A plant genomic DNA DSB may be repaired by two major competing mechanisms, namely, cNHEJ and HDR. The GT approach is based on the HDR mechanism; therefore, to increase its efficacy, blocking the cNHEJ pathway is a good option (Van Vu et al., 2019). A number of studies have been published regarding the uses of chemical treatments for blocking cNHEJ to enhance GT efficiency in mammals (Chu et al., 2015; Maruyama et al., 2015; Robert et al., 2015; Zhang et al., 2017). However, only limited information regarding the applications of the chemicals to plant GT is currently available. Therefore, we selected SCR7 (an inhibitor of DNA ligase IV), NU7441 [a DNA-dependent protein kinase (DNA-PKcs) inhibitor], and KU0060648 (a dual inhibitor of DNA-PKcs and phosphatidylinositol-3 kinase (PI-3K)) to determine their effects on plant GT. Previously, a replicon-based CRISPR-Cas-mediated targeted DNA insertion system was successfully developed with SlANT1 as a visible marker (Cermak et al., 2015; Vu et al., 2020). We used SpCas9 (pTC217) and LbCas12a (pHR01 and pMR01) carrying GT vectors to assess the effects of chemical treatments around the replicon peak (3–5 days post-transformation) and GT scoring followed Vu et al. (2020) since the scoring total GT callus number provides richer data than simply scoring GT cotyledon number.

During the GT reaction performed using the SpCas9-based pTC217 vector, treatment with the DNA ligase IV inhibitor led to an improvement in the GT efficiency compared to the mock control (Figure 1A, upper panel). The GT efficiency was increased to around 3.3% with the treatments of either 10 or 50 μM SCR7 compared to the mock 0 μM treatment or the 1 μM SCR7 treatment (Supplementary Table 1). However, Fisher’s LSD test for comparison of the GT efficiency between the treated concentrations and the mock control did not return significant p-values to determine whether the GT enhancement was strong enough (Figure 1A, upper panel). The data indicate that 10 μM SCR7 was the best concentration for SpCas9-based GT in tomato. By contrast, the LbCas12a-based pHR01 construct exhibits a decreasing trend from the mock control to the highest SCR7 concentration. There were mild GT efficiency changes among the mock and 1 or 10 μM SCR7 treatments. Nevertheless, the GT efficiency was dramatically reduced to 4.14 ± 0.66% at 50 μM SCR7 from 6.95 ± 0.91% in the mock treatment (68% reduction) (Figure 1A, bottom panel and Supplementary Table 1). Our observation during the experiment revealed a mildly negative impact of the SCR7 on the viability of the explants at 1 and 10 μM, but more severe at the 50 μM concentration.

To test the impacts of NU7441 and KU0060648 on CRISPR-Cas-based GT in tomato, we employed both single and multiple replicon systems to carry the SpCas9 or LbCas12a constructs. In the case of SpCas9, the GT efficiency was slightly increased (Figure 1B, top panel) when 0.2 μM KU0060648 or 1 μM NU7441, which were within the optimal concentration ranges tested in human cells (Robert et al., 2015), was applied to NSEL medium (Supplementary Figure 1) and incubated from days 3 to 8 post-transformation. However, pairwise comparison resulted in p-values that were not small enough to reject the null hypothesis. A similar situation also occurred for the LbCas12a-based single replicon (pHR01), although NU7441 treatment led to a smaller p-value. In this case, 1 μM NU7441 treatment resulted in GT efficiency at 11.07 ± 1.50% compared to 7.17 ± 1.67% of that of the mock control, which represented a 1.54-fold enhancement (Figure 1B, middle panel and Supplementary Table 2). The effect is considerably more clear with NU7441 treatment using the LbCas12a-based multiple replicon system (pMR01). Blocking cNHEJ with 1 μM NU7441 significantly increased the multireplicon-based GT efficiency from 11.01 ± 0.93% in the mock control to 15.89 ± 1.15%, representing an approximately 1.44-fold change (Figure 1B, bottom panel and Supplementary Table 2). There was almost no GT efficiency change in the case of KU0060648 treatment using both the single- and multi-replicon.

Recently, polyamines, such as putrescine, spermidine, and spermine, were shown to enhance HDR by facilitating RAD51-mediated homologous strand annealing and synaptic complex formation in mice (Lee et al., 2019). We added 1 mM putrescine, spermidine, or spermine to the NSEL medium and incubated the transformed explants for 5 days after cocultivation with agrobacteria carrying the multireplicon pMR01 plasmid. In contrast to the observations made in animals, no improvement in GT efficiency was achieved by any of the polyamines in the two replicates (Supplementary Table 3). In fact, the GT efficiency was reduced with supplementation with longer chain polyamines (i.e., spermidine and spermine) (Supplementary Figure 2). We also observed a higher proliferation of the purple GT calli and a delay in shoot formation from the explants treated with spermidine or spermine compared to the mock control. We surmised that direct supplementation with synthetic polyamines, especially spermidine and spermine, at high concentrations might not facilitate organ regeneration. Thus, we next used silver nitrate (AgNO₃) to increase endogenous polyamines. Silver nitrate was shown to stimulate the activity of arginine decarboxylase (ADC; EC 4.1.1.9), one of the key enzymes of putrescine and polyamine biosynthesis (Hanfrey et al., 2001). Using the multireplicon pMR01 vector for this experiment, we treated the transformed explants with 30 μM AgNO₃ for 5 days on NSEL medium. In two replicates, we observed a significant increase in the number of purple GT spots per explant compared to that of the mock control at 21-day post-transformation (dpt) (Supplementary Figure 3A and Supplementary Table 4). Furthermore, the addition of AgNO₃ at a later stage, in the first subculture with the SEL4C medium, stimulated shoot regeneration from purple calli (Supplementary Figure 3B). This finding is consistent with that silver nitrate enhances somatic embryogenesis (Roustan et al., 1990; Giridhar et al., 2004).

In our previous report, LbCas12a was shown to mediate GT more effectively than the SpCas9 system. The comparisons were conducted at the SlANT1 locus under various experimental conditions (Vu et al., 2020). However, the comparison using the single replicon-based pTC217 and pHR01 constructs might have flawed parameters, such as slightly different SSN-mediated binding and cutting sites and donors. Furthermore, the promoter and terminator driving the expression of SpCas9 and LbCas12a might also contribute to the differences. To better characterize and compare the GT performance of SpCas9 and LbCas12a, we designed a GT system with specific SlANT1 cutting sites that are accessible to both nucleases (Figure 2A and Supplementary Figure 4) and the same promoter and terminator to drive the transcription of Cas nucleases. The comparisons were conducted using single sgRNAs (sgR2^ANT1 vs. crR1.23^ANT1 and crR3.20^ANT1, sgR3^ANT1 vs. crR3.23^ANT1) (Figure 2B and Supplementary Data 1). The 20-nt cRNA was also used for the LbCas12a (crR3.20^ANT1) to mimic the exact binding sequences of the SpCas9 gRNAs (Supplementary Figure 4). Each of the GT tools was expressed from both the replicon and T-DNA to compare the GT efficiency of the two delivery methods (Figure 2B). Our data demonstrate the superiority of the replicon system compared to the T-DNA in plant GT, as the GT efficiencies were enhanced fivefold to eightfold with the replicons compared to that of the T-DNA tools (Figure 2C and Supplementary Figure 5).

Moreover, in keeping with our previous data (Vu et al., 2020), all of the LbCas12a-based GT constructs, except the 20-nt crRNA, exhibited significantly higher GT efficiencies than the SpCas9-based GT tools (Figure 2C, 8.00 ± 1.11% of pHRC01 vs. 5.10 ± 0.35% of pHRC04, 9.53 ± 0.63% of pHRC03 vs. 6.73 ± 0.52% of pHRC05). These results also indicate that the 23-nt crRNA LbCas12a mediated GT more effectively than the 20-nt crRNA (pHRC03 vs. pHRC02). Surprisingly, analysis of the indel mutation efficiency of the guide RNAs at the plant stage demonstrated a reverse correlation between the indel mutation efficiency and the GT efficiency at cutting site 1 and cutting site 2 (Figure 2D, pHRC01 vs. pHRC04, and pHRC03 vs. pHRC05 and Supplementary Figure 5). This result indicates that a considerably stronger indel mutation activity may negatively affect the GT reactions. Since cNHEJ-mediated indel mutations are favored throughout the life cycle of the cells and HDR is limited to the S-G2 phases (Van Vu et al., 2019), DNA DSBs that form at cell cycles other than S-G2 may lead to permanent modifications at the cutting sites, resulting in the inhibition of any further cleavage at these sites in the HDR-favorable cell cycles. Additionally, or alternatively, although relatively low indel-forming activity of LbCas12a compared to SpCas9, recurrent cutting by LbCas12a exposes more frequently the 3′ends of broken DNA, which initiate to copy donor template, consequently enhancing GT frequency. Our data confirmed the stronger activity of SpCas9 in generating DNA DSBs compared to that of LbCas12a in tomato (Figure 2D and Supplementary Figure 5). In the case of the pHRC02 and pHRC03 constructs, the indel efficiency of 20-nt crRNA seemed to higher compared to that of the 23-nt crRNA at the same cutting site but not at the significant p-value (0.0587) (Figure 2D). The highest indel mutation rate was 49.5% on average with sgR2^ANT1 (Supplementary Figure 5). Most of the mutation traces appear to be deletions (Supplementary Figure 6). In the experiment, we obtained GT events with a typical purple phenotype due to the overexpression of SlANT1 and the subsequent accumulation of anthocyanin in the plants (Supplementary Figure 7).

The cleavage activity of Cas nucleases was shown to be temperature-dependent, especially in the case of LbCas12a (Malzahn et al., 2019). Previously, we showed the temperature dependency of CRISPR-Cas-mediated GT in tomato (Vu et al., 2020). In that case, LbCas12a exhibited considerably better GT support at temperatures as high as 31°C compared to that at 19°C or 25°C. This result partially explained why SpCas9 was superior in indel mutation formation compared to LbCas12a at RT. Recently, Puchta’s group reported a temperature-tolerant LbCas12a variant that significantly enhanced indel mutation (Schindele and Puchta, 2020) and GT (Merker et al., 2020a; Huang et al., 2021) efficacy in plants. The group used 24-nt single crRNAs that were released by ribozyme flanking sequences. It is interesting to characterize and compare the ttLbCas12a nuclease using our replicon system and on somatic cells of tomato. Here, we used a multiplexed editing approach with a single array of crRNAs and the gRNA length may affect the processing activity of the LbCas12a (Zetsche et al., 2017). We investigated the impacts of single and dual cleavages using 20- or 23-nt crRNA on the GT performance of both the wild-type and ttLbCas12a variants at the well-characterized SlANT1 locus (Figure 3A, Supplementary Figure 4, and Supplementary Data 1). To contribute to the comparison data, another dual crRNA (crR1-2.23^ANT1) was also tested with LbCas12a_gRNA2 (Supplementary Figure 4), which was previously used by Vu and coworkers (Vu et al., 2020). Our data collected from 6 biological replicates were processed and compared using uncorrected Fisher’s LSD test. The statistical analysis demonstrated very mild changes in GT efficiency among the GT constructs using single crRNAs with both LbCas12a variants (Figure 3B; pHRC11, pHRC12, and pHRC13 for LbCas12a; and pHRC17, pHRC18, and pHRC19 for ttLbCas12a). The ttLbCas12a-based dual crRNA constructs (pHRC20, pHRC21, and pHRC22) showed higher GT efficiencies than LbCas12a (pHRC14, pHRC15, and pHRC16, respectively), although the p-values were only close to 0.05 and significant (Figure 3B). The highest difference in GT efficiency was found between ttLbCas12a-containing pHRC21 (9.74 ± 1.49%) and LbCas12a-containing pHRC15 (6.44 ± 0.75%) with crR1-3.20^ANT1, exhibiting a 1.51-fold change (p = 0.0714) (Figures 3B,C).

In parallel with this experiment, we utilized targeted deep sequencing to analyze the cleavage activity of the nucleases with each of the gRNAs (Figure 3C) at 10 dpt. Due to the large size (∼0.3 × 0.3 cm) of the cotyledon explants used in the previous study, the majority of the cells were not in direct contact with the agrobacteria and thus were not all transformed. Therefore, we reduced the size of the cotyledon explant to ∼0.1 × 0.3 cm (Supplementary Figure 8) to reduce the untransformed cell portion for assessing the editing efficiency by targeted deep sequencing. The data collected from two biological replicates showed strong enhancement of indel mutation efficiencies of a gRNA if it was used in a dual crRNA construct, regardless of the LbCas12a variants. The largest enhancement was from LbCas12a_gRNA1 (23 nt) in the single crRNA crR1.23^ANT1-carrying pHR11 (0.05%) and the dual crRNA crR1-3.23^ANT1-expressing pHRC16 (1.66%) in replicate 1, representing a 33.2-fold increase (Figure 3C). The cleavage activity of ttLbCas12a was also higher than that of the WT variant when single cuts were used. For example, in the case of the crR1.23^ANT1-expressing constructs, the indel mutation efficiency was 5.00- to 7.29-fold increased with ttLbCas12a (pHRC17) compared to LbCas12a (pHRC11) (Figure 3C). For the cases of the dual crRNAs, ttLbCas12a showed more balanced indel mutation efficiencies between the two gRNAs used in the same construct, especially when the dual crRNA crR1-3.20^ANT1 with 20-nt gRNAs was examined (Figure 3C, pHRC21, and pHRC22 compared to pHRC15, and pHRC16, respectively). This activity might be one of the reasons that better enhancement of GT efficiency was mediated by ttLbCas12a compared to the WT version.

To further validate and utilize ttLbCas12a in our research on GT, we compared its performance at the two loci without using any donor-associated selection marker, and no allele-associated selection was employed during the experiments. The salt-tolerant allele SlHKT1;2 (N217D) (Figure 4A) was successfully edited using our LbCas12a-based GT system (Vu et al., 2020), although at low efficiency. We introduced two glyphosate-resistant alleles of tomato 5-enolpyruvylshikimate-3-phosphate synthase 1 (SlEPSPS1): (1) T178I and P182S (TIPS allele, corresponding to T103I and P107S in maize, patent no. US6566587B1) (Lebrun et al., 2003) (Figure 4B) and (2) G177A and A268T (GAAT allele, corresponding to G102A and A193T in maize, patent no. US 6225114 B1) (Eichholtz et al., 2001) (Supplementary Figure 9). Extensive comparisons of GT efficiency between LbCas12a and ttLbCas12a were conducted with single crRNAs (crR1.20^HKT1;2 and crR2.20^HKT1;2) and dual crRNAs (crR1-2.20^HKT1;2 and crR1-2.23^HKT1;2) (Figure 4A, Supplementary Figure 10, and Supplementary Data 1) at the SlHKT1;2 locus. A T-DNA-based GT construct was also used in parallel with the replicon-based construct and the dual crR1-2.20^HKT1;2 construct. Additional comparisons were performed with the dual crRNAs crR1-2.23^EPSPS1 and crR1-3.23^EPSPS1 (Figure 4B, Supplementary Figure 10, and Supplementary Data 1) to replace the TIPS and GAAT alleles, respectively, at the SlEPSPS1 locus.

Gene targeting efficiency calculated by GT deep sequencing reads only showed precise gene editing at both loci in cases of dual crRNAs, although at low efficiencies (Figure 4C). No GT read was obtained from the replicon-based single crRNA constructs and the T-DNA-based dual crRNA regardless of the LbCas12a variants. At the SlHKT1;2 locus, the GT efficiency was higher for ttLbCas12a using the dual 20-nt crRNA (crR1-2.20^HKT1;2), but contrasting results were obtained with the longer dual crRNA (crR1-2.23^HKT1;2) (Figure 4C). However, we obtained higher GT efficiency with the 23-nt dual crRNA (crR1-2.23^EPSPS1) for the replacement of the TIPS allele using ttLbCas12a (Figure 4C). These data indicate that the targeted deep sequencing method may help to identify the GT reads among the constructs, although the read numbers were still too low to be used for statistical comparisons. The indel mutation efficiency obtained from the T-DNA construct was up to 15-fold lower than that of the replicon using the same crR1-2.20^HKT1;2 (Figure 4C). In this experiment, the indel mutation efficiencies of the single crRNAs were also lower than those of the dual crRNAs. The highest indel mutation activity was obtained with crR1-2.20^HKT1;2 and LbCas12a, yielding values of 4.32% of LbCas12a_gRNA1 and 2.56% of LbCas12a_gRNA2. The gap between the indel mutation efficiencies of LbCas12a_gRNA1 and LbCas12a_gRNA2 from crR1-2.20^HKT1;2 was also lower in the case of ttLbCas12a. However, ttLbCas12a did not perform well when it was expressed with the 23-nt dual crRNA (crR1-2.23^HKT1;2); hence, no GT read was obtained at 10 dpt using the combination (Figure 4C). In general, between the two replicates, higher indel activity in a construct was correlated with higher GT efficiency.

Additional experiments for the exchange of the GAAT allele using the 23-nt dual crRNAs with two (crR1-3.23^EPSPS1) or four cleavage sites (crR1-2.23^EPSPS1 and crR2-4.23^EPSPS1) and the LbCas12a variants (Supplementary Figures 9, 10) demonstrated that ttLbCas12a performed better in either case. In triplicate, the highest GT efficiency was obtained with ttLbCas12a-expressing constructs (pHRES2.11 compared to pHRES2.9) using the dual crRNA crR1-3.23^EPSPS1, up to 0.015 ± 0.015% for G177A and 0.018 ± 0.009% for A268T (Figure 5A). Notably, the four crRNA constructs (pHRES2.10 and pHRES2.12) showed a mild reduction in the gRNA1 and gRNA3 indel mutation efficiencies, which were expected to be higher than those of the dual gRNA-expressing constructs due to the synergistic effects of two close cleavage sites (Figure 5A).

To further validate the performance of the LbCas12a variants, we screened and analyzed transformants obtained from the transformation of the LbCas12a- and ttLbCas12a-based constructs for the exchange of the TIPS allele. CAPS screening by BpiI digestion of the PCR products flanking SlEPSPS1-targeted site 1 demonstrated potential TIPS allele-carrying events (Figure 5B). The BpiI site located 84 bp downstream of the T178 codon was modified during the cloning of the homologous DNA donor through the Golden gate cloning method. Therefore, the undigested bands that appeared on the agarose gel were potentially derived from the TIPS-carrying alleles. Using the PCR products of the events that showed undigested bands in the CAPS assay for targeted deep sequencing analysis revealed several transformants carrying the precise GT allele at various frequencies and up to 9.003% in the case of the H281.82 plant generated with the ttLbCas12a construct (Table 1). Interestingly, the GT frequency revealed from ICE Synthego decomposition of the Sanger data of the same event showed only at 7% (Supplementary Figure 11). Two of the transformants (#H281.50 and #H281.82) obtained by the ttLbCas12a-based construct show targeted deep sequencing GT frequency at more than 5% that may be potential for obtaining inherited offsprings (Table 1). Based on that threshold frequency, we revealed that only the ttLbCas12a construct resulted in 2.78% GT efficiency at the plant stage (Figure 5C). Besides, Sanger sequencing of the purified PCR products from 77 and 72 transformants of the wt LbCas12a and ttLbCas12a variant, respectively, and analysis of their.ab1 files by ICE Synthego (Hsiau et al., 2019) identified similar indel mutation efficiencies induced by the Cas proteins (Figure 5D). The average indel mutation efficiency was 29.36 ± 2.84% with LbCas12a and a similar rates for ttLbCas12a (28.03 ± 3.24%) (p = 0.5389, Student’s t-test). Notably, the SlEPSPS1 TIPS allele-carrying GT plants showed no phenotypic changes compared to the WT plant (Supplementary Figure 12, pHRES2.8 events #82). Several events that contained high indel mutation rates exhibited phenotypic defects compared to WT (Supplementary Figure 12, pHRES2.7 events #27 and 28).

Previously, NU7441 exhibited enhancement of LbCas12a-based GT with SlANT1 as a visible marker (Figure 1B). To further optimize ttLbCas12a-based GT in tomato, we investigated the impacts of various NU7441 concentrations on the GT process using ttLbCas12a constructs with crR1-2.20^HKT12. Targeted deep sequencing data demonstrated the enhancement of GT efficiency with 1 μM NU7441 added to the culture medium albeit at low read numbers. However, when the NU7441 concentrations were higher, the GT efficiency was lower than that of the mock control (Figure 5E). Notably, the NU7441 treatments at various concentrations showed only mild changes in the indel mutation efficiency of gRNA1 and gRNA2 of the dual crRNA (Figure 5E).

The SlANT1, SlHKT1;2, and SlEPSPS1 loci were used to conduct HDR-based DNA insertion and allele replacement experiments. The chemical treatment experiments used a single replicon (pTC217 and pHR01) and multiple-replicon tool (pMR01) vectors from previous works (Cermak et al., 2015; Vu et al., 2020) to target the SlANT1 gene. The SpCas9 containing the pTC217 vector was ordered from Addgene (Plasmid #70018) (Cermak et al., 2015). Further works used the single replicon system as a vector for the delivery of guide RNA and CRISPR-Cas expression cassettes and GT donor templates in this study (Figure 2A and Supplementary Data 1).

For comparison of the SpCas9 and LbCas12a complexes in GT performance, we designed an HDR-mediated insertion of selection markers at the SlANT1 locus that was well studied in our laboratory (Vu et al., 2020). The gRNA binding and cutting sites were selected in a manner that could be used for both SpCas9 (SpCas9_gRNA1 and SpCas9_gRNA2) and LbCas12a (LbCas12a_gRNA1 and LbCas12a_gRNA3) (Figure 2A). We used a single gRNA for each of the GT vectors with either SpCas9 or LbCas12a (Figure 2B). At the cutting site of LbCas12a_gRNA3, two different gRNA lengths (20 and 23 nt) were tested (Figure 2B). All the editing constructs in the experiment were delivered by Agrobacterium-mediated transformation of tomato cotyledon explants, and two sets of constructs were cloned: one set with the single geminiviral replicon system (Vu et al., 2020) for the amplification of homologous DNA donors and the other T-DNA set for comparison.

For the assessment and characterization of the activities of ttLbCas12a in GT in comparison with the wild-type version of LbCas12a, similar constructs were designed for targeting the SlANT1 locus with single LbCas12a crRNA expression cassettes. In addition, dual guide RNA expression cassettes combining LbCas12a_gRNA1 (20 and 23 nt) and LbCas12a_gRNA3 (20 and 23 nt) were also used for the GT experiments (Figure 3A and Supplementary Data 1).

A similar system for comparison of ttLbCas12a and wtLbCas12a was designed for targeting the SlHKT1;2 and SlEPSPS1 loci (Figures 4A,B). Single and dual cutting sites were used for SlHKT1;2 (Figure 4A and Supplementary Data 1). The gRNA lengths (20 and 23 nt) were also evaluated at the same locus. For SlEPSPS1, only dual gRNAs 23 nt in length were assessed for GT with the LbCas12a variants (Figure 4B and Supplementary Data 1).

In all the GT vectors, the expression of SpCas9 and LbCas12a variants was driven by a long CaMV 35S promoter that contains an intron (Trp1) at the 5′UTR. A copy of AtUBQ10 intron 1 was also inserted in the coding sequence of the LbCas12a variants (Supplementary Data 1), which was tested previously for GT experiments by Vu et al. (2020). The crRNA and sgRNA expression cassettes were transcribed with the support of the core sequence of the AtU6 promoter (Nekrasov et al., 2013).

Agrobacterium-mediated tomato transformation was conducted following our protocol (Supplementary Figure 1) published previously by Vu and coworkers (Vu et al., 2020). For the treatment of chemicals, the chemicals at the tested concentrations were added to the NSEL medium. The 30 μM of silver nitrate was also added to the first subcultured SEL4C medium at the later stage for stimulating somatic embryogenesis. The total treatment time was 5 days [day 3 to day 8 post transformation (dpt)]. For targeted deep sequencing, cotyledon samples were collected at 10 dpt.

For assessment of the GT efficiency at SlANT1, purple spot counting was conducted at 21 dpt, and purple plants were recorded at the hardening stage. The GT efficiency reflected by the purple spot numbers was calculated with normalization to the SlANT1 overexpression tool that was transformed in parallel with the GT tool. Our calculation method normalized the number of cells/calli that was edited (and hence carrying p35S-ANT1 expression cassette) by the HDR tools to the number of cells that were successfully transformed with a T-DNA harboring over-expression cassette of the ANT1 gene driven by p35S in the same experimental conditions. This calculation may reflect more actual HDR frequency regarding the editing at the cell levels. A similar calculation was used and accepted elsewhere (Puchta et al., 1993, 1996). This calculation method was also explained in our previous report (Vu et al., 2020).

For assessment of the GT efficiency at the SlHKT1;2 and SlEPSPS1 loci, targeted deep sequencing was conducted with thin cotyledon explants collected at 10 dpt. Sanger sequencing was performed to screen and validate GT plants.

Genomic DNAs were isolated from the cotyledon explants (30 explants per sample) or mature leaves (three distinct leaflets from three different young leaves per plant per sample) using the CTAB method. The MiSeq sequencing service (MiniSeq^TM System, Illumina, San Diego, CA, United States) was used. MiSeq samples were prepared in three PCRs according to the manufacturer’s guidelines with genomic DNAs as templates for the first PCR. The first and second PCRs used primers listed in Supplementary Table 5, whereas the third PCRs were performed with the manufacturer’s primers to assign sample IDs. The first PCR primers were designed for binding to the upstream and downstream sequences from the homologous donor sequence junctions to avoid amplifying the donor DNA sequences. The second PCR primers were designed to amplify 150–180-bp flanking the targeted base changes at the targeted sites. High-fidelity DNA Taq polymerase (Phusion, NEB, Ipswich, England) was used for PCR. The MiSeq raw data FASTQ files were analyzed by the Cas-Analyzer tool (Park et al., 2016). The indel analysis window was set to 5 bases, with a comparison range covering both read ends. The GT efficiency was assessed using the corresponding donor sequence as the input HDR donor sequence.

All comparison experiments were conducted with at least three replicates, and data were recorded by purple spot counting, targeted deep sequencing, and plant event screening. Some of the experiments using targeted deep sequencing were conducted in two replicates. The editing data, statistical analysis, and plots were further processed by the MS Excel and GraphPad Prism programs and explained in detail in the legends of figures and/or tables. Pairwise comparison data were tested with Student’s t-test with unequal variance and two-tailed parameters. Similar parameters were applied for multiple comparisons using Fisher’s LSD test. A difference was considered to be significant when the statistical tests returned a p-value < 0.05.