Fine‐tuning the amylose content of rice by precise base editing of the Wx gene

The genetic diversity and phenotypic variability of crop agronomic traits is valued by breeders for their benefits in crop breeding but are limited for most target traits. Genome editing has proved to be a powerful tool for quick and efficient creation of continuous beneficial genetic variation for crop breeding (Eshed and Lippman, 2019). The rice Waxy (Wx) gene (LOC_Os06g04200) encodes granule-bound starch synthase I (GBSSI), which determines the amylose content (AC) of endosperm by controlling amylose synthesis. This is one of the major contributors for the eating and cooking quality (ECQ) of rice (Li et al., 2016), an attribute that is receiving increased attention in society because of the improvement in people's living standards. Rice AC ranges from 0 to ~30% depending on the presence of different Wx alleles, with Wxa(relatively high AC of more than 20%) and Wxb (intermediate AC of 14 to ~18%) being the major alleles found in the indica and japonica varieties, respectively (Teng et al., 2012). Amino acid changes in the Wx/GBSSI protein can affect the AC of rice grain, as in the well-known 'soft rice' varieties (AC of 7%–10%) with genotypes Wxop/hp, WxmqorWxmp (Zhu et al., 2015), which all have non-synonymous mutations in the N-terminal domain of Wx/GBSSI (Momma and Fujimoto, 2012). As rice varieties with moderately low AC (<12%), that is the 'soft rice' varieties, have become more popular commercially and for breeders (Li and Gilbert, 2018), both traditional and molecular breeding approaches including CRISPR/Cas9-mediated gene knockout (Ma et al., 2015; Zhang et al., 2018) have been used to mutate Wx to reduce the AC of rice grain. However, only a limited number of Wx mutants have been generated, far fewer than needed to meet the diverse demands of ECQ. We hypothesized that the AC of rice grain could be fine-turned by generating a series of novel amino acid substitution(s) close to the 'soft rice' allele responsible sites (such as the residues 158th inWxmq or Wxmp, 191th inWxmq and 165th in Wxop/hp allele) in the N-terminal domain of the Wxb allele by state-of-the-art base editing. Based on the requirements of cytidine base editors (CBEs) (Zong et al., 2017), we designed three sgRNAs targeting the third (target site1, TS1), fourth (target site 2, TS2) or fifth (target site3, TS3) exons of Wxb (Figure 1a), which were all close to the mentioned 'soft rice' allele responsible sites. The three sgRNAs were cloned into vector pH-nCas9-PBE to generate vectors PBE-TS1, PBE-TS2 and PBE-TS3, respectively. The resulting plasmids were individually introduced into the japonica rice cultivar Nipponbare (NIP) by Agrobacterium-mediated transformation. A total of 5, 10 and 7 independent T0 transgenic lines, respectively, were generated, and 2, 5 and 2 representative edited lines (Figure 1b) were taken to the T1 generation; only T-DNA-free homozygous individuals were then chosen and analysed in detail. We observed a variety of T1 mutation types depending on the number and position of the base changes and substitutions within the editing window; these reflected the changes present in the parental lines, suggesting that the T0alleles were faithfully transmitted to the next generation (Figure 1b). Using TS1, one line, Wxm5 (from T0 line B7-2/6), carrying a C2, 3, 5-to-T2, 3, 5 transition that led to P124F and R125W mutations was obtained; using TS2, four lines including Wxm6 (from T0 line B6-29, a G6, 7-to-A6, 7 transition leading to a G159K mutation), Wxm7 (from T0 line B2-25, a G6-to-C6 transversion leading to a G159A mutation), Wxm8 (from T0 line B2-25, with a G1-to-A1 transition and G6-to-C6 transversion, leading to G159A and D161N mutations) and Wxm9 (from T0 line B1-68, a G4-to-T4 transversion and G6-to-A6 transition, leading to G159E and V160F mutations) were identified; in TS3, two lines including Wxm10 (from T0 line B2-21, a C5, 6-to-T5, 6 transition, leading to a T178I mutation) and Wxm11 (from T0 line B2-21, a C5-to-G5 transversion and C6-to-T6 transition, leading to a T178S mutation) were obtained (Figure 1c). In addition, for all seven T1 edited lines (Wxm5-Wxm11), we failed to find any mutations in any of the potential off-target sites (Figure 1d). To determine the effect of these mutations on AC, we measured the apparent amylose contents (AACs) of grains from the seven mutant lines (Wxm5-Wxm11), NIP (Wxb) and a 'soft rice' control Nangeng9108 (NG9108) (Wxmp) (Figure 1e). Notably, Wxm5 had an AAC (1.4 ± 0.2%) as low as the glutinous rice. The AACs of Wxm6 (11.9 ± 0.1%), Wxm7 (11.3 ± 0.1%), Wxm10 (9.8 ± 0.2%) and Wxm11 (7.9 ± 0.1%) were all moderately but significantly lower than that of NIP (14.4 ± 0.2%), but comparable with that of NG9108 (9.6 ± 0.2%). The AACs of Wxm8 (5.8 ± 0.2%) and Wxm9 (4.2 ± 0.1%) lay between those of NG9108 and Wxm5. The GBSSI activities in developing seeds of the Wx-edited lines 10 days after flowering ranged from 231.5 ± 16.5 to 712.1 ± 54.1 nmol/g/min (Figure 1f), all lower than in NIP. The reduced GBSSI activities are likely due to the lower total amount of GBSSI protein (Figure 1g). These results demonstrate that amino acid substitutions in TS1-TS3 indeed can reduce the total GBSSI abundance and activity and decrease the AC of seeds. In general, the quality of the appearance of the milled rice (especially the transparency of the grain) is negatively correlated with AAC (Li et al., 2018). The milled rice grains of the 'soft rice' varieties with 7%–10% AAC (e.g. NG9108) are semi-translucent while the glutinous rice grains with AAC < 2% are opaque. We compared the appearance of the milled rice grains (10% moisture) of the seven Wx-edited lines (T2 generation) with those of NIP and NG9108. As indicated in Figure 1h, the milled grains of Wxm5 and Wxm9 were opaque and glutinous rice-like, consistent with their low AAC. The milled grains of Wxm8, and Wxm11 were semi-translucent like those of NG9108. Interestingly, the appearance of the milled grain of Wxm6, Wxm7 andWxm10, with AACs of 9.8%–11.9%, tended to be like that of NIP rather than NG9108, being almost transparent rather than semi-translucent, indicating that we successfully generated novel germ plasms with moderately reduced AC (~10%) but without affecting the quality of the appearance of the milled rice. The results achieved in NIP were confirmed in two other japonica varieties, Jingeng818 (JG818) and Suijing18 (SJ18), by generating T-DNA-free and homozygous T1 mutants like those observed in NIP, for exampleWxm5, Wxm6, Wxm7 and Wxm10 (Figure 1i), indicating that the strategy used in this study is reliable and can be used to fine-tune AC in elite japonica varieties. In summary, we have used a base-editing system to create a series of mutants with AACs of 1.4%–11.9% and have achieved the goal of fine-tune rice AC over the range of 0%–12% to enrich the range of breeding materials available to breeders. Furthermore, we speculated that base-editing other sites (e.g. the C-terminal domain) and/or base editing of the varieties with other Wx alleles (e.g. Wxa) could be available to further extend the range of AC. This study shows that it is possible to obtain a range of mutations by substituting many individual amino acids in the critical domains of genes controlling economically important traits. This provides an important new strategy for crop breeding. This work was supported by grants from the National Transgenic Science and Technology Program (2019ZX08010-003), the National Natural Science Foundation of China (31701511), the National Science Foundation of Jiangsu Province (BK20170610), the National Key R&D Program of China (2018YFA0900600) and State Key Laboratory of Plant Cell and Chromosome Engineering (PCCE-KF-2020-01). The authors have submitted a patent application based on the results reported in this paper. J. Y., C. G. and Y. X. designed the research; Y. X., Q. L., X. L., F. W., Z. C., J. W., W. L., F. F., Y. T., Y. J., X. W. and R. Z. performed the research; Q. B. and C. G. contributed to the writing; and Y. X. and Q-H. Z. wrote the manuscript.