It has been reported that pre-assembled CRISPR/Cas9 ribonucleo proteins can be delivered into protoplasts to induce mutations, without the need for stable integration of CRISPR/Cas9 genes into the host-plant genome. Particle bombardment has also been used to deliver CRISPR/Cas9 ribonucleo proteins to wheat and maize cells, producing non-transgenic mutants. 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, plant regeneration from leaf, hypocotyl, epicotyl, shoot, root, cotyledon, or callus explants has been well established for the majority of crop plant species, including many that are recalcitrant to regeneration from protoplasts. It has also been shown that proteins can be produced following transient expression of Agrobacterium T-DNA genes. In addition, Agrobacterium inoculation protocols have been developed for many perennial crop species. 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,drainage pot 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 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 efficient screening protocol makes it possible to efficiently obtain non-transgenic mutant plants, a method that should be applicable to heterozygous perennial crop species.Using tobacco as a model plant and an intron-containing GUS gene as a marker , 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 , 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 Fig. 1e is indicative of transient GUS expression, demonstrating that there are high levels of transient expression of the genes in the T-DNA region. The results 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 targets the beginning of the fourth exon of the endogenous tobacco phytoene desaturase gene , as shown in Figure 1c.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 .

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 . 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 high throughput 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 high throughput nature of Illumina sequencing, and the second step, a fine identification of mutants, makes use of the high resolution of HRM analysis.Although our mutagenesis rate was relatively high per explant, without a visible phenotype, it would be diffificult 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 , pds-12, with leaf tissues from independently derived non-mutant shoots , 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 . 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 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 42- plant pooled tissue sample containing the pds-12 mutant showed a drastically elevated nucleotide variant frequency at positions 45–51 , which is consistent with verified mutations at these nucleotide positions . NVF is a measure of the frequency of abnormal 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 , 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, high-throughput sequencing uncovered four types of mutations in the pds-12 mutant line: a 1-bp deletion at position 51, a 5-bp deletion at positions 45–49, a 2-bp deletion at positions 49–50, and a 4-bp deletion at positions 46–49 . 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 . Using a single-blind approach ,growing raspberries containers 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 , with 100% accuracy . 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.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 .

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 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 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 , 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 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 .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 TDNA fragment . Mutant plants were considered to be non-transgenic if they lacked a PCR product for all three primer sets . Approximately 17.2% of the tested pds mutant lines were determined to be non-transgenic following PCR analysis . As shown in Supplementary Figure 3, a non-transgenic plant , along with a transgenic plant , 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.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/Cas9- mediated non-transgenic mutant plants, which should be 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 high throughput 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 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% . 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 shoot growth disadvantages, which may have contributed to lower rates of mutant shoot production. Additionally, multiple sgRNA sequences may be used to target the same gene to increase the efficiency of mutant production. Protoplast-mediated delivery of CRISPR/Cas9 ribonucleoproteins offers advantages for creating non-transgenic mutant plants. 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, grape, and apple. Furthermore, regeneration protocols have not been successfully demonstrated in many other plant species.