Genome editing in plants using CRISPR type I-D nuclease

Genome editing in plants has advanced greatly by applying the clustered regularly interspaced short palindromic repeats (CRISPRs)-Cas system, especially CRISPR-Cas9. However, CRISPR type I—the most abundant CRISPR system in bacteria—has not been exploited for plant genome modification. In type I CRISPR-Cas systems, e.g., type I-E, Cas3 nucleases degrade the target DNA in mammals. Here, we present a type I-D (TiD) CRISPR-Cas genome editing system in plants. TiD lacks the Cas3 nuclease domain; instead, Cas10d is the functional nuclease in vivo. TiD was active in targeted mutagenesis of tomato genomic DNA. The mutations generated by TiD differed from those of CRISPR/Cas9; both bi-directional long-range deletions and short indels mutations were detected in tomato cells. Furthermore, TiD can be used to efficiently generate bi-allelic mutant plants in the first generation. These findings indicate that TiD is a unique CRISPR system that can be used for genome engineering in plants.


Results and discussion
TiD composed of Cas effector proteins with a Cas10d can be used for genome editing. The CRISPR/Cas TiD locus consisting of eight Cas genes (Cas1d-Cas7d, Cas10d) followed by an array of repeat-spacer units, was identified from Microcystis aeruginosa 1,2 .
The typical Cas8 gene-the common effector in CRISPR type I-A, B, C, E, and F 2,16 -is missing from the CRISPR/Cas TiD locus of M. aeruginosa, predicting different mechanisms of cascade complex stability and in vivo DNA cleavage activity in TiD compared with other type I sub-types (Fig. 1a). To identify the PAM in the TiD system in M. aeruginosa, we performed a depletion assay using the negative selection marker ccdB. pCmMa567d10 ( Supplementary Fig. 7), containing expression cassettes for Cas5d, Cas6d, Cas7d and Cas10d carrying a mutation in the HD-like domain [dCas10d (H177A)] and gRNAs targeted to the ccdB promoter, was introduced into Escherichia coli strain BL21-AI followed by a PAM library plasmid, pPAMlib-ccdB (Fig. 1b, Supplementary Fig. 9). ccdB negative selection revealed the PAM: 5′-GTH-3′ (H = A or C or T) adjacent to the target sequence ( Fig. 1b lower panel, Supplementary Data 1 and 2). When pCmMa567 ( Supplementary Fig. 8), which carries expression cassettes for Cas5d, Cas6d and Cas7d, was used for screening instead of pCmMa567d10, GTH PAMs were screened out, but the resulting transformants were unstable, and growth of the E. coli cells was very weak. These results suggested that Cas10d requires the correct PAM for full repression, and that Cas10d is a functional counterpart of Cas8 for PAM recognition and stabilization. We did not find any similar amino acid sequences shared between the Cas10d and Cas8 protein families.
To detect the genome editing activity of the TiD and Cas10d nuclease function, we then performed the luciferase single-strand annealing (SSA) recombination system using HEK293T cells (Fig. 1c). This system consists of NanoLuc luciferase containing 300 bp homology arms separated by a stop codon and a target gene fragment. First, the human AAVS1 gene fragment containing the TiD target site (Supplementary Table S2) was used to evaluate the TiD complex using the 35-bp crRNA spacer sequence. In this assay, HEK293T cells were transfected simultaneously with TiD Cas effectors, gRNA, and NanoLuc interrupted with target gene fragment and firefly luciferase expression vectors, and then the luc reporter assay was carried out 72 hr after transfection. Deletion of either Cas3d or Cas10d abolished TiD genome editing activity in the luc reporter assay (Fig. 1d, Supplementary Data 1), suggesting that both Cas3d or Cas10d have essential roles in genome editing activity. In the original CRISPR locus of M. aeruginosa strain PCC9808, both 35bp and 36-bp spacer sequences are used to target specific genomic DNAs; both spacers function in genome editing in human cells (Fig. 1e, Supplementary Data 1). To evaluate the genome editing activity of TiD for plant genes, we next performed the luc assay for several target rice and tomato gene sequences, SlIAA9 (important in parthenocarpy) 17 and NADK2 (OsNADK2) (Fig. 1f, Supplementary Fig. 1b, Supplementary Data 1). The results showed that there were several targets with GTT or GTC PAM with higher activity in the luc assay than GTA PAM.
Targeted mutagenesis by TiD in plants. Next, we constructed TiD expression vectors with plant-cell-specific-promoters for expression of codon-optimized Cas genes and gRNA were employed to induce site-directed mutagenesis in tomato plant: pTiDP1.2, an all-in-one vector, harboring a single CaMV35S promoter driving all 5 ORFs of Cas effectors separated by 2A selfcleaving peptide, and pMGTiD20, in which two expression cassettes under two promoters (CaMV35S and Parsley UBIQUITIN 4-2) are used to express Cas effector genes ( Supplementary  Fig. 1a). The separated cassettes in pMGTiD20 were designed to eliminate decreasing C-terminal expression levels in the long single cassette for multiple ORFs in pTiDP1.2. gRNAs targeted a 35-bp sequence in the tomato genes, SlIAA9 and RIN (SlRIN; involved in fruit ripening) 18 . For the SlIAA9 gene, the selected gRNAs in the luc reporter assay, GTT_gRNA5-A(-) and GTC_gRNA1(+) (Fig. 1f), both a single gRNA for GTC_gRNA1 (+) and multiplex gRNAs for GTT_gRNA5-A(-) and GTT_gRNA5-B(+) (Supplementary Table 1) were used for further analysis. The TiD vector, pTiDP1.2, containing the designed gRNAs, a single gRNA for GTC_gRNA1(+) or multiplex gRNAs for GTT_gRNA5-A(−) and GTT_gRNA5-B(+) were then transformed into the tomato cultivar Micro-Tom by Agrobacterium-mediated transformation, respectively. We analyzed TiD-induced mutation efficiency in transgenic tomato calli by Cel-1, PCR-RFLP using AccI 17 , and sequencing. In the T0 transgenic tomato calli and shoots, the small indels mutations were detected by these analyses (Fig. 2, Supplementary  Figs. 2, 3).
Cel-1 assay and sequence analysis of PCR products to determine small indels mutations induced by SlIAA9 GTC_gRNA1(+) revealed somatic mutation in 7/11 transgenic Micro-Tom calli (Fig. 2a, Supplementary Fig. 2a). We further analyzed the mutation efficiency in regenerated tomato shoots using PCR-RFLP for SlIAA9 GTC_gRNA1(+), identifying undigested bands with AccI in 14/15 transgenic Micro-Tom shoots (Fig. 2b, upper panel, Supplementary Fig. 2b). Together with the sequence analysis, the results indicated that these transgenic shoots contained 100% mutated DNA (Fig. 2e, Fig. 1 Genome editing activity of CRISPR type I-D detected by luc reporter assay. a The CRISPR type I-D (TiD) structure. Upper; subunit organization of TiD and schematic of gRNA (black) of TiD. Middle; schematic of gRNA (blue) of TiD and target DNA (black). The PAM of the target is shown in red. b PAM identification by the E. coli negative selection screening using ccdB expression system. PAM library was inserted in front of the target sequence of ccdB promoter. PAM frequency was determined using the survived E. colli cells. Data are means ± S.E. of independent experiments (n = 3). c Scheme of the luciferase reporter assay used to detect genome editing in human HEK293T cells. The Cas expression vectors and a LUC reporter vector, in which the target sequence was introduced, were transfected into HEK293T cells, and endonuclease cleavage was detected by luminescence. d Luc reporter assay showed both Cas3d and Cas10d were required for the TiD activity. Black bar; non-target gRNA in the luc reporter assay. gRNAs were target to the human AAVS locus listed in Supplementary Table 1. Data are means ± S.E. of independent experiments (n = 4). *P < 0.05 and **P < 0.01 are determined by Student's t tests. e, Effect of the gRNA target sequence length in the TiD activity. Data are means ± S.E. of independent experiments (n = 4). *P < 0.05, **P < 0.01, and ***P < 0.005 are determined by Student's t tests. f Luc reporter assay to determine targets for the SlIAA9 gene. gRNAs were target to the tomato IAA9 gene (SlIAA9) listed in Supplementary Table 1. Data are means ± S.E. of independent experiments (n = 4) and *P < 0.05, **P < 0.01, and ***P < 0.005 are determined by Student's t tests. Supplementary Fig. 3a). Thus, mutation rates were increased in transgenic tomato shoots during regeneration from Micro-Tom calli, and sequence analysis of cloned PCR products from the shoot target DNA revealed bi-allelic mutations. Homozygous mutants were effectively isolated in the T1 generation (Fig. 2b, upper panel, 2e, Supplementary Fig. 3a SlIAA9-tid gRNA(+) MT T1_#A-1, #A-2, #B-1). Bi-allelic mutants were also generated using the commercial tomato cultivar Ailsa Craig, as indicated by PCR-RFLP and sequencing analyses from clone-based sequencing and MiSeq next-generation sequencing (Fig. 2b, lower panel, 2e, 2f). Mature bi-allelic tomato plants exhibited clear typical SlIAA9 disruption phenotypes, such as parthenocarpy (fertility without seeds) and changes in leaf morphology 17 (Ailsa Craig; Fig. 2c, d, Micro-Tom; Supplementary Fig. 3b). Together with the SlIAA9 experiments, TiD induced small indels at the target site (Fig. 2). In the previous study by Ueta et al., the CRISPR/Cas9 that targeted the SlIAA9 exon2located very near the site of the SlIAA9 GTC_gRNA1(+)induced biallelic mutations in tomato calli 17 . When comparing the mutation frequencies of CRISPR/Cas9 and TiD in calli, the TiD activities were slightly lower than those of Cas9 (63.6% for TiD and 73.0% for Cas9) 17 ; however, the TiD SlIAA9 GTC_gRNA1 (+) could not induce biallelic mutations in calli (Fig. 1a, Supplementary Fig. 2a). Thus, TiD activity in inducing somatic mutations in calli was lower than that of Cas9. On the contrary, in the shoot samples, TiD could induce biallelic mutations at target sites with efficiency levels similar to those of Cas9 (Fig. 2b, 17 . Together, these results suggest there might be tissue specificity in the mechanism of TiD-mediated mutagenesis. Further analyses of other targets will be required to test this hypothesis; for example, detecting mutation patterns in cell lineages during shoot regeneration, and investigating tissuespecific mutagenesis might provide clues to further improvement of the TiD system in plant genome editing. Detection of the long-range deletion mutations by TiD in plants. Type I-E CRISPR-Cas can induce long-range deletion at target sites in the mammalian genome [13][14][15] . To detect TiD activity in a plant genome, we performed long-range PCR on TiD transgenic tomato calli. Long-range PCR was performed using specific primers located around 2-4 kbp upstream and downstream of the target sequence, respectively (Fig. 3, upper panel, Supplementary Table 4). Figure 3 showed that the several types of long-range deletion induced by SlIAA9 GTC_gRNA1(+) were detected in transgenic calli by PCR, and sequencing of cloned DNA identified bi-directional deletion (Δ2463 nt) from the mixed PCR product, with a mutation rate of 6.7% (1/15 sequencing clones) in one callus line (#5; Fig. 3, upper-left panel lane 5, Supplementary Fig. 6). Using SlIAA9 GTT+GTT_gRNA5-(-)(+), specific deletion bands were detected by the nested PCR in 1/20 and 1/30 transgenic calli, respectively (#3; Fig. 3, lower-left panel lane 3, Supplementary Fig. 6). Sequence analysis showed the same 100% mutated fragments in these clones, with bi-directional deletions of Δ4305 nt (Fig. 3). Interestingly, these results indicated that the deletion mutations generated by TiD in tomato genome were bi-directional, which, together with the generation of small indels mutations by TiD, is the unique feature of TiD that differs from mutation by type I-E [13][14][15] , although recent work suggests that Cas9 induced rare complex large deletions in addition to the desired small indels in mouse ES cells, and that these large deletions were bi-directional, similar to TiD but with lower frequency 19 . Furthermore, microhomology and insertions were observed in TiD mutation sites in long-range deletion mutations (Fig. 3), suggesting that specific DNA repair pathways function in these mutations.
Next, we tried to generate a CRISPR TiD targeting another locus in the tomato genome, the SlRIN gene, using pMGTiD20 vector, and analyzed mutations in the transgenic calli and regenerated shoots (Fig. 4). Long-range PCR was performed using specific primers located around 3 kbp upstream and downstream of the target sequence, respectively (Fig. 4a, Supplementary Fig. 6) and the sequence analysis showed the same 100% mutated fragments in these clones, with bi-directional deletions of Δ4930 nt (Fig. 4a). In the analysis for the regenerated shoots, 4 regenerated shoot lines from 12 individual transgenic shoots exhibited specific bands by long-range PCR (Fig. 4b, left panel, Supplementary Fig. 6). Interestingly, similar band patterns were detected in the individual shoot lines #4, #5 and #12 (Fig. 4b, left panel, Supplementary Fig. 6) and sequence analysis of the cloned PCR products indicated two types of long-range deletions (Δ4930 nt and Δ7257 nt); on the other hand, a single type of long-range deletions of Δ7257 was found in line #6 (Fig. 4c, d). The results indicated that the forward PCR primers annealed to homologous sequences 4.6 kbp upstream of the target sequence and could detect the Δ7257 mutation in these lines. Together, these results suggest that the bi-allelic mutations were effectively induced by the CRISPR TiD in tomato shoots, and the mature mutant plants for RIN were effectively obtained in the first generation (T0) by CRISPR TiD (Fig. 4b, right panel). From these results, we can see that the small indels were not detected in these loci using SlIAA9 GTT+GTT_gRNA5-(-)(+) and SlRIN GTC_4003-4238(+), indicating that varied mutation patterns were induced by each gRNA in the CRISPR TiD system.
Off-target effects generated by TiD in the plant genome. We next analyzed TiD off-target effects in plant genome. The TiD targets that has the 5′-GTH -3′ PAM in the whole genome of Arabidopsis and rice, and entire region of tomato chromosome 4 and 5, and each SlIAA9 and SlRIN gene were counted and compared to those of Cas9 (5′-NGG -3′ PAM) (Fig. 5a, b,  Supplementary Fig. 4, on-target). In this analysis, tomato chromosomes were selected as being representative of the tomato whole genome. The results indicate the more target sites for TiD The PCR amplified fragments separated on agarose gels indicate CRISPR TiD induced long-range deletions at the tomato RIN locus in the mutant shoots (Micro-Tom, T0 generation). WT; wild-type, 1-12; the transgenic shoot lines. The large deletions were detected in the lines #4, 5, 6 and 12. The bands with the same length as those of wild-type were non-specific bands. Arrows indicate the specific bands that were subjected to further sequencing analyses; red arrows indicate the fragment1 and blue arrows indicate the fragment2 as shown in c. Young mutant shoots for tomato RIN (Right; Micro-Tom, T0 generation #6) generated by CRISPR TiD. Bar = 1 cm. c Sanger sequencing using the cloned DNA from the CRISPR TiD transgenic tomato shoots (T0; #4, 5, 6, and 12) indicated the large deletion mutations occurred identically, however, the mutation frequencies were varied in the lines. d The mutation sequences of the cloned DNA from a. The nucleotide positions from the PAM were indicated on the sequence. exist in both the target genes and chromosome levels in tomato and Arabidopsis than those for Cas9. On the contrary, the rice genome has more Cas9 targets than those of TiD, this might result from the higher GC content in the rice genome and Cas9 PAM than in other species and the TiD PAM. Furthermore, when the off-target candidate sequences which contain 0 to 5 mismatches were also counted in tomato whole genome for each SlIAA9 and SlRIN gene, in Arabidopsis and rice whole genomes, and in the representative chromosomes of tomato for the ontargets in the same chromosome, respectively, there are less mismatch sequences for TiD than those for Cas9 (Fig. 5a, b,  Supplementary Fig. 4, mismatch numbers 0-5). In rice chromosomes, the decreasing tendency of TiD targets compared with Cas9 was clearer in off-targets. These data show that there are less off-target sequences for TiD in plant genomes, suggesting a TiD advantage in plant genome editing.
Although the gRNA target sequences used in this study, SlIAA9 GTC_gRNA1(+), SlIAA9 GTT + GTT_gRNA5-(-)(+), and SlRIN GTC_4003-4238(+), do not have highly similar sequences, and have fewer mismatches in the tomato genome, we next evaluated the off-target mutations in the T0 generation of tomato plants exhibiting clear SlIAA9-gene knock-out phenotypes; Three potential off-target sites for SlIAA9 GTC_gRNA1(+) with 9-11 mismatches, two potential off-target sites for SlIAA9 GTT + GTT_gRNA5-(-)(+) with 11 mismatches, and two potential offtarget sites for SlRIN GTC_4003-4238(+) with 6 and 7 mismatches, respectively, which are the sites with lowest mismatches for each on-target, were selected and further analyzed (Fig. 5c, d, Supplementary Table 5, Supplementary Figs. 5, 6). MiSeq analysis of PCR products around the potential off-target sites for SlIAA9 GTC_gRNA1(+) showed that there was little-tono off-target mutation in the T0 generation of tomato plants (Fig. 5c). Long-range nested PCR of the potential off-targets for SlIAA9 GTC_gRNA1(+) and SlIAA9 GTT + GTT_gRNA5-(-) (+) was also performed using specific primers located around 5-8 kbp upstream and downstream of the target sequence, and the results suggested there were no obvious effects ( Supplementary  Fig. 5a, b). Also, the off-target effects of long-range deletion mutations were evaluated for SlRIN GTC_4003-4238(+) in the T0 transgenic plants using specific primers located around 3 kbp upstream and downstream of the target sequence, respectively. As before, no off-target mutations were found in the T0 generation of tomato plants (Fig. 5d, Supplementary Fig. 6). The Cel-1 assay to evaluate small indels also showed no digested bands in the SlIAA9 GTT + GTT_gRNA5-(-)(+) and SlRIN GTC_4003-4238 (+) lines, indicating no mutations in these off-target sites ( Supplementary Fig. 5c). Together with a comprehensive analysis of many other on-targets for TiD, further work in vivo to evaluate off-target effects for fewer mismatches will be required in order to precisely elucidate the mechanisms of the TiD system when used in conjuction with the advanced unbiased technologies, i.e. CIRCLE-seq 20 and DISCOVER-seq 21 .
Conclusions. Although there are eight subtypes of CRISPR type I families identified from bacteria and archaea 2 , type I-D CRISPR-Cas, namely TiD, remains less well characterized. We showed that the CRISPR/Cas TiD locus from M. aeruginosa strain PCC9808 consists of eight Cas genes (Cas1d-Cas7d, Cas10d) followed by an array of 36 repeat-spacer units. In the TiD system, the HD domain, a functional DNA cleavage domain that has been identified in CRISPR type I-A, B, C, E, and F 7,22-26 , is lacking in Cas3d. Instead of the active Cas3 nuclease, TiD has Cas10d, which has an HD-like nuclease domain in the N-terminal region 1 . Interestingly, the Cas10d in TiD was highly divergent compared with Cas10s in the type III CRISPR-Cas family; instead, the Cas10d HD domain was similar to the Cas3 HD domains of type I-B, C, E, and F 1 . In the present study, we first developed a CRISPR TiD system as a genome editing tool for site-directed mutagenesis yielding both short indels and long-range deletion mutations in plant cells. Notably, the desired phenotypes in TiD transgenic mutated tomato were identified. Plant genomes, especially those of crop plants, have complex genome gene structures, with highly duplicated and redundant functional genes, as well as clusters of miRNA and non-coding RNA regions. The specific features of CRISPR type I-D CRISPR, which produces diverse and long-range deletion in genomic regions of interest, could be an effective genome editing tool kit with which to remove complex genome gene structures with low off-target effects. In this study, we used two types of TiD vectors, both of which induced mutations at their respective targets in the tomato genome. Further improvement of the TiD vector will still be important in developing this efficient tool. The unique TiDinduced mutation patterns suggest that the specific DNA cleavage mechanism and subsequent DNA repair pathway may differ from those of other genome editing tools. The diverse range of large deletions that can be generated from a single target site by TiD would enable long-range chromosome engineering; thus allowing expansion of the types of plant genome engineering that are possible using novel technologies in the CRISPR-Cas system.

Methods
Vector construction. All plasmid DNAs used in this study are shown in Supplementary Figs. 7 to 22. In the construction of plasmid DNAs, PCR amplification for cloning was carried out using PrimeSTAR Max (TaKaRa), cloning for assembling was performed using Quick ligation kit (NEB), NEBuilder HiFi DNA Assembly (NEB), and Multisite gateway Pro (Thermo Fisher Scientific).
Bacterial vectors. Gene fragments corresponding to E. coli codon-optimized Cas effector genes consisted of Cascade; Cas5d, Cas6d, Cas7d and dCas10d (H177A) the expression cassette for gRNA consists of the crRNA spacer corresponding to the target ccdB promoter sequence flanked on both sides by the 37-bp CRISPR repeat, were synthesized (gBlocks®) (IDT), assembled, and cloned into the pACYC184 vector (Nippon Gene) separately. Expression of Cas genes and gRNAs was driven by the T7 promoter. For gRNA expression, a DNA fragment containing a T7 promoter-repeat-spacer-repeat sequence was cloned into pACYC184. After confirming sequences of each gene expression cassette, Cas gene and gRNA cassettes were re-assembled into pACYC184 to yield pCmMa567 containing Cas5d, Cas6d, Cas7d, and gRNA expression cassettes, and to yield pCmMa567d10 containing Cas5d, Cas6d, Cas7d, dCas10d (H177A), and gRNA expression cassettes as shown in Supplementary Fig. 7. For PAM screening, the PAM screening reporter plasmid was constructed, assembling the sequence of the lacI gene, the lacI promoter to the 129th codon of the lacZ gene following the ccdB gene in pMW219 (Nippongene) to yield pPAM-ccdB. To generate a PAM screening reporter plasmid library pPAMlib-ccdB, the gene fragment corresponding to the T7 promoter, lac operator, and PAM with 4-nt randomized nucleotide was synthesized (IDT) and inserted in front of the lacZ-ccdB gene of pPAM-ccdB.
Luc reporter assay plasmids. The NanoLUxxUC expression vector was constructed for the luc reporter assay. First, DNA fragments of "NLUxxUC_Block1" and "NLUxxUC_Block2" were synthesized (IDT). "NLUxxUC_Block1" includes the 5′ end of the NanoLUC gene (351 bp) and Multi Cloning Site, and an XbaI site was attached to the 5′ end of the NanoLUC gene. "NLUxxUC_Block2" includes 465 bp of the 3′ end of the NanoLUC gene and an XhoI site was attached to the 3′ end. "NLUxxUC_Block1" and "NLUxxUC_Block2" fragments were then assembled and replaced with EGxxFP fragments in the pCAG-EGxxFP vector (Addgene; #50716).
Plasmid interference assay. Escherichia coli strain BL21-AI [F-ompT hsdSB (rB − mB − ) gal dcm araB:: T7RNAP-tetA] (Thermo Fisher Scientific) was used in the plasmid interference assay. E. coli cells harboring pCmMa567d10 were grown in LB medium supplemented with chloramphenicol (30 mg/ml). To calculate the library size of 4-nt PAM, pPAMlib-ccdB was introduced into E.coli cells harboring pCmMaTiD567d10, and the E.coli cells were plated onto LB agar medium supplemented with chloramphenicol (30 mg/mL) and kanamycin (25 mg/mL) and grown at 37°C overnight. The amount of pPAMlib-ccdB DNA used in the PAM screening experiment was decided as that leading to the formation of approximately 26,000 colonies, which is 100 times the theoretical-library size of the 4-nt PAM. The appropriate amount of pPAMlib-ccdB DNA was introduced into E. coli cells harboring pCmMaTiD567d10. After transformation, the E.coli cells were precultured in SOB liquid medium supplemented with 0.2% arabinose at 37°C for 2 h, then plated onto LB agar medium supplemented with 0.2% arabinose, 0.4 mM IPTG, chloramphenicol (30 mg/mL) and kanamycin (25 mg/mL) for growth at 37°C overnight. Plasmid DNA was extracted from surviving E. coli colonies, and the PAM sequence was amplified with adapters for Illumina sequencing using extracted plasmid DNAs as templates. The 4-nt PAM regions from 300-400 reads were analyzed with Miseq and counted manually.
Luc reporter assay. Human embryonic kidney cell line 293 T (HEK293T, RIKEN BRC) was used in luc reporter assay. Cells are cultured in Dulbecco's modified Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum (Thermo Fisher Scientific), GlutalMAX™ Supplement (Thermo Fisher Scientific), 100 units/ mL penicillin, and 100 μg/mL streptomycin in a 60 mm dish at 37°C with 5% CO 2 incubation. Cells (2.0 × 10 4 cells/well) were seeded onto 96-well plates (Corning) the day before transfection and transfected using TurboFect Transfection Reagent (Thermo Fisher Scientific) following the manufacturer's protocol. A total of 200 ng plasmid DNAs including (1) pGL4.53 vector encoding Fluc gene (Promega) used as an internal control, (2) pCAG-nLUxxUC vector interrupted with target DNA fragment (Supplementary Table 2), (3) plasmid DNAs encoding TiD components ( Supplementary Fig. 6, pEFs vectors), and (4) pAEX-hU6gRNA for the gRNA expression vector were used in each well of a 96-well plate. NanoLuc and Fluc luciferase activities were measured 3 days after transfection using the Nano-Glo® Dual-Luciferase® Reporter Assay System (Promega). The NanoLuc/Fluc ratio was calculated for each sample. The NanoLuc/Fluc ratio of the sample transfected with non-targeting gRNAs was used as the control and the relative activity was calculated for each sample to evaluate the gRNA activity. The experiments were repeated three to four times independently with similar results.
Plant transformation. Tomato plants (Solanum lycopersicum L.) cv. Micro-Tom and Ailsa Craig were used for site-directed mutagenesis experiments. Plants were grown under conditions of 24°C with 16 h light at 4000-6000 lx/8 h dark in the growth chamber. Transgenic tomato plants were generated using TiD vectors for plants. Leaf disks from tomato cotyledons were transformed with Agrobacterium tumefaciens strain GV2260 harboring the TiD vector. Transgenic calli and shoots were selected and regenerated to plantlets on MS medium containing 100 μg/mL kanamycin according to the method of Ueta et al. 17 .
DNA deletion analysis by long-range PCR. For mutation analysis by long-range PCR in tomato plants, genomic DNA were isolated independently from 20 T0 transgenic TiD tomato calli for SlIAA9 GTC_gRNA1(+) and GTT + GTT_gRNA5 (−) (+), respectively, and 30 T0 calli and 12 T0 shoots for SlRIN GTC_4003-4238 (+) using NucleoSpin ® Plant II (TaKaRa Bio). To analyze the large deletions, a region of about 5-6 kbp including the target site of each gRNA was amplified by first PCR for SlIAA9 GTC_gRNA1(+) and by nested-PCR for SlIAA9 GTT + GTT_gRNA5(−) (+) and SlRIN GTC_4003-4238(+) using PrimeSTAR GXL DNA Polymerase (TaKaRa Bio) and several kinds of primer sets for long-range PCR under the following conditions: 35 cycles of 10 s at 98°C, 15 s at 60°C, and 7 min at 68°C. The PCR products were analyzed by 1% agarose gel electrophoresis and stained with ethidium bromide. The first round PCR products for SlIAA9 GTC_gRNA1(+) were pooled and purified for cloning. The purified PCR products were cloned into pMD20-T vector using Mighty TA-cloning Kit (TaKaRa). Nested PCR for SlIAA9 GTT + GTT_gRNA5(−) (+) and the SlRIN GTC_4003-4238(+) transgenic calli was carried out, and only small DNA fragments separated in the agarose gel were extracted and purified from the gel for further analyses. For the SlRIN GTC_4003-4238(+) transgenic shoots, nested PCR was performed twice using the same primer sets and the small DNA fragments were also extracted after gel electrophoresis. The cloning of extracted fragments was carried out as mentioned above. Each cloned plasmid was analyzed by Sanger sequencing. The clone numbers for sequencing were varied for each sample as described in the results. All primers used for long-range PCRs used in the mutation analyses are listed in Supplementary Table 4.
Mutation analyses in short-range PCR products. To evaluate mutations introduced in transfected transgenic tomato calli and shoots, a region of about 400 bp surrounding the target locus of gRNA was amplified by short-range PCR using a PCR kit as described above. In the Cel-1 assay, PCR products from transgenic plants were digested using a Surveyor® Mutation Detection Kit (IDT). In PCR-RFLP, the PCR products from transgenic tomato plants were digested with AccI. Mutated and the wild-type DNA fragments were separated by 2-2.5% agarose gel electrophoresis and stained by GelRed (Biotium). PCR amplicons were also cloned into the TA cloning vector (TaKaRa Bio) to determine their sequences by the Sanger method. Amplicon deep sequences for on-and off-targets mutation analyses were performed using Multiplex identifiers-labeled PCR 17 . PCR products were subjected to Truseq on the MiSeq platform (Illumina). MiSeq data was analyzed using CLC Genomics Workbench software version 7.5.1 (CLC bio). All primers used for short-range PCRs used in the mutation analyses are listed in Supplementary Table 3.
In silico analysis for TiD target sites. Target sites of TiD and SpCas9 and the DNA sequences (on-target sequences) were mined in the tomato chromosome 4 and 5, and the whole genome of Arabidopsis and rice by an in-house Perl script, respectively. For each on-target sequence, the off-target sites with up to five mismatches were identified by using a tool Cas-OFFinder 28 . Then, the total numbers of on-target sites and off-target sites in the chromosome 4 were calculated. Similarly, on-target sites and sequences within the entire genomic regions of the SlIAA9 and RIN genes were detected by an in-house Perl script, respectively. For each ontarget sequence of the gene, all off-target sites were identified in the tomato genome sequence by Cas-OFFinder. The total numbers of on-target sites of the gene and off-target sites in the genome were obtained.