“Omics” technologies allow us to identify a list of genes responsible for specific traits in plants (Alonso and Ecker, 2006). To characterize gene function accurately, loss- or gain-of-function mutants of interest are required. The CRISPR (clustered regularly interspaced short palindromic repeats) system is extremely useful for this reverse genetic approach, because it can be used to generate indel (insertion and deletion) mutations or nucleotide substitutions at the target loci in any plant species with high accuracy and efficiency (Zhu et al., 2020). The most widely used CRISPR system is derived from Streptococcus pyogenes; this system consists of CRISPR-associated protein 9 (SpCas9) and a single-stranded guide RNA (gRNA) (Jinek et al., 2012). For plant genome editing, SpCas9 and gRNA are expressed in cells, and a whole plant is regenerated from the edited callus. These processes, which are called transformation and regeneration, often take from several months to over a year (Ahn et al., 2013; Altpeter et al., 2016). In addition, because tissue culture media and growth conditions need to be optimized depending on genotype, even for the same species, establishing a standard tissue culture method for one species is difficult (Altpeter et al., 2016).

To generate gene-edited plants without tissue culture, plant viral vectors have been engineered; these vectors express SpCas9 proteins and/or gRNA in plants (Ali et al., 2015; Ariga et al., 2020; Ellison et al., 2020; Lei et al., 2021; Li et al., 2021; Wang et al., 2017; Zhang et al., 2020). This system is called the virus-induced genome editing (VIGE) system (Oh et al., 2021a). The delivery capacity of viral vectors is negatively correlated with the size of the cargo involved. Although sonchus yellow net rhabdovirus can deliver a large cargo, SpCas9 (~4.2 kbp) with gRNA (~100 bp), into the systemic tissues (Ma et al., 2020), the movement of commonly used viruses, such as tobacco rattle virus (TRV) (Ali et al., 2018), tobacco mosaic virus (Cody et al., 2017), and geminivirus (Yin et al., 2015), is hampered by the size of SpCas9.

Rather than focus on the delivery of both SpCas9 and gRNA, scientists have focused on the delivery of gRNA into SpCas9-expressing meristematic cells. This approach results in the production of gene-edited seeds (Ellison et al., 2020; Lei et al., 2021; Li et al., 2021). Guide RNA fused with the mobile Flowering Locus T (FT) mRNA sequence enabled the acquisition of heritable mutants with a mutation frequency of 65%-100% from the TRV-infected Nicotiana benthamiana (Ellison et al., 2020), which expresses a SpCas9 gene under the control of the 35SPPDK promoter (the 35S promoter fused to the maize C4PPDK basal promoter; Baltes et al., 2015; Li et al., 2013; Yoo et al., 2007). FT-fused gRNA induced the heritable mutation with a frequency of 4%-9% from the cotton leaf crumple virus-infected Arabidopsis (Lei et al., 2021), which expressed SpCas9 under the control of YAO and 35S promoter. A barley stripe mosaic virus-derived gRNA also generated heritable mutations (13%-100%) in the transgenic wheat, which expresses SpCas9 under the control of maize ubiquitin promoter (Li et al., 2021). However, whether the promoter used for SpCas9 expression is critical to induce heritable gene editing with virus-delivered gRNA remains unknown.

In this study, we developed a TRV-based VIGE system for the ecological model plant Nicotiana attenuata. Previous Arabidopsis studies showed that the expression of SpCas9 in shoot apical meristem or germ cells is critical for heritable gene editing (Mao et al., 2016; Yan et al., 2015). Ribosomal protein S5 A (RPS5A) promoter strongly confers expression of transgenes in meristematic cells and germlines (Tsutsui and Higashiyama, 2017). Therefore, we hypothesized that the efficacy of the VIGE system would be enhanced when SpCas9 proteins were expressed under the control of the RPS5A promoter in N. attenuata.

Plant materials

The wild-type N. attenuata Utah seeds were a kind gift from the Department of Molecular Ecology at the Max Plank Institute for Chemical Ecology. Seeds were germinated following protocol as previously described (Krügel et al., 2002). All M₁ and F₁ seeds were pooled from three infected plants and randomly germinated. The plants were grown in a plant growth chamber under a 16/8 h light/dark photoperiod at 25°C with ± 2°C. To generate SpCas9-expressing transgenic plants, the pHAtC (35S:Cas9; accession No. KU213971; Kim et al., 2016) and pRPS5A:Cas9 binary vectors were transformed into the Agrobacterium tumefaciens strain LBA4404 by the thaw-freeze method. The hypocotyl tissues were used for transformation and tissue culture as described previously (Kang et al., 2020).

Vector construction and guide RNA design

The N. attenuata RPS5A promoter was amplified from the genomic DNA of N. attenuata with the RPS5A-F and RPS5A-R primers. The coding sequence of SpCas9 was synthesized by Macrogen (Korea) and amplified with Cas9-F and Cas9-R primers. Heat shock protein terminator was amplified from the pKI1.1R vector (Tsutsui and Higashiyama, 2017) with HSP-F and HSP-R primers. These amplified polymerase chain reaction products were then recombined into pHAtC (Kim et al., 2016) to acquire pRPS5A:Cas9 vector (Supplementary Table S1) using an In-Fusion HD Cloning Kit (Takara, Japan). The U6 sequences were removed from pHAtC and pRPS5A:Cas9 vectors with SacI (NEB, USA).

The TRV1 and TRV2 vectors were a gift from Sir David Baulcombe (University Cambridge). The subgenomic promoter of pea early-browning virus was used to express N. attenuata phytoene desaturase (NaPDS)-targeted gRNA (TRV2:gPDS), N. attenuata chalcone synthase 2 (NaCHAL2)-targeted gRNA (TRV2:gCHAL2), and N. attenuata chalcone synthase 3 (NaCHAL3)-targeted gRNA (TRV2:gCHAL3) (Supplementary Table S2). The subgenomic promoter of TRV was used to express NaPDS-targeted gRNA (T:gPDS and TH:gPDS) and TH:gPDS fused with Arabidopsis FT mRNA sequence (TH:gPDS-FT) (Supplementary Table S2). The coding sequence of FT, which was described in Ellison et al. (2020), was synthesized by Macrogen. The target sequence of NaPDS (A4A49_02763), NaCHAL2 (A4A49_34074), and NaCHAL3 (A4A49_39367) was designed using Cas-Designer program implemented in CRISPR RGEN Tools (http://www.rgenome.net/) (Bae et al., 2014; Park et al., 2015). Primers used in this study were synthesized by Macrogen. Genome sequence information for N. attenuata is available in the N. attenuata Data Hub (http://nadh.ice.mpg.de/NaDH/). All primer sequences are summarized in Supplementary Table S3.

Agroinfiltration

N. attenuata infiltration was performed as previously described (Senthil-Kumar and Mysore, 2014) with minor modifications. Briefly, the TRV1 and TRV2 derivatives were introduced into Agrobacterium tumefaciens strain AGL1. The AGL1 containing TRV2 derivatives, was mixed with AGL1 carrying TRV1 in a ratio of 1:1 before inoculation; the mixed solution was co-infiltrated into the leaves of 4- to 5-week-old SpCas9-expressing N. attenuata. Three plants were used for each agroinfiltration. After agroinfiltration, N. attenuata plants were maintained under a 48 h dark at 22°C and then transferred into the growth chamber (a 16/8 h light/dark photoperiod at 25°C).

Targeted deep sequencing

Targeted deep sequencing was performed as previously described (Oh et al., 2020). Briefly, the genomic DNA was extracted from leaves, petals, anthers, and ovaries using HiGene genomic DNA prep kit (BIOFACT, Korea). All primer sequences are summarized in Supplementary Table S3. High-throughput sequencing was performed in the KAIST BIOCORE center (http://biocore.kaist.ac.kr/) with Illumina Miniseq (Illumina, USA). The indel frequency and patterns were analyzed using the Cas-Analyzer tool, which is available at CRISPR RGEN Tools (Park et al., 2017). The indel frequency (%) was calculated by dividing the number of sequencing reads containing mutations by the number of total sequencing reads.

No gene-edited plants were obtained from infected 35S:Cas9 plants

To develop a VIGE system for N. attenuata, we first generated transgenic N. attenuata expressing SpCas9 under the control of the 35S promoter (35S:Cas9) (Oh et al., 2021b). A gRNA was designed to target the N. attenuata phytoene desaturase (NaPDS) gene and cloned into the TRV2 vector under the subgenomic promoter of the pea early-browning virus (TRV2:gPDS) (Fig. 1). We infected 35S:Cas9 leaves with TRV2:gPDS and then collected M₁ seeds from the infected plants. If germ cells were mutated by the SpCas9-gRNA complex and viruses were not detected in the next generation, we would expect three distinct genotypes in the seeds (M₁ generation): wild-type allele, monoallelic mutation (one wild-type allele and one mutated allele), and biallelic mutation. To screen gene-edited seeds, we pooled equal amounts of leaf tissue from M₁ seedlings and performed targeted deep sequencing across the target site (Fig. 1). If one monoallelic mutant was found in pooled 20 M₁ seedlings, the mutation frequency would be expected to be near 2.5% (1 monoallelic mutation/40 total alleles = 2.5%). We then examined the mutation frequencies in 174 M₁ progenies from TRV2:gPDS-infected 35S:Cas9 (Supplementary Fig. S1). Although the mutation was observed in the infected leaves (19.7% ± 2.7%, Oh et al., 2021b), we found no gene-edited seeds in the pooled M₁ seedlings of TRV2:gPDS-infected 35S:Cas9 (Supplementary Fig. S1, Oh et al., 2021b).

We next manipulated gRNA expression or movement to induce heritable gene editing in seeds. Firstly, we tested whether a NaPDS-targeted gRNA expressed by the TRV subgenomic promoter (T:gPDS) performed better than TRV2:gPDS. Among 223 M₁ progenies from T:gPDS-infected 35S:Cas9, we could not obtain gene-edited plants in the pooled M₁ seedlings (Supplementary Fig. S2). Next, we inserted a hammerhead ribozyme sequence between the subgenomic promoter of TRV2 and NaPDS-targeted gRNA (TH:gPDS), because we found that TH:gPDS induced more indel mutations in the infected leaves than did T:gPDS (Oh et al., 2021b). A total of 123 M₁ progenies from TH:gPDS-infected 35S:Cas9 contained no mutations in the target site (Supplementary Fig. S3). To increase the movement of gRNA into germ cells, a NaPDS-targeted gRNA was fused with Arabidopsis FT mRNA and cloned into the TRV2 vector (TH:gPDS-FT) (Ellison et al., 2020). TH:gPDS-FT generated a similar mutation frequency compared with TH:gPDS in the infected leaves (Supplementary Fig. S4A). No mutated seeds were found from TH:gPDS-FT-infected plants (Supplementary Fig. S4B). These results suggest that the 35S promoter-driven SpCas9 was not able to produce gene-edited seeds with TRV-delivered gRNA in N. attenuata.

Heritable virus-induced gene editing in pRPS5A:Cas9 plants

We hypothesized that the germline expression of SpCas9 would be critical to increase the editing efficiency of plant viruses in germ cells. For Arabidopsis germline mutation, the RPS5A-driven SpCas9 performed much better than the 35S-driven SpCas9 (Tsutsui and Higashiyama, 2017). In addition, the adenine base editor (ABE) is known to produce the edited T₁ plants of Arabidopsis when the RPS5A promoter is used to express ABE rather than the 35S and the YAO promoter (Kang et al., 2018). Although the YAO promoter was also developed for editing germ cells in Arabidopsis (Yan et al., 2015), the YAO promoter did not perform better than the 35S promoter for virus-mediated germline editing in Arabidopsis (Lei et al., 2021). Therefore, we generated the transgenic N. attenuata plants expressing SpCas9 driven by the N. attenuata RPS5A promoter (pRPS5A:Cas9).

To explore the effects of the RPS5A promoter-driving SpCas9, we infected pRPS5A:Cas9 leaves with TRV2:gPDS. The mutation frequency was 6.4% ± 0.2% in the pRPS5A:Cas9 leaves infected with TRV2:gPDS (Fig. 2A). The major mutation patterns were T and A insertions in the infected leaves (Fig. 2B). Next, several hundred seeds were collected to investigate the heritable editing ability of TRV2:gPDS in pRPS5A:Cas9. A total of 206 M₁ progenies were divided into 11 groups for pooling genomic DNA (Fig. 2C). The targeted deep sequencing analysis showed that in all three groups, the editing frequency was more than 2%: group #3, 2.2% (n = 17); group #9, 5.9% (n = 15); and group #10, 4.6% (n = 16) (Fig. 2C). We then examined the editing frequencies in individual plants of three groups: one monoallelic mutant in group #3, three monoallelic mutants in group #9, and one monoallelic mutant in group #10 were identified (Fig. 2D). All five mutants had monoallelic mutations and no biallelic mutation (Fig. 2E). The mutation patterns of five M₁ mutants were single nucleotide insertions; M₁-3-1, M₁-9-7, M₁-9-8, M₁-9-9, and M₁-10-11 had a G, T, A, A, and C insertion, respectively, at the cleavage site (Fig. 2E). The total mutation frequency of TRV2:gPDS was putatively about 2.4% (5 mutants/206 total progenies = 2.4%) in pRPS5A:Cas9 (Fig. 2D). Next, we examined the heritability of mutations found in an M₁-3-1 plant to the next generation. The 145 M₂ seeds from a M₁-3-1 plant were germinated, and the obvious albino phenotype of seedlings was observed to occur at a ratio of 3.5:1 (113 green: 32 white seedlings) (Fig. 2F), which was consistent with the segregation ratio of the monoallelic mutation.

In addition, we targeted the chalcone synthase 2 (NaCHAL2) and chalcone synthase 3 (NaCHAL3) gene of N. attenuata to demonstrate the feasibility of our VIGE system. A gRNA targeting NaCHAL2 and NaCHAL3 was cloned into TRV2:gCHAL2 and TRV2:gCHAL3, respectively. We infected pRPS5A:Cas9 leaves with TRV2:gCHAL2 and TRV2:gCHAL3. The mutation frequency of TRV2:gCHAL2 and TRV2:gCHAL3 was 1.6% ± 0.6% and 49.6% ± 21.9% in the infected leaf, respectively (Figs. 3A and 3D). The major mutation patterns were single nucleotide insertions and deletions at each cleavage site in the infected leaves (Figs. 3B and 3E). We examined the mutation frequencies in individual M₁ plants collected from the infected pRPS5A:Cas9 and identified one monoallelic M₁ mutant of NaCHAL2 (G deletion) (Fig. 3C, Supplementary Fig. S5A), and one monoallelic M₁ mutant of NaCHAL3 (ten nucleotide deletion) (Fig. 3F, Supplementary Fig. S5B).

Male and female germ cells were edited in pRPS5A:Cas9 plants

We further examined the editing frequencies in floral organs to determine which organs were responsible for heritable mutations in pRPS5A:Cas9 plants. The editing frequencies in floral tissues of pRPS5A:Cas9 plants infected with TRV2:gPDS were 4.7% ± 0.4% in petals (Fig. 4A), 5.2% ± 1.2% in anthers (Fig. 4B), and 3.3% ± 0.3% in ovaries (Fig. 4C). The major mutation patterns were single nucleotide insertions at the cleavage site in floral tissues (Figs. 4D-4F) as shown in infected leaves (Fig. 2B). These results suggest that a few microsporocytes or megasporocytes are edited by TRV2:gPDS. To clarify which germ cells are edited, we pollinated wild-type N. attenuata flowers with the pollen from TRV2:gPDS-infected pRPS5A:Cas9 (Cross A, Fig. 5A), and we also pollinated TRV2:gPDS-infected pRPS5A:Cas9 flowers with the pollen from wild-type plants (Cross B, Fig. 5B). We removed all anthers before hand pollination. A total of 109 F₁ progenies from Cross A were divided into 5 groups (Fig. 5A). We found that group #1 of F₁ seedlings contained mutations with a frequency of 1.2% (n = 24) (Fig. 5A). The editing frequencies in individual plants of group #1 were examined; one mutant harboring an A insertion was obtained (Fig. 5C). Next, 35 F₁ progenies from Cross B were divided into 2 groups (Fig. 5B). The mutation frequencies of group #1 and group #2 were 3.4% (n = 17) and 2.7% (n = 18), respectively (Fig. 5B). The editing frequencies in individual plants of group #1 were examined: two monoallelic mutants were identified (Fig. 5D). Both mutants had a T insertion at the cleavage site (Fig. 5D). These results suggest that both female and male germ cells were edited by TRV-delivered gRNA in the pRPS5A:Cas9 plant.

Mei et al. (2019) successfully edited the target genes by delivering gRNA with foxtail mosaic virus in SpCas9-expressing maize and Setaria viridis under the control of a maize ubiquitin promoter, and N. benthamiana expressing SpCas9 with a double 35S promoter. Uranga et al. (2021) also delivered gRNA with potato virus X into N. benthamiana expressing SpCas9 under the control of the 35S promoter. However, they failed to obtain gene-edited seeds from virus-infected plants. Recently, Ellison et al. (2020) found that gRNA fused to FT sequence dramatically increased the efficacy of germline editing in N. benthamiana. Despite our attempts to develop a VIGE system for wild-type species in the genus Nicotiana, we failed to obtain gene-edited progenies from 35S:Cas9 plants infected by gRNA fused with FT sequence (Supplementary Fig. S4). This result suggests that the FT mRNA movement might have occurred only in N. benthamiana, not in N. attenuata. Another possible explanation is that the level of SpCas9 proteins driven by the 35S promoter was insufficient to induce the mutation in N. attenuata, because the 35S promoter activity in germ cells is species-specific. For instance, GUS activity derived by 35S promoter was not detected in Arabidopsis pollen (Wilkinson et al., 1997) but was detected in Nicotiana tabacum pollen (Patro et al., 2012; Wilkinson et al., 1997; Zheng et al 2007). In cotton, the 35S promoter was not active during the early stage of embryogenesis (Sunilkumar et al., 2002). Here we showed that the promoter for expressing SpCas9 is crucial for heritable genome editing. To develop a VIGE system for a wide range of plants, species- and tissue-specific promoters should be examined carefully.

Previous studies reported the mutation efficacy of VIGE system: 13%-100% seeds from infected wheat (Li et al., 2021), 65%-100% seeds from infected N. benthamiana (Ellison et al., 2020), and 4%-9% seeds from infected Arabidopsis (Lei et al., 2021). The efficiency of VIGE in N. attenuata (2.4%-6.3%) was similar to its efficiency in Arabidopsis and much lower than its efficiency in wheat and N. benthamiana. In wheat, the editing efficiency in anther cells including pollen and pollen sacs was more than 75%. This result suggests that a virus infects the shoot apical meristem or floral meristem in wheat, which in turn produces mutated stamens and/or carpels. Recently, Wu et al. (2020) reported that WUSCHEL inhibits the spread of cucumber mosaic virus and TRV in the shoot meristematic region of Arabidopsis. Therefore, TRV would not express gRNA in meristematic cells of N. attenuata. TRV may infect microsporocytes or megasporocytes and mutate a few of them, which explains the mutation efficacy of M₁ seeds (2.4% for TRV2:gPDS, 6.3% for TRV2:gCHAL2, and 4.5% for TRV2:gCHAL3 in pRPS5A:Cas9) that we observed. Virus infection and replication are also host-plant dependent. For instance, N. benthamiana has a natural mutation at RNA-dependent RNA polymerase (RdRP), which enhances the susceptibility of plants to virus infection (Yang et al., 2004). That might be the reason for high heritable mutation efficacy in virus-infected N. benthamiana. Transient silencing of RdRPs may help to enhance the spreading of gRNA in a plant.

The mutation patterns induced by TRV2:gPDS were exclusively single nucleotide insertions in leaves, petals, anthers, ovaries, and M₁ plants (Figs. 2, 4, and 5). TRV2:gCHAL2 and TRV2:gCHAL3 mainly induced single nucleotide insertions and deletions in infected leaves (Fig. 3). These results suggest that mutation patterns are associated with the context of gRNA or target DNA sequences. We did not detect chimeric mutations in M₁ edited seeds, which suggests that VIGE occurs in the ovaries and pollen and/or the progenitor cells of ovules and anthers in N. attenuata. Although both egg and pollen cells were edited in TRV-infected pRPS5A:Cas9 (Figs. 4 and 5), we could not obtain biallelic M₁ seeds from infected plants. A low mutation frequency in each haploid cell might be responsible.

The transformation and regeneration procedure of N. attenuata takes more than six months to generate T₀ plants (Kang et al., 2020). In contrast, our TRV-derived VIGE system takes an average of three to four months to obtain monoallelic M₁ seeds without sterilized equipment, culture media, and intensive labor (Fig. 1). Thus, the time and effort required to obtain gene-edited plants are dramatically reduced. However, it is worthwhile mentioning that at least one transgenic plant should be generated using a traditional transformation method when applying this technique to other plants. In this study, we developed a TRV-derived VIGE system for producing heritable mutants in N. attenuata expressing SpCas9-driven by the RPS5A promoter. The mutation frequencies of our VIGE system were about 2%-6%, which are similar to the frequency of Arabidopsis transformation (1%-5%). Just as the floral dip method enables scientists to explore gene function in Arabidopsis, our VIGE system allows us to evaluate the fitness consequences of altering target gene sequences in non-model plant species.

We thank Emily Wheeler, for editorial assistance and Joohee Kim, for technical assistance. The targeted deep sequencing was supported by Bio Core facilities in KAIST. Y.O. and S.G.K. have filed a patent application based on this work. This work was supported by KAIST-funded Global Singularity Research Program (N11200021) and the Samsung Science & Technology Foundation (SSTF-BA1901-10).

S.G.K. conceived the research. Y.O. and S.G.K. designed the experiments. Y.O. performed the experiments. Y.O. and S.G.K. wrote the manuscript. All authors read and approved the final manuscript.

The authors have no potential conflicts of interest to disclose.