Rapid breeding of parthenocarpic tomato plants using CRISPR/Cas9

Parthenocarpy in horticultural crop plants is an important trait with agricultural value for various industrial purposes as well as direct eating quality. Here, we demonstrate a breeding strategy to generate parthenocarpic tomato plants using the CRISPR/Cas9 system. We optimized the CRISPR/Cas9 system to introduce somatic mutations effectively into SlIAA9—a key gene controlling parthenocarpy—with mutation rates of up to 100% in the T0 generation. Furthermore, analysis of off-target mutations using deep sequencing indicated that our customized gRNAs induced no additional mutations in the host genome. Regenerated mutants exhibited morphological changes in leaf shape and seedless fruit—a characteristic of parthenocarpic tomato. And the segregated next generation (T1) also showed a severe phenotype associated with the homozygous mutated genome. The system developed here could be applied to produce parthenocarpic tomato in a wide variety of cultivars, as well as other major horticultural crops, using this precise and rapid breeding technique.

time compared to classical techniques, it is important to develop and optimize highly efficient editing systems for important crop plants.
Here, we applied the CRISPR/Cas9 system to the production of parthenocarpic tomato plants via disruption of the tomato (Solanum lycopersicum) IAA9 (SlIAA9) gene in both Micro-Tom and the commercial cultivar Ailsa Craig. Introduction of the mutation was highly efficient, with almost 100% of genomic DNA isolated from tomato cells of the regenerated generation (T0) having the desired mutation, with no off-target mutations. Tomato plants with the SlIAA9 gene knocked out by our CRISPR/Cas9 system exhibited simple leaves instead of wild-type compound leaves, and fruit development was triggered before fertilization, giving rise to parthenocarpy. The mutations were heritable in subsequent generations. These results suggest that precise and effective genome editing can be used to guide molecular breeding in tomato.

CRISPR/Cas9 constructs and design of target gRNA.
To express gRNA targeted to SlIAA9 in Micro-Tom and Ailsa Craig (see Supplementary Fig. 1) in tomato plants, we constructed two new CRISPR/Cas9 vectors with different promoters for the expression of Cas9 (Fig. 1a). In both vectors, the Arabidopsis U6-26 promoter was used to express the gRNA constitutively, and either the 2 × CaMV35S promoter or parsley Ubi4-2 promoter was introduced to express an Arabidopsis codon-optimized Cas9 (AtCas9). Both these two vectors differ in codon usage, number of NLSs (three), and the type of U6 promoter used compared to the former vector pEgP526-2A-GFBSD2 26 (Supplementary Fig. 2). We designed three gRNAs (gRNA1, gRNA2, and gRNA3) with Tissues indicated by white circles (white dotted-circles in lower panels "GFP") were used for mutation analysis or plantlet-regeneration. (d) Heteroduplex mobility assay with the MultiNA electrophoresis system. Multiple heteroduplex peaks (red arrows) were detected in PCR amplicons from the CRISPR/Cas9 transgenic tomato calli, whereas a single peak was detected from the wild-type control (blue arrow). VC; vector control. M; internal marker. (e) PCR-RFLP analysis of the genome editing in Micro-Tom calli (RNA2-20b). +; Acc I digested PCR products, −; non-digested PCR products. Numbers show the regenerated T0 plant lines. (f) Determination of mutation ratios in transgenic calli and shoots. The mutation ratios were calculated by dividing number of mutant calli or shoots by the total number of transformed calli or shoots. a target sequence within the second exon of the SlIAA9 8 gene in the tomato genome (Fig. 1b) to introduce mutations in the SlIAA9 gene using the CRISPR/Cas9 system. Joung's group reported that 17 bp is the minimum size for a gRNA target sequence in this system (excluding the PAM sequence) 27 , and also showed that, whereas a 20-bp gRNA has a strong capacity to bind to genomic DNA with mismatches in non-target sequences (leading to off-target effects), 17-and 18-bp gRNAs [truncated-gRNA (tru-gRNA)] have high specificity for the target sequence with minimal cleavage activity 27 . Accordingly, we designed a 20-bp gRNA and a17-18 bp tru-gRNA for the target mutagenesis of SlIAA9. Target sequences were selected in silico using the gRNA design website "focas" 26 . In this website, the CasOT algorithm is used to find potential off-target sites, suggesting an off-target score, and an "on_target_score_calculator.py" algorithm is used to evaluate on-target activity 28 . Using these tools, we selected target sequences that scored high on on-target activity and low on off-target activity. In total, we selected five gRNAs in three sites (Fig. 1b) and inserted these into our CRISPR/Cas9 vectors, and the vectors were used to transform Micro-Tom. Transformed calli were selected using GFP fluorescence (Fig. 1c) and resistance to kanamycin. Transformed calli and shoots with strong GFP fluorescence and good cell growth were then analyzed for the CRISPR/Cas9-induced mutation ( Supplementary Fig. 9).

Detection of CRISPR/Cas9-induced mutations in tomato.
To detect mutations in the CRISPR/Cas9 transgenic tomato plants, we used HMA (heteroduplex mobility analysis) (Fig. 1d), PCR-RFLP (restriction fragment length polymorphism), (Fig. 1e and Supplementary Fig. 3), or Cel1-assay (data not shown) of a PCR amplification product of the SlIAA9 sequence including the target sequences ( Supplementary Fig. 9). A heteroduplex of mutant and wild type DNAs from the CRISPR/Cas9 transgenic Micro-Tom calli was detected in HMA as shifted peaks following electrophoresis (Fig. 1d). To determine the mutation rates in the transgenic tomato lines, we performed PCR-RFLP by digestion of the PCR product with the restriction enzyme Acc I, looking for uncleaved bands indicating the presence of mutated sequences (Fig. 1e). The representative results of PCR-RFLP analysis of Micro-Tom calli showed that somatic cell level mutation was detected in almost 100% of the transgenic calli induced by gRNA2 (20 bp) using both pEgP237-2A-GFP and pEgPubi4_237-2A-GFP vectors (e.g., tomato lines #4, #5, and #7 with pEgPubi4_237-2A-GFP vector, and #7 and #8 with pEgP237-2A-GFP) (Fig. 1e). Figure 1f shows the mutation frequencies indicated as the rates of mutated lines to those of CRISPR/Cas9 transgenic tomato lines analyzed by PCR-RFLP (as shown in Fig. 1e). gRNA1 did not induce the mutation in either calli or shoots (data not shown). The mutation rates in Micro-Tom calli were 11.1% for gRNA3 (17 bp) and 40.0-46.2% for gRNA2 (18 bp, 20 bp) and gRNA3 (19 bp) in the pEgP526-2A-GFBSD2 vector 26 ( Supplementary  Fig. 2). On the other hand, for gRNA2 (18 bp and 20 bp) and gRNA3 (17 bp), the mutation frequency was 66.7-77.8% in the pEgP237-2A-GFP and pEgPubi4_237-2A-GFP vectors. In Micro-Tom shoots, the highest mutation rates (100%) were detected using gRNA2 (20 bp) in both vectors (Fig. 1f). Thus, pEgP237-2A-GFP and pEgPu-bi4_237-2A-GFP vectors both enabled the introduction of high-efficiency mutations in tomato. gRNA2 (20 bp) induced the highest mutation efficiency of all the gRNAs tested at the somatic cell level, possibly representing bi-allelic mutation. gRNA2 (20 bp) was also transferred into Ailsa Craig that have the same target sequences in the SlIAA9 genome ( Supplementary Fig. 1).

Analysis of mutation pattern and off-target mutations.
PCR-RFLP detected a mutation rate of almost 100% at the somatic cell level induced by gRNA2 (20 bp). We next followed the mutation rates of these high-efficiency mutated shoots in all the regenerated transgenic tomato plants with mutations, and found a high level of mutation (30-90%) in T0 plants of both Micro-Tom and Ailsa Craig (Fig. 2a). To study these mutations in more detail, we then analyzed the mutated sequences in Micro-Tom using Sanger's method and next-generation sequencing (NGS) by Mi-seq (Fig. 2b,c). We analyzed the sequence of PCR products of 30-60 clones by the Sanger method. From analysis of the sequences derived from shoots of pEgPubi4-gRNA2-20b line #8, which had shown a mutation rate of almost 100% at the somatic level detected by PCR-RFLP (Fig. 1e), we found a plurality of base substitutions, intercalations, deletions, and long stretches of base deletions upstream of the PAM sequence in the cloned DNA (Fig. 2b). In contrast, upon analysis of the cloned DNAs, pEgP237-gRNA2-20b line #9 (Fig. 1e) showed only two types of mutation, at similar levels, suggesting the presence of bi-allelic mutation (Fig. 2b). Using Mi-seq, the amplicon sequence of the mature leaf of pEgPubi4-237-gRNA2-20b line #8 showed that the mutation rate at the somatic cell level was 97.2%, with base insertions and deletions detected 4-5 bp upstream of the PAM sequence (Fig. 2b,c). We also analyzed a mature leaf from tomato line pEgP237-gRNA2-20b #9, and showed that the mutation rate at the somatic cell level was 99.5% (Fig. 2b,c). Base insertion and deletions were detected upstream of the PAM sequence in both transgenic plants (Fig. 2b,c). These data suggest that both our CRISPR/ Cas9 vectors can effectively induce mutation with high frequency (near 100%) in regenerated T0 plants in tomato.
To compare the off-target effects obtained with the tru-gRNA compared to the normal 20 bp gRNA, we selected two sequences that have a high risk of off-target effects for each gRNA2. We used Mi-seq to analyze the mutation rate at the somatic cell level at both on-and off-target sites. gRNA2 off-target_1, with 4 mismatches to the gRNA2 on-target sequence located on chromosome 9 (chr9: 68093279-68093302), and gRNA2 off-target_2, with 5 mismatches to the on-target site (chr6: 26946923-26946946), were analyzed (Fig. 2d). The results revealed that no off-target mutations were detected at any of the off-target sites.
SlIAA9 knockout and knockdown have been associated with fruit development prior to pollination, or parthenocarpy, and abnormal leaf morphology (typically exhibiting a simple leaf) 8 . Simple leaves were observed in Micro-Tom CRISPR/Cas9 mutant lines of both pEgPubi4-gRNA2-20b and pEgP237-gRNA2-20b plants (Fig. 3b,c and Supplementary Fig. 5a). Several leaf samples were isolated randomly from mutant plants of pEgPu-bi4_237-gRNA2-20b T0 #8 and used for analysis of mutation efficiency by PCR-RFLP. The results showed a mutation rate of almost 100% at the somatic cell level in each leaf sample (Fig. 3d). Similarly, in other individual plants with simple leaves, mutated sequences were present at almost 100% at the somatic cell level (data not shown). The CRISPR/Cas9 mutants of Ailsa Craig (Fig. 3e and Supplementary Fig. 4a) also showed altered leaf morphology (Fig. 3f) and likely increased plant growth rates by observation of plants (Fig. 3g). We further analyzed fruit development and parthenocarpy in the Micro-Tom SlIAA9-crispr mutants (Fig. 3h-j). Accelerated enlargement of the fruit without self-pollination has been suggested to enhance the development of seedless fruits, and such parthenocarpic characters have been reported in other SlIAA9 mutants 8 . Seedless fruits were also observed in Micro-Tom SlIAA9-crispr mutants, whereas wild-type fruits set seed adequately (Fig. 3h,i). The parthenocarpic fruits were detected in the mutants, whereas no parthenocarpic fruits were observed in the wild type (Fig. 3j). Fruit morphology of the mutants was otherwise similar to that of wild-type tomato (Fig. 3c). To obtain successive generations of SlIAA9-crispr mutants effectively, artificial crossing of the SlIAA9 mutant and wild-type is an important process in yielding the seedless phenotype ( Supplementary Fig. 5b). We then analyzed the mutation efficiency in pollen, representing the haploid male gametophyte of Micro-Tom SlIAA9-crispr mutants, by PCR-RFLP. The results indicated a high mutation efficiency, suggesting that efficiency is sufficiently high to generate heterozygotes in the next generation ( Supplementary Fig. 5c-e). Furthermore, in the few collected seeds of SlIAA9-crispr mutants, a high mutation frequency was also detected (T1 seeds; Supplementary Fig. 6), and next generation plants showed the typical simple leaf phenotype upon germination (T1 seeds; Supplementary Fig. 6b). We further analyzed the fruit phenotypes and the expression levels of several marker genes in the mutants, and showed that the development of mutant fruits was similar to that of wild-type ( Supplementary Fig. 7). Together, these results suggest that our highly efficient CRISPR/Cas9 system for tomato was successful in generating almost 100% mutagenesis in the T0 generation, and of establishing heritable mutation effectively to obtain mutant tomato lines in successive generations.

Discussion
Current advances in new breeding techniques such as genome editing are now available also in crop plants. Here, we developed highly efficient genome editing using the CRISPR/Cas9 system in tomato, demonstrating for the first time the feasibility of this genome editing-based breeding technique to produce parthenocarpic tomato plants by disruption of the SlIAA9 gene. Wang et al. 8 reported that the SlIAA9 gene is expressed throughout leaf and fruit ontogeny ( Supplementary Fig. 8) and that downregulation of the IAA9 gene using antisense RNA caused parthenocarpy, leading them to suggest that the IAA9 protein prevents ovary development prior to pollination in wild-type tomato. In this study, we optimized the CRISPR/Cas9 system in tomato in terms of the promoters used for gRNA and Cas9 expression, the number of NLS in Cas9, and the length of gRNA target sequence, and established a set of expression cassettes yielding highly efficient site-directed mutagenesis on the specified target in both a model tomato and the commercial cultivar Ailsa Craig (Figs 1 and 3). Our work has shown that several optimizations of the basal cassettes in CRISPR/Cas9 enhanced the mutation efficiency in both tomato cultivars dramatically. While it is difficult to ascribe the improvement to a specific or multiple elements in the vectors from our current data, the enhancement can be attributed to the new vectors as a whole. Our results suggest that the CRISPR/Cas9 system is indeed one of the strongest tools available for gene editing to date, with more potential for further improvements in this effective system.
In transgenic plants regenerated from callus propagated on excised leaf disks, both gRNA and Cas9 expression need to be maintained at high levels during callus formation and propagation to ensure efficient mutagenesis. The off-target effects of the CRISPR/Cas9 system have been studied extensively, and it was found that the tolerance of the gRNA to multiple mismatches within the protospacer sequence also induced DSBs in some non-target genomic sequences [30][31][32][33] . Regeneration and micropropagation are important techniques for breeding of some useful crops, such as tree species whose seeds are difficult to produce. In such cases, a high frequency of off-target effects at the somatic level that would cause undesired mutations in the regenerated plants is one of the high risks during the generation of individual plants with newly induced on-target mutations. Therefore, the reduction of off-target effects is highly desired, especially in crop species. In this study, we designed highly specific gRNAs aimed at avoiding off-target effects; one such strategy was to use truncated gRNA (tru-gRNA) targets with 17-to 18-b target sequences. tru-gRNAs have been used in mammalian and Arabidopsis cells, and indeed have been shown to reduce the risk of off-target effects 26 . In tomato, our tru-gRNA targeting SlIAA9 introduced on-target mutations efficiently with no off-targets as judged by amplicon sequence analysis (Fig. 2). Interestingly, the 20 b gRNA2 target sequence in SlIAA9 also led to high on-target mutation with no off-target effects, suggesting that we had selected highly specific gRNAs. Further application of in silico tools, such as the on-target score calculator 26,28 , and improvements of the vectors, will continue to contribute the development of efficient genome editing systems without off-target mutations in various useful crop species.
We obtained bi-allelic and homozygous mutations in T0 regenerated plants of Micro-Tom as well as in the finest cultivar Ailsa Craig. Using this system, T0 plants with the desired phenotypes and 100% mutation in the target gene in general cultivars were established with high efficiency (Fig. 2). Mosaic mutations were also found in these cultivars; however, several lines exhibited almost 100% mutated target sequences at the somatic level (Fig. 2) in individual mutant plants with strong phenotypes, since no normal IAA9 protein molecules were generated in the knock-out mutants. The SlIAA9 CRISPR/Cas9 knock-out tomato exhibited the typical phenotypes of parthenocarpy, which processes fruiting without fertilization and leads to seedless fruits. In SlIAA9-crispr, seedless tomato fruits were generated; however, in a very few cases, small numbers of fertilized fruits developed a few seeds, which grew with phenotypes exhibiting the heritable mutation ( Supplementary Fig. 6). In vegetative tissues, SlIAA9-crispr exhibited the leaf phenotype typical of other SlIAA9 mutants generated by chemical, physical, or RNAi induced mutation 8 . These abnormal leaf phenotypes do not negatively affect plant growth. New developments in delivery systems for CRISPR-Cas9 in plant cells 34,35 using our efficient system would also advance the further application of parthenocarpy and other important traits, such as stress tolerance, in various useful crops in future. Agrobacterium-mediated leaf disk method (Supplementary Fig. 9). Briefly, cut cotyledon pieces (7-10 days after seeding) were transformed with A. tumefaciens strain GV2260 harboring the CRISPR/Cas9 plasmid. Transgenic calli were selected on MS medium containing 0.1 μg/mL blasticidin S or 100 μg/mL kanamycin, and by observing GFP fluorescence. Around 20-40 leaf disks transformed with CRISPR/Cas9 were cultured in each individual experiment. The three independent experiments were performed. We finally regenerated around 40 CRISPR/ Cas9 mutant plants that showed similar parthenocarpic phenotypes.

Design of gRNA and CRISPR/Cas9 vectors.
To design target gene sequences with low off-target effects in the second exon of SlIAA9 genome DNA (Solyc04g076850) (Supplementary Fig. 1 and Supplementary Fig. 8), we used the web-tool "focas" (http://focas.ayanel.com) (Osakabe et al. 26 ) and analysis using Cas-OT software (Xiao et al., 2014). Three candidate sequences of tru-gRNA targets (17-18 bp) were selected and named gRNA1-17b, gRNA2-18b, and gRNA3-17b. An additional 2 bp was added to each tru-gRNA to target 20 bp sequences at the sites for gRNA2 and gRNA3; these additional gRNAs were named gRNA2-20b and gRNA3-20b. At the 5′-end of gRNA3-20b, a G that does not exist in the wild-type genome was added to its 19-bp sequence to allow efficient expression from the U6 promoter (Fig. 1c). Stop codons that would be generated by a frame shift were located downstream of each target sequence. The annealed oligos for gRNAs were transferred into CRISPR/Cas9 vectors using Golden Gate Cloning methods. In pEgP526-2A-GFBSD2 26 , a gRNA cassette harboring the Arabidopsis U6 snRNA-1 (U6-1) promoter, the fungal and plant codon-optimized Streptococcus pyogenes Cas9 (fcoCas9), and GFBSD2 was placed under the control of the 2 × CaMV35S promoter. GFBSD2 was fused to fcoCas9 via a self-cleaving 2A peptide derived from Thosea asigna, to monitor the GFP fluorescence used to select Cas9-expressing plant cells. In pEgP237-2A-GFP, gRNA under the Arabidopsis U6 snRNA-26 (U6-26) promoter, a Arabidopsis codon-optimized S. pyogenes Cas9 (AtCas9) and GFP were driven by the 2 × CaMV35S promoter. For pEgPubi4_237-2A-GFP, the U6-26 promoter was used for gRNA expression, and a parsley ubi4-2 promoter was used for Cas9 expression.

Detection of CRISPR/Cas9-induced mutation.
To detect mutations in the target sequences, genomic DNA was isolated from the transformed calli, regenerated shoots, leaf tissues, fruits, and pollen of plants selected by GFP fluorescence and chemical tolerance. To evaluate mutations introduced in the CRISPR/Cas9 transgenic plants, a region of about 300 bp surrounding the target locus of gRNA was amplified by PCR. To detect mutations in the PCR product, HMA using MCE202 MultiNA (Shimadzu, Japan), Cel-1 assay, or PCR-RFLP was first used. In HMA, the PCR fragments were analyzed directly using a microchip electrophoresis system with MCE202 MultiNA. In the Cel-1 assay, the PCR products for gRNA1 and gRNA3 were digested with Guide-it ™ Mutation Detection Kit (Takara, Japan) or Surveyor ® Mutation Detection Kits (Transgenomic, USA). In PCR-RFLP, the PCR products for gRNA2 were digested with Acc I. The mutated and the wild-type DNA fragments were separated by agarose-gel electrophoresis.
Sequencing analyses. In Sanger's sequencing method, the PCR product was cloned into a plasmid T-Vector pMD 20 (Takara, Japan). The resulting plasmids were used to transform E. coli strain DH5α, about 30-50 white colonies were selected from which to extract and sequence the plasmid DNA containing the mutation. In next-generation sequencing by Mi-seq (illumina, Japan), genomic DNA including the mutation was used to amplify the region including gRNA2 and gRNA3 target sequences by PCR. Similarly, two selected sequences with high risk of off-target effects from off-target candidate sequences of gRNA2 (chr9: 68093279-68093302 and chr6: 26946923-26946946) or gRNA3 (chr12: 23706808-23706831 and chr3: 63407250-63407273) selected by Cas-OT were examined. First PCR products were extracted from the gel using the Wizard ® SV Gel and PCR Clean-Up System (Promega, Japan) and used as templates for a second round of PCR. Second PCR primers were subjected to Truseq (illumina, Japan). Mi-seq data (about 11000-30000 reads) was analyzed using CLC Genomics Workbench software version 7.5.1 (CLC bio, Japan), mapped on the SlIAA9 sequence using Integrative Genomics Viewer (IGV) (Broad Institute) and graphing of deletion, insertion and substitution on the target sequence was generated using Microsoft Excel and CRISPResso (http://crispresso.rocks).
Phenotypical characterization of CRISPR/Cas9-induced SlIAA9 mutant. Plant phenotypes (leaves and fruits) in the CRISPR/Cas9 SlIAA9 mutants with highly mutated sequences were evaluated. To evaluate tomato parthenocarpy, seed numbers and formation were measured in the cut fruits. The average seed numbers was calculated in 5-18 fruits from the individual plant lines. Number of the parthenocarpic fruits in three plants of each construct and WT was counted and the average numbers were calculated.
qRT-PCR analysis. Total RNA was extracted using RNAiso Plus (Takara, Japan) from buds, flowers, mature green fruits, and red fruits. Reverse transcription was performed using the purified RNA with SuperScript III and Random hexamer primer (Invitrogen, Carlsbad, USA). qRT-PCR was performed in a LightCycler system (Roche Molecular Biochemicals, Germany) using SYBR Premix Ex Taq (Takara, Japan). The qRT-PCR primers are listed in Supplementary Table 1. Other experimental conditions were as previously described 36,37 .