Production of High Amylose and Resistant Starch Rice Through Targeted Mutagenesis of Starch Branching Enzyme Iib by Crispr/cas9


 Rice is the staple food for half of the world’s population. Starch accounts for 80-90% of the total mass of rice seeds, and rice starch is low in resistant starch (RS) with a high glycemic index (GI). RS has gained important since it is beneficial in preventing various diseases. Starch branching enzyme IIb (SBEIIb) plays a key role in the amylopectin synthesis pathway in the endosperm of cereals. Down-regulation of SBEIIb in several important crops has led to high amylose, high RS and low GI starch. In this study, we mutated OsSBEIIb in the japonica rice cultivar TNG82 through CRISPR/Cas9 and investigated the molecular and physicochemical modifications in OsSBEIIb mutant lines, e.g., gene expression, enzyme activity, apparent amylose content (AAC), RS and GI. As expected, the levels of modification in these starch related traits in heterozygous mutant lines were about half as those of homozygous mutant lines. Gene expression and enzyme activity of OsSBEIIb were down-regulated significantly while AAC and RS contents increased progressively from 17.4% and 0.5% in WT, respectively, to as high as 25.0% and 7.5% in heterozygous mutant lines and 36.0% and 12.0% in homozygous mutant lines. Consequently, with increased RS and decreased rate of reducing sugar production, GI progressively decreased in heterozygous and homozygous mutant rice endosperms by 11% and 28%, respectively. Our results demonstrate that it has huge potential for precise and efficient generation of high RS and low GI rice through CRISPR/Cas9 to provide a more suitable source of starch for type II diabetes.


Introduction
Diabetes mellitus is a chronic disease and type 2 diabetes (T2DM) is the most common form, representing 90-95% of diagnosed diabetes cases (IDF DIABETES ATLAS Ninth edition 2019). Glycemic index (GI) is a measure of postprandial blood glucose response after consumption of foods containing carbohydrate. GI is affected by starch chemical structure (e.g., amylose and amylopectin), and the level of GI affects the health of the human body. Foods with low GI can be absorbed slowly and help maintain blood sugar steady state, preventing various chronic diseases (Bjorck et al., 1994). Low GI foods have consistently shown bene cial effects on glycemic control in both the short term and the long term, also improve blood lipid concentration and prevent further diabetic complications. Thus, GI is a useful reference for individuals with T2DM to use as a dietary guideline (Riccardi et al., 2008).
Starch is the main carbohydrate in human diet and the main source of energy. Starch can be divided into amylose and amylopectin according to its structure. The difference in structure affects the rate of starch being digested. Starch with high amylose content (AC) and resistant starch (RS) has a lower GI value. RS refers to starch that cannot be digested and absorbed in the human small intestine (Englyst et al., 1992).
RS functions similarly to soluble ber; it helps feed the probiotic bacteria in gut and increases the production of short-chain fatty acids such as butyrate (Upadhyaya et al., 2016). Therefore, generating high amylose and RS content crops by traditional breeding or genetic technology is important to produce healthier foods and prevent diabetics.
Rice is a major staple food for half of the world's population, and rice starch accounts for 80-90% of the total mass of rice seeds (Vandeputte et al., 2004). The physical and chemical properties of starch granules, such as AC, gelatinization temperature (GT), starch paste viscosity and amylopectin structure, affect the GI, eating and cooking quality of rice (Singh et al., 2006). Amylose is consisted of glucose residues linked by α (1→4) glycosidic bonds and form long linear chains with a degree of polymerization (DP) < 5,000. In contrast, amylopectin has amylose-like chains with additional branches formed by α (1→6) glycoside bonds with DP 5,000-50,000. Rice amylose is in the range of DP 230-10,000 and amylopectin chains are in the range of DP 6-120 (Fitzgerald et al., 2009). Cultivated rice in the world is mainly divided into japonica rice and indica rice. Starch from indica rice contains a higher amylose than that of japonica rice and has higher melting and pasting temperatures, gel hardness and lower pasting viscosity. It is well documented that amylose formation is controlled by one major locus Wx in the rice genome which encodes the granule-bound starch synthase I (GBSSI) protein. Wx has three major Wx alleles, Wx-I, Wx-II and Wx-III. Wx-I bears a loss-of-function mutation that results in glutinous rice varieties with extremely low AC (< 2%), Wx-II shows a leaky phenotype that leads to a medium level of AC (< 20%) which is widely found in japonica, and Wx-III functions as the WT allele with high AC (> 20%) mainly distributed in indica rice (Tian et al., 2009). Starch synthase IIa (SSIIa) and starch synthase IIIa (SSIIIa) are mainly present in endosperm. The main function of SSIIa is to use medium chains to synthesize long chains of amylopectin with DP of 16-21 and SSIIIa is mainly synthesize long chains of amylopectin with DP > 30. Down-regulation of OsSSIIa or OsSSIIIa signi cantly increases the expression level of OsGBSSI, apparent amylose content (AAC) and AC but decreases the amylopectin content in rice (Fujita et al., 2007). Starch branching enzyme (SBE) is the only enzyme capable of forming the branch linkages in amylopectin. The main function of SBE enzyme is to cut the α(1→4) glycosidic bonds and pass the broken short chain for reconnection to the acceptor chain through α(1→6) glycosidic bonds to form amylopectin (Tian et al., 2009). There are two classes of SBEs: SBEI and SBEII. In the endosperm of monocots, SBEII is divided in SBEIIa and SBEIIb (Nakamura et al., 2018). In rice and maize endosperms, SBEIIb is the major isoform (Ohdan et al., 2005). The mount of SBEIIa is 2-3 fold higher than SBEIIb in wheat endosperm (Regina et al., 2015), while the activity of SBEIIa is equal to that of SBEIIb in barley endosperm (Regina et al., 2010).
Among most cereals, amylose is the main linear component of cereal starch and usually accounts for 15-25% (Ball et al., 1998). The AC of wild and cultivated rice ranges from 0-30%. Overexpression of the appropriate Wx gene which encodes the GBSSI can further increase the level of amylose (Hanashiro et al., 2008) while down-regulation of the expression of the enzymes involved in the biosynthesis of amylopectin like SBEI and SBEIIb thereby reduces the synthesis of amylopectin and increases the proportion of amylose and RS in cereals (Wei et al., 2010a;Butardo et al., 2011). Inhibiting the expression of SBEIIb genes affects the synthesis and structure of amylopectin. Study in wheat showed that inhibition of both TaSBEIIa and TaSBEIIb at the same times signi cantly increases long side-chain branching of amylopectin with DP > 24, AC and RS (Regina et al., 2015). Inhibiting StSBEIIb in potatoes also increases the average length of amylopectin branch chains in its starch, leading to an increase in AAC (Jobling et al., 1999). In addition, overexpressing StSBEII in potato leads to an increase in the number of amylopectin chains of DP ≤ 12 and causes a particularly large increase in those of DP6 with signi cantly decreased AC (Brummell et al., 2015). The rice OsSBEIIb mutants through antisense RNA show signi cant change in the structure of amylopectin and content of short side chains with a signi cantly reduced DP of 8 ~ 12 in the endosperm (Zhu et al., 2012). Compared with WT, the endosperm of the transgenic rice with suppressed expression of OsSBEIIb gene has a higher AC, RS, gelatinization temperature, and stronger resistance to acid hydrolysis and enzymatic hydrolysis. Starch granules with a higher ratio of long chains of amylopectin are more resistant to gelatinization (Jane et al., 1999). Consistently, deletion or suppression of OsSBEIIb by chemical mutagenesis, RNAi and CRISPR/Cas9 targeted editing in rice causes a signi cant increase in the amylose and RS contents and a signi cant decrease in amylopectin content. Signi cant increase of AC also changes starch chemical properties, e.g. decreased swelling power and viscosity or increased gelatinization temperature (Wei et al., 2010a;Butardo et al., 2011;Yang et al., 2012;Sun et al., 2017;Baysal et al., 2020). These studies suggest that overexpression of GBSSI to increase amylose biosynthesis or down-reulation of SBEIIb, SBEI, SSIIa and SSIIIa to inhibit amylopectin biosynthesis can effectively increase AC in the endosperms of cereal crops.
Foods rich in RS can lead to lower blood sugar and insulin response, and reduce the risk of type II diabetes, obesity and cardiovascular disease (Rahman et al., 2007). A positive relationship between RS and AC has been established through several studies. Generating the high amylose crops through inhibition of the SBEIIb gene leads to increase RS with a strong anti-digestive properties, with bene cial to human health. A large number of high amylose crops have been generated through down-regulation of SBEII gene by chemical mutagenesis, RNAi and CRISPR/Cas9, including high amylose and RS wheat Traditional cross and mutation breeding are restricted by the large-scale screening, time-consuming, low e ciency and genetic barrier and cannot keep pace with the demand for increased crop production. On the other hand, genetic engineering provides big potential to increase crop productivity by overcoming these limitations (Mehta et al., 2020). In particular, genome editing techniques that can change speci c positions in the genome have gained popularity in recent years for improving crop traits. CRISPR/Cas mediated gene editing is simple and easy to implement with a high e ciency, and therefore become a new tool leading the gene editing technology. CRISPR/Cas9 is suitable and e cient for crop breeding and exploring the functions of plant genes for crop improvement (Bortesi and Fischer, 2015). Thus, CRISPR/Cas9-mediated gene editing technology has the potential to greatly facilitate plant breeding. AC is controlled by a single dominant Waxy gene, which encodes a GBSS in rice. The loss of function caused by CRISPR/Cas9 of rice wx genes reduces the AC (Zhang et al., 2018). On the other hand, CRISPR/Cas9mediated loss-of-function mutations of starch branching enzymes SBEI and SBEIIb in rice leads to higher AC and RS (Sun et al., 2017). Strong reduction in both StSBE1 and StSBE2 in tetraploid potatoes through CRISPR/Cas9 also creates high amylose potato with recreation in short amylopectin chains, an increase in long chains and a large reduction in branching frequency relative to WT starch (Tuncel et al., 2019). Targeted mutagenesis of TaSBEIIa by CRISPR/Cas9 in both winter and spring wheat varieties has generated transgene-free high-amylose wheat which also signi cantly increased RS, protein and soluble pentosane contents (Li et al., 2020).
In this study, we report the feasibility of creating high amylose, high RS and low GI rice plants through CRISPR/Cas9-mediated target gene editing of OsSBEIIb in the japonica rice cultivar TNG82. CRISPR/Cas9 has advantages over other means of gene editing, including low cost, easy operation, and high e ciency, etc. We also de ne the role of SBEIIb in determining the different molecular and physical modi cations of OsSBEIIb gene between homozygous and heterozygous mutants, as shown in gene expression level, enzyme activity, amylose, RS, and physicochemical properties of starch. Thus, this work enables the provision of a more suitable source of starch for type II diabetes and advances the work of breeding in the future.

Vector construction
The sgRNA-Cas9 plant expression vector was kindly provided by Prof. Yaoguang Liu, South China Agriculture University and reconstructed by Academia Sinica for this work. The vector was reconstructed by inserting synthesized oligos into a BsaI site of the vector pYLCRISPR/Cas9Pubi-H, which contains a codon-optimized Cas9 driven by a maize ubiquitin promoter, a sgRNA scaffold directed by a rice U6a promoter, a selectable marker gene (hptII) driven by a CaMV 35S promoter and the backbone of the binary vector pCAMBIA1300 (CAMBIA, Canberra, Australia) (Ma et al., 2015). The sgRNA and off target site analysis were designed by CRISPR-P 2.0 design tool . The 2 editing target sites were designed on exon 3 and intron 3 of OsSBEIIb gene.

Rice transformation
Rice transformation through A. tumefaciens was carried out according to previous methods (Ku et al., 1999;Yeh et al., 2015). Calli induced from immature seeds of the japonica cultivar rice TNG82 (Oryza sativa L.) were used for rice transformation. The compact embryogenic calli were co-cultured with A. tumefaciens strain AGL1 carrying the plasmid pYLCRISPR/Cas9Pubi-H and incubated at 28°C in dark for 3-4 days. After co-cultivation, calli were thoroughly washed with 250 mg L − 1 cefotaxime in sterile distilled water and transferred to a selection medium containing 50 mg L − 1 hygromycin B (Invitrogen®) and 250 mg L − 1 cefotaxime at 28°C in light for one month. Healthy hygromycin-resistant calli were subsequently transferred to a 1/2 MS regeneration medium supplemented with 2.5 mg L − 1 kinetin, 1 mg L − 1 NAA and 50 mg L − 1 hygromycin. Transgenic plantlets were transferred to a hormone-free 1/2 MS medium without hygromycin in magenta boxes for 10-14 days to promote root growth. Calli selection and plant regeneration were conducted in a growth chamber at 28°C and 60% relative humidity under a 16 h/8 h light/dark photoperiod. Transgenic seedlings were transplanted in soil, cultured in the greenhouse or eld between May and October. Self-pollinated seeds obtained from greenhouse-grown T0 plants were used for analysis of gene expression level and SBEIIb enzyme activity while the seeds obtained from eld-grown plants were used for analysis of grain composition and molecular structure of starch.

DNA extraction, PCR, restriction enzyme digestion and screening of mutant plants by sequencing
To detect mutations in transgenic lines, genomic DNA was isolated from leaf tissue or calli of transgenic and WT (wild type) plants, as described (Sheu et al., 1996). Two hundred ng of genomic DNA was used as template for PCR ampli cation using Taq DNA Polymerase Master Mix RED (AMPLIQON®). Speci c primers used in PCR analysis (Supplemental Table 1) included the primers gSBE3-F2/gSBE3-R2 for anking the sequence of target site 1 (exon 3) and the primers gSBE3-F1/gSBE3-R1 for anking the sequence of target site 1 and target site 2. The PCR products ampli ed by the primers gSBE3-F2/gSBE3-R2 using calli genomic DNA as template were used for restriction enzyme digestion analysis to con rm mutations. PCR was performed in a thermal cycler under the following conditions: 94°C/5 min, 30 cycles of 94°C/0.5 min, 55°C/0.5 min and 72°C/0.5 min, and a nal extension of 72°C/5 min. The PCR products were digested with DdeI; and mutants displayed distinct band patterns on agarose gel, as compared to WT. Regenerated T0 plants from con rmed calli mutant lines were subjected to sequence analysis. The DNA fragment across the target site 1 and target site 2 was ampli ed by the primers gSBE3-F1/gSBE3-R1 using leaf genomic DNA in PCR reaction: 94°C/5 min, 30 cycles of 94°C/0.5 min, 60°C /0.5 min and 72°C/0.5 min, and a nal extension of 72°C/5 min. PCR products were cloned into the TA cloning vector T&A™ (Yeastern Biotech Co., Ltd) for sequencing Analysis of grain composition Seeds were harvested and dried at 37°C for at least 3 days. Hundred-grain weights (g) of selected representative transgenic lines was measured in triplicate. The absolute digimatic caliper (Mitutoyo, Japan) was used to determine grain appearance and dimension. The opacity of seeds was investigated using the chalkiness index following the standard of Australian rice industry.

Analysis of OsSBEIIb gene expression by qRT-PCR
Total RNA was isolated from endosperm tissues, as described by Wang and Vodkin (1994). Total RNA was rst treated with RNase-free DNase I to remove genomic DNA and the resulting RNA was used for rst-strand cDNA synthesis with a random primer (Supplemental Table 1) and M-MLV reverse transcriptase (Promega) (Chen et al., 2006). Two µg cDNA was used as a template for gene expression assay by qRT-PCR; and the protocols of Livak and Schmittgen (2001) and Yeh et al., (2015) were followed using speci c primers qcSBE3-F2/qcSBE3-R2 for OsSBEIIb and primers qcSBE1-F1/qcSBE1-R1 for OsSBEI (Supplemental Table 1). Expression of 17S rRNA (accession number X00755) was used as an internal control for normalization of expression levels.
Activity assay of starch branching enzyme (SBE) The method for assaying the total activity of SBE in rice grains (Sun et al., 2011) includes the activities of SBE1, SBEIIa, and SBEIIb together.
Analysis of total starch, amylose content (AC), resistant starch (RS) and amylopectin content The total starch, AC and RS contents in the rice our were measured in vitro with the starch assay kits Megazyme K-STAR, K-AMYL and K-RSTAR (Megazyme, Wicklow, Ireland), respectively. The amylopectin content was indirectly determined by subtracting the amylose percentage from the total starch percentage of the sample.

Measurement of the rate of reducing sugar production
The in vitro digestibility of rice our was determined as described previously (Butardo et al., 2011). The rate of reducing sugar production was calculated as follows: (R 180 -R 10 )/170 (mg − 1 min − 1 ); S 180 , reducing sugar content at 180 min (mg); R 10 , reducing sugar content at 10 min (mg).

Rapid visco analysis
The American Association of Cereal Chemists (AACC) 61 − 02 standard measurement method was slightly modi ed for viscosity analysis. Three g of rice our was mixed with 25 mL pure water in a measuring cup for measurement of viscosity using RVA-4 (Newport Scienti c, Australia) rapid viscosity analyzer. After incubating at a xed temperature of 50℃ for 1 min, the temperature was increased to 95℃ at a rate of 12℃/min and maintained at 95℃ for 2.5 min. Then the temperature was reduced to 50℃ at 12℃/min and maintained at 50℃ constant temperature to the end of the measurement, the

Measurement of glycemic index (GI) through static in vitro digestion
The consensus INFOGEST protocol was followed for the static in vitro digestion (Minekus et al., 2014). The obtained values of starch available were normalized to percentage of hydrolyzed starch.

Calculation of glycemic index (GI)
Glucose concentration quanti ed during the digestion process was normalized to percentage of total starch hydrolyzed, so that the GI index of the samples can be determined using the area under the curve (AUC) (Fernandes et al., 2020).

Data Analyses
Grain morphology, starch molecular and physiochemical properties of TNG82 (untransformed wild type), TNGS14 (indica rice variety) and OsSBEIIb mutant lines were analyzed by one-way variance analysis (ANOVA). Standard deviation was used to represent error values or error bars. Student's t-test was used to test signi cant difference in starch molecular and physiochemical properties between TNG82 and OsSBEIIb mutant lines.

CRISPR/Cas9-mediated mutagenesis of OsSBEIIb in rice and production of transgenic plants
We targeted the third exon (Target 1) and third intron (Target 2) of rice OsSBEIIb gene (EnsemblPlants no Os02g0528200) to make long deletion mutations through CRISPR/Cas9 mediated editing (Fig. 1A). Target 1 sequence contains a restriction site for screening of mutations using PCR-based restriction enzyme (PCR/RE) digestion (Fig. 1A). The CRISPR/Cas9 construct for rice transformation, kindly provided by Academia Sinica, contains two gRNA cassettes in the vector pYLCRISPR/Cas9Pubi-H (Fig. 1B).
The CRISPR/Cas9 vector was transformed into embryogenic calli of the japonica rice cultivar (TNG82), through Agrobacterium-mediated method, and multiple transgenic lines were obtained after three rounds of selection on hygromycin-containing medium (Yeh et al., 2015). In total, 15 independent OsSBEIIb mutant rice lines were selected from 46 hygromycin-resistant transgenic calli through PCR/RE assay for detailed analysis. Our sequence analysis indicated that, of these 15 independent transgenic plants, 1 line (6.6%), 11 lines (73.3%), and 3 lines (20.0%) are genotypically homozygous, bi-allelic, and heterozygous mutants, respectively, at the rst target site on exon 3 of OsSBEIIb (Fig. 1C, D).
We obtained long deletion mutants at a single or both alleles in which Cas9 cuts the two target sites at the same time, e.g., Line-34, 60 and 64 (Table 1). However, some mutants harbored large insertions, e.g., Line-60 had 210 bps insertion at the rst target site that contains a DdeI cutting site. In addition, there were several short insertion and deletion mutants with 1-26 bps. Line-64 is considered as heterozygous for phenotype because it has 221 bps deletion at one allele and 3 bps deletion at the other allele of Target 1 (Table 1), which may not change OsSBEIIb function. Representative OsSBEIIb homozygous and heterozygous mutant lines were chosen for comparison with WT and indica cultivar TNGS14. At the rst target we selected ve homozygous (Line-33, 34, 42, 43, 60) and two heterozygous (Line-37, 64) OsSBEIIb mutant lines for phenotypic analyses, such as gene expression, grain morphology and starch physicochemical properties. CRISPR-Cas9-based genetic screens is a powerful new tool in biology, but the on-target activity and off-target effects of individual sgRNAs can vary widely. CRISPR-P2 tool was used to predict potential off-target site editing of the two target sites for OsSBEIIb, and site-speci c genomic PCR and DNA sequencing were used to investigate whether the predicted off-target sites were also edited. The results showed that no mutations are detected at the putative off-target loci in the genome of the seven OsSBEIIb mutant lines obtained in this study (Supplemental Table 2).
OsSBEIIb expression and SBE activity during rice grain lling in WT and seven OsSBEIIb mutant lines Quantitative real-time PCR (qRT-PCR) analysis showed that the expression levels of OsSBEIIb in both homozygous and heterozygous mutant lines were signi cantly lower than that of WT (Figure 2A), with very low levels in the ve homozygous mutant lines (Line-33, 34, 42, 43, 60) to about 2.5-fold lower in the two heterozygous mutant lines (Line-37, 64) in comparison to that in WT. As expected, the expression of OsSBEIIb in the homozygous mutants was almost completely knocked down since rice genome contains only one copy of the gene (Yang et al., 2012). The expression levels of OsSBEI were not altered in these mutant lines (Supplemental Figure 1). The total enzyme activities of SBE in young rice grains 20 and 25 days after owering were compared between WT and mutant lines ( Figure 2B). For both WT and the mutant lines, the total activities of SBE at 20 days were higher than those of SBE at 25 days after owering. Both homozygous and heterozygous OsSBEIIb mutant lines showed signi cantly lower activities than that of WT ( Figure 2A). Compared to WT, the ve homozygous and two heterozygous mutant lines exhibited nearly 3 and 2-fold decreases in total SBE activities, respectively. The decreases in total SBE activity among the mutants were mainly due to the lower OsSBEIIb expression levels with a positive correlation coe cient of r = +0.95 (Supplemental Figure 2A). The substantial activities remaining in the mutant grains are presumably due to the normal expression of other SBEs (Baysal et al., 2020).
Grain morphology and starch physicochemical properties WT, indica cultivar TNGS14, 2 heterozygous mutant lines (Line37, 64) and 2 homozygous mutant lines (Line-34, 60) that have long deletion at the rst target site, and Line-42 that has 5 bps deletion at each alleles at the rst target site were selected for comparison of their grain morphology and starch physiochemical properties. The grain weight, length, width and thickness were signi cantly lower in the homozygous mutant lines than those of WT, while there were no signi cant differences in these traits between the heterozygous mutant lines and WT. The grains of the heterozygous mutant lines were chalky, while the grains of the homozygous mutant lines were opaque throughout (Fig. 3). In contrast, WT grains appeared uniformly translucent. The Onset gelatinization temperature of homozygous mutant lines was signi cantly higher than that of WT whereas, it was slightly higher in heterozygous mutant lines than WT ( Table 2). The predicted glycemic indexes (GI) of the homozygous (average 55.5) and heterozygous mutant lines (average 68.5) were signi cantly lower than that of WT (77); and the GI values of the two heterozygous mutant lines were similar to that of the indica cultivar TNGS14 ( Table 2). The decreases in GI among the mutants are mainly due to the increases in RS content with a negative correlation coe cient of r = -0.97 (Supplemental Fig. 2K).
Apparent amylose content (AAC), amylose content (AC), total starch content and the ratio of amylopectin/amylose in indica rice cultivar TNGS14, TNG82 (WT) and seven OsSBEIIb mutant lines The ACs of WT and the indica rice cultivar TNGS14 were 15.8% and 26.8%, respectively. As expected, indica rice tends to have a much higher AC than japonica rice. Both AAC and AC increased signi cantly in the seven mutated lines, with AAC ranging from 23.1 to 23.6% in the two heterozygous and from 34.0 to 36.0% in the ve homozygous mutant lines (Fig. 4A). On average, the AC increased from 15.8% in the WT to 24.1% (1.53 fold) in the heterozygous and 30.8% (1.95 fold) in the homozygous mutant lines (Fig. 4B). The increases in AC in the grains of these mutant lines did not affect the total starch contents; the total starch contents in the mutant lines were similar to that in WT (Fig. 4C). As expected, the ratio of amylose/amylopectin increased signi cantly in the mutant lines, relative to WT, ranging from 21.0 in TNG82 to 32.7-38.4 in the heterozygous mutant lines and 50.4-54.4 in the homozygous mutants ( Fig. 4D). Taken together, these results clearly show that a total knockout of OsSBEIIb in rice can increase AC from 15.8 to 31-35% in TNG82 which are mainly due to the decreases in OsSBEIIb expression (r = -0.98) and OsSBEIIb enzyme activity (r = -0.95) in the mutants (Supplemental Fig. 2B, C). Consistently, the increases in AAC are also related to the decreases in OsSBEIIb expression (r = -0.94) and OsSBEIIb enzyme activity (r = -0.95) in the mutants (Supplemental Fig. 2D, E). Clearly, suppression of OsSBEIIb expression for amylopectin biosynthesis in rice leads to increased AC.

RS content and rate of reducing sugar production in indica rice cultivar TNGS14, TNG82 (WT) and seven OsSBEIIb mutant lines
The RS contents in the grains in WT and indica cultivar TNGS14 were very low (0.4-1.0%) but signi cantly increased to 6.7-7.5% in the two heterozygous and 11.6-12.0% in the ve homozygous mutant lines (Fig. 5A). Again, the increases in RS are mainly due to the decreases in OsSBEIIb expression (r = -0.99) and OsSBEIIb enzyme activity (r = -0.95) among the mutants (Supplemental Fig. 2F, G). The rate of reducing sugar production, a gradual depletion of starch during digestion, was estimated after α-amylase treatment. The rates were much lower in the grains of the two heterozygous (Line-37, Line-64) and ve homozygous OsSBEIIb mutant lines than that of WT (Fig. 5B, C); and they are positively correlated to their GIs (r = 0.93) ( Table 2, Supplemental Fig. 2J). Again, the rates of reducing sugar production were lower in the ve homozygous lines than those of two heterozygous lines and indica cultivar TNGS14 (Fig. 5B, C). The decreases in the rate of reducing sugar production in the mutants are entirely due to the decreases in OsSBEIIb expression (r = 0.99) and OsSBEIIb enzyme activity (r = 0.96) (Supplemental Fig. 2H, I).

In uences of OsSBEIIb mutations on starch viscosity
Rapid viscosity analysis (RVA) was performed with the grain ours of indica rice cultivar TNGS14, WT and seven CRISPR-edited mutant lines to test their resistance to alkaline gelatinization. Compared to WT, the peak viscosities (PKV) in all seven OsSBEIIb mutant lines decreased signi cantly and the varicosities for the ve homozygous mutant lines were much lower than those of the two heterozygous mutant lines (Fig. 6). Both hot viscosities (HV) and cool paste viscosities (CPV) of OsSBEIIb mutant lines were signi cantly lower than those of WT, but there were no differences between the homozygous and heterozygous mutant lines. Analogously, the starch breakdown viscosities (BDV, PKV-HV) and setback viscosities (SBV, CPV-HV) of the two heterozygous mutant lines were similar to those of WT, but much higher than those of the ve homozygous mutant lines. Whereas all these values in the seven OsSBEIIb mutant lines were signi cantly lower than those in the indica cultivar TNGS14, the two heterozygous mutant lines showed higher PKV and BDV values than the indica cultivar TNGS14 (Fig. 6). In summary, these results indicate that the structure of starch is signi cantly modi ed by down-regulation of OsSBEIIb in the rice endosperm that leads to changes of starch viscosity and palatability.

Discussion
Starch is the major source of carbohydrate for humans. Starch accounts for up to 75% of the mass of cereal grains (Ordonio et al., 2016). Rice starch is low in RS and metabolic engineering of rice starch is mainly driven by the health bene ts of RS, which has a high content of amylose and is more di cult to digest (Englyst et al., 1992). RS has been reported to provide many health bene ts for humans, and high RS foods have lower glycemic and insulin responses and reduce the risk of type II diabetes, cardiovascular disease and obesity (Baxter et al., 2019). The number of death caused by diabetes increased by 80% in the past twenty years, and it is projected to increase to over 6 million people by 2040 (IDF DIABETES ATLAS Ninth edition 2019). Thus, it's important to generate high AC, high RS and low GI rice as a better source of starch for diabetes patients.
SBE is a key enzyme in the process of starch biosynthesis, forming the branched structure of amylopectin by catalysis of glucose monomer binding through α-1,6 bonds (Tian et al., 2009). SBEIIb de cient mutants and transgenic rice have been shown to accumulate greatly modi ed with as higher AC, higher RS content and higher onset temperature for gelatinization (Nakamura et al., 2018). SBEIIa and SBEIIb usually produce short chains while SBEI usually produces long chains of amylopectin (Satoh et al., 2003). Down-regulating or eliminating the expression of SBEII in several major crops also leads to increased contents of amylose and RS. For example, editing the exon 12 or exon 18 of OsSBEIIb in the japonica cultivar Nipponbare through CRISPR/Cas9 signi cantly decreases its expression level and signi cantly increases the contents of amylose and RS in the endosperm (Baysal et al., 2020). In the present study, we developed seven OsSBEIIb mutant lines including both heterozygous and homozygous lines through CRISPR/Cas9 in the low amylose and low RS japonica rice cultivar TNG82 for molecular and physiological studies. We designed two gRNAs respectively targeting exon 3 and intron 3 of OsSBEIIb to generate long fragment deletion (Fig. 1A). Knock-out and knock-down of OsSBEIIb lead to decreased gene expression, OsSBEIIb enzyme activity and total starch content and increased amylose and RS contents. Expression of OsSBEIIb in homozygous lines is almost totally suppressed and is about 50% lower in heterozygous lines than that in WT ( Fig. 2A). As expected, OsSBEIIb expression in the homozygous mutants is almost completely knocked down since rice genome contains only one copy of the gene (Yang et al., 2012). The transcript for OsSBEI is unaffected in OsSBEIIb RNAi mutant rice seed (Butardo et al., 2011;Baysal et al., 2020). Consistently, in the current work the mRNA level of SBEI is not affected at all in the mutant lines (Supplemental Fig. 1). Following the suppression of OsSBEIIb expression, SBE activities in heterozygous and homozygous lines are signi cantly reduced (Fig. 2B). The substantial SBE activity remaining in the mutant lines (Fig. 2B) is presumably due to the normal expression of other SBEs as reported previously (Butardo et al., 2011; Baysal et al., 2020).
In transgenic barley, almost all the increase in AC has been obtained by inhibiting the SBE genes. HvSBEIIa and HvSBEIIb activities are equal in barley endosperm, and silencing both HvSBEIIa and HvSBEIIb in barley at the same time increases AC from 28.5 to 89.3% (Regina et al., 2010). Moreover, amylose-only barley is generated when the three SBE genes are simultaneously suppressed (Carcio et al., 2012). In wheat, TaSBEIIa is the major isoform of SBEII in the endosperm and RNAi inhibition of TaSBEIIa increases AC from 25.5% to more than 70% (Regina et al., 2006(Regina et al., , 2015. In potatoes, AAC is increased from 22% to as high as 60-87% through small interfering RNAs (siRNA) to simultaneously inhibit the expression of StSBEI and StSBEII together (Andersson et al., 2006). The AC of ordinary maize starch is about 30% (Huang et al., 2015) and its RS content is as low as 1-2% (Chanvrier et al., 2007).
ZmSBEIIb is the predominant isoform in the maize endosperm which is at least 50 times the abundance of ZmSBEIIa (Gao et al., 1997). In the maize amylose extender (ae) mutant AAC and RS are as high as 85.6% and 34.3%, respectively (Li et al., 2008).
OsSBEIIb is also the predominant SBEII isoform in rice endosperm, and chemical mutagenesis of OsSBEIIb in the japonica rice cultivar Jiangtangdao 1 signi cantly increases AC from 16.2 to 31.1% and the RS content from 0.4 to 11.7% (Yang et al., 2012). Suppression of OsSBEIIb expression mediated by arti cial microRNA (ami-RNA) in the japonica cultivar Nipponbare signi cantly increases AC from 19.6 to 41.2% and increases RS from 0.1 to 4.4% while GI decreases from 85 to 44 (Butardo et al., 2011). Targeted editing of OsSBEIIb by CRISPR/Cas9 in the japonica cultivar Kinmaze increases its AC from 15 to 25% and RS from 0.1 to 6.5 with a 30% reduction in grain weight (Sun et al., 2017). In this study, the AAC and AC in OsSBEIIb homozygous mutant endosperms are 2 fold higher (36.0% and 31.8%, respectively) than WT (17.4% and 15.8%, respectively) and 1.2 fold higher than the indica variety TNGS14 (28.0% and 26.8%, respectively). The corresponding values in heterozygous mutant lines (25.0% AAC, 23.4% AC) are 1.44 fold and 1.48 fold higher than those of WT, respectively (Fig. 3A, B). Thus, our results are consistent with previous studies. Satoh et al., (2003) showed that OsSBEIIb mutant has signi cantly reduced weight of seeds with opaque endosperm and the starch grains are abnormal in size, shape, and distribution. The seed weight and size of our homozygous lines are also signi cantly decreased, but these traits are very similar between heterozygous lines and WT ( Table 2). The degree of chalkiness of our mutant grains is similar to that of OsSBEIIb mutant rice obtained in the previous study (Butardo et al., 2011).
Usually, starch with a high proportion of amylose has a high RS content. Compared with amylopectin, amylose has a smaller molecular weight, tighter arrangement and its chain structure has fewer space obstacles to regenerate crystal structure, making it di cult for α-amylase to digest (Shu et al., 2006;Gani et al., 2017). In this study we nd 17.8-fold and 29.5-fold increases in RS content in the heterozygous and homozygous mutant lines, respectively, relative to WT (Fig. 4A). Expression of OsSBEIIb is negatively correlated with RS (Supplemental Fig. 2F (Sajilata et al., 2006) because the hydrogen bonds that exist in the glucose chain connecting amylose molecules make it more resistant to α-amylase (Alhambra et al., 2019). The digestibility or rate of reducing sugar production in the homozygous mutant lines obtained in this study is signi cantly lower than those of WT and the high AC indica cultivar TNGS14 (Fig. 4B), due to increased AC and RS. In contrast, the rates of reducing sugar production in heterozygous mutant lines are very similar to that of indica cultivar TNGS14 but lower than that of the WT (Fig. 4C). This is related to the higher proportion of amylose and long-chain amylopectin and high semi-compound starch which is consisted of many individual granules held together with a smaller speci c surface area. Therefore, composite particles bind less effectively to α-amylase than individual particles, which limits hydrolysis (Wei et al., 2010b;Zhu et al., 2012). In our mutant lines, the averaged GIs in heterozygous and homozygous mutant lines are 68.5 and 55.5, respectively; both are signi cantly lower than the GI for WT (77). The GIs of heterozygous mutant lines are similar to that of indica cultivar TNGS14. In spite of the fact that our homozygous mutant lines are in a japonica background (TNG82) that has low AC (15.8%) and the indica rice TNGS14 that has a high AC (26.8%), the homozygous mutant lines have a lower GI value (55.5) than WT (77) and TNGS14 (69) ( Table 2). Sun et al., (2017) nd signi cantly higher RS contents (5.2-9.8%) in their OsSBEIIb mutant lines with long fragment insertion and mutant lines with only 1 bp insertion at the editing site by CRISPR/Cas9. In our study, a similar RS content is found in homozygous mutant lines with long fragment insertion (Line-34, 60) and short fragment insertion (Line-33, 42, 43) as well (Fig. 4A). The levels of change in molecular and physicochemical properties in heterozygous mutant lines are almost half of those in homozygous mutant lines, indicating that the non-editing allele in heterozygous lines is being used to transcript and translate functional OsSBEIIb. Interestingly, Line-64 with deletion 211 bp and 3 bp in each allele has similar molecular and physicochemical properties to other heterozygous lines. Apparently, this allele with 3 bp deletion can lead to loss of only a single amino acid in the enzyme protein in the translation step buy does not affect the ability of the mutated OsSBEIIb to function (Fig. 1D). Comparing the target 2 (intron) mutations between Line-34 (-219/-221) and Line-60 (-219/WT), we show that the mutation on intron doesn't seem to lead to any effect on starch properties. Clearly, choosing mutant lines with high RS content, high AAC and low GI is the goal in a breeding program.
The RVA has been widely used for assessing the pasting of starch and it is often used to identify a particular characteristic of the rice variety (Kesarwani et al., 2016). The AC in rice is negatively correlated with breakdown (BDV) and peak viscosities (PKV) but positively correlated with cool paste (CPV) and setback viscosity (SBV: CPV-PKV). The low viscosity and high gelatinization temperature of high amylose rice and corn can be attributed to their high ACs (Shu et al., 1998). The swelling degree of starch is inhibited by the increase of AC. It is generally believed that starch with a low AC will show a higher viscosity and a lower gelatinization temperature (Noda et al., 2004). The high helix conformation of amylose with more hydrogen bonds requires a higher temperature to gelatinize, leading to an increase in gelatinization temperature (McGrane et al., 2004). With signi cantly higher AC in the endosperm starch of our OsSBEIIb mutant lines, their PKV, BDV and gelatinization temperature are signi cantly decreased but SBV is signi cantly increased (Fig. 5).
In this study, we further demonstrate the advantages of CRISPR/Cas9-mediated genome editing technology in breeding high-amylose rice to meet the increasing demand caused by diet-related chronic diseases such as diabetes. Compared with conventional breeding, chemical-or physical-induced mutagenesis and transgene-based strategies, CRISPR/Cas9-mediated genome editing technology can modify target genes more precisely in a very effective way (Jones et al., 2015). A previous study demonstrates that mutant lines with simultaneous repression of OsSS IIIa and OsSBEIIb in rice by RNAi has increased the expression levels of both OsGBSSI and ADP-glucose pyrophosphorylase (OsAGPase) genes, resulting in 1.5-fold accumulation of amylose in the endosperm compared to OsSSIIIa or OsSBEIIb single mutant line (Asai et al., 2014). Also, it has been shown that the starch synthase SSIIa and SSIIIa double mutant rice has signi cantly increased AC by 2-fold and increased medium-long chains of amylopectin (DP 7-11) in the japonica cultivar Kitaake (Zhang et al., 2011

Conclusion
In the present study, high AAC and RS rice with a signi cantly lower GI was generated through mutated OsSBEIIb by CRISPR/Cas9 in the japonica rice cultivar TNG82. We showed that the molecular and physicochemical modi cations in OsSBEIIb homozygous and heterozygous mutant lines are signi cantly different from WT. The levels of modi cation in the molecular and physiochemical properties of starch in heterozygous mutant lines were about half as those in homozygous mutant lines. The level of gene expression and enzymic activity of OsSBEIIb decreased more signi cantly in the homozygous than heterozygous mutant lines while the AAC in the homozygous and heterozygous mutant lines were 2.1and 1.4-fold higher than that in WT (TNG82), respectively. On the contrary, the RS contents in the homozygous and heterozygous mutant lines were 25-and 15.6-fold higher than those in WT (TNG82), respectively. With increasing RS, the GI was also progressively decreased in heterozygous and homozygous mutant endosperms. These results demonstrate that it has huge potential for precise and e cient generation of transgene-free high RS and low GI rice through CRISPR/Cas9 for the provision of a more suitable source of starch for chronic disease patients.       (C) Changes in reducing sugar after α-amylase treatment. Error bars indicate SD of means (t-test, 3 replicates of samples, ** P < 0.005, *** P < 0.001).