We tested whether targeted integration of multiple copies of a reporter expression cassette can be achieved in human cells through single crossover recombination upon the introduction of double-strand breaks by ZFN or CRISPR/Cas9. As proof of concept, we assembled a pair of ZFNs targeting the human CCR5 gene based on a previous study [9, 10], and designed an sgRNA targeting the same locus on the human CCR5 gene (Fig. 1a and Additional file 1: Table S1). To examine the in vivo cleavage activities of the designed ZFNs and CRISPR/Cas9, we transfected HeLa cells with the vector pX330, which co-expresses Cas9 and sgRNA, or with vectors expressing the ZFN pair. Two days after transfection, genomic DNA was prepared for a T7 endonuclease I (T7E1) assay (Additional file 1: Table S2) and TA cloning analysis. The results of the assays show that both sgRNA and ZFN were able to induce NHEJ at their target sites with comparable efficiencies (Fig. 1b and c).
Subsequently, we designed a donor plasmid containing an EGFP reporter and a 1.6-kb fragment homologous to the CCR5 locus, with the ZFN or CRISPR/Cas9 targeting sites located in the middle (Fig. 2a). We speculated that the in vivo cleavage of a donor plasmid by ZFN or CRISPR/Cas9 could facilitate the integration of the entire plasmid into the CCR5 locus through a single crossover event at the homology region on the plasmid (Fig. 2a). Alternatively, it is also possible that when the chromosome and the donor plasmid are cut by ZFN or CRISPR/Cas9, the entire plasmid can be integrated into the target site in either forward or reverse orientation through the NHEJ repair pathway, as previously reported [11] (Fig. 2a and b). To test whether our donor plasmid was integrated into the targeted locus through single crossover recombination or NHEJ, HeLa cells stably expressing EGFP were examined through fluorescent imaging and flow cytometry analysis 12 days after transfection. We detected the targeted knock-in of the donor plasmid by ZFN or CRISPR/Cas9 in the forward orientation, but not in the reverse orientation (Fig. 2c) through PCR, using integration site- and donor-specific primers (Fig. 2a and Additional file 1: Table S3). The forward integration of EGFP was replicated in HEK293T cells (Additional file 2: Figure S1A). A previous study has shown that when the target site was cleaved by ZFN in vivo, NHEJ can capture the linearized donor plasmid in both forward and reverse orientations with almost equal frequencies [11]. The failure to detect reverse integration events may indicate the existence of a longer homology region containing nuclease target sites on the donor plasmid, resulting in HDR through single crossover recombination being favored as the main repair pathway. This in turn results in direction-dependent integration and reduces the number of direction-independent integration events mediated by NHEJ. Subsequent analysis of the sequences of the junction between target loci and knocked-in donors in targeted HeLa cells revealed indel events that typically occur after DSB repair by classical NHEJ (Fig. 2d). Interestingly, 90% (9/10 sequences) of the 5′ junction sequences of ZFN-driven knock-ins were found to have an additional spacer inserted in its sequence, and 10% (1/10 sequences) were found to have deletions. On the other hand, 100% of 3′ junction sequences had an additional spacer inserted in its sequence. Among the CRISPR/Cas9-driven knock-ins, 80% (8/10) of 5′ junction sequences were found to have single base insertions, and 20% (2/10) were found to have deletions. On the other hand, 90% (9/10) of 3′ junction sequences had single base insertions and 10% (1/10) had deletions. Similar results were obtained from the analysis of junction sequences in targeted HEK293T cells (Additional file 2: Figure S1B). To evaluate the frequencies of targeted forward integration induced by ZFN or CRISPR/Cas9, we screened for clones derived from single cells stably expressing EGFP by sorting the cells through fluorescence-activated cell sorting (FACS). Compact clonal populations of cells were observed after approximately 9 days of continuous culture (Fig. 2e). The clonal cells were then expanded for an additional 11 days and harvested for junction PCR analysis to detect targeted integration events. Out of 50 clones obtained from CRISPR/Cas9-edited cells, 5 (10%) yielded amplified DNA segments of the expected size (Fig. 2f), while 2 out of 20 clones (10%) obtained from ZFN-edited cells yielded the expected amplicon. Therefore, the knock-in of a 6.4-kb DNA fragment through single crossover recombination in our study is highly efficient (10%). We further investigated whether multiple copies of donor plasmids can be integrated into target sites cut by ZFN or CRISPR/Cas9 through single crossover recombination as is often achieved in yeast. We designed a second donor plasmid based on the first donor plasmid by replacing the EGFP coding sequence with the DsRed coding sequence and keeping the other sequences unchanged (Fig. 3a). The two donor plasmids were separately co-transfected with either ZFN or Cas9/sgRNA expression plasmids into HeLa cells. Twenty days after transfection, cells stably expressing dual fluorescent proteins (Additional file 3: Figure S2A) were collected through FACS and subjected to junction PCR analysis. EGFP-positive and DsRed-negative cells were collected as controls for analysis. In addition to the two pairs of primers used to amplify the 5′ and 3′ junctions, a pair of donor-specific primers were designed to amplify the internal junction in set-ups where multiple donor plasmids were integrated into the target sites (Fig. 3a). The expected amplicons from both 5′ junction and 3′ junction PCR were obtained from both ZFN-driven and CRISPR/Cas9-driven knock-in HeLa cells. However, the expected amplicon from the internal junction was obtained only from ZFN-edited cells (Fig. 3b). The subsequent sequence analysis (Fig. 3c) revealed that both the 5′ and 3′ junction sequences in cells with ZFN-driven knock-in of only the EGFP donor plasmid retained an intact ZFN target site, highlighting the potential for the integration of another donor plasmid. In contrast, both the 5′ and 3′ junction sequences in cells with CRISPR/Cas9-driven knock-in of only the EGFP donor plasmid had single base insertions at the cutting site. This would likely abolish the binding of the sgRNA/Cas9 complex for further cutting and, subsequently, the integration of another donor plasmid. Indels were observed at the 5′, internal, and 3′ junction sequences of ZFN-driven knock-ins of both EGFP and DsRed donor plasmids; more variable indel patterns were observed at both the 5′ and 3′ junction sequences of CRISPR/Cas9-driven knock-ins of both EGFP and DsRed donor plasmids. Similar results were obtained from targeted HEK293T cells (Additional file 3: Figure S2B and C). To evaluate the frequencies of multiply targeted forward integration induced by ZFN, we screened for clones derived from single cells stably expressing both EGFP and DsRed by sorting cells through fluorescence-activated cell sorting (FACS). Compact clonal populations of cells were observed after approximately 9 days of continuous culture (Fig. 3d). The clonal cells were then expanded for an additional 11 days of culture and harvested for PCR analysis to detect multiple targeted integration events. Out of 20 clones obtained from ZFN-edited cells, 2 (10%) yielded 5′ and 3′ junction PCR products with the expected size and 13 clones (65%) yielded the expected internal junction amplicon. Only clone number 11 (5%) yielded all the expected 5′, internal, and 3′ junction PCR products (Fig. 3e).
To further characterize the integration events by single crossover, six single cell clones of ZFN-driven knock-in of the EGFP donor plasmid were randomly selected for Southern blot analysis. Genomic DNA isolated from each clone was digested with Bam HI and then hybridized with a DIG-labeled probe binding to CMV promoter region to check the 5′ junction of around 4 kb DNA fragment, or digested with Hpa I and hybridized with a probe binding to EGFP downstream sequence to check the 3′ junction of 8.6 kb DNA fragment (Fig. 4a). When multi-copy integration occurs, the 6.4 kb plasmid fragment can be released from genomic DNA either by Bam HI or Hpa I, and detected by hybridization with according probes (Fig. 4b & c). The results showed that junction PCR seemed to underestimate the frequency of targeted integration. Out of the six clones, two (#3 and #5, 33.33%) presented expected size of 5′ junction fragment, and four (#1, #3, #5 and #6, 66.67%) presented expected size of 3′ junction fragment (Fig. 4d & e). Two clones (#3 and #5) presented expected size of both junction fragments, and multi-copy integration, which indicates targeted multi-copy integration through single crossover mediated by ZFN was able to reach to 33.33%. It should be noted that, random integration may happen in addition to targeted multi-copy integration, as blotting signals of DNA fragments with sizes out of the expected sizes of junction fragments was observed in #3 clone (Fig. 4d & e). Incomplete single copy integration at target site seems to happen in #1 clone, as only expected size of 3′ junction fragment was observed, and plasmid fragment and expected size of 5′ junction fragment was not detected (Fig. 4d & e), which implies imperfect recombination could happen at the 5′ junction. Complicated integration of donor plasmid may happen in #6 clone. After Hpa I digestion, expected size of 3′ junction fragment was observed, but expected size of 5′ junction fragment and plasmid fragment was not detected by hybridization (Fig. 4e), which implies incomplete integration of single copy of donor plasmid may happen at one allele of CCR5. After Bam HI digestion, plasmid fragment was detected but expected size of 5′ junction fragment was not detected by hybridization (Fig. 4d). Which presumably suggests incomplete integration of two copies of donor plasmid with one copy breaks between Bam HI and Hpa I restriction sites flanking the EGFP coding region may happen at another allele of CCR5 or random site. Thus Bam HI digestion was able to release the plasmid fragment (Fig. 4d), however, Hpa I was unable to release the plasmid fragment (Fig. 4e). Random integration of single copy of donor plasmid at a region lack of Bam HI and Hpa I restriction sites may happen in #2 clone, as blotting signal of expected size of both 5′ and 3′ junction fragments, together with the plasmid fragment was not detected (Fig. 4d & e). Random integration of multiple copies of donor plasmid possibly happened in #4 clone, as expected size of both 5′ and 3′ junctions was not detected, but strong blotting signal of plasmid fragment, and DNA fragments with size out of expected size of junction fragments was detected (Fig. 4d & e).