CRISPR-Cas is an efficient method for genome editing in organisms from bacteria to human cells. We describe a transgene-free method for CRISPR-Cas-mediated cleavage in nematodes, enabling RNA-homology-targeted deletions that cause loss of gene function; analysis of whole-genome sequencing indicates that the nuclease activity is highly specific.
THE ability to target genes for the creation of heritable, stable loss-of-function mutations is a powerful tool in studies of animal biology. In the nematode Caenorhabditis elegans, which is amenable to the handling of large numbers of animals in liquid cultures or on solid media and the creation of frozen libraries, this has long been possible by PCR screening of large randomly generated libraries containing mobilized endogenous transposons (Rushforth et al. 1993), chemically induced small deletions (Jansen et al. 1997; Gengyo-Ando and Mitani 2000; Edgley et al. 2002), or heterologous transposons (Bessereau et al. 2001). Another random approach is the construction of libraries of sequenced mutagenized strains (C. elegans Deletion Mutant Consortium 2012; Thompson et al. 2013). More recently developed technologies have offered to replace such random and large-scale approaches with engineered nucleases designed to induce deletions and provide double-strand DNA breaks at specific desired points in the genome. This was first done using zinc-finger nucleases (Urnov et al. 2010) and transcription activator-like effector nucleases (TALENs) (Miller et al. 2011), which have proved effective in Caenorhabditis nematodes (Wood et al. 2011). More recently, technologies have been developed that use clustered regularly interspaced short palindromic repeats (CRISPR) (Jinek et al. 2012; Cong et al. 2013; Mali et al. 2013).
Methods based on CRISPR-directed nuclease activity offer the potential for maximally convenient engineering of target specificity to effect changes to the genome. In this method, a Cas9 nuclease forms a complex with two RNAs, called the CRISPR RNA (crRNA) and the trans-activating crRNA (tracrRNA), or with a synthetic fusion guide RNA (sgRNA) containing elements of both RNAs; these RNAs provide targeting information for site-specific cleavage by the nuclease (Jinek et al. 2012, 2013). This site-specific nuclease activity can be targeted by simply incorporating into the sgRNA a 20-nucleotide sequence corresponding to a target site in the genome containing this identical 20-nucleotide sequence fused at its 3′ end to a proto-spacer adjacent motif (PAM), which has the sequence 5′-NGG-3′. The C. elegans coding sequence is ∼50% GC (C. elegans Sequencing Consortium 1998), and thus almost one-eighth of exonic nucleotides will be candidates for CRISPR-Cas-mediated cleavage, counting both strands. The double-strand break created by CRISPR-Cas activity will then be repaired by nonhomologous end joining (NHEJ) or by homology-directed recombinational repair (Lemmens and Tijsterman 2011). NHEJ is prone to the creation of deletions and small insertions that disrupt gene function; homology-directed recombinational repair can be exploited by the inclusion of a repair template that will be incorporated into the repaired double-strand break to effect precisely designed changes and can be used to insert large constructs such as fluorophore reporter fusions and rescuing transgene markers (Robert and Bessereau 2007; Lo et al. 2013).
To implement CRISPR-Cas-mediated cleavage in nematodes, we adapted existing protocols for using CRISPR-Cas-mediated cleavage in cultured mammalian cells (Mali et al. 2013) with reference to methods used for the expression of engineered zinc finger and TALEN nucleases in Caenorhabditis (Wood et al. 2011). In this method, we injected into the germline syncytia of C. elegans hermaphrodites in vitro-synthesized capped and polyadenylated message RNA for humanized Cas9 nuclease, with 5′- and 3′-UTRs optimized for C. elegans germline expression (Figure 1, A and B, and Supporting Information, File S1). With this mRNA we included an in vitro-synthesized hybrid sgRNA containing 20 nucleotides of identity with a target site. To test this system, we selected as targets genes with recognizable, viable null phenotypes. F2 progeny of injected animals were inspected for the presence of these phenotypes; by this method, 10 independent alleles of dpy-11 and 1 allele of unc-4 were isolated (Figure 1, C and D, and Table 1). Efficiency was sometimes very high but showed considerable variability: dpy-11 alleles were recovered at the rate of one independent mutant for every five animals injected, while only a single unc-4 mutation was recovered from injection of 15 P0’s. The differing efficiencies observed with mutagenesis using CRISPR-Cas-mediated cleavage to target different sites may reflect differences between the 20-nucleotide targeting sequences inserted into the sgRNAs, perhaps indicating preferences of nucleotide composition or effects of secondary structure; the differing efficiencies might alternatively result from effects of genomic context.
The CRISPR-Cas-induced mutations we recovered were characterized by a prevalence of large deletions and possible chromosomal rearrangements. The 4 smallest of our 11 deletions removed 7, 23, 565, and 1031 nucleotides. Five more alleles were deletions of >2 kbp; the remaining mutations, sy745 and sy750, may be more complex genomic rearrangements (Table 1 and File S1). These results contrast with another recent report of phenotype-based recovery of CRISPR-Cas-induced mutations in C. elegans that showed a strong preference for extremely small insertions or deletions, ranging from a single nucleotide to <20 (Friedland et al. 2013). Even at n = 11, and even comparing only to the four mutations reported by Friedland et al. as having been isolated through phenotypic screening, the difference in mutant profile is striking. The most obvious procedural difference is that we induced mutations by delivering in vitro-synthesized RNAs, an inherently transient mechanism; the other group’s smaller changes were induced by transfection of expression constructs that endured for at least a full generation. It seems at least possible that the effect of the transient CRISPR-Cas nuclease activity provided by synthetic RNA was largely confined to a brief interval in germline development that was unusually susceptible to double-strand breaks being repaired by a mechanism that results in the generation of large deletions, while the CRISPR-Cas nuclease activity provided by transgene expression may been effective at other points in germline development at which double-strand breaks are frequently repaired by NHEJ to leave only very small lesions. It remains to be determined whether the generation of large deletions by transient application of CRISPR-Cas nuclease activity to the germline syncytium is mediated by NHEJ repair.
To determine whether mutations were simultaneously being induced at other sites in the genome, we sequenced genomic DNA of sy740 and sy745, two independent but closely related dpy-11 mutants generated using CRISPR-Cas-mediated cleavage, recovering 67,212,770 and 74,587,841 unpaired 50-bp reads, respectively (33× coverage and 37× coverage, respectively). Any mutation present in only one of the two strains compared to the published sequence of C. elegans could be a de novo mutation induced by off-site cleavage by the sgRNA-guided Cas9 nuclease. We performed two analyses to identify candidate mutations: the GATK pipeline (DePristo et al. 2011) identified 1592 mismatches, small insertions, or deletions predicted to be present in either or both strains compared to the reference C. elegans genome, and split-read analysis identified 48 candidate large deletions or insertions (Table 2). Of the 1592 candidate changes identified by the GATK pipeline, 324 were predicted in one strain but not in the other (Table S1). After manual inspection of the aligned reads at the sites of these 324 candidate strain-specific changes, 313 were found to be present in both strains, 3 were false positives, and 8 were strain-specific changes (Table S1). None of the 8 changes unique to a strain was homozygous, and none was near a sequence with homology to the sequence used to target CRISPR-Cas-mediated nuclease activity. Five of the 8 altered the extent of a mononucleotide or trinucleotide repeat; 2 of these had extremely low coverage in the sample that did not contain the mutation (Table S1). Of 48 candidate deletions detected by split-read analysis, 24 were homozygous in both strains; for the other 24, manual inspection of aligned reads could demonstrate only that wild-type sequence was present at the site in both strains, indicating the deletion was not homozygous (Table S2). If a large deletion identified by split-read analysis was present but was not homozygous, it would not be detected in manual curation of the aligned sequencing reads. We tested 7 of these 24 using PCR, and saw no evidence that CRISPR-Cas-mediated mutagenesis had caused off target genomic lesions (see File S1, Table S2, and Table S4).
Thus, neither of two analysis methods detected CRISPR-Cas-induced mutations new to either strain, suggesting that to the limit of their ability to detect mutations, CRISPR-Cas-mediated nuclease activity is highly specific in its effects—a result that stands intriguingly in contrast to a recent report of CRISPR-Cas causing frequent off-target changes in cultured human cells (Fu et al. 2013). We tested the detection power of both analyses by assessing their ability to detect deletions they should possess when the analyses were performed using a version of the C. elegans reference genome into which a large number of short sequences had been inserted at known sites. This comparison enabled us to assess the false-negative rate of each analysis (Table S3), suggesting that we should have detected at least two-thirds of deletions, and at least three-fourths of deletions in nonrepetitive sections of the genome, such as coding DNA. We suggest that the failure of these analyses to identify de novo CRISPR-Cas-induced mutations in comparisons with the unmodified reference genome reflects an actual absence of such mutations. A healthy sense of caution nonetheless suggests that mutant strains generated using CRISPR-Cas should be outcrossed to remove potential off-target mutations that might result from CRISPR-Cas nuclease activity.
This transgene-free, rapidly reconfigurable system for targeted nuclease activity in C. elegans might be applicable more broadly in nematodes, particularly in nematode species less readily transformed with expression constructs than is C. elegans. A similar system for the expression of ZFNs can be used not only in C. elegans but also in C. briggsae, a species more evolutionarily distant from C. elegans than mouse is from human (Stein et al. 2003; Wood et al. 2011). The double-strand breaks generated by CRISPR-Cas-mediated nuclease activity in this report were repaired in a mutagenic fashion, leaving deletions. Such double-strand breaks could instead be repaired by homologous recombination with a targeting construct supplied by microinjection along with the nuclease mRNA and guide RNA; this has been done for the double-strand breaks produced by transposon excision (Robert and Bessereau 2007) and has recently been done for double-strand breaks generated using TALENs and CRISPR-Cas (Lo et al. 2013). Such a system might provide a rapid expedient for the generation of transgenes in nematode species that currently lack such resources. The availability of relatively species-independent tools for the rapid generation of lesions in specific genes is likely to facilitate the study of gene function in species previously found to be inconvenient for such approaches, including species in which RNA interference (RNAi) is less effective than in C. elegans (Moshayov et al. 2013). An ability rapidly to inactivate selected genes or to make precisely engineered genomic changes in a broad range of species should make it possible to explore the functional significance of identified evolutionary changes. Because this technique should make genome modification and transgenesis readily feasible in a broad range of species, it should be possible to select a species uniquely suited to a biological question of interest, rather than being constrained to model this biological question in a species that is well studied and has excellent molecular resources, but whose biology is less well suited to the question.
Our work is one of several nearly simultaneous reports, each of which uses somewhat different approaches to achieve genome modification using CRISPR-Cas (Cho et al. 2013; Dickinson et al. 2013; Friedland et al. 2013; Katic and Grosshans 2013; Lo et al. 2013; Tzur et al. 2013; Waaijers et al. 2013); a comparison and synthesis of these works should yield CRISPR-Cas genome modification methods suitable for most applications.
We thank Mihoko Kato and David Greenstein for helpful comments on the manuscript, participants in the Engineered Nucleases workshop at the 2013 International C. elegans meeting for collegial discussions, and WormBase. High-throughput DNA sequencing was performed at the Millard and Muriel Jacobs Genetics and Genomics Laboratory at California Institute of Technology. P.W.S. is an investigator of the Howard Hughes Medical Institute, which supported this work.
Note added in proof: The vectors generated for use in this manuscript have been submitted to Addgene (http://addgene.org).
Communicating editor: O. Hobert
- Received July 29, 2013.
- Accepted August 13, 2013.
- Copyright © 2013 by the Genetics Society of America