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‐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). 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 (Li et al., 2017; Rodríguez‐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/plantcare/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 (Xie et al., 2017) 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 Agrobacterium‐mediated transformation of TFB.

From transgenic (T₁) segregating families, we PCR‐selected transgene‐free plants and further identified 23 homozygous mutant T₂‐lines (Figure 1b). These lines had base insertions or deletions at the targets, or fragment deletions (or a 36‐bp substitution) between the paired targets, which removed the putative CREs, respectively (Figure 1b). In two lines (UISS‐1, UISS‐6) by the 5’UISS‐editing, the intronic splicing site (GT) was deleted (Figure 1c). We named some of these Wx mutant alleles that showed obviously down‐regulated expression largely affecting AC (see below) as follows: Wx^a‐dE (Endosperm‐box deleted in T1T2‐2 line), Wx^a‐dU (unknown element deleted in T3T4‐2), Wx^a‐dA (A‐box deleted in T5T6‐5), Wx^a‐dC1, Wx^a‐dC2 and Wx^a‐dC3 (CAAT‐box deleted in T7T8‐4, T7T8‐5 and T7T8‐6, respectively), and Wx^dS1, Wx^dS2 and Wx^dS3 (splicing‐site deleted/impaired in UISS‐1, UISS‐2 and UISS‐6) (Figure 1b). We selected a T10‐edited mutant (TFBg) for analyses.

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 transcript (Wx^dS2‐1 with 58.3%) (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’UISS‐edited 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 (Li et al., 2019). However, our strategy of editing exon/intron border sequences within 5’UTRs can quantitatively regulate gene activity and phenotypic performance (see below).

Next, we measured AC and GC of these lines. The lines without the CAAT‐box showed reduced ACs, from 24.6 % in TFB to moderate levels of 19.6% (T7T8‐4), 18.1% (T7T8‐5) and 17.8% (T7T8‐6) (Figure 1e). While ACs of UISS‐1, UISS‐2, UISS‐6 were 10.6%, 9.8% and 11.5%, respectively (Figure 1e); these lines are novel valuable ‘soft rice’ germplasms. TFBg had 2.4% AC, a typical glutinous rice. Some other edited lines also produced slightly reduced ACs, such as 23.8% (T3T4‐2), 22.8% (T5T6‐5) and 22.9% (UISS‐5). These AC variations were related to the corresponding Wx mRNA levels. In accordance with the AC reductions, T7T8‐4, T7T8‐5, T7T8‐6, UISS‐1, UISS‐2 and UISS‐6 showed increased GC values (62–83 mm, comparing to 56 mm of TFB) (Figure 1e). In addition, the polished grain appearances of T7T8‐4, T7T8‐5 and T7T8‐6 were similar to that of TFB (Figure 1f). However, due to the lower ACs in UISS‐1, UISS‐2 and UISS‐6, their polished grains had lower endosperm transparency with milky‐white appearances (Figure 1f).

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.

Finally, we observed that the major agronomic traits (1000‐grain weight, grain length, grain width, plant height and plant morphology) of these edited lines were similar to TFB, except for slightly decreased 1000‐grain weight in UISS‐1, UISS‐2 and UISS‐6 (93%–95% of TFB).

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.

在谷物作物中，谷物淀粉由不同比例的直链淀粉和支链淀粉组成，它们决定了烹饪和食用质量。直链淀粉的合成受Waxy（Wx）基因控制，该基因编码颗粒结合的NDP-葡萄糖-淀粉葡糖基转移酶（Shure等，1983）。在稻（稻属），天然的改变活动蜡质等位基因调节不同直链淀粉含量（AC），凝胶稠度（GC）和谷物淀粉的粘度粘贴; 这些因素共同影响谷物的外观，蒸煮/进食质量和淀粉的物理特性（Zhang等，2019）。WX^一是很强的等位基因主要分布在籼稻（一个栽培稻亚种）栽培品种高的AC（25％-30％）（王等人。，1995）。尽管蜡质^b，主要呈现在粳稻（另一亚种）品种，是一种弱等位基因生产中等的AC（15-18％）（一色等人。，1998）。通常，具有较高AC和较低GC值的米粒的进食质量较差，而具有中等AC（15–20％）和较高GC值（60–80 mm）的米粒对大多数消费者而言味道更好。使用连续的回交方法，Wx ^b可以渗入in品种提高了谷物品质。然而，传统的育种方法耗时且难以打破具有不良性状的紧密连锁亲缘。

我们先前使用CRISPR / Cas9靶向Wx编码区以产生糯米（Ma et al。，2015）。但是，这种功能敲除策略仅产生无效的基因等位基因，并且当靶向Wx时，通常会产生糯系。通过基因组编辑产生性状的各种定量变化的研究很少。为了迅速提高稻米品质，在这里，我们开发了CRISPR / Cas9编辑策略，以产生新的蜡质等位基因其表达的定量调节生产各种ACS，使用精英籼品种TianFengB（TFB）作为测试。TFB因其高产表现而在杂交水稻育种中被广泛使用，但由于较高的AC（约25％）和较低的GC（56毫米，见下文），其谷物品质（及所得杂交种）质量较差。

通过基因组编辑破坏启动子序列可能会改变农艺性状（Li等，2017 ;Rodríguez-Leal等，2017）。因此，我们选择一个ca。TFB中用于定位的Wx ^a的2.0kb上游序列，其中包含一个0.9kb的启动子调控区和一个1.1kb的内含子的5'非翻译区（UTR）（图1a）。第一种策略基于转录调控，因此我们使用Plant-CARE（http://bioinformatics.psb.ugent.be/webtools/plantcare/html/）分析了启动子序列，并确定了三个推定的顺式调控元素（CRE），胚乳盒，A盒和CAAT盒。我们使用CRISPR‐GE（Xie et al。，2017）在该区域（图1a）设计了四对靶标（T1-T8 ）进行多重编辑。我们探索的第二种策略是通过靶向Wx ^a的5'UTR内含子剪接位点（5'UISS）和目标T9来进行转录后调控，以改变内含子剪接模式和效率。此外，还进行了编码外显子编辑（目标T10）来生产糯米。使用我们的CRISPR / Cas9系统（Ma et al。，2015），我们准备了用于双靶或单靶编辑的6种构建体（图1a，b），并将其用于农杆菌TFB介导的转化。

从转基因（T ₁）分离家族中，我们进行PCR选择不含转基因的植物，并进一步鉴定出23个纯合突变体T ₂系（图1b）。这些品系在靶标处有碱基插入或缺失，或在成对的靶标之间存在片段缺失（或36 bp取代），分别去除了假定的CRE（图1b）。通过5'UISS编辑在两行（UISS-1，UISS-6）中，删除了内含子剪接位点（GT）（图1c）。我们将其中一些明显显着下调表达的Wx突变等位基因命名，这些表达明显影响了AC（见下文），如下所示：Wx ^a-dE（胚乳盒在T1T2-2品系中删除），Wx ^a-dU（未知元素已在T1T2-2中删除）。 T3T4-2），WX^一个-DA，（A-框T5T6-5删除）WX^一个-DC1，WX^一个-DC2和WX^一个-DC3（CAAT盒在T7T8-4，T7T8-5和T7T8-6分别删除，），和Wx ^dS1，Wx ^dS2和Wx ^dS3（在UISS-1，UISS-2和UISS-6中拼接站点已删除/受损）（图1b）。我们选择了T10编辑的突变体（TFBg）进行分析。

为了研究5'UISS编辑的系的剪接模式，我们进行了逆转录（RT）-PCR和cDNA测序。UISS-1（Wx ^dS1），UISS-2（Wx ^dS2）和UISS-6（Wx ^dS3）产生了多个不同频率的交替或非典型剪接的转录本，类似于粳稻品种KY131中的Wx ^b（图1c）。三个成绩单（Wx ^dS1-3，Wx ^dS3-4和Wx ^b‐4）被发现保留了非剪接的内含子。这些主要的选择性剪接事件具有大小差异，已通过凝胶电泳确认（图1d）。显然，在UISS-1和UISS-6中剪接位点的缺失导致内含子剪接模式的改变（并抑制了某些转录本的剪接）。但是，在UISS-2中5'UISS附近的2 bp缺失也产生了一个替代的转录本（Wx ^dS2-1占58.3％）（图1c），这表明该2 bp缺失可能会改变影响mRNA前的构象。正确的内含子剪接。

然后，我们使用定量RT-PCR（qRT-PCR）来测量发育中的胚乳中这些品系的成熟mRNA水平。在T7T8-4（Wx ^a-dC1），T7T8-5（Wx ^a- dC2）和T7T8-6（Wx ^a-dC3）），表达水平分别下调至TFB的37.4％，32.7％和24.9％（图1e）。胚乳盒（T1T2-2），未知元件（T3T4-2）和A盒（T5T6-5）缺失的品系也显示出mRNA的显着降低（TFB的85.2％，67.4％和60.5％） ， 分别）。然而，在目标处具有碱基变异的其余系几乎没有表达变化。这些结果证实了这些推定的CRE对转录的调控作用。此外，所有5'UISS编辑的品系均表现出降低的mRNA水平。特别是Wx ^dS1（UISS-1），Wx ^dS2（UISS-2）和Wx ^dS3（UISS-6）只有大约 TFB的mRNA水平为10％，表明这些错误剪接，非典型剪接和非剪接可能会大大降低转录本的稳定性。编码区内的内含子剪接位点的碱基编辑可能引起异常的mRNA剪接和基因功能敲除（Li等人，2019）。但是，我们在5'UTR中编辑外显子/内含子边界序列的策略可以定量调节基因活性和表型表现（见下文）。

接下来，我们测量了这些管线的AC和GC。没有CAAT-box的生产线显示AC降低，从TFB中的24.6％降至中度水平19.6％（T7T8-4），18.1％（T7T8-5）和17.8％（T7T8-6）（图1e）。而UISS-1，UISS-2，UISS-6的AC分别为10.6％，9.8％和11.5％（图1e）；这些系是新颖的有价值的“软稻”种质。TFBg具有2.4％的AC（典型的糯米）。其他一些经编辑的行也产生了稍微减少的AC，例如23.8％（T3T4-2），22.8％（T5T6-5）和22.9％（UISS-5）。这些AC变化与相应的Wx相关mRNA水平。根据AC减少量，T7T8-4，T7T8-5，T7T8-6，UISS-1，UISS-2和UISS-6显示出增加的GC值（62-83 mm，而TFB为56 mm）（图1e） ）。此外，T7T8-4，T7T8-5和T7T8-6的抛光晶粒外观与TFB相似（图1f）。但是，由于UISS-1，UISS-2和UISS-6中的AC较低，它们的抛光谷物的胚乳透明度较低，呈乳白色外观（图1f）。

我们进一步使用快速粘度分析（RVA）来评估淀粉质量（Fitzgerald等，2003）。编辑行的RVA粘度指数与其AC有关，变化程度不同（图1g）。其中，PT7T8-4，PT7T8-5和PT7T8-6表明显著降低粘度指数，更接近的籼稻HHZ（携带蜡质^b具有高谷物质量用17.1％AC）。

最后，我们观察到这些编辑品系的主要农艺性状（1000粒重，籽粒长度，籽粒宽度，植物高度和植物形态）与TFB相似，只是在UISS-1，UISS中1000粒重略有减少‐2和UISS‐6（占TFB的93％–95％）。

总而言之，我们开发了高效的CRISPR / Cas9介导的启动子/ 5'UISS工程设计策略，以产生具有精细调整的基因表达活性的转录和转录后调控的新的定量性状等位基因。我们希望这些具有AC和GC水平的谷物改良品系的应用及其在杂交水稻育种中的开发将为稻米产品提供更好的质量，以满足消费者的喜好。由于CRE和5'UTR内含子存在于许多基因中，因此我们的研究提供了一种有望改善作物和其他生物重要性状的育种方法。