Easi-CRISPR: Efficient germline modification with long ssDNA donors

CRISPR/Cas9 technology efficiently produces short insertions or deletions (indels) and can insert short exogenous sequences at Cas9 cut sites. However, targeting long inserts is still a major technical challenge. To overcome this challenge, we developed Easi-CRISPR (Efficient additions with ssDNA inserts-CRISPR), a method that uses long, in vitro-synthesized, single-stranded DNAs with 50-100 base homology arms as repair templates. We demonstrate that Easi-CRISPR can generate knock-in and floxed alleles in mice with an efficiency at many loci as high as 100%. The simple design requirements for donor DNAs and the reproducibly high-efficiency of Easi-CRISPR enables rapid development of many types of commonly used animal and cell models.

The two most common types of knock-in animals needed for biomedical research and molecular genetic studies are those carrying: i) targeted insertions of commonly used sequences encoding recombinases such as Cre and Flp; transcriptional regulators such as rtTA and tTA; and reporters such as EGFP and tdTomato; and importantly: ii) Conditional Knock-Out (CKO) alleles in which two loxP sites flank one or more critical exons (i.e. "floxed" alleles). The mouse research community still depends on lengthy ES cell-based approaches to develop such models because simple CRISPR-based strategies are not available for efficient insertion of longer foreign DNAs.
To develop a simple method suitable for generating many common types of knock-in animal models, we tested a long ssDNA repair template strategy in mice. The rationale was as follows: short ssODN templates (up to 200 bases) are quite efficient in directing insertions at Cas9 cut sites and the same repair-mechanism might be exploited for delivering larger cargo if the length of the ssDNA could be extended. We demonstrated recently that ~500 base ssDNAs containing ~400 bases of synthesized microRNA sequence flanked by ~55 base homology arms directed efficient targeting of the mouse genome (10-83% efficiency when analyzed during embryonic stages) (14). In the present study, we tested the suitability of this method for i) inserting 1-1.5 kb protein coding cassettes into different genes, and ii) generating floxed alleles. The method reported here, termed Easi-CRISPR (Efficient additions with ssDNA inserts-CRISPR; pronounced Easy-CRISPR) can be readily adapted to generate even longer insertions at Cas9 cut sites.
We developed four knock-in mouse strains using Easi-CRISPR. Three models consisted of an ~1.4 kb cassette encoding P2A-FlpO recombinase inserted immediately before the stop codons of Fgf8, Slc26a5 and Mafb. The fourth model consisted of ~1 kb reverse tetracycline transactivator (rtTA)-polyA cassette inserted in-frame immediately after the initiation codon of Otoa. In each case, the cassettes contained flanking homology arms to the target loci that ranged between 72 and 105 bases. Long ssDNA donor molecules were synthesized using the IvTRT (in vitro Transcription and Reverse Transcription) method described previously (14), or obtained from a commercial source (single-stranded gBlocks® gene fragments, Integrated DNA Technologies). Schematics of the ssDNA cassettes are shown in Figure 1a The founder G0 pups were generated following standard CRISPR/Cas9 mouse genome engineering protocols (15) with some modifications described in the online methods. Details of the microinjections and the results are shown in Supplementary Table 1. Genotyping of G0 pups using 5' and 3' primers (Supplementary Table 2) indicated that the overall insertion efficiency at all four loci was 28% (7 out of 25 pups contained the respective insertion cassettes; Figure 1b, Supplementary Figures 1b, 2b and 3b). The individual insertion efficiencies, for different genes, ranged from 25% to 33% (Supplementary Table  1). Notably, one founder contained bi-allelic insertions of the knock-in cassette (Figure 1b). The fidelity of the insertions, including the junctions, was confirmed by sequencing (Figure 1c and Supplementary  Figure 1c, 2c and 3c). All targeted animals contained precisely targeted insertions, except for one Slc26a5 P2A-FlpO founder in which genomic DNA sequence duplications were present at the 3' end (Supplementary Figure 1b). Two of the founders tested to date produced heterozygous F1 offspring (Supplementary Figure 8). Furthermore, F1 offspring of the Fgf8-P2A-FlpO founder #4 showed the expected expression of FLPO in cochlear inner hair cells (Figure 1d). Taken together, these results indicate that Easi-CRISPR can insert sequences of ~1 kb or longer at high efficiency and that the technique is reproducible at multiple genomic loci.
CKO mice are among the most highly used genetically engineered models, but unfortunately there are no compelling strategies for generating such models using the CRISPR/Cas9 system. In order to test whether Easi-CRISPR can generate floxed alleles, we chose three candidate genes. To achieve this goal, we designed two guide RNAs to cut and replace the selected genomic exons with a floxed exon cassette. The details of the target exons, lengths of the ssDNA repair templates, and genotyping strategies are given in Figure 2a Table 1. We injected Cas9 protein and separated crRNA and tracr RNAs as ribonucleoprotein (RNP) complexs for each of three floxed exon cassettes (from the Col12a1, Paf1 and Ppp2r2a genes) (17). Genotyping of the 9 founder pups born (three for each gene) showed that all pups contained at least one floxed allele, a surprising 100% efficiency. Two of these pups also contained biallelic insertions of the cassettes (Figure 2b Table 1). The fidelity of the inserts and the correct fusions were confirmed by sequencing ( Supplementary Figures 11-13). We observed that one of the pups contained only one loxP site, suggesting that partial insertion of the ssDNA cassette can also occur. A similar targeting strategy was designed to flox an exon at a fourth locus but in this case Cas9 mRNA and a standard sgRNA were included in the injection mix. This resulted in some founder pups containing only one of the two loxP sites, indicating partial cassette insertions (data not shown). Notably, all of the floxed alleles that were generated at 100% efficiency were obtained RNP complex microinjection (17),suggesting that the high efficiency of full-length cassette insertion is attributable to RNP injections. However, more experiments involving side-by-side comparisons at the same set of loci, using RNP injections vs Cas9 mRNA and sgRNA injections would be necessary to reach a clear conclusion.
Currently, there are two main strategies for generating floxed alleles; insertion of two loxPs in the same allele using two separate ssODNs and two separate sgRNAs (16), or through homologous recombination of long donor dsDNA with at least 0.5 kb of homology arms (4). Both approaches are highly challenging and are not efficient: while the published reports demonstrate about 5 to 16% efficiency (4,16) many labs are still struggling to apply these strategies for other loci. In addition, some of these methods also require specially designed targeting vectors. In contrast, our Easi-CRISPR approach i) does not require specially designed targeting vectors, and ii) achieves highly efficient insertion of floxed donor cassettes (up to 100%), especially when RNP-based delivery of CRISPR components is used. Although higher efficiency of targeting when using RNPs with separated crRNA and tracrRNAs instead of a sgRNA has been reported previously (17), to our knowledge, such RNP strategies have not been used in combination with ssDNA repair templates so far. Therefore, we suggest that Easi-CRISPR with RNP injections is by far the simplest and most efficient way to create floxed alleles.
Many types of knock-in animal models are routinely used in biomedical research. For example, several thousands of knock-in mice were designed to express commonly used protein-coding sequences, the majority of which are about 1 to 1.5 kb long. Even though the CRISPR/Cas9 system has made several paradigm shifts in animal transgenesis (18), the research community still depends on lengthy ES cellbased approaches to develop knock-in and floxed animal models because there are no simple CRISPR/Cas9-based strategies available for efficient insertion of longer sequences directly in zygotes. Given the simple design requirements for the donor DNA (addition of 50 to 100 base homology arms to a cassette) and the high efficiency of Easi-CRISPR (25 to 33% efficiency for insertion of a protein coding cassette at four different genomic loci, and 7 to 100% efficiency for generating floxed alleles at three different loci), this method provides a simple and efficient means of developing many of the commonly used transgenic mouse models and should be readily adaptable to other species or cells in culture.
The Easi-CRISPR method could also offer an appealing strategy for inserting multiple point mutations situated in close proximity in the genome as, for example, to test gene regulatory elements. Lastly, whilst the mechanism by which ssDNA molecules are used as templates in DNA repair processes is not well understood, ssDNA-mediated repair mechanisms and the cellular proteins required for such repair processes (19,20) are emerging. In this context, Easi-CRISPR can serve as a tool to interrogate molecular mechanisms of DNA repair by ssDNA donors. a. The alleles that did not contain the inserts were not analyzed for presence of indels because genotyping assays were designed to identify mainly the targeted-insertion alleles (knock-in or floxed). However, noticeable sized additions or deletions were observed for some samples in these genotyping assays (e.g., additions in Slc26a5 founder #1; Supplementary figure 1b; lane 1, and deletions in the non-targeted alleles in floxing experiments; figure 1b and supplementary figures 9b and 10b) b. Pup #1 had additional sequences at the 3' junction (sequence not characterized completely) and pup #1 had a precise insertion at both junctions (see Supplementary Figure 1c) c. The pups #1 and #2 were heterozygous for both 5' and 3' loxP sites and they carried deletions in their second allele and the pup #3 was biallelic for both the 5' and 3' loxP sites (see Figure 2b). d. Pup #1 was heterozygous for both 5' and 3' loxP sites and the second allele had a deletion; pup #2 was heterozygous for 5' loxP and homozygous for 3' loxP and the pup #3 was heterozygous for both loxPs (see Supplementary Figure 9c). e. Pup # 1 and 2 are heterozygous for both loxPs with deletions in the second allele and the pup #3 is biallelic (see Supplementary Figure 10c).

CRISPR reagents
CRISPR guide RNAs were designed using CRISPR.mit.edu and were used as single guide RNAs (sgRNAs) for Otoa or as annealed 2-part synthetic crRNA and tracrRNA molecules for Fgf8, Slc26a5, Mafb, Paf1, Ppp2r2a (Alt-R TM CRISPR guide RNAs, Integrated DNA Technologies). The sgRNA was transcribed from a template generated by annealing two primers using the HiScribe™ T7 Quick High Yield RNA Synthesis Kit (New England Biolabs (NEB)) following the manufacturer's instructions. Cas9 mRNA was prepared using the pBGK plasmid as described in Harms et al., 2014. The plasmid was linearized with XbaI, gel purified and used as the template for in vitro transcription using the mMESSAGE mMACHINE T7 ULTRA kit (Ambion: AM 1345). Cas9 protein was obtained from IDT (Alt-R TM S.p. Cas9 Nuclease 3NLS). Singlestranded DNA donors were prepared using the IvTRT method described previously (14) or obtained from Integrated DNA Technologies. The floxed Col12a1 mice were generated using chemically synthesized crRNAs and tracrRNA (Genome Craft Type CT, FASMAC), and Cas9 protein (NEB).

Microinjection of one-cell embryos
All animal experiments performed were approved by the institutional IACUC protocols. C57BL/6 mice at 3-4 weeks of age (Charles River Laboratories) were superovulated by intraperitoneal injection of 5 IU pregnant mare serum gonadotropin, followed 48 hours later by injection of 5 IU human chorionic gonadotropin (both hormones from National Hormone & Peptide Program, NIDDK). Mouse zygotes were obtained by mating C57BL/6 stud males with superovulated C57BL/6 females. One-cell stage fertilized mouse embryos were injected with 20 ng/µl Cas9 protein (or 10 ng/µl of Cas9 mRNA; for the Otoa locus), 20 ng/µl of annealed crRNA and tracrRNA (or 10 ng/µl of each sgRNA; for the Otoa locus) and 5-10 ng/µl of ssDNA. Both cytoplasmic and pronuclear injections were performed. The surviving embryos were surgically implanted into pseudo-pregnant CD-1 females.

Mouse genomic DNA extraction, genotyping and sequencing
Mouse genomic DNA was extracted from toe samples using the Qiagen Gentra Puregene Tissue Kit. Primers were designed to amplify the correctly targeted junctions. Genomic DNA was subjected to flanking primer PCR and internal (donor oligo-specific) and external primer PCR. PCR reactions were performed using the Promega Hot Start green mix. The amplicons were separated on a 1% agarose gel. The gel-purified amplicons were subjected to sequencing using one of the PCR primers.  The gel on the right shows PCR bands amplified using the external primers to both loxPs, which shows that Pup #1 was heterozygous for both 5' and 3' loxP sites and the second allele had a deletion (shown by an asterisk); pup #2 was heterozygous for 5' loxP and homozygous for 3' loxP and the pup #3 was heterozygous for both loxPs.