Although CRISPR‐Cas9 has revolutionized our ability to generate site‐specific double‐strand breaks, precise editing of the genome remains challenging in most eukaryotes, including plants (Shan et al., 2013). In plants, homology‐directed repair is inefficient, limiting our ability to make precise edits of the DNA sequence (Ali et al., 2020; Butt et al., 2017). Moreover, cytosine and adenine base editors have serious drawbacks including lower efficiency, unclean edited sequence and the possibility of off‐target mutations at other loci (Rees and Liu, 2018). Chimeric single‐guide RNAs (sgRNAs) can provide editing information, in RNA form, but this modality suffers from several limitations including lower efficiency, less versatility and the need for long homology arms (Butt et al., 2017).

In contrast to genome editing methods that use just a Cas nuclease to generate double‐strand breaks, prime editing employs a Cas9 nickase (nCas9) fused with reverse transcriptase (RT). The desired edits are encoded on a prime editing guide RNA, which guides the nCas9‐RT complex to the target site (Anzalone et al., 2019). There, the nCas9 generates a single‐strand break (Shrivastav et al., 2008) on the non‐complimentary strand and the RT domain transfers the desired edits from the pegRNA to the DNA (Anzalone et al., 2019). Researchers have developed several prime editing strategies: in PE1, wild type M‐MLV RT fused to the C terminus of Cas9 (H840A) nickase; in PE2, Cas9 (H840A) with pentamutant M‐MLV RT (D200N/ L603W/ T330P/ T306K/ W313F); in PE3, a PE2 prime editor with additional simple gRNA to simultaneously nick the non‐edited strand (Anzalone et al., 2019). Prime editing has several advantages over other methods, such as enabling precise sequence deletion, addition and substitution. However, although it has been tested in human cell lines, prime editing remains to be tested in plants.

To test prime editing in rice (Oryza sativa), we first attempted to engineer herbicide resistance by targeting rice ACETOLACTATE SYNTHASE (OsALS). ALS catalyses the initial step common to the biosynthesis of the branched‐chain amino acids and is primary target site for herbicides like Bispyribac sodium. A single amino acid change (W548L) in ALS results in a BS‐resistant phenotype (Butt et al., 2017). We cloned the PE2 fragment containing Cas9 (H840A) with pentamutant M‐MLV RT under the control of the OsUBIQUITIN promoter in rice vectors. We therefore designed a pegRNA to edit the OsALS sequence. The RT template with a length of 15 bp has two substitutions, a G‐to‐T substitution that converts tryptophan 548 to leucine and a silent G‐to‐A substitution that destroys the PAM site thus preventing re‐targeting by the pegRNA‐nCas9‐RT machinery (Figure 1a). These nucleotide modifications result in the loss of a BsaXI site and generation of an MfeI site. The primer binding site (PBS) was designed with a length of 13 bp. The pegRNA was expressed in rice vectors under the OsU3 promoter.

We transformed rice via Agrobacterium and after two weeks of selection, we collected four independently growing calli from different selection plates. We performed the DNA extraction from these calli and amplified the target DNA by PCR. We pooled the amplicons in equimolar concentrations and performed deep sequencing. Our data showed that the prime editing successfully edited OsALS at the target site with an efficiency of 0.26 to 2% (Figure 1b). The different editing efficiencies among two pools are possibly due to varied number of non‐edited WT cells between these calli. The editing efficiencies are further validated when we enriched the edited DNA from the four calli by cutting with BsaXI (which cuts the unedited sequence) and conducted PCR/restriction enzyme analysis (PCR/RE) using MfeI (Figure 1c). The digestion of amplicons by MfeI indicated the frequency of editing in the samples. We used Sanger sequencing to confirm these edits (Figure 1d). Most of the reads were fully edited and repaired according to the RT template. Interestingly, some of the reads showed an A‐to‐G substitution, which converts tyrosine 553 to cysteine. This substitution is not the part of the RT template and probably came from the scaffold RNA, as the first nucleotide of the scaffold RNA adjacent to the RT template (a ‘G’) can be used for DNA repair (Figure 1d).

We also targeted rice IDEAL PLANT ARCHITECTURE 1 (OsIPA) using prime editing (Figure 1e). The OsIPA transcription factor reduces the number of unproductive tillers and improves rice yield. We designed a pegRNA for two consecutive substitutions (AG to GA) to convert S163 to D in IPA with length of RT 20 bp and PBS 13 bp. Two silent substitutions (CGC to AGA) destroy the PAM site. These mutations destroy a PvuII site and generate Pst1 and BbsI sites. We transformed rice via Agrobacterium and regenerated shoots. We analysed the plantlets after enriching for edited DNA with PvuII digestion by Sanger sequencing. We found that prime editing successfully edited OsIPA at the target site, (Figure 1f).

Similarly, we targeted rice TEOSINTE BRANCHED 1 (OsTB1), a member of the TEOSINTE BRANCHED1, CYCLOIDEA AND PCF TRANSCRIPTION FACTOR gene family (Figure 1g). OsTB1 negatively regulates lateral branching by repressing axillary bud outgrowth. We designed a pegRNA to target the OsTB1 promoter with length of RT 20 bp and PBS 13 bp. A C‐to‐G substitution destroyed the PAM to prevent re‐targeting and two consecutive insertions (AA) and one substitution (C to T) destroyed and RsaI site and created an SspI site. We analysed the shoots by enriching the DNA with RsaI digestion and by Sanger sequencing and observed partial repair and different types of reads (Figure 1h). The possible reason for chimeric cells is that prime editing machinery could be still functional in the non‐edited cells and continuously modified the targeted region.

To test whether we could improve the editing efficiency, we tried the PE3 strategy, where a second sgRNA is used to nick the complimentary strand. We designed the sgRNA to target OsALS at a distance of +55 from the pegRNA and expressed this sgRNA from the OsU3 promoter using the polycistronic tRNA‐gRNA system (Butt et al., 2017; Xie et al., 2015) (Figure 1a). We transformed the rice callus with ALS‐PE2 (containing just the pegRNA and RT‐nCas) and ALS‐PE3 (containing the sgRNA, pegRNA, and RT‐nCas) plasmids. After selection, we regenerated shoots on media supplemented with 0.75 µm BS (Figure 1i). For both ALS‐PE2 and ALS‐PE3, we recovered shoots resistant to BS. Sanger sequencing showed that these plantlets were successfully edited (Figure 1j). We recovered almost equal numbers of shoots from PE2 and PE3 (Figure 1k), suggesting that (unlike mammalian systems) PE3 did not increase editing efficiency in plants.

In the present study, we successfully used prime editing technology on three loci in plants. While this work was prepared for publication, similar findings were reported in pants (Li et al., 2020; Lin et al., 2020; Tang et al., 2020). We engineered herbicide resistance trait in rice via nucleotide substitutions; however, the system requires further improvements and assessments on its ability to enable diverse editing modalities for different trait engineering applications in plants.

尽管CRISPR-Cas9彻底改变了我们产生位点特异性双链断裂的能力，但在包括植物在内的大多数真核生物中，基因组的精确编辑仍然面临挑战（Shan等人，2013）。在植物中，同源性修复效率低下，限制了我们对DNA序列进行精确编辑的能力（Ali等，2020 ; Butt等，2017）。此外，胞嘧啶和腺嘌呤碱基编辑器还存在严重的缺陷，包括效率较低，编辑序列不干净以及其他基因座可能发生脱靶突变的可能性（Rees and Liu，2018）。）。嵌合单向导RNA（sgRNA）可以以RNA形式提供编辑信息，但是这种模式存在一些局限性，包括效率低，通用性差和需要长同源臂（Butt等人，2017年）。

与仅使用Cas核酸酶来产生双链断裂的基因组编辑方法相反，主要编辑使用与逆转录酶（RT）融合的Cas9切口酶（nCas9）。所需的编辑均在主要的编辑指南RNA上编码，该指南将nCas9-RT复合体引导至目标位点（Anzalone等，2019）。在那里，nCas9在非互补链上产生单链断裂（Shrivastav等，2008），RT结构域将所需的编辑从pegRNA转移到DNA（Anzalone等，2019）。）。研究人员开发了几种主要的编辑策略：在PE1中，野生型M-MLV RT与Cas9（H840A）切口酶的C末端融合；在PE2，Cas9（H840A）和五突变M‐MLV RT（D200N / L603W / T330P / T306K / W313F）中; 在PE3中，是PE2的主要编辑者，具有附加的简单gRNA来同时刻痕未编辑的链（Anzalone等人，2019年）。相对于其他方法，主要编辑具有多个优点，例如可以进行精确的序列删除，添加和替换。但是，尽管已经在人类细胞系中对其进行了测试，但是主要编辑仍有待在植物中进行测试。

为了测试水稻（Oryza sativa）的原始编辑，我们首先尝试通过靶向水稻乙酰丙酸合酶（OsALS）来设计除草剂抗性。ALS催化了支链氨基酸生物合成所共有的起始步骤，并且是Bispyribac钠等除草剂的主要目标位点。ALS中的单个氨基酸变化（W548L）导致BS抗性表型（Butt等人，2017）。我们在OsUBIQUITIN启动子的控制下，用五突变M-MLV RT克隆了含有Cas9（H840A）的PE2片段。因此，我们设计了pegRNA来编辑OsALS顺序。长度为15 bp的RT模板具有两个取代，一个G-to-T取代将色氨酸548转换为亮氨酸，一个沉默的G-to-A取代破坏了PAM位点，从而防止了pegRNA-nCas9的重新靶向‐RT机械（图1a）。这些核苷酸修饰导致BsaXI位点的丢失和MfeI位点的产生。设计引物结合位点（PBS），其长度为13 bp。pegRNA在OsU3启动子下在水稻载体中表达。

我们通过农杆菌转化水稻，经过两周的选择，我们从不同的选择板上收集了四个独立生长的愈伤组织。我们从这些愈伤组织中提取了DNA，并通过PCR扩增了目标DNA。我们汇集了等摩尔浓度的扩增子，并进行了深度测序。我们的数据显示主要编辑成功编辑了OsALS在目标位置的效率为0.26至2％（图1b）。两个池之间不同的编辑效率可能是由于这些愈伤组织之间未编辑的WT信元数量不同所致。当我们用BsaXI切割（剪切未编辑的序列）并使用MfeI进行PCR /限制酶分析（PCR / RE）（图1c）时，从四个愈伤组织中富集了编辑的DNA，从而进一步验证了编辑效率。MfeI对扩增子的消化表明了样品中编辑的频率。我们使用Sanger测序来确认这些编辑（图1d）。大多数读物已根据RT模板进行了全面编辑和修复。有趣的是，一些读物显示了A到G取代，将酪氨酸553转化为半胱氨酸。这种取代不是RT模板的一部分，可能来自支架RNA，

我们还使用主要编辑（图1e）针对水稻IDEAL PLANT ARCHITECTURE 1（OsIPA）。OsIPA转录因子减少了非生产性分till的数量并提高了水稻产量。我们设计了一个pegRNA，用于两个连续的替换（从AG到GA），以将S163转换为IPA中的D，长度为RT 20 bp，PBS为13 bp。两个无声替换（从CGC到AGA）破坏了PAM站点。这些突变破坏PvuII位点并生成Pst1和BbsI位点。我们通过农杆菌转化水稻并再生了芽。我们通过Sanger测序通过PvuII消化富集了可编辑的DNA，从而分析了小植株。我们发现主要编辑成功地在目标站点上编辑了OsIPA（图1f）。

同样，我们针对的是水稻TEOSINTE BRANCHED 1（OsTB1），它是TEOSINTE BRANCHED1，CYCLOIDEA和PCF转录因子基因家族的成员（图1g）。OsTB1通过抑制腋芽的生长来负面调节侧枝。我们设计了一个靶向OsTB1的pegRNA启动子的长度为RT 20 bp，PBS为13 bp。从AC到G的替换破坏了PAM以防止重新定向，并破坏了两个连续的插入（AA）和一个替换（从C到T），并破坏了RsaI位点并创建了一个SspI位点。我们通过用RsaI消化和Sanger测序富集了DNA来分析枝条，并观察到部分修复和不同类型的读数（图1h）。嵌合细胞的可能原因是主要的编辑机制仍可以在未编辑的细胞中发挥作用，并不断修饰目标区域。

为了测试我们是否可以提高编辑效率，我们尝试了PE3策略，其中使用了第二个sgRNA来刻划互补链。我们设计了一个sgRNA，使其靶向距pegRNA +55处的OsALS，并使用多顺反子tRNA-gRNA系统从OsU3启动子表达了该sgRNA （Butt等人，2017 ; Xie等人，2015）（图1a） 。我们用ALS‐PE2（仅包含pegRNA和RT‐nCas）和ALS‐PE3（包含sgRNA，pegRNA和RT‐nCas）质粒转化了水稻愈伤组织。选择后，我们再生辅以0.75μ媒体芽米BS（图1i）。对于ALS-PE2和ALS-PE3，我们回收了对BS具有抗性的芽。Sanger测序表明这些小植株已成功编辑（图1j）。我们从PE2和PE3中回收了几乎相等数量的芽（图1k），这表明（与哺乳动物系统不同）PE3不会增加植物的编辑效率。

在本研究中，我们成功地在植物的三个基因座上使用了主要的编辑技术。尽管准备发表这项工作，但在裤子上也报道了类似的发现（Li等，2020； Lin等，2020； Tang等，2020）。我们通过核苷酸取代工程改造了水稻的除草剂抗性。但是，该系统需要对其功能进行进一步的改进和评估，以使其能够针对植物中的不同性状工程应用应用多种编辑方式。