CRISPR/Cas9-mediated mutagenesis of homologous genes in Chinese kale

The clustered regulatory interspaced short palindromic repeat-associated protein 9 (CRISPR/Cas9) system has developed into a powerful gene-editing tool that has been successfully applied to various plant species. However, studies on the application of the CRISPR/Cas9 system to cultivated Brassica vegetables are limited. Here, we reported CRISPR/Cas9-mediated genome editing in Chinese kale (Brassica oleracea var. alboglabra) for the first time. A stretch of homologous genes, namely BaPDS1 and BaPDS2, was selected as the target site. Several stable transgenic lines with different types of mutations were generated via Agrobacterium-mediated transformation, including BaPDS1 and BaPDS2 double mutations and BaPDS1 or BaPDS2 single mutations. The overall mutation rate reached 76.47%, and these mutations involved nucleotide changes of fewer than 10 bp. The clear albino phenotype was observed in all of the mutants, including one that harbored a mutation within an intron region, thereby indicating the importance of the intron. Cleavage in Chinese kale using CRISPR/Cas9 was biased towards AT-rich sequences. Furthermore, no off-target events were observed. Functional differences between BaPDS1 and BaPDS2 were also assessed in terms of the phenotypes of the respective mutants. In combination, these findings showed that CRISPR/Cas9-mediated targeted mutagenesis can simultaneously and efficiently modify homologous gene copies of Chinese kale and provide a convenient approach for studying gene function and improving the yield and quality of cultivated Brassica vegetables.

which flanks the 3′ end of the target sequence, and then achieve the cleavage of the target site 13 . The resulting DSBs trigger non-homologous end joining (NHEJ) or homologous recombination (HR) repair pathways, which may lead to gene mutations 14 . If an insertion, deletion, or substitution occurs at the coding region of the gene, it will probably cause a frameshift mutation in the target gene and may remarkably affect gene function 15 . CRISPR/Cas9 technology has been widely and successfully applied to host DNA mutagenesis in a variety of plants, such as Nicotiana benthamiana 16 , Arabidopsis 17 , wheat 18 , rice 11 , Zea 19 , sorghum 20 , potato 21 , sweet orange 22 , tomato 23 , grape 24 , populus 25 , citrus 26 , and apple 27 . However, despite the existence of a diverse range of Brassica plants, the CRISPR/Cas9 system has only been successfully applied to oilseed rape 28,29 and Brassica oleracea 30 . Brassica vegetables are a large group of agriculturally important vegetables that have high nutritional and economic value. Chinese kale (B. oleracea var. alboglabra) is an original Chinese Brassica vegetable that is widely distributed in southern China and Southeast Asia. In addition to its good flavor, Chinese kale also has high nutritional value due to its high levels of antioxidants and anticarcinogenic compounds, including vitamin C, carotenoids, total phenolics, and glucosinolates 31,32 . Here, we report the application of the CRISPR/Cas9 technology in Chinese kale (a cultivated Brassica vegetable) for the first time.
The phytoene desaturase (PDS) gene, which encodes a key enzyme involving carotenoid biosynthesis, is commonly selected as the target gene for CRISPR/Cas9 experiments in plant, because the albino phenotype caused by the disruption of PDS can be easily recognized 16 . Previously, we cloned and analyzed two homologous gene copies in Chinese kale, namely BaPDS1 and BaPDS2 33 . In this study, we utilized these two BaPDSs as target genes in Agrobacterium-mediated genetic transformation. The present study showed the simultaneous site-directed mutagenesis of these homologous genes in Chinese kale, which provided a theoretical foundation for the application of the CRISPR/Cas9 gene editing technique in Chinese kale and other Brassica vegetables.

Results
SgRNA design and vector construction. The common target site was selected in the fourth exon sequence of both BaPDS1 and BaPDS2. The target sequence is GATGGAGATTGGTATGAAACCGG (Fig. 1A). The target sequences for BaPDS1 and BaPDS2 are completely identical, and the GC content of gRNA is 40%. The recombinant plasmids of pSG-BaPDS and pCC were digested with EcoRI-HF and XbaI-HF, respectively (Fig. 1B). The fragment of the sgRNA expression cassette was purified by gel recovery and ligated into the digested pCC. The recombinant plasmid of pCC-target-sgRNA was successfully obtained for subsequent analysis (Fig. 1C).
CRISPR/Cas9 transformation efficiency in Chinese kale. Using the established genetic transformation protocol of our laboratory (Fig. 2), about 5,000 explants were used for Agrobacterium-mediated transformation. Seventy-five hygromycin-resistant plants were successfully obtained (1.5%). To confirm the existence of the transformed constructs in the transgenic lines, genomic DNA was extracted from each hygromycin-resistant plant. PCR amplification with specific primers targeting the hygromycin gene was conducted. A 558-bp target band was successfully amplified using the empty vector, plasmid (positive control), and the 68 resistant plants. No products were obtained with the template in water, wild-type plants (negative control), and seven out of the 75 plants ( lines 2, 5, 15, 18, 28, 32, and 49) (Fig. 3, Supplemental Fig. 1). These results indicated that the target expression cassette was successfully transferred into the 68 Chinese kale lines. In addition, the false positive rate in this experiment was only 9.33%. Statistical analysis of the BaPDSs mutation efficiency. To detect the mutation efficiency of CRISPR/ Cas9, PCR was performed using specific primers (Table 1) and the 68 transgenic plants, followed by sequencing of the resulting PCR products. When both genes were simultaneously mutated, they were called double mutations, and when there was only one gene mutated, they were called single mutations. Approximately 76.47% of the transgenic plants harbored mutations, including 14 that were BaPDS1 and BaPDS2 double mutants, accounting for 20.59% of the total number of transgenic plants; 10 BaPDS1 single mutants, accounting for 14.70%; and 28 BaPDS2 single mutants, accounting for 41.18% (Table 2). Overall, the ratio of double mutations accounted for about one-fifth of the transgenic plants, and the proportion of the single mutant-type was more than a half, indicating that CRISPR/Cas9 system-induced gene editing has superior mutation efficiency in Chinese kale.
Analysis of the BaPDSs mutations. Fifty-two BaPDSs mutants were obtained by sequencing confirmation. The distribution of mutations of CRISPR/Cas9-editing of the two homologous gene copies was investigated (Fig. 4). Among the BaPDS1 and BaPDS2 double mutants, the proportion of both mutations at the target site or within introns respectively accounted for 21.43% (3/14) of the total number of mutations, and the proportion of mixed mutations was 57.14% (8/14). Among the BaPDS1 single mutants, 80% (8/10) were situated at the  target site, whereas 20% (2/10) occurred in intronic area. Among the BaPDS2 single mutants, the frequency of mutations at the target site and within introns was 17.86% (5/28) and 82.14% (23/28), respectively (Fig. 4). These results suggest that the CRISPR/Cas9 system can simultaneously edit both copies of the genes using the sgRNA, which targets the consensus sequence of the two homologous genes in Chinese kale.
Analysis of the BaPDS1 mutations. The mode of the BaPDS1 mutation was further analyzed. A total of 24 BaPDS1 mutations were identified, including 18 that have one-to nine-base single-nucleotide substitutions at the target site (Fig. 5A). Interestingly, 17 mutants showed a G to A substitution at the ninth base upstream of the PAM locus. This substitution is predicted to result in an amino acid change from tryptophan (Trp) to a stop codon, thereby generating a truncated (nonsense) BaPDS1 protein. A single mutant harbored a G to T substitution six bases upstream of the PAM locus. Similarly, the encoded tyrosine (Tyr) was substituted by a stop codon, thereby also resulting in a truncated protein (Fig. 5A). In addition, we found that among the remaining six BaPDS1 mutants, a 10-bp (ATTAATATAT) insertion was detected in the third intron of BaPDS1 (Fig. 5B). However, regardless of the position of the mutation (target site or intron), the plants showed the same albino phenotype (Fig. 6D,G).

Analysis of the BaPDS2 mutations.
Similarly, the mode of the BaPDS2 mutations was also investigated.
Among the 42 BaPDS2 mutants, nine were found to be mutated at the target site, resulting in a frameshift mutation or amino acid substitution (Fig. 5C). Thirty-one mutants had an AG deletion in the fourth intron and a TA deletion in the third intron, and the remaining two had an AG deletion in the fourth intron. (Fig. 5D). Similarly, regardless of the site of the mutation, BaPDS2 single mutants exhibited a similar albino phenotype (Fig. 6E,H).

BaPDS2-CRISPR test-R GCTATAGAAGATAAGAGCCGAGCCT
Hyg-F CGATTGCGTCGCATCGACC Detection of the hygromycin resistance gene Hyg-R TTCTACAACCGGTCGCGGAG  Patterns and frequency of the BaPDSs mutations. A variety of mutation types were observed in this survey, including substitutions, insertions, and combinatorial mutagenesis (substitutions and deletions occurring simultaneously) (Fig. 7). Among these, up to 51.52% of the mutations were combined mutations, followed by substitutions (39.39%). Insertion mutations accounted for only 9.09% of the observed genomic alterations. All of the CRISPR/Cas9-mediated mutations were short nucleotide changes (≤10 bp) (100%), and 5-bp-long combined mutations were the predominant type, accounting for 45.45% of the observed mutations. All of the insertion mutations were involved in the ATTAATATAT sequence.
Analysis of potential off-targets. To assess whether the CRISPR/Cas9 system specifically edits the BaPDS genes, three potential off-target sequences were selected using the CRISPR-P web tool. Reports have indicated that the 'seed sequence' , which consists of 12 nucleotides located at the target site and adjoining the PAM, is crucial for specific recognition and efficient targeted cleavage of the Cas9 protein 15,34 . Among the three potential off-target sites selected, off-target sites 1 and 3 only showed a 1-bp mismatch in the seed sequence, and off-target site 2 had a 2-bp mismatch in the seed sequence (Table 3). All three potential off-target sites were PCR amplified using specific primers (Table 1). Sequencing did not detect any mutations within all of the tested sites. These results indicate that CRISPR/Cas9-induced target mutagenesis is highly specific to the BaPDSs in Chinese kale and does not have off-target effects.
Albino phenotype of the BaPDSs mutants. All of the obtained 52 PDS mutants showed a clear albino phenotype compared to the wild-type (Fig. 6). Interestingly, the leaves of plants with edited PDS1 and PDS2 were pure albino (Fig. 6C), whereas those with PDS1 or PDS2 single mutations either at the target loci or within introns resulting in a mosaic albino phenotype belonged to chimeric mutants (Fig. 6D-H). These results indicate that the gene function of BaPDSs had been knocked-out or knocked-down.

Discussion
Brassica vegetables have drawn increasing attention owing to its high economic value. Based on morphological characteristics, they are divided into three types of B. rapa, B. oleracea, and B. juncea. Chinese kale is one of the most important vegetables within B. oleracea 31,32 . It has 18 chromosomes and belongs to the CC genome type 35 . As a diploid, Chinese kale is suitable for gene modification and offspring inheritance studies. The screening of targeted mutants has long been employed as a useful strategy for plant functional genomics and crop improvement. However, this approach is extremely laborious and time-consuming as it generally entails traditional mutagenesis methods 36,37 . In recent years, precise, efficient, and multi-site mutagenesis has been achieved in plants using genome editing technology. As the third-generation genome editing technology, the CRISPR/Cas9 system has been applied in various plants since its emergence. Numerous studies have used protoplast transient transformation to integrate the CRISPR/Cas9 system, which can rapidly and efficiently screen the target site and detect mutations 13,18 . However, unlike model plants such as Arabidopsis and tobacco, protoplast technology (isolation, purification, and especially regeneration) remains a major obstacle in accessing mutated Brassica vegetables, thereby limiting its practical application to this particular crop species. Therefore, Agrobacterium-mediated stable transformation was selected in this study, despite the efficiency of transient transformation being higher than that of stable transformation 38 . As expected, a variety of Chinese kale mutants were obtained using the CRISPR/Cas9 system, with relatively high efficiency (76.47%).
Multi-site precise editing is very important for investigations into complex traits and functionally redundant genes. However, the high genetic redundancy in Brassica crops makes it difficult to alter a trait by random mutagenesis and to obtain corresponding mutants, especially recessive mutants 28,39 . The innate ability of Cas9 to simultaneously edit multiple loci within the same individual has numerous potential applications, such as the mutation of multiple members of a gene family or functionally related genes that control complex traits 40 . Targeted mutagenesis of duplicated genes in Camelina sativa was conducted using CRISPR/Cas9 gene editing to generate triple mutants of the three delta-12-desaturase (FAD2) genes 39 . A four-allele mutated line of the granule-bound starch synthase (GBSS) gene obtained by CRISPR/Cas9 resulted in the complete knockout of GBSS enzyme activity in potato 14 . In this study, double mutants of BaPDS1 and BaPDS2 were obtained, and their proportion amounted to about 20% of the total number of transgenic plants ( Table 2). In this study, we selected only one common and conservative co-target site in two homologous genes (BaPDS1 and BaPDS2) of Chinese kale rather than using the tandem multiplex gRNA method that has been employed in previous studies 13,41 because the mutation efficiency induced by multiplex gRNA is significantly lower than that of a single gRNA 13 . Our results confirmed the powerful capability of the CRISPR/Cas9 system in inducing multiplex mutagenesis simultaneously, and furtherly broadened its application in various plant species. Mutation types are affected by different factors and vary among plant species. Previous reports have shown that 1-bp insertions and deletions predominantly occur in rice 11 and Arabidopsis 17 , deletions involving a few bases are the major types of mutations in tobacco 13 , and the proportion of substitutions induced by BnaA6. RGA-sgRNA1 in B. napus are much higher than other sgRNAs 42 . In this study, short nucleotide substitutions predominantly occurred in Chinese kale. This discrepancy may be caused by differences in intrinsic DNA-repair mechanisms among plant species, transformation methods, target genes, or target sites 11,42 .
In this study, some mutations did not occur precisely at the target site, but within flanking regions. The BaPDS2 intronic mutation in 33 lines involved the sequence AAG (Fig. 5D), which is another PAM (NAG) 43 besides NGG 44 for Cas9 cleavage. Conversely, there were six lines that harbored a 10-bp insertion (ATTAATATAT). The mutation involved the third intron of BaPDS1, which is AT-rich (Fig. 5B). Furthermore, there were 31 BaPDS2 mutations with a 2-bp deletion (TA) in the third intron, which is also AT-rich (Fig. 5D). Most of these mutations occurred at AT-rich sites. These findings indicate that Cas9 preferentially cuts and edits AT-rich sequences in Chinese kale, which is similar to that observed in rice 11 , citrus 45 , and Camelina 46 . However, this hypothesis requires further confirmation.
Off-target effects are usually caused by the base-mismatches between sgRNA and the genomic DNA sequences, which hinders further application of the CRISPR/Cas9 technology 47,48 . However, there is growing evidence that off-targets might not be a critical problem in plants because its actual risk may be low during tissue culture-based transformation or other mutagenesis treatments 40 . In this study, we also assessed for potential off-target sites, but found none. These results corroborate the previous conclusion that Cas9/sgRNA specificity is determined by the position and number of mismatches in sgRNA target pairing 15 , and a well-designed specific sgRNA, such as the sgRNA in this study, would not target non-targeted sites 11 . However, this study was not a comprehensive evaluation of the specificity for targeted gene editing using the CRISPR/Cas9 system in Chinese kale, and more candidate off-target sites need to be analyzed in the future.
Numerous studies have shown that intronic sequences are closely related to the expression of plant functional genes. Recent investigations also have revealed that a large number of microRNAs are located within the introns of protein-coding genes, linking their expression to the promoter-driven regulation of the host gene 49,50 . For instance, miR838 that is derived from a hairpin within intron 14 of the DC1 could mediate gene expression by auto-regulation in A. thaliana 51 . In this study, we found several mutations within the third and fourth introns of BaPDSs, and these mutation modalities were relatively higher in frequency (Fig. 5B,D). Furthermore, all of the intron mutations resulted in mosaic albino phenotypes (Fig. 6F-H), indicating a knock-down in the function of the BaPDS genes. These introns may be involved in gene expression in Chinese kale. Nevertheless, the reasons why intron mutations cause loss of gene function remain unclear and will be further studied in subsequent studies.
Generating distinct mutants of multiple copies of homologous genes could be helpful in elucidating the function of the respective genes, and successful examples in plants, such as Camelina sativa 39 and potato 14 , were previously reported. Although we have cloned BaPDS1 and BaPDS2 in Chinese kale, as well as analyzed their expression patterns in an earlier study 33  this study. The pure albino phenotype of the double mutants and the mosaic albino phenotype of the BaPDS1 and BaPDS2 single mutants (Fig. 6) indicate that the functions of BaPDS1 and BaPDS2 are distinct and also partially compensated. In addition, the loss of function of either gene would be lethal to Chinese kale plants, suggesting that both are essential for carotenoid biosynthesis. The rapid development of the CRISPR/Cas9 system has provided a robust platform for plant genomic studies. The present investigation achieved fixed-point editing of the Chinese kale genome using the CRISPR/Cas9 system via stable transformation for the first time. Additionally, a single target site for simultaneously mutating multi-genes to create various mutation materials was also achieved. The findings of the present study strengthen the application of the CRISPR/Cas9 system in cultivated Brassica vegetables, and enrich our understanding of the mutation mode of CRISPR/Cas9 in Brassica crops, and provide a foundation for the further study of gene function through the creation of mutants.

Materials and Methods
Plant materials. The Chinese kale cultivar 'Sijicutiao' which is a multi-generation inbred from our laboratory was used in the study. Sterile seedlings grown on 1/2 Murashige and Skoog (1/2 MS) medium with 0.8% phytagar, and they were placed on tissue culture chambers with temperature of 25/20 °C (day/night), light cycle of 16/8 h (day/night), and light intensity of 100 μmol m −2 s −1 . After seven days, cotyledons with 1-to 2-mm long petioles were cut and used as explants in Agrobacterium-mediated transformation.
Vector construction. The CRISPR/Cas9 vector supports from our laboratory 52 . The plasmid pX330, pCAM-BIA1302, and pUC19 were used as materials for the construction of the CRISPR/Cas9 vector using recombinant DNA technology. The mgfp sequence in pCAMBIA1302 was successfully replaced by the Cas9 sequence from pX330, and the recombinant plasmid was constructed, hereafter referred to as pCC. The sgRNA sequence in pX330 was imported into the multiple cloning site region of the pUC19 plasmid, and the recombinant plasmid was designated as pSG. pCC and pSG comprised the two basic vectors of the CRISPR/Cas9. According to our previous report, the BaPDS gene family has two members, BaPDS1 (GenBank Accession No. KX426039) and BaPDS2 (GenBank Accession No. KX426040) 33 . Sequence analysis of BaPDSs was conducted using the web tools of CRISPR DESIGN (http://crispr.mit.edu/) and ZiFiT Targeter Version 4.2 (http:// zifit.partners.org/ZiFiT/CSquare9 Nuclease.aspx), and the common target sequence of BaPDS1 and BaPDS2 was designed. The synthesized oligos were annealed and inserted into the BbsI sites of the pSG. Then, the recombinant plasmid (pSG-BaPDS) was digested by EcoRI-HF and XbaI-HF, and the sgRNA expression cassette was inserted into the multiple cloning site region of the pCC vector to generate pCC-target-sgRNA for Chinese kale transformation. All of the above vectors were verified by restriction and sequencing.

Agrobacterium-mediated transformation of Chinese kale plants. The transformation of Chinese
kale was conducted as previously described, with modifications 53,54 . The explants pre-cultured for 3 d in Murashige and Skoog (MS) medium with 0.5 mg/L 2,4-D and 0.8% phytagar were infected with the Agrobacterium strain GV3101 (optical density at 600 nm between 0.6 and 0.8) by immersion for 1-2 min. The explants were co-cultivated with Agrobacterium strain GV3101 for three days in MS medium with 0.03 mg/L naphthaleneacetic acid (NAA), 0.75 mg/L boric acid (BA), and 0.8% phytagar, and then transferred to the MS medium supplemented with 0.03 mg/L NAA, 0.75 mg/L BA, 0.8% phytagar, 325 mg/L carbenicillin, and 325 mg/L timentin for a week. Hygromycin-resistant shoots regenerated on the same medium with 12 mg/L hygromycin B were transferred to tissue culture bottles containing the subculture medium. After three months, hygromycin-resistant plantlets were obtained and prepared for subsequent analysis.
Transformation efficiency detection. Genomic DNA was extracted from the shoots of hygromycin-resistant and wild-type plants using a standard cetyltrimethylammonium bromide (CTAB) method. To confirm the Chinese kale transformants, PCR amplification using the specific primers HygR-F and HygR-R (Table 1), which were designed to flank the hygromycin sequence contained in the vector, was performed. The PCR products were detected using electrophoresis to calculate transformation efficiency. The transformants without the hygromycin gene were not used for further analysis.

Mutation detection.
To evaluate mutations introduced in the CRISPR/Cas9 transgenic plants, genomic DNA of each positively transgenic shoot was amplified using specific primers, which were designed to amplify the 816-bp and 702-bp flanking regions of the target site of BaPDS1 and BaPDS2, respectively. For genotyping of the transgenic plants, the PCR products were sequenced. DNAMAN software (version 6.0; Lynnon Corporation, Canada) was used to compare the sequence, the mutation rate was calculated, and all of the sequencing data were collected to analyze the mutation type.
Off-target analysis. Potential off-target sites were identified with the CRISPR-P 55 online software (http:// crispr.hzau.edu.cn/CRISPR2/) using the full 20-bp target sequence in a BLAST analysis. These top-ranking potential off-target sites containing 1-2 bp mismatches in the 12-bp seed sequence were selected ( Table 3). The sequences surrounding the potential off-target sites were amplified using specific primers (Table 1), and the PCR products were analyzed by Sanger sequencing.