Functional disruption of cell wall invertase inhibitor by genome editing increases sugar content of tomato fruit without decrease fruit weight

Sugar content is one of the most important quality traits of tomato. Cell wall invertase promotes sucrose unloading in the fruit by maintaining a gradient of sucrose concentration between source leaves and fruits, while invertase inhibitor (INVINH) regulates this process. In this study, knock-out of cell wall INVINH in tomato (SlINVINH1) was performed by genome editing using, CRISPR/Cas9 and Target-AID technologies. Most of the genome-edited lines set higher soluble solid content (SSC) fruit than the original cultivar ‘Suzukoma’, while fruit weight was different among the genome-edited lines. From these genome-edited lines, three lines (193–3, 199–2, and 247–2), whose SSC was significantly higher than ‘Suzukoma’ and fruit weight were almost the same as the original cultivar, were selected. The fruit weight and overall plant growth of the two lines were comparable to those of the original cultivar. In contrast, the fructose and glucose contents in the mature fruits of the two lines were significantly higher than those of the original cultivar. The mature fruits of genome edited line 193–3 showed the highest sugar content, and the fructose and glucose contents were 29% and 36% higher than that of the original cultivar, respectively. Whole genome sequence data showed no off-target mutations in the genome-edited lines. Non-target metabolome analysis of mature fruits revealed that fructose was the highest loading factor in principal component analysis (PCA) between the genome-edited line and the original cultivar, and no unexpected metabolites appeared in the genome-edited line. In this study, we succeeded in producing tomato lines with high sugar content without a decrease in fruit weight and deterioration of plant growth by knock-out of SlINVINH1 using genome editing technology. This study showed that functional disruption of SlINVINH1 is an effective approach to produce tomato cultivars with high sugar content.


Results and discussions
Production of cell wall INVINH knock-out tomato. Two genes encoding cell wall INVINHs, SlIN-VINH1 (Solyc12g099200) and SlINVINH2 (Solyc12g099210), are present in the genome of tomato Solanum lycopersicum 'Heinz 1706' (Sol Genomics Network, https:// solge nomics. net/). SlINVINH1 and SlINVINH2 have high sequence homology with each other at both the nucleotide and amino acid levels, and both genes consist of two exons and one intron. SlINVINH1 and SlINVINH2 are tandemly present in the tomato genome, and the distance between them is only 3,809 bp (Fig. 1a). Therefore, these two genes may have appeared due to gene duplication. The expression level of SlINVINH2 shown by the Tomato eFP browser (http:// bar. utoro nto. ca/ efp2/ Tomato/ Tomato_ eFPBr owser2. html, 28 ) was lower in all organs and tissues. A study reported that specific silencing of SlINVINH1 increased hexose levels in tomato fruits 11 . Therefore, we considered that SlINVINH2 is a pseudogene or does not play an important role in tomato fruit. To be sure, we selected three consensus sequences between SlINVINH1 and SlINVINH2 as target sequences for CRISPR/Cas9 and Target-AID, expecting knockout of not only SlINVINH1 but also of SlINVINH2 (Fig. 1a,d). The efficiency of mutations by CRISPR/Cas9 was reported to depend on the target sequence 29 . In this study, two target sequences in the first exon (Target 1 and Target 2) and one target sequence in the second exon (Target 3), a total of three target sequences were used for multiplex genome editing (Fig. 1a,d). The proline-lysine-phenylalanine (PKF) motif in INVINH competes with sucrose in the binding site of INV; thus, the PKF motif in INVINH is important for the inhibition of INV activity 6 . Target 1 and Target 2 exist upstream of the PKF motif (Fig. 1a); therefore, the mutation in Target 1 or Target 2 can lead to loss of function of INVINH.
To produce SlINVINH knock-out tomato lines, cotyledon explants of tomato 'Suzukoma' were transformed with Agrobacterium tumefaciens GV2260 harbouring CRISPR/Cas9 vector (Fig. 1b, Fig. S1) or Target-AID vector (Fig. 1c, Fig. S2). Gene transformation in the differentiated shoots from callus was checked by PCR, and only transformants were selected. Tomato frequently produces polyploid plants during de-differentiation and differentiation processes. Therefore, ploidy of the transformants was confirmed, and only diploid plants were Mutation patterns in SlINVINH1 by CRISPR/Cas9 or Target-AID. The mutation pattern of SlIN-VINH1 by CRISPR/Cas9 or Target-AID was confirmed by direct sequencing of SlINVINH1 in the transformants. Therefore, five transformants by CRISPR/Cas9 and eight transformants by Target-AID, which have nucleotide(s) insertion or deletion in the exon of SlINVINH1, were obtained. These transformants (T 0 generation) were selfpollinated and T 1 seeds were obtained. Among the T 1 plants, null segregants with homozygous mutations in SlINVINH1 were selected. In the CRISPR/Cas9 genome-edited lines, nucleotide(s) insertion or deletion at 3 bp downstream from the Proto-spacer Adjacent Motif (PAM) sequence occurred frequently, such as the adenine insertion 3 bp downstream of the PAM sequence of Target 1 in line 247-2 (Fig. 2). This trend was consistent with the report by Jinek et al. 21 , which shows that the mutations by CRISPR/Cas9 occurred frequently around 3 bp downstream of the PAM sequence.
In contrast, in the genome-edited lines by Target-AID using nCas9, insertion or deletion at 3 bp downstream of the PAM sequence, but also substitution of cytosine around 18 bp downstream of PAM sequence occurred frequently, such as the substitution from cytosine to guanine of Target 1 in line 193-3 (Fig. 2). This trend was consistent with the report by Nishida et al. 26 , where mutations by Target-AID using nCas9 occurred frequently around 3 bp downstream of the PAM sequence and at 15-19 bp downstream from the PAM sequence. which are null segregants with homozygous mutations in SlINVINH1, fruit weight and soluble solid content (SSC) of mature fruit were measured. Although the SSC average was higher in the genome-edited lines than the original cultivar 'Suzukoma' (Fig. S3a), fruit weight differed among the genome-edited lines (Fig. S3b). From the 13 genome-edited lines, three lines (193-3, 199-2, and 247-2), whose SSC average was higher than 'Suzukoma' and fruit weight almost equal to 'Suzukoma' (Fig. S3) were selected for further analysis.
Overall plant growth and visual vegetative phenotype of the three selected genome-edited lines (T 2 generation) showed no difference compared with the original cultivar (Fig. 3). On comparison with ''Suzukoma' , lines 193-3, 199-2, and 247-2 showed 37.6%, 14.5%, and 21.4% significantly higher SSC, respectively (Fig. 4a), while no significant difference, 13.7% significantly lower, and no significant difference in fruit weight was observed, respectively ( Fig. 4b). Thus, we succeeded in producing tomato lines with high sugar content (193-2 and 247-2) without a decrease in fruit weight and deterioration of plant growth. These results indicate that knock-out of SlINVINH1 by genome editing is an effective approach to produce high-sugar content tomato cultivars.
Mutation patterns of SlINVINH1 and SlINVINH2 in the three genome-edited lines (193-3, 199-2, and 247-2) are shown in Fig. 2 and Figure S4, respectively. As described above, frameshift mutations presumably lead to loss of function in SlINVINH1 in the three lines (Fig. 2), and in SlINVINH2 in lines 199-2 and 247-2 ( Fig. S4). Only base substitution occurred in SlINVINH2 in line 193-3, in which SSC of mature fruit was the highest among the three lines. These results reinforce our hypothesis that SlINVINH2 does not play an important role, and the knock-out of only SlINVINH1 is sufficient to increase the sugar content of tomato fruit.
Knock-down of SlINVINH1 by RNAi technology in tomato increases the hexose content of fruits 11 . Among the three genome-edited lines (193-3, 199-2, and 247-2), two lines (193-3 and 247-2) showed significantly higher fructose and glucose content in mature fruits than the original cultivar 'Suzukoma' (Fig. 5). The mature fruits of line 193-3 showed the highest sugar content in mature fruit with fructose and glucose contents 29% and 36% higher than 'Suzukoma' , respectively (Fig. 5). The increase in hexose content of the SlINVINH1 genomeedited lines might have been caused by higher sucrose hydrolysis in the apoplast of fruits by retention of high cell wall INV activity, as reported in earlier studies 11 .
Transcript level of SlINVINH1 and mutation in the SlINVINH1 transcript. To determine the transcript level of SlINVINH1 in the genome-edited lines, quantitative RT-PCR analysis of SlINVINH1 was performed using RNA from mature fruit, which showed high SlINVINH1 expression. Therefore, the transcript level of SlINVINH1 decreased in the genome-edited lines compared to the original cultivar 'Suzukoma' . However, the difference in SlINVINH1 transcript levels between the genome-edited lines and 'Suzukoma' was not significant (Fig. S5). This was expected because the target sequences of genome editing were chosen in the coding region of SlINVINH1, but not in the promoter region.
To prove that the transcript of SlINVINH1 lost its function in the genome-edited lines, the sequence of the SlINVINH1 transcript was determined. Similar patterns of insertions, deletions, and substitutions, which were found in the genome of the genome-edited lines (Fig. 2), were found in the SlINVINH1 transcripts of the www.nature.com/scientificreports/ genome-edited lines (Fig. S6), indicating that SlINVINH1 mRNA lost its function. This result supported that the phenotypic effect was due to genome editing of the SlINVINH1 gene.
Off-target mutation analysis of the SlINVINH1 knock-out tomato. Genome editing by CRISPR/ Cas9 can induce off-target mutations as well as in the sequences similar to those of the target sequence 29 . Therefore, to confirm the presence of any off-target mutations in the genome-edited lines (193-3, 199-2, and 247-2), whole genome sequencing was performed and the sequences with mismatches of three bases or less with the target sequences (Target 1, Target 2, and Target 3) in the tomato genome were checked. Therefore, no off-target mutations were found in the SlINVINH1 genome-edited lines in the checked sequences (Table S1).

Non-target metabolome analysis of the SlINVINH1 knock-out tomato.
To evaluate the fruit quality of the SlINVINH1 knock-out tomato, non-target metabolome analysis was performed using GC-TOF/MS and fruit metabolites between the SlINVINH1 genome editing line 193-3 and the original cultivar 'Suzukoma' were compared. Metabolome analysis revealed 148 metabolite peaks, including unidentified peaks. Among them, 76 metabolites were identified, and 19 metabolite levels, including fructose, glucose, and sucrose levels, were significantly different between the genome-edited line and the original cultivar 'Suzukoma' . The relative abundance of the 19 metabolites were 0.51-4.01 fold-change (Table S2). Among the 16 metabolites, excluding fructose, glucose, and sucrose, 10 metabolites significantly increased and 6 metabolites significantly decreased in the genome-edited line compared to the original cultivar (Table S2). Metabolites, such as Trehalose, glucose-6-phosphate, isomaltose, and myo-inositol were found to increase in the genome-edited line. These metabolites are synthesised from glucose. Therefore, the higher glucose levels may have increased the biosynthesis of these metabolites in the genome-edited line (Fig. 5). Two amino acids (asparagine and glutamine) and one organic acid (malic acid) were found to be the decreased metabolites in the genome-edited line. Although the reason for the decrease in these metabolites is unclear, the taste of tomato fruits may be affected because these metabolites are associated to taste 30,31 . No unexpected metabolite peak appeared in the genome-edited line, indicating that no unexpected metabolite accumulated in the fruits of the genome-edited line.
The metabolome data obtained were subjected to variable standardisation and principal component analysis (PCA). In the PCA score scatter plot, SlINVINH1 genome-edited line 193-3 and original cultivar 'Suzukoma' www.nature.com/scientificreports/ were separated in the direction of the first principal component axis (Fig. 6a). In the PCA loading scatter plot, fructose was found to have the highest loading factor (80.2%) and the other metabolites had a low loading factor (less than 30%) in the first principal component separation (Fig. 6b). These results show that fructose was the highest contributing factor for principal component separation between SlINVINH1 genome edited line 193-3 and 'Suzukoma' .
Approaches to produce high sugar content tomato. For the production of tomato with high sugar content, two approaches, including breeding and cultivation, have been applied. Under cultivation approach, water stress and salt stress have been applied to tomato cultivation to increase starch accumulation in fruit at an early developmental stage 32 and to induce the concentration effect of metabolites in fruit at a later developmental stage 33 . However, the application of water stress or salt stress induces cracking of fruit and blossom-end rot. Therefore, a high cultivation technique to control water status and labour for this control is necessary for such cultivation. The most serious disadvantage of water stress or salt stress in cultivation to produce high sugar content tomato is a large decrease in fruit weight 34 , with more than 50% decrease in some cases. However, in the case of SlINVINH1 genome-edited tomato, no high cultivation technique and labour are necessary, and there was no decrease in fruit weight (Fig. 4b).  www.nature.com/scientificreports/ Cross breeding and mutation breeding have been performed as breeding approaches. By gamma ray irradiation, the mutant of tomato dwarf cultivar 'Micro-Tom' , whose SSC in fruit is four degrees higher than the original cultivar, was screened 35 . Among the introgression lines of S. lycopersicum 'M82' containing chromosome fragments of the wild tomato species S. pennellii, the lines with higher sugar content in the fruit were screened 35 . Later, cell wall INV 37,38 and ADP-glucose pyrophosphorylase (AGPase) 39 from S. pennellii were identified as the genes responsible for high sugar content of the introgression lines.
Transgenic technology has also been used to produce tomato cultivars with high sugar content. Introduction of the apple hexose transporter gene (MdHT2.2) 40 or peach sucrose non-fermenting-1-related protein kinase 1 gene (SnRK1) 41 into tomato increased the sugar content of fruit. Overexpression of the tomato phosphoenolpyruvate carboxykinase gene (SlPEPCK) 42 or auxin response factor 10 gene (SlARF10) 43 also promoted sugar accumulation in fruits.
Advantages of the SlINVINH1 knock-out tomato. Jin et al. 11 reported an increase in the hexose content of tomato fruit by knock-down of SlINVINH1 using RNAi technology. In this study, we succeeded in producing tomatoes with high sugar content by knock-out of SlINVINH1 using genome editing technology, without deterioration in plant growth and decrease in fruit weight. The SlINVINH1 knock-out tomato can produce high sugar content fruit without a high cultivation technique that needs control water and labour. These characteristics of the SlINVINH1 knock-out tomato show that it can be a useful high sugar content tomato cultivar, and the knock-out of SlINVINH1 by genome editing technology is an effective approach to produce them.
To sell genome-edited crops on the market, it is necessary to provide sufficient and reliable information to consumers. In this study, we confirmed that the SlINVINH1 knock-out tomato lines are transgene-free and free of off-target mutations using whole genome sequencing. In addition, to confirm metabolome equivalence between the genome-edited line and the original cultivar, non-target metabolome analysis was performed. Therefore, although the levels of 19 metabolites, including fructose, glucose, and sucrose, were significantly different between the genome-edited line and the original cultivar 'Suzukoma' , the differences were not so large, and no unexpected metabolite accumulation appeared in the genome-edited line. These data may help consumers accept genome-edited tomatoes.
In this study, we succeeded in producing high sugar content tomato by knock-out of cell wall INVINH using genome editing technology. Cell wall INVINH is highly conserved among various plant species. Hence, knock-out of cell wall INVINH by genome editing can increase the sugar content of various other fruits, and this technology is expected to be widely used in breeding to produce high sugar content fruit crops.

Materials and methods
All experimental research on plants complied with relevant institutional, national, and international guidelines and legislation. Plant material. Tomato (Solanum lycopersicum 'Suzukoma') was grown in the greenhouse at Nagoya University (Nagoya, Japan). Peat Pot P (Hokkaido Peat Moss, Saitama, Japan) was used as the culture soil. Water containing liquid fertiliser (Otsuka House No. 1 and No. 2, Otsuka Chemical, Osaka, Japan) was applied to plants by bottom irrigation. Five plants from each line and three fruits from each plant were used for phenotyping.
Genetic transformation of tomato. Transformation of tomatoes was performed by modifying the method described by Amemiya et al. 44 . Cultured A. tumefaciens harbouring CRISPR/Cas9 (Fig. 1b, Fig. S1) or Target-AID vectors (Fig. 1c, Fig. S2) was suspended in liquid MS medium containing 100 µM acetostringone and 10 µM 2-mercaptoethanol. Cotyledon explants from tomato seedlings at ca. 10 days after sowing were immersed in the A. tumefaciens suspension for 10 min. The explants were co-cultured with A. tumefaciens on MS basal medium (MS medium containing 2% (w/v) sucrose and 0.8 (w/v) agar, pH 5.8) containing 50 µM acetosyringone in the dark. The explants were then transferred to the callus induction medium (MS basal medium containing 50 µg/mL kanamycin, 375 µg/mL augmentin, and 1.5 µg/mL trans-zeatin) for selection of transformants, callus induction, and sterilisation of A. tumefaciens. After shoot induction from the callus, the calluses were transferred to shoot elongation medium (basal MS medium containing 50 µg/mL kanamycin, 375 µg/mL augmentin, and 1 µg/mL trans-zeatin) for selection of transformants, shoot elongation, and sterilisation of A. tumefaciens. After shoot elongation, the shoots were transferred to the root induction medium (MS medium containing 2% (w/v) sucrose, 50 µg/mL kanamycin, 375 µg/mL augmentin, and 0.3% (w/v) gelrite, pH 5.8) for selection of transformants, root induction, and sterilisation of A. tumefaciens. After rooting, the shoots with roots were transferred to the soil and acclimated. Explants, calluses, and shoots were grown under 16  www.nature.com/scientificreports/ Görlitz, Germany). Ploidy of transformants was confirmed by comparison with the peak of the original cultivar 'Suzukoma' .

Confirmation of genetic transformation and copy number of the transformed gene. Genomic
DNA was extracted from young leaves according to the method described in Edwards et al. 45 . The DNA precipitate was added to 200 µL isopropanol and centrifuged at 18,000 × g for 5 min at 20 °C, 200 µL 70% (v/v) ethanol was added, and the mixture was centrifuged at 18,000 × g for 5 min at 20 °C for purification. The DNA precipitate was suspended in TE buffer (1 mM EDTA, 10 mM Tris-HCl, pH 8.0). Genetic transformation was confirmed by PCR amplification of the Cas9 and PcUbi promoter regions (Fig. 1b,c) with the primers under the conditions described in Table S3.
The copy number of the transformed gene was confirmed by quantitative PCR (qPCR) amplification of the Cas9 fragment (Fig. 1b,c) with the primers under the conditions described in Table S4. qPCR analysis was performed using the Step One Plus Real-Time PCR System (Applied Biosystems, Massachusetts, USA) and TB Green Premix Ex Taq II (Takara Bio, Shiga, Japan). Copy number was estimated by calculating the relative value of Cas9 to LAT52, which was reported to be a single copy gene in the tomato genome 46 .
Confirmation of mutation of SlINVINH1 in genome and transcript. The fragments covering the three target sequences in SlINVINH1 (Fig. 1a) were amplified by PCR using genomic DNA or cDNA prepared for quantitative RT-PCR analysis as described below. The primers and PCR conditions for genomic DNA or cDNA are described in Table S5 and S6, respectively. The obtained PCR product was purified using ExoSAP-IT for PCR Product Cleanup kit (Applied Biosystems). Sequencing reaction was performed using the Big Dye Terminator v3.1 Cycle Sequence Kit (Thermo Fisher Scientific, Massachusetts, USA) and, DNA sequencing was performed using a 3730xl DNA Analyser (Thermo Fisher Scientific).
Quantitative RT-PCR analysis of SlINVINH1. Total RNA was extracted from the 100 mg pericarp of mature fruit using the innuPREP Plant RNA Kit (Analytik Jena, Jena, Germany), according to the manufacturer's protocol. cDNA was synthesised from total RNA using the PrimeScript RT reagent kit (Takara Bio). Quantitative PCR 0.4 μL of ROX II dye (50X) and 7 μL of sterilised water. The primers and PCR conditions are described in Table S7. Amplification was performed using the Step One Plus Real-Time PCR System (Applied Biosystems). Relative expression levels were normalised to SlUBQ3 as the reference gene and calculated using the standard curve method. Three biological replicates of each genome-edited line were analysed.
Measurement of SSC and sugar content. SSC of the pericarp of mature fruit was measured using a Digital Refractometer PR-101 (ATAGO, Tokyo, Japan). Fifteen fruits from five plants of each line were used for the measurement.
Sugar content was measured as described below: 800 µL of ethanol and 50 µL of 1 M mannitol were added to 200 mg pericarp of mature fruit and incubated at 80 °C for 30 min, then centrifuged at 21,000 × g for 5 min at 20 °C. The supernatant was dried under vacuum and suspended in water.
The suspension was passed through a Sep-Pak light C18 (Waters, Massachusetts, USA) and 0.45 µm microfilter (ADVANTEC, Tokyo, Japan) and used as a sample for high-performance liquid chromatography (HPLC) analysis. A Chromaster UV-VIS system (Hitachi High-Tech Science, Tokyo, Japan) and Shodex Sugar SP0810 (SHOWA DENKO, Tokyo, Japan) were used for HPLC analysis. Column temperature was maintained at 80 °C. A refractive index detector (Refractive Index Detector RI-3H, Japan Analytical Industry, Tokyo, Japan) was used.
Off-target sites with a mismatch of three bases or less were searched using CRISPR RGEN Tools Cas-OFFinder (http:// www. rgeno me. net/ cas-offin der/, 47 ). The presence or absence of mutations at the off-target site was confirmed by IGV viewer (http:// softw are. broad insti tute. org/ softw are/ igv/).

Metabolome analysis.
Metabolome analysis was performed according to [48][49][50][51] . Five fruits from each line were used for the analysis. Metabolites in the pericarp of mature fruit were crushed and then extracted by adding methanol/chloroform/water (3:1:1, v/v/v) containing 10 stable isotope references to a concentration of 25 mg FW/mL. After centrifugation, the supernatant was decompressed and dried. To derivatise metabolites, 30 µL of 20 mg/mL of methoxylamine hydrochloride in pyridine solution was added and kept at room temperature for 24 h. Subsequently, to trimethylated, 30 µL of N-methyl-N-trimethylsilyltrifluoro acetamide was added and incubated at 37 °C for 1 h. Prior to analysis, 30 uL of n-heptane was added to each sample (18.5 µg FW equivalent per µL of the solution). The derivatised metabolites were analysed by Agilent 6890 N GC gas chromatography