Gene Editing Profiles in 94 CRISPR-Cas9 Expressing T0 Transgenic Tobacco Lines Reveal High Frequencies of Chimeric Editing of the Target Gene

Chimeric editing is often reported in gene editing. To assess how the general chimeric editing is, we created a transgenic tobacco line carrying a marker, beta-glucuronidase gene (gusA), introduced a CRISPR-Cas9 editing vector into the transgenic tobacco line for knocking out gusA, and then investigated the gusA editing efficiencies in T0 and subsequent generations. The editing vector carried a Cas9 gene, which was driven by the cauliflower mosaic virus 35S promoter, and two guide RNAs, gRNA1 and gRNA2, which were driven by Arabidopsis U6 (AtU6) and U3 (AtU3) promoter, respectively. The two gRNAs were designed to knock out a 42-nucleotide fragment of the coding region of gusA. The editing vector was transformed into gusA-containing tobacco leaves using Agrobacterium tumefaciens-mediated transformation and hygromycin selection. Hygromycin-resistant, independent T0 transgenic lines were used to evaluate gusA-editing efficiencies through histochemical GUS assays, polymerase chain reactions (PCR), and next-generation sequencing of PCR amplicons. Profiles of targeted sequences of 94 T0 transgenic lines revealed that these lines were regenerated from non-edited cells where subsequent editing occurred and created chimeric-edited cells in these lines during or after regeneration. Two of them had the target fragment of 42 bp pairs of nucleotides removed. Detail analysis showed that on-target mutations at the AtU6-gRNA1 site and the AtU3-gRNA2 site were found in 4.3% and 77.7% of T0 transgenic lines, respectively. To overcome the issue of extremely low editing efficiencies in T0 lines, we conducted a second round of shoot induction from the chimeric line(s) to enhance the success of obtaining lines with all or most cells edited. The mutation profiles in T0 transgenic lines provide valuable information to understand gene editing in plant cells with constitutively expressed CRISPR-Cas9 and gRNAs.


Introduction
The clustered regularly interspaced short palindromic repeats (CRISPR)-associated protein 9 (Cas9) system uses the Cas9 endonuclease led by a guide RNA (gRNA) to target DNA sites through nucleotide base pairing and induce DNA double-strand breaks for short insertion/deletion mutations [1][2][3]. This system has become a very powerful gene editing tool and has been widely used for modifying genes in various plant species [4][5][6][7].
Among other requirements, a desirable gene editing system should allow the effective induction of on-target mutations with minimum occurrences of off-target changes. Mutation frequencies (e.g., percentages of the cells with mutations) are often used to describe gene editing efficiencies. Gene editing frequencies are also described as the percentage of T 0 regenerants in which on-target editing is detected in all cells. As known from many editing studies, many T 0 editing lines or regenerants were chimeric [7][8][9][10][11]. The development of a method for increasing the proportion of non-chimeric/chimeric T 0 editing events is important for enhancing the effectiveness of a gene editing research project. 2 of 11 In a previous gene editing study in which we intended to knock out the reporter gene beta-glucuronidase (gusA) in gusA transgenic blueberry, we observed very low mutation frequencies in the T 0 calli (<6% in the best callus cluster) [8]. While successfully edited plants were recovered through the second round of shoot regeneration from leaf explants of 10 selected T 0 lines, the overall editing efficiency was low (~15%) [8]. In a recent effort to knock out a 10.5 kb transposon from the promoter region of a grape MybA1 gene (VvMybA1), only one out of hundreds T 0 regenerants screened was a non-chimeric edited line, suggesting an extremely low efficiency in the production of putative editing lines even in stable transgenic lines where Cas9 and gRNAs were constitutively expressed [9]. To investigate whether the low editing efficiency observed in the blueberry study was speciesspecific and whether the second round of regeneration would help enhance the recovery of non-chimeric edited plants, we transformed the same editing vector into tobacco and evaluated the gusA editing efficiencies in transgenic calli and lines in T 0 and subsequent second round of regenerants. Because tobacco can easily be transformed, we were able to produce close to 100 independent T 0 lines to evaluate by sequencing in this study. This study also benefited from using gusA as a target gene as it provides an excellent marker to discern edited, non-edited, or chimeric plants through histochemical GUS staining assay. Our results confirmed that most T 0 transformants were chimeric for the target gene editing and the second round of regeneration was useful for increasing the chance of obtaining non-chimeric edited lines.

Results
Hygromycin selection at 20 mg/L was very effective in inhibiting the regeneration of non-transformed cells. All transformed leaf disks produced hygromycin-resistant calli and shoots, and the shoots from different explants or different positions of the same explant were labeled as independent lines because there was barely any chance that they were from the same transformed cells. Histochemical GUS assay revealed that most hygromycinresistant regenerants (hereafter: H-tobacco) showed a mixture of blue and white tissues, indicating that loss of the gusA function occurred in the transformed tissues due to the editing, as intense blue staining was observed in the shoots of kanamycin-resistant tobacco without transformation of the P35S-Cas9-GUS-gRNAs (hereafter: K-tobacco) ( Figure 1A).
To evaluate the editing efficiencies of gusA, 2-3 individual hygromycin-resistant shoots from each leaf explant, selected 10 weeks after inoculation, were subjected to GUS staining.
In two experiments, one had~8.3% and the other had~6.0% of the H-tobacco T 0 lines producing shoots or leaf disks without visible blue staining (Figure 1), which likely resulted from knocking out the gusA gene. This agrees with 7/94 (7.4%) lines having over 98% editing efficiency for gRNA2 in experiment 2 (Table 1). About 50% of the regenerants in each experiment were likely chimeric with both edited and non-edited gusA, because they showed blue color much lighter than that in the K-tobacco tissues containing an active gusA in all cells. The tissues showing either no visible blue (presumably fully edited) or a partial blue (presumably chimeric) were not likely caused by poor penetration of the GUS staining solution because dark blue was shown in all control shoots, and the vacuum infiltration must have enabled well penetration of the GUS staining solution to cells.  To evaluate the editing efficiencies of gusA, 2-3 individual hygromycin-re shoots from each leaf explant, selected 10 weeks after inoculation, were subjected staining. In two experiments, one had ~8.3% and the other had ~6.0% of the H-tob lines producing shoots or leaf disks without visible blue staining (Figure 1), which resulted from knocking out the gusA gene. This agrees with 7/94 (7.4%) lines havin 98% editing efficiency for gRNA2 in experiment 2 (Table 1). About 50% of the rege in each experiment were likely chimeric with both edited and non-edited gusA, b they showed blue color much lighter than that in the K-tobacco tissues containing tive gusA in all cells. The tissues showing either no visible blue (presumably fully or a partial blue (presumably chimeric) were not likely caused by poor penetration GUS staining solution because dark blue was shown in all control shoots, and the v infiltration must have enabled well penetration of the GUS staining solution to cel  The presence of Cas9 and hpt in the H-tobacco transformants was verified using PCR before these transformants were subjected to PCR amplification of the gusA fragments covering the two targeted sites. Of the H-tobacco lines screened, we did not identify any lines showing only the single PCR band representing the target gusA~42-bp fragment removed. This suggested that none of the transgenic lines were regenerated from a single edited cell with two target sites edited simultaneously. However, we did identify two transgenic lines which showed PCR bands corresponding to both edited and non-edited gusA, indicating that some cells had the target region removed in these two lines. Overall, the efficiency of simultaneously editing the two target sites was low.
To investigate the molecular features of gusA editing in the H-tobacco lines, sequences (~50 K reads per sample, Q > 30) of the PCR amplicons from newly developed leaf tissues of 94 randomly selected transgenic lines were produced and analyzed. Non-edited K-tobacco was used as a control (Table S1). For either gRNA1 or gRNA2, the K-tobacco control had insertion (Ins) and deletion (Del) (hereafter: Indel) frequencies less than 0.1%. We arbitrarily used Indel frequencies of above 3% as a criterion to define detectable edited transgenic lines for the analysis in this study.
For the gRNA1 site, 10.6% of the H-tobacco lines were edited and had both insertions and deletions-they were all chimeric. Of these chimeric lines, 16.1% of cells had deletions and 1.2% had insertions. Of the top Indels (> 1000 reads per mutation) in each edited transgenic line, most transgenic lines had 1-bp or 2-bp deletions and a 1-bp insertion with a thymine ("T"). 40-bp and 42-bp removal were the major large fragment deletions detected, which occurred in 1.1% and 2.1% of the transgenic lines, respectively (Table 2). However, there were a total of only three on-target mutations that occurred in four T 0 H-tobacco lines ( Table 2 and Table S1), including both 40-bp and 42-bp removal detected in three chimeric transgenic lines and one insertion line. The overall percentage of plants containing detectable on-target edited cells (editing frequency) was 4.3%. Notably, the 42-bp removal detected in two chimeric transgenic lines, which had a frequency of 36.5% and 2.6%, respectively, was an on-target deletion where gRNA1 and gRNA2 made cuts simultaneously ( Figure 2A). Those off-target mutations for the gRNA1 were due mainly to the gRNA2. Table 2.
Summary of PCR amplicon sequences from 73 gusA edited H-tobacco lines (Indel frequencies > 3%). The editing positions were identified using the online Cas-Analyzer (http://www.rgenome.net/cas-analyzer/#!) (accessed on 10 December 2022) [12]. For each line, sequences with a total of reads > 1000 were included in the calculation. % of T 0 plants = number of H-tobacco T 0 lines with the particular target sequence variant divided by 94 (total number of the sequenced T 0 lines) × 100. Avg % of reads = Average percentage of reads out of the total reads from all edited lines that had the target sequence variant. Chance (%) of obtaining a non-chimeric edited variant = % of T 0 plants × Avg % of cells × 100. This represents the chance of obtaining a non-chimeric edited line for the specific sequence variant in the second round of regeneration (given no continuous editing during the regeneration). WT: non-edited sequence. Del: Deletion. Ins: Insertion. Underlined letters show PAM sequences of the gRNAs.   lines (Tables 2 and S1), including both 40-bp and 42-bp removal detected in three chimer transgenic lines and one insertion line. The overall percentage of plants containing detec able on-target edited cells (editing frequency) was 4.3%. Notably, the 42-bp removal d tected in two chimeric transgenic lines, which had a frequency of 36.5% and 2.6%, respe tively, was an on-target deletion where gRNA1 and gRNA2 made cuts simultaneous (Figure 2A). Those off-target mutations for the gRNA1 were due mainly to the gRNA2.  For the gRNA2 site, there were a total of 10 H-tobacco lines which had over 90% of cells with editing at the gRNA2 target site (Table 2), while about 18% (17/94) of H-tobacco lines had mutation frequencies (<1%) similar to that of the K-tobacco control. Overall, 77.7% of the H-tobacco lines were edited, and seven (7.4%) transgenic lines showed editing frequencies greater than 98% (Table S1). Notably, all mutations detected for the gRNA2 site were on-target mutations ( Table 2). Of these transgenic lines, 12.2% of cells had deletions, and 12.7% had insertions when sequences with a total of reads > 1000 for each line were included in the calculation ( Table 2). Most of the transgenic lines had 1-bp (60.6%) or 2-bp (34.0%) deletions or a 1-bp (44.7%) insertion with thymine ("T"). A 40-bp deletion was detected in 2.1% (2/94) of the H-tobacco lines (Table 2). Interestingly, a single "T" insertion was the major form of insertion at both gRNA target sites. In fact, there was only one H-tobacco line showing an adenine ("A") insertion, and there were no H-tobacco lines with detectable insertion of cytosine ("C") or guanine ("G"). This does not seem to be a random event, because, in tomato, the most abundant insertion was "A" followed by "T", "C", and "G" [7]. Overall, the gRNA2 produced a higher number of edited transgenic lines than the gRNA1 (77.7% versus 4.3%).
We further checked profiles of the mutations in seven (7.4%) transgenic lines, each showing editing frequencies greater than 98%, and identified the major mutations with mutation frequencies greater than 10%. A total of 2-4 major mutations were found in each line, and none of these lines seemed to be produced from a single edited cell because over 20% of the Indels for each line were composed of multiple minor mutations (mutation frequencies < 10%) (Table 3). Apparently, chimeric editing of the target gene occurred in most, if not all, of the edited T 0 lines, suggesting that the efficiency of obtaining nonchimeric edited T 0 lines was very low. To determine if GUS staining results were correlated with mutation frequencies detected by sequencing, we analyzed the staining and sequencing data from one K-tobacco and 94 H-tobacco lines ( Figure 1B). A correlation analysis was conducted between the mutation frequency and the score of GUS staining for each line. The results showed little correlation (R 2 = 0.0008), likely due to the chimeric nature of these transgenic lines ( Figure S1). In other words, GUS staining in leaf disks was not a reliable criterion to determine whether an H-tobacco transformant was a non-chimeric edited or a chimeric line in the T 0 generation because continuous editing is expected in the transgenic plants due to the Cas9 being driven by a constitutive promoter. Indeed, we observed that some samples had high mutation frequencies in young leaves while their older leaf disks still showed blue staining.
A second round of shoot regeneration experiments was performed by culturing leaf explants from different chimeric editing lines on the regeneration medium. When the induced young shoots were stained in GUS solution, it was obvious that more H-tobacco regenerants from the parent transgenic line with weak staining showed no blue staining than those from the parent line with intense staining (Figures 1A and S1). The staining assay suggests that it is possible to increase the chance of obtaining lines from single Cas9-edited cells by conducting a second round of regeneration from chimeric-edited lines. This was further confirmed by the profiles of the mutations for the gRNA2 site in 30 transformants produced from six selected T 0 transgenic lines from experiment #1 (Table 4). For the plants produced from three light blue lines containing presumably both edited and non-edited cells, ten out of 15 plants (66.7%) were non-chimeric edited lines. Of the 15 plants from three white T 0 lines showing no GUS staining, they were all non-chimeric, and one plant with a 42-bp deletion was non-chimeric for both gRNA1 and gRNA2 sites ( Table 4). The results demonstrate that a second round of regeneration from chimeric-edited lines can increase the potential for the generation of non-chimeric-edited plants.

Discussion
The phytoene desaturase (PDS) gene is often used in plant species as a candidate gene to determine gene editing efficiencies [13][14][15][16][17][18] because disruption of this gene causes albino leaves by impairing chlorophyll, carotenoid, and gibberellin biosynthesis [19,20]. However, spontaneous mutations can also cause albino tissues in plant regeneration, which may complicate PDS as a system to determine gene editing frequencies in some plants [21]. The gusA gene and the green fluorescent protein (GFP) gene are major screenable markers for plant genetic engineering. Both have been used in gene editing studies. For example, gusA was recently used to monitor the expression of the CRISPR-Cas9 [22] and was also used as an editing target for testing different gene editing platforms in a transgenic blueberry line [8]. In this study, we used gusA as an editing target for evaluating the efficiencies of obtaining edited lines in the T 0 transgenic tobacco. As demonstrated, the GUS staining worked effectively in showing various types of editing outcomes in transgenic tobacco.
There are different ways to describe gene editing efficiencies. In editing studies using protoplasts, the mutation frequency is usually used to show the percentage of the edited and non-edited gene target(s). In editing studies using plant regenerants, the percentage of either plants/shoots or the molecules containing on-target editing is often used to indicate gene editing efficiency. In an editing study in transgenic tomato, the average mutation rate across 63 target genes in T 0 was used as an estimate of editing efficiency [7]. The editing efficiencies for the two gRNA editing sites varied much in this study, with 10.6% for gRNA1 and 77.7% for gRNA2, although most of the edited plants were chimeric. The cause for the difference was unknown, but whether the AtU6 and AtU3 promoters had different strengths of promoter activities is an interesting question to examine. It was likely that the gRNA1 driven by the AtU6 promoter had a lower expression level than that of the gRNA2 driven directly by the AtU3 and indirectly by the upstream AtU6 promoter. As expected, the editing efficiency for targeting two editing sites simultaneously was detected in two transgenic lines. Even with a high editing efficiency of 77.7% for gRNA2, T 0 transgenic plants regenerated initially from a single edited cell were not found in the 94 sequenced lines.
There are many factors, e.g., plant species, Cas9 sources, gRNAs, target cells, and editing approaches (i.e., error-prone nonhomologous end joining and homology-directed repair), that can affect gene editing frequencies [17,[23][24][25]. This was supported by our sequencing data of 94 CRISPR-Cas9 expressing T 0 transgenic tobacco lines, of which seven lines had over 98% of their cells edited at the gRNA2 site. Apparently, it is easier to regenerate plants from individual edited cells of these seven lines. To enhance the chance of obtaining non-chimeric edited plants, we demonstrated that a second round of regeneration would be very helpful. This information is important for those who are working on gene editing for clonally propagated plant species.

Plant Materials and Growth Conditions
The gusA vector, named pBISN1, was transformed into tobacco Nicotiana tabacum cv. Samsun for generating gusA transgenic tobacco lines. The vector contains a neomycin phosphotransferase II (nptII) gene as a selectable marker for kanamycin selection. The transformation was done following a published protocol (Duan et al. 2016). A T 0 transgenic tobacco containing an active gusA was propagated on Murashige and Skoog (MS) medium containing 50 mg/L kanamycin [26] and used for gusA editing in this study (Murashige and Skoog 196). This T 0 transgenic line showed one hybridization band in Southern Blot analysis and a ratio of transgenic to nontransgenic close to 3:1 in its first-generation seedlings. We considered this T 0 transgenic line a line with a single copy of the transgenes. The cultures were maintained in our lab at 25 • C under a 16 h photoperiod of 30 µE m −2 s −1 from cool white fluorescent tubes.
Leaf explants,~0.8 cm × 0.8 cm, from in vitro cultured kanamycin-resistant, gusAcontaining tobacco plants of a single-copy K-tobacco transgenic line were used for tobacco transformation by following a published protocol [29]. About 200 leaf disks were transformed and subject to selection in a medium containing 20 mg/L hygromycin in each of the two transformations conducted separately. The hygromycin-resistant shoots from separate explants were labeled as independent transgenic lines. They were grown on an MS medium containing 50 mg/L hygromycin.
To induce new shoot regeneration from selected T 0 plants of the H-tobacco lines, leaf explants were cultured on a regeneration medium containing 50 mg/L hygromycin. Regenerated shoots were grown on an MS medium containing 50 mg/L hygromycin.

Histochemical GUS Staining Assay
Leaf disks/pieces or shoots were evaluated by histochemical GUS staining assay [30]. They were stained in 2 mM 5-bromo-4-chloro-3-indoyl-β-d-glucuronide (X-Gluc) (Phy-toTech Labs, Overland Park, KS, USA) in 100 mM phosphate buffer for 24 h at 37 • C after a two-minute vacuum at 80,000 Pa, and chlorophyll was removed with 70% ethanol washes. The blue staining for each leaf disk was graded on a scale of 0 (no blue for wild-type control) to 4 (all blue for non-edited control). Three leaf disks/pieces for each plant of 100 randomly selected T 0 lines were stained and scored; 95 of these lines were sampled for PCR amplificon sequencing.

PCR Amplificon Sequencing and Identification of Edited Cells
Amplicon amplification was run as follows: the first round of PCR reaction was conducted to produce a GUS amplicon covering both Cas9 target sites with the GUSspecific primers-forward ATGTTACGTCCTGTAGAAA and reverse GCTCCATCACTTC-CTGATTAT. The second PCR reaction was conducted to add adaptor sequences using the GUS-specific primers (upper case) with Illumina adaptor sequences (lower case): forward primer acactgacgacatggttctacaTCGTCCGTCCTGTAGAAA and reverse primer tacggtagca-gagacttggtctGCTCCATCACTTCCTGATTAT. The PCR products were barcoded and sequenced using the Illumina platform in the RTSF Genomics Core of Michigan State University (https://rtsf.natsci.msu.edu/genomics/sample-requirements/illumina-sequencingsample-requirements/) (accessed on 10 November 2022).

Conclusions
All T 0 transgenic plants were regenerated before a gene editing event took place. Our sequence data lay a foundation that a second round of regeneration from T 0 chimeric lines can increase the chance for the production of putative editing lines.

Data Availability Statement:
The sequence data generated in this study are available from the corresponding author on reasonable request.