ and HEX−. (b) Validation of the assay with synthetic DNA spiked into WT genomic DNA. Analysis of unspiked WT genomic DNA without genome-editing showed only the FAM+ and HEX+ WT (green) allele on the left 2D droplet scatter plot. The assay robustly detected a spike-in of 5% of synthetic HDR (orange) and NHEJ (blue) alleles added to the genomic DNA. (c) Assay sensitivity established by 2-fold serial dilution of HDR and NHEJ synthetic template in a constant (100 ng) background of WT genomic DNA. The limit of detection (LoD) was ~0.1% for NHEJ and <0.05% for HDR, as established by comparison with WT genomic DNA-only wells (nonoverlap of 95% confidence intervals, dotted line). Data represent two merged wells per dilution point and four merged wells for WT genomic DNA-only negative control. The 95% confidence interval is shown. (d) Simultaneous detection of HDR (0.6%) and NHEJ (5.3%) induced by Cas9 in HEK293T cells.

Figure 2. Design of point mutagenesis and assay systems for RBM20.

(a) Design of point mutagenesis for RBM20. The locations of gRNAs (F1, F2, R1, or R2), TALENs, and donor DNAs are shown. The mutation sites are highlighted in green. The locations and sequences of sense and antisense strand oligonucleotide donors (60 nt) are also shown. (b–e) Design of assay systems for RBM20. The locations of HDR and NHEJ probes are shown for the Cas9 with gRNA-F1/R2 (b), Cas9 with gRNA-F2/R1 (c), FokI-dCas9 (d), and TALEN (e) assays. Note that two NHEJ probes were included in dual Cas9 assays (see Supplementary Fig. S1 and Supplementary Table S6). For simplicity, primers and reference probes are not included here (see Supplementary Table S5 for their sequences). Red triangles indicate the predicted cut sites by nucleases. HDR probes specifically bind to alleles induced by HDR, whereas NHEJ probes lose their binding sites when insertions or deletions are created by NHEJ.

Figure 3. Measurement of HDR and NHEJ induced by sequence-specific nucleases in HEK293T cells.

(a) Scatter plots of droplets showing HDR- and NHEJ-inducing activities of nucleases at RBM20. Representative 2D scatter plots of droplets positive for HDR (orange), NHEJ (blue), WT (green) alleles were obtained from HEK293T cells transfected with indicated nucleases targeting RBM20. Unedited WT genomic DNA was used as a negative control. Conditions that gave equivalent or more HDR than NHEJ are highlighted in bold italic. (b–d) Quantified genome-editing outcomes at RBM20 (b), GRN (c), and ATP7B (d) in HEK293T cells. HDR (red) and NHEJ (blue) allelic frequency (%) is shown. Conditions that gave equivalent or more HDR than NHEJ are highlighted in bold italic. Values are mean ± SEM. (n = 6). The difference between the HDR and NHEJ frequencies were evaluated by Student’s T-test (*p < 0.05). S, sense strand oligonucleotide donor; AS, antisense strand oligonucleotide donor. The background signals of assays from equivalent amounts of unedited WT genomic DNA were subtracted from edited genomic DNAs (Supplementary Fig. S6).

Figure 4. Measurement of HDR and NHEJ induced by sequence-specific nucleases at RBM20 in HeLa cells and iPSCs.

(a) Scatter plots of droplets showing HDR- and NHEJ-inducing activities of dual Cas9-H840A and TALEN at RBM20. Representative 2D scatter plots of droplets positive for NHEJ (blue) and WT (green) alleles were obtained from HeLa cells transfected with dual Cas9-H840A or TALENs targeting RBM20. Both systems failed to induce HDR. (b) Quantified genome-editing outcomes at RBM20 in HeLa cells. HDR (red) and NHEJ (blue) allelic frequency (%) is shown. (c) Scatter plots of droplets showing HDR- and NHEJ-inducing activities of Cas9 and TALEN at RBM20. Representative 2D scatter plots of droplets positive for HDR (orange), NHEJ (blue), WT (green) alleles were obtained from human iPSCs transfected with Cas9 or TALENs targeting RBM20. Only TALENs induced more HDR than NHEJ (highlighted in bold italic). (d) Quantified genome-editing outcomes at RBM20 in iPSCs. HDR (red) and NHEJ (blue) allelic frequency (%) is shown. Conditions that gave equivalent or more HDR than NHEJ are highlighted in bold italic. For (b,d), values are mean ± SEM. (n = 6). The difference between the HDR and NHEJ frequencies were evaluated by Student’s T-test (*p < 0.05). S, sense strand oligonucleotide donor; AS, antisense strand oligonucleotide donor. The background signals of assays from equivalent amounts of unedited WT genomic DNA were subtracted from edited genomic DNAs (Supplementary Fig. S6). Conditions with low activity are shown in a different scale (highlighted in green).

Figure 5. Little correlation between HDR and NHEJ frequencies induced by genome-editing.

Scatter plots of HDR- and NHEJ-inducing activities of genome-editing conditions. (a–i) The frequency of HDR and NHEJ induced by all tested conditions (a), by Cas9-D10A (b) (blue), by Cas9-H840A (c) (red), by FokI-dCas9 (d) (purple), by Cas9 (e) (green), by TALEN (f) (black), at GRN (g), at ATP7B (h), and at RBM20 (i) are plotted. The raw data are shown in Figs 3 and 4 and Supplementary Figs S7,S9, and S10. The shapes of the data points represent GRN in HEK293T cells (circle) or HeLa cells (pentagon), ATP7B in HEK293T cells (cross) or HeLa cells (bar), and RBM20 in HEK293T cells (square), HeLa cells (star) or human iPSCs (triangle). The R² values for the plots are shown. There is little correlation between HDR and NHEJ in general. Note that data points with the 0 value are not shown in the plots but were included for calculation of the R² values.

Figure 6. Schematic of nuclease-induced genome modifications and heat map of HDR-inducing activity.

The HDR- and NHEJ-inducing activities of tested conditions are summarized as a heat map. The raw data are shown in Figs 3 and 4 and Supplementary Figs S7, S9 and S10. The conditions from the best (red) to the worst (black): >0.1% HDR and HDR>NHEJ, <0.1% HDR and HDR>NHEJ, 2× HDR>NHEJ>1× HDR, >0.1% HDR and 2× HDR