Precision genome editing using synthesis-dependent repair of Cas9-induced DNA breaks

Abstract

The RNA-guided DNA endonuclease Cas9 has emerged as a powerful tool for genome engineering. Cas9 creates targeted double-stranded breaks (DSBs) in the genome. Knockin of specific mutations (precision genome editing) requires homology-directed repair (HDR) of the DSB by synthetic donor DNAs containing the desired edits, but HDR has been reported to be variably efficient. Here, we report that linear DNAs (single and double stranded) engage in a high-efficiency HDR mechanism that requires only ∼35 nucleotides of homology with the targeted locus to introduce edits ranging from 1 to 1,000 nucleotides. We demonstrate the utility of linear donors by introducing fluorescent protein tags in human cells and mouse embryos using PCR fragments. We find that repair is local, polarity sensitive, and prone to template switching, characteristics that are consistent with gene conversion by synthesis-dependent strand annealing. Our findings enable rational design of synthetic donor DNAs for efficient genome editing.

Keywords: CRISPR; HDR; PCR repair template; SDSA; short homology arms.

Fig. 1.

Tagging of the mouse Adcy3 locus with mCherry using a PCR donor with short homology arms. (A) Schematic representation of the mouse Adcy3 locus repair strategy using a PCR donor: mCherry (red), homology arms (blue), locus (gray lines), and DSB (blue line). (B) Agarose gel showing representative PCR reactions using primers flanking the DSB at the Adcy3 locus (primers correspond to sequence outside the homology arms from the PCR donor). The Upper bands (“insert” arrow) correspond to the mCherry insertion. Materials and Methods and SI Appendix, Table S1 provide experimental details and sequences.

Fig. 2.

PCR fragments with short homology arms are efficient donors to create GFP knockins in HEK293T cells. (A) Diagrams showing PCR donors for GFP insertion at the Lamin A/C and RAB11A loci. Locus, gray; GFP, green; homology arms, blue; and DSB, vertical line. GFP was inserted at the DSB in Lamin A/C and 11 bp upstream of the DSB in RAB11A. (B) Graphs showing percentage of GFP⁺ cells obtained with PCR donors with homology arms of the indicated lengths (33/33 refers to a right homology arm and a left homology arm, each 33 bp long). Insert size in all cases was 714 bp. Each bar represents the average insertion efficiency from two or more independent experiments (SI Appendix, Table S1). Error bars represent the ±SD. PCR fragments were nucleofected in HEK293T cells at the concentration indicated and counted by flow cytometer 3 d later. For this and all other figures, SI Appendix, Table S1 provides details. (C) Graphs showing percentage of GFP⁺ cells obtained with PCR or plasmid donors with homology arms of the indicated lengths. Insert size in all cases was 714 bp. Each bar represents the average insertion efficiency from two or more independent experiments (SI Appendix, Table S1). Error bars represent the ±SD. PCR fragments were nucleofected in HEK293T cells at the concentration indicated and cells were counted by flow cytometer 3 d later. (D) Confocal images of cells 3 d after nucleofection. GFP, green; DNA, blue. The GFP subcellular localizations are as expected for in-frame translational fusions.

Fig. 3.

Editing efficiency increases with decreasing insert size. Graphs showing percentage of GFP⁺ cells obtained with PCR donors with homology arms and inserts of the indicated lengths. Each bar represents the average insertion efficiency from two or more independent experiments (SI Appendix, Table S1). Error bars represent the ±SD. (A) Knockin of donors containing full-length GFP at the Lamin A/C locus. PCR fragments were nucleofected in HEK293T cells at the concentration indicated and cells were counted by microscopy 3 d later. (B) Knockin of donors containing full-length GFP or GFP11 at the Lamin A/C locus. PCR fragments were nucleofected at the concentration indicated in HEK293T (expressing GFP1–10) and cells were counted by microscopy 3 d later. (C) Knockin of donors containing full-length GFP or GFP11 at the RAB11A locus (11 bp upstream of DSB). PCR fragments were nucleofected at the concentration indicated in HEK293T (expressing GFP1–10), and cells were counted by flow cytometer 3 d later.

Fig. 4.

Repair is a polarity-sensitive process. (A) Synthesis-dependent strand annealing (SDSA) model for gene conversion (15, 16). In this, and all other schematics, each line corresponds to a DNA strand. Locus DNA is in gray, donor homology arms are in blue, donor insert is in green, and arrows indicate 3′ ends. Donor DNA strands of opposite polarity are shown above and below the locus for clarity. PCR donors contain both strands, ssODNs donors would contain either a sense or antisense strand. Dotted lines represent DNA synthesized during the repair process. Resection of DSB: DSB is resected, creating 3′ overhangs on each side of the DSB. Strand invasion and DNA synthesis: The overhangs pair with complementary strands in the donor and are extended by DNA synthesis. Annealing: The newly synthesized strands withdraw from the donor and anneal back at the locus. Ligation (not shown) seals the break. (B) Diagrams showing donor ssODNs with only one homology arm (same conventions as in A). The ssODNs contain a 126-bp insert (green) coding for 3×Flag and GFP11 and homology arm targeting either the right or left side of the DSB (SI Appendix, Table S1). (C) Normalized editing efficiency of ssODNs containing only one homology arm at the Lamin A/C and RAB11A loci. The polarity that allows pairing between the ssODN and resected ends (as shown in diagram in A) is favored. Sense and antisense ssODNs were tested in parallel experiments and their efficiency were normalized as follows: normalized efficiency of sense ssODN (light blue) = % GFP⁺ cells with sense ssODN/[% GFP⁺ cells with sense ssODN + % GFP⁺ cells with antisense ssODN]. Normalized efficiency of antisense ssODN (dark blue) = % GFP⁺ cells with antisense ssODN/[% GFP⁺ cells with sense ssODN + % GFP⁺ cells with antisense ssODN]. Numbers on top of each column indicate the nonnormalized % of GFP⁺ cells for each ssODN determined by microscopy (Lamin A/C) or flow cytometer (RAB11A).

Fig. 5.

Polarity of ssODNs affects incorporation of distal edits. (A) Schematics showing possible pairing interactions between resected locus (gray) and ssODNs (light or dark blue for sense and antisense ssODN, respectively, arrows indicate 3′ ends) coding for a distal insert (green). Sequences between the DSB and insert were recoded to help integration of the distal insert and prevent cutting of edited locus by Cas9. (B) Normalized efficiency of sense versus antisense ssODNs calculated as in Fig. 4 (SI Appendix, Table S1 provides detailed results). Distance from the DSB, locus, and guide RNA polarity are indicated Below each experiment. ssODN polarity has little effect on editing efficiency for proximal edits, but has a larger effect for distal edits. The favored polarity changes, depending on whether the distal edit is positioned to the left or right of the DSB. Note that the favored ssODN polarity does not correlate with crRNA polarity (for example, first two columns in the graph show crRNA1776 and crRNA1777, which cut at the same position but have opposite polarity). Experiments involving the PYM1 locus were done on HEK293T that were cloned out and genotyped by PCR genotyping (size shift) for 3×Flag insertion (Fig. 6). All other experiments were performed on HEK233T (GFP1–10) cells that were directly scored for GFP⁺ by flow cytometer or microscopy 3 d after nucleofection. Numbers Above each column indicate the overall percentage of edits. Note that overall frequency decreases with increasing distance from the DSB (also see SI Appendix, Fig. S6).

Fig. 6.

Recoding of sequences between the DSB and the edit increases recovery of distal edits. (A) Schematics showing resected locus (gray with arrow at the 3′ ends, PYM1 locus) and ssODN donor (blue with arrow at the 3′ end) coding for a proximal edit (green, restriction enzyme site, 1 bp to the right of the DSB) and a distal edit (red, 3×Flag, 23 bp to the left of the DSB). Double arrows represent the region between the proximal and distal edits that is recoded (silent mutations). (B) Graphs showing percentage of edited cells containing proximal + distal edits (purple), proximal only (green), or distal only (red), using a ssODN donor with or without a recoded region. More than 50 cell clones were analyzed by PCR genotyping (size shift) and RE digestion. (C) Schematics showing resected locus (gray with arrow at the 3′ ends, Lamin A/C locus) and PCR donor (blue, thick bar) coding for a proximal edit (green, GFP11 inserted at the DSB) and a distal edit (red, tagRFP, 33 bp to the right of the DSB). Double arrows represent the region between proximal and distal edits that is recoded (silent mutations). (D) Graphs showing percentage of edited cells containing proximal + distal edits (purple), proximal only (green), or distal only (red), using a PCR donor with or without a recoded region. Edits were determined by direct examination of >1,000 cells by microscopy.

Fig. 7.

Repair is prone to template switching between donors. (A) Schematics showing repair of a DSB at the RAB11A locus (gray) with two ssODN donors. Arrows indicate 3′ ends. Donor 1 contains GFP11 (green) with a stop codon (red cross) and two homology arms (blue). Donor 2 contains GFP11 with no stop codon and no homology arm. Double arrows indicate identical sequence shared between the donors. (B) Graphs showing the percent of GFP⁺ cells (y axis, as determined by flow cytometer) for each donor combination (x axis). Each bar represents the average insertion efficiency from two independent experiments (SI Appendix, Table S1). Error bars represent the ±SD. For comparison, an ssODN identical to donor 1 but without the stop codon gives 17.2% edits (discontinuous Rightmost bar). (C) Schematics showing repair of a DSB at the RAB11A locus as in diagram A but with two PCR donors (thick bars). (D) Graphs showing the percent of GFP⁺ cells as in graphs B but with two PCR donors. Each bar represents the average insertion efficiency from two independent experiments (SI Appendix, Table S1). Error bars represent the ±SD. (E) Schematics showing repair of a DSB at the Lamin A/C locus (gray) with two ssODN donors. Arrows represents 3′ ends. Donor 1 contains GFP11 (green) and two homology arms (blue). Donor 2 contains a recoded GFP11 (stars) with no homology arm. Double arrows indicate identical sequence shared between the donors. In this experiment, the edits were amplified en masse by PCR using a locus-specific primer and an insert-specific primer and sequenced by Illumina sequencing (Materials and Methods). (F) Graph showing the percentage of reads with evidence of template switching (y axis) for each donor combination (x axis). Donor 1 + donor 2 without mutations and donor 1 + donor 2 with one mutation every 3 nucleotides (1/3) show no evidence of template switching (0%), whereas donor 1 + donor 2 (1/6) and donor 1 + donor 2 (1/12) show evidence of template switching (0.5% and 1.4%, respectively). SI Appendix, Fig. S7 and Table S6 provides details.

Fig. 8.

Guidelines for donor design. (A) Schematic showing a typical editing experiment using a PCR fragment (thick line) with two homology arms (blue) to introduce an edit (green) at a distance from the DSB (stippled line). (B) Recommendations based on results presented in this study. We refer readers to refs. and for additional recommendations for ssODNs designed to insert edits at the DSB.