Efficient genome editing in plants using a CRISPR/Cas system

Abstract

Dear Editor, In the past few years, the development of sequence-specific DNA nucleases has progressed rapidly and such nucleases have shown their power in generating efficient targeted mutagenesis and other genome editing applications. For zinc finger nucleases (ZFNs) and transcription activator-like effector nucleases (TALENs), an engineered array of sequence-specific DNA binding domains are fused with the DNA nuclease Fok1 1,2 . These nucleases have been successful in genome modifications by generating double strand breaks (DSBs), which are then repaired through non-homologous end joining (NHEJ) or homologous recombination (HR) in different species, including mouse, tobacco and rice 3,4,5 . Recently, another breakthrough technology for genome editing, the CRISPR/Cas system, was developed. CRISPR (clustered regulatory interspaced short palindromic repeats) loci are variable short spacers separated by short repeats, which are transcribed into non-coding RNAs. The non-coding RNAs form a functional complex with CRISPR-associated (Cas) proteins and guide the complex to cleave complementary invading DNA 6 . After the initial development of a programmable CRISPR/Cas system, it has been rapidly applied to achieve efficient genome editing in human cell lines, zebrafish and mouse 7,8,9,10 . However, there is still no successful application in plants reported. We report here that the CRISPR/Cas system can be used to efficiently generate targeted gene mutations and corrections in plants. The Cas9 gene was driven by the CaMV 35S promoter and the chimeric single guide RNA (sgRNA) was driven by the AtU6-26 promoter in Arabidopsis or the OsU6-2 promoter in rice. We show that the engineered CRISPR/Cas was active in creating DSBs when transiently expressed in Arabidopsis protoplasts and stably expressed in transgenic Arabidopsis and rice plants. Our results demonstrate the feasibility of using engineered CRISPR/Cas as molecular scissors to create DSBs at specific sites of the plant genome to achieve targeted genome modifications in both dicot and monocot plants. We used the optimized coding sequence of hSpCas9 9 driven by the CaMV 35S promoter. For the non-coding RNA components of CRISPR, we expressed the sgRNA using native promoters for U6 RNAs in Arabidopsis (Figure 1A and Supplementary information, Figure S1A) or rice (Supplementary information, Figure S1A). The target site precedes an NGG, the requisite protospacer adjacent motif (PAM). To improve co-delivery, both the sgRNA and hSpCas9 were subcloned into one expression vector (Figure 1A). A split yellow fluorescent protein (YFP) reporter system, YF-FP, was used to test the functionality of the engineered CRISPR/Cas system in Arabidopsis protoplasts (Figure 1B). Co-transformation of the YF-FP reporter and the CRISPR/Cas construct led to the production of strong YFP signal with gene correction rate by HR at 18.8% ((4.76%–0.78%)/21.23%) (Figure 1C). The results suggest that the engineered CRISPR/Cas system is highly functional in generating DSBs on target DNA sequences in plant cells and that the DSBs can be repaired by HR to achieve gene correction. Having successfully targeted a reporter gene in protoplasts, we started to target endogenous loci in plants. The Arabidopsis genes BRASSINOSTEROID INSENSITIVE 1 (BRI1), JASMONATE-ZIM-DOMAIN PROTEIN 1 (JAZ1) and GIBBERELLIC ACID INSENSITIVE (GAI) and the rice genes Rice Outermost Cell-specific gene5 (ROC5), Stromal Processing Peptidase (SPP) and Young Seedling Albino (YSA) were selected for CRISPR/Cas-based disruption (Supplementary information, Figure S1B). These genes were selected owing to obvious growth phenotypes when they are dysfunctional. We designed sgRNAs to target these genes (Supplementary information, Figure S1C). The targets contained restriction enzyme sites close to the PAM sequences, so that the restriction sites may be disrupted when successfully targeted by the CRISPR/Cas (Supplementary information, Figure S2), and RFLP (Restriction Fragment Length Polymorphism) analysis can be used to detect mutations in the target region. The vector containing the Cas9 and sgRNA expression cassette was introduced into plants by Agrobacterium-mediated transformation using floral dipping in Arabidopsis and tissue culture in rice. More than 50 T1 and 20 T0 transgenic plants were generated for each target in Arabidopsis and rice, respectively (Figure 1D). We observed that a high percentage of the Arabidopsis T1 transgenic plants showed growth phenotypes at a very young stage (one week after transplanting in soil) (Figure 1D). For BRI1, more than 50% plants displayed retarded growth and rolling leaves (Figure 1D and 1E), which are expected for bri1 mutant plants. More than a quarter of the T1 plants for GAI also showed a dwarf phenotype (Figure 1D). At later stages, some continued to exhibit a dwarf phenotype that was similar to bri1 or gai mutant plants (Figure 1F and Supplementary information, Figure S1D). The designed target for GAI is located in the DELLA domain (Supplementary information, Figure S1C), which is important for GA-induced degradation of the GAI protein. It is known that amino acid substitutions or deletions in the DELLA domain of GAI would result in insensitivity to GA-induced degradation, leading to a dwarf phenotype. About 10% of T0 transgenic rice plants targeting YSA showed the expected albino leaf phenotype at the seedling stage (Figure 1D and 1G). We genotyped transgenic plants first by RFLP analysis. Clear undigested bands were observed (Figure 1H and 1I). The failure of restriction enzyme digestion suggested the occurrence of DNA sequence mutations in the target regions. We then sequenced the PCR products to see whether there are additional sequence peaks in the target. Results from the two tests showed that the mutation frequency was very high in both Arabidopsis and rice, ranging from 26% (8 out of 31) to 84% (16 out of 19), except for the SPP sgRNA1 target (5%, 1 out of 21) (Figure 1D). Furthermore, the undigested bands from RFLP analysis were cloned and sequenced. We found that in 24 out of the 27 Arabidopsis T1 transgenic plants and 14 out of the 24 rice T0 transgenic plants subjected to sequencing, there were 2 or more different mutated alleles in one single transgenic plant (Figure 1J–1K, Supplementary information, Tables S1 and S2). These plants all contained mutant alleles with small insertions or deletions (indels) at the target sites (Supplementary information, Figures S3–S11). The presence of multiple mutated alleles in the Arabidopsis transgenic plants indicated that in these plants the CRISPR/Cas did not function or certainly did not complete the genome editing during the fertilization stage, and the editing activity continued after the division of fertilized eggs. Regardless, the high frequency of Arabidopsis T1 transgenic plants showing the expected mutant phenotypes suggests that some of the mutations must have been generated very early in development and possibly in early meristematic cells. Therefore, germ line transmission of some of the mutations into T2 plants is expected for many, if not all, of the T1 plants. The identification of 3 bp deletions (which would result in an amino acid deletion) in 2 out of the 3 GAI sgRNA1 T1 transgenic plants (Supplementary information, Figure S6) could well explain the high-frequency dwarf phenotype observed (Supplementary information, Figure S1D). It is also worth noting that one rice T0 transgenic line for ROC5 sgRNA1 (data not shown) and two each for YSA sgRNA1 (Figure 1I, lane 13 and data not shown) and sgRNA2 (data not shown) showed only mutated alleles and no wild-type allele in the RFLP analysis. Sequencing of individual clones revealed that the plants contained only or mostly mutated alleles (Supplementary information, Table S2, Figures S8, S10, S11). Especially for the ROC5 sgRNA1 and YSA sgRNA1 lines, they contained one or two types of mutated alleles only. Importantly, the YSA sgRNA1 rice plants showed the expected albino leaf phenotype (Figure 1G). The result suggests that these rice plants are likely homozygous or bi-allelic mutants, which implies that in this case the CRISPR/Cas may have completed the generation of DSBs in the first meristematic cell during regeneration of the rice plants from transgenic calli. To our knowledge, this is the first study demonstrating highly efficient targeted mutagenesis in multiple genes in Arabidopsis and rice using engineered CRISPR/Cas. Although future studies are needed to examine the germ line transmission and heritability of the CRISPR/Cas-induced mutations and to evaluate any potential off-target effects of the CRISPR/Cas, our results here suggest that the CRISPR/Cas technology will make targeted gene editing a routine practice not only in model plants but also in crops. Detailed methods are described in the Supplementary information, Data S1 and Table S3.