CRISPR technology is an established tool for the generation of knockout plants (Zhang et al., 2019), yet limitations remain. First, the manipulation of individual genes may fail to produce phenotypes for groups of genes with redundant or synergistic functions. While this has been partially addressed by multiplexing guide RNAs (gRNAs), there is concern that as the number of targets increases, the chances of obtaining higher‐order knockouts diminish (Zhang et al., 2016). Second, knocking out fundamentally important genes can cause severe pleiotropic phenotypes or lethality. Tissue‐specific knockout of genes in somatic tissues can overcome this limitation (Decaestecker et al., 2019; Liang et al., 2019; Wang et al., 2020). However, the efficiency of simultaneously targeting more than three genes in a tissue‐specific context is unexplored. Here, by multiplexing gRNAs in Arabidopsis thaliana plants expressing Cas9 either ubiquitously (pPcUBI) or root cap specifically (pSMB), we show that six genes can be simultaneously mutated with high efficiency, generating higher‐order mutant phenotypes already in the first transgenic generation (T1). The mutation frequencies for all target genes were positively correlated and unaffected by the order of the gRNAs in the vector, showing that efficient higher‐order mutagenesis in specific plant tissues can be readily achieved.

We selected six efficient gRNAs (Decaestecker et al., 2019 and unpublished results) to target the coding sequences of six genes (GFP, and the Arabidopsis genes SMB, EXI1, GL1, ARF7 and ARF19; Figure 1a) whose knockout lead to easy‐to‐score phenotypes in T1 seedlings (gfp: loss of GFP signal, smb: accumulation of root cap cells (Fendrych et al., 2014), gl1: absence of trichomes (Herman and Marks, 1989)) and do not severely affect plant growth or reproduction. Since position effects within gRNA arrays had been a concern regarding mutation efficiency (Zhang et al., 2016), we generated two vectors (hereafter, pPcUBI(I) and pPcUBI(II)) combining Cas9‐mTagBFP2 driven by the ubiquitous pPcUBI promoter and the six gRNAs in an inverted order (Figure 1b) and transformed these into an Arabidopsis line with ubiquitous expression of a nuclear‐localized GFP (pHTR5:NLS‐GFP‐GUS (Decaestecker et al., 2019) hereafter, NLS‐GFP).

Forty‐nine out of 96 pPcUBI(I) and 52 out of 95 pPcUBI(II) T1 seedlings displayed both gfp and smb phenotypes in roots, indicating simultaneous mutations (Figure 1c,d). Additionally, 44 out of 96 pPcUBI(I) and 45 out of 95 pPcUBI(II) T1 seedlings also lacked trichomes on the first two true leaves, revealing a high mutation frequency for GL1. Altogether, 79% of the pPcUBI(I) and 68% of the pPcUBI(II) T1 seedlings with at least one detectable knockout phenotype also showed triple gfp smb gl1 mutant phenotypes. When selecting plants based on the loss of GFP, 90% of the pPcUBI(I) and 85% of the pPcUBI(II) T1 seedlings displayed triple mutant phenotypes, indicating a strong correlation of mutagenesis efficiencies.

We quantified indel frequencies in 48 pPcUBI(I), 47 pPcUBI(II) and a control NLS‐GFP plant. The targeted loci were PCR‐amplified from root tips and sequenced using Illumina sequencing. Plants showing total or partial gfp and smb phenotypes had high indel frequencies in GFP (27%–100%) and SMB (38%–98%), as well as in all other target genes. Hierarchical clustering showed that transgenic T1 plants fell in two major classes that had either high or low levels of mutagenesis for all target genes (Figure 1e). In agreement with previous reports (Feng et al., 2014), 1‐bp indels were the predominant repair outcome (50%–80% and 1%–15% respectively), in‐frame indels were rare (2%–8%), and 6%–26% of mutations were bigger deletions (>6‐bp), insertions (>3‐bp) or complex repair outcomes (Figure 1f).

We compared indel frequencies for each target between the two constructs to test the effect of the gRNA position (Figure 1g). The overall indel frequencies were higher for pPcUBI(II), though the difference was only significant for GFP. As all other gRNAs had no substantial changes in indel frequencies, our data do not support a position effect in gRNA arrays, thus reducing the complexity of future experimental design.

We then tested whether six genes can be efficiently mutated in a tissue‐specific context by making two vectors expressing Cas9‐P2A‐mTagBFP2 under the root cap‐specific pSMB promoter with the same arrangement of gRNAs (hereafter, pSMB(I) and pSMB(II)). Plants were grown in the presence of 1 µM brassinazole (BRZ) to facilitate smb phenotyping. This treatment leads to a root covered by living root cap cells in smb mutants (Fendrych et al., 2014) and was easily recognizable due to the presence of nuclear mTagBFP2 signal in living root cap cells (Figure 1h).

Thirty‐two out of 86 pSMB(I) and 46 out of 88 pSMB(II) T1 seedlings showed both gfp and smb phenotypes, as well as a strong mTagBFP2 signal specifically in root cap nuclei as determined by confocal microscopy (Figure 1i). In agreement with our previous report (Decaestecker et al., 2019), mTagBFP2 signal intensity could be used as a proxy for the penetrance of gfp and smb knockout phenotypes. To determine mutagenesis efficiency in all target genes specifically in Cas9‐expressing root cap cells, we collected root‐tip protoplasts expressing mTagBFP2 (BFP⁺, Cas9 expressing cells) using fluorescence‐activated cell sorting from T2 seedlings of ten pSMB(I) and eight pSMB(II) independent lines. We chose four pSMB(II) lines (19, 25, 35 and 48) with weak or chimeric gfp and smb T1 mutant phenotypes and four pSMB(I) and(II) lines with highly penetrant smb and gfp T1 mutant phenotypes.

The targeted loci were PCR‐amplified directly from sorted protoplast populations and sequenced by NGS. In pSMB(I) and (II), T2 seedlings coming from a T1 parent with strong smb and gfp phenotypes, the Cas9‐expressing BFP⁺ populations had indel frequencies between 51% and 92% for all six target loci (Figure 1j). As expected, the BFP⁺ populations of the pSMB(II) lines that with weak or chimeric gfp and smb phenotypes in T1 had lower indel frequencies (2%–50%). These results confirmed that in lines with high GFP and SMB mutagenesis activity, all genes were simultaneously mutated with high efficiency. Similarly to the ubiquitous lines, the alleles generated were largely consistent across events, with 1‐bp indels being the predominant repair outcome (50%–87% and 2%–10%), in‐frame insertion or deletions were rare (0%–5%), and 3%–21% of mutations were bigger indels (>3‐ and >6‐bp) or combination of indels (Figure 1k).

In conclusion, we show that ubiquitous CRISPR and CRISPR‐TSKO approaches allow fast and simultaneous disruption of six genes in the first transgenic generation with high efficiency. As mutation efficiencies over all loci are correlated, we suggest the use of a target gene with an easy‐to‐score, non‐detrimental loss‐of‐function phenotype as a proxy for highly mutagenized lines. As an alternative to endogenous genes (Li et al., 2020), loss of GFP in a reporter line can also be used as a proxy. We foresee this approach to be a powerful tool to dissect genetic networks in model and crops species alike.

CRISPR 技术是用于产生基因敲除植物的成熟工具（Zhang等人，2019 年），但仍然存在局限性。首先，单个基因的操作可能无法为具有冗余或协同功能的基因组产生表型。虽然这已通过多路复用指导 RNA (gRNA) 得到部分解决，但人们担心随着目标数量的增加，获得高阶敲除的机会会减少 (Zhang et al ., 2016 )。其次，剔除根本上重要的基因会导致严重的多效性表型或致死性。体细胞组织中基因的组织特异性敲除可以克服这一限制（Decaestecker et al ., 2019; 梁等人，2019；王等人，2020）。然而，在组织特异性环境中同时靶向三个以上基因的效率尚未得到探索。在这里，通过在拟南芥植物中多路复用 gRNAs，无论是普遍表达 ( pPcUBI ) 还是特异性表达 Cas9 ( pSMB )），我们证明了六个基因可以同时高效突变，在第一代转基因（T1）中已经产生了更高阶的突变表型。所有靶基因的突变频率呈正相关且不受载体中 gRNA 顺序的影响，表明在特定植物组织中可以很容易地实现有效的高阶诱变。

我们选择了六种有效的 gRNA（Decaestecker等人，2019 年和未发表的结果）来靶向六种基因（GFP和拟南芥基因SMB、EXI1、GL1、ARF7和ARF19；图 1a）的编码序列，这些基因的敲除导致容易- T1 幼苗中的to-score 表型（gfp：GFP 信号丢失，smb：根帽细胞的积累（Fendrych等人，2014 年），gl1：没有毛状体（Herman 和 Marks，1989)) 并且不会严重影响植物的生长或繁殖。由于 gRNA 阵列内的位置效应一直是突变效率的一个问题（Zhang等人，2016 年），我们生成了两个载体（以下称为pPcUBI(I)和pPcUBI(II) ），结合了由普遍存在的pPcUBI启动子驱动的 Cas9-mTagBFP2和六个 gRNA 以倒序排列（图 1b）并将它们转化为普遍表达核定位 GFP（pHTR5:NLS-GFP-GUS（Decaestecker等人，2019 年）以下，NLS-GFP）的拟南芥品系。

96 个pPcUBI(I)中有 49 个和 95个 pPcUBI(II) T1 幼苗中有 52 个在根中显示gfp和smb表型，表明同时发生突变（图 1c，d）。此外，96 个pPcUBI(I)中的 44 个和 95个 pPcUBI(II) T1 幼苗中的 45 个在前两个真叶上也缺乏毛状体，这表明GL1的突变频率很高。总共有 79% 的pPcUBI(I)和 68% 的pPcUBI(II) T1 幼苗具有至少一种可检测的敲除表型，也显示出三重gfp smb gl1突变表型。当根据 GFP 的损失选择植物时，90%pPcUBI(I)和 85% 的pPcUBI(II) T1 幼苗表现出三重突变表型，表明诱变效率具有很强的相关性。

我们量化了 48 个pPcUBI(I)、47 个pPcUBI(II)和一个对照 NLS-GFP 植物中的插入缺失频率。目标基因座从根尖进行 PCR 扩增，并使用 Illumina 测序进行测序。显示全部或部分gfp和smb表型的植物在GFP (27%–100%) 和SMB (38%–98%) 以及所有其他靶基因中具有高插入缺失频率。层次聚类显示转基因 T1 植物分为两个主要类别，对所有靶基因具有高或低水平的诱变（图 1e）。与之前的报道一致（Feng et al ., 2014)，1 bp 插入缺失是主要的修复结果（分别为 50%–80% 和 1%–15%），框内插入缺失很少（2%–8%），6%–26% 的突变更大缺失（>6-bp）、插入（>3-bp）或复杂的修复结果（图 1f）。

我们比较了两个构建体之间每个目标的插入缺失频率，以测试 gRNA 位置的影响（图 1g）。pPcUBI(II)的总体插入缺失频率较高，尽管差异仅对GFP显着。由于所有其他 gRNA 在插入缺失频率方面没有实质性变化，我们的数据不支持 gRNA 阵列中的位置效应，从而降低了未来实验设计的复杂性。

然后，我们通过在具有相同 gRNA 排列的根帽特异性pSMB启动子下制作两个表达 Cas9-P2A-mTagBFP2 的载体（以下称为pSMB(I)和pSMB(二））。植物在 1 µM 芸苔素 (BRZ) 存在下生长，以促进smb表型分析。这种处理导致smb突变体中的根被活根帽细胞覆盖（Fendrych等人，2014 年），并且由于活根帽细胞中存在核 mTagBFP2 信号而易于识别（图 1h）。

通过共聚焦显微镜确定，86个 pSMB(I)中的 32 个和 88 个pSMB(II) T1 幼苗中的 46 个显示gfp和smb表型，以及在根冠核中特异性的强 mTagBFP2 信号（图 1i）。与我们之前的报告（Decaestecker等人，2019 年）一致，mTagBFP2 信号强度可用作gfp和smb敲除表型外显率的代表。为了确定所有靶基因的诱变效率，特别是在表达 Cas9 的根帽细胞中，我们收集了表达 mTagBFP2 (BFP ⁺, Cas9 表达细胞) 使用荧光激活细胞分选从 10 个pSMB(I)和 8 个pSMB(II)独立系的 T2 幼苗中进行。我们选择了具有弱或嵌合gfp和smb T1 突变表型的四个pSMB(II)系 (19, 25, 35 和 48) 以及具有高渗透性smb和gfp T1 突变表型的四个pSMB(I)和(II)系。

目标基因座直接从分选的原生质体群体中进行 PCR 扩增，并通过 NGS 测序。在pSMB(I)和(II)中，来自具有强smb和gfp表型的 T1 亲本的 T2 幼苗，表达 Cas9 的 BFP ⁺群体在所有六个目标基因座上的插入缺失频率在 51% 和 92% 之间（图 1j）。正如预期的那样，在 T1中具有弱或嵌合gfp和smb表型的pSMB(II)系的 BFP ⁺种群具有较低的插入缺失频率 (2%–50%)。这些结果证实，符合高GFP和SMB诱变活性，所有基因同时高效突变。与普遍存在的系相似，产生的等位基因在各个事件中基本一致，1-bp 插入缺失是主要的修复结果（50%–87% 和 2%–10%），框内插入或缺失很少见（0% –5%）和 3%–21% 的突变是更大的插入缺失（>3- 和 >6-bp）或插入缺失的组合（图 1k）。

总之，我们表明普遍存在的 CRISPR 和 CRISPR-TSKO 方法允许在第一代转基因中快速和同时破坏六个基因，并且效率很高。由于所有基因座的突变效率是相关的，我们建议使用具有易于评分、无害的功能丧失表型的靶基因作为高度诱变系的代表。作为内源基因的替代方案 (Li et al ., 2020 )，报告基因系中 GFP 的丢失也可以用作代理。我们预计这种方法将成为剖析模型和作物物种中的遗传网络的有力工具。