Common wheat has a large genome with three subgenomes (A, B and D), making it challenging to create mutations at multiple genomic sites simultaneously. The CRISPR/Cas9 system offers a game‐changing tool for editing crop genomes (Chen et al., 2019). Three main strategies have been developed to produce multiple single‐guide RNAs (sgRNAs), including the conventional multiplex system with tandem repeats of separate U3 or U6 promoters (TRSP), the tRNA‐processing system (Xie et al., 2015) and the ribozyme‐processing system (Gao and Zhao, 2014). Although CRISPR/Cas9‐mediated genome editing was previously achieved by biolistic (Wang et al., 2014, 2018) and Agrobacterium transformation (Zhang et al., 2019a), a most efficient CRISPR/Cas9 system for multiplex editing in wheat remains elusive. To address this important question, we designed three multiplex editing constructs corresponding to these three systems, based on the pBUE411 vector (Figure 1a). For the TRSP system, wheat Pol III promoters, TaU3, TaU6.3 and TaU6.1 (Zhang et al., 2019a), were used to drive sgRNA expression independently. For the tRNA system, a TaU3 promoter was also used to express the tRNA‐sgRNA cassettes in a single transcript unit. For the ribozyme system, a Pol II promoter, Cestrum yellow leaf curling virus (CmYLCV) promoter (Cermak et al., 2017), was employed for expressing hammerhead ribozyme (HH)–sgRNA–hepatitis delta virus (HDV) ribozyme cassettes in a single transcript unit. A longer sgRNA scaffold was applied in three vectors to optimize the sgRNA structure (Dang et al., 2015). Wheat codon‐optimized Cas9 was driven under a maize (Zea mays) ubiquitin promoter (Ubip). Three genes, TaDA1, TaPDS and TaNCED1, were selected for simultaneous editing. The sgRNA for TaDA1 could target its homoeologous genes on A and B chromosomes, while the sgRNAs for TaPDS and TaNCED1 were designed to target all three homoeologous genes, respectively (Zhang et al., 2019b). In total, three sgRNAs could target 8 genomic sites in common wheat (Figure 1b). The sgRNA cassettes in the vectors were all arranged in the same order for close comparison. These T‐DNA vectors were introduced into hexaploid wheat Fielder via Agrobacterium tumefaciens‐mediated transformation.

A total of 22, 26 and 27 T0 plants were generated from the transformed calli of TRSP, tRNA and ribozyme systems, respectively. The genotype of each plant was characterized by Hi‐TOM sequencing of the PCR amplicons with primers flanking each target site (Liu et al., 2019). The editing efficiency of individual genes was first analysed. Edits of TaDA1‐A and TaDA1‐B were detected in all three systems, and the ribozyme system was most effective (Figure 1c). The superior editing ability of the ribozyme system was also observed at TaPDS where mutations could cause albino phenotype. Impressively, 22 out of 27 plants showed albino phenotype in the ribozyme system, whereas only 5 plants displayed albino phenotype in either the TRSP or tRNA system. Sequencing results supported the observation as the ribozyme system achieved the highest efficiency, up to 100.00% (Figure 1c). Although fewer plants showed albino phenotype, the editing efficiencies in TRSP and tRNA systems still reached to 86.36% and 92.31%, respectively. For TaNCED1, all three vectors exhibited low activity, and the tRNA and ribozyme systems resulted in higher gene editing rates than the TRSP system (Figure 1c). The three systems showed similar mutation profiles for individual genes where small deletions and 1bp insertions predominated (Figure 1d).

The ability to target multiple genomic sites was further analysed for three systems. Compared with the TRSP system, more plants with over 4 edited sites were identified in the tRNA and ribozyme systems (Figure 1e), and the ribozyme system generated the highest simultaneous editing rates. The efficiencies of simultaneous editing in three genes in the tRNA and ribozyme systems were 34.62% and 37.04%, respectively, which were about twofold higher than that in the TRSP system (Figure 1f). Thus, the tRNA and ribozyme systems are more effective than TRSP, and the ribozyme system appeared to be most robust. The high editing efficiency of the ribozyme system might partly result from the use of the Pol II promoter, CmYLCV, for sgRNA expression.

The phenotype caused by gene editing depends on the genotype in individual plants. Therefore, we investigated the ratio of mutated reads at each target site in each plant through Hi‐TOM sequencing (Liu et al., 2019). No significant differences were detected at all targeted sites between the TRSP and tRNA systems, but the ratios of edited reads were significantly increased in the ribozyme system except for TaNCED1 (Figure 1g). Ratios of edited reads in the ribozyme system were about threefold higher for TaDA1 and twofold higher for TaPDS than those in the TRSP and tRNA systems, respectively. The results suggested that the ribozyme system greatly decreased the proportions of unedited reads at multiplex chromosomes and therefore increased the probability of the loss‐of‐function phenotype in T0 generation. This might explain the discrepancy between the high editing efficiency and less albino phenotype caused by TaPDS mutation in the TRSP and tRNA systems. Although over 86.00% of the plants carried edited TaPDS in the TRSP and tRNA systems, the editing ratios in most plants were not enough to display albino phenotype. To further quantify the relationship between ratios of the edited reads and observable phenotype, we compared the ratios of edited reads for TaPDS‐A, TaPDS‐B and TaPDS‐D among three groups of plants as no albino, chimeric and albino phenotype, respectively (Figure 1h). Positive correlation between the phenotype and editing ratio was observed (Figure 1i). The lowest average ratio for plants with albino phenotype was 80.59%, indicating an editing threshold for displaying loss‐of‐function phenotype. Wild type alleles were not detected at TaPDS in 6 lines of the ribozyme system despite different levels of chimerism (Figure 1j). These results collectively revealed that the phenotype caused by targeted mutagenesis occurred only when the ratios of edited homoeologous genes achieved a higher level at all homoeologous chromosomes simultaneously.

In summary, we compared three multiplex CRISPR/Cas9 systems for simultaneous genome editing at 8 target sites in common wheat. The tRNA and ribozyme systems were more effective than the TRSP system in multiplex genome editing. Furthermore, the ribozyme system could significantly increase the ratios of edited homoeologous genes at multiplex chromosomes in individual plants and therefore generated more plants with loss‐of‐function phenotypes. The ribozyme system established in our study would greatly aid fundamental and translational research in wheat.

普通小麦的基因组很大，具有三个亚基因组（A、B 和 D），因此在多个基因组位点同时产生突变具有挑战性。 CRISPR/Cas9 系统为编辑作物基因组提供了一种改变游戏规则的工具（Chen等人， 2019 ）。已开发出三种主要策略来产生多个单引导 RNA (sgRNA)，包括具有独立U3 或 U6启动子 (TRSP)串联重复的传统多重系统、tRNA 处理系统 (Xie等人， 2015 ）和核酶加工系统（Gao和Zhao， 2014 ）。尽管 CRISPR/Cas9 介导的基因组编辑之前是通过生物射弹 (Wang et al ., 2014 , 2018 ) 和农杆菌转化 (Zhang et al ., 2019a ) 实现的，但用于小麦多重编辑的最有效的 CRISPR/Cas9 系统仍然难以捉摸。为了解决这个重要问题，我们基于 pBUE411 载体设计了与这三个系统相对应的三个多重编辑结构（图 1a）。对于 TRSP 系统，小麦 Pol III 启动子 TaU3、TaU6.3 和 TaU6.1（Zhang等人， 2019a ）用于独立驱动 sgRNA 表达。对于 tRNA 系统，还使用 ​​TaU3 启动子在单个转录单元中表达 tRNA-sgRNA 盒。对于核酶系统，采用 Pol II 启动子、Cestrum 黄叶卷曲病毒 (CmYLCV) 启动子（Cermak等人， 2017 ）在单个转录单位。 在三个载体中应用了更长的 sgRNA 支架来优化 sgRNA 结构 (Dang et al ., 2015 )。小麦密码子优化的 Cas9 由玉米 ( Zea mays ) 泛素启动子 (Ubip) 驱动。选择三个基因TaDA1 、 TaPDS和TaNCED1进行同时编辑。 TaDA1的 sgRNA 可以靶向 A 和 B 染色体上的同源基因，而TaPDS和TaNCED1的 sgRNA 被设计为分别靶向所有三个同源基因（Zhang等， 2019b ）。总共，三个 sgRNA 可以靶向普通小麦中的 8 个基因组位点（图 1b）。载体中的 sgRNA 盒均按相同顺序排列，以便进行仔细比较。这些 T-DNA 载体通过根癌农杆菌介导的转化引入六倍体小麦 Fielder 中。

由TRSP、tRNA和核酶系统转化的愈伤组织分别产生总共22、26和27个T0植物。每个植物的基因型通过每个目标位点两侧的引物对 PCR 扩增子进行 Hi-TOM 测序来表征（Liu et al ., 2019 ）。首先分析了单个基因的编辑效率。在所有三个系统中均检测到TaDA1-A和TaDA1-B的编辑，并且核酶系统最有效（图 1c）。在TaPDS中也观察到了核酶系统卓越的编辑能力，其中突变可能导致白化表型。令人印象深刻的是，27 株植物中有 22 株在核酶系统中显示出白化表型，而只有 5 株植物在 TRSP 或 tRNA 系统中显示出白化表型。测序结果支持了这一观察结果，因为核酶系统实现了最高效率，高达 100.00%（图 1c）。尽管表现出白化表型的植物较少，但TRSP和tRNA系统的编辑效率仍然分别达到86.36%和92.31%。对于TaNCED1 ，所有三个载体均表现出低活性，并且 tRNA 和核酶系统比 TRSP 系统具有更高的基因编辑率（图 1c）。这三个系统显示出相似的单个基因突变谱，其中小缺失和 1bp 插入占主导地位（图 1d）。

进一步分析了三个系统靶向多个基因组位点的能力。与TRSP系统相比，tRNA和核酶系统中鉴定出了更多具有超过4个编辑位点的植物（图1e），并且核酶系统产生了最高的同时编辑率。 tRNA和核酶系统中三个基因的同时编辑效率分别为34.62%和37.04%，比TRSP系统高出约两倍（图1f）。因此，tRNA 和核酶系统比 TRSP 更有效，并且核酶系统似乎最稳健。核酶系统的高编辑效率可能部分归因于使用 Pol II 启动子 CmYLCV 进行 sgRNA 表达。

基因编辑引起的表型取决于个体植物的基因型。因此，我们通过Hi-TOM测序调查了每株植物中每个靶位点的突变reads的比例（Liu et al ., 2019 ）。 TRSP 和 tRNA 系统之间的所有目标位点均未检测到显着差异，但除TaNCED1外，核酶系统中编辑读数的比率显着增加（图 1g）。核酶系统中编辑读数的比率， TaDA1和TaPDS分别比 TRSP 和 tRNA 系统高三倍和两倍。结果表明，核酶系统大大降低了多染色体上未编辑读数的比例，因此增加了 T0 代功能丧失表型的可能性。这可能解释了 TRSP 和 tRNA 系统中TaPDS突变引起的高编辑效率和较少白化表型之间的差异。尽管超过86.00%的植物在TRSP和tRNA系统中携带编辑过的TaPDS ，但大多数植物中的编辑比率不足以显示白化表型。为了进一步量化编辑读数的比率和可观察表型之间的关系，我们分别比较了三组植物中无白化、嵌合和白化表型的TaPDS-A 、 TaPDS-B和TaPDS-D的编辑读数比率（图1h）。观察到表型和编辑率之间呈正相关（图 1i）。具有白化表型的植物的平均比例最低为80.59%，表明显示功能丧失表型的编辑阈值。 尽管嵌合水平不同，但在核酶系统的 6 个品系中， TaPDS未检测到野生型等位基因（图 1j）。这些结果共同表明，只有当所有同源染色体上编辑的同源基因的比例同时达到较高水平时，才会出现定点诱变引起的表型。

总之，我们比较了三种多重 CRISPR/Cas9 系统，用于在普通小麦的 8 个靶位点同时进行基因组编辑。在多重基因组编辑中，tRNA 和核酶系统比 TRSP 系统更有效。此外，核酶系统可以显着增加单个植物中多染色体上经过编辑的同源基因的比例，从而产生更多具有功能丧失表型的植物。我们研究中建立的核酶系统将极大地帮助小麦的基础和转化研究。