CRISPR/Cas9 activity in the rice OsBEIIb gene does not induce off-target effects in the closely related paralog OsBEIIa

Genome editing with the CRISPR/Cas9 system allows mutations to be induced at any 20-bp target site in the genome preceded by the short protospacer adjacent motif (PAM) 5′-NGG-3′. The brevity and degeneracy of the PAM ensures that the motif occurs every ~10 bp in plant genomes, and all plant genes therefore contain many targetable sites. However, the CRISPR/Cas9 system tolerates up to three mismatches in the target site, so the ability to target genes in a specific manner requires the design of synthetic guide RNAs (sgRNAs) that do not bind off-target sites anywhere else in the genome. This is straightforward for single-copy genes but more challenging if a target gene has one or more paralogs because the principles that balance targeting efficiency (the frequency of on-target mutations) and accuracy (the absence of off-target mutations) are not fully understood and may be partially species-dependent. To investigate this phenomenon in rice, we targeted the rice starch branching enzyme IIb gene (OsBEIIb) with two sgRNAs designed to differ at two and six positions, respectively, from corresponding sites in the close paralog OsBEIIa. In each case, half of the mismatches were in the essential seed region immediately upstream of the PAM, where exact pairing is thought to be necessary, and the other half were in the distal part of the target. The sgRNAs also differed in predicted targeting efficiency (39 and 96 %, respectively). We found that the sgRNA with the low predicted efficiency was actually the most efficient in practice, achieving a mutation frequency of 5 % at the target site, whereas the sgRNA with the high predicted efficiency generated no mutations at the second target site. Furthermore, neither of the sgRNAs induced an off-target mutation in the OsBEIIa gene. Our data indicate that efficiency predictions should be tested empirically because they do not always reflect the experimental outcome and that a 1-bp mismatch in the seed region of a sgRNA is sufficient to avoid off-target effects even in closely related rice genes.