Comparative analysis of single-stranded DNA donors to generate conditional null mouse alleles

The International Mouse Phenotyping Consortium is generating null allele mice for every protein-coding gene in the genome and characterizing these mice to identify gene-phenotype associations. While CRISPR/Cas9-mediated null allele production in mice is highly efficient, generation of conditional alleles has proven to be more difficult. To test the feasibility of using CRISPR/Cas9 gene editing to generate conditional knockout mice for this large-scale resource, we employed Cas9-initiated homology-driven repair (HDR) with short and long single stranded oligodeoxynucleotides (ssODNs and lssDNAs).

Our studies demonstrate that Cas9-mediated HDR with pairs of ssODNs can generate conditional null alleles at many loci, but reveal inefficiencies when applied at scale. In contrast, lssDNAs are amenable to high-throughput production of conditional alleles when they can be employed. Regardless of the single-stranded donor utilized, it is essential to screen for sequence errors at sites of HDR and random insertion of donor sequences into the genome.

Using pairs of single guide RNAs and short ssODNs to introduce loxP sites around a critical exon or exons, we obtained putative conditional allele founder mice, harboring both loxP sites, for 23 out of 30 targeted genes. LoxP sites integrated in cis in at least one mouse for 18 of 23 genes. However, loxP sites were mutagenized in 4 of the 18 in cis lines. HDR efficiency correlated with Cas9 cutting efficiency but was minimally influenced by ssODN homology arm symmetry. By contrast, using pairs of guides and single lssDNAs to introduce loxP-flanked exons, conditional allele founders were generated for all four genes targeted, although one founder was found to harbor undesired mutations within the lssDNA sequence interval. Importantly, when employing either ssODNs or lssDNAs, random integration events were detected. Conclusions: Our studies demonstrate that Cas9-mediated HDR with pairs of ssODNs can generate conditional null alleles at many loci, but reveal inefficiencies when applied at scale. In contrast, lssDNAs are amenable to high-throughput production of conditional alleles when they can be employed. Regardless of the single-stranded donor utilized, it is essential to screen for sequence errors at sites of HDR and random insertion of donor sequences into the genome.