A method for the production and expedient screening of CRISPR/Cas9-mediated non-transgenic mutant plants

Developing CRISPR/Cas9-mediated non-transgenic mutants in asexually propagated perennial crop plants is challenging but highly desirable. Here, we report a highly useful method using an Agrobacterium-mediated transient CRISPR/Cas9 gene expression system to create non-transgenic mutant plants without the need for sexual segregation. We have also developed a rapid, cost-effective, and high-throughput mutant screening protocol based on Illumina sequencing followed by high-resolution melting (HRM) analysis. Using tetraploid tobacco as a model species and the phytoene desaturase (PDS) gene as a target, we successfully created and expediently identified mutant plants, which were verified as tetra-allelic mutants. We produced pds mutant shoots at a rate of 47.5% from tobacco leaf explants, without the use of antibiotic selection. Among these pds plants, 17.2% were confirmed to be non-transgenic, for an overall non-transgenic mutation rate of 8.2%. Our method is reliable and effective in creating non-transgenic mutant plants without the need to segregate out transgenes through sexual reproduction. This method should be applicable to many economically important, heterozygous, perennial crop species that are more difficult to regenerate. A group led by Dr. Yi Li of University of Connecticut, USA has developed a technique to reliably create desirable mutations in crop plants without introducing any foreign DNA, thus generating non-transgenic mutant plants. Their technique will be particularly important for application of genome editing technologies in perennial crop plants. Genome editing technologies have been used to introduce desired mutations into plants but the approach normally incorporates foreign genes such as Cas9 into the plant’s genome. To avoid this, Dr. Li’s team has used Agrobacterium to transiently express the CRISPR/Cas9 components, a bacterium commonly used to genetically engineer plants. With their method, CRISPR/Cas9 from the bacteria edit the plant genome without introducing any foreign DNA, and then a two-step screening process is used to identify non-transgenic mutant plants. This new method provides a reliable and efficient alternative for producing non-transgenic genetically engineered crops.


Introduction
Transgenic technologies provide powerful tools for crop improvement. However, the application of these technologies has been hampered by public apprehension toward potential food safety and gene flow concerns, resulting from the presence and/or expression of transgenes 1,2 . Recent development of CRISPR/Cas9-mediated genome editing has made targeted mutagenesis an attractive alternative to traditional transgenic technologies 3,4 . In plants, the most widespread application of CRISPR/Cas9 entails the stable integration of Cas9 endonuclease and single-guide RNA (sgRNA) genes into host genomes [5][6][7][8][9][10][11][12][13][14][15][16] . When CRISPR/Cas9-mediated gene editing is used in sexually propagated annual crop plants, transgenes (Cas9, sgRNA, and so on) can be eliminated from host genomes following sexual reproduction and screening of segregating populations. This segregation of CRISPR/Cas9 transgenes from mutations of interest can result in non-transgenic mutant plant progeny 17-20 . However, this strategy is rarely feasible or practical for vegetatively propagated perennial plants. These plants generally require years to reach sexual maturity; thus, multiple years are needed before sexual reproduction is feasible 21 . Additionally, these plants are highly heterozygous for genes controlling many important traits, and these traits will segregate and recombine following sexual reproduction, resulting in non-transgenic mutant progeny likely lacking a combination of desirable traits 22 .
Developing a method to generate CRISPR/Cas9-mediated non-transgenic mutants is highly desirable for many applications of genome editing, particularly for asexually propagated, heterozygous, perennial crop plants. It has been reported that pre-assembled CRISPR/Cas9 ribonucleoproteins can be delivered into protoplasts to induce mutations, without the need for stable integration of CRISPR/Cas9 genes into the host-plant genome [23][24][25] . Particle bombardment has also been used to deliver CRISPR/Cas9 ribonucleoproteins to wheat and maize cells, producing non-transgenic mutants [26][27][28][29] . However, working with protoplasts, as well as utilizing biolistics, limits the potential for full-plant regeneration to some species and tissue types. Therefore, it is important to also develop alternative methods to produce non-transgenic CRISPR mutants of perennial crop plant species.
In contrast to the limited success of plant regeneration from protoplasts 30,31 , plant regeneration from leaf, hypocotyl, epicotyl, shoot, root, cotyledon, or callus explants has been well established for the majority of crop plant species 32 , including many that are recalcitrant to regeneration from protoplasts 33,34 . It has also been shown that proteins can be produced following transient expression of Agrobacterium T-DNA genes 35,36 . In addition, Agrobacterium inoculation protocols have been developed for many perennial crop species 37 . To circumvent the limited regeneration potential when using CRISPR ribonucleoproteins to produce non-transgenic mutants, we report a method for using Agrobacterium to transiently express the Cas9 and sgRNA genes in plant cells, using tobacco as a model plant and PDS as a model target gene. We have also developed a high-throughput screening protocol utilizing next-generation sequencing in combination with high-resolution DNA melting (HRM) analysis to efficiently identify mutants from a population of shoots regenerated in the absence of selection pressure. We demonstrate that the combination of Agrobacteriummediated transient CRISPR/Cas9 expression with a highly

Results
Production of mutants via Agrobacterium-mediated expression without antibiotic selection Using tobacco (Nicotiana tabacum Xanthi) as a model plant and an intron-containing GUS gene as a marker (Fig. 1a), we observed that transient expression of T-DNA genes in inoculated leaf discs peaked 3-4 days following Agrobacterium infection in the absence of kanamycin selection. Figure 1d shows the GUS activity in tobacco leaf discs 2-6 days post infection (dpi), and Fig. 1e shows the GUS activity in leaf discs after 5 additional days in timentin-containing media. The antibiotic timentin was used to suppress Agrobacterium growth following an initial 2-6 day co-incubation; thus, the GUS activities shown in Fig. 1e should result from stable integration of the GUS gene into the tobacco genome. The difference in GUS expression between explants in Fig. 1d and those in  The total number of independent pds mutant shoots as indicated by an albino phenotype 3 The mutation efficiency was calculated as follows: ð no:of pds mutant shoots no:of explants Þ 100 in Fig. 1d, e indicated that a 3-day or 4-day co-incubation was optimal for Agrobacterium-mediated transient expression of the GUS gene. Three days of Agrobacterium co-incubation was subsequently used for transient expression of CRISPR/Cas9 genes. The sgRNA used in these experiments (Fig. 1b) targets the beginning of the fourth exon of the endogenous tobacco phytoene desaturase gene (PDS), as shown in Figure 1c. Previous studies have shown that disruption of this gene leads to an albino phenotype 38 . We used this phenotype as a visual marker to identify tobacco mutants whose PDS gene had been edited following the expression of Cas9 and PDS-targeting sgRNA genes. We infected 415 tobacco leaf-disc explants in three independent experiments using 3 days of Agrobacterium co-incubation without any selection for transgenic cells or shoots (Table 1). A total of 197 shoots regenerated from infected explants exhibited the albino phenotype, indicating a mutation in PDS, demonstrating a mutation rate of 0.475 pds mutants per explant. However, due to the lack of chemical selection, the total number of shoots regenerated from each explant was very high, and therefore, the mutation rate per total regenerated shoots was quite low (2.57%, Supplementary Table 1). These results indicate that pds mutant plants can be produced via Agrobacterium-mediated expression of Cas9 and sgRNA genes without using antibiotic selection. Ten independent pds mutant plants were randomly chosen for further analysis, and six are shown in Table 2. The specific genetic mutations in these plant lines were identified via highthroughput sequencing, which demonstrated that all plants contained tetra-allelic mutations, meaning that all four alleles were mutagenized and no wild-type alleles could be detected. Microscopic analysis of plant tissues was unable to uncover the presence of green cells in albino pds mutants, suggesting that all PDS genes in all cells were mutated in these plants.
The albinism resulting from the disrupted PDS gene enabled us to conveniently identify pds mutant shoots at early stages of shoot development in this study. However, the vast majority of desirable mutations for crop improvement are unlikely to display any visually identifiable phenotypes at the early stages of shoot development. When no selection pressure is applied during callus and shoot regeneration following Agrobacterium infection, the vast majority of regenerated shoots or plantlets should be non-mutant, as demonstrated above. Therefore, the ability to efficiently identify mutants lacking any visually identifiable phenotype from a population of regenerated shoots is essential for using the abovedescribed Agrobacterium-mediated transient mutagenesis system. Toward this end, we tested the effectiveness and efficiency of a two-step screening method using the newly produced pds mutants. The first step, an initial identification of mutants, takes advantage of the highthroughput nature of Illumina sequencing, and the second step, a fine identification of mutants, makes use of the high resolution of HRM analysis.
Initial screening of CRISPR-mediated mutants using highthroughput DNA sequencing analysis Although our mutagenesis rate was relatively high per explant, without a visible phenotype, it would be difficult to identify mutant plants due to the high number of regenerated shoots in the absence of chemical selection. Additionally, other CRISPR mutagenesis projects could have an even lower mutation rate than the one reported here. Therefore, we have developed a two-step method for high-throughput screening of shoots to identify the presence of targeted mutations. We first mixed leaf tissue from an albino pds mutant (MT), pds-12, with leaf tissues from independently derived non-mutant shoots (WT), regenerated from Agrobacterium-infected explants, at MT-to-WT ratios of 1:20, 1:41, and 1:83. We isolated genomic DNA from these pooled tissue samples and performed PCR reactions to amplify a 186-bp fragment that contained the sgRNA-target region on the fourth PDS exon (Fig. 1c). PCR products were sequenced on an Illumina platform to~×60,000 to ×100,000 coverage. Next, we measured the amount of PCR product derived from the MT-to-WT ratios of 1:20, 1:41, and 1:83 mixed tissues and diluted it with ×6 the amount of PCR product (ng) derived from WT plant tissue. The diluted PCR product was also used for Illumina sequencing analysis.
We observed that the PCR products derived from a 42plant (1MT: 41WT) pooled tissue sample containing the pds-12 mutant showed a drastically elevated nucleotide variant frequency (NVF) at positions 45-51 (red triangles, Fig. 2, showing a 1MT: 41WT pooled tissue sample), which is consistent with verified mutations at these nucleotide positions (Table 2). NVF is a measure of the frequency of abnormal (compared to WT reference) nucleotides detected by DNA sequencing at a given position due to mutations or sequencing error. When we diluted the same PCR products with 6× WT PCR products (blue squares, Fig. 2), we observed significant reductions in NVF at positions 45-51 relative to the undiluted PCR product. The observed elevations and reductions of NVF before and after a 6× WT DNA dilution further verified the presence of mutations at nucleotide positions 45-51, as NVF resulting from sequencing error would be unaffected by dilution. As shown in Table 2 (Table 2). Similar results were observed using the 1MT: 20WT and 1MT: 83WT pooled tissue samples, with a more drastic elevation of NVF for the 1MT: 20WT samples and reduced elevated NVF for the 1MT: 83WT samples compared to the 1MT: 41WT samples (data not shown).
Using a single-blind approach (the researchers who conducted the experiment did not know which 42-plant groups contained pds mutants), we tested the accuracy of the mutant screening method based on elevations and reductions of NVF before and after a 6× WT DNA dilution as described above. We created eight 42-plant pools, five of which contained plant tissue from a single pds mutant plant line and three of which contained 100% WT plants. The five pds mutants used were pds-9, pds-10, pds-11, pds-13, and pds-14. We confirmed that the screening method was reliable for identifying all 42-plant pools that contained pds mutant plants at a ratio of 1MT: 41WT (Supplementary Figure 1), with 100% accuracy (Table 3). Thus, the elevations and reductions of NVF before and after a 6× WT DNA dilution were excellent indicators of 42-plant pools that contained mutant plants.

Fine identification of mutants using DNA high-resolution melting analysis
After identification of mutant-containing 42-plant pools, HRM analysis was used to identify individual mutant plant lines within each of the 42-plant pools (Supplementary Figure 2). To determine the sensitivity of HRM analysis, we performed HRM analysis on PCR products amplified from various DNA templates combined at different ratios. These template mixes were created using one-part pds-12 plant tissue combined with different parts independently regenerated non-mutant (WT) tissues in the following ratios: 1:1, 1:6, 1:19, and 1:29. Figure 3a shows that mutant-containing PCR products at ratios of 1:1, 1:6, and 1:19 (MT:WT) could be distinguished from a wild-type DNA reference. The plant pool size we chose for subsequent HRM analysis was 7; thus, each 42-plant pool containing DNA from mutants could be divided into six pools of seven plants each.
We also used a single-blind experiment approach to test the accuracy of HRM analysis to identify mutant plants. We created eight additional 7-plant pools, five containing a single pds mutant each (pds-9, pds-10, pds-11, pds-13, or pds- 14), and the remaining three negative control pools containing only wild-type plants. The HRM analysis results are shown in Fig. 3b and demonstrate that all pooled samples containing pds mutant plants could be identified with 100% accuracy. Finally, upon the identification of mutant-containing 7-plant pools, individual mutant(s) within each pool were identified via HRM based on a 1:1 mix between each putative mutant and a WT plant. Through this method, we successfully identified all pds mutant plants ( Table 2).

Determination of non-transgenic pds mutants
To distinguish transgenic from non-transgenic mutants, we performed PCR on 29 randomly selected pds mutant lines using primers targeted to three regions in the T-DNA fragment (Fig. 1b). Mutant plants were considered to be non-transgenic if they lacked a PCR product for all three primer sets (Fig. 4). Approximately 17.2% of the tested pds mutant lines were determined to be nontransgenic following PCR analysis ( Table 4). As shown in Supplementary Figure 3, a non-transgenic plant (pds-7), along with a transgenic plant (pds-9), was cultured on MS media containing 100 mg/L kanamycin. The nontransgenic pds-7 plant died under kanamycin selection, while the transgenic pds-9 plant grew normally.

Discussion
Producing non-transgenic mutants of heterozygous perennial crop plants using CRISPR/Cas9 technology is highly desirable but challenging. We developed an effective method for producing and identifying CRISPR/Cas9mediated non-transgenic mutant plants, which should be  (Table 2) applicable to many perennial heterozygous crop plants.
We have demonstrated that we can use Agrobacterium to transiently express CRISPR/Cas9 genes, and such expression can lead to the production of tetra-allelic, nontransgenic mutant plants. We have also demonstrated that the first step of our mutant identification, based on elevated and reduced nucleotide variance frequencies before and after a WT DNA dilution, using highthroughput DNA sequencing analysis, is reliable and highly efficient. Furthermore, the second step of mutant identification, using HRM analysis, is simple and effective.
With one sgRNA targeting the tobacco PDS gene, we achieved a 47.5% mutation rate (i.e., 0.475 pds mutants per explant) using no selective pressure during callus or shoot regeneration. At least 17.2% of the pds mutant plants produced this way were non-transgenic, for an overall non-transgenic mutation rate of 8.2% (i.e., 0.082 pds mutants per explant). We expect that the rate of recovery for non-transgenic mutant plants following Agrobacterium-mediated transient expression of CRISPR/ Cas9 genes could be much higher than we reported here. One reason is that the albino phenotype caused by pds mutations used in this study can result in cell-or shootgrowth disadvantages, which may have contributed to lower rates of mutant shoot production. Additionally, multiple sgRNA sequences may be used to target the same gene 39 to increase the efficiency of mutant production.
Protoplast-mediated delivery of CRISPR/Cas9 ribonucleoproteins offers advantages for creating non-transgenic mutant plants [23][24][25][26][27][28]40 . However, regenerating plants from protoplasts can be difficult and has not been demonstrated to be possible for many important crop species. Low efficiency of plant regeneration from protoplasts has been reported in economically important crops such as avocado (Persea americana) 41 , grape (Vitis vinifera) 42 , and apple (Malus domestica) 43 . Furthermore, regeneration protocols have not been successfully demonstrated in many other plant species 31,44 . In the case of citrus, for instance, a globally important fruit crop that is often heterozygous for important trait genes, protoplast regeneration can be achieved only if protoplasts are derived from juvenile tissues 45 . However, plants generated from juvenile citrus tissues require years to reach a mature fruit production stage, and this delay is an impediment to the efficient evaluation of gene functions in fruit and for other genes pertinent to citrus breeding programs 46,47 . On the other hand, tissues from mature citrus trees, such as shoot segments, can be transformed with Agrobacterium 48 .
Jacobs et al. 49 discussed the possibility of using Agrobacterium-mediated transient expression of Cas9 and sgRNA genes to produce non-transgenic mutant plants. Iaffaldano et al. 50 reported the production of CRISPR/ Cas9-mediated mutant plants using Agrobacterium without any selection of stable transgenic plants. However, Iaffaldano et al. 50 did not characterize whether their mutant plants were transgenic or non-transgenic, nor did they calculate the percentage of each type. The nontransgenic mutation rate reported in this study is lower than mutation rates for the previously reported protoplast microinjection methods 25 but higher than for particle bombardment methods (per explant) [26][27][28] . However, Agrobacterium can be a more versatile tool because of its   Table 2) ability to efficiently infect many different tissues types across most plant species 51,52 , and it is the most widely used method for plant transformation. Thus, protocols for Agrobacterium infection and subsequent callus and shoot regeneration are readily available for many crop plant species. Furthermore, even when other methods are applicable, Agrobacterium infection procedures are relatively simple and easy to perform for many plant species. Thus, the method presented here provides an alternative method for the production of non-transgenic mutant plants.
Our method of using Agrobacterium to transiently express CRISPR/Cas9 genes, without the use of chemical selection such as kanamycin, also offers some other advantages. The lack of a chemical selection agent following Agrobacterium infection allows for even greater rates of plant regeneration compared to when chemical selection is used and therefore likely enhances mutant callus/shoot production. This is because selection agents, such as kanamycin, hygromycin, and various herbicides, can suppress shoot regeneration for many plant species, even when those shoots express relevant resistance genes 53 . Furthermore, our high-throughput screening method, based on a novel DNA sequencing strategy in combination with HRM analysis, makes mutant identification fast and easy to perform. The drawback of our We have observed that the threshold of mutant plant detection using the Illumina sequencing method can be much more sensitive than the detection of one mutant out of 42 plants (the 1MT: 41WT) described in this study. The use of the 42-plant sample pools, therefore, assures that the presence of a mutant plant can be readily detected in pooled samples. Additionally, the threshold of HRM analysis to detect mutant plants can be as low as 1 of 19 plants, but we chose one out seven (1MT: 6WT). Again, this pooling choice provides a safe margin for error by reducing the chance of missing mutant plants due to false negatives. Furthermore, a small number of false positives from the sequencing analysis of 42-plant pools are not a concern because they can be easily identified in the subsequent step of screening. Thus, our method, which combines a unique DNA sequencing strategy with HRM analysis, provides a reliable protocol to screen for CRISPR/Cas9-mediated mutant shoots from a population of shoots regenerated in the absence of selection pressure. The overall protocol for mutant identification is schematically represented in Supplementary Figure 4. Using this protocol, it would take~14 days to identify all mutants from 1000 independent shoots that were subjected to Agrobacterium-mediated transient expression of CRISPR/Cas9 genes.
In conclusion, we have developed a highly useful method utilizing Agrobacterium-mediated transient expression of Cas9 and sgRNA genes combined with multi-step pooled screening, enabling the reliable production and efficient identification of non-transgenic mutants regenerated in the absence of selection pressure. Due to the versatility of CRISPR/Cas9 and Agrobacterium-mediated infection, as well as the ease of plant regeneration from leaf, shoot, or root explants, this method is applicable to many economically important plant species, particularly heterozygous perennial plant species that are recalcitrant to regeneration from protoplasts or following biolistic bombardment.  Table 4 Rate of non-transgenic pds mutants among all pds mutants intron-containing GUS gene under the control of a CaMV 35S promoter and a NOS terminator. We modified this construct by adding a kanamycin resistance gene (NPTII) under the control of a CaMV 35S promoter and CaMV 35S terminator using the KpnI and XbaI restriction enzymes (Fig. 1a). For CRISPR/Cas9-based disruption of the tobacco PDS gene, we made a number of modifications to the pK7WGF2::hCas9 plasmid (Addgene plasmid #46965). First, we added an IV2 intron to a Cas9 gene to prevent Cas9 expression in Agrobacterium, this gene was used to replace the original Cas9 DNA in the vector using the SpeI and ApaI restriction enzyme sites. Second, we added an sgRNA targeting the fourth exon of the PDS gene in Nicotiana benthamiana 54 under the control of the Arabidopsis thaliana U6-26 gene promoter (accession # At3G13855), using the XbaI and KpnI restriction enzymes. All of the added polynucleotides were synthesized by Genscript Corporation (Piscataway, NJ, USA).

Agrobacterium-mediated transformation for transient expression
A. tumefaciens (EHA105) was used to deliver Tiplasmid DNA into tobacco leaf-disc explants. Agrobacterium cells were cultured overnight (16 h, 200 r.p.m., 28°C) in 5 mL of liquid LB medium containing 100 mg/L spectinomycin. The overnight culture was diluted with 50 mL of fresh liquid LB media (1:10 dilution) and then grown for 6-8 h (200 r.p.m., 28°C). Agrobacterium cells were collected at an OD 600 of 0.6-0.8, centrifuged at 5000 r.p.m. for 15 min, and finally re-suspended in 50 mL of liquid MS medium containing 100 µM acetosyringone (AS). Re-suspended bacterial cells were shaken (180 r.p. m., 28°C) for 1 h before use. Tobacco leaf discs (0.5 cm 2 , from vegetatively propagated clonal plants) were incubated in Agrobacterium cell solution for 20 min. Explants were blotted dry on sterile filter paper and transferred onto solid MS media containing 100 µM AS and no antibiotics.

Histochemical GUS activity assay
To optimize the time needed for robust transient T-DNA expression, tobacco leaf explants inoculated with Agrobacterium (containing the pCAMBIA1305.1 vector) were incubated in the dark at 25°C for 2, 3, 4, 5, or 6 days. Half of the leaf discs (six) from each treatment were stained for GUS activity 55 . The remaining explants were transferred to solid MS medium containing 150 mg/L timentin to suppress Agrobacterium growth. These explants were stored in the dark for an additional 5 days. Leaf discs from each treatment were histochemically stained to estimate stable versus transient expression levels of the GUS gene.

Callus and shoot regeneration
For transient expression of the CRISPR/Cas9 genes, tobacco leaf discs were incubated in the dark at 25°C for 3 days before being transferred to solid media containing MS salts, 2.0 mg/L benzylaminopurine (to promote shoot initiation), and 150 mg/L timentin (to repress Agrobacterium growth). No antibiotics for the selection of transgenic plant cells were included in the culture media. The plates were stored at 25°C under a 16-h photoperiod. After 4 weeks, visual screening was performed to identify white-colored mutant shoots. The white shoots were then transferred to solid MS media containing 150 mg/L timentin for rooting. Genomic DNA was isolated using the NucleoSpin plant II kit (Macherey-Nagel, Dueren, Germany). The primer pair PDS-F (5′-CTGAAGCAGT-CACCAAGA-3′) and PDS-R (5′-AGTACGCATTCTT-GAGGAGTC-3′) was used to amplify the sgRNA-target region. A second round of PCR was performed, with the primer pair PDS-FA (5′-ACACTCTTTCCCTA-CACGACGCTCTTCCGATCTCTGAAGCAGTCACC-AAGA-3′) and PDS-RA (5′-GTGACTGGAGTTCA-GACGTGTGCTCTTCCGATCTAGTACGCATTCTT-GAGGAGTC-3′), using the PCR product of the first round as the template to add general adapter sequences to the PCR products. A final round of PCR was performed to add a barcode tag to the PCR products, with the primer pair PDS-FAI (5′-AATGATACGGCGACCACCGA-GATCTACAC-index-ACACTCTTTCCCTACACGA-3′) and PDS-RAI (5′-CAAGCAGAAGACGGCA-TACGAGAT-index-GTGACTGGAGTTCAGACGTG-3′). Sequencing reactions were performed using the barcoded PCR products and the Illumina MiSeq platform (paired-end ×2 150 bp read length). The raw Illumina reads were mapped to the 186-bp reference sequence in the tobacco genome using bwa mem (-c 300000 -v 2) of BWA v0.5.9 56 to determine the nature of mutations.

High-throughput sequencing
For the 21-, 42-, and 84-plant pools used in the highthroughput sequencing screen, leaf tissue from pds-12 mutant plants was combined with an equal amount of tissue from each of 20, 41, or 83 independently regenerated non-mutant shoots, respectively. One control pool was created using the tissue from 84 independently regenerated non-mutant shoots. Genomic DNA was isolated, amplified, tagged, and barcoded as described previously. For dilution, the PCR product amplified from the 42-plant pool was quantified using a NanoDrop and was then diluted with ×6 the amount of PCR product (ng) amplified from a pool composed of 84 independently regenerated non-mutant shoots. Sequencing reactions were performed in triplicate using the barcoded PCR products and the Illumina MiSeq platform (paired-end ×2 150 bp read length). The raw Illumina reads were mapped to the 186-bp reference sequence in the tobacco genome using bwa mem (-c 300000 -v 2) of BWA v0.5.9 56 . NVFs were calculated over a 77-bp region. To confirm the high-throughput screening method, five additional mutant-containing 42-plant pools were created, each comprising tissue from a single mutant plant (pds-9, pds-10, pds-11, pds-13, or pds-14) in combination with tissue from 41 independently regenerated nonmutant shoots. Three WT shoot control pools were created using the tissue from 42 independently regenerated non-mutant shoots. All eight pools were given randomly generated labels to assure that high-throughput sequencing screening could be performed in a single-blind manner.

HRM analysis
For 2-, 7-, 20-, and 30-plant pools used in the highresolution melting analysis screen, leaf tissue from pds-12 mutant plants was combined with an equal amount of tissue from each of 1, 6, 19, or 29 independently regenerated non-mutant shoots, respectively. One control pool was created using the tissue from 30 independently regenerated non-mutant shoots. Genomic DNA was isolated using the NucleoSpin plant II kit (Macherey-Nagel). We used primers PDS-F and PDS-R (described previously) to amplify the sgRNA-target region. We also tested 85bp (F primer: 5′-ATCTGTTCTGCACCTGAATAC-3′, R primer: 5′-AACCCAGTCTCATACCAA-3′) and 146 bp (F primer: 5′-CTGAAGCAGTCACCAAGA-3′, R primer: 5′-AACCCAGTCTCATACCAA-3′) PCR products for HRM analysis. The 186-bp fragments were used for our experiments, as shown in our results. PCR amplifications were performed in 10 µL volumes containing 5 µL of Precision Melt Supermix (Bio-Rad, USA), 0.5 µL of each 2 µM primer, and 50 ng of genomic DNA. PCR was carried out in triplicate using 96-well white-walled PCR plates and a CFX96™ Real-Time PCR Detection System (Bio-Rad). The amplification started with an initial denaturation step at 95°C for 5 min, followed by 50 cycles of 95°C for 10 s, 60°C for 30 s, and 72°C extension for 1 min. High-resolution melting analyses of the PCR amplicons were carried out in the same plate immediately following PCR amplification, using the following cycle: 95°C for 30 s and 60°C for 1 min. Melting curves were generated over a 65-95°C range with 0.2°C increment. Melting curves were analyzed using Precision Melt Analysis software (Bio-Rad). To confirm our HRM screening method, five mutantcontaining 7-plant pools were created, each comprised tissue from a single mutant plant (pds-9, pds-10, pds-11, pds-13, or pds-14) in combination with tissue from six independently regenerated non-mutant shoots. Three control pools were created using the tissue from seven independently regenerated non-mutant shoots. The eight pools were labeled randomly to assure that HRM screening was performed in a single-blind manner.

Determination of non-transgenic mutant plants
Stable integration of the CRISPR/Cas9 T-DNA was determined by PCR analysis of the tobacco genomic DNA. Primers 1F (5′-AGGTGGCGAAGTCATCTGC-3′) and 1R (5′-TGTCGTTTCCCGCCTTCAG-3′) were used to amplify a 701-bp region of the T-DNA fragment. Primers 2F (5′-GCCTGTTTGGTAATCTTATCGC-3′) and 2R (5′-TCTTTCCACTCTGCTTGTCTCG-3′) were used to amplify a 1326-bp fragment of the Cas9 gene. Primers 3F (5′-ACTGGGCACAACAGACAATC-3′) and 3R (5′-ACCGTAAAGCACGAGGAA-3′) were used to amplify a 668-bp fragment of the kanamycin resistance gene. The absence of PCR products from all three primer sets was deemed to be indicative of a lack of stable transgene integration into the tobacco genome, thus indicating a non-transgenic mutant plant. Shoots from pds mutant plants were cultured on MS media containing 100 mg/L kanamycin and allowed to grow for 35 days, producing further evidence for the lack of transgene insertion into the genome.