Production of novel beneficial alleles of a rice yield‐related QTL by CRISPR/Cas9

In crop plants, many beneficial alleles of yield-related quantitative trait loci involve changes in transcription, rather than changes in protein coding sequence. However, these alleles have multiple mutations and discerning the causal mutation and then recapitulating this mutation in elite crop cultivars has remained challenging. Now, CRISPR-Cas9 approaches allow researchers to address both issues. Our study of a locus involved in resistance to lodging demonstrated that saturation editing of the qSCSA3-1/OsTB1/SCM3 region led to loss-of-function, normal-like, and gain-of-function plant types, based on OsTB1 expression and stem cross-section area. One gain-of-function allele recapitulated a previously described beneficial allele of OsTB1 carrying a TGTG insertion in the 5' region of OsTB1 and enhanced OsTB1 expression at the inflorescence formation stage. These studies indicate that genome editing technology, combined with information on agronomically important genes, can accelerate genetic improvement of crops.

Breeding high-yield crop cultivars has improved agronomic performance for key grain crops (Hirano et al., 2017). However traditional breeding takes years. Molecular genetic studies have identified beneficial trait-associated alleles in elite cultivars providing a platform for gene-editing techniques such as the clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated 9 (Cas9) system (Rothan et al., 2019).
Most current 'super rice' varieties with high yields have several beneficial agronomic traits. For example the variety Liang-You-Pei-Jiu has strong culms for lodging resistance and large panicles for high yield. Our previous study using recombinant inbred lines (RILs) from a cross between 93-11 and PA64 identified 43 quantitative trait loci (QTLs) associated with many agronomic traits including heading date spikelet number per panicle and grain shape (Gao et al., 2013). However no QTL for culm strength or panicle size was detected.
In addition to affecting SCSA introgression of the NPB allele resulted in a decrease in bending stress stem area and number of spikelets per panicle and an increase in tiller number but had no effect on plant height ( Figure 1f). These observations indicated that OsTB1 has pleiotropic effects on different traits. Therefore we examined the effect of OsTB1 using mutant and complemented plants (Figure 1g). The null mutant of OsTB1 fc1-2 showed lower bending stress smaller panicles and increased tiller number compared with rescued plants expressing OsTB1 from 93-11 or Kasalath ( Figure 1h). These observations confirmed that qSCSA3-1 is allelic to SCM3/OsTB1 and has pleiotropic effects.
We compared the sequence of OsTB1 among rice cultivars NPB (japonica) Koshihikari (japonica) and four varieties with large SCSA: 93-11 (indica) Kasalath (indica) Zhongchao123 (japonica) and Chugoku117 (indica) and found polymorphisms in the 5'flanking coding sequence (CDS) (Figure 1i,j). Interestingly the four varieties with large SCSA contained a TGTG insertion at + 219 in the 5'-noncoding region suggesting that this insertion might be important for OsTB1 expression and consequently affect SCSA.
Based on this prediction we introduced mutations in the promoter and 5'-noncoding regions of SCM3/OsTB1 in NPB using CRISPR-Cas9. We designed six single-guide RNAs (sgRNAs) four targeting the promoter sequence and two sgRNA5 and 6 targeting the region proximal to the TGTG insertion in the 5'flanking CDS. These six sgRNAs were integrated into one plasmid and transformed into NPB using Agrobacterium. We obtained 23 plants carrying 9 different mutations (Figure 1j). Relative to wildtype (WT) NPB Type 1 had more tillers and smaller culms and panicles; Type 2 had similar phenotypes to WT; Type 3 had fewer tillers and larger culms and panicles (Figure 1k). The Type 1 plants 67-1-1 63-1-6 71-3-2 and 66-2-3 were similar to the null mutant of OsTB1 fc1-2 (Minakuchi et al., 2010). Indeed 71-3-2 and 66-2-3 contained deletions in the 5'-noncoding region of OsTB1. OsSPL14/IPA1 positively regulates OsTB1 expression through binding to a GTAC motif (Lu et al., 2013) and this motif was deleted in these plants likely resulting in the null phenotypes.
The Type 3 plants 83-4-1 and 76-1-2 mimicked the phenotypes of NIL SCM3 (Yano et al., 2015) and 93-11 suggesting that mutations in these plants enhance OsTB1 expression. Indeed these plants showed higher OsTB1 expression than the control whereas the Type 1 and Type 2 plants showed reduced and unchanged expression levels respectively (Figure 1l). 76-1-2 had a TGTG insertion which also occurs in varieties with large SCSA (Figure 1i j) confirming that the TGTG insertion enhances OsTB1 expression. Although 83-4-1 had mutations at six sites in the 5'flanking CDS region it showed a gain-of-function phenotype (Figure 1k). Thus different sequence(s) in the 5'-flanking CDS region in WT may repress or enhance transcription. In 84-5-1 a 36-bp deletion around the TGTG insertion did not increase OsTB1 expression suggesting that the TGTG insertion did not disrupt a repressive site. Instead the TGTG insertion may create a binding site for transcription activator(s) hypotheses that will require further validation.
For Type 2 plants which showed similar phenotypes to the control plants (Figure 1k) OsTB1 expression did not differ from the control although they had various deletions and singlenucleotide polymorphisms upstream of the gene (Figure 1j). These results demonstrated that targeted editing of OsTB1 cisregulatory elements could produce alleles having different expression levels and phenotypes (Figure 1l m).
The application of CRISPR-Cas9 for studying cis-elements has been discussed but a detailed strategy has not been established in rice (Rodriguez-Leal et al., 2017). In this study we introduced mutations in cis-regulatory sequences which led to alterations in gene expression and phenotypes and reproduced beneficial alleles in new rice varieties. QTL studies have identified beneficial alleles in rice elite germ plasm including alleles affecting gene expression. For example the molecular mechanism for major QTLs controlling plant structure such as GN1A IPA1/WFP NUMBER OF GRAINS 1 (NOG1) FZP1/qSRN7 and also SCM2/APO1 as mentioned above depends on the difference in the expression level of these genes (Huo et al., 2017). These spontaneous alleles have been introgressed into elite varieties by breeding which is timeconsuming and labour-intensive. CRISPR-Cas9 however easily produced alleles with similar expression levels as shown for OsTB1. Furthermore we may be able to isolate desirable alleles with more appropriate levels of expression for optimizing breeding targets. In this regard saturation editing of known  , 18, 1987-1989 genes associated with agronomic traits of interest may accelerate molecular design of desirable traits.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article.

Figure S1
Mapping of SCSA (Stem Cross-Section Area) QTLs CSSLs. Figure S2 Identification of hygromycin in different editedplants. Table S1 Primer sequences used in this study. Data S1 Materials and Methods.