Tools and targets: The dual role of plant viruses in CRISPR–Cas genome editing

Abstract

The recent emergence of tools based on the clustered, regularly interspaced, short palindromic repeats (CRISPR) and CRISPR-associated (Cas) proteins have revolutionized targeted genome editing, thus holding great promise to both basic plant science and precision crop breeding. Conventional approaches for the delivery of editing components rely on transformation technologies or transient delivery to protoplasts, both of which are time-consuming, laborious, and can raise legal concerns. Alternatively, plant RNA viruses can be used as transient delivery vectors of CRISPR–Cas reaction components, following the so-called virus-induced genome editing (VIGE). During the last years, researchers have been able to engineer viral vectors for the delivery of CRISPR guide RNAs and Cas nucleases. Considering that each viral vector is limited to its molecular biology properties and a specific host range, here we review recent advances for improving the VIGE toolbox with a special focus on strategies to achieve tissue-culture-free editing in plants. We also explore the utility of CRISPR–Cas technology to enhance biotic resistance with a special focus on plant virus diseases. This can be achieved by either targeting the viral genome or modifying essential host susceptibility genes that mediate in the infection process. Finally, we discuss the challenges and potential that VIGE holds in future breeding technologies.

Abbreviations
• AtRH8
• Arabidopsis thaliana RNA helicase
• BeYDV
• Bean yellow dwarf virus
• BSMV
• Barley stripe mosaic virus
• BYSMV
• Barley yellow striate mosaic virus
• Cas
• CRISPR-associated
• CP
• coat protein
• CRISPR
• clustered, regularly interspaced, short palindromic repeats
• crRNA
• CRISPR RNA
• DBP
• DNA-binding protein phosphatase
• DDX
• DEAD-box RNA helicase
• DSB
• double-strand break
• eEF1A
• eukaryotic translation elongation factor 1A
• eIF
• eukaryotic translation initiation factor
• FoMV
• Foxtail mosaic virus
• FT
• Flowering locus T
• GMO
• genetically modified organism
• HDR
• homology-directed repair
• InDels
• insertion–deletion mutations
• mRNA
• messenger RNA
• NHEJ
• nonhomologous end joining
• PABP
• poly-A binding protein
• PAM
• protospacer adjacent motif
• PPV
• Plum pox virus
• PVX
• Potato virus X
• RNP
• ribonucleoprotein
• sgRNA
• single guide RNA
• SSN
• site-specific nuclease
• SYNV
• Sonchus yellow net virus
• TALEN
• transcription activator-like effector
• TBSV
• Tomato bushy stunt virus
• TEV
• Tobacco etch virus
• TMV
• Tobacco mosaic virus
• tracrRNA
• trans-activating crRNA
• TRV
• Tobacco rattle virus
• TuMV
• Turnip mosaic virus
• VIGE
• virus-induced genome editing
• VPg
• viral protein genome-linked
• ZFN
• zinc-finger nuclease

A BRIEF RETROSPECTIVE ON STRATEGIES FOR TARGETED GENOME EDITING IN PLANTS

Plant breeding for the improvement of crops essential to human nutrition dates back to the very beginnings of agriculture (Hartung & Schiemann, 2014). Farmers have long relied on traditional breeding to obtain new cultivars by crossing major edible crops with wild plants harboring the desired traits (Scheben et al., 2017). Although genetic recombination increases variability, thousands of years of directed evolution have fixed large portions of crop genomes, thus limiting the potential for improving many traits. Since genes were identified as the underlying elements controlling qualitative and quantitative traits in plants, mutation breeding has been used to increase genetic diversity by introducing random modifications through physical irradiation or chemical mutagens (Pacher & Puchta, 2017). However, these procedures are characterized by their stochastic nature so that large numbers of mutants must be screened to identify the desired modifications. Such nonspecific and time-consuming breeding programs cannot fulfil demands for increased crop production even when marker-assisted breeding is used to enhance selection efficiency. In the mid-1990s, recombinant DNA technology was adopted to break the bottleneck of reproductive isolation. In transgenic breeding, exogenous genes are transferred into elite cultivars to generate the desired traits (Prado et al., 2014). Transgenesis greatly increases genetic variability beyond conventional techniques but also raises concerns regarding the potential impact of genetically modified organisms (GMOs) on human health and the environment. Transgene insertion occurs at random positions into the genome, which may also have unintended effects on the target organism. Consequently, the commercialization of GMOs has been strictly limited in many countries through long and costly regulatory processes (Hartung & Schiemann, 2014; O'Connell et al., 2014).

The remarkable progress in functional genomics over the last several decades presents the opportunity to modify genomes predictably, thereby boosting the transition to precision plant breeding (Osakabe et al., 2010; Sikora et al., 2011). This is possible because of the discovery of site-specific nucleases (SSNs) that promote deletions, insertions, or replacements of a DNA sequence into the target genome (Voytas, 2013).

Core Ideas

• Viral vectors are useful tools for the transient delivery of guide RNAs and Cas nucleases into plant cells.
• Virus-mediated delivery of CRISPR–Cas reaction components avoids limitations associated to stable transformation.
• Resistance to virus diseases can be achieved by CRISPR–Cas editing of viral genes or host susceptibility factors.

Plants are exposed to diverse environmental (e.g., UV radiation, natural radioactivity, and pathogen infection) and biological (e.g., DNA replication and metabolic processes) conditions that cause DNA damage, with double-strand breaks (DSBs) being the most mutagenic lesions. To maintain genome stability, DSBs are repaired by one of the two major host repair mechanisms (Que et al., 2019) (Figure 1). Nonhomologous end joining (NHEJ) is by far the preferred pathway in somatic cells. It involves minimal end processing, thus generating small insertion–deletion mutations (InDels) at the junction point. These modifications can cause a frameshift mutation or alter key amino acids in the gene product, leading to gene disruption (Chang et al., 2017; Seol et al., 2018) (Figure 1, left panel). Because of its high efficiency, NHEJ is broadly used for large-scale knock-out experiments, although it lacks sufficient accuracy for sophisticated engineering. The second repair pathway, known as homology-directed repair (HDR), is activated when a template with significant homology to the DSB is present such as the sister chromatid or exogenous DNA (Ceccaldi et al., 2016) (Figure 1, right panel). Exact rejoining of the broken ends can be used to precisely introduce specific point mutations or to insert the sequence of interest into the target gene. However, HDR-based approaches remain challenging because of their low efficiency and the limitations of donor DNA delivery in plants.

FIGURE 1. Double-strand break (DSB) repair mechanisms. DSBs that occur in the genome are usually repaired by error-prone nonhomologous end joining (NHEJ), generating small insertions or deletions that eventually lead to gene disruption (left). In the presence of a DNA template homologous to the sequence surrounding the DSB, homology-directed repair (HDR) enables an accurate gene correction or insertion (right)

Decades of progress in genome editing have culminated in the development of four major classes of SSNs that can specifically bind to a user-selected genomic region and induce precise modifications in the target gene: meganucleases, zinc-finger nucleases (ZFNs), transcription activator-like effectors (TALENs), and clustered, regularly interspaced, short palindromic repeats (CRISPR)–CRISPR-associated (Cas) systems (Figure 2). The state of this field, with a special focus on the programmability and editing specificity of each SSN, is further explained below.

FIGURE 2. Site-specific nucleases (SSNs) used in genome editing. (a) Meganucleases are homodimeric proteins that recognize target DNA sequences 20–40 bp in length. (b) Zinc-finger nucleases (ZFNs) act as dimers, where each ZFN is composed of a zinc finger protein (ZFP) DNA-binding domain at the N-terminal and a FokI nuclease domain at the C-terminal. The linker between both domains is represented by a black line (spacer). The ZFNs typically recognize target DNA sequences 18–36 bp in length excluding spacers. (c) Transcription activator-like effector nucleases (TALENs) are dimeric enzyme SSNs similar to ZFNs. Each TALEN consists of a TALE DNA-binding domain (highly conserved 34-amino-acid sequence except for positions 12 and 13) at the N-terminal and the FokI domain at the C-terminus. TALENs typically recognize target DNA sequences 30–40 bp in length excluding spacers. (d) A CRISPR–Cas system is composed of the monomeric Cas endonuclease and a single guide RNA (sgRNA), which is complementary to a 20-bp target DNA sequence upstream of the protospacer-adjacent motif (PAM). The SSNs and their corresponding target DNA sequences are not to scale

Meganucleases

Meganucleases are derived from naturally occurring enzymes encoded by mobile introns. They are natural mediators of DNA targeting and were the first SSNs to be used for genome editing in plants, with yeast I-SceI meganucleases being the best characterized (Puchta & Fauser, 2014; Voytas, 2013). These proteins recognize large DNA sequences (20–40 bp) and typically act as a dimer of two identical subunits (Figure 2a). Despite their small size (∼165 amino acids for a meganuclease monomer), difficulties arise in engineering such a large recognition sequence by protein redesign; consequently, meganucleases are now used much less than other SSNs.

Zinc-finger nucleases

Zinc-finger nucleases are artificial bipartite enzymes (∼310 amino acids in size) consisting of a specific zinc-finger protein domain that binds to DNA and is linked to the catalytic domain of the FokI endonuclease (Urnov et al., 2010) (Figure 2b). The zinc-finger protein module contains three to six Cys₂–His₂ arrays derived from a human transcription factor and recognizes a 3-bp region in the DNA. Subsequently, FokI dimerization is needed to cleave an 18–24 bp target sequence and induce DSBs with staggered cut ends. Zinc-finger nucleases have been successfully deployed for genome modification of various plant species (Ainley et al., 2013; Curtin et al., 2011). Nevertheless, the construction of modular enzymes is time-consuming and expensive; along with low specificity and a high rate of off-target mutations, this has limited their application.

Transcription activator-like effector nucleases

Transcription activator-like effector nucleases comprise a customizable array of TALEs artificially fused to the FokI cleavage domain (Christian et al., 2010) (Figure 2c). The DNA-binding domain contains 13–28 copies of a tandemly arrayed 34-amino acid sequence known as TALE repeat, which is highly conserved except for positions 12 and 13. These two residues specifically recognize a single base within the target sequence and thus determine the DNA-binding specificity. Transcription activator-like effector nucleases possess a high editing efficiency and are considered the most specific of all SSNs since their target sequence is 50–60 bp long. As for drawbacks, a new TALE array must be designed for each target sequence (Joung & Sander, 2013).

CRISPR–CAS SYSTEMS: THE REVOLUTION OF GENOME EDITING

All of the protein-dependent DNA cleavage systems described above rely on repetitive chimeric proteins that must be customized for each target sequence through an expensive, time-consuming process. In 2012, the field of genome editing was revolutionized by the emergence of systems based on CRISPR and Cas proteins (Cong et al., 2013; Gasiunas et al., 2012; Jinek et al., 2012; Mali et al., 2013). The CRISPR–Cas two-component system consists of the monomeric Cas endonuclease and a customizable single guide RNA (sgRNA) that combines the functions of the CRISPR RNA (crRNA) and the trans-activating crRNA (tracrRNA) (Figure 2d). Unlike previously developed SSNs, sequence specificity is achieved because of a complementary sgRNA that can be easily engineered to specifically target any DNA sequence. Moreover, Cas itself possesses intrinsic nuclease activity, obviating the need to fuse it to an additional protein (Wang et al., 2017a). The design flexibility of this system has favored its rapid adoption in laboratories with no expertise in protein engineering. Consequently, applications of CRISPR–Cas systems have promptly proven to be effective for genome engineering in eukaryotic cells including plant model species and economically important crops (Li et al., 2017; Nekrasov et al., 2013; Zhang et al., 2016).

The CRISPR–Cas system is an RNA-mediated adaptive immune system in bacteria and archaea that provides defense against phage infection (Barrangou et al., 2007; Brouns et al., 2008; Mojica et al., 2005). A CRISPR locus consists of clusters of Cas genes and CRISPR arrays, where series of 21–40 bp direct repeats (i.e., scaffolds) are interspaced with 25–40 bp variable sequences (i.e., protospacers). These protospacers are traces of past invasions of foreign DNA and share a common end sequence known as the protospacer adjacent motif (PAM). With every infection, new fragments of phage DNA are incorporated into the CRISPR array, which can be transcribed into small RNAs that guide Cas proteins to cleave the invader's genetic material. Based on their Cas genes and interference complex, CRISPR–Cas systems are divided into two classes that can be further subdivided into six types (Shmakov et al., 2015). Class-1 systems (types I, II, and IV) employ multiple Cas effectors for pre-crRNA processing, protospacer loading, and target cleavage, while in Class-2 systems (types II, V, and VI) all steps are performed by a single protein.

Class-2 type-II CRISPR–Cas from Streptococcus pyogenes was the first system shown to specifically cleave DNA both in vitro and in eukaryotic cells (Barrangou et al., 2007; Jinek et al., 2012; Mali et al., 2013). This system was repurposed for gene editing by the creation of an artificial sgRNA consisting of a fusion between a crRNA and a fixed tracrRNA via a short loop (Figure 3a). First, the Cas nuclease (Streptococcus pyogenes Cas9, hereinafter referred as Cas9) binds to the scaffold region of the sgRNA to create a catalytically active complex. Then 20 nucleotides at the 5′ end of the crRNA direct the complex to a specific target DNA site adjacent to the 5′-NGG-3′ PAM. Upon genome binding, Cas9 catalyzes the formation of a DSB through its two nuclease domains (RuvC and HNH), which cleave the sgRNA-bound complementary and PAM-containing noncomplementary DNA strands, respectively. The DSBs created by Cas9 activate the host DNA repair mechanisms, thus leading to targeted genomic modifications (see Section 1.2 and Figure 1). Since Cas9 remains bound to the target sequence for a long time, it tends to produce considerably smaller InDels than the large deletions generated by ZFNs and TALENs (Richardson et al., 2016). Moreover, multiple sites can be targeted simultaneously using several sgRNAs while expressing a single Cas9 (i.e., multiplexing), a property difficult to achieve with other SSNs.

FIGURE 3. Schematic overview of the primary Class-2 CRISPR–Cas systems for genome editing. (a) In the Cas9-sgRNA system, CRISPR RNA (crRNA) (light purple) and the trans-activating crRNA (tracrRNA) (green) are fused by a short loop to form a synthetic single guide RNA (sgRNA). When Cas9 recognizes a 5′-NGG-3′ protospacer-adjacent motif (PAM) in the target DNA, the 20-bp protospacer of the crRNA (yellow) base-pairs with its complementary DNA site, and the endonucleolytic cleavage generates a PAM-proximal blunt double-strand break (DSB). (b) In the Cas12a-crRNA system, the mature crRNA (light purple) is generated by intrinsic RNase activity. When Cas12a recognizes a 5′-TTTV-3′ PAM in the target DNA, the 23-bp protospacer of the crRNA (yellow) base-pairs with its complementary DNA site, and the endonucleolytic cleavage generates a PAM-distal DSB with 5′ overhangs

Exploring the diversity of bacterial CRISPR–Cas systems

The potential applications of the CRISPR–Cas9 system in gene editing have motivated research in this field at an accelerating pace. In this context, Cas9 variants including SpCas9-VQR, SpCas9-EQR, Cas9-NG, and xCas9 3.7 with PAM requirements of 5′-NGA-3′, 5′-NGAG-3′, 5′-NG-3′, and 5′-NG/GAA/GAT-3′, respectively, have shown efficacy in several plant species (Kleinstiver et al., 2016; Zhang et al., 2019). Additionally, studies on CRISPR–Cas9 diversity in bacteria have led to the identification of orthologs from Neisseria meningitidis (NmCas9) (Hou et al., 2013), Staphylococcus aureus (SaCas9) (Ran et al., 2015), Streptococcus thermophilus (St1Cas9) (Xu et al., 2015), and Campylobacter jejunii (CjCas9) (Kim et al., 2017a). Each orthogonal Cas9 system holds unique characteristics regarding protein size, PAM requirements, and sgRNA scaffolds for performing gene editing (Anders et al., 2016). Both SaCas9 and StCas9 have demonstrated success in diverse plant species with relatively high editing efficiency (Jia et al., 2017; Kaya et al., 2016; Steinert et al., 2015; Veillet et al., 2020). However, the use of orthologs is restricted primarily by their PAM complexity (i.e., 5′-NNGRRT-3′ and 5′-NNAGAAW-3′ for SaCas9 and St1Cas9, respectively), making SpCas9 the SSN of choice for most genome engineering approaches.

The applicability of the CRISPR–Cas9 system depends mostly on the 5′-NGG-3′ PAM being conveniently located within the target gene, thereby restricting the editing machinery to G-rich regions in the genome. This limitation can be counteracted with an alternative Class-2 type V CRISPR system encoding Cas12a as the effector protein (formerly Cpf1) (Zetsche et al., 2015) (Figure 3b). Some evidence indicates that types II and V CRISPR systems evolved through independent but remarkably similar pathways (Koonin et al., 2017; Shmakov et al., 2017). This gives Cas12a features unique from those of Cas9 (Zaidi et al., 2017): (a) a single crRNA uncoupled from the tracrRNA directs the cleavage of target DNA, which is notably shorter than the Cas9 sgRNA (42 vs. 100 nt); (b) target recognition is based on a T-rich PAM sequence (5′-TTTV-3′); (c) a staggered cleavage with 4–5 bp 5′ overhangs is generated, as opposed to the blunt ends induced by Cas9, which may facilitate HDR-mediated knock-in of target genes; (d) the intrinsic RNase activity that Cas12a uses for processing the pre-crRNA can also be exploited to target multiple genes at once; and (e) the smaller size of the Cas12a nuclease may facilitate its cellular delivery. Targeted mutagenesis has already been achieved in plants using Cas12a orthologs from Francisella novicida U112 (FnCas12a), Acidaminococcus sp. BV3L6 (AsCas12a), and Lachnospiraceae bacterium ND2006 (LbCas12a) (Endo et al., 2016; Tang et al., 2017; Wang et al.,2017c; Xu et al., 2017). Direct comparisons among these orthologs show that LbCas12a possesses the highest efficiency of all (Tang et al., 2017). Deep sequencing analysis also shows that LbCas12a mostly induces 6- to 13-bp deletions, in contrast to the editing pattern of Cas9, which is characterized by 1- to 3-bp deletions (Tang et al., 2017).

The large size of Cas proteins currently used for genome editing often restricts delivery into cells, since, for example, vectors have limited packaging capacity. Consequently, researchers are now focusing on engineering compact and versatile CRISPR–Cas systems that could facilitate the next generation of genome editing applications. The first step toward this direction was the discovery of CasΦ, a hyper compact type-V system from huge bacteriophages composed of a single Cas protein half the size of Cas9 or Cas12a (700–800 amino acids) (Pausch et al., 2020). CasΦ recognizes the 5′-TBN-3′ PAM (where B = G, T, or C) and generates staggered cut ends similar to Cas12a. This new system was found to be active when delivered as ribonucleoproteins (RNPs) into Arabidopsis thaliana (L.) Heynh. protoplasts, albeit with a very low editing efficiency (0.85%). More recently, Xu et al. (2021) subjected the exceptionally compact types V-F Cas12f (formerly Cas14) system from uncultivated archaea to RNA and protein engineering. When fused to a transcriptional activator, the miniature system—named CasMINI (only 529 amino acids in size)—could drive high levels of gene activation comparable with that of Cas12a in mammalian cells. This work also demonstrated that CasMINI allows for robust base editing and gene editing with a unique InDel pattern.

CRISPR specifications in plants

Successful CRISPR-mediated editing in plants requires specific vectors, efficient target sites, and suitable delivery methods to be used in each plant species. A crucial factor is the choice of optimal promoters for regulating the expression of the editing machinery (Arora & Narula, 2017). Single guide RNAs are usually regulated by tissue-specific RNA polymerase III promoters driving the expression of small RNAs; for instance, the U6 promoter from A. thaliana (AtU6) in dicots or from maize (Zea mays L.) (TaU6) in monocots. Likewise, Cas9 is placed downstream of RNA polymerase II promoters, such as that of ubiquitin, and tagged with a nuclear localization signal to target the nuclease to the host-cell nucleus. Most work in eukaryotic cells has been conducted using codon-optimized versions of Cas9. This requirement seems to lack particular relevance in plants, since either human (Li et al., 2013; Miao et al., 2013) or plant codon-optimized versions of Cas9 (Feng et al., 2013; Nekrasov et al., 2013; Xie & Yang, 2013) have been validated for gene editing in several plant species.

An additional factor of outstanding importance in gene editing approaches is the selection of a specific target site harboring a short PAM sequence on its 3′ end. Target sites should be determined according to two primary selection criteria: (a) on-target activity, which is defined by the cleavage specificity and predicted editing efficiency of the sgRNA, and (b) minimal or nonexistent off-target effects in order to prevent unintended cuts in the genome. Many bioinformatic tools can help researchers in designing highly specific sgRNAs, such as COSMID (Cradick et al., 2014) or CRISPR-PLANT (Xie et al., 2014), but it is unclear to which extent their predictions agree with actual measurements. Another web tool (named CRISPOR) ranks potential sgRNAs for an input sequence according to their predicted on-target activity and potential off-targets in the genome (http://crispor.org) (Concordet & Haeussler, 2018). Since its creation, the popularity of CRISPOR within the scientific community has greatly increased, as it not only provides an accurate scoring model but also helps with cloning, expressing, and validating sgRNAs. Off-target effects associated with gene editing may not cause serious issues, since the related ethical issues are almost negligible in plants as compared with other organisms. Several studies have reported no off-target cleavage in CRISPR–Cas9-induced mutant plants subjected to whole-genome sequencing (Jacobs et al., 2015; Nekrasov et al., 2017). These findings suggest that off-targeting seems to be particularly low in plants. However, attention must be paid to gene paralogs with almost identical sequences to the targets, which may also be cleaved, thus producing undesired modifications.

CRISPR–Cas approaches beyond DSB-mediated genome editing

Although CRISPR–Cas can efficiently create DSBs at specific target sites, the occurrence of unintended InDels and the low rate of HDR in higher eukaryotes can lead to imprecise modifications. These limitations can be overcome by employing CRISPR–Cas-mediated base-editing systems that induce specific base changes at target sites (Komor et al., 2016; Nishida et al., 2016). Base editors consist of a cytidine or adenosine deaminase artificially fused to a nickase Cas9 (nCas9). Cytosine base editors can convert targeted cytosine to uracil resulting in C-to-T substitution in the genome, whereas adenine base editors mediate the A-to-G conversion. Both types of base editors have been widely used for genome editing in various plant species (Li et al., 2018a, 2018b; Zong et al., 2018). Since base editors do not create a DSB, production of on- and off-target InDels is limited, which in turn greatly increases the precision of genome editing.

Prime editing has recently been developed to generate precise base edits beyond the four transition modifications (Anzalone et al., 2019). Prime editors are composed of an engineered reverse transcriptase fused to a nCas9 and a prime-editing guide RNA, which contains not only the protospacer sequence that directs the nickase to the target genomic region but also and additional sequence spelling the desired specific changes. Prime editors have been adapted successfully to plants through codon, promoter, and editing-condition optimization (Abdullah et al., 2020; Lin et al., 2020, 2021).

Last but not least, the application range of CRISPR–Cas technology can be further expanded using a dead Cas9 (dCas9), where both nuclease domains are inactivated but the enzyme can still target a specific genomic region. dCas9 serves as a scaffold for recruiting effector proteins such as chromatin modifiers or epigenetic effectors, thus leading to programmable gene regulation and epigenome editing (Selma & Orzáez, 2021).

DELIVERY OF CRISPR–CAS COMPONENTS TO PLANTS

The presence of a functional Cas–sgRNA complex in the cell nucleus is crucial for efficient genome editing. Direct delivery of the editing complex may seem to be the simplest option, but transferring such a large protein across the cell membrane is exceptionally difficult. Moreover, plants pose unique challenges such as the presence of a rigid cell wall, the common occurrence of polyploidy, and recalcitrant regeneration in many species (Varanda et al., 2021). Therefore, the delivery and expression of CRISPR–Cas components within a plant cell are crucial steps in the genome editing process.

From GMO approaches to DNA-free genome editing

Current CRISPR–Cas approaches in plants have largely focused on delivering the editing machinery via transformation technologies (Nekrasov et al., 2013) (Figure 4a). By far, the standard method for obtaining transgenic plants is based on Agrobacterium tumefaciens. This plant pathogen possesses the ability to transfer its T-DNA to the host cell nucleus, where it becomes randomly integrated (Nester, 2015). The A. tumefaciens-based approach has been widely used for the delivery of CRISPR–Cas components because of its simplicity and low economic cost. Several studies have confirmed that the stable integration of CRISPR–Cas DNA leads to high editing efficiencies in diverse plant species such as A. thaliana, rice (Oryza sativa L.), tomato (Solanum lycopersicum L.), maize, and grapevine (Vitis vinifera L.) (Miao et al., 2013; Feng et al., 2014; Pan et al., 2016; Char et al., 2017; Tian et al., 2018). The other method commonly used for genetic transformation, particularly in monocot species, is particle bombardment. This consists of coating metallic microprojectiles (generally gold, silver, or tungsten) with DNA constructs that are subsequently fired into plant cells at high pressure. By using gold microparticles carrying CRISPR–Cas components that stably integrate into the host genome, targeted mutations have been successfully produced in several plant species (Wang et al., 2014; Li et al., 2015).

FIGURE 4. Delivery strategies for CRISPR–Cas components into plants. (a) Traditional transformation technologies for the stable expression of CRISPR–Cas DNA combined with herbicide or antibiotic selection. Breeding techniques, such as selfing and crossing, are required for the genetic segregation of the transgene before edited plants can be commercialized. The integration of CRISPR–Cas DNA and genome editing are represented as a blue and a red band within the chromosomes, respectively. (b) Transient expression strategies to obtain transgene-free, edited plants bypassing the selection steps. CRISPR–Cas components can be delivered as DNA into plant cells (top) or as messenger RNA (mRNA) or ribonucleoprotein (RNP) in protoplasts (bottom). Details are as in the legend in Figure 4a. PEG, polyethylene glycol

Despite their high efficiency and multiplexing capacity, the greatest disadvantage of transformation technologies lies in their random integration of foreign DNA, which is likely to produce undesirable off-target mutations. Moreover, crop plants carrying coding regions for Cas9 and sgRNAs are subject to GMO regulation (Gao et al., 2015) and require traditional breeding techniques to remove unwanted DNA before their commercialization (Pyott et al., 2016; Peng et al., 2017). Alternatively, transient expression of CRISPR–Cas components, combined with the elimination of the canonical selection steps, can result in the regeneration of plants free from foreign DNA (Nekrasov et al., 2013; Zhang et al., 2016) (Figure 4b, top panel). This strategy significantly reduces transgene integration, although it does not completely eradicate it, since degraded DNA fragments may still become incorporated into the host genome.

The use of RNPs for the delivery of preassembled Cas9–sgRNA into protoplasts constitutes a promising approach for achieving DNA-free editing in plants (Woo et al., 2015; Liang et al., 2017; Andersson et al., 2018; González et al., 2020) (Figure 4b, bottom panel). Protoplasts are uncovered plant cells that result from the enzymatic digestion of the polysaccharide-rich cell wall. They closely resemble animal cells and can be transfected via electroporation or polyethylene glycol treatment. In the RNP strategy, the gene-editing complex can cleave target DNA immediately upon delivery without requiring the cellular transcription and translational machinery; the complex is rapidly degraded afterward. Consequently, off-target cleavage is considerably reduced as compared with DNA-based expression of CRISPR–Cas. When this strategy is used for gene knock-out, it can yield DNA-free-edited plants that may not be subjected to GMO regulatory issues. More recently, the RNP editing toolbox in plants has also expanded to include Cas12a by delivering both LbCas12a- and AsCas12a-crRNA RNPs into soybean [Glycine max (L.) Merr.] and tobacco (Nicotiana tabacum L.) protoplasts (Kim et al., 2017b). However, regenerating a whole plant from single-celled protoplasts requires tissue culture procedures that frequently generate undesired somaclonal mutations.

Plant viruses as delivery vehicles of CRISPR–Cas components

Plant viruses have been implemented as heterologous gene expression vectors since the beginning of genetic engineering several decades ago. The advent of molecular biology and reverse genetics, as well as the discovery of RNA silencing and the development of high-throughput sequencing technologies, have enabled the manipulation of viral genomes to express heterologous proteins and RNAs in plants (Scholthof et al., 1996). Since then, a cultivar of A. tumefaciens-based viral vectors have been developed for loading into plant cells via simple inoculation methods. Recent studies highlight the potential use of viral vectors as transient delivery vehicles for CRISPR–Cas components in many biological systems including plants (Platt et al., 2014; Senís et al., 2014; Lau & Suh, 2017; Xu et al., 2019). This strategy is known as virus-induced genome editing (VIGE); it has been lauded as a game-changer in CRISPR-based genome editing. Virus-induced genome editing presents several advantages over conventional delivery methods (Cody & Scholthof, 2019). First, owing to prolific viral replication, CRISPR–Cas components can accumulate to high levels within the host cell, leading to efficient (up to 60–70%) and fast (3–7 d) genome editing. This transient expression platform provides an ideal screening tool for assessing the efficacy and specificity of sgRNA designs. Second, when using RNA viruses, targeted modifications can be obtained without the integration of any foreign material into the plant genome, which should avert additional regulatory and ethical issues. Finally, viral vectors can be deployed even if editing the target gene negatively affects plant fitness because they are directly delivered into adult plants and editing can occur rapidly—prior to the onset of any severe physiological effects. Nevertheless, the application of each viral vector encounters challenges related to molecular biology and the specific host range. Since the first VIGE reports, researchers have made great progress in engineering viral vectors that enable efficient delivery of CRISPR–Cas components into plants, which has helped expand the VIGE toolbox to additional plant species and purposes. In the following lines, we discuss the most relevant studies carried out on plant viral-assisted CRISPR–Cas editing (Table 1).

TABLE 1 Plant virus vectors developed for the delivery of CRISPR–Cas components in virus-induced genome editing strategies

Virus	Viral vector	Virus family, genus	Nuclease	Single guide RNA type	Plant species	Systemic expression	Heritability of gene editing	Reference
+Single-stranded DNA	BeYDV	Geminiviridae, Mastrevirus	ZFN, TALEN, SpCas9	AtU6-sgRNA	Tobacco (Nicotiana tabacum)	No	Not determined	Baltes et al., 2014
			TALEN, SpCas9	AtU6-sgRNA	Tomato (Solanum lycopersicum)	No	Not determined	Čermák et al., 2015
					Potato (Solanum tuberosum)	No	Not determined	Butler et al., 2016
			SpCas9	AtU6-sgRNA	Potato (S. tuberosum)	No	Not determined	Butler et al., 2015
					Tomato (S. lycopersicum)	No	Not determined	Dahan-Meir et al., 2018
	WDV	Geminiviridae, Mastrevirus	SpCas9	OsU6-sgRNA	Rice (Oryza sativa)	No	Not determined	Wang et al., 2017b
				TaU6-sgRNA	Wheat (Triticum aestivum)	No	Not determined	Gil-Humanes et al., 2017
	CaLCuV	Geminiviridae, Begomovirus	–	AtU6-sgRNA	N. benthamiana	Yes	Not determined	Yin et al., 2015
	CLCrV	Geminiviridae, Begomovirus	–	AtU6-sgRNA (+/- FT)	Arabidopsis thaliana	Yes	Low frequency (progeny of infected plants)	Lei et al., 2021
+Single-stranded RNA	TRV	Virgaviridae, Tobravirus	Mega-nuclease	–	N. alata	Yes	Low frequency	Honig et al., 2015
			ZFN, TALEN	–	Tobacco (N. tabacum), petunia (Petunia hybrida)	Yes	Not determined	Marton et al., 2010
			–	PEBV-sgRNA	N. benthamiana	Yes	Low frequency (progeny of infected plants)	Ali et al., 2015a
			–	PEBV-sgRNA	A. thaliana, N. benthamiana	Yes	Not determined	Ali et al., 2018
			–	PEBV-sgRNA (+/- FT)	N. benthamiana	Yes	High frequency (progeny of infected plants)	Ellison et al., 2020
	TMV	Virgaviridae, Tobamovirus	–	TMV-sgRNA-ribozyme	N. benthamiana	No	Not determined	Cody et al., 2017
			SpCas9	TMV-sgRNA-ribozyme	N. benthamiana	No	Not determined	Chiong et al., 2021
	PEBV	Virgaviridae, Tobravirus	–	PEBV-sgRNA	A. thaliana, N. benthamiana	Yes	Not determined	Ali et al., 2018
	BSMV	Virgaviridae, Hordeivirus	–	BSMV-sgRNA	N. benthamiana, wheat, maize (Zea mays)	Yes	N. benthamiana: yes (via tissue regeneration)	Hu et al., 2019
				BSMV-sgRNA (+/- FT)	Wheat (Triticum aestivum)	Yes	Yes (progeny of infected plants)	Li et al., 2021
	BNYVV	Benyviridae, Benyvirus	–	p31-sgRNA	N. benthamiana	Yes	Not determined	Jiang et al., 2019
	FoMV	Alphaflexiviridae, Potexvirus	–	FoMV-sgRNA	N. benthamiana, maize (Zea mays), foxtail (Setaria viridis)	Yes	Not determined	Mei et al., 2019
			SpCas9	AtU6-sgRNA	N. benthamiana	Yes	Not determined	Zhang et al., 2020
	PVX	Alphaflexiviridae, Potexvirus	SpCas9	PVX- sgRNA	N. benthamiana	Yes	High frequency (via tissue regeneration)	Ariga et al., 2020
			–	PVX- sgRNA (+/- FT)	N. benthamiana	Yes	High frequency (via tissue regeneration or progeny of infected plants)	Uranga et al., 2021a
	TEV	Potyviridae, Potyvirus	LbCas12a	PVX- sgRNA (in PVX vector)	N. benthamiana	Yes	Not determined	Uranga et al., 2021b
			AcrIIA4	PVX- sgRNA-FT (in PVX vector)	N. benthamiana	Yes	Not determined	Calvache et al., 2022
−Single-stranded RNA	BYSMV	Rhabdoviridae, Cytorhabdovirus	SpCas9	BYSMV-sgRNA	N. benthamiana	No	Not determined	Gao et al., 2019
	SYNV	Rhabdoviridae, Betanucleo-rhabdovirus	SpCas9	SYNV-sgRNA-tRNA	N. benthamiana	Yes	Yes (via tissue regeneration)	Ma et al., 2020

Note. AcrIIA4, Streptococcus pyogenes anti-CRISPR/Cas9 protein; BeYDV, Bean yellow dwarf virus; BNYVV, Beet necrotic yellow vein virus; BSMV, Barley stripe mosaic virus; BYSMV, Barley yellow striate mosaic virus; CaLCuV, Cabbage leaf curly virus; CLCrV, Cotton leaf crumple virus; FoMV, Foxtail mosaic virus; LbCas12a, Lachnospiraceae bacterium ND2006; PEBV, Pea early browning virus; PVX, Potato virus X; SpCas9, Streptococcus pyogenes CRISPR-associated protein 9; SYNV, Sonchus yellow net virus; TALEN, transcription activator-like effector; TEV, Tobacco etch virus; TMV, Tobacco mosaic virus; TRV, Tobacco rattle virus; WDV, Wheat dwarf virus; ZFN, zinc-finger nuclease.

DNA viruses as proof of concept for VIGE

The first VIGE reports were based on the use of geminiviruses for generating plant knock-out lines. Geminiviruses are a widespread group of plant DNA viruses characterized by single-stranded, circular DNA with monopartite or bipartite genomes (Lozano-Durán, 2016). Their small size facilitates easy manipulation of the viral genome; on the other hand, it physically limits their cargo capacity. The coat protein (CP) of some bipartite geminiviruses can be replaced by a heterologous DNA of up to 800–1,000 bp while maintaining most of the features required for viral movement and replication. Although this modification enables the virus to produce large amounts of sgRNA, it is nevertheless insufficient for carrying long DNA fragments such as genes encoding Cas nucleases.

With the aim of further increasing geminivirus cargo capacity, further removal of CP and movement protein coding sequences results in noninfectious replicons. Since deconstructed viruses are unable to move systemically, they must be delivered into plants by A. tumefaciens-mediated transformation in contrast to other techniques such as agroinfiltration or mechanical inoculation, where infectious replicons are required. Bean yellow dwarf virus (BeYDV) was the first to be adapted for the expression of ZFNs and TALENs, as well as Cas9 nuclease and sgRNA in plants (Baltes et al., 2014; Čermák et al., 2015; Butler et al., 2016), showing a considerable cargo capacity and notably higher editing efficiency than conventional A. tumefaciens T-DNA delivery. Another application of geminiviruses in VIGE relates to gene knock-in. Because replication of DNA viruses occurs in the host-cell nucleus with high copy numbers, it leads to the production of vast numbers of repair templates that are required for HDR to outcompete NHEJ. As a result, an increase in the frequency of HDR has been documented in several works with BeYDV (Dahan-Meir et al., 2018), Cabbage leaf curly virus (CaLCuV) (Yin et al., 2015), and Wheat dwarf virus (WDV) (Wang et al., 2017b; Gil-Humanes et al., 2017).

sgRNA delivery with positive-strand RNA viruses

The use of DNA viruses for gene editing inevitably creates the possibility of accidentally integrating foreign genetic material into the host genome. Conversely, RNA viruses present the advantage of developing their infectious cycles exclusively in the cytoplasm, thus resulting in plants free from foreign DNA, which should avoid raising regulatory and ethical issues.

Several plant RNA-virus-based vectors have been tested as vectors for the delivery of sgRNAs into plants. Tobacco rattle virus (TRV; genus Tobravirus) is a bipartite positive-strand RNA virus composed of TRV1 and TRV2 genomes (Varanda et al., 2021). TRV1 is essential for viral replication and movement, whereas TRV2 encodes the CP and other nonstructural proteins. This nonessential region has been engineered for the expression of heterologous proteins, as well as fragments of host-plant genes for virus-induced gene silencing (Senthil-Kumar & Mysore, 2014). Compared with geminivirus, TRV carries various advantages for VIGE including a broad host range (>400 plant species) and the ability to migrate to growing plant tissues (Cody & Scholthof, 2019).

Tobacco rattle virus-based vectors were first used in plant genome engineering to deliver ZFNs and TALENs into tobacco and petunia (Petunia ×hybrida hort. ex E. Vilm.) (Marton et al., 2010). In this system, targeted editing enables the recovery of a reporter gene; most importantly, modifications are transmitted to the next generation. A similar approach has been used for the expression of meganucleases in flowering tobacco (Nicotiana alata Link & Otto), yielding mutant seeds from infected plants (Honig et al., 2015). Regarding CRISPR–Cas technology, Ali et al. (2015a) have developed a TRV-based vector as an sgRNA delivery vehicle to successfully edit A. thaliana and Nicotiana benthamiana Domin genomes. The detection of targeted modifications in the progeny of infected plants confirms the ability of TRV to infect germline cells. However, these same authors demonstrated in later work that Pea early browning virus (PEBV; genus Tobravirus) delivered sgRNAs into the plant more efficiently than TRV, while also being able to infect meristematic tissue, which allowed the recovery of seeds with the desired modifications although at a low rate (Ali et al., 2018).

Tobacco mosaic virus (TMV; genus Tobamovirus) is a monopartite positive-strand RNA virus that expresses large amounts of CP from a viral subgenomic promoter. It can be easily manipulated via partial substitution of CP with heterologous genes, which then allows for high-level gene expression in several hosts and prolonged integrity of its derived vectors (Gleba et al., 2007). Based on this potential, a TMV-based vector has been developed for sgRNA delivery (Cody et al., 2017). Although CP deletion impairs the systemic movement of the virus, high concentrations of sgRNA can be delivered, leading to efficient editing in N. benthamiana plants previously infiltrated with a plasmid expressing Cas9.

Transgenic plants constitutively expressing Cas9 are a useful tool for exploring the feasibility for VIGE of plant viral vectors. For instance, Beet necrotic yellow vein virus (BNYVV; genus Benyvirus) replicons previously engineered for simultaneous expression of foreign proteins have been shown to deliver sgRNAs for efficient genome editing in N. benthamiana (Jiang et al., 2019). This approach has also been expanded for targeted mutagenesis in nonmodel plant species using Barley stripe mosaic virus (BSMV; genus Hordeivirus) in wheat (Triticum aestivum L.) and maize (Hu et al., 2019) or using Foxtail mosaic virus (FoMV; genus Potexvirus; family Alphaflexiviridae) in N. benthamiana, maize, and green foxtail (Setaria viridis (L.) P. Beauv.) (Mei et al., 2019). A later report with BSMV also showed that inoculation with a pool of viruses harboring different sgRNA was effective in inducing simultaneous editing in multiple target genes in wheat, most of them being transmitted to the next generation (Li et al., 2021).

Virus-mediated expression of Cas proteins

All of the previously mentioned VIGE attempts were based on plant virus-mediated expression of sgRNAs in transgenic plants constitutively expressing Cas9 nuclease. Researchers had long failed in their efforts to develop a plant virus-derived vector able to deliver the entire CRISPR–Cas system and move systemically, since the expression of large heterologous genes with positive-strand RNA viruses is known to affect vector stability and cell-to-cell movement (Avesani et al., 2007). This limitation was overcome when two independent reports demonstrated the use of negative-strand RNA viruses to deliver both Cas9 nuclease and sgRNAs at the whole-plant level.

Rhabdoviruses form a diverse family of negative-strand RNA viruses that show stable expression of foreign genes up to 6 kb in vertebrates (Jackson & Li, 2016). This property creates the possibility of delivering Cas9 proteins into plant cells, even though plant rhabdoviruses have been poorly exploited because of difficulties in genetic manipulation (Zhou et al., 2019). In a first report, Gao et al. (2019) validated targeted mutagenesis in N. benthamiana using Barley yellow striate mosaic virus (BYSMV) to express both Cas9 and the sgRNA. Moreover, work by Ma et al. (2020) shows that Sonchus yellow net virus (SYNV)-mediated expression of all CRISPR–Cas9 components led to an effective DNA-free genome edit, either in single or multiple genes. Unlike BYSMV, the SYNV vector generated modifications at the whole-plant level that were inherited by subsequent generations of the edited plants. However, since rhabdoviruses rarely infect germline cells, tissue culture is still necessary to obtain individual edited plants.

More recently, two positive-strand RNA viruses in the genus Potexvirus have been developed to express Cas9 proteins with the help of RNA interference suppressor p19, derived from Tomato bushy stunt virus. Because of their filamentous flexible structure, gene insert size may not be physically limited in potexviruses, as is the case with small isometric viruses (Kendall et al., 2008). A first approach based on FoMV confirms that N. benthamiana was edited at the whole-plant level when viral vectors harboring both Cas9 and the sgRNA were delivered together (Zhang et al., 2020). Nonetheless, no mutations were transmitted to germline cells, and it is unclear how two FoMV replicons could infect the same cell. Alternatively, Potato virus X (PVX) has been engineered to deliver the entire CRISPR–Cas9 system into plants, and the recombinant vector was both agroinfiltrated and mechanically inoculated in N. benthamianai in a study by Ariga et al. (2020). In their study, most cells were infected with PVX-Cas9 RNA and showed high levels of Cas9 expression, while the integration of viral DNA into the host genome was scarce. Targeted modifications were inherited in the next generation with no PVX transmission through seeds, resulting in foreign-DNA-free edited progeny. Later work by Uranga et al. (2021a) demonstrates that PVX allows for the delivery of unspaced sgRNA arrays, thus leading to highly efficient multiplex editing. Moreover, virus-free edited progeny carrying biallelic mutations were obtained from plants regenerated from infected tissue as well as when plants were inoculated with mobile sgRNAs.

The genus Potyvirus includes more than 200 species that infect a wide range of host plants from diverse botanical families and are known to establish synergistic interactions with a broad range of unrelated viruses, resulting in the accumulation of the nonpotyviral partner (Wylie et al., 2017). This property was recently explored to develop a two-virus vector system for the simultaneous delivery of all CRISPR–Cas reaction components into plants (Uranga et al., 2021b). In this system, PVX-mediated sgRNA delivery was combined with the coinoculation of a Tobacco etch virus (TEV) vector transiently expressing a Cas12a nuclease smaller than that of SpCas9. Interestingly, the interaction between PVX and TEV resulted in highly efficient genome editing in N. benthamiana. Additionally, gene editing can be temporarily controlled by using phage-derived inhibitors of CRISPR–Cas immunity, known as anti-CRISPR proteins (Borges et al., 2017; Trasanidou et al., 2019; Marino et al., 2020). Recently, Calvache et al. (2022) successfully demonstrated that TEV-mediated delivery of anti-CRISPR proteins completely abolished the high gene-editing levels obtained via the PVX–sgRNA system. This innovative two-virus vector system constitutes a major improvement of the VIGE toolbox for achieving transformation-free genome editing in plants.

Toward tissue-culture-free editing of plants

Except for A. thaliana and its close relatives that are suitable for floral dip transformation, tissue culture is required in most plant species to regenerate whole plants from somatic cells. Besides being a time-consuming and labor-intensive process, this strategy can generate undesired genetic or epigenetic variations. In tissue culture, hormones are added to the medium to stimulate cell division and maintain the resulting callus at a similar stage as meristematic cells, although the necessary levels of these hormones vary between species and must be optimized individually. The finding that transient expression of developmental regulators such as Wuschel, Shoot meristemless, and Isopentenyl transferase induces somatic embryogenesis in plants constituted a breakthrough advance in the field (Lowe et al., 2016). Very recently, a novel approach explored the simultaneous delivery of developmental regulators and sgRNAs in Cas9-expressing N. benthamiana plants, either by coculturing seedlings germinated in liquid culture with A. tumefaciens or by agroinfiltration of soil-grown plants (Maher et al., 2020) (Figure 5a). Ectopic expression of Wuschel-, Shoot meristemless-, and Isopentenyl transferase-induced de novo meristem formation and edited shoots could be regenerated, thereby bypassing tissue culture. In a different study, the simultaneous expression of Growth-regulating factor 4 and GRF-interacting factor 1 with CRISPR–Cas9 notably increased the frequency of genome editing (Debernardi et al., 2020). These results suggest the potential for the combined delivery of CRISPR–Cas components and developmental regulators as a promising future method for DNA manipulation in a broad range of recalcitrant species.

FIGURE 5. Tissue culture-free strategies for plant genome editing. (a) Overexpression of developmental regulators can induce de novo formation of meristem carrying the targeted genome modifications. These shoots can either be propagated for regeneration or can set flowers directly for the collection of edited seeds. (b) In Cas9-expressing plants, systemic infection with a recombinant RNA virus carrying mobile single guide RNAs (sgRNAs) can cause mutations in the germline, thus resulting in a high rate of edited seeds.

Another recent approach aiming to bypass tissue culture used a positive-strand RNA virus (in this case, TRV) to deliver mobile sgRNAs into Cas9-expressing plants (Ellison et al., 2020) (Figure 5b). In this study, the viral vector was engineered to produce an sgRNA fused on its 3′ end to mobile RNA elements like Flowering locus T (FT) messenger RNA (mRNA), which promoted the mobility of the editing complex to apical meristems. Surprisingly, these apical parts of the plants showed a higher editing efficiency than that of the initially infected tissues. Moreover, the induction of mutations in the germline resulted in a high rate of biallelic mutations, with no evidence of viral transmission in the progeny of infected plants. This strategy is of particular interest when using plants with a single copy of Cas9, as transgene-free edited plants can be easily obtained by the segregation of the edited progeny. However, the effect of FT on heritable editing may be partly dependent on the biological properties of the employed viral vector. For instance, a recent work showed that the unique ability of BSMV to enter apical meristems leads to the efficient transmission of edits to the next generation, while the addition of FT to the sgRNA failed to promote, and even decreased, the efficiency of heritable mutagenesis. Moreover, it was demonstrated that transgene- and virus-free edited progeny can be obtained by crossing wild-type wheat with Cas9-transgenic wheat pollen previously infected with BSMV replicons (Li et al., 2021).

ENHANCING VIRUS RESISTANCE IN PLANTS VIA CRISPR–CAS TECHNOLOGY

Plant viruses pose a serious threat to many economically relevant crops. They are responsible for nearly half of all plant diseases, leading to massive losses to agricultural production every year (Zaidi et al., 2016). Moreover, they are difficult to control and cannot be eradicated through the chemical application. In the context of a rapidly growing population, there is a major need to develop novel disease-resistance strategies in order to meet global food demand. In recent years, impressive progress in next-generation sequencing techniques and several omics (i.e., transcriptomics and proteomics) has provided valuable insight into the defense and virulence pathways of both plants and pathogens (Barakate & Stephens, 2016). Key genes involved in pathogenesis have been identified by studying changes in gene expression, protein modifications, and protein–protein interactions. By incorporating this knowledge into genome editing tools, researchers have been able to improve biotic stress resistance in crops. In this section, we discuss recent studies demonstrating the capacity of CRISPR–Cas technology against virus infection based on two different strategies: (a) targeting the viral genome (Table 2) or (b) modifying host factors essential in the virus life cycle (Table 3).

TABLE 2 Representative viral factors targeted to induce virus resistance in plants

Affected virus	Viral factor	Gene function	Plant species	References
BSCTV	LIR, Rep/RepA	RCA mechanism	Nicotiana benthamiana	Baltes et al., 2015
BeYDV	CP, Rep, IR	RCA mechanism	Arabidopsis thaliana, N. benthamiana	Ji et al., 2015
TYLCV, BCTV, MeMV	CP, Rep, IR	RCA mechanism	N. benthamiana	Ali et al., 2015b
CMV, TMV	ORF1, 2, 3; CP; 3′ UTR	Replication mechanism	A. thaliana, N. benthamiana	Zhang et al., 2018
TuMV	HC-Pro, CP	Replication mechanism	N. benthamiana	Aman et al., 2018

Note. BCTV, Beet curly top virus; BeYDV, Bean yellow dwarf virus; BSCTV, Beet severe curly top virus; CMV, Cucumber mosaic virus; CP, coat protein; HC-Pro, helper component proteinase; IR, intergenic region; LIR, long intergenic region; MeMV, Merremia mosaic virus; ORF, open reading frame; RCA, rolling-circle amplification; Rep, replication association protein; TMV, Tobacco mosaic virus; TuMV, Turnip mosaic virus; TYLCV, Tomato yellow leaf curl virus; UTR, untranslated region.

TABLE 3 Representative host factors (S genes) for inducing virus resistance in plants

Host factor	Gene function	Affected virus and viral factor	Plant species	References
eIF(iso)4E	Formation of the translation initiation complex	TuMV VPg	Arabidopsis thaliana	Lellis et al., 2002
		PVY, PVMV, TEV VPg	Pepper (Capsicum spp.)	Ruffel et al., 2002, 2006
		TEV VPg	A. thaliana, tomato (S. lycopersicum)	Contreras-Paredes et al., 2013; Estevan et al., 2014
		CVYV, ZYMV, PRSV-W VPg	Cucumber (Cucumis sativus)	Chandrasekaran et al., 2016
eIF4G	Formation of the translation initiation complex	RTSV VPg	Rice (Oryza sativa)	Macovei et al., 2018
eEF1A	Translation elongation and unfolded protein response	TuMV NIb	A. thaliana	Thivierge et al., 2008
		TMV 126K and 3′ UTR	Nicotiana benthamiana	Yamaji et al., 2010
		BaMV 3′ UTR	N. benthamiana	Lin et al., 2007
PABP2, PABP4, PABP8	Translation initiation	TuMV RdRp, VPg	A. thaliana	Dufresne et al., 2008
RH8 and DDXL	mRNA binding and processing	TuMV, PPV VPg	A. thaliana, peach (Prunus persica)	Huang et al., 2010
DBP1	Proteosome-mediated regulation of eIF(iso)4E	TuMV, PPV undetermined	A. thaliana	Castelló et al., 2010
PDL1, PDL2, PDL3	Cell-to-cell trafficking	TMV Cg, ToMV 130K and 180K	A. thaliana, N. benthamiana	Amari et al., 2010
PCaP1	Microtubule depolymerization	TuMV P3N-PIPO	A. thaliana	Vijayapalani et al., 2012
SEC24A	Intracellular protein transport	TuMV 6K2	A. thaliana	Jiang et al., 2015

Note. CVYV, Cucumber vein yellowing virus; DBP, DNA-binding protein phosphatase; DDX, DEAD-box RNA helicase; eEF1A, eukaryotic translation elongation factor 1A; eIF, eukaryotic translation initiation factors; ; NIb, nuclear inclusion b; PABP, Poly-A binding protein; PDL, plasmodesmata-located protein; PPV, Plum pox virus; PRSV-W, Papaya ringspot mosaic virus-W; PVY, Potato virus Y; PVMV, Pepper veinal mottle virus; RdRp, RNA-dependent RNA polymerase; RTSV, Rice tungro spherical virus; TEV, Tobacco etch virus; TMV, Tobacco mosaic virus; TuMV, Turnip mosaic virus; UTR, untranslated region; VPg, viral genome-linked protein; ZYMV, Zucchini yellow mosaic virus; 126K, 126-KDa protein.

Editing the viral genome

Most studies related to CRISPR–Cas editing for disease resistance have focused on targeting single-stranded DNA viral genomes (Baltes et al., 2014; Yin et al., 2015; Ali et al., 2016) (Figure 6, left panel). Geminiviridae is a large family of plant single-stranded DNA viruses that cause huge losses in agricultural production among relevant crop families including Cucurbitaceae, Euphorbiaceae, Solanaceae, Malvaceae, and Fabaceae (Zaidi et al., 2016). Initial studies have focused on disrupting essential viral replication genes to inhibit viral spread (Table 2). Following this aim, modifications in coding and noncoding regions of Beet severe curly top virus (BSCTV) have conferred significant disease resistance in A. thaliana and N. benthamiana (Ji et al., 2015). Similarly, a significant reduction in viral load was reported when targeting the replication initiator protein of BeYDV (Baltes et al., 2015), as well as when editing the viral replication initiator protein and CP or the conserved intergenic region in Tomato yellow leaf curl virus (TYLCV), Beet curly top virus (BCTV), and Merremia mosaic virus (MeMV) (Ali et al., 2015b). Results from this last work suggest that modifications in the coding regions of geminiviruses gave rise to virus variants capable of replicating and moving systemically to escape the CRISPR–Cas machinery. Conversely, the disruption of the intergenic region led to durable, broad-spectrum resistance against multiple viruses without viral variant escapes.

FIGURE 6. Development of CRISPR–Cas-mediated virus resistance in plants. (a) Protection against single-stranded DNA viruses can be achieved by disrupting essential viral replication genes, which then inhibits viral spread. The CRISPR–Cas machinery must be stably integrated into the plant to edit the pathogen's genome whenever it enters the host. (b) Protection against RNA viruses can be achieved by editing receptors (R genes) located on the plant cell surface that recognize the pathogen and trigger defense or by editing host proteins (S genes) that are recruited by the pathogen to complete the infection cycle. When host genes are targeted, the CRISPR–Cas transgene can be transiently expressed or segregated to obtain nontransgenic edited plants

On the other hand, protection against RNA viruses was long considered unachievable because SpCas9 is a DNA cutter. This limitation was recently overcome by the discovery of CRISPR nucleases capable of cleaving RNA. By expressing the Cas variant from Francisella novicida (FnCas9) and RNA-targeting sgRNAs, a notable reduction (40–80%) in the viral accumulation of Cucumber mosaic virus (CMV) or Tobacco mosaic virus (TMV) was observed in A. thaliana and N. benthamiana (Zhang et al., 2018) (Table 2). A parallel study also confirmed that the RNA-guided Cas13a nuclease from Leptotrichia wadei (LwaCas13a) can target the Turnip mosaic virus (TuMV) genome, thus reducing viral replication in tobacco (Aman et al., 2018).

Editing the host plant genome

All of the aforementioned strategies rely on the stable expression of Cas9 nuclease in the host plant to achieve a disease-resistant phenotype, which inevitably subjects the edited plants to GMO regulation. On the contrary, when host genes are targeted, it is possible to segregate the CRISPR–Cas transgene and release nontransgenic edited plants in the field (Chandrasekaran et al., 2016; Pyott et al., 2016; Macovei et al., 2018). The most-used strategy in disease resistance breeding is based on the introduction of dominant or semidominant Resistance (R) genes into elite cultivars via crosses with wild relatives or transfer between plant species (Win et al., 2012; Dangl et al., 2013). Resistance genes primarily encode nucleotide-binding and oligomerization domain-like receptors, which recognize pathogen effectors that alter molecular processes within the plant to support pathogen growth (Dodds & Rathjen, 2010) (Figure 6, right panel). Notwithstanding its efficacy against a plethora of pathogens, nucleotide-binding and oligomerization domain-like receptors tend to have a narrow recognition spectrum. Moreover, the strong selection pressure that large monocultures impose on the pathogen population is counteracted by effector diversification. Consequently, pathogens can hinder host defenses and easily adapt to new resistant cultivars.

Susceptibility (S) genes have emerged as an alternative to R-gene-mediated disease resistance breeding. S genes encode either negative regulators of immunity or host proteins involved in plant endogenous pathways that, when controlled by the pathogen, lead to the suppression of defense mechanisms, thus promoting the spread of infection (Hashimoto et al., 2016; Garcia-Ruiz, 2018) (Figure 6, right panel). The removal or inactivation of an S gene directly interferes with pathogen propagation and leads to a phenotype characterized by milder symptoms or even a complete absence of infection on occasion. The recessive nature of S genes offers the potential to generate a more durable resistance in the field (Kang et al., 2005).

To date, some of the most studied S genes in monocots and dicots are the eukaryotic translation initiation factors (eIF) 4E and 4G, along with their isoforms. Translation of viral RNA is a crucial step in a pathogen's life cycle before viral replication can occur, and it relies completely on the host translational machinery. In eukaryotic organisms, mRNA translation is orchestrated by a multicomponent complex consisting of eIF4E and eIF4G, which form the eIF4F core, as well as other factors that recruit ribosomes to the 5′ untranslated region (Lellis et al., 2002). It is well-documented that the viral genome-linked proteins (VPg) from potyviruses interacts with both the 5′ untranslated region of the viral genome and host eIFs to hijack the host translation complex (Ruffel et al., 2006; Ling et al., 2009). Unlike cellular mRNAs that can employ both eIF4E and eIF(iso)4E, various potyviral VPg have evolved to bind specifically to a single isoform. Recessive resistance related to eIF(iso)4E deficiency was first described in A. thaliana mutants exhibiting a loss of susceptibility to TEV (Lellis et al., 2002) (Table 3). Subsequent studies have confirmed eIF4Es-mediated resistance against potyviruses in various crops including lettuce (Lactuca sativa L.), pepper (Capsicum annuum L. var. annuum), and wild tomato (Nicaise et al., 2003; Ruffel et al., 2002, 2006). The fundamental role of plant eIF4Es suggests their potential as candidates for targeted editing via CRISPR–Cas. Along these lines, the modification of two independent sites in eIF4E conferred resistance to potyviruses including Cucumber vein yellowing virus (CVYV), Zucchini yellow mosaic virus (ZYMV), and Papaya ringspot mosaic virus-W (PRSV-W) in homozygous mutant cucumber plants (Chandrasekaran et al., 2016). In contrast, heterozygous plants were sensitive to all viruses, as expected, because of the recessive nature of resistance. In a study by Pyott et al. (2016), similar results were obtained for site-specific mutagenesis of the closely related eIF(iso)4E locus in A. thaliana. Plants harboring biallelic 1-bp InDels presented immunity against TuMV, while no developmental alterations with respect to wild-type plants were observed.

Conservation of plant eIF4Es suggests that they may function as susceptible factors for a wide range of viruses. However, because of partial functional redundancy, disrupting a single eIF4E does not always confer virus resistance, and knock-out mutations of both eIF4E and its corresponding isoform eIF(iso)4E result in embryo lethality (Martínez-Silva et al., 2012; Patrick et al., 2014). For this reason, it is necessary to improve the available genetic resources to identify and functionally characterized alternative S genes that enable recessive-resistance-based breeding against a broad spectrum of plant viruses. Results from diverse approaches—including forward and reverse genetics, genome-wide analysis, and screenings of naturally occurring resistant cultivars—have revealed >100 host factors that can be affected by the infection of a single plant virus (Nagy & Pogany, 2012; Hyodo & Okuno, 2014; Wang, 2015). Other factors conforming to the host translational machinery beyond eIF4Es have also been shown to be involved in viral replication. For example, Eukaryotic translation elongation factor 1A (eEF1A) forms ternary complexes with aminoacylated elongator transfer RNAs and recruits them to the ribosome (Andersen et al., 2003). This protein has been found to interact with the viral replicase in TMV (Yamaji et al., 2010) and TBSV (Li et al., 2009), as well as with viral RNA in Turnip yellow mosaic virus (Dreher et al., 1999), TMV (Zeenko et al., 2002), and TBSV (Li et al., 2009). Downregulation of eEIF4A in N. benthamiana reduces TMV accumulation without affecting translation or the number of infection loci (Table 3). Additionally, Poly-A binding protein (PABP) interacts with the eIF4F complex to form a protein bridge that brings the 5′ and 3′ ends of the mRNA into proximity to improve translation (Sachs, 2000). Among the eight PABP genes encoded in the A. thaliana genome, PABP2, PABP4, and PABP8 are highly expressed and probably the most important for cellular functions. Within TuMV replication vesicles, the viral RNA-dependent RNA polymerase and VPg interact primarily with PABP2, but also with PABP4 and PABP8 to a lesser extent (Dufresne et al., 2008) (Table 3). However, because both eEF1A and PABP are also involved in essential processes such as plant growth, gene expression, and hormone signaling, mutations in these genes may negatively affect plant performance (Abdul-Razzak et al., 2009). Therefore, when seeking disease resistance, it is imperative to select host factors that can be modified without any adverse effects on plant development.

DEAD-box RNA helicases (DDXs) are a large family of proteins involved in almost every step of RNA metabolism (Cordin et al., 2006). Regarding viral infection, they participate in viral RNA synthesis, selection, and recruitment of the RNA template for translation, ribosome biogenesis, and assembly of the RNP complex for RNA degradation. As previously described for eIF4Es, the potyviral VPg was found to interact with an A. thaliana RNA helicase (AtRH8) and a Prunus persica DDX-like protein (Huang et al., 2010) (Table 3). These two DDXs are not required for plant growth and development, although they are necessary for infection by potyviruses. Arabidopsis thaliana RNA helicase colocalized with TuMV, TEV, and Plum pox virus (PPV) replication vesicles, thus suggesting its possible role in viral translation and replication. Moreover, the overexpression of the VPg-binding region of either AtRH8 or Prunus persica DDX-like protein results in a significant reduction of viral accumulation. Subsequent studies on the role of other DDX members have confirmed that AtRH2 and AtRH5 are components of tombusvirus replication particles, whereas AtRH20 stimulates Tomato bushy stunt virus (TBSV) RNA synthesis (Kovalev & Nagy, 2014).

DNA-binding protein phosphatases (DBPs) form a unique family of protein phosphatases that are distinctively capable of binding DNA. Tobacco DBP1, the first-characterized member of this family, has been shown to be involved in the transcriptional regulation of a defense-related gene during compatible plant–virus interactions (Carrasco et al., 2005). In A. thaliana, knock-out of DNA-binding protein phosphatase1 (DBP1) conferred resistance to potyviruses TuMV and PPV without altering plant growth in a study by Castelló et al. (2010) (Table 3). DNA-binding protein phosphatase1 was found to form a stabilizing interaction with eIF(iso)4E, which could lead to a reduction in eIF(iso)4E levels and the consequent loss-of-susceptibility phenotype. Moreover, it has been proposed that DBP1 interacts with the 14-3-3 family protein GRF6, thereby altering the phosphorylation status of eIF(iso)4E and leading to complete resistance to PPV in grf6 mutant plants (Khan & Goss, 2004; Carrasco et al., 2014).

Plant viruses move cell to cell through plasmodesmata by using movement proteins that increase the plasmodesmata size exclusion limit and form microtubules (Taliansky et al., 2008). Although some viruses remain confined to the vascular system, most of them tend to infect distal tissues such as roots and young leaves. Both cell-to-cell and long-distance movement of viruses are mediated by viral proteins and host factors. Plasmodesmata-located proteins (PDL1, PDL2, and PDL3) are known to promote the systemic movement of Grapevine fanleaf virus and Cauliflower mosaic virus by interplaying with viral movement proteins. A mutational analysis in which such interactions were disrupted found considerably reduced microtubule formation and delayed onset of viral infection, resulting in milder symptoms in host plants compared with wild-type individuals (Amari et al., 2010).

Systemic movement of potyviruses through xylem and phloem does not rely on microtubules, being based primarily on replication vesicles. Turnip mosaic virus movement is dependent on interactions between P3N-PIPO and 6K2 with the host proteins PCaP1 and SEC24A, respectively, which facilitate intracellular trafficking of viral vesicles. Following this model, TuMV movement has been found to be inefficient in pcap1 and sec24a knock-out plants, leading to reduced virus accumulation, a lack of systemic movement, and mild symptoms (Vijayapalani et al., 2012; Jiang et al., 2015).

CONCLUDING REMARKS

Their simplicity, versatility, and robustness in generating targeted genome modifications suggest CRISPR–Cas systems as powerful tools for advancing basic plant biology and crop improvement. Using viral vectors as delivery vehicles for CRISPR–Cas reaction components significantly improves editing efficiency and eliminates the tedious backcrossing process when the whole Cas nuclease-sgRNA complex is transiently expressed via an RNA virus. Furthermore, various studies have demonstrated the absence of the virus in the progeny of infected plants unless vegetatively propagated (Ali et al., 2015a; Ellison et al., 2020; Ariga et al., 2020; Lei et al., 2021; Li et al., 2021; Uranga et al., 2021a), thus avoiding regulatory concerns.

An efficient transfer of editing technologies from the bench to the field requires reconsideration of certain aspects. Since proof of concept in most VIGE studies is based on model plant species such as N. benthamiana, there is an urgent need to expand these systems across agronomically relevant crops. Fortunately, the growing number of available viral vectors is helping overcome species- or cultivar-specific challenges to genome editing. The current bottleneck stems from the inability of viral vectors to edit the germline, consequently leading to minimal or nonexistent recovery of plants with heritable modifications. Strategies discussed earlier, such as using mobile sgRNAs, significantly increase the probability of obtaining mutated progeny (Ellison et al., 2020; Lei et al., 2021; Li et al., 2021; Uranga, et al., 2021a). However, further research is needed to obtain heritable edits before VIGE gains popularity over traditional transformation techniques involving time-consuming tissue-culture procedures.

Genome editing technologies can provide molecular immunity against devastating plant viruses by altering regions of the viral genome involved in viral replication or host genes that confer susceptibility to the pathogen. Additionally, by specifically disrupting key genes involved in plant–pathogen interactions, it is possible to unravel the molecular mechanisms of pathogenesis, such as signaling molecules or receptor proteins. The rapid expression of virus-delivered CRISPR–Cas reaction components enables identification of the phenotypic effects associated with gene editing prior to seed set and plant selection. In this way, it is possible to identify appropriate gene targets and optimize sgRNA sequences. Like the well-established virus-induced gene silencing, it seems reasonable to affirm that VIGE approaches can also be employed to study plant genetic responses to pathogens.

In conclusion, viral vectors carry enormous potential for CRISPR–Cas-mediated genome editing in plants. We envision growing popularity for viral delivery systems that will expand their use to many laboratories, thus contributing to future advances in plant functional genomics and agricultural biotechnology.

ACKNOWLEDGMENTS

This work was supported by the Ministerio de Ciencia e Innovación (Spain) grant PID2020-114691RB-I00 through the Agencia Estatal de Investigación (cofinanced European Regional Development Fund). Mireia Uranga was supported by a predoctoral contract (FPU17/05503) from the Ministerio de Ciencia, Innovación y Universidades (Spain)

AUTHOR CONTRIBUTIONS

Mireia Uranga: Conceptualization; Investigation; Writing – original draft; Writing – review & editing. José-Antonio Daròs: Conceptualization; Funding acquisition; Investigation; Writing – original draft; Writing – review & editing.

CONFLICT OF INTEREST

The authors declare no conflict of interest.