Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)/Cas systems have revolutionized plant breeding. Different Cas enzymes have been widely used to introduce deletions or insertions into plant genomes. However, many important agronomical traits in plants are associated with single-nucleotide polymorphisms. For example, the Lin5 SNP causes an Asn366 to Asp change, resulting in a higher sugar content in tomato fruits (Tieman et al., 2017). Thus, it is important to develop tools to efficiently introduce precise base changes in crops.

CRISPR/Cas-mediated base editing systems can introduce precise base mutations at target genomic sites (Gaudelli et al., 2018). Adenine base editors (ABEs) can convert adenine to guanine (A to G). The ubiquitin promoter-driven ABE works well in rice and wheat (Hua et al., 2018; Li et al., 2018; Zeng et al., 2019). However, the low efficiencies of ABEs in dicot plants remain a significant problem (Kang et al., 2018). A previous study showed that 35S promoter-driven ABE barely has activity, but the AtRPS5A promoter-driven ABE successfully causes A to G conversions in Arabidopsis, suggesting that the promoter used in the ABE system can substantially affect the editing efficiencies (Kang et al., 2018). More work is needed to identify promoters that can be used efficiently to drive ABEs in dicots. Here, we report that the tomato SlEF1α promoter can be used to drive ABEs to achieve efficient adenine base editing in tomato and soybean. This promoter can also be used to drive nCas9 variant-mediated ABEs, such as ABE-nCas9-NG and ABE-XNG-nCas9 to perform adenine base editing at some NG PAM loci in dicot plants.

To screen for effective promoters to drive ABEs in tomato, we expressed ABE7.10-nCas9 in tomato (Ailsa Craig) with seven different promoters, including the 35S, AtUbi, AtRPS5A, SlRPS5A1 (Solyc11g042610), SlRPS5A2 (Solyc10g078620), SlTCTP (Solyc01g099770) and SlEF1α (Solyc06g005060) promoters (Figure 1a). SlTCTP and SlEF1α promoters showed high activities in most tissues of tomato according to published RNA seq data (http://ted.bti.cornell.edu). We compared the activities of the seven ABE systems at six single guide RNA (sgRNAs) sites (Appendix S1). The six sgRNAs were designed to target SlRIN, a transcription factor important for tomato fruit ripening. For each tested promoter and each sgRNA, we obtained about 20 T₀ tomato transformants to examine the editing efficiencies. By performing Sanger sequencing of targeted loci, we identified T₀ transgenic plants that carry base substitutions.

Among the seven tested promoters, p35S- and pAtUbi-driven ABEs barely had activities at all tested sgRNA sites in tomato (Figure 1b). The pAtRPS5A- and pSlRPS5A1-driven ABEs caused A to G conversions at sgRNA1, sgRNA3, sgRNA4 and sgRNA6 with editing efficiencies of 15%–40%; the pSlRPS5A2-ABE-nCas9 was active at only sgRNA1 and sgRNA6 with much lower editing efficiencies (< 15%) compared to pAtRPS5A- and pSlRPS5A1-driven ABEs (Figure 1b). The pTCTP-ABE-nCas9 had low editing efficiencies (<10%) at sgRNA1, sgRNA3 and sgRNA4. In contrast, pSlEF1α-ABE-nCas9 had high editing efficiencies at sgRNA3 (71.4%), sgRNA4 (53.3%) and sgRNA6 (53.5%), and a moderate editing efficiency at sgRNA1 (25.8%) (Figure 1b,c). Among all the seven tested systems, the pSlEF1α-ABE-nCas9 was the most efficient, and its editing efficiency was about two times of that of pSlRPS5A1-ABE-nCas9 (Figure 1b). In addition, our results showed that pSlEF1α-ABE-nCas9 had much higher editing efficiency than p35S- and pAtUbi-driven ABEs at sgRNA24 (GCATTATGCAGGAGAAGAGT) targeting SlMBP7 (Figure 1d). We examined potential off-target editing of pSlEF1α-ABE-nCas9 with sgRNA1, sgRNA3, sgRNA4 and sgRNA6, and did not find mutations at any of the tested potential off-target sites identified using CRISPR-GE (http://skl.scau.edu.cn/). We examined the editing windows of pSlEF1α-, pAtRPS5A- and pSlRPS5A1-driven ABEs in tomato, and found overlapping A/G signals at the A4-A11 positions (Figure 1e,f).

We compared the activities of p35S-, pSlRPS5A1- and pSlEF1α-driven ABEs at GmHPPD (sgRNA19) and GmELF3A (sgRNA20) loci in soybean (Glycine max Zhonghuang 302). We expressed these ABEs in soybean hairy roots. The pSlEF1α-driven ABE has much higher editing efficiencies than p35S- or pSlRPS5A1-driven ABEs at sgRNA19 and sgRNA20 (Figure 1g). These results show that the pSlEF1α-ABE7.10-nCas9 system is efficient in generating A to G conversions in dicot crops, such as tomato and soybean.

The nCas9-mediated base editors require an NGG PAM close to the desired editing site, which severely limits the application of base editors for crop improvements. Previous study reported two SpCas9 variants, spCas9-NG and XNG-spCas9, which can recognize a broad range of NG PAMs in the tomato genome (Niu et al., 2020). We tested the pRPS5A-ABE7.10-nCas9-NG and pRPS5A-ABE7.10-XNG-nCas9 with the nickase mutation (D10A), and they failed to cause A to G conversions at sgRNA14 in T₀ transgenic tomato plants or at sgRNA15-18 in T₁ Arabidopsis (Col-0) transgenic plants (Figure 1h). However, the pSlEF1α-ABE7.10-nCas9-NG showed activity at sgRNA14 in tomato, and both the pSlEF1α-ABE7.10-nCas9-NG and pSlEF1α-ABE7.10-XNG-nCas9 systems showed adenine editing activities at sgRNA15 (GGG PAM) and sgRNA17 (TGT PAM) in T₁ transgenic Arabidopsis plants (Figure 1h,i). In soybean, we designed five sgRNAs to target GmHPPD (sgRNA19 and sgRNA23), GmELF3A (sgRNA20) and GmFAD2 (sgRNA21 and sgRNA22) genes, and pSlEF1α-ABE7.10-nCas9-NG showed adenine editing at sgRNA23 (AGA PAM) and pSlEF1α-ABE7.10-XNG-nCas9 showed activity at sgRNA19 (AGG PAM) (Figure 1j,k). These results suggest that the SlEF1α promoter could be used to drive spCas9 variant-mediated ABE to perform adenine base editing at some NG PAM loci.

Recently, a modified ABE system, TadA8e, was reported to have higher editing efficiency in rice (Yan et al., 2021). We introduced pSlEF1α-ABE7.10-nCas9-NG and pSlEF1α-TadA8e-nCas9-NG into tomato to target the SlRIN gene. Sanger sequencing results showed that both systems were active in catalysing A to G conversions at target site of AGG PAM (sgRNA6) and AGA PAM (sgRNA14) with comparable editing efficiencies in tomato (Figure 1l).

In conclusion, we found that the tomato pSlEF1α promoter can be used to construct very efficient ABEs for dicot crops. Using pSlEF1α-ABE-nCas9, high-efficiency adenine base editing was achieved at genomic sites with NGG PAMs in tomato and soybean. Using pSlEF1α-ABE-XNG-nCas9 or pSlEF1α-ABE-nCas9-NG system, adenine base editing was attained at sites with relaxed NG PAMs such as the TGT PAM in Arabidopsis and AGA PAM in soybean. Further work is required to expand adenine base editing at other non-NGG PAMs.

成簇规则间隔短回文重复序列 (CRISPR)/Cas 系统彻底改变了植物育种。不同的 Cas 酶已被广泛用于在植物基因组中引入缺失或插入。然而，植物中许多重要的农艺性状都与单核苷酸多态性有关。例如，Lin5 SNP 导致 Asn366 变为 Asp，导致番茄果实中的糖分含量更高（Tieman等人，2017 年）。因此，重要的是开发工具以有效地在作物中引入精确的碱基变化。

CRISPR/Cas 介导的碱基编辑系统可以在目标基因组位点引入精确的碱基突变 (Gaudelli et al ., 2018 )。腺嘌呤碱基编辑器 (ABE) 可以将腺嘌呤转化为鸟嘌呤（A 转化为 G）。泛素启动子驱动的 ABE 在水稻和小麦中表现良好（Hua et al ., 2018 ; Li et al ., 2018 ; Zeng et al ., 2019）。然而，双子叶植物中 ABE 的低效率仍然是一个重大问题（Kang等人，2018 年）。之前的一项研究表明，35S启动子驱动的 ABE 几乎没有活动，但AtRPS5A启动子驱动的 ABE 成功地导致拟南芥中 A 到 G 的转换，表明 ABE 系统中使用的启动子可以显着影响编辑效率（Kang等人，2018）。需要做更多的工作来确定可有效用于驱动双子叶植物 ABE 的启动子。在这里，我们报告番茄SlEF1α启动子可用于驱动 ABE，以在番茄和大豆中实现高效的腺嘌呤碱基编辑。该启动子还可用于驱动 nCas9 变体介导的 ABE，例如 ABE-nCas9-NG 和 ABE-XNG-nCas9，以在双子叶植物的某些 NG PAM 位点进行腺嘌呤碱基编辑。

为了筛选在番茄中驱动 ABE 的有效启动子，我们在番茄 (Ailsa Craig) 中用七种不同的启动子表达了 ABE7.10-nCas9，包括35S、AtUbi、AtRPS5A、SlRPS5A1 (Solyc11g042610)、SlRPS5A2 (Solyc10g078620)、SlTCTP (Solyc01g099770 ) 和SlEF1α (Solyc06g005060) 启动子（图 1a）。根据已发表的 RNA 序列数据 (http://ted.bti.cornell.edu)，SlTCTP和SlEF1α启动子在番茄的大部分组织中表现出高活性。我们比较了七个 ABE 系统在六个单向导 RNA (sgRNA) 位点的活动（附录 S1）。六个 sgRNA 被设计为靶向SlRIN，一种对番茄果实成熟很重要的转录因子。对于每个测试的启动子和每个 sgRNA，我们获得了大约 20 个 T ₀番茄转化体来检查编辑效率。通过对目标基因座进行 Sanger 测序，我们确定了携带碱基替换的 T ₀转基因植物。

在七个测试的启动子中，p35S和pAtUbi驱动的 ABE 在番茄中所有测试的 sgRNA 位点几乎没有活动（图 1b）。pAtRPS5A和pSlRPS5A1驱动的ABE 在 sgRNA1、sgRNA3、sgRNA4 和 sgRNA6 处引起 A 到 G 的转换，编辑效率为 15%–40%；与pAtRPS5A和pSlRPS5A1驱动的 ABE 相比， pSlRPS5A2 -ABE-nCas9 仅在 sgRNA1 和 sgRNA6 上具有活性，编辑效率低得多 (< 15%) （图 1b）。pTCTP - ABE -nCas9 在 sgRNA1、sgRNA3 和 sgRNA4 上的编辑效率较低 (<10%)。相反，pSlEF1α-ABE-nCas9 在 sgRNA3 (71.4%)、sgRNA4 (53.3%) 和 sgRNA6 (53.5%) 上具有高编辑效率，在 sgRNA1 (25.8%) 上具有中等编辑效率（图 1b、c）。在所有七个测试系统中，pSlEF1α -ABE-nCas9 的效率最高，其编辑效率约为pSlRPS5A1 -ABE-nCas9 的两倍（图 1b）。此外，我们的结果显示pSlEF1α -ABE-nCas9 在靶向SlMBP7的 sgRNA24 (GCATTATGCAGGAGAAGAGT) 上比p35S和pAtUbi驱动的 ABE 具有更高的编辑效率（图 1d）。我们检查了pSlEF1α的潜在脱靶编辑-ABE-nCas9 与 sgRNA1、sgRNA3、sgRNA4 和 sgRNA6，并且未在使用 CRISPR-GE (http://skl.scau.edu.cn/) 识别的任何测试的潜在脱靶位点发现突变。我们检查了番茄中pSlEF1α-、pAtRPS5A-和pSlRPS5A1-驱动的 ABE 的编辑窗口，并在 A4-A11 位置发现了重叠的 A/G 信号（图 1e、f）。

我们比较了大豆（Glycine max Zhonghuang 302）中GmHPPD (sgRNA19) 和GmELF3A (sgRNA20) 基因座上p35S、pSlRPS5A1和pSlEF1α驱动的 ABE 的活性。我们在大豆毛状根中表达了这些 ABE。pSlEF1α驱动的ABE在 sgRNA19 和 sgRNA20 上的编辑效率比p35S或pSlRPS5A1驱动的 ABE 高得多（图 1g）。这些结果表明，pSlEF1α -ABE7.10-nCas9 系统可有效地在双子叶作物（例如番茄和大豆）中产生 A 到 G 的转化。

nCas9 介导的碱基编辑器需要靠近所需编辑站点的 NGG PAM，这严重限制了碱基编辑器在作物改良中的应用。之前的研究报告了两种 SpCas9 变体，spCas9-NG 和 XNG-spCas9，它们可以识别番茄基因组中广泛的 NG PAM（Niu et al ., 2020）。我们测试了带有切口酶突变 (D10A) 的pRPS5A -ABE7.10-nCas9-NG 和pRPS5A -ABE7.10-XNG-nCas9，它们未能在 T ₀转基因番茄植物的 sgRNA14 或 sgRNA15引起 A 到 G 的转化-18 在 T ₁ 拟南芥(Col-0) 转基因植物中（图 1h）。然而，pSlEF1α-ABE7.10-nCas9-NG 在番茄中显示出对 sgRNA14 的活性，并且pSlEF1α -ABE7.10-nCas9-NG 和pSlEF1α -ABE7.10-XNG-nCas9 系统都在 sgRNA15 (GGG PAM) 和 sgRNA17 显示出腺嘌呤编辑活性(TGT PAM) 在 T ₁转基因拟南芥植物中（图 1h，i）。在大豆中，我们设计了五个 sgRNA 来靶向GmHPPD（sgRNA19 和 sgRNA23）、GmELF3A（sgRNA20）和GmFAD2（sgRNA21 和 sgRNA22）基因，并且pSlEF1α -ABE7.10-nCas9-NG 在 sgRNA23（AGA PAM）和pSlEF1α显示腺嘌呤编辑-ABE7.10-XNG-nCas9 在 sgRNA19 (AGG PAM) 上显示出活性（图 1j、k）。这些结果表明SlEF1α启动子可用于驱动 spCas9 变体介导的 ABE 在某些 NG PAM 位点进行腺嘌呤碱基编辑。

最近，据报道，一种改良的 ABE 系统 TadA8e 在水稻中具有更高的编辑效率（Yan等人，2021 年）。我们将pSlEF1α -ABE7.10-nCas9-NG 和pSlEF1α -TadA8e-nCas9-NG 引入番茄中以靶向SlRIN基因。Sanger 测序结果表明，这两个系统在 AGG PAM (sgRNA6) 和 AGA PAM (sgRNA14) 的靶位点催化 A 到 G 转化方面都很活跃，在番茄中的编辑效率相当（图 1l）。

总之，我们发现番茄pSlEF1α启动子可用于构建非常有效的双子叶作物 ABE。使用pSlEF1α -ABE-nCas9，在番茄和大豆的基因组位点使用 NGG PAM 实现了高效腺嘌呤碱基编辑。使用pSlEF1α -ABE-XNG-nCas 9或pSlEF1α -ABE-nCas9-NG 系统，在具有松弛的 NG PAM（例如拟南芥中的 TGT PAM 和大豆中的 AGA PAM）的位点实现腺嘌呤碱基编辑。需要进一步的工作来扩展其他非 NGG PAM 的腺嘌呤碱基编辑。