Efficient targeted mutagenesis in allotetraploid sweet basil by CRISPR/Cas9

Abstract Sweet basil (Ocimum basilicum) is an economically important herb and its global production is threatened by basil downy mildew caused by the obligate biotrophic oomycete Peronospora belbahrii. Effective tools are required for functional understanding of its genes involved in synthesis of valuable secondary metabolites in essential oil and disease resistance, and breeding for varieties with improved traits. Clustered regularly interspaced short palindromic repeat (CRISPR)/Cas9 gene editing technology has revolutionized crop breeding and functional genomics. The applicability and efficacy of this genomic tool in the allotetraploid sweet basil were tested by editing a potential susceptibility (S) gene ObDMR1, the basil homolog of Arabidopsis DMR1 (Downy Mildew Resistant 1) whose mutations conferred nearly complete resistance against Arabidopsis downy mildew pathogen, Hyaloperonospora arabidopsidis. Two single guide RNAs targeting two different sites of the ObDMR1 coding sequence were designed. A total of 56 transgenic lines were obtained via Agrobacterium‐mediated stable transformation. Mutational analysis of 54 T0 transgenic lines identified 92.6% lines carrying mutations at target 1 site, while a very low mutation frequency was detected at target 2 site. Deep sequencing of six T0 lines revealed various mutations at target 1 site, with a complete knockout of all alleles in one line. ObDMR1 homozygous mutant plants with some being transgene free were identified from T1 segregating populations. T2 homozygous mutant plants with 1‐bp frameshift mutations exhibited a dwarf phenotype at young seedling stage. In summary, this study established a highly efficient CRISPR/Cas9‐mediated gene editing system for targeted mutagenesis in sweet basil. This system has the capacity to generate complete knockout mutants in this allotetraploid species at the first generation of transgenic plants and transgene‐free homozygous mutants in the second generation. The establishment of this system is expected to accelerate basil functional genomics and breeding.


| INTRODUC TI ON
Basil belonging to the genus Ocimum in Lamiaceae family is a popular herb widely used in culinary, therapeutic, and cosmetic industries (da Costa et al., 2015;Wyenandt et al., 2015). There are nearly 35-150 species of Ocimum distributed across the globe, among which Ocimum basilicum, commonly known as sweet basil, is highly demanded with significant economic value (Wyenandt et al., 2015).
Essential oil extracted from sweet basil contains diverse secondary metabolites with a wide spectrum of biological activities, such as high antioxidant, antimicrobial, anticancer, and insecticidal activities, and thus used for culinary flavoring, aromatherapy, preservation of various foods, and eco-friendly insect control (Li & Chang, 2016). Different ploidies have been reported across varied species of Ocimum, with O. basilicum considered tetraploid (2n = 4× = 48) (Pyne, Honig, Vaiciunas, Wyenandt, & Simon, 2018).
Global production of sweet basil is severely threatened by an obligate biotrophic oomycete pathogen Peronospora belbahrii, which causes basil downy mildew (BDM) (Cohen, Ben Naim, Falach, & Rubin, 2017;Wyenandt et al., 2015). Current control strategies for BDM mainly rely on the use of limited available fungicides, whose repeated use poses the risk of evolving fungicide-resistant strains (Cohen et al., 2017;Wyenandt et al., 2015). In addition, frequent applications of these chemicals increase production cost, and adversely affect human health and the environment. The most potent and cost-effective control strategy to halt the rapid dissemination of BDM is to utilize disease-resistant cultivars. Resistance and tolerance have been found in Ocimum species differing vastly from sweet basil (Cohen et al., 2017). Traditional breeding involving interspecific hybridization to transfer naturally existing disease resistance genes to sweet basil has been a hassle as it is largely met with sexual incompatibility, hybrid F1 sterility, and difficulty in segregating out undesirable traits (Ben-Naim, Falach, & Cohen, 2018;Cohen et al., 2017). Genetic engineering, with the advent of cutting-edge clustered regularly interspaced short palindromic repeat (CRISPR)/ CRISPR-associated protein (Cas) gene editing technology, offers a promising platform to functionally understand the molecular basis of basil-P. belbahrii interactions, broaden resistance resources, and accelerate the breeding process.
Adapted from bacterial immune systems, CRISPR/Cas has revolutionized the way scientists perform functional genomics studies and crop breeding as this technology provides an unparalleled tool to precisely edit DNA sequences with ease, high efficiency, and high fidelity (Zhou, Wang, & Liu, 2018). The first used CRISPR system for genome editing was CRISPR/Cas9 (Jinek et al., 2012(Jinek et al., , 2013, which has been successfully used in genome editing of many plant species (Haque et al., 2018;Jaganathan, Ramasamy, Sellamuthu, Jayabalan, & Venkataraman, 2018). This gene editing system requires Cas9 and a single guide RNA (sgRNA), which is a fusion of CRISPR RNA (crRNA) containing a 20-nt DNA target sequence upstream of a Cas9 protospacer adjacent motif (PAM, 5′-NGG-3′) and trans-activating CRISPR RNA (tracrRNA) (Jinek et al., 2012). This technology relies on specific base pairing of the 20-bp sequence of the sgRNA with the target DNA, which directs Cas9 endonuclease to cleave the target DNA at 3-nt upstream of the PAM motif (Jiang & Doudna, 2017). The double-stranded breaks (DSBs) generated by Cas9 activate innate DNA repair by either non-homologous end-joining (NHEJ) or homology-directed repair (HDR) mechanism (Jiang & Doudna, 2017).
Without a homologous DNA template, the cell repairs the DSB through NHEJ, which is error-prone causing short insertions or deletions (indels) around the cleavage site. With a homologous DNA template, the cell will repair the DSB through HDR, leading to precise mutations. As this approach can generate homozygous or complete knockout mutants as early as in the first generation of transgenic lines for both diploid and polyploid species (Gumtow, Wu, Uchida, & Tian, 2018;Pan et al., 2016;Wang et al., 2016), it greatly speeds up functional genomics studies and shortens the breeding process. Furthermore, the use of CRISPR/Cas9-mediated gene editing allows for the development of foreign DNA-free crops, which is more acceptable by consumers, as opposed to the conventional way of developing genetically modified (GM) crops. Traditional genetic engineering generates GM crops that embody the transgene within the genome to express new traits. In contrast, the trait generated through CRISPR-mediated gene editing can be segregated from the introduced transgenes; or the desired trait can be achieved via DNA-free approach for delivery of gene-editing reagents. The resultant transgene-free plants can bypass the regulatory restrictions set for GM crops by USDA (Waltz, 2016). Many proof-of-concept studies have used CRISPR for crop nutritional improvement and enhanced resistance to biotic and abiotic stresses (Borrelli, Brambilla, Rogowsky, Marocco, & Lanubile, 2018;Jaganathan et al., 2018;Langner, Kamoun, & Belhaj, 2018). The modification or removal of S genes prevents pathogen infection, and therefore represents an effective strategy to achieve resistance (Dong & Ronald, 2019;Zaidi, Mukhtar, & Mansoor, 2018).
Targeted mutagenesis of S genes using CRISPR/Cas technology has recently emerged as a desirable approach to generating broadspectrum disease resistance (Borrelli et al., 2018;Zaidi et al., 2018).
The success of this approach relies on the targeting of a suitable S gene. Due to the lack of an effective functional genomics approach for sweet basil, a S gene that contributes to susceptibility to BDM has not been identified. However, a number of S genes have been identified in Arabidopsis that are required for susceptibility to its downy mildew-causing oomycete pathogen Hyaloperonospora arabidopsidis, one of which is DMR1 (Downy Mildew Resistant 1) (van Damme, Huibers, Elberse, & Van den Ackerveken, 2008;van Damme et al., 2009;Hok et al., 2011;Van Damme et al., 2005). DMR1 encodes a homoserine kinase (HSK) catalyzing phosphorylation of homoserine. dmr1 mutants contain high levels of homoserine and are highly resistant to H. arabidopsidis (Van Damme et al., 2005, 2009. DMR1 seems to be conserved in various plant species and its homologs in multiple plant species have been shown to be a determining factor for susceptibility to various pathogens (Huibers et al., 2013;Sun et al., 2016). Therefore, the DMR1 homolog in sweet basil, ObDMR1, represents a candidate S gene with a potential role in susceptibility to BDM.
In the present study, we targeted ObDMR1 to establish an effective CRISPR/Cas9-mediated gene editing system in sweet basil.
Sweet basil cultivar Genoveser was transformed, respectively, with two sgRNA/Cas9 binary constructs targeting one or two sites of ObDMR1 via Agrobacterium-mediated stable transformation. High frequency of small insertion/deletion (indel) mutations and complete knockout of ObDMR1 were achieved in the first generation (T0) of transgenic plants. Homozygous ObDMR1 mutants with some being transgene free were identified in the second generation (T1).
T2 homozygous mutants exhibited a dwarf phenotype at the young seedling stage, confirming the efficient mutagenesis of ObDMR1 by CRISPR/Cas9.

| Plant materials and growth conditions
Sweet basil cultivar Genoveser (Enza Zaden) plants were used for Agrobacterium-mediated transformation, and routinely grown in a controlled growth chamber set at 25°C with a photoperiod of 12 hr.
The same conditions were applied to grow newly regenerated transgenic T0, T1, and T2 seedlings. Older T0 and T1 plants were grown in the greenhouse at 25-27°C with a photoperiod of 16 hr for seed production. Selfing bags were mounted on flower stalks at the beginning of flowering to avoid cross-pollination.

| Identification of the homolog of Arabidopsis DMR1 in sweet basil
The homolog of Arabidopsis DMR1 (AtDMR1) in sweet basil (Ocimum basilicum), ObDMR1, was identified by TBLASTX search against the non-redundant transcriptomic sequence dataset generated from two sweet basil varieties Red Rubin and Tigullio using Trinity assembly (Torre et al., 2016) using the protein encoding sequence of AtDMR1 (GenBank accession: NM_127281) as a query. A single significant hit was identified and the transcript sequence was retrieved. To amplify ObDMR1 from sweet basil cv. Genoveser by PCR, we designed a pair of primers (DMR1-Gen-F: 5′-CGTCCCCTATTCTCTCACTATGGC-3′; DMR1-Gen-R: 5′-AAAACCCAGAGACCATGCAAATG-3′) targeting 5′-UTR and 3′-UTR of this transcript. PCR was performed using Genoveser genomic DNA (gDNA) as the template and Phusion High-Fidelity DNA Polymerase (NEB) with PCR conditions as: initial denaturation at 94°C for 3 min; followed by 35 cycles at 94°C for 15 s, 54°C for 30 s and 72°C for 75 s; with a final extension at 72°C for 7 min. The resultant PCR product was gel purified using QIAquick Gel Extraction Kit (QIAGEN) and subjected to Sanger sequencing. The amino acid sequences of ObDMR1 in Genoveser and AtDMR1 were aligned using BLASTP to assess their homology. The amino acid sequence alignment was also generated using CLUSTALX 2.1 (Larkin et al., 2007) and displayed using BOXSHADE (https://embnet.vital -it.ch/softw are/BOX_form.html).
The non-redundant transcriptomic assembly of two sweet basil varieties Red Rubin and Tigullio generated using Trinity (Torre et al., 2016) was uploaded as a custom genome to identify potential off-targets. The candidate target sequences with total score as well as efficiency score of more than 0.50 were selected and further subjected to secondary structure analyses using the web tool RNAstructure (Reuter & Mathews, 2010) (http://rna.urmc.roche ster. edu/RNAst ructu reWeb /Serve rs/Predi ct1/Predi ct1.html). The ones with no more than three hydrogen bonds were chosen for sgRNAs.

C A T T T C C A A C A T C A C G T T T T A G A G C T A G A A
The PCR product was purified using QIAquick PCR Purification Kit (QIAGEN) and then digested by BsaI. The digested PCR product was ligated into BsaI-digested pKSE401 vector. The resulting plasmid was named as pKSE401-sgRNA1+2. Both pKSE401-sgRNA1 and pKSE401-sgRNA1+2 were introduced into Agrobacterium tumefaciens strain EHA105 for basil transformation.

| Agrobacterium-mediated transformation of sweet basil
Sweet basil cultivar Genoveser was transformed with A. tumefaciens EHA105 harboring pKSE401-sgRNA1 or pKSE401-sgRNA1+2 based on the method described previously (Deschamps & Simon, 2002;Phippen & Simon, 2000) with modifications. The Agrobacteria, The plants were grown under 100% relative humidity for 3-4 days in a tray covered with a plastic dome in a growth chamber set at 25°C with a 12-hr photoperiod. Humidity was gradually reduced over next 2-3 days and then the plants were transferred to the greenhouse to produce seeds.

| Accession numbers
The ObDMR1 gDNA sequence was deposited in NCBI GenBank under accession number MT000722.

| Identification of the homolog of Arabidopsis DMR1 in sweet basil
To determine whether a homolog of Arabidopsis DMR1 (AtDMR1, At2g17265) is present in sweet basil, we did a local TBLASTX search using the 1113-bp protein coding sequence of AtDMR1 as a query against the non-redundant transcriptomic sequence dataset generated from two sweet basil varieties Red Rubin and Tigullio using Trinity assembly (Torre et al., 2016). We identified one significant hit with an E value of 9e-147. All other hits were insignificant with E values higher than 0.11. The single significant hit comp43301_c0_

| Selection of sgRNA target sequences and generation of constructs for Agrobacterium-mediated transformation
To test the efficiency of CRISPR/Cas9-mediated gene editing in sweet basil, ObDMR1 was targeted for mutagenesis. 20-nt candidate sgRNA target sequences were identified using the online tool EuPaGDT (http://grna.ctegd.uga.edu/) (Peng & Tarleton, 2015) and their RNA secondary structures were analyzed using RNAstructure had total score and efficiency score higher than 0.5, and no offtargets were found in the transcriptome of Red Rubin and Tigullio (Table S1). The GC content is 40% and 55% for target 1 and target 2, respectively. The predicted RNA structures of both targets were relatively less complex with three hydrogen bonds (Table S1).
To perform sweet basil gene editing using Agrobacteriummediated transformation, we utilized pKSE401, which is a plant binary vector developed for gene editing in dicots (Xing et al., 2014), to generate two constructs pKSE401-sgRNA1 and pKSE401-sgRNA1+2 that express one sgRNA and two sgRNAs, respectively. pKSE401-

| Generation of transgenic sweet basil plants expressing gene-editing reagents
To generate transgenic basil plants expressing sgRNA(s) and Cas9, leaf discs prepared from the first pair of true leaves from 3-weekold Genoveser plants were used as explants. The explants were infected and co-cultivated with A. tumefaciens EHA105 harboring either pKSE401-sgRNA1 or pKSE401-sgRNA1+2 on callus and shoot induction (SI) media supplemented with acetosyringone for 3 days (Figure 3a).
Then the explants were transferred to SI media containing kanamycin to selectively induce the formation of transgenic calli and shoots.
Calli were seen 2 weeks after culturing the explants on kanamycincontaining SI media and tiny shoot buds emerged from calli later after

| Targeted mutagenesis of ObDMR1 in T0 transgenic plants
To
To identify transgene-free homozygous mutants, we determined the presence of transgene in the seven T1 homozygous mutants by amplifying a fragment of sgRNA1 expression cassette delivered by the plasmid pKSE401-sgRNA1 (Figure 2b). As a template integrity control, the 383-bp ObDMR1 fragment was successfully amplified from all plants (Figure 6c). No amplification of the sgRNA1 expression cassette was observed in T1 line #9-2 and #9-9, whereas its amplification in other lines was successful (Figure 6c), demonstrating that the transgene was segregated out from #9-2 and #9-9. These results demonstrated the heritability of ObDMR1 mutations to next generation independent of the presence of transgene, which leads to transgene-free homozygous mutants.

| Dwarf phenotype of homozygous ObDMR1 mutants at seedling stage
In order to observe the morphological phenotypic differences under normal growth conditions, T2 homozygous ObDMR1 mutants along with the WT plants were sown and grown in the same controlled growth chambers. Homozygous mutant seeds germinated normally and no clear morphological difference between the mutants and WT was observed within 2 weeks after sowing. However, varying levels of stunted growth started to appear in seedlings of ObDMR1 mutants after that. At 18 days old, all mutant lines displayed apparent dwarfism compared with WT, with 9-2, 9-9, 11-5, 11-7, and 11-8 more drastic than 16-4 ( Figure 7). The notable stunting in plant height was observed at young seedling stage. The difference in height gradually reduced along with age. Mutant plants of 45 days old or older showed similar growth and expansion of leaves as WT.  WT, T1 homozygous lines #9-2, #9-9, #11-5, #11-7, and #11-8 that carry a "T" insertion displayed greater dwarfing phenotypic aberration than #16-4 that carries a "G" insertion ( Figure 7). It is unclear what caused the difference. The "T" and "G" insertion are located at the same location of the gene and the resultant frameshifts lead to the change to the same amino acid sequence except for the codon containing the nucleotide insertion ( Figure 6b). Insertion of "T" did not change the amino acid isoleucine (I), while insertion of "G" led to the change from I to methionine (M) (Figure 6b). This single amino acid difference may somehow relieve the dwarfing effect. Another possibility is that the integration of the transgene in #16-4 disrupted a gene that plays a role in regulating plant height. The second possibility can be tested using a transgene-free line with "G" homozygous insertion, which is currently being identified from a T1 population of The asterisks (*) indicate statistically significant differences (p < .001) between the mutant line and WT, determined by one-tailed t test. #9-2, #9-9, #11-5, #11-7 and #11-8 contain 1-bp "T" insertion and #16-4 carries 1-bp "G" insertion in ObDMR1. Similar stunted growth was consistently observed when T2 seeds of these mutant lines were sown and grown side by side with WT Although mutations were effectively generated at sgRNA1 target site, very low to negligible frequency of mutations occurred at target site 2. It is unclear what caused the significant difference.

| D ISCUSS I ON
The target sequence could be one contributing factor. Previous studies showed that CRISPR/Cas9-induced mutation efficiency in plants is affected by the GC content and secondary structure of sgRNAs (Ma et al., 2015;Tang et al., 2018). In our study, we used two 20-nt sgRNA target sequences with varying GC content of 40% and 55%, respectively, and different secondary structures (Table S1).
Another factor could be due to the effectiveness of the promoter that drives the expression of sgRNAs. A report demonstrated that AtU6-26 promoter displayed a much higher transcriptional activity in model plant Arabidopsis than AtU6-29 promoter (Li, Jiang, Yong, & Zhang, 2007). In our study AtU6-26 and AtU6-29 were used to drive the expression of sgRNA1 and sgRNA2, respectively. The low expression of sgRNA2 may have led to the low efficiency of mutations at target 2. Determining the expression levels of both sgRNA1 and sgRNA2 using reverse transcription quantitative PCR (RT-qPCR) will clarify this and determine whether AtU6-29 promoter is suitable to be used for multiplexing gene editing in sweet basil.
At this point, since AtU6-26 promoter was shown to be highly effective in sweet basil gene editing, multiplexing gene editing can be achieved by expressing an array of sgRNAs interspaced with tRNA as a single polycistronic gene under AtU6-26 promoter, which is then processed to multiple sgRNAs through the endogenous tRNA processing system (Xie, Minkenberg, & Yang, 2015). Alternatively, multiplex gene editing may be achieved using CRISPR/Cas12 system as the dual nuclease activity of Cas12 allows the processing of a single transcript containing multiple guide RNAs and simultaneous editing of their targeted loci (Wang, Mao, Lu, Tao, & Zhu, 2017).
The whole genome sequence assembly for sweet basil has been generated, but not yet publicly available (Dudai et al., 2018). As such, we were not able to analyze the off-targets in detail. We tried to limit the off-targets based on the transcriptomic sequences by selecting the target sequences without the off-target in the transcriptome through sgRNA design tool EuPaGDT (Peng & Tarleton, 2015).
However, detailed analysis of off-targets need to be done in the future when the whole genome sequence of sweet basil is available.

ACK N OWLED G M ENTS
We sincerely thank Dr. Qi-Jun Chen from China Agricultural University for providing the plasmid pKSE401 (Addgene Plasmid

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest to this work.

AUTH O R CO NTR I B UTI O N S
M. Tian conceived and designed the overall study. N. Navet designed and performed the experiments, and analyzed the data.
N. Navet wrote the initial draft of the manuscript. M. Tian reviewed and edited the manuscript. Both authors approved the final manuscript.