Quantitative regulation of Waxy expression by CRISPR/Cas9‐based promoter and 5'UTR‐intron editing improves grain quality in rice

In cereal crops, grain starches are composed of different proportions of amylose and amylopectin, which determine the cooking and eating qualities. The amylose synthesis is controlled by the Waxy (Wx) gene encoding a granulebound NDP-glucose-starch glucosyltransferase (Shure et al., 1983). In rice (Oryza sativa L.), the varied activities of natural Wx alleles regulate different amylose contents (AC), gel consistency (GC), and pasting viscosity of grain starches; these factors together influence the grain appearance, cooking/eating quality, and starch physical characters (Zhang et al., 2019).

In cereal crops, grain starches are composed of different proportions of amylose and amylopectin, which determine the cooking and eating qualities. The amylose synthesis is controlled by the Waxy (Wx) gene encoding a granule bound NDP-glucosestarch glucosyltransferase (Shure et al., 1983). In rice (Oryza sativa L.), the varied activities of natural Wx alleles regulate different amylose contents (AC), gel consistency (GC) and pasting viscosity of grain starches; these factors together influence the grain appearance, cooking/eating quality and starch physical characters (Zhang et al., 2019). Wx a is a strong allele mainly distributing in indica (an O. sativa subspecies) cultivars producing high ACs (25%-30%) (Wang et al., 1995). While Wx b , presenting mainly in japonica (another subspecies) cultivars, is a weak allele producing moderate ACs (15-18%) (Isshiki et al., 1998). Generally, rice grains with higher ACs and lower GC values have poor eating quality, while those with moderate ACs (15-20%) and higher GC values (60-80 mm) give better taste for most consumers. Using the successive backcrossing methods, Wx b can be introgressed into indica varieties to improve the grain quality. However, the traditional breeding methods are time consuming and difficult to break close linkage drags with undesirable traits.
We previously employed CRISPR/Cas9 to target the Wx coding region to generate glutinous rice (Ma et al., 2015). However, this kind of function-knockout strategy produces only null gene alleles, and when Wx is targeted generally glutinous lines are generated. Studies on generating various quantitative variations of traits by genome editing are rare. To rapidly improve rice grain quality, here, we developed CRISPR/Cas9 editing strategies to generate new Wx alleles producing various ACs by quantitative regulation of its expression, using an elite indica variety TianFengB (TFB) as a test. TFB is a widely used parent in hybrid rice breeding for its high-yield performance, but its grain quality (and of the resultant hybrids) is poor due to higher AC (ca. 25%) and lower GC (56 mm; see below).
Disruption of promoter sequences by genome editing may change agronomic traits Rodr ıguez-Leal et al., 2017). Therefore, we selected a ca. 2.0-kb upstream sequence of Wx a in TFB for targeting, which contains a 0.9-kb promoter regulatory region and a 1.1-kb intron-containing 5'untranslation region (UTR) (Figure 1a). The first strategy is based on transcriptional regulation, thus we analysed the promoter sequence using Plant-CARE (http://bioinformatics.psb.ugent.be/webtools/plantca re/html/) and identified three putative cis-regulatory elements (CREs), Endosperm-box, A-box and CAAT-box. We designed four pairs of targets (T1-T8) in this region ( Figure 1a) using CRISPR-GE  for multiplex editing. The second strategy we explored is for post-transcriptional regulation by targeting the 5'UTR intronic splicing site (5'UISS) of Wx a with a target T9 to alter the intron-splicing pattern and efficiency. In addition, a coding-exon editing (with a target T10) was done to produce glutinous rice. Using our CRISPR/Cas9 system (Ma et al., 2015), we prepared six constructs for the double-target or single-target editing ( Figure 1a, b), and used them for Agrobacteriummediated transformation of TFB.
To investigate the splicing patterns of the 5'UISS-edited lines, we performed reverse transcription (RT)-PCR and cDNA-sequencing. UISS-1 (Wx dS1 ), UISS-2 (Wx dS2 ) and UISS-6 (Wx dS3 ) generated multiple alternatively or atypically spliced transcripts with various frequencies, similar to Wx b in a japonica variety KY131 (Figure 1c). Three transcripts (Wx dS1 -3, Wx dS3 -4 and Wx b -4) were found to retain the non-spliced intron. These major alternative splicing events with size differences were confirmed by gel electrophoresis ( Figure 1d). Obviously, the splicing-site deletion in UISS-1 and UISS-6 resulted in the altered intron-splicing patterns (and suppressed splicing of some transcripts). However, the 2-bp deletion near the 5'UISS in UISS-2 also produced an alternative   (Figure 1c), suggesting that this 2-bp deletion might change the pre-mRNA conformation affecting correct intron splicing.
Then, we used quantitative RT-PCR (qRT-PCR) to measure mature mRNA levels of these lines in developing endosperm. In T7T8-4 (Wx a-dC1 ), T7T8-5 (Wx a-dC2 ) and T7T8-6 (Wx a-dC3 ), the expression levels were down-regulated to 37.4%, 32.7% and 24.9% of TFB, respectively (Figure 1e). The lines with deletions of the Endosperm-box (T1T2-2), unknown element(s) (T3T4-2) and A-box (T5T6-5) also showed significantly decreased mRNA (85.2%, 67.4% and 60.5% of TFB, respectively). However, the rest lines with base variations at the targets had little or no expression changes. These results verified the regulatory roles of these putative CREs on transcription. In addition, all the 5'UISSedited lines exhibited decreased mRNA levels; especially, the lines Wx dS1 (UISS-1), Wx dS2 (UISS-2) and Wx dS3 (UISS-6) had only ca. 10% mRNA levels of TFB, suggesting that these mis-splicing, atypical splicing and non-splicing might largely reduce the transcript stability. The base editing of intronic splicing sites within coding regions could cause aberrant mRNA splicing and gene function knockout . However, our strategy of editing exon/intron border sequences within 5'UTRs can quantitatively regulate gene activity and phenotypic performance (see below).
We further used Rapid visco analysis (RVA) to assess the starch quality (Fitzgerald et al., 2003). The RVA viscosity indexes of the edited lines varied to various degrees relating to their ACs (Figure 1g). Among them, PT7T8-4, PT7T8-5 and PT7T8-6 showed significantly decreased viscosity indexes, closer to those of indica HHZ (carrying Wx b with 17.1% AC) that has high grain quality.
In summary, we developed high-efficient CRISPR/Cas9-mediated promoter/5'UISS-engineering strategies for generating new quantitative trait alleles with fine-tuned transcriptional and post-transcriptional regulations of gene expression activity. We expect that application of these grain-improved lines having desirable AC and GC levels and their exploitation in hybrid rice breeding will provide rice products with better quality to meet consumer's preferences. As CREs and 5'UTR introns are present in many genes, our study provides a promising breeding method for improvement of important traits in crops and other organisms.