Fast and accurate gene editing in carrot cells using CRISPR technology

We used a carrot callus line characterized by an intensive purple color due to anthocyanin accumulation. White callus line from which the purple callus was derived was also used as the reference for anthocyanin content analysis. Callus was maintained in Petri dishes with a phytagel solidified Gamborg B5 salts with vitamins medium (Duchefa, Haarlem, The Netherlands) supplemented with 1 mg/L 2,4-dichlorophenoxyacetic acid (2,4-D, Sigma, St. Louis, USA), 0.025 mg/L kinetin (Duchefa, Haarlem, The Netherlands), and 30 g/L sucrose; pH 5.8. Callus was cultured at 26 °C in the dark and was subcultured every 3 weeks to a fresh medium.

CRISPR/Cas9 vectors for multiplex gene targeting were developed according to Lowder et al. (2015) and then transferred to Agrobacterium tumefaciens strain LBA 4404 by electroporation (Main et al. 1995). Vector T-DNA contained at its right border a Cas9 gene driven by the 2 × 35S CaMV promoter followed by two single guide RNAs, each controlled by the AtU3 promoter. The left T-DNA fragment contained the hygromycin resistance aph gene with 35S CaMV promoter. Three Cas9 entry clones (pYPQ154, pYPQ166, and pYPQ167) containing three codon-optimization variants of SpCas9 genes were used: AteCas9 (Fauser et al. 2014; Schiml et al. 2014), zCas9 (Xing et al. 2014) and Cas9p (Ma et al. 2015). Two gRNAs, gRNA3—ATTAGAGCCCGGGACTACT and gRNA4—AAGTTTTGTCAGAGGCCAT, were designed to target the second exon of the F3H gene at sites adjacent to CGG-PAM and GGG-PAM, respectively, based on the sequence of the reference F3H gene (NCBI Acc. no. XM_017385173↗) deposited within a doubled haploid DH1 carrot genome sequencing project (Iorizzo et al. 2016).

Agrobacterium tumefaciens was growing in lysogeny broth supplemented with 50 mg/L of kanamycin at 26 °C and using a rotary shaker with 260 rpm. An overnight culture was centrifuged at 6000 rpm for 10 min. The pellet was resuspended in callus culture liquid medium with addition of 100 µM acetosyringone (Sigma, St. Louis, USA) and diluted to OD₆₀₀ = 0.5. Two-week-old callus clumps of about 5 mm diameter were placed in a Petri dish and immersed in A. tumefaciens inoculum for 20 min. Then, clumps were blotted on a sterile filter paper and placed on a solidified culture medium, followed by 3-day co-cultivation at 26 °C in the dark. Subsequently, for Agrobacterium elimination, calli were rinsed with sterile water containing 800 mg/L cefotaxime and 400 mg/L timentin, and were transferred to a fresh callus culture medium supplemented with 400 mg/L cefotaxime and 200 mg/L timentin. After 4 weeks, callus clumps were transferred and spread on a fresh medium supplemented with 25 mg/L hygromycin. Single, small new callus clumps developing after a 4-week selection were picked up and transferred periodically to the same selection medium until the end of the experiments.

Genomic DNA was extracted from callus using a CTAB method described by Rogers and Bendich (1988) with modifications. Small fragments of transgenic callus were placed in 2 ml eppendorf tubes with 100 µl of CTAB buffer and two 3 mm beads, and then grinded in the Retsch Mixer Mill MM400 (Retsch GmbH, Haan, Germany) for 3 min at room temperature. PCR was performed using the Eppendorf Mastercycler thermocycler (Eppendorf, Hamburg, Germany) and primers listed in Table 1, and the reaction was set up in a volume of 10 µl containing 0.1 µM of each primer, and 2 × buffer including Taq polymerase and dNTPs (PCR Mix Plus, A&A Biotechnology, Gdynia, Poland). The PCR conditions were as follows: the initial denaturation at 94 °C for 4 min, 35 cycles of 45 s at 94 °C, 30 s at 56 °C or 60 °C (depending on primers used), and 60 s at 72 °C, and the final extension for 5 min at 72 °C. PCR products amplified with F3H primers were 100 × diluted and used as templates for the next PCR round of a nested PCR. Final PCR products were digested with NcoI High-Fidelity enzyme (NEB R3193S, Ipswich, USA) 37 °C for 3 h. PCR products and restricted fragments were visualized using MidoriGreen Advance (Nippon Genetics) after electrophoresis in 1% agarose (Prona Plus, Conda, Spain) gel.

Undigested fragments were eluted from the gel and purified using the Promega Wizard^® SV gel and PCR clean-up system, cloned into pGEM plasmids (Promega, Madison, USA) after genetic transformation to E. coli, plasmids were isolated using GeneJET Plasmid Miniprep Kit (Thermo Scientific, Waltham, USA) according to the manufacturer’s instruction. Samples were subjected to Sanger sequencing with F3H_FI primer (Genomed, Warsaw, Poland). Reads were manually aligned to the reference sequence using the BioEdit software.

Anthocyanin content was determined using a modified pH differential method (Lee et al. 2005). Small callus clumps were homogenized in the extraction buffer (70% ethanol; v/v, acidified with 0.1% 1N HCl; v/v) and filled up to 1 ml. Homegenized tissue was vortexed and shaked for 15 min for complete anthocyanin extraction. To remove any cell debris, extracts were centrifuged for 10 min at 14,000 rpm. Acquired extracts were diluted in pH 1.0 (50 mM KCl) and pH 4.5 (400 mM sodium acetate) buffers and vortexed. Five independent extractions were performed for each callus line. Absorbance of each sample at 520 and 700 nm was measured using the NanoDrop™ 2000c spectrophotometer (Thermo Scientific, Waltham, USA). Anthocyanin content was calculated based on molar weight (449.39) and molar extinction coefficient (26,900) of cyanidin-3-glucoside. Results are presented in mg of cyanidin-3-glucoside (C3G) equivalent per 100 g of callus tissue.

Protoplasts from small callus clumps were isolated according to the protocol previously described (Grzebelus et al. 2012). Due to low amount of tissue available for isolation, volumes of used solutions were reduced eight times. Purified protoplasts were resuspended in 30 µl of the protoplast culture medium. A drop of medium containing protoplasts was placed on a microscopic slide and observed under bright-field microscope (Zeiss Axiovert S100, obj. 10 ×). White and purple protoplasts were counted in five vision fields. Proportions of calli and protoplasts differing in color were compared using statistical tests for significance level between two proportions and Chi-square, respectively.

The utilization of CRISPR/Cas9 for plant genome editing is very promising in both basic and applied research. This system, however, has not been used in carrot before. Therefore, we intended to validate the system as well as to develop a model that can be easily and efficiently used in further research enabling carrot genome editing. We first wanted to verify that CRISPR/Cas9 system is able to induce mutations in the carrot genome. For this purpose, we defined a model system that consisted of a carrot callus accumulating a high level of anthocyanins. We chose to compare three SpCas9 vectors differing in codon-optimization (Fig. 1a). We hypothesized that Cas9 can be used to induce mutations in the F3H gene leading to blockage of the anthocyanin biosynthesis pathway and resulting in discolored callus cells as compared to the original purple callus cells that contain functional F3H. To validate this approach, we designed two gRNAs targeting the F3H exon2 (Fig. 1b). After callus transformation, new hygromycin resistant calli had started to develop within 4 weeks of culture, and then, they were selected for further growth. The presence of introduced aph and Cas9 genes in these calli was confirmed by PCR (Supplementary Fig. S1), suggesting successful transformation.

The developing calli clearly exhibited color variation. Some calli retained deep purple color similar to a non-transformed wild-type (WT) calli, while others were completely white (Fig. 2a). Several calli had mosaic color with white, light purple, or dark purple parts (Fig. 2b). In general, the color of individual calli did not change during the whole culture process. In contrast, WT purple carrot calli grew continuously with deep purple color during the whole experiment when maintained on hygromycin-free medium, while on the selection medium, it did not grow, turned brown, and eventually died due to sensitivity to hygromycin.

Interestingly, we found the frequency of calli differing in color depended on the delivered vector constructs. Most calli developed after transformation with the Arabidopsis codon-optimized AteCas9 gene were discolored or showed considerably lighter pigmentation (Fig. 2a). Such phenotypes were observed for 89.5% of calli, and the frequency of completely white only calli was 49% (Fig. 2c). Similar frequencies were observed after transformation with the maize optimized-codon zCas9 gene. Neither the frequency of white callus nor the frequency of combined white and mosaic calli differed statistically from those observed for AteCas9 calli (p = 0.198 and 0.293, respectively). However, significantly lower frequency of completely white callus (14.8%, p < 0.01) was observed for transformants containing the plant codon-optimized Cas9p variant. As the frequency of mosaic callus was similar for all three Cas9 variants used, the combined frequency of white and mosaic callus using Cas9p (64.8%) was also lower than AteCas9 (p = 0.025) and zCas9 (p = 0.002).

Color differences in transgenic callus may likely result from either lower accumulation of anthocyanins in all cells or repressed accumulation in only some of them. To reveal reasons for observed variation, we measured anthocyanin content in callus and then screened these callus tissues at the cell level. The obtained calli were grouped to phenotypically different classes based on color, i.e., deep purple, completely white, and mosaic regardless, which Cas9 vector was used for transformation. All protoplasts released from WT white calli and observed under the microscope were transparent white, while protoplasts of WT purple calli were mainly purple, and those of intense hue dominated. Protoplasts with faint color were rare, though a small fraction of completely white protoplasts was observed (Table 2; Fig. 3a). All protoplasts released from white transgenic calli were completely free of pigmentation (Fig. 3b). Mosaic calli were heterogeneous and the mean frequency of purple protoplasts was only 25.8% (Fig. 3c). Hue of most transgenic purple protoplasts was usually less intense than that of protoplasts from WT purple calli, and differentiation between very faint and completely white protoplasts was not possible in this experiment. White protoplasts were also observed in transformed purple calli and their frequency (27%) was significantly higher (p < 0.001) than in WT purple calli (Table 2). Taken together, these discoloring phenotype results from protoplasts are consistent with discoloring phenotype of callus clumps that we examined earlier (Fig. 2).

We then tried to measure anthocyanin content to validate the discoloring phenotype which is due to loss of anthocyanin pigments. Purple WT calli contained high amounts of anthocyanins equivalent to 13.4 mg C3G per 100 g of fresh callus weight as determined spectrophotometrically (Fig. 3d). As a control, we used the white callus line from which the purple carrot line was derived. No anthocyanins were detected in cells of this white callus line. In transgenic white calli, no anthocyanins were detected, but in purple calli, the anthocyanin content was, on average, at the same level as in the purple WT calli (Fig. 3d). Mosaic calli had, on average, 78% less anthocyanins (2.9 mg C3G per 100 g FW) than WT purple calli that corresponded to a lower fraction (25.8%) of anthocyanin accumulating cells (93.9% in WT purple callus). Thus, the pigment content in calli is highly correlated with the frequency of purple protoplasts (r = − 0.89; p = 0.019).

These combined data indicate that callus discoloration after transformation with CRISPR/Cas9 vectors is, indeed, due to loss of anthocyanin accumulation, presumably due to targeted knockout of the F3H gene.

To validate that callus discoloration was due to the occurrence of intended mutations at target sites of F3H, we performed molecular genotyping analysis. It is known that Cas9 usually generates DSBs at the third-nucleotide proximal to PAM in the target sequence, which is the hotspot for mutation occurrence. In our design, the gRNA4 sequence contains a restriction site located proximal to PAM and recognized by the NcoI nuclease (Fig. 1b). We expected that successful targeting of F3H with gRNA4 should result in mutations that destroy the NcoI restriction site, which can be confirmed by the restriction fragment length polymorphism (RFLP) assay. We first performed a nested PCR to ensure correct amplification of the desired 538 bp-long F3H fragment (Fig. 1b) from carrot genome and then digested final PCR products with NcoI nuclease.

As expected, amplicons derived after PCR with DNA of WT callus were completely digested resulting in two fragments of 391 and 147 bp (Fig. 4). In contrast, digestion of PCR products from transgenic calli resulted in either two DNA fragments, uncleaved DNA or a mixture of both cleaved and uncleaved DNAs (Fig. 4). The presence of uncleaved products for transgenic calli indicated that NcoI recognition site was mutated. Samples where NcoI did not digest PCR products usually originated from white transgenic calli and most frequently represented AteCas9 transformed calli. Samples with complete digestion predominantly represented Cas9p calli, consistent with the phenotypic data (Fig. 2). Calli transformed with AteCas9 resulted in most completely undigested PCR products. Although zCas9-expressing calli resulted in similar numbers of completely undigested products to Cas9p-experssing calli, they contain more partly undigested products than Cas9p samples (Fig. 4). These results again suggest that AteCas9 is most efficient in carrot cells for generating mutations, followed by zCas9 and Cas9p. Independent on the Cas9 variant used, undigested products were also present in both purple and mosaic callus, indicating that at least some cells were mutated (e.g., samples 5 or 15 from the second gel in Fig. 4).

To confirm that the NcoI restriction site was, indeed, mutated, we sequenced products of the expected 538 bp length and those which had larger deletions. The results revealed mutations that destroy NcoI recognition site in calli transformed by all three Cas9 variants (Fig. 5). The most frequent changes were single-nucleotide insertions, mainly A or T, between the third- and fourth-nucleotide upstream the PAM. Short deletions up to 18 nt were the second most frequent type of mutations. In all calli, deletions usually occurred upstream PAM, but there were also deletions, terminating or eliminating the PAM sequence (Fig. 5).

It is well known that CRISPR/Cas9 may result in off-targeting at unintended sites which share sequence homology with the target site (Fu et al. 2013, Hsu et al. 2013). In this study, we used two gRNAs for simultaneous targeting of F3H. The second gRNA (i.e., gRNA3) was originally designed using the reference F3H gene of DH1 carrot. Sequencing of F3H fragments from WT purple callus used in our study revealed that gRNA3 had a A/G mismatch at the sixth-nucleotide upstream the PAM which is within a seed region crucial for Cas9/gRNA binding. Despite this mismatch, a quarter of sequences mutated at the NcoI site at the gRNA4 target had additional mutations at the gRNA3 target (Fig. 5). These mutations were detected with all three Cas9 versions. Given the relatively high off-targeting activity of gRNA3, these multiplexing Cas9 constructs should result in large deletions due to simultaneous cleavage at both gRNA3 and gRNA4 target sites. We, indeed, identified such large deletions in transformed calli (Fig. 6a). Sequencing of these products confirmed deletions of 116–119 nt in length (Fig. 6b). The deleted chromosomal region encompassed gRNA3 and gRNA4 target sequences and the whole fragment spanning between them. Two of these chromosomal deletions (− 119 nt) were canonical with three nucleotides proximal to PAM left at both target sites, representing perfecting ligation after Cas9 cleavage at both sites. These results suggest that CRISPR/Cas9 can tolerate 1 nt mismatch even in the seed sequence of the protospacer in carrot cells.

MK-Ch, RB, and YQ designed the experiment, MK-Ch, TO, and LGL performed the experiments. MK-Ch, TO and RB analyzed the results, MK-Ch, RB, and YQ wrote the manuscript. All authors read and approved the manuscript.

Below is the link to the electronic supplementary material.

The authors declare that they have no conflict interests.