The CRISPR/Cas‐mediated genome editing technology has been widely applied to create knockout alleles of genes by generating short insertions or deletions (indel) in various plant species. Due to the low efficiency of homology‐directed repair (HDR) and difficulties in the delivery of DNA template for HDR, precise genome editing remains challenging in plants (Mao et al., 2019). A tandem repeat‐HDR method was developed very recently for sequence replacement in rice, which is most useful for monocots (Lu et al., 2020). Base editors developed from Cas9 nickase fusion with cytosine and adenine deaminases enable targeted C‐to‐T or A‐to‐G substitutions, but are restricted to specific types of base replacements and target site selections (Mao et al., 2019). A ‘search‐and‐replace’ method, also known as prime editing, was developed in mammalian cells, which enables user‐defined sequence changes on a target site without requiring DSBs or the delivery of DNA repair templates (Anzalone et al., 2019). Several research groups have adopted this method for use in monocotyledonous plants, including rice and wheat (Butt et al., 2020; Hua et al., 2020; Li et al., 2020; Lin et al., 2020; Tang et al., 2020; Xu et al., 2020). For reasons that are still unclear, although base editing has been highly efficient in monocots such as rice, its efficiencies are very low in dicots (Kang et al., 2018; Mao et al., 2019). Whether prime editing can be used for dicotyledonous plants such as tomato, is unknown. Here, we report successful adoption of prime editors for use in tomato through codon and promoter optimization.

The prime editing system consists of three parts: an nCas9‐MMLV (engineered Moloney murine leukaemia virus reverse transcriptase) fusion protein, a prime editing guide RNA (pegRNA) and a small guide RNA (sgRNA) for nicking. We incorporated the mammalian prime editing system into a plant binary vector for expression in tomato, generating pCXPE01. As shown in Figure 1a, the commonly used CaMV 35S promoter (2x35S) was used to express the nCas9‐hMMLV (human codon‐optimized MMLV) fusion protein while pegRNA and sgRNA were driven by the U6 promoter of Arabidopsis. In order to test whether the system may work in tomato, we constructed a dual‐luciferase reporter system, where the NanoLuc, an engineered super sensitive luciferase, was completely disabled by introducing frame‐shift mutations, a two nucleotide deletion and six nucleotide substitution (NanoLucM). Only precise editing on NanoLucM can restore its luciferase activity, and the efficiency could be sensitively quantified through luminescence measurement, using the firefly luciferase as an internal control (Figure 1b). Two pegRNAs, pegRNA‐12 and pegRNA‐13, were designed with 13 and 14 nt PBS (primer binding site), respectively, and a 23 nt RT (reverse transcription) template. Each was accompanied by a sgRNA for nicking at a site located 32‐nt or 49‐nt downstream from the pegRNA nicking sites (Figure 1b). The two pCXPE01 constructs were each introduced into tomato leaves together with the Dual‐LucM reporter using biolistic bombardment. Five days later, we detected the restored luminescence in both samples of pegRNA‐12 and pegRNA‐13, with an average efficiency of 0.26% compared with the control Dual‐Luc that was counted as 100%. These results indicated that the primer editor can be used in tomato.

Previous studies on base editing in dicots showed that improvement of nCas9 expression level could significantly increase the editing efficiency (Kang et al., 2018). Therefore, we sought to optimize pCXPE01 to improve editing efficiency by increasing nCas9‐MMLV expression level. We replaced the hMMLV with a plant codon‐optimized MMLV (pMMLV), generating pCXPE02. Transient expression assays on the pegRNA‐12 and pegRNA‐13 sites using the same Dual‐LucM reporter described above resulted in a 3.2‐fold improvement compared with that of pCXPE01 (0.85% vs. 0.26%). Then, we replaced the 35S promoter with the ribosomal protein S5A (RPS5A) promoter of tomato (pCXPE03), which increased the average efficiency to 2.6%, approximately 10 times higher than that of the original pCXPE01 (Figure 1c). Such improvements are consistent with previous reports of using the RPS5A promoter to improve base editing efficiencies in the dicotyledonous plant Arabidopsis (Kang et al., 2018). The improved prime editing frequency in tomato leaves was comparable to that in monocots reported recently.

To determine whether the optimized primer editor pCXPE03 may be used to edit endogenous genes in tomato, three tomato genes, GAI (Solyc11g011260), ALS2 (Solyc03g044330) and PDS1 (Solyc03g123760), were tested. In order to make a clear distinction between prime editing and the random indels caused by nCas9, multiplex base substitutions and/or insertions were designed for introduction into these genes using a total of seven pegRNAs (Figure 1f). Each pegRNA contains a 14 nt PBS and a 13–21 nt RT template. Corresponding plasmids were constructed using pCXPE03 and introduced into tomato (Micro‐Tom) using Agrobacteria. After six weeks of selection on hygromycin‐containing medium, 280 regenerated shoots (Figure 1d) were mixed together for DNA extraction and genotyping using next‐generation sequencing (NGS). Analysis of a total of 12,820,501 NGS reads detected desired prime editing sequences in four pegRNA sites, including pegRNA‐22, pegRNA‐24, pegRNA‐25 and pegRNA‐27, with frequencies ranging from 0.025% to 1.66% (Figure 1f). Such efficiencies were comparable with the reported prime editing results in rice determined using NGS. Similar to the prime editing in rice, undesired by‐product sequences were also observed at all targets in tomato at frequencies ranging from 0.5% to 4.9%, possibly due to the nicking activity of nCas9.

We regenerated hundreds of T0 tomato plants (Figure 1e) for the three genes and genotyped 124 transgenic seedlings using Sanger sequencing. According to the sequencing results, we detected desired edits at two genes, ALS2 and PDS1. We found that 2 out of 30 ALS2 lines (PESG25#13 and #27, 6.7%) contained the desired CAG‐to‐GAT multi‐nucleotide substitutions at the pegRNA‐25 site, resulting in the S642I amino acid change in ALS2. For the PDS1 that encodes the essential phytoene desaturase, we identified one mutant PESG27#17 (1 out of 29, 3.4%) having the designed CG‐insertion at the pegRNA‐27 site. In all previous reports of prime editing on endogenous genes in wheat and rice, only chimeric or heterozygous edited plants were produced except for one pegRNA targeting OsALS2 that produced one homozygous rice line. Accordingly, no plant phenotype results were reported in these studies. Similarly, our prime‐edited T0 tomato plants were chimeras and did not display any obvious phenotypes (Figure 1e, g and h). Therefore, for both monocots and dicots, assessment of the utility of prime editing awaits future analysis of large populations of edited lines and their off‐springs. Editing frequencies vary at the seven sites. Higher efficiencies at the pegRNA‐22, pegRNA‐25 and pegRNA‐27 sites may be due to the nicking positions of sgRNA (+4, +4, +3) that were located closer to the pegRNAs, consistent with the PE3 design strategy of prime editing in mammalian cells. We note that multiplex base substitutions and/or insertions were tested here; it is possible prime editing may yield higher efficiencies for simpler base changes (e.g. one‐nucleotide substitution; Anzalone et al., 2019). Regardless, the editing results here suggest that we can use pCXPE03 for prime editing in tomato.

Compared to that in monocots, base editors do not function well in dicots and thus need to be improved (Kang et al., 2018). Here, through codon and promoter changes, we have improved the efficiency of prime editing considerably in tomato, to levels comparable to those in rice. Further improvements would make prime editing a useful tool for precise genome editing in plant research and breeding.

CRISPR/Cas介导的基因组编辑技术已被广泛应用于通过在各种植物物种中产生短插入或缺失（indel）来创建基因的敲除等位基因。由于同源定向修复（HDR）的效率低下以及HDR DNA模板的传递困难，精确的基因组编辑在植物中仍然具有挑战性（Mao et al ., 2019 ）。最近开发了一种串联重复 HDR 方法，用于水稻中的序列替换，这对于单子叶植物最有用（Lu等人， 2020 ）。由 Cas9 切口酶与胞嘧啶和腺嘌呤脱氨酶融合开发的碱基编辑器可实现靶向 C 至 T 或 A 至 G 替换，但仅限于特定类型的碱基替换和目标位点选择（Mao等人， 2019 ）。在哺乳动物细胞中开发了一种“搜索和替换”方法，也称为初始编辑，该方法可以在目标位点上进行用户定义的序列更改，而无需 DSB 或 DNA 修复模板的传递（Anzalone等人， 2019 ） ）。几个研究小组已采用这种方法用于单子叶植物，包括水稻和小麦（Butt等人， 2020 ；Hua等人， 2020 ；Li等人， 2020 ；Lin等人， 2020 ；Tang等人，2020）。 ， 2020 ；徐等人， 2020 ）。由于尚不清楚的原因，尽管碱基编辑在水稻等单子叶植物中非常高效，但在双子叶植物中效率却很低（Kang et al ., 2018 ；Mao et al . 2018）。， 2019 ）。引物编辑是否可用于双子叶植物（例如番茄）尚不清楚。在这里，我们报告通过密码子和启动子优化成功采用 Prime 编辑器用于番茄。

引物编辑系统由三部分组成：nCas9-MMLV（工程莫洛尼鼠白血病病毒逆转录酶）融合蛋白、引物编辑引导RNA（pegRNA）和用于切口的小引导RNA（sgRNA）。我们将哺乳动物引物编辑系统整合到植物二元载体中以在番茄中表达，生成pCXPE01。如图1a所示，常用的CaMV 35S启动子（2x35S）用于表达nCas9-hMMLV（人类密码子优化的MMLV）融合蛋白，而pegRNA和sgRNA则由拟南芥的U6启动子驱动。为了测试该系统是否可以在番茄中发挥作用，我们构建了一个双荧光素酶报告系统，其中NanoLuc（一种工程超灵敏荧光素酶）通过引入移码突变、两个核苷酸缺失和六个核苷酸取代而被完全禁用（纳米卢克M）。只有对 NanoLucM 进行精确编辑才能恢复其荧光素酶活性，并且可以使用萤火虫荧光素酶作为内部对照，通过发光测量灵敏地量化效率（图 1b）。两种 pegRNA，pegRNA-12 和 pegRNA-13，分别设计有 13 nt PBS（引物结合位点）和 23 nt RT（逆转录）模板。每个都伴有一个 sgRNA，用于在 pegRNA 切口位点下游 32-nt 或 49-nt 处进行切口（图 1b）。使用基因枪轰击将两个 pCXPE01 构建体与 Dual-LucM 报告基因一起引入番茄叶片中。五天后，我们检测到 pegRNA-12 和 pegRNA-13 样品的发光恢复，与算作 100% 的对照 Dual-Luc 相比，平均效率为 0.26%。这些结果表明引物编辑器可用于番茄。

先前对双子叶植物碱基编辑的研究表明，nCas9表达水平的提高可以显着提高编辑效率（Kang et al ., 2018 ）。因此，我们试图通过增加 nCas9-MMLV 表达水平来优化 pCXPE01 以提高编辑效率。我们用植物密码子优化的 MMLV (pMMLV) 替换 hMMLV，生成 pCXPE02。使用上述相同的 Dual-LucM 报告基因对 pegRNA-12 和 pegRNA-13 位点进行瞬时表达测定，与 pCXPE01 相比，结果提高了 3.2 倍（0.85% 与 0.26%）。然后，我们用番茄核糖体蛋白S5A（ RPS5A ）启动子（pCXPE03）替换了35S启动子，这将平均效率提高到2.6％，比原始pCXPE01高约10倍（图1c）。这种改进与之前关于使用 RPS5A 启动子提高双子叶植物拟南芥碱基编辑效率的报道一致（Kang et al ., 2018 ）。番茄叶片中提高的引物编辑频率与最近报道的单子叶植物中的相似。

为了确定优化的引物编辑器 pCXPE03 是否可用于编辑番茄中的内源基因，测试了三个番茄基因： GAI (Solyc11g011260)、 ALS2 (Solyc03g044330) 和PDS1 (Solyc03g123760)。为了明确区分prime编辑和nCas9引起的随机插入缺失，设计了多重碱基替换和/或插入，总共使用七个pegRNA引入这些基因（图1f）。每个 pegRNA 包含一个 14 nt PBS 和一个 13-21 nt RT 模板。使用 pCXPE03 构建相应的质粒，并使用农杆菌导入番茄 (Micro-Tom)。在含潮霉素的培养基上进行六周的选择后，将 280 个再生芽（图 1d）混合在一起，进行 DNA 提取并使用下一代测序 (NGS) 进行基因分型。对总共 12,820,501 个 NGS 读取的分析在四个 pegRNA 位点（包括 pegRNA-22、pegRNA-24、pegRNA-25 和 pegRNA-27）中检测到所需的引物编辑序列，频率范围为 0.025% 至 1.66%（图 1f）。这种效率与使用 NGS 确定的水稻中报道的 Prime 编辑结果相当。与水稻中的prime编辑类似，在番茄的所有靶标中也观察到了不需要的副产物序列，频率范围为0.5%至4.9%，这可能是由于nCas9的切口活性所致。

我们针对这三个基因再生了数百株 T0 番茄植株（图 1e），并使用桑格测序对 124 株转基因幼苗进行了基因分型。根据测序结果，我们在两个基因ALS2和PDS1上检测到所需的编辑。我们发现 30 个 ALS2 系中有 2 个（PESG25#13 和 #27，6.7%）在 pegRNA-25 位点包含所需的 CAG 至 GAT 多核苷酸取代，导致 ALS2 中的 S642I 氨基酸发生变化。对于编码必需的八氢番茄红素去饱和酶的PDS1 ，我们鉴定了一种突变体 PESG27#17（29 个中的 1 个，3.4%）在 pegRNA-27 位点具有设计的 CG 插入。在之前对小麦和水稻内源基因进行初等编辑的所有报告中，除了一种靶向OsALS2的 pegRNA 产生了一种纯合水稻品系外，仅产生了嵌合或杂合编辑植物。因此，这些研究中没有报告植物表型结果。同样，我们的prime编辑的T0番茄植株是嵌合体，没有表现出任何明显的表型（图1e、g和h）。因此，对于单子叶植物和双子叶植物，对原代编辑效用的评估有待未来对大量编辑品系及其后代的分析。七个站点的编辑频率各不相同。 pegRNA-22、pegRNA-25 和 pegRNA-27 位点的较高效率可能是由于 sgRNA 的切口位置（+4、+4、+3）更靠近 pegRNA，这与 PE3 的设计策略一致哺乳动物细胞中的主要编辑。我们注意到这里测试了多重碱基替换和/或插入；对于更简单的碱基变化，引物编辑可能会产生更高的效率（例如单核苷酸取代；Anzalone等人， 2019 ）。 不管怎样，这里的编辑结果表明我们可以使用 pCXPE03 在番茄中进行prime编辑。

与单子叶植物相比，碱基编辑器在双子叶植物中的功能不佳，因此需要改进（Kang et al ., 2018 ）。在这里，通过密码子和启动子的改变，我们显着提高了番茄中引物编辑的效率，达到了与水稻相当的水平。进一步的改进将使prime编辑成为植物研究和育种中精确基因组编辑的有用工具。