Production of very-high-amylose cassava by post-transcriptional silencing of branching enzyme genes

High amylose starch, a desired raw material in the starch industry, can be produced by plants deficient in the function of branching enzymes (BEs). Here we report the production of transgenic cassava plants with starches containing up to 50% amylose due to the constitutive expression of hair-pin dsRNAs targeting the BE1 or BE2 genes. A significant decrease in BE transcripts was confirmed in these transgenic plants by quantitative real-time RT-PCR. The absence of BE1 protein in the BE1-RNAi plant lines (BE1i) and a dramatically lower level of BE2 protein in the BE2-RNAi plant lines (BE2i) were further confirmed by Western blot assays. All transgenic plant lines were grown up in the field, but with reduced biomass production of the above-ground parts and storage roots compared to wild type (WT). Considerably high amylose content in the storage roots of BE2i plant lines was achieved, though not in BE1i plant lines. Storage starch granules of BE1i and BE2i plants had similar morphology as WT, however, the size of BE1i starch granules were bigger than that of WT. Comparisons of amylograms and thermograms of all three sources of storage starches revealed dramatic changes to the pasting properties and a higher melting temperature for BE2i starches. Glucan chain length distribution analysis showed a slight increase in chains of DP>36 in BE1i lines and a dramatic increase in glucan chains between DP 10-20 and DP>40 in BE2i lines, compared to that of WT starch. Furthermore, BE2i starches displayed a B-type X-ray diffraction pattern instead of the A-type pattern found in BE1i and WT starches. Therefore, cassava BE1 and BE2 function differently in storage root starch biosynthesis; silencing of cassava BE1 or BE2 caused various changes to starch physico-chemical properties and amylopectin structure. We also report that remarkably high amylose content in cassava starch has been first obtained in transgenic cassava by silencing of BE2 expression, thus showing a high potential for future industrial utilization.


Introduction
Cassava (Manihot esculenta Crantz) is one of the most important starchy crops worldwide, providing staple food for more than 700 million people in the tropics and subtropics (Hershey, 2017). Compared with other starches (e.g., from maize and potato), pure-white cassava starch contains low levels of fat (0.08-1.54%), proteins (0.03-0.6%), and phosphorus (0.75-4%), making it an excellent starch resource for various industrial applications (Richard et al., 1991;Moorthy, 2002;Balagopalan, 2002;Sharma et al., 2016). Modified starches of cassava have been broadly used in the food, textile, / 2 0 pharmaceutical, paper manufacturing, and other industries . Promotion of cassava-based bioethanol production in China and other South-east Asian countries also made the crop an important source of bioenergy development over the last decade (Marx, 2019).
To meet the growing industrial demands for value-added starches, the native starch with novel physico-chemical property is needed in cassava . Unlike the traditional physical and/or chemical starch modifications, breeders intend to produce new cassava varieties with modified starches to expand their application spectrum through genetic approaches, such as starch-related mutant identification and reverse-genetic/biotechnological approaches (Schwall et al., 2000;Jobling et al., 2002;Raemakers et al., 2005;Ceballos et al., 2007;Zhao et al., 2011;Bull et al., 2018).
Therefore, the development of cassava varieties having novel starches including waxy (amylose-free) and high-amylose starches are important objectives for cassava breeders.
The target genes are mainly involved in starch biosynthesis, such as granule bound starch synthases (GBSS), soluble starch synthases (SSS), starch branching enzyme (SBE or BE), and debranching enzyme (DBE) (Zeeman et al., 2010;Bahaji et al, 2014). More specifically, the SSS, BE, and DBE are involved in amylopectin synthesis, and GBSS is the key enzyme for amylose biosynthesis in plants including cassava (Zhao et al., 2011;Bull et al., 2018).
Since the BE gene was first studied by Mendel with the wrinkled (rugosus, r) phenotype in the pea (Pisum sativum L.) mutants (Bhattacharyya et al., 1990;Tetlow and Emes, 2014) and identified from maize and rice using the well-known amylose-extender (ae -) mutants (Hedman and Boyer, 1982;Mizuno et al., 1993;Kim et al., 1998), its important role in amylopectin biosynthesis has been studied in many plants like potato (Rydberg et al., 2001), maize (Yao et al., 2004), wheat (Regina et al., 2005), and rice (Nakamura et al., 2012). BE has the unique activity of producing α -1, 6 glycosidic bonds, the basis in forming branching chains of amylopectin. BEs usually comprise two classes, BE1 and BE2, which can be distinguished by their amino acid sequences (reviewed by Tetlow and Emes, 2014). Enzymatic activity of BEs is regulated by phosphorylation and protein-protein interactions (Tetlow et al., 2004). Suppression of BE function results in the increase of amylose content in many plants, for example, potato (Schwall et al., 2000), wheat (Sestili et al., 2010), rice (Sun et al., 2017), and sweet potato (Shimada et al., 2006).
Amylose content of native cassava starch is about 21% on average, with a range from 15%-27% (Sánchez et al., 2009). Although cassava mutants having amylose-free or small-granule starch with relatively higher amylose (30%) have been identified from self-pollinated generations in germplasm collections at the International Center for Tropical Agriculture (CIAT, Ceballos et al., 2007Ceballos et al., , 2008, there are still no reports on very-high-amylose cassava. Due to the difficulty and time-consuming processes needed to generate high amylose cassava cultivars via traditional breeding programs, transgenic approaches became an alternative means of achieving this goal, and have proven very effective for the rapid acquisition of cassava lines with novel starches (Raemakers et al., 2005;Ihemere et al., 2006;Zhao et al., 2011;Radhika et al., 2014). Furthermore, GBSS-RNAi transgenic cassava lines have generated many more diversified starches than the gbss mutant AM206-5 from CIAT (Zhao et al., 2011, Rolland-Sabaté et al., 2013, providing us with sound evidence that this approach will be effective.
In this study, we successfully developed transgenic cassava lines with high amylose starch in their storage roots by the silencing of BE1 and BE2 expression. The transgenic cassava shows altered phenotypes and properties of storage starch, indicating a different function for each of these genes in starch biosynthesis. Importantly, these high amylose starches showed B-type XRD diffraction-grams, with higher melting temperatures and potentially useful pasting properties, thus providing very promising raw materials for industrial applications.
Genomic integration of T-DNA in the transgenic cassava plants was confirmed using standard protocols for Southern blot analysis with DIG-labeled PCR products of the hygromycin phosphotransferase gene (HPT) for hybridization (Zhao et al., 2011).
The MeBE1 and MeBE2 RNAi transgenic lines (BE1i and BE2i for short, respectively) and wild type (WT) cassava TMS60444 were propagated by in vitro shoot culture followed by transfer to pots in the greenhouse for macropropagation (16 h/8 h of light/dark, 30 °C/22 °C day/night). In field trials, ten stem cuttings each per transgenic line and WT were planted in the field in early May at the Wushe Plantation for Transgenic Crops, Shanghai, China (31°13948.0099N, 121°28912.0099E). Phenotypic data on the performance of plants was recorded and the storage roots were harvested in early November for agronomic trait evaluation and subsequent experiments.

Transcriptional and translational expression analysis of MeBE1 and MeBE2 genes
Leaves from three plants per line were harvested from 5-month-old plants and then ground in liquid nitrogen for mRNA extraction. To quantify the expression of MeBE1 and MeBE2 genes, real-time qRT-PCR was performed as described previously (Xu et al., 2013). The Total protein was extracted from cassava leaves as described by Ritte et al. (2000) and quantified according to the method of Bradford (Bradford, 1976). Approximately 60 μ g of protein was loaded on an SDS-PAGE gel and Western blotting was performed with rabbit antisera against MeBE1 and MeBE2 proteins, respectively.

Starch extraction and analyses of amylose content and starch granule size
The starch was extracted from storage roots that were harvested from six-month-old cassava plants grown in the field, washed twice in 75% ethanol, and air-dried overnight at 40°C (Zhao et al., 2011). The amylose content of starch was determined using a previously described colorimetric method (Knutson and Grove, 1994). Amylose type III and amylopectin (Sigma A0512 and Sigma 10118, St. Louis, MO, USA) from potato were used to establish the standard curve for amylose quantification.
To observe starch granules by scanning electron microscopy (SEM), isolated starch / 2 0 granules in distilled water were placed or sprinkled on double-sided sticky tape, air-dried, and coated with gold powder. Samples were observed and photographed by SEM (JSM6360lv, JEOL, Tokyo, Japan). The granule size distribution of starch was determined as described (Zhou et al., 2015). A Master-size 2000 laser diffraction instrument (Malvern Instruments Ltd., Worcestershire, UK) was used in wet-well mode. The starch was added to the reservoir and sonicated for 30 s at 6 W until an obscuration value of 12-17% was achieved. The refractive indices used for the water and starch were 1.330 and 1.50, respectively.

Physicochemical property assay of cassava storage starches
The chain length distribution was determined following the protocol previously described by Ritte et al. (2000), using high performance anion exchange chromatography (HPAEC) that was equipped with a CarboPac PA1 column. Sample preparation included boiling 5 mg starch and enzymatic digestion with isoamylase (I5282, Sigma, St. Louis, MO, USA).
Oligosaccharides with a polymerization degree of 4-7 (47265, Sigma, St. Louis, MO, USA) were used as a standard.
X-ray diffraction analysis of the extracted starches was performed using a D8 Advance Bruker X-ray diffractometer (Bruker AXS, Karlsruhe, Germany) and scanned through the 2θ range of 5-60° at a rate of 4° min −1 .The starch pasting properties were analyzed using a rapid viscosity analyzer (model RVA-Super 4; Newport Scientific Pty. Ltd., Australia). The starch was suspended in distilled water (5% w/v, dry weight basis, 25 ml) and tested using a dedicated program for cassava. The temperature of the starch slurry was increased from 30°C to 95°C at a rate of 5°C min −1 and held at 95°C for 6 min, followed by cooling to 50°C at the same rate and maintenance for 10 min. The rotating speed of the paddle remained constant (160 rpm) throughout the analysis, excluding the speed of 960 rpm applied during the first 10 s. The thermal properties of the starch samples were analyzed using a differential scanning calorimeter (DSC, Q2000; TA Instruments Ltd., Crawley, UK) at a temperature scanning range of 30-95°C.

Statistical analysis
Root samples were collected from three independent plants per line and then pooled for further analyses. Data from at least three replicates were presented as the mean ± SD.
Analysis of variance (ANOVA) by Duncan's multiple comparison tests or independent samples Student's t-test was performed using SPSS software, version 17 (SPSS Inc., Chicago, IL, USA). An alpha value of P< 0.05 was considered statistically significant. The average root diameter was 2.8 cm for WT, and 2.3 cm (BE1i-18) to 2.6 cm (BE1i-13) for BE1i lines, and 2.3 cm (BE2i-58) to 2.5 cm (BE2i-24) for BE2i lines (Fig1 a, f). Thus, for both BE1i and BE2i transgenic lines, the total biomass was reduced due to retarded growth of both aerial and subterranean organs.

Down-regulation of MeBE1 and MeBE2 expression in cassava leads to reduced
To verify the phenotypic changes correspond with expression of BEs, transcriptional analysis by real-time qRT-PCR and translational analysis using Western blot assays were carried for MeBE1 and MeBE2 genes using leaf samples from field-grown cassava plants.
Relative transcript levels of BE1 and BE2 genes were drastically reduced by 10 -fold compared to WT in the BE1i or BE2i lines, respectively (Fig. 2a, b). No obvious change was detected for MeBE1 in BE2i lines and MeBE2 in BE1i lines (Fig2a,b). Furthermore, the protein bands of MeBE1 were undetectable in BE1i lines, while the MeBE2 bands in BE2i lines similarly showed a dramatic reduction (Fig. 2c). The protein levels of MeBE1 in BE2i lines and MeBE2 in BE1i lines did not change (Fig2c). These results suggested that the expression levels of BE1 and BE2 genes in cassava were specifically and individually silenced in the BE1i and BE2i transgenic lines, respectively.

MeBE2 but not MeBE1 influences the amylose content in storage roots of cassava
Using colorimetric measurement of amylose-binding stain as a proxy for starch content, 22% amylose was detected in storage starch of WT. Amylose content in BE1i was almost the same as in WT (Fig. 3). However, for BE2i lines, the percentage of amylose in total starch was 50% in BE2i-24, 46% in BE2i-58 and 31% in BE2i-59, a dramatic increase over WT. These data support the hypothesis that BE2 plays key role in amylopectin biosynthesis in cassava storage roots, and that down-regulation of MeBE2 expression by 8 / 2 0 RNAi is an effective way of producing high-amylose starch in cassava.

MeBE1 predominantly affects starch granule size in cassava storage roots
Observation by scanning electron microscopy (SEM) revealed no clear difference in the starch granule shape among WT and BE1i or BE2i transgenic lines. All of these starch granules were a mixture of round, truncated, and dome-shaped granules (Fig. 4a)

MeBE2 rather than MeBE1 leads to dramatic changes in pasting property in cassava storage starch
The BE1i starches exhibited similar pasting patterns to that of WT starch, while the BE2i starches exhibited different patterns (Fig. 5a). More specifically, all three BE1i starches showed increased peak viscosity (PV), hot viscosity (HV), breakdown (BD), and cold viscosity (CV), while decreasing in peak time and pasting temperature (PT) compared to WT starch. Since the value of parameter setback (SB) was correlated with amylose content (Zhou et al., 2015), BE1i starches displayed similar SB values commensurate with the similar amylose content in WT starch (Fig. 3). Meanwhile, due to the increase in amylose content in BE2i-38, BE2i-58, and BE2i-59, starch in these lines showed increased PV, HV, CV, PT, peak time, and SB, but a decrease in BD. These findings indicate that high amylose cassava starch might also be easier to be retrograde than that of WT. HV values represent the resistance to shear thinning. Starch from three BE1i lines (BE1i-13, BE1i-18, BE1i-26) and three BE2i lines (BE2i-38, BE2i-58, BE2i-59) all showed an increased HV over WT. This is especially true for BE2i starches, which ranged from 1129 cP for BE2i-59 to 1136 cP for BE2i-38, which is roughly 2 times higher than that of WT starch (642 cP), suggesting that high amylose cassava starch is superior in terms of resistance to shear thinning. The values of peak time (ranging from 5.5 min for BE2i-59 to 7.0 min for BE2i-58) and PT (ranging from 71.0°C for BE2i-59 to 77.3°C for BE2i-58) in the three BE2i transgenic starches (BE2i-38, BE2i-58, and BE2i-59) were all higher than that of WT starch (5.0 min and 68.9°C, respectively), which indicated that their gelatinization process was slower (Fig. 5a). In addition, the distinctive pasting profile of BE2i-24 starch indicated that its hardness to paste under given conditions renders this line unsuitable for comparison with other starches. However, the severe delay in the / 2 0 gelatinization process suggests the high amylose content may intensively alterthe BE2i-24 starch structure (Fig. 5a).
Thermal property analysis of storage starches showed that the melting process requires higher temperature in BE2i starches and lower temperature in BE1i starches compared to WT (Fig. 5b) showing a lower energy requirement for BE2i starches during the melting process.

MeBE2 function affects the semi-crystallinity property of storage starch via modified amylopectin structure
X-ray diffraction (XRD) analysis of normal cassava storage starch showed typical A-type crystals with one doublet around 17° (2θ) and 2 singlets predominantly around 15° and 23° (2θ) (Defloor et al., 1998, Zhao et al., 2011. Compared to WT storage starch, BE1i storage starches did not distinctly change in crystallinity pattern (Fig. 6a), whereas all BE2i starches underwent pronounced changes in their diffraction grams, including two new singlets emerging around 5°(2θ) and 20 °(2θ), replacement of a doublet with a singlet appearing around 17 ° (2θ), and replacement of a singlet around 23° (2θ) by one doublet.
The peak intensities around 15° (2θ) of BE2i starches were weaker than that of WT and BE1i starches. These XRD features indicated that BE2i storage starches display patterns typical of B-type crystallinity.
Alteration of starch crystal structure and properties in these transgenic starches suggested changes to their amylopectin structure. Chain length distribution analysis of amylopectin by HAPEC-PAD showed differences in the degree of polymerization (DP) between the amylopectins of BE1i or BE2i and WT (Fig. 7). Compared to WT, all three BE1i amylopectins showed a slight increase for long chains of DP >36, and the extent of the difference was less than 0.2% (Fig. 7a). Amylopectins from all three BE2i starches  demonstrated an increase in glucan chains around DP 10-20 and DP >40, but a slight decrease around DP 22-38, with the extent of the difference reaching about 0.5% in BE2i-58 and BE2i-59 (Fig. 7b). The changes in BE2i starches 1 0 / 2 0 reflect the key function of cassava BE2 during amylopectin biosynthesis.

Discussion
High amylose content in starch is a desirable trait in starchy crop breeding and a key characteristic of industrial applications for starch (Jobling 2004). Unlike grains, it is difficult to obtain starch mutants by artificial mutagenesis in root crops like cassava due to their propensity for vegetative propagation, self-pollination, incompatibility, and heterozygosity (Ceballos et al., 2004). Only recently, natural waxy and small granule starch cassava lines were identified from self-pollinated cassava lines in the CIAT cassava breeding program (Ceballos et al., 2007(Ceballos et al., , 2008(Ceballos et al., , 2017. Instead, successful approaches using transgenesis to obtain waxy cassava have been reported in the last decade by down-regulating expression of the GBSSI gene either constitutively or tissue-specifically (Raemakers et al., 2005;Zhao et al., 2011;Koehorst-van Putten et al., 2012), or by deletion mutating GBSSI and PTST1 via CRISPR/Cas9-based mutagenesis (Bull et al., 2018). Very-high-amylose cassava starch has not yet been identified through current traditional breeding programs.
This need therefore provides an opportunity to generate novel cassava germplasm with high-amylose starch content in storage roots by down-regulating the BE genes that are responsible for amylopectin biosynthesis. In this report, we successfully generated transgenic cassava with very-high-amylose content up to 50% by post-transcriptional silencing of MeBE2, but not MeBE1. Starch physico-chemical analyses confirmed the drastic alteration of starch properties predominately in BE2i starches, indicating the importance of MeBE2 in amylopectin biosynthesis in cassava storage roots.
In higher plants, the starch branching enzymes can be classified into SBEI and SBEII subfamilies which play different roles in amylopectin biosynthesis through distinct differences in their polymer substrate preferences (Tetlow and Emes, 2014). In cassava, the exact functions of BE1 and BE2 are still unclear, although their tissue and spatial expression patterns have been studied under different growth and environmental conditions (Salehuzzaman et al., 1992;Bauama et al., 2003;Pei et al., 2015), or in response to day/night cycle ( Supplementary Fig. S1). Our study showed that although down-regulation of MeBE1 or MeBE2 expression in cassava exhibits similar remarkable phenotypic changes and altered circadian oscillation of sugars ( Supplementary Fig. S2 Silencing MeBE1 expression in cassava has no effect on the amylose content of storage roots, indicating a different role for MeBE1 in cassava starch biosynthesis. Overall, the BE1i starch did not have dramatic changes in starch physico-chemical properties compared to WT except for enlargement of the starch granule size; a similar observation of no significant effects on starch has been reported in maize lacking SBE1 activity (Blauth et al., 2002). Rydberg et al. (2001) reported that potato BE2 was more active than BE1 on an amylopectin substrate, whereas BE1 was more active than BE2 on an amylose substrate. Further investigation is required to decipher the enzymatic process of cassava BE1 protein and its interactions with other starch biosynthetic enzymes (Tetlow et al., 2015). Nevertheless, similar reduction in biomass for both BE1i and BE2i plants indicated that MeBE1 and MeBE2 are equally important for cassava growth and development.
Amylose content is a dominant factor affecting applied properties of starch (Van Hung et al., 2006;Zhou et al., 2015). Important and distinct changes in high-amylose starches from BE2i lines include changes in the paste properties, thermograms, amylopectin chain lengths, and crystallinity. High-amylose cassava starches have increased PV, HV, CV, PT, peak time, and SB; however, BD is decreased, indicating a high capacity for resistance to shear thinning and a slower gelatinization process. These starches also need a higher temperature during the melting process. The transformation of crystal features of high-amylose starch from the A-type to the B-type concurrent with an increase of amylose content in cassava showed a similar trend to that of sweet potato (Zhou et al., 2015), showing three significant alterations at 5 ° (2θ), 17° (2θ) and 20 ° (2θ) in XRD analysis (Fig.6).
The chain length distribution can reflect the branching pattern of amylopectin, which is related to the activity of BEs (Tester et al., 2004;Wang et al., 2017) (Table II), as reported in other high-amylose starches (Jane et al., 1999;Kong et al., 2015;Zhou et al., 2015). More investigations are required to determine the mechanisms of BE1 and BE2 participation in cassava starch biosynthesis.
In conclusion, our study reports the generation of high-amylose cassava by down-regulating expression of the BE2 gene. The high-amylose cassava starch showed       transgenic plant lines. Dx10, Dx50, and Dx90, the projected equivalent diameter below which 10%, 50% and 90% of the total volume of all particles analyzed is represented.