Previously, an existing CRISPR/Cas9 vector pKSE401 was successfully applied into Arabidopsis and B. napus to generate mutants with a high efficiency (Xing et al., 2014; Yang et al., 2017). However, verification of positive transformants and isolation of transgene-free offspring are tedious. To overcome this drawback, an optimized CRISPR/Cas9 vector, named pKSE401G, was made by inserting a 35S::sGFP cassette into pKSE401 (Figure 1A) (Xing et al., 2014). Other parts of the vector remained intact. Specifically, zCAS9-NLS was under the control of two constitutive 35S promoters, and expression of the two sgRNA scaffolds was driven by the Arabidopsis U6-26 and U6-29 promoter, respectively (Figure 1B) (Xing et al., 2014). To make constructs, two sgRNAs should be designed for one target gene. The fragments could be amplified by over-lapping PCR using the vector pCBC-DT1T2 (Xing et al., 2014) as template, and then inserted into pKSE401G that was linearized by the type II restriction enzyme BsaI through the Golden Gate Assembly method.

In order to assess the editing efficiency of pKS401G, two sgRNAs were first designed for mutating the Arabidopsis gene RECEPTOR-LIKE PROTEIN KINASE 1(RPK1) (Figure 2A). This gene encodes a leucine rich repeat receptor like kinase that positively regulates abscisic acid (ABA) signaling transduction in Arabidopsis (Osakabe et al., 2005). The construct was transformed into wild-type Arabidopsis Col-0. In the T1 generation, some green seeds were easily identified under the fluorescence microscope, which are potential positive primary transformants harboring the GFP-expressing cassette (Figure 2B). When these green seeds germinated, the green fluorescence signal was still present in the young seedlings (Figure 2C). These CR-rpk1 (CRISPR-rpk1) plants were further confirmed by PCR based genotyping using the gene specific primers of Cas9 (Supplementary Figure S1A). As a result, all the GFP-positive lines contained the Cas9 sequence.

Next, the mutation frequency was analyzed by direct sequencing in the T1 generation. All sequencing results of each amplicon derived from individual transformants were decoded by using the website DSDecode¹ (Liu et al., 2015). The results showed that both target sites were successfully edited (Figure 2D), however, the editing efficiencies were significantly different. At sgRNA1 target site, 33.3% (9 of 27) of the T1 plants contained detectable mutations. In contrast, the mutation rate was much lower (7.4%, 2 of 27) at sgRNA2 target site. A low proportion of homozygous mutations (3.7%, 1 of 27) could also be identified.

Cas9-free mutants provide reliable materials for gene functional studies and crop improvement. To isolate Cas9-free mutants, the seeds from nine T1 transgenic lines were harvested for screening. The seeds that did not contain the CRISPR/Cas9 insertion were selected based on the absence of GFP fluorescence (Figure 3A). In total, 251 seedlings were screened. GFP signal was detected in 82.1% (206/251) of transgenic plants, while only 17.3% (45/251) of plants did not have the GFP signal (Supplementary Table S1). As lack of green fluorescence could be due to gene silencing (Gao et al., 2016), the GFP-negative plants were further checked by PCR using Cas9-, GFP-, and NPTII-specific primers, respectively (Supplementary Figure S2, left panel). Consequently, 37 plants were confirmed as Cas9-free. Then genotyping of these 37 Cas9-free plants revealed that 56.8% (21/37) were homozygous mutants, and 21.6% (8/37) were bi-allelic mutants. Thus, a total of 78.4% had defects in both alleles (Figure 3B and Supplementary Table S2). Among all the homozygotes and bi-allelic mutants, 100.0% (29/29) of them carried a 1-bp insertion (i1) (Supplementary Table S2). The frequency of heterozygotes was 16.2% (6/37), and 5.4% (2/37) were not edited (Supplementary Table S2). The results demonstrated that screening Cas9-free material is more efficient owing to application of the GFP reporter.

To determine whether the mutations in the Cas9-free plants could be stably transmitted to offspring, 20 individual plants obtained from Cas9-free lines L8-3 and L18-1 in T2 were genotyped. All detected plants were homozygous at sgRNA1 target site with the same mutations as their parents, and no new mutation was identified at sgRNA2 target site (Supplementary Table S3). These results suggested that the mutations in Cas9-free plants could be stably transmitted to the next generation following a Mendelian law in Arabidopsis.

Next, we examined the expression level of RPK1 in the CR-rpk1 mutants. Several independent lines were examined, in all of which the transcript level was greatly reduced (Figure 3C and Supplementary Figure S4A), probably caused by non-sense mediated decay. As previously reported, one important phenotype of the RPK1 mutant is less sensitive to ABA in Arabidopsis (Osakabe et al., 2005). Therefore, we measured the stomatal aperture after ABA treatment for 1 h in three independent CR-rpk1 mutants. As expected, the stomatal aperture became much smaller in wild type after ABA treatment; in contrast, stomatal closure was largely inhibited in the CR-rpk1 mutants (Figure 3D). These data demonstrated that the pKSE401G vector could be used for gene functional studies in Arabidopsis.

We then tested genome editing efficiency of pKSE401G in the allotetraploid plants B. napus. BnaARF2 was selected as the target gene, which is the ortholog of Arabidopsis AUXIN RESPONSE FACTOR 2 (ARF2) gene, a repressor in auxin signaling (Schruff et al., 2006). There are four BnaARF2 paralogous genes in the B. napus genome, namely, BnaA6.ARF2, BnaA9.ARF2, BnaC3.ARF2, and BnaC9.ARF2 (Supplementary Figure S5). Two sgRNAs were designed to target the conserved sequences among the four paralogous genes (Figure 4A). GFP signal was tracked as a selection marker of positive calli and regenerated plants during the transformation process (Figures 4B,C), which greatly increased work efficiency. Finally, a total of five individual GFP-positive plants were isolated. All of these transformants were further confirmed by PCR using Cas9-specific primers (Supplementary Figure S1B).

Next, we used Sanger sequencing to assess the editing efficiency. Each of the four BnaARF2 genes was, respectively, examined in the five transgenic plants. As a result, the mutation frequency at sgRNA1 target site is always 20% for each of the four genes, while the mutation frequency at sgRNA2 target site ranged from 60.0 to 100.0% (Table 1). With a combination of the two target sites, three transgenic plants (L34, L35, and L37) were quadruple mutants, one (L38) was triple mutant (BnaA9.ARF2 BnaC9.ARF2, and BnaC3.ARF2), and one (L39) was double mutant (BnaC9.ARF2 and BnaC3.ARF2). Among all of the types of mutations, 47.2% (17/36) were deletions, 25.0% (9/36) were insertions, 13.9% (5/36) were combined mutations, and 13.9% (5/36) were substitutions (Figure 4D and Supplementary Figure S3). The number of deleted nucleotides ranged from 1 to 10 (Supplementary Figure S3). The average mutation frequency in the T0 generation in B. napus was much higher than that in the T1 generation in Arabidopsis (52.5% vs. 20.4%).

The five positive transgenic plants were self-crossed to obtain the T1 population. In T1, GFP signal was examined in a total of 95 seedlings from L35, L37, L38, and L39 (Figure 5A). Consequently, 24 seedlings (25.7%) lost the GFP signal (Supplementary Table S1). All of these GFP-negative plants were further confirmed by using Cas9-, GFP- and NPTII-specific primers, respectively (Supplementary Figure S2, right panel). Next, mutations in BnaARF2s at the sgRNA2 target site were examined in 20 Cas9-free plants in T1 from L35, L37, and L38 (Figure 5B, Table 2, and Supplementary Table S4). Homozygous mutations were identified in BnaC9.ARF2, BnaA6.ARF2, and BnaC3.ARF2 (Supplementary Table S4). Bi-allelic mutations were identified in BnaA9.ARF2 and BnaC9.ARF2 (Supplementary Table S4). Heterozygotes were found in BnaA6.ARF2, BnaA9.ARF2, and BnaC3.ARF2 (Supplementary Table S4). The results suggested that fluorescence screening would speed up getting Cas9-free mutants in B. napus.

To test whether the mutations in the Cas9-free plants could be stably transmitted to the next generation in B. napus, we compared the mutations between one T1 plant with its 10 offspring in T2 (L35-1). As shown in Supplementary Table S5, the mutations in the four BnaARF2 genes were either homozygous or bi-allelic in the one T1 plant at sgRNA2 target sites, and these mutations remained the same in the 10 T2 plants. In addition, there is no newly induced mutation at sgRNA1 target site. These results indicated that the mutations in Cas9-free mutants were inheritable following a Mendelian law in B. napus.

Among the transgenic plants, the quadruple mutants of BnaARF2s, called CR-bnaarf2, were identified in L35 and L37, respectively. Furthermore, all of the four BnaARF2 genes were remarkably down-regulated in these mutants compared to wild type (Figure 5C and Supplementary Figure S4B). In Arabidopsis, arf2 mutants are defective in several aspects, such as reduced fertility, delayed senescence, enlarged rosette leaves, seeds and cotyledons (Schruff et al., 2006). We also examined the phenotypes of CR-bnaarf2. The primary roots were much shorter than WT (Figures 5D,E), the seed size and 1000-seed weight were significantly increased compared to WT (Figures 5F,G), indicating that BnaARF2 is a positive regulator of root elongation and a negative regulator of seed enlargement in B. napus. Together, these data suggested that the pKSE401G vector performed well in targeted genome editing in B. napus.

The U6 and U3 promoters used for driving the two sgRNAs have a major influence on genome editing efficiency (Bortesi et al., 2016; Pan et al., 2016; Yang et al., 2017). Arabidopsis U6 and U3 promoters worked well in B. napus, perhaps because the two species are closely related. Then we would like to test the genome editing efficiency of pKSE401G in other dicot plant species which is evolutionarily far from Arabidopsis. First, we tried it in the diploid woodland strawberry Fragaria vesca in Rosaceae family. The well-studied gene FveMyb10 was chosen as the target, as transient knock-down of FveMyb10 provoked loss of coloration in fruit receptacle (Lin-Wang et al., 2014). Similarly, two sgRNAs were designed for targeting FveMyb10 (Figure 6A). Ten individual fruits were injected, and the GFP signal was observed 7 days after injection (Figure 6B, upper panel). Compared to control fruit (Figure 6B, EV), less anthocyanin was accumulated in the injected fruit (Figure 6B, down panel). Moreover, mutations could be identified in FveMyb10 at both the sgRNA1 and sgRNA2 target sites in all the injected fruit (Figure 6C and Supplementary Figure S6). These results indicated that pKSE401G is effective in strawberry by transient transformation.

We further assessed the genome editing efficiency of pKSE401G in soybean. Two sgRNAs were designed for GmNFR1a (Nod Factor Receptor 1a) (Figure 7A), encoding a LysM receptor-like kinase required for nodulation in the roots of legumes (Indrasumunar et al., 2011). The GFP fluorescence was visualized in the roots after being transfected by agrobacteria (Figure 7B, left panel). In contrast to wild type, the nodules were completely inhibited (Figure 7B, right panel and Supplementary Figure S7). The genotyping result showed that GmNFR1a was mutated at both the sgRNA1 and sgRNA2 target sites (Figure 7C and Supplementary Figure S8). The results in strawberry and soybean indicated that pKSE401G could sufficiently generate mutations in a wide range of plant species.

Low-frequency cases of off-target cleavage have been reported for CRISPR/Cas9 in plant research (Zhang et al., 2014; Pan et al., 2016). To detect the off-target events, potential off-target loci that are highly homologous to sgRNAs of RPK1 and BnaARF2 were predicted by the online tool CRISPR-P² (Lei et al., 2014). Previous reports revealed that the 12 nucleotides adjoining the PAM, designated as “seed sequence,” are critical for recognition specificity and cleavage efficiency of Cas9 (Xie and Yang, 2013; Zhang et al., 2014; Yan M. et al., 2015). The off-target sequences usually bear 1–3 mismatches in the ‘seed sequence.’ At least three of the most likely off-target sites for each sgRNA were examined in total 50 randomly selected plants by using gene specific primers (Supplementary Table S6). No mutations were found in the putative off-target sites (Table 3), indicating that sequence editing induced by the usage of pKSE401G is highly specific.

The B. napus variety Westar; Arabidopsis variety Col; the strawberry variety Ruegen (Ru F7-4, red-fruited) (Slovin et al., 2009), and the soybean variety Williams 82 (W82) were used as wild-types in this study. The plants were cultivated in a growth room under a light intensity of 100 μmol m^-2 s^-1 with a 16/8 h light/dark photoperiod at 22°C (B. napus, Arabidopsis and strawberry) or 25°C (soybean).

For Arabidopsis transformation, the CRISPR/Cas9 constructs were transformed into Arabidopsis wild-type Col-0 through floral dipping (Clough and Bent, 1998). The procedure of Agrobacterium-mediated B. napus transformation was carried out as previously described (Zhang et al., 2015). Briefly, the explants were incubated in the Agrobacterium-infection buffer (MS 4.43 g L^-1; sucrose 30 g L^-1; acetosyringone 100 mM L^-1; pH 5.8–5.9) for 20 min, then transferred to M1 medium plates (MS 4.43 g L^-1; sucrose 30 g L^-1; acetosyringone 100 mM L^-1; mannitol 18 g L^-1; 2,4-D 1 mg L^-1; kinetin 0.3 mg ml^-1; pH 5.8–5.9) and kept in dark for 48 h. Afterwards, the explants were transferred to M2 medium plates (MS 4.43 g L^-1; sucrose 30 g L^-1; acetosyringone 100 mM L^-1; mannitol 18 g L^-1; AgNO₃ 4 mg L^-1; 2,4-D 1 mg L^-1; kinetin 0.3 mg mL^-1; Timentin 270 mg L^-1; pH 5.8–5.9) with proper selection antibiotics to induce callus growth. The calli were transferred to M3 (MS 4.43 g L^-1; glucose 10 g L^-1; xylose 0.25 g L^-1; zeatin 2 mg L^-1; IAA 0.1 mg L^-1; Timentin 270 mg L^-1; pH5.8-5.9) and followed by M4 (MS 2.22 g L^-1; sucrose 10 g L^-1; IBA 0.5 mg L^-1; Timentin 135 mg L^-1; pH 5.8–5.9) medium to let shoot and root regenerate, respectively.

The strawberry fruit transient transformation assay was performed as described (Luo et al., 2018). Briefly, a single agrobacterium (GV3101) colony was picked and grown in 2 ml of liquid LB medium until OD₆₀₀ reached around 0.8 to 1.0. Then the culture was spun down and resuspended in the infiltration buffer (MS 4.43 g L^-1; sucrose 20 g L^-1) to reach an OD₆₀₀ of 0.8. Fruit at the white stage were used for injection. The color phenotype was examined 1 week after injection.

Agrobacterium rhizogenes (K599) mediated hairy root transformation was performed on soybean W82 as previously described (Kereszt et al., 2007). In brief, a single colony was picked and grown in 10 ml of liquid LB medium until OD₆₀₀ reached around 0.4 to 0.6. Then the culture was spun down and resuspended in the CCM buffer (MS 0.443 g L^-1; 3.9 g L^-1 morpholino ethanesulfonic acid; 150 mg L^-1 cysteine; and 150 mg L^-1 dithiothreitol) to reach an OD₆₀₀ of 0.2 to 0.3. Then the roots were incubated in the solution for 2–3 days. The phenotype was observed 4 weeks after transformation.

The CRISPR/Cas9 vector pKSE401G was modified from pKSE401 (Xing et al., 2014). The vector map was shown in Figure 1. The 35S-sGFP-terminator cassette was amplified from pK7GWIWG2D (Karimi et al., 2007) using primers 35S-GFP-Ter-F and 35S-GFP-Ter-R, and inserted into the PmeI site of pKSE401 by the Gibson assembly method (Gibson et al., 2009).

The GENE-sgRNA plant expression vectors were constructed as previously described with minor modifications (Xing et al., 2014). The target sgRNA sequences were designed using the web server CRISPR-P³ (Lei et al., 2014), and then the sequences were further analyzed by the software CRISPR Primer Designer (Yan M. et al., 2015). Using pCBC-DT1T2 as the template, two AtU6 promoter-sgRNA-AtU6 terminator cassettes were amplified by PCR using the primers listed in Supplementary Table S6. Then the PCR fragments were inserted into pKSE401G by Golden Gate Assembly (Gao et al., 2013), and confirmed by Sanger sequencing. These vectors were used for plant transformation.

The Arabidopsis and B. napus mutants were first screened by GFP fluorescence, which was examined using a dissecting microscope equipped with a GFP filter (Olympus, SZX16). Then, the candidates were further confirmed by Cas9 specific primers (Supplementary Table S6).

To analyze the mutations caused by CRISPR/Cas9, genomic DNA was extracted from leaves, fruit, or root using the CTAB method (Molecular Cloning, 3rd edition). The flanking sequences of the CRISPR target sites were amplified by PCR using gene-specific primers (Supplementary Table S6). Then, most of the amplicons were directly sequenced. To decode the mutations, the online tool DSDecode⁴ (Liu et al., 2015) was used for chromatogram decoding. Specifically, the sequence files (xxx.abi) and the reference gene sequences were uploaded to the server and analyzed using default settings. Subsequently, the results were aligned with the reference sequences to ensure that the mutations were in the sgRNA target sites. For complex mutations, the amplicons were first sub-cloned into the pGEM-T easy vector (A3600, Promega, United States), and about 10 clones for each amplicon were individually sequenced.

Stomatal assays were performed essentially as previously described (Desikan et al., 2002). Fully expanded leaves (3–4 weeks) were used for this assay. Abaxial epidermal strips from similar rosette leaves were first floated in 10 mM MES buffer (pH 6.15) containing 20 mM KCl for 2.5–3 h under white light (95 E m^-2 s^-1) to allow the stomata open. Afterwards, added 1 μM ABA to the buffer and incubated for another 1hr. The stoma images were taken by a ZEISS microscope equipped with a digital camera (AxioCam ICc5, Zeiss). Stomatal aperture was measured using Image J.

Total RNA was extracted using a Plant Total RNA Isolation Kit (Sangon Biotech, Shanghai, China, No. SK8631) following the manufacturer’s instructions. Approximately 1 μg of total RNA was used for cDNA synthesis using a PrimeScript^TM RT reagent kit (TaKaRa, Japan, Cat#RR047A). For qPCR, a total volume of 10 μl reaction mixture was used containing 5 μl of 2 × SYBR Green master mix (Cat# 172-5124, Bio-Rad), 0.5 μl of 5× diluted cDNA, 0.25 μl of each primer, and 4 μl ddH₂O. Amplification was performed using a CFX Connect^TM system (Bio-Rad, United States). The amplification program consisted of one cycle of 95°C for 5 min, followed by 50 cycles of 95°C for 15s, 60°C for 20 s, and 72°C for 20 s. The fluorescent product was detected at the third step of each cycle. The expression level of each gene was calculated using the 2^-ΔΔCT method (Livak and Schmittgen, 2001). All analyses were repeated three times using biological replicates. The ACTIN and GAPDH genes served as the internal control. All primers are listed in Supplementary Table S6.

The protein sequences of genes in Arabidopsis thaliana, B. napus, B. rapa, and B. oleracea were obtained from the website http://planttfdb.cbi.pku.edu.cn/index.php (Jin et al., 2017). The sequence alignment was performed using Clustal Omega⁵. An unrooted phylogenetic tree was constructed using MEGA7⁶(Kumar et al., 2016) with the neighbor-joining statistical method and bootstrap analysis (1000 replicates).