Simultaneous inhibition of DNA-PK and Polϴ improves integration efficiency and precision of genome editing

Abstract

Genome editing, specifically CRISPR/Cas9 technology, has revolutionized biomedical research and offers potential cures for genetic diseases. Despite rapid progress, low efficiency of targeted DNA integration and generation of unintended mutations represent major limitations for genome editing applications caused by the interplay with DNA double-strand break repair pathways. To address this, we conduct a large-scale compound library screen to identify targets for enhancing targeted genome insertions. Our study reveals DNA-dependent protein kinase (DNA-PK) as the most effective target to improve CRISPR/Cas9-mediated insertions, confirming previous findings. We extensively characterize AZD7648, a selective DNA-PK inhibitor, and find it to significantly enhance precise gene editing. We further improve integration efficiency and precision by inhibiting DNA polymerase theta (Polϴ). The combined treatment, named 2iHDR, boosts templated insertions to 80% efficiency with minimal unintended insertions and deletions. Notably, 2iHDR also reduces off-target effects of Cas9, greatly enhancing the fidelity and performance of CRISPR/Cas9 gene editing.

Fig. 1. High-throughput small molecule compounds screen identifies modulators of end joining (EJ) and homology-directed repair (HDR) DNA double-strand break (DSB) repair pathway choice.

a Outline of the small molecule screen in HEK293T cells harbouring a stable integrated traffic light reporter (TLR) construct in the AAVS1 safe harbour locus. The reporter consists of a mutated enhanced green fluorescence protein (eGFPm) followed by out-of-frame T2A and Discosoma sp. red fluorescent protein (DsRed) sequences. Reading frames are specified based on the start codon position in eGFP. Flow cytometry analysis quantifies DSB repair events. DsRed fluorescence represents EJ events, while eGFP fluorescence depicts HDR. CMV cytomegalovirus promoter. b Illustration of the screening cascade that identified DNA-PK inhibitors efficiently enhance HDR and reduce EJ repair events. pXC₅₀ value signifies the negative logarithm activity value of in vitro measured IC₅₀s, EC₅₀s, Ki, Kd, or percent inhibition endpoints. c Dot-plot illustrating the outcome of the primary screen assessing 20,548 compounds at 2 µM (n = 1). Only wells with cell counts above 1000 are illustrated. DMSO-treated cells serve as negative control, and 1 µM KU0060648 treatment as positive control. Compounds highlighted in black were selected for knock-in sequencing (KI-Seq) analysis. Dashed lines indicate set thresholds to categorise differences in EJ and HDR frequencies compared to DMSO control. Source data are provided as a Source Data file.

Fig. 2. Knock-in sequencing (KI-Seq) determines DNA repair outcomes at SpCas9-induced DNA double-strand breaks (DSBs).

a Schematic of plasmid-based genome editing strategy to study effects of small molecules on DNA repair patterns in HEK293T cells. Deep-targeted amplicon sequencing data is analysed with KI-Seq. Alterations to the reference DNA sequence are assigned to different DNA DSB repair pathways: ±1 bp insertions/deletions = non-homologous end joining (NHEJ), deletions with microhomologies (MH-del), i.e., ≥ 2 bp = alternative end joining (alt-EJ), integration of double-stranded DNA (dsDNA) donor without homology arms = NHEJ-KI, integration of a single-stranded DNA (ssDNA) donor with homology arms = HDR-KI, all remaining events are summarised in “other”. b Representative InDel profiles of ssDNA integration in HEK293T cells visualised by KI-Seq for target sites selected for their main mutation. gDel: 1 bp deletion, targeting STAT1, gIns: 1 bp insertion and gMej: 3 bp deletion associated to microhomologies, both targeting CD34. The graphical representation of the top ten variants shows ±30 bp around the SpCas9 cut-site. Each variant is annotated with absolute frequencies (Freq.) = fraction of mapped reads and relative frequencies (Rel.) in brackets = fraction of mutated reads. PAM protospacer adjacent motif. c Distribution of different repair events at the selected target sites, without DNA donor, with dsDNA or ssDNA donor, analysed with KI-Seq. Each transfection condition includes treatment with 1 µM DNA-PK inhibitor AZD7648 or corresponding DMSO control. Bar graphs represent mean values ± standard deviation (n = 3, technical replicates). d, e Mean fraction of mutated reads ± standard deviation (n = 3, technical replicates) by repair pathways. HEK293T cells were treated with different concentrations (0.3–10 µM) of DNA repair inhibitors from the traffic light reporter (TLR) screen or DMSO control and transfected with SpCas9, gIns and (d) ssDNA or (e) dsDNA donor. The horizontal line denotes the mean knock-in efficiency in DMSO-treated cells. Asterisk indicates treatments affecting cell confluency. Compounds and their associated targets are shown. Source data are provided as a Source Data file.

Fig. 3. Characterisation of cell-line-specific SpCas9-mediated DNA double-strand break (DSB) repair with KI-Seq. AZD7648 promotes efficient integration of short single-stranded DNA donors in dividing cells.

a Workflow of knock-in sequencing (KI-Seq) using ribonucleoprotein (RNP)-based genome editing to study DSB repair and different knock-in strategies in various cell lines in the presence of 1 µM DNA-PK inhibitor AZD7648 or DMSO control. b–g Deep targeted amplicon sequencing data is analysed with KI-Seq. Alterations to the reference DNA sequence are assigned to different DNA DSB repair pathways: ±1 bp insertions/deletions = non-homologous end joining (NHEJ), deletions with microhomologies (MH-del), i.e., ≥ 2 bp = alternative end joining (alt-EJ), integration of double-stranded DNA (dsDNA) donor without homology arms = NHEJ-KI, integration of a single-stranded DNA (ssDNA) donor with homology arms =  HDR-KI, all remaining events are summarised in “other”. Left: fraction of mutated reads of different repair events at depicted target sites without donor or compound treatment. Right: knock-in efficiencies of dsDNA or ssDNA in the presence of 1 µM AZD7648 or DMSO for three immortalised (b) HEK293T, (c) Jurkat, (d) HepG2, and three primary cell lines (e) SpCas9-inducible human induced pluripotent stem cells (hiPSC), (f) primary human CD4 + T cells and (g) primary human hepatocytes (PHH). Bar graphs represent mean values ± standard deviation (n = 3, biological replicates) calculated as percent of mapped reads. Significance level of ssDNA-mediated knock-ins was evaluated using Student’s paired t test (two-tailed) *P  < 0.05, **P < 0.01, ***P < 0.001. Calculated P values: DMSO vs AZD7648 (HEK293T gDel = 0.0152, gIns = 0.0001, gMej = 0.0183; Jurkat gDel = 0.0852, gIns = 0.0178, gMej = 0.0141; HepG2 gDel =  0.0203, gIns = 0.0084, gMej = 0.087; hiPSC gDel = 0.0316, gIns = 0.0077, gMej = 0.0365; primary human CD4 + T cells gDel = 0.0417, gIns = 0.0121, gMej = 0.0294; PHH gDel = 0.2543, gIns =  0.1686, gMej = 0.6828). Source data are provided as a Source Data file.

Fig. 4. Simultaneous Polϴ and DNA-PK inhibition increases the frequency and precision of single-stranded DNA-mediated integration.

a–c Fraction of mutated reads of different repair events (±1 bp insertions/deletions = non-homologous end joining (NHEJ), deletions with microhomologies (MH-del), i.e., ≥2 bp = alternative end joining (alt-EJ), integration of double-stranded (dsDNA) donor without homology arms = NHEJ-KI, integration of a single-stranded (ssDNA) donor with homology arms = HDR-KI, all remaining events are summarised in “other”) at the gMej target site with indicated donor DNA types for HEK293T wild-type (wt) cells and three knock-out pools of enzymes involved in alt-EJ repair. TIDE analysis estimated knockout efficiencies of cell pools were PARP1 = 94%, POLQ = 92% and LIG3 = 88%. Experiments were performed with 1 µM AZD7648 or DMSO control treatment 3 h before plasmid transfections. Horizontal lines illustrate the mean knock-in efficiency in DMSO or AZD7648-treated wt cells. Bar graphs show mean values ± standard deviation (n = 3, technical replicates). d Knock-in efficiencies of ssDNA donor integration at selected target sites with 1 µM AZD7648 or DMSO for HEK293T wt and POLQ-KO cell pools. Bar graphs represent mean values ± standard deviation (n = 3, technical replicates). Statistical differences were evaluated using Student’s paired t test (two-tailed) *P < 0.05, **P < 0.01, ***P < 0.001. Calculated P values: wt DMSO vs wt AZD7648 (gDel = 0.0135, gIns = 0.0003, gMej = 0.0143); wt AZD7648 vs POLQ KO AZD7648 (gDel = 0.0037 gIns = 0.0037 gMej = 0.0058). e Frequencies of different repair events at the gMej target site in plasmid and ssDNA transfected HEK293T cells. Cells were treated 1–3 h before transfections with DMSO,1 µM AZD7648, and 1 µM AZD7648 in combination with 3 µM Polϴ inhibitor PolQi1 or PolQi2. Bar graphs represent mean values ± standard deviation (n = 3, biological replicates). f Percentage of different repair events in SpCas9-inducible hiPSC transfected with sgRNA gMej in the presence of 1 µM DNA-PK inhibitor AZD7648 or 1 µM AZD7648 in combination with 3 µM PolQi1 or PolQi2. Bar graphs depict the mean percentage of mutated reads ±standard deviation (n = 5, biological replicates). Source data are provided as a Source Data file.

Fig. 5. 2iHDR treatment reduces off-target mutations and translocations.

a Heatmaps show mean fraction of mutated reads (InDel frequencies) for HEK3 and HEK4 on- and off-target sites (n = 3, technical replicates). Cells were treated with DMSO, 1 µM AZD7648 and 1 µM AZD7648 with 3 µM PolQi1 or PolQi2. Ctrl refers to a non-targeting control guide. Divergent bases are indicated on the right (small red letters). b Quantification of large deletions from long-read sequencing data of HBEGF-edited HEK293T cells treated with indicated compounds in the following concentrations 1 µM AZD7648, 3 µM PolQi1, 3 µM PolQi2 and specified DNA donors. Bar graphs present the percentage of deletion index (n = 1). ssDNA: single-stranded DNA donor with homology arms; HDR donor: plasmid with long homology arms. c Simultaneous genome editing in HEK293T cells at HBEGF and PCSK9 loci for translocation analysis in the presence of inhibitors (1 µM AZD7648, 3 µM PolQi1 and 3 µM PolQi2) or DMSO control. Ctrl refers to a non-targeting control guide. Dot-plots depict mean translocation frequencies analysed with ddPCR ±standard deviation (n = 3, technical replicates). Statistical differences were evaluated using one-way ANOVA (with Dunnett ´s post-hoc correction), *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.001. Calculated P values: AZD7648 = 0.9424, PolQi1 = 0.9424, PolQi2 = 0.8571, AZD7648 + PolQi1 = <0.0001, AZD7648 + PolQi2 = <0.0001). Source data are provided as a Source Data file.

Fig. 6. 2iHDR boosts integration of different DNA donor templates with various integration-to-cut-site distances in diverse cell lines.

a–c Quantification of luminescence signal via HiBiT lytic assay to estimate protein tagging efficiencies with HiBiT and Halo-HiBiT tags at indicated genomic loci and cell lines. Experiments were performed with 1 µM AZD7648 and 1 µM AZD7648 in combination with 3 µM PolQi1 or PolQi2. Data represent mean luminescence signal after background subtraction ±standard deviation (Hela HiBiT, n = 4; Jurkat HiBiT, n = 3; Jurkat HaloTag-HiBiT, n = 3; all technical replicates). RLU relative light unit d Bar graphs represent the percentage of Jurkat cell clones containing HaloTag-HiBiT knock-in assessed with the Nano-Glo HiBiT lytic detection system for the indicated target genes. Cells were treated with DMSO or 1 µM AZD7648 and 3 µM PolQi2. Numbers show knock-in (KI) positive clones per total analysed clones. e ddPCR analysis of HaloTag-HiBiT KI-positive Jurkat cell clones. Bar graphs show the percentage of homozygous HaloTag-HiBiT integration at different loci using DMSO or 1 µM AZD7648 and 3 µM PolQi2. Numbers represent homozygous per heterozygous clones. f Flow cytometry analysis to measure GFP integration into the TRAC locus in primary human CD3 + T cells derived from three individual human donors in the presence of 1 µM AZD7648, and 1 µM AZD7648 in combination with 3 µM PolQi1 or PolQi2. Knock-in efficiencies were evaluated 7 days post-electroporation. Data represent the percentage of GFP-positive cells in the viable cell population. Source data are provided as a Source Data file.