Avoid off-targets

Target site selection

Single-molecule studies and the structure of the Cas9/gRNA complex suggest that binding of the gRNA to the target sequence is required for subsequent DNA cleavage (9-11). The sequence closest to the PAM (also called the ‘seed sequence’) is hereby most important. However, most gRNAs with a 20 nt homology to the target site can tolerate several mismatches. It is therefore of paramount importance to select target sites which are unique and do not have closely homologous sequences elsewhere in the genome. There are several excellent target finders available on the internet which will suggest unique sites for most target loci (see links).

Truncated gRNAs

Keith Joung’s group has recently demonstrated that slightly truncated gRNAs (18 nt instead of the natural 20 nt homology) retain normal activity in mammalian cells, but significantly reduce the risk of off-target mutagenesis (12). This suggests that for most targets the 20 base-pairing interactions between gRNA and genomic target site provide more energy than required for Cas9 binding and activation and therefore can tolerate some mismatches. Truncated gRNAs might only provide the necessary energy when all nucleotides bind the target and hence require perfect homology. Truncated gRNAs have so far not been tested in Drosophila, but are likely to act in a similar fashion than in mammalian cells.

High fidelity Cas9 versions

Two variants of Cas9 have been used to engineer CRISPR/Cas systems with higher specificity. Point mutations in either one of the two active sites can turn Cas9 into a nickase that only cuts a single DNA strand. By using two gRNAs that bind in close vicinity but on opposite strands it is possible to create double strand breaks using the Cas9 nickase (13,14). The advantage is that off-target sites of the two individual gRNAs are unlikely to be close together and therefore only lead to DNA nicks, which have low mutagenic potential. We have recently demonstrated that an act-cas9D10A nickase can be used in combination with our pCFD4 vector for fly genome engineering (15).
The development of Fok1-dCas9 takes this concept one step further (16,17). Here, dCas9 itself is unable to cut DNA, but is fused to a Fok1 endonuclease domain. Fok1 only cuts DNA as a dimer, which can again be achieved by using two gRNAs with closely juxtaposed target sites. This system is more specific than double-nicking as a single Fok1-dCas9/gRNA complex does not induce any DNA lesions. We have generated Fok1-dCas9 expression vectors for Drosophila, which are available from Addgene, and have generated transgenic fly lines that are available through the Bloomington Stock Center. A downside of both systems is that both double-nicking and dimeric Fok1-dCas9 are reported to have lower efficiency than wildtype Cas9. Furthermore, the requirement for gRNA binding sites within a defined interval restricts possible target sites.

Redundancy, backcrossing and rescue

Potential CRISPR/Cas off-targets can generally be dealt with in the same way as RNAi off-targets or second hits induced by traditional mutagenesis methods. RNAi is chronically plagued with off-target problems and it is common practice to validate RNAi results with a second independent RNAi construct. An analogous approach is recommended for CRISPR/Cas where the ease of gRNA cloning makes it practical to produce the same or similar mutation with a second gRNA. These independent mutations are very unlikely to share the same off-targets and can be analyzed independently or in a trans-heterozygous combination. Further confidence in the specificity of observed phenotypes could be gained by genetic rescue experiments.
Mutations generated by traditional unspecific mutagenesis methods such as EMS or X-rays are accompanied by often thousands of unwanted mutations elsewhere in the genome. Therefore, desired mutations have to be isolated by multiple backcrosses into a wildtype background. Potential CRISPR/Cas off-targets are likely to be much rarer and therefore genetic cleanup operations can be smaller. It might be a good idea to remove the chromosomes not harboring the designed mutation while establishing stable fly lines.

Summary

Current evidence suggests that CRISPR/Cas genome engineering in flies does not cause widespread off-target effects. However, a genome-wide analysis of possible genome alterations following CRISPR/Cas treatment has so far not been performed. It is therefore necessary to take steps to reduce the risk of off-target cuts and to take into account the possibility that mutations at secondary sites might occur. We recommend to use unique target sites that have been designed with one of the publicly available design tools and to create at least two mutations using independent gRNAs. This will drastically reduce the chances that observed phenotypes are caused by off-target events. Truncated gRNAs and high-fidelity Cas9 variants are additional possibilities to further reduce the risk of off-tragets, but have so far not been systematically evaluated in Drosophila.