CRISPR/Cas9‐mediated efficient targeted mutagenesis in grape in the first generation

Summary The clustered regularly interspaced short palindromic repeats‐associated protein 9 (CRISPR/Cas9) system is a powerful tool for editing plant genomes. Efficient genome editing of grape (Vitis vinifera) suspension cells using the type II CRISPR/Cas9 system has been demonstrated; however, it has not been established whether this system can be applied to get biallelic mutations in the first generation of grape. In this current study, we designed four guide RNAs for the VvWRKY52 transcription factor gene for using with the CRISPR/Cas9 system, and obtained transgenic plants via Agrobacterium‐mediated transformation, using somatic embryos of the Thompson Seedless cultivar. Analysis of the first‐generation transgenic plants verified 22 mutant plants of the 72 T‐DNA‐inserted plants. Of these, 15 lines carried biallelic mutations and seven were heterozygous. A range of RNA‐guided editing events, including large deletions, were found in the mutant plants, while smaller deletions comprised the majority of the detected mutations. Sequencing of potential off‐target sites for all four targets revealed no off‐target events. In addition, knockout of VvWRKY52 in grape increased the resistance to Botrytis cinerea. We conclude that the CRISPR/Cas9 system allows precise genome editing in the first generation of grape and represents a useful tool for gene functional analysis and grape molecular breeding.


Introduction
Grape (Vitis vinifera L.) is a widely cultivated perennial fruit crop that has great economic value as it is a source of many products, including wine, jam, juice and jelly, grape seed extracts, raisins, vinegar and grape seed oil (Tu et al., 2016). However, the yield and berry quality of grape is limited by a range of biotic and abiotic stresses (Li, 2015). There is therefore considerable interest in understanding the molecular mechanisms that grape has evolved to resist such stresses. Most functional studies of grape resistance genes have used overexpression analysis (Dai et al., 2015); however, as this approach does not always accurately reflect normal gene function (Lloyd, 2003), gene silencing methods are now also being widely employed. To date, RNAi, VIGS (virus-induced gene silencing) and gene editing have been the most commonly used means for gene silencing (Pandey et al., 2016;Shan et al., 2013;Wei et al., 2017).
As the CRISPR/Cas9 technology has proven so effective, an increasing number of gene functional studies are based on this system. In A. thaliana, for example, the cbf1 cbf3 and cbf1 cbf2 cbf3 (cbfs) mutants were generated by mutating CBF1 and CBF1/ CBF2 in a cbf3 T-DNA insertion mutant via CRISPR/Cas9-mediated system, and the resulting lines used to study responses to chilling and freezing stresses (Jia et al., 2016). An example from rice is the knockout of the ERF transcription factor gene, OsERF922, via CRISPR/Cas9, which gave rise to improved resistance to the rice blast fungus . In maize, a ARGOS8 mutant generated by the CRISPR-Cas9 system showed increased resistance to drought stress (Shi et al., 2017), and in tomato, the mutations of RIN gene induced by CRISPR/Cas9 affected fruit ripening of T0 plants, while the mutations were shown to be stably inherited in the T1 generation (Ito et al., 2015).
In grape, five CRISPR/Cas9 target sites were identified and characterized (Wang et al., 2016b), and other reports have shown that this genome editing approach can be used with grape suspension cells or protoplast (Malnoy et al., 2016;Ren et al., 2016). However, little is known regarding the efficiency of CRISPR/Cas9-mediated targeted mutagenesis in the first generation of grape mutants. For example, it is important for gene functional analysis and molecular breeding strategies to establish whether homologous mutations can be efficiently generated. In this study, we chose to target a gene (VvWRKY52) from the grape WRKY transcription factor family, which has been shown to play roles in biotic stress responses . Four specific targets were designed in the first exon region of VvWRKY52, and 22 independent mutant strains were generated, of which 15 were homozygous. We analysed the variation caused by four different target sites and confirmed that the technology can be efficiently used to generate homozygous mutations in grape in the first generation. In addition, we also demonstrated that the CRISPR/Cas9 system can be used for precise genome editing in the first generation of grape.

Target selection and vector construction
The VvWRKY52 gene is located on chromosome 16 in grape reference genome ( Figure S1), and no other copy of the VvWRKY52 was found in the Grape Genome Database (129; http://www.genoscope.cns.fr). In addition, the whole genome of the V. vinifera Thompson Seedless cultivar was sequenced in 2014 (Di Genova et al., 2014), and the copy number variation (CNV) regions were identified between the grape reference genome (PN40024) and Thompson Seedless (Cardone et al., 2016). However, no CNV was found at the region of VvWRKY52. This indicated that VvWRKY52 only has a single copy in the genome of Thompson Seedless. We also found two kinds of sequences at the target region of VvWRKY52, which are shown in Figure S1, allele I and allele II. According to the sequence results, 102 of 203 clones were allele I and others were allele II. This indicated that the sequences of these two sites are equal in the genome. We also analysed the mutant information at target 1 and target 4 in the mutant lines (Tables S3 and S4), and the results showed that only one kind of mutant type or wild-type (WT) type was found in each allele of the mutant lines. Given that the mutation caused by this CRISPR/Cas9 system is random, each allele should have more than one mutant type at every target site in most mutant lines if the VvWRKY52 has other copy in the grape genome. Therefore, we think that VvWRKY52 only has a single copy in the genome of Thompson Seedless. The sequence of VvWRKY52 gene was amplified with gene-specific primers from genomic DNA of the V. vinifera Thompson Seedless cultivar, and used for target selection. The four targets were designed using the online tools CRISPR-P (http://cbi.hzau.edu.cn/crispr/) and CRISPR RGEN (http://www.rgenome.net/), based on their GC content and putative off-target sites (Table 1). Then, we confirmed that no SNPs were detected at four target sites compared with the Thompson Seedless genome sequence. All selected targets were located in the first exon and their relative position is shown in Figure 1a. Four A. thaliana promoter sequences, AtU3d, AtU3b, AtU6-1 and AtU6-29, were used to drive expression of the T4, T1, T2 and T3 targets, respectively ( Figure 1b). The expression cassettes including the four targets were inserted into the binary vector pYLCRISPR/Cas9P35S-N  using BsaI (Figure 1b).

Grape transformation and identification of transgenic mutant lines
To generate the proembryonal masses (PEM) used for grape transformation, flower buds of Thompson Seedless were selected for embryo callus induction ( Figure 2a) (Dhekney et al., 2012). The embryo callus was then transferred to X6 medium for 1-3 weeks and maintained in the dark at 26°C. After 3-5 weeks, the PEM developed on the X6 medium ( Figure 2b). Agrobacterium tumefaciens strain EHA105, containing the CRISPR/Cas9 vector, was used to transform the PEM as previously described (Dhekney et al., 2012). After co-culturing, the PEM was transferred to solid DM medium for callus induction, and the reinduced callus was then transferred to X6 medium with 200 mg/L carbenicillin, 200 mg/L cefotaxime and 75 mg/L kanamycin (X6CCK75) (Figure 2c). The kanamycin-resistant PEM (Figure 2d) was obtained after maintaining the callus in the dark at 26°C on X6CCK75 medium for 4-6 months, and the somatic embryos (SE) (Figure 2e) developed from the PEM were selected for transgenic plant recovery (Figure 2f,g). The entire experiment cycle was approximately 12 months, dating from the target design to the identification of mutant transgenic lines ( Figure 3). Genomic DNA was extracted from leaves of transgenic lines and WT plants, and PCR was performed to confirm presence of the transgene using vector-specific primers (NPTII-F: 5 0 -AGAGGC-TATTCGGCTATGACTG-3 0 ; NPTII-R: 5 0 -CAAGCTCTTCAGCAA-TATCACG-3 0 ). All 72 putative transgenic plants tested positive in this regard ( Figure S2). In order to detect the mutations in the transgenic plants, the putative edited area of VvWRKY52 was amplified by gene-specific primers (VvWRKY52-Target-F; VvWRKY52-Target-R) from all the transgenic lines and WT plants ( Figure S3), and the PCR products were sequenced. Of the 72  Figure S1. In view of VvWRKY52 as a single copy gene, we think these sequences come from two different alleles, one inherited from the mother and another from the father. The information was used to identify the biallelic mutant lines. Fifteen of 22 mutant lines carried biallelic mutations, while the remaining had only mutations on one allele (Figures 4 and 7a). The morphologies of the different transgenic lines and WT plants are shown in Figure 5. No significant differences were found between the phenotypes of the WT and transgenic lines #20, #26, #40, #45, #38 and #42.
In addition, we observed that the large deletions were often induced by a combination of effects from the different targets. As shown in Figure 6b, four different large deletions were found in five transgenic lines (#38, #41, #47, #52 and #69). A 52-bp deletion was detected in #38 and #52 between T3 and T4, a 100bp deletion in lines #47 and #69 between T1 and T3, a 71-bp deletion in line #52 between T1 and T2 and a 152-bp deletion in line #41 between T1 and T4 (Figure 6b,c). This indicated that the CRISPR/Cas9 system can be used for precise genome editing in the first generation.

Characterization of mutations in mutant lines in the first generation
In order to further understand the mutation efficiency of various targets, we counted the mutagenesis of four targets sites in the transgenic T0 plant. The mutation efficiency in the sites T1 (27.78%), T3 (16.67%) and T4 (25.00%) were much higher than T2 (5.55%). Thirteen of 18 transgenic lines with mutation at target 4 were biallelic mutation and 6 of 20 for target 1 (Table 2). In addition, no biallelic mutation was found in T2 and T4 sites. This indicated that the target with high efficiency would improve the opportunity to get the biallelic mutation lines in the first generation.
In order to identify the biallelic mutant lines, 5-24 randomly selected clones for every independent line were sequenced. Fifteen of 22 transgenic lines were found to contain biallelic mutations and seven were monoallelic mutation ( Figure 4). However, three of 15 biallelic mutation lines were ORF-preserving mutations and only 12 lines were loss of function in two alleles ( Figure 7a). The deduced amino acid sequences from the mutated DNA sequences of six biallelic mutant lines (#37, #42, #53, #60, #61 and #68) are shown in Figure 7b. A mutation that would result in the early termination of translation of two different alleles was found in four selected lines (#37, #42, #60 and #68). However, in #53 and #61, this mutation was only found in one allele, and the other was a transcoding mutation within the region of the selected 87 amino acids (Figure 7). This indicated that biallelic mutant lines can be efficiently obtained using this CRISPR/Cas9 system in grape.

Off-target analysis
To detect off-target events, the 12 transgenic lines with biallelic mutations and WT plants were selected for sequence analysis. The putative off-target sites for each target were predicted with the online tool CRISPR-P. Six putative off-target sites, which may produce the mutations in the six different genes, were selected for further study (Table 3). The primers used for amplifying the offtarget areas are listed in Table S1. The PCR products were purified and sequenced using the forward primer, which had also been used for the amplification. No mutation was found in any of the 72 (12 lines 9 6 putative off-target sites) sequenced samples (Table 3).  To prove whether the targeted mutations of VvWRKY52 affect the resistance of Thompson Seedless against Botrytis cinerea, the phenotypes of the WT and transgenic lines #20, #40, #45, #38 and #42 were selected for further analysis. We observed that in the detached leaves of four transgenic lines (#40, #45, #38 and #42) with mutant, the resistance to B. cinerea was found to be increased compared with WT plant at 5 days postinoculation. Interestingly, transgenic lines with biallelic mutant (#38 and #42) showed higher resistance than lines with monoallelic mutant (#40 and #45) and line without mutant (#20) showed no significant difference compared with WT plant (Figure 8a). We also observed the cell death and fungal structures of leaves from different phenotypes at 5 days postinoculation (Figure 8b,c). The largest cell death was found in WT plant and line #20, followed by lines #40 and #45, and lines #38 and #42 had the smallest (Figure 8b). The B. cinerea colonies growing on the WT plant and line #20 were the largest, followed by those on lines #40 and #45, while lines #38 and #42 had the smallest colonies (Figure 8c). These results are consistent with the phenotype in Figure 8a. In addition, we counted the percentage of spreading lesions on the phenotypes of the WT and transgenic lines at 5 days after inoculation (Figure 9). The results are consistent with disease phenotype in Figure 8a and the cell death in Figure 8b.

Discussion
The use of the CRISPR/Cas9 system has developed rapidly in recent years , and the technology is being applied to gene functional analysis and molecular breeding (Cao  Yang, 2017). In grape, it has been shown to be effective with both protoplasts (Malnoy et al., 2016) and transgenic suspension cells (Ren et al., 2016), and appropriate target sites have been identified in the V. vinifera genome (Wang et al., 2016b). However, it has not been established whether this system could efficiently produce biallelic mutations, or the extent of offtargeting in the first generation: the former issue is especially important for woody plants with long reproductive cycles (Fan et al., 2015).
Grape is one of the most economically important fruit trees grown worldwide, and the efficient production of biallelic mutations in the first generation would be highly significant. In recent years, many CRISPR/Cas9 toolkits for multiplex genome editing in plants have been reported Xing et al., 2014), making it possible to design multiple targets for one gene and significantly improve the mutation efficiency, accelerating the application of this technique. In this current study, we designed four targets for one gene (Figure 1a) and the expression cassettes were inserted into a binary vector (Figure 1b), which had previously been shown to work in poplar (Fan et al., 2015).
Organogenesis and somatic embryogenesis are two commonly used strategies for grape transformation (Martinelli et al., 1993;Robacker, 1993). Compared to the organogenesis pathway, somatic embryogenesis has a very low chimera rate and is the more frequently used method for regeneration in grape (Gambino et al., 2007), so we used this approach, following a protocol described in Dhekney et al. (2012), with minor changes. An overview is provided in Figures 2 and 3. We used the PEM for transformation instead of the SE, which has been previously reported (Zhou et al., 2014b), and which we found to be effective.
affected by factors such as plant species, target sequence, Cas9 promoter and sgRNA sequences and transformation method . In our study, we obtained 72 regeneration lines through Agrobacterium-mediated transformation, all of which had the T-DNA insertion ( Figure S1), indicating that the method described here is optimal. Twenty-two (31%) of the 72 transgenic lines contained mutations in the target sites, which is a high mutation rate for grape. CRISPR/Cas9 is known to cause biallelic mutations in several plant species, such as rice , A. thaliana  and maize (Svitashev et al., 2015). Moreover, in our study, 15 of the 22 transgenic lines with mutations were shown to be biallelic (Figures 4 and 7), making this an attractive method to use in grape. We observed no difference in phenotype between the different genotypes ( Figure 5), indicating that knocking out this gene did not influence the growth and development of the plants. WRKY  transcription factor family has been shown to play roles in biotic stress responses (Guo et al., 2014). To identify the function of VvWRKY52 in biotic stress, we compared the resistance of the phenotypes of WT and five transgenic lines against B. cinerea.
We found that knocking out VvWRKY52 in grape increased the resistance to B. cinerea (Figures 8 and 9). It indicated that the CRISPR/Cas9 system represents a useful tool for gene functional analysis in grape.
The CRISPR/Cas9 system generates insertion or deletion mutations; for example, a study in apple revealed that all the mutations were short insertions or short deletions (Nishitani et al., 2016). Here, we found that most mutations were short deletions (Figure 6a), while a previous study in grape showed that most of the mutations were short insertions (Ren et al., 2016). We speculate that this difference reflects the use of different genotypes (Chardonnay/Thompson Seedless) or different sample refer to the two alleles. '*' indicates that translation is terminated. The sequence in the box refers to a transcoding mutation compared to wild type (WT). The numbers of clones detected with this mutation and the numbers of predicted amino acids are shown in black.  (Fan et al., 2015) and, indeed, we identified four different large deletions (Figure 6b,c), which were induced by cutting at two or more target sites. This indicated that the CRISPR/Cas9 system can be used for precise genome editing in the first generation. We concluded that the CRISPR/Cas9 system can be used to efficiently generate biallelic mutant lines in the first generation. However, as off-target mutation events may influence the application of this system (Ren et al., 2016), we also investigated their occurrence. A previous study showed that a GC content of 50%-70% in the target sequences usually results in a high editing efficiency , which is consistent with our results. The GC contents of the four selected targets are listed in Table 1. The mutation efficiency was much higher in the T1 and T4 sites, which had GC contents of 55% and 65%, respectively. Targets with high editing efficiency may also be associated with a greater risk of off-targeting, which we hypothesize might be the case with the T1 and T4 sites. We selected six potential off-target sites for further analysis (Table 3) and screened all 12 transgenic lines with biallelic mutations, as well as WT plants. No mutation was found in any of the 72 (12 lines 9 6 putative off-target sites) sequenced samples (Table 3), suggesting a very low number of off-target events in grape. To summarize, our study demonstrated that the CRISPR/Cas9 system is a useful tool for efficiently generating biallelic mutation lines in the first generation of grape transformants, and will likely accelerate grape gene functional research and molecular breeding.

Plant material and cultures
Thompson Seedless seedlings were grown in the grape germplasm resources orchard at the Northwest A & F University, Yangling, Shaanxi, China. The embryogenic callus was induced from floral explants of Thompson Seedless according to a previously published protocol (Gribaudo et al., 2004). The embryogenic calli were transferred to X6 medium (Caisson, MSP24-1LT) to form PEM, which were used for grape transformation. All cultures described above were maintained in the dark at 26°C.

Plant transformation
The PEM, maintained in the dark at 26°C, was transferred to fresh X6 medium for 1 week and then used for grape transformation. The binary vector was introduced into A. tumefaciens strain EHA105 using the freeze-thaw method (Wise et al., 2006). The A. tumefaciens culture used for co-cultivation was prepared as previously described (Zhou et al., 2014b). Agrobacteriummediated transformation of the PEM was performed according to Dhekney et al. (2012) with minor modifications. Briefly, the bacterial culture (OD600, 0.4-0.6) was incubated with the PEM for 7 min, and the PEM was then transferred onto filter paper to remove excess bacteria. Blotted PEM was transferred to a Petri dish containing two layers of filter paper with liquid DM medium (DKW basal salts, 2.0 mg/L each of thiamine-HCl and glycine, 1.0 mg/L nicotinic acid, 0.3 g/L KNO 1.0 g/L myo-inositol, 30 g/L sucrose, 5.0 mM 6-benzyladenine, 2.5 mM 2-naphthoxyacetic acid and 2.5 mM 2,4-dichlorophenoxyacetic acid, pH 5.7) and cocultivated in the dark at 26°C for 3 days. After 3 days, the PEM was transferred to solid DM medium containing 200 mg/L carbenicillin, 200 mg/L cefotaxime and 75 mg/L kanamycin, for 1 month in the dark at 26°C. The resulting callus was transferred to X6 medium with 200 mg/L carbenicillin, 200 mg/L cefotaxime and 75 mg/L kanamycin. Petri dishes were placed in the dark for the development of transgenic SE lines. SE at the late cotyledonous stage was transferred onto MS1B medium (MS salts and vitamins, 0.1 g/L myo-inositol, 20.0 g/L sucrose, 1.0 mM 6-benzyladenine, and 7.0 g/L TC agar, pH 5.8) under a 16-h photoperiod with white fluorescent lights, to regenerate the plants.

Detection of mutations
To identify stable transgenic lines, vector-specific primers (NPTII -F: 5 0 -AGAGGCTATTCGGCTATGACTG-3 0 ; NPTII -R: 5 0 -CAAGCTCTT CAGCAATATCACG-3 0 ) were used. The potential edited area of VvWRKY52 was amplified using gene-specific primers (VvWRKY52-Target-F; VvWRKY52-Target-R) with the PrimeSTAR â Max DNA Polymerase kit (Takara) from stable transgenic lines and WT plants. The PCR product was purified and sequenced using the specific primer, VvWRKY52-Target-F. All mutant transgenic lines were selected for the next identification step. PCR products from transgenic lines whose sequence showed a bimodal pattern were inserted into the pClone007 Simple Vector (TSINGKE). Single clones were sequenced and 5-20 single clones from each stable transgenic line with mutants were selected for further analysis to identify the mutation. DNAMAN (version 4.0; Lynnon Biosoft, Inc., San Ramon, CA) was used for alignment analysis.

Off-target analysis
The stable transgenic lines with biallelic mutants were used for off-target analysis. The potential off-target sites of the four targets, predicted using the online CRISPR-P (http://cbi.hzau.ed u.cn/crispr/) tool, were selected for further analysis (Table 3). Specific primers were designed and used to amplify the genomic DNA fragments with potential off-target sites (Table S2), and then, the genomic DNA fragments were sequenced.

Inoculation of grape with pathogen
Botrytis cinerea isolated from grape was maintained on potato glucose agar medium in the dark at 25°C. After 21 days, conidia were used for inoculation. The B. cinerea conidial suspension (1.5 9 10 6 conidia/mL) was prepared and used for inoculation by spraying as previously described . Detached Figure 9 The percentage of spreading lesions counted at 5 days after inoculation. Lines #20 had the T-DNA insertion, but no mutations. Lines #40 and #45 contained single allele mutations, and lines #38 and #42 contained biallelic mutations. The mean AE SD from three independent experiments is represented by bars. Asterisks indicate the statistical significance between transgenic lines and wild-type (WT) plants (**P < 0.01, Student's t-test). leaves with the same size were selected and transferred to a bed of 0.8% agar in trays quickly. The leaves were sprayed with conidial suspension. Then, preservative film was used to cover the trays to ensure a relative humidity of 90%-100% as previously described (Wan et al., 2015). Spraying with distilled water was used as the control. All trays were maintained in the dark for 24 h, then in a light/dark (16/8-h) regime at 22°C. For each different lines and WT plant, at least 18 leaves from three biological replicates were tested. The percentage of spreading lesions was counted at 5 days after inoculation (Wan et al., 2015). Cell death and the fungal structures were stained with trypan blue as previously described .

Statistical analysis
Microsoft Excel (Microsoft Corporation, Redmond, WA) and Sigma plot (v. 10.0; Systat Inc., Point Richmond, CA) were used for data analysis. SPSS Statistics 17.0 software (IBM China Company Ltd., Beijing, China) was used to assess the significant differences through paired t-tests. All experiments were repeated three times as independent analyses.

Figure S1
Schematic map of VvWRKY52 location and two alleles of VvWRKY52 in Thompson Seedless. Allele I and Allele II are part of the coding sequences of VvWRKY52. '*' and red line indicated the difference. Figure S2 Identification of T-DNA insertion of 72 transgenic lines. 'P' means positive control and WT (wild type) was negative control. Figure S3 The DNA fragments from independent transgenic lines were amplified for sequencing. The number indicated different transgenic lines. WT indicated non-transgenic line. Table S1 Primers used for off-target analysis.