Dear Editor,

Classical CRISPR‐Cas systems introduce a DNA double‐strand break (DSB) at target genomic loci. In plant, DSBs are typically repaired through the error‐prone nonhomologous end joining (NHEJ) pathway and result in small InDels (Chen et al. , 2019). Recently, base editors (BEs), including cytosine BEs (CBEs) and adenine BEs (ABEs), were developed to introduce precise nucleotide substitutions by combining the CRISPR‐Cas system with engineered nucleotide deaminases (Gaudelli et al. , 2017; Komor et al. , 2016). These BEs enable precise conversions between A∙T and G∙C pairs in the eukaryotic genome without introducing DSBs. ABE was developed by fusing directly evolved E. coli TadA tRNA adenosine deaminase (ecTadA*7.10) to Sp Cas9 nickase (Gaudelli et al. , 2017). In plants, the efficiency of ABE tools is generally limited and varies greatly among different targets (Hua et al. , 2019). Several efforts have been made to enhance the A∙T‐to‐G∙C conversion activity of ABE by adjusting the number and location of nuclear location signals (NLSs) or the architecture of ABE (Hua et al. , 2019; Li et al. , 2018). However, the efficiency improvement of ABE is still eagerly wanted for robust base conversions in plant genome. Here, we report selection‐based enrichment strategies to achieve highly efficient adenine editing in rice.

We previously established a plant ABE tool (the pHUN411‐ABE vector) exhibiting limited efficiency (Li et al. , 2019). Another study indicated that by using an ACC‐1 single guide RNA (sgRNA), the ABE system can introduce a T‐to‐C conversion corresponding to the dominant C2186R mutation of OsACC, which confers aryloxyphenoxypropionate (APP) herbicide resistance in rice (Li et al. , 2018). This process may provide a marker to select base‐edited plants under herbicide pressure thus may enhance editing efficiency by enriching cells with a functional ABE system. To test this hypothesis, the ACC‐1 sgRNA and the sgRNA for desired target were constructed into single pHUN411‐ABE vector for simultaneous expression (Figure 1a). The vectors were introduced into rice (Oryza sativa cv. Nipponbare) calli via Agrobacterium‐ mediated transformation. The resistant calli were selected by hygromycin with or without the APP herbicide haloxyfop‐R‐methyl for 14 days. For each transformant, ~200 newly emerged calli were collected as a sample to analyse editing by amplicon‐based next generation sequencing (NGS, BioProject PRJNA588580). The frequency of base conversion in T₉ of the Pid3‐1 site and T₅/T₆ of the WX‐2 site was increased under double selections relative to that selected by hygromycin alone (Figure 1b). However, no significant enhancement in the frequency of nucleotide substitution was obtained at A₆ of the WX‐1 site. Plants were regenerated under continuous selections, and then mutations were identified. Up to 31.3% of the hygromycin‐selected plants carried base conversion at the desired targets, whereas the mutant frequencies were greater than 50.0% under double selection of hygromycin and the herbicide (Figure 1c). The results demonstrate that the selection of ACC‐1‐edited plants might enhance base editing of the ABE system at some targets. Because the activity of ABEs may vary greatly among different targets (Hua et al. , 2018; Kang et al. , 2018; Li et al. , 2019; Yan et al. , 2018), the presence of a functional ABE for ACC‐1 editing cannot guarantee the editing of the coexpressed sgRNA in the same cells. In addition, the coedited herbicide selection marker needs to be removed through segregation by self‐crossing or backcrossing in plants, which may limit the practical application of this strategy.

BE expression level is closely correlated with editing efficiency (Koblan et al. , 2018). Thus, base conversions should be more easily produced in plants with higher expression levels of BE. To obtain regenerated plants with high ABE expression levels, single transcriptional and translational unit (STTU) ABE systems were constructed by fusing hygromycin phosphotransferase (HPT) to the N or C terminus of the ecTadA‐ecTadA*7.10‐nSp Cas9 region with a self‐cleavage 2A peptide (Figure 1d). The ABE coding region in pHUN411‐ABE was first replaced by HPT‐ABE (HABE ) or ABE‐HPT (ABEH) fusion. To avoid duplicate HPTs , the original HPT in the vector backbone was replaced by a Mannose‐6‐phosphate isomerase (PMI ) marker, generating the pPUN411‐HABE and pPUN411‐ABEH vectors (Figure 1d). The constructs were introduced into rice by Agrobacterium ‐mediated transformation and selected by hygromycin. In resistant calli after 14‐day‐selection, amplicon‐based NGS indicated HABE generated significantly higher editing frequencies at 4 out of 6 desired sites compared with ABE (Figure 1e, BioProject PRJNA576084). At T₉ of the Pid3‐1 site, the frequency of the conversion generated by HABE was increased as much as 2.7‐fold that achieved by ABE. Compared with the editing efficiencies of ABE, those of ABEH were significantly increased by 1.9‐ to 4.5‐fold at all 6 sites. The results indicate that the editing activity of ABE could be enhanced by HPT fusion. Interestingly, the editing efficiencies of ABEH were significantly higher than those of HABE at 4 sites, suggesting that C‐terminal HPT fusion may enhance the editing efficiency of the plant ABE system to a great extent than N‐terminal fusion. The base editing induced by HABE and ABEH was further determined in the regenerated plants (Figure 1f). Unlike the limited mutants generated by the unmodified ABE, the majority of transgenic plants of the HPT‐fused ABEs carried base conversions in the editing window of the target. Using the ABEH tool, >97.9% plants were edited at all targets, which suggests that the pPUN411‐ABEH vector can provide robust and efficient base editing in rice. During stable transformation, the cells with stronger expression of the selection marker will grow faster under selection pressure and have a greater opportunity to regenerate plants. However, in standard ABE systems, such as the pHUN411‐ABE system, ABE and HPT are typically expressed in different cassettes. Therefore, antibiotic‐resistant cells with high levels of HPT may not have enough expression of ABEs. In this report, we provide a strategy to fuse ABE and HPT in a STTU system, which can be expected to result in synchronization of their expression levels. Antibiotic selection on HPT expression thus would enrich cells with high ABE expression level. To test this hypothesis, the transcript levels of the different ABEs in the regenerated plants were determined by quantitative reverse transcription PCR (qRT‐PCR). The ABE expression levels of the pPUN411‐HABE/ABEH plants were significantly higher than those of the pHUN411‐ABE plants (Figure 1g, one‐way ANOVA), confirming the stronger expression of ABE in the STTU system. Although ABE expression did not significantly differ between the ABEH and HABE plants, we believe that the C‐terminal HPT of the ABEH fusion may avoid the incomplete transcription or translation of ABE in the resistant cells, thereby providing increased editing activity in hygromycin‐resistant cells.

CRISPR genome editing systems are frequently introduced into plants by Agrobacterium ‐mediated stable transformation. The inserted T‐DNA fragment may need to remove in edited lines by segregation. Most of plant CRISPR systems use negative selection, such as an antibiotic or herbicide selection. Generally, negative selection is lethal to untransformed cells. If negative selection is applied to segregate the T‐DNA‐free lines in the T₁ generation, desirable plants without T‐DNA insertions may be subject to toxicity and thus hardly to be recovered. In contrast, positive selection methods (e.g. mannose) typically inhibit but do not kill untransformed cells. To screen T‐DNA‐free progeny lines, the T₀ seeds of pPUN411‐ABEH were selected by hygromycin or mannose pressure for 4 days, and then, sensitive seeds were rescued without selection for another 4 days (Figure 1h). We found that the recovery of mannose‐selected seeds was much easier than that of seeds selected by hygromycin. Furthermore, PCRs with sequence‐specific primers confirmed that 97.8% (45 out 46) rescued mannose‐sensitive plants were T‐DNA‐free. These results suggested the PMI‐mannose selection in genome editing system can facilitate the segregation of T‐DNA‐free plants. Taken together, our results provide an efficient and easy‐to‐use ABE system for plant adenine base editing. More importantly, the study offers a general strategy to optimize efficiency of stable transformation‐based plant genome editing.

亲爱的编辑，

经典CRISPR‐Cas系统在目标基因组位点引入DNA双链断裂（DSB）。在植物中，通常通过易错的非同源末端连接（NHEJ）途径修复DSB，并产生较小的InDels（Chen等人，2019）。最近，开发了碱基编辑器（BEs），包括胞嘧啶BEs（CBEs）和腺嘌呤BEs（ABEs），通过将CRISPR‐Cas系统与工程核苷酸脱氨酶结合使用来引入精确的核苷酸取代（Gaudelli等，2017 ; Komor等。，2016）。这些BE可以在真核基因组中实现A∙T和G∙C对之间的精确转换，而无需引入DSB。ABE是通过融合直接进化而开发的大肠杆菌TadA tRNA腺苷脱氨酶（ecTadA * 7.10）转为Sp Cas9切口酶（Gaudelli等，2017）。在植物中，ABE工具的效率通常受到限制，并且在不同目标之间差异很大（Hua et al。，2019）。通过调整核定位信号（NLSs）的数量和位置或ABE的结构，为增强ABE从A∙T到G∙C的转化活性做出了一些努力（Hua等人，2019 ; Li等人。，2018）。然而，仍然急切需要ABE的效率提高以在植物基因组中进行可靠的碱基转化。在这里，我们报告了基于选择的富集策略，以实现水稻中高效的腺嘌呤编辑。

我们先前建立了效率有限的工厂ABE工具（pHUN411‐ABE载体）（Li等，2019）。另一项研究表明，通过使用ACC-1单向导RNA（sgRNA），ABE系统可以引入与OsACC的显性C2186R突变相对应的T到C转化，赋予水稻（Li）苯氧基丙氧基丙酸酯（APP）除草剂抗性等人，2018）。该过程可能为在除草剂压力下选择基础编辑植物提供了标记，因此可以通过使用功能性ABE系统富集细胞来提高编辑效率。为了验证该假设，将ACC-1 sgRNA和所需靶标的sgRNA构建到单个pHUN411-ABE载体中以同时表达（图1a）。的载体导入水稻（稻栽培品种日本晴）的愈伤组织通过农杆菌介导的转化。潮霉素在有或没有APP除草剂haloxyfop-R-methyl的情况下选择了14天的抗性愈伤组织。对于每个转化体，收集了约200个新出现的愈伤组织作为样本，以通过基于扩增子的下一代测序（NGS，BioProject PRJNA588580）来分析编辑。T _9中的基数转换频率在两次选择下，Pid3-1-1位点和WX-2位点的T ₅ / T ₆相对于单独用潮霉素选择的位点有所增加（图1b）。然而，在A ₆处没有获得核苷酸取代频率的显着提高。WX-1站点的名称。在连续选择下再生植物，然后鉴定突变。多达31.3％的潮霉素选择植物在所需的靶标上进行了碱基转化，而在潮霉素和除草剂的双重选择下，突变频率大于50.0％（图1c）。结果表明，在某些目标上选择ACC-1编辑的植物可能会增强ABE系统的基础编辑。由于ABE的活性在不同的靶标之间可能有很大差异（Hua等人，2018 ; Kang等人，2018 ; Li等人，2019 ; Yan等人，2018），用于ACC-1编辑的功能性ABE的存在不能保证在同一细胞中对共表达的sgRNA进行编辑。此外，需要通过在植物中通过自交或回交进行分离来去除共同编辑的除草剂选择标记，这可能会限制该策略的实际应用。

BE表达水平与编辑效率紧密相关（Koblan等，2018）。因此，在具有较高BE表达水平的植物中应该更容易产生碱基转化。为了获得具有高ABE表达水平的再生植物，通过将潮霉素磷酸转移酶（HPT）融合到ecTadA-ecTadA * 7.10-n Sp Cas9区域的N或C末端来构建单转录和翻译单位（STTU）ABE系统裂解2A肽（图1d）。pHUN411-ABE中的ABE编码区首先被HPT-ABE（HABE）或ABE-HPT（ABEH）融合取代。为了避免重复HPTS，原来HPT载体主链中的甘露糖被6-甘露糖异构酶（PMI）标记所取代，生成了pPUN411-HABE和pPUN411-ABEH载体（图1d）。通过农杆菌介导的转化将构建体引入水稻，并通过潮霉素进行选择。在14天选择后的抗性愈伤组织中，基于扩增子的NGS表明，HABE在6个所需位点中的4个位点产生了比ABE更高的编辑频率（图1e，BioProject PRJNA576084）。在T ₉在Pid3-1位点中，HABE产生的转换频率增加了ABE所获得的转换频率的2.7倍。与ABE的编辑效率相比，在所有6个位置，ABEH的编辑效率均显着提高了1.9到4.5倍。结果表明，通过HPT融合可以增强ABE的编辑活性。有趣的是，ABEH的编辑效率在4个位点上明显高于HABE，这表明C末端HPT融合可能比N末端融合在很大程度上增强了植物ABE系统的编辑效率。由HABE和ABEH诱导的碱基编辑在再生植物中进一步确定（图1f）。与未经修饰的ABE生成的有限突变体不同，HPT融合的ABE的大多数转基因植物在靶标的编辑窗口中进行了碱基转化。使用ABEH工具，在所有靶标上均编辑了> 97.9％的植物，这表明pPUN411‐ABEH载体可为水稻提供强大而有效的碱基编辑。在稳定转化过程中，具有较强选择标记表达的细胞在选择压力下将生长得更快，并具有更大的植物再生机会。但是，在标准ABE系统（例如pHUN411-ABE系统）中，ABE和HPT通常在不同的盒中表达。因此，高水平HPT的抗生素耐药细胞可能没有足够的ABE表达。在此报告中，我们提供了在STTU系统中融合ABE和HPT的策略，可以预期这将导致它们的表达水平同步。因此，对HPT表达的抗生素选择将使细胞具有较高的ABE表达水平。为了验证这一假设，通过定量逆转录PCR（qRT-PCR）确定了再生植物中不同ABE的转录水平。pPUN411‐HABE / ABEH植物的ABE表达水平显着高于pHUN411‐ABE植物的ABE表达水平（图1g，单向ANOVA），证实了ATU在STTU系统中的表达更强。尽管ABEH和HABE植物之间的ABE表达没有显着差异，但我们认为ABEH融合体的C末端HPT可以避免抗性细胞中ABE的不完全转录或翻译，从而在潮霉素抗性细胞中提供增强的编辑活性。为了验证这一假设，通过定量逆转录PCR（qRT-PCR）确定了再生植物中不同ABE的转录水平。pPUN411‐HABE / ABEH植物的ABE表达水平显着高于pHUN411‐ABE植物的ABE表达水平（图1g，单向方差分析），证实了ATU在STTU系统中的表达更强。尽管ABEH和HABE植物之间的ABE表达没有显着差异，但我们认为ABEH融合体的C末端HPT可以避免抗性细胞中ABE的不完全转录或翻译，从而在潮霉素抗性细胞中提供增强的编辑活性。为了验证这一假设，通过定量逆转录PCR（qRT-PCR）确定了再生植物中不同ABE的转录水平。pPUN411‐HABE / ABEH植物的ABE表达水平显着高于pHUN411‐ABE植物的ABE表达水平（图1g，单向ANOVA），证实了ATU在STTU系统中的表达更强。尽管ABEH和HABE植物之间的ABE表达没有显着差异，但我们认为ABEH融合体的C末端HPT可以避免抗性细胞中ABE的不完全转录或翻译，从而在潮霉素抗性细胞中提供增强的编辑活性。pPUN411‐HABE / ABEH植物的ABE表达水平显着高于pHUN411‐ABE植物的ABE表达水平（图1g，单向ANOVA），证实了ATU在STTU系统中的表达更强。尽管ABEH和HABE植物之间的ABE表达没有显着差异，但我们认为ABEH融合体的C末端HPT可以避免抗性细胞中ABE的不完全转录或翻译，从而在潮霉素抗性细胞中提供增强的编辑活性。pPUN411‐HABE / ABEH植物的ABE表达水平显着高于pHUN411‐ABE植物的ABE表达水平（图1g，单向ANOVA），证实了ATU在STTU系统中的表达更强。尽管ABEH和HABE植物之间的ABE表达没有显着差异，但我们认为ABEH融合体的C末端HPT可以避免抗性细胞中ABE的不完全转录或翻译，从而在潮霉素抗性细胞中提供增强的编辑活性。

CRISPR基因组编辑系统经常通过农杆菌介导的稳定转化引入植物。插入的T-DNA片段可能需要通过分离在编辑的行中删除。大多数植物CRISPR系统使用否定选择，例如抗生素或除草剂选择。通常，否定选择对未转化的细胞是致命的。如果采用负选择来分离T ₁代中不含T-DNA的品系，则没有插入T-DNA的理想植物可能会受到毒性作用，因此很难恢复。相反，阳性选择方法（例如甘露糖）通常会抑制但不会杀死未转化的细胞。要筛选不含T‐DNA的后代系，T ₀通过潮霉素或甘露糖压力选择pPUN411‐ABEH的种子4天，然后将敏感种子挽救而无需再选择4天（图1h）。我们发现，甘露糖选择的种子的恢复比潮霉素选择的种子的恢复容易得多。此外，使用序列特异性引物进行的PCR证实97.8％（46种中的45种）拯救的甘露糖敏感植物不含T-DNA。这些结果表明，在基因组编辑系统中选择PMI-甘露糖可以促进无T-DNA植物的分离。综上所述，我们的结果为植物腺嘌呤碱基的编辑提供了一个高效且易于使用的ABE系统。更重要的是，该研究提供了一种总体策略，可优化基于稳定转化的植物基因组编辑的效率。