Targeted mutagenesis with sequence-specific nucleases for accelerated improvement of polyploid crops: Progress, challenges, and prospects

Abstract

Many of the world's most important crops are polyploid. The presence of more than two sets of chromosomes within their nuclei and frequently aberrant reproductive biology in polyploids present obstacles to conventional breeding. The presence of a larger number of homoeologous copies of each gene makes random mutation breeding a daunting task for polyploids. Genome editing has revolutionized improvement of polyploid crops as multiple gene copies and/or alleles can be edited simultaneously while preserving the key attributes of elite cultivars. Most genome-editing platforms employ sequence-specific nucleases (SSNs) to generate DNA double-stranded breaks at their target gene. Such DNA breaks are typically repaired via the error-prone nonhomologous end-joining process, which often leads to frame shift mutations, causing loss of gene function. Genome editing has enhanced the disease resistance, yield components, and end-use quality of polyploid crops. However, identification of candidate targets, genotyping, and requirement of high mutagenesis efficiency remain bottlenecks for targeted mutagenesis in polyploids. In this review, we will survey the tremendous progress of SSN-mediated targeted mutagenesis in polyploid crop improvement, discuss its challenges, and identify optimizations needed to sustain further progress.

Abbreviations
• ADH
• alcohol dehydrogenase
• CAPS
• cleaved amplified polymorphic sequences
• Cas
• CRISPR-associated
• CE
• capillary electrophoresis
• CPPs
• cell-penetrating peptides
• CRISPR
• clustered regularly interspaced short palindromic repeat
• DHI
• double haploid inducer
• DSBs
• double-stranded breaks
• MSNs
• mesoporous silica nanoparticles
• PAM
• protospacer adjacent motif
• PCR
• polymerase chain reaction
• QTL
• quantitative trait locus
• RNAi
• RNA interference
• RNP
• ribonucleoprotein
• sgRNA
• single-guide RNA
• SSNs
• sequence-specific nucleases
• SWCNTs
• single-walled carbon nanotubes
• TALENs
• transcription activator-like effector nucleases
• UTR
• untranslated region
• ZFNs
• zinc-finger nucleases

INTRODUCTION

Climate change and population growth pose monumental challenges to food security (Clarke & Zhang, 2013). Plant breeding has made significant contributions to the improved productivity of crops. However, the rate of genetic improvement must double to meet future demands (Voss-Fels et al., 2019). Genome editing, genomic selection, advances in sequencing technologies, high-throughput phenotyping, and novel speed breeding techniques should significantly accelerate conventional breeding of high-yielding and stress-tolerant cultivars (Hickey et al., 2019).

Genome-editing techniques using sequence-specific nucleases (SSNs) were developed and refined over the last three decades. SSNs are now being widely deployed as powerful plant breeding tools. Targeted mutagenesis technologies that have been applied in plants include meganucleases, zinc-finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs), and the clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR-associated protein (Cas) system. Genome editing also leverages genomic data, including causal mutations and homologous sequences, but uses genetic transformation technologies to install desired changes to DNA sequence in a genetic background of interest. In canonical genome editing, nucleases are deployed to cellular nuclei and cleave DNA at desired chromosomal locations, creating double-stranded breaks (DSBs). Lack of a homologous DNA repair template frequently results in imprecise DNA repair, following nonhomologous end joining. Insertions or deletions close to the targeted DSB typically lead to a shift in the reading frame causing loss of gene function or “knockout.”

Because genome editing can be accomplished in a single generation without the need for crossing or selfing, it is, in theory, superior to marker-assisted breeding methods for introgression of a single or small number of favored alleles (Figure 1A,B). In the context of polyploid plants, this is particularly significant, as many polyploid cultivars are highly heterozygous and clonally propagated; in such cases, full recovery of the genotype of the recurrent parent is impossible using marker-assisted backcross breeding. Genome editing also enables simultaneous co-targeting of multiple functionally redundant homeoalleles that would otherwise assort independently in a conventional breeding scenario. Since the initial report of its application in Arabidopsis and rice (Feng et al., 2013), the CRISPR/Cas platform has dominated plant genome editing. In recent years, modifications to the catalytic activity of Cas enzymes have expanded the scope of potential nucleic acid targets as well as broadened applications of the technology in plants beyond targeted sequence changes.

FIGURE 1. Trait introgression in polyploids using conventional breeding versus gene editing. Gene-editing technologies offer several advantages over conventional breeding methods and marker-assisted backcrossing for trait introgression in polyploid crops. Depicted here are the results of initial crosses between a donor and recipient parent in a conventional breeding scenario (at left) and targeted mutagenesis mediated by SSN (e.g., CRISPR/Cas9; at right) for both an allopolyploid species (A) and an autopolyploid species (B)

This review surveys the key achievements of targeted mutagenesis, or gene knockout, in polyploid plant species enabled by genome editing. Impediments to its wider application as well as successful strategies for generation and identification of efficient multiallelic mutagenesis are examined. Future directions in polyploid crop improvement and synthetic biology are discussed in the context of emerging technologies including third-generation sequencing platforms, novel Cas nucleases, alternative reagent delivery, and directed evolution.

THE IMPORTANCE OF POLYPLOIDY IN GLOBAL AGRICULTURE

The nuclei of polyploid plant cells exceed two complete sets of chromosomes in their euploid form, a condition that is heritable. This phenomenon is common in flowering plants, of which over 70% of species are polyploid or have experienced polyploidization events during the course of their evolutionary histories (Meyers & Levin, 2006). The masking of deleterious alleles through gene redundancy, neofunctionalization of duplicated genes, fixed heterosis, and emergent self-fertilization and asexual reproductive capabilities are among the beneficial consequences of increased ploidy and have contributed to the evolutionary success of polyploid plants (Comai, 2005).

Core ideas

• Genome editing has revolutionized crop improvement.
• Many of the world's most important crops are polyploid.
• Progress and bottlenecks for targeted mutagenesis in polyploids are discussed.
• Approaches for elevation of multi-allelic editing efficiency in polyploids are discussed.

Two forms of polyploidy are observed among plants. Autopolyploid species such as alfalfa (Medicago sativa L.; 2n = 4x = 32) and potato (Solanum tuberosum L.; 2n = 4x = 48) contain multiple sets of chromosomes originating from the same taxon and exhibit polysomic inheritance in sexual progeny. Species categorized as allopolyploid, including peanut (Arachis hypogaea L.; 2n = 4x = 40), bread wheat (Triticum aestivum L.; 2n = 6x = 42), and sugarcane (Saccharum spp. hybrid; 2n = 10–13x = 100–130), are the consequence of interspecific hybridizations either preceded or followed by fusions of unreduced gametes. Allopolyploidy results in multiple sets of chromosomes originating from different species (homeologous chromosomes belonging to sub-genomes) that do not pair during prophase I of meiosis.

Polyploidization is theorized to have facilitated the domestication of crop species due to generation of novel genetic functions, combinations, interactions, and epigenetic changes that resulted in enhanced adaptability (Renny-Byfield & Wendel, 2014; Udall & Wendel, 2006). Phylogenetic analyses have revealed the prevalence of polyploidization among domesticated species compared to their wild relatives, as well as a tendency for polyploidy events to precede domestication (Salman-Minkov et al., 2016), lending support to this hypothesis.

The positive impacts of polyploidy on plant vigor and quality are well-documented. Comparisons of polyploid and diploid hybrids of alfalfa revealed stronger heterosis in polyploids leading to improved trait performance, increased leaf size, and higher forage yield (Bingham et al., 1994; Kidwell et al., 1994). In addition, packaging the increased amounts of DNA within polyploid plant cells can cause increases in cell size (Melaragno et al., 1993). There is also evidence of a higher photosynthetic rate per cell in polyploid leaves compared to the leaves of diploids (Coate et al., 2012; Warner & Edwards, 1993). Increased ploidy has also been associated with significant improvements in fiber quality in allotetraploid cotton (Jiang et al., 1998) as well as increased diversity in grain hardness in tetraploid and hexaploid wheats (Chantret et al., 2005).

Not surprisingly, polyploid species are among the world's most important food, fiber, forage, oilseed, and biofuel crops. Six polyploid species, including sugarcane, wheat, potato, banana (Musa spp.; 2n = 3x = 33), sweet potato (Ipomoea batatas [L.] Lam.; 2n = 6x = 90), and cotton (Gossypium hirsutum L.; 2n = 4x = 52), rank among the top 20 crops worldwide in terms of gross production, representing a combined 3.4 billion tonnes of annual production valued at nearly $500 billion (FAO, 2021).

CHALLENGES AND OPPORTUNITIES FOR GENOME EDITING IN POLYPLOIDS

To date, genome-editing techniques have been applied in over 20 polyploid plant species and interspecific hybrids including cereal and horticultural crops, forages, fiber crops, oilseeds, ornamental plants, and turf and bioenergy grasses (Tables S1 and S2). Despite these achievements, genome editing has yet to reach its full potential for polyploid crop improvement. The highly complex and repetitive nature of polyploid genomes poses significant challenges for computational biology in terms of phasing, annotation, full chromosome assembly, and differentiation between homologs and homeologs during generation of polyploid genome sequences (Kyriakidou et al., 2018). Chromosomal rearrangements and epigenetic shifts during the evolution of polyploid species have led to global transcriptome changes including activation of transposable elements, neo- and subfunctionalization of duplicated genes, and biased expression of homeologs leading to expression-level dominance (Wendel et al., 2018). These factors present difficulties for connecting genotype to phenotype in polyploid plants (Bourke et al., 2018) as well as for molecular characterization of the edited genes or genomes.

Furthermore, all or a large number of functionally redundant alleles or copies must be successfully co-edited to generate a mutant phenotype in polyploids. Efficient co-editing requires refined genome editing reagents and protocols. Apomixis or vegetative propagation exhibited by many polyploids prevents transgene removal from the edited plant's genome through Mendelian segregation. However, removal of editing tools following targeted mutagenesis is required for release of nonregulated cultivars in several regulatory frameworks (Thygesen, 2019; USDA-APHIS, 2020; Viera et al., 2021). More efficient, higher throughput, and nonintegrative transformation platforms will accelerate the development and commercialization of edited polyploid crop cultivars (Altpeter et al., 2016; Ghogare et al., 2021; Gordon-Kamm et al., 2019).

While multiallelic editing does pose several challenges, the larger numbers of gene copies present in polyploid genomes offer opportunities for obtaining editing outcomes that are not possible in diploid systems. The presence of multiple functional alleles in polyploid plants enables generation of a range of phenotypes by targeting partial loss of expression across hom(e)ologous loci (Eid et al., 2021). This also facilitates targeted manipulation of genes for which complete loss-of-function variants result in impaired plant performance in diploids. Furthermore, the buffering effect against negative mutations and potential for high biomass production make polyploid plants a promising platform for the application of genome editing-mediated directed evolution, either as heterologous hosts or for evolved function of native genes.

GENOME-EDITING TECHNOLOGIES FOR TARGETED MUTAGENESIS APPLIED IN POLYPLOIDS Protein–DNA interaction-based methods

Meganucleases, ZFNs, and TALENs rely on protein–DNA interactions for sequence recognition and cleavage. Meganucleases were the first genome-editing tools to be applied in plants (Puchta et al., 1996). Meganucleases and ZFN are no longer preferred platforms due to difficulties in re-engineering their targeting activity (Prieto et al., 2007; Ramirez et al., 2008). However, the sizes of these nucleases, approximately 165 amino acids (aa) for meganucleases and 300 aa per ZFN monomer (Baltes & Voytas, 2015), allow for delivery via vector systems that require shorter coding sequences, including viruses (Marton et al., 2010). Meganuclease- and ZFN-mediated gene editing have been reported both for proof-of-concept and trait development in tobacco (Baltes et al., 2014; Chujo et al., 2017; Petolino et al., 2010; Townsend et al., 2009; Yanagawa et al., 2020) and wheat (Bilichak et al., 2020; Cigan et al., 2017). TALENs are the most specific of all the genome-editing tools in terms of site recognition (Mussolino et al., 2014), but their assembly for targeted edits is far more time consuming compared to CRISPR/Cas, and their large size (∼950 aa per monomer) limits multiplexed gene editing (Baltes & Voytas, 2015). Published reports of targeted mutagenesis using TALENs are available for sugarcane, peanut, canola, tobacco, wheat, and potato (Tables S1 and S2).

CRISPR/Cas system

CRISPR/Cas-mediated genome editing is dependent upon base pairing between a single-guide RNA (sgRNA) molecule and a 20- to 24-nucleotide “spacer” neighboring a protospacer adjacent motif (PAM), which varies in sequence content and length depending on the form of Cas nuclease used (Anzalone et al., 2020). The short length of CRISPR/Cas target sequences can facilitate co-targeting of short, conserved stretches of sequence across multiple alleles in polyploid genomes. Class 2 bacterial CRISPR immune systems have been adapted for research applications because they utilize single large proteins to cleave nucleic acids (Makarova et al., 2020). These are further classified into types II, V, and VI: Cas9 (type-II) and Cas12 (type-V) proteins are RNA-guided, DNA-targeting endonucleases, while type-IV Cas13 proteins are RNA-guided and target RNA. Cas13 variants have been deployed in Nicotiana benthamiana, sweet potato, and potato for conferring resistance to RNA viruses (Cao et al., 2021; Mahas et al., 2019; Yu et al., 2022; Zhan et al., 2019; Zhang, Zhao, et al., 2019).

Cas9 nucleases

Most reports of genome editing in polyploid plant species have used the 1368-aa Cas9 nuclease from Streptococcus pyogenes (SpCas9) (Tables S1 and S2), which generates a blunt DSB 3 bp upstream of a 5′-NGG-3′ PAM (Doudna & Charpentier, 2014). Cas9 sgRNAs are an artificially engineered fusion of CRISPR RNA (crRNA) and trans-activating CRISPR RNA (tracrRNA) into one continuous sequence (Jinek et al., 2012). Off-target activity presents a challenge because Cas9 will cleave DNA targets with up to three mismatches in base pairing between the sgRNA and spacer (Zhu et al., 2017). This can be mitigated to some degree through selection of unique target sequences, given the availability of genomic data. Mismatches in the eight to 12 bases preceding the PAM (i.e., the “seed sequence”) reduce sgRNA–Cas9 complex affinity for the target region (Doench et al., 2014; Liu et al., 2016).

Modifications to Cas expression cassettes can dramatically influence editing outcomes through Cas protein abundance or altered catalytic activity (Hassan et al., 2021; Mattiello et al., 2022). Mutations in the HNH or RuvC nuclease domains (thereby forming nCas9-H840A/D10A) result in single-stranded cleavage or “nickase” activity. When used with pairs of sgRNAs to induce DSBs, nCas9 provides a greater level of specificity to the target region, but at considerable cost to editing efficiency (Fauser et al., 2014; Mikami et al., 2016). Variants of SpCas9 have also been engineered with less stringent PAM requirements and lower off-target activity (Collias & Beisel, 2021; Karvelis et al., 2017).

Cas9 from Staphylococcus aureus (SaCas9) is smaller (1053 aa) than SpCas9, enabling nonintegrative viral delivery of editing reagents (Friedland et al., 2015; Ran et al., 2015), and recognizes a 5′-NGRRT-3′ PAM (where R represents A or G), thereby expanding the number of potential genomic targets. Targeted mutagenesis mediated by SaCas9 has been reported for tobacco (Hsu et al., 2019; Kaya et al., 2016, 2017) and potato (Veillet et al., 2020).

Cas12 nucleases

Among type-V CRISPR nucleases, Cas12a (formerly known as Cpf1, for CRISPR from Prevotella and Francisella 1) has been the most widely used. Cas12a is guided by a single crRNA, recognizes the T-rich PAM sequences 5′-TTN-3′ (FnCas12a) or 5′-TTTV-3′ (LbCas12a and AsCas12a), and its DNA cleavage occurs 18 and 23 base pairs from the 3′ end of the PAM (Zetsche et al., 2015), which generates larger deletion mutations than typically observed with Cas9 with a higher likelihood of eliminating gene function (Bernabé-Orts et al., 2019). Cas12a also possesses RNase activity and is capable of processing its own crRNAs (Zetsche et al., 2017), and exhibits lower off-target activity than Cas9 (Kim et al., 2016; Tang et al., 2018). Cas12a nucleases from Lachnospiraceae bacterium (LbCas12a), Francisella novicida (FnCas12a), and Acidaminococcus (AsCas12a) and Cas12b from Alicyclobacillus acidoterrestris (AacCas12b) have been used for targeted mutagenesis in canola (Sidorov et al., 2022), cotton (Li et al., 2019; Li, Liang, et al., 2021), Nicotiana benthamiana (Bernabé-Orts et al., 2019; Calvache et al., 2022; Uranga, Vazquez-Vilar, et al., 2021), tobacco (Endo et al., 2016; Hsu et al., 2019; Vazquez-Vilar et al., 2021), banana (Wu et al., 2020), and wheat (Kim et al., 2021; Liu, Wang, et al., 2020; Wang, Tian, et al., 2021). An improved Cas12a with eight introns exhibited enhanced editing efficiency in canola according to a recent report (Lawrenson et al., 2022). AacCas12b (formerly C2c1) derived from Alicyclobacillus acidoterrestris is guided by a chimeric sgRNA formed by crRNA and tracrRNA, and requires a higher optimal temperature for efficient cleavage (40–55 °C) (Liu et al., 2017; Yang et al., 2016). One group of researchers capitalized on the heat inducibility of the CRISPR/AacCas12b system to achieve efficient multiallelic mutagenesis in cotton (Wang et al., 2020).

TRAIT DEVELOPMENT IN POLYPLOID CROPS ENABLED BY TARGETED MUTAGENESIS

Targeted mutagenesis outcomes in polyploids allow identification of phenotypes with the desired level of target gene suppression, similar to RNA interference (RNAi). However, gene knockouts generated by SSNs are an improvement over RNAi “knockdowns” because sustained expression of a transgene is not required to maintain partial or complete loss-of-function. The following section summarizes the current progress in polyploid trait development through SSN-mediated gene knockout. Targeted mutagenesis has been used in polyploids to eliminate or reduce function of genes involved in susceptibility to pathogens, yield components, and end-use quality (Table 1). These approaches have resulted in, and will continue to inform, accelerated production of higher yielding, higher value polyploid cultivars.

TABLE 1 Highlights of trait development enabled by targeted mutagenesis in polyploid crops

Species	Trait category	Gene target(s)	Nuclease used	Nuclease/gRNA cassette modifications	Phenotype(s) obtained	Mutagenesis efficiencya (%)	Co-editing range in T0b (%)	Reference(s)
Brassica napus	Increased yield	INDEHISCENT (BnaIND); ALCATRAZ (BnaALC)	SpCas9	Cas9: CaMV 35s promoter, codon-optimized; gRNA: AtUbi3b/d and AtUbi6-1/6-29 promoters	Pod shatter resistance	BnaIND: 4 (8/202); BnaALC: 1 (2/206)	BnaIND: 25–100; BnaALC: 25–75	Zhai et al. (2019)
	Increased yield	MORE AXILLARY GROWTH 1 (BnaMAX1)	SpCas9	Cas9: CaMV 35s promoter, Nos terminator; gRNA: AtUbi6-26 promoter	Semi-dwarf; increased branching, silique number, and yield	Site 1: 72 (101/141); Site 2: 69 (82/119)	25–100	Zheng et al. (2020)
	Disease resistance	CYTOCHROME B-C1 COMPLEX SUBUNIT 8 (BnQCR8)	SpCas9	Cas9: AtUbi10 promoter, Nos terminator; gRNA: AtUbi10 promoter	Resistance to white mold (Sclerotinia sclerotiorum) and gray mold (Botrytis cinerea)	61 (14/23)	6.25–75	Zhang, Cheng, et al. (2021)
Musa × paradisiaca	Disease resistance	ENDOGENOUS BANANA STREAK VIRUS (eBSV-B)	SpCas9	Cas9: A. thaliana codon-optimized, PcUb4-2 promoter, pea3A terminator; gRNA: OsU6 promoter	Inactivation of eBSV in B subgenome	95 (19/20)c	NA	Tripathi et al. (2019)
Nicotiana benthamiana	End-use quality	β(1,2)-XYLOSYLTRANSFERASE (NbXylT); α(1,3)-FUCOSYLTRANSFERASE (NbFucT)	SpCas9	Cas9: Plant codon-optimized, CaMV 35s promoter; gRNA: AtUbi6 promoter	Absence of xylose and fucose in recombinant proteins	NbXylT: 80 (43/54); NbFucT: 83 (60/73); XylT+FucT: 73 (54/74)	NbXylT: ≤100; NbFucT: 75–100	Jansing et al. (2019)
Oryza alta	Increased yield	SHATTERING QTL-1 (OaqSH1-C, -D); AWN-1 (OaAn-1-C, -D); SEMIDWARF (OaSd-1-C, -D); GRAIN SIZE 3 (OaGS3-C, -D); GRAIN NUMBER, PLANT HEIGHT, AND HEADING DATE7 (OaGhd7-C, -D); DAYS TO HEADING7 (OaDTH7-C, -D); GRAIN WIDTH AND WEIGHT 5 (OsGW5-C, -D); GRAIN NUMBER 1a (OsGn1a-C, -D); BLACK HULL4 (OaBh4-C, -D); WAXY (OaWx-C, -D); TILLER ANGLE CONTROL 1 (OaTAC1-C, -D)	SpCas9	Cas9: Plant codon-optimized, ZmUbi promoter, Nos terminator; gRNA: OsU6 promoter	Reduced shattering, awn length, and plant height; increased grain length; altered daylength sensitivity; increased grain width and grain number; white seed hull; reduced amylose content and tiller angle	NR	≤100	Yu et al. (2021)
Saccharum spp. hybrid	End-use quality	CAFFEIC ACID O-METHYL TRANSFERASE (COMT)	TALEN	CmYLCV promoter; AtHSP or NtHSP 3′UTR	Reduced lignin content; improved saccharification efficiency and ethanol production	10 (4/39)	8–99	Jung and Altpeter (2016); Kannan et al. (2018); Ko et al. (2018)
Solanum tuberosum	End-use quality	VACUOLAR INVERTASE (StVInv)	TALEN	Nos promoter; Nos terminator	Improved cold storage; reduced acrylamide levels	3 (18/600)	25–100	Clasen et al. (2016)
	Disease resistance	DOWNY MILDEW RESISTANCE 6-1 (StDMR6-1); BASIC HELIX-LOOP-HELIX TRANSCRIPTION FACTOR 7 (bHLH7/StCHL1)	SpCas9	Cas9: A. thaliana codon-optimized, CaMV 35s promoter, HSP terminator; gRNA: CmYLCV promoter	Resistance to late blight (Phytophthora infestans)	StDMR6-1: 12 (147/1271); StCHL1: 75 (166/221)	≤100	Kieu et al. (2021)
Triticum aestivum	Disease resistance	MILDEW-RESISTANCE LOCUS (TaMLO1)	TALEN	ZmUbi-1 promoter; Nos terminator	Resistance to powdery mildew	5 (35/687)	17–50	Wang et al. (2014)
	Disease resistance	HISTIDINE-RICH CALCIUM-BINDING PROTEIN (TaHRC-B)	SpCas9	Cas9: Plant codon optimized, ZmUbi1 promoter; gRNA:TaUbi6 promoter	Resistance to Fusarium head blight (Fusarium graminearum)	NR	50	Su et al. (2019)
	Increased yield	GRAIN WIDTH AND WEIGHT 2 (TaGW2)	SpCas9	Cas9: Maize and wheat codon-optimized, ZmUbi promoter, 2xNLS; gRNA: TaU3 promoter, gRNA–tRNA multiplexing	Increased grain size and weight	Monocistronic: 35 (6/17); Polycistronic: 27 (6/22)	17–100	Wang, Pan, et al. (2018)
	End-use quality	α-Gliadin gene family	SpCas9	Cas9: Wheat codon-optimized, ZmUbi1 promoter, OCS terminator, 2xNLS; gRNA: TaU6 promoter	Reduced gluten content	NR	NR	Sánchez-León et al. (2018)

Abbreviations: NA, not applicable; NR, not reported; SpCas9, Cas9 from Streptococcus pyogenes; TALEN, transcription activator-like effector nuclease.

a

Calculated as (total number mutant plants / total number transgenic plants) × 100%; values in parentheses indicate the number of mutants and number of transgenic plants.

b

Calculated as (number edited copies/alleles / total number copies/alleles) × 100% and presented as a range from lowest to highest observed.

c

Calculated from subset of transgenic lines.

Disease resistance

Plant pathogens cause significant reductions in global crop yields, including annual losses of 21.5% and 17.2% in polyploid wheat and potato, with climate change only expected to exacerbate plant disease outbreaks (Newbery et al., 2016; Savary et al., 2019). Genome editing has enabled production of disease-resistant lines in six polyploid species (Tables S1 and S2) and will continue to be a major asset in the battle for global food security. Two strategies have predominated for generating disease-resistant crops via loss-of-function mutations. The first tactic involves mutating the pathogen's genome, as described earlier for CRISPR/Cas13- mediated resistance to RNA viruses (Aman et al., 2018). Cas9 has also been used in Nicotiana benthamiana (Ali, Abulfaraj, et al., 2015; Ji et al., 2015; Mehta et al., 2019; Roy et al., 2019; Yin et al., 2019; Zhang, Zheng, et al., 2018) to develop virus-resistant plants, primarily through targeted disruptions of the intergenic region of DNA virus genomes, which contain critical viral replication sequence elements (Zaidi et al., 2016).

The second approach entails knockout of disease susceptibility genes (or S-genes) in the host, including salicylic acid, jasmonic acid, and brassinosteroid signaling genes, as well as eIF4E translation initiation factors and mycotoxin-responsive genes. The earliest report of co-targeting S-genes in a polyploid crop described TALEN-mediated mutagenesis of the A-, B-, and D-subgenome homeoalleles of MILDEW-RESISTANCE LOCUS (TaMLO) to generate powdery mildew (Blumeria graminis f. sp. tritici; Bgt) resistance in wheat. Only TaMLO aabbdd triple mutants obtained in the T2 generation exhibited significant reductions in mildew microcolonies compared to wild type and did not support growth of three different races of Bgt (Wang et al., 2014). CRISPR/Cas9 was used to target the homeoalleles of ENHANCED DISEASE RESISTANCE 1 (TaEDR) in wheat and generate transgene-free homozygous T1 Bgt-resistant mutants (Zhang et al., 2017). So far, a single report describes disease resistance conferred by transgene-free, ribonucleoprotein (RNP)-mediated targeted mutagenesis of a S-gene in a polyploid crop (Makhotenko et al., 2019). The knockout of multiple alleles of COILIN, a scaffolding protein involved in formation of Cajal bodies in potato, produced heterozygous plants with resistance to potato virus Y and salt and osmotic stress tolerance (Makhotenko et al., 2019).

A small number of publications report recovery of homozygous disease-resistant mutants in the T0 generation. Co-editing of 14-3-3D protein homeologs (Gh14-3-3d-A, -D) in allotetraploid cotton produced T0 aadd double mutants whose progeny exhibited Verticillium wilt (Verticillium dahliae) resistance both in vitro and in a whole-plant disease screen (Zhang, Ge, et al., 2018). In triploid banana hybrid Sukali Ndizi (Musa × paradisiaca; AAB), tri-allelic mutants of DOWNY MILDEW RESISTANCE 6 (DMR6) were obtained in the T0. A vegetative progeny line showed no Xanthomonas wilt (Xanthomonas campestris) symptoms when challenged by the pathogen (Tripathi et al., 2021).

In some cases, co-mutagenesis of over a dozen alleles including multiple paralogs of S-gene families was required to achieve disease resistance in polyploids. Loss-of-function modifications in the sixteen alleles of eight homologous copies of CYTOCHROME B-C1 COMPLEX SUBUNIT 8 (BnQCR8-A10, -C09, -A03, -A05, -A01, -C01, -C05, -C02) were attempted in order to develop dual resistance to white mold (Sclerotinia sclerotiorum) and gray mold (Botrytis cinerea) in canola. Editing progressed across generations at the BnQCR8 loci, and T2 progeny lines with seven to 15 co-mutated copies of BnQCR8 were found to be resistant to both pathogens. Lines with more mutated copies displayed higher disease resistance (Zhang, Cheng, et al., 2021). Knockout of 15 of the 16 BnQCR8 copies did not impair plant performance, indicating one copy is sufficient for normal biological function (Zhang, Cheng, et al., 2021). Targeted perturbation of two discrete deoxynivalenol-responsive NF-X1-LIKE 1 (TaNFXL1) paralogs was carried out in wheat to develop resistance to Fusarium head blight (FHB). Transgene-free T2 plants displayed up to 11 mutant TaNFXL1 alleles and exhibited reductions in percent infected spikelets (Brauer et al., 2020).

Comparisons of CRISPR knockout mutants across multiple putative S-genes have enabled the identification of useful disease resistance alleles. CRISPR/Cas9-targeting of several putative S-genes in potato, including DOWNY MILDEW RESISTANCE 6 (StDMR6-1,2), DEFENSE NO DEATH 1 (StDND1), MILDEW-RESISTANCE LOCUS (StMLO), HMBPP SYNTHASE (StHDS), TRIPHOSPHATE TUNNEL METALLOENZYME 2 (StTTM2), and BASIC HELIX-LOOP-HELIX TRANSCRIPTION FACTOR 7 (bHLH7/StCHL1), was used to explore their impact on late blight (Phytophthora infestans) resistance. Although editing of six of these genes failed to generate late blight resistance, tetra-allelic knockout lines of StDMR6-1 and StCHL displayed resistant phenotypes with no detectable growth penalty (Kieu et al., 2021). Targeted mutagenesis of WRKY 11 (BnaA.WRKY11.a, BnaC.WRKY11.a) and WRKY 70 (BnaA.WRKY70.a, BnaA.WRKY70.b, BnaC.WRKY70.a, BnaC.WRKY70.b) in canola was evaluated for white mold resistance. Lines with mutated WRKY70 alleles exhibited slightly smaller lesion sizes than control lines. However, edits in WRKY11 did not confer resistance to white mold (Sun et al., 2018).

Other reports have described a more targeted approach in which only a single hom(e)olog is identified as the causal factor and selected for knockout. For example, a nonfunctional B genome homeolog of HISTIDINE-RICH CALCIUM-BINDING PROTEIN (TaHRC-B) was identified as a putative FHB resistance gene in wheat. Homozygous TaHRC-B CRISPR/Cas9 mutants were obtained in T1 through selfing and exhibited significantly lower FHB severity than wild type (Su et al., 2019). Similarly, CRISPR/Cas9-mediated mutagenesis of the S homeolog of tobacco EUKARYOTIC TRANSLATION INITIATION FACTOR 4E (NtEIF4E-S) generated hetero- and homozygous mutants resistant to potato virus Y (Liu, Zeng, et al., 2019). CRISPR/Cas9 was also used to deactivate endogenous banana streak virus (eBSV) found in the plantain (Musa × paradisiaca; AAB) B genome, which triggers production of infectious virions when plants are stressed. The edited plantain lines lacked the infectious episomal form of BSV and were asymptomatic under BSV-activating water stress conditions (Tripathi et al., 2019).

Increased yield

Genes influencing yield components, including those affecting fruit, leaf, and seed morphology, plant architecture, and root traits, have been selected for loss-of-function to increase the yield of polyploid crops. Strategies to facilitate hybrid seed production could improve crop yields through heterosis. Several publications have reported male sterility and haploid induction in wheat by eliminating expression of homeologs of MALE STERILE 26 (TaMS26), MATRILINEAL (TaMTL), NO POLLEN1 (TaNP1), and CENTROMERIC HISTONE 3α (TaCENH3α) (Cigan et al., 2017; Li, Wang, et al., 2020; Liu, Wang, et al., 2020; Lv et al., 2020).

Other examples include efforts to develop shatter-resistant crops by eliminating expression of genes involved in pod dehiscence zone or abscission layer formation. Braatz et al. (2017) mutated all four alleles of ALCATRAZ (BnA.ALC.a, BnC.ALC.a), but only larger sized siliques of the edited lines displayed enhanced shatter resistance. Zhai et al. (2019) found CRISPR mutants of one INDEHISCENT (BnA03.IND) homeolog, which was more highly expressed in the silique wall, had an intermediate level of shatter resistance that was superior to the BnC03.IND single mutant, BnALC double mutant, and wild-type control lines. SHATTERPROOF1 and 2 (SHP1/2) affect expression of both BnALC and BnIND. Co-targeting of SHP1/2 generated a line with biallelic edits at five loci, which exhibited reductions in dehiscence-associated structures in its siliques (Zaman et al., 2021). CRISPR/Cas9 mutagenesis of both homeologs of quantitative trait locus (QTL) of SEED SHATTERING in CHROMOSOME 1 (qSH1) in wild tetraploid rice (Oryza alta) eliminated the abscission layer between grain and pedicel present in wild type plants. This example demonstrates the utility of gene editing for rapid domestication of orphan crops (Yu et al., 2021).

Knockouts of negative genetic regulators of seed size, number, and weight have also been analyzed for their impact on yield. The function of GRAIN WIDTH AND WEIGHT 2, which negatively regulates grain width and thousand grain weight (TGW), has been well-characterized through targeted mutagenesis in wheat. A triple aabbdd gw2 CRISPR mutant generated in an initial study displayed significant increases in grain size and TGW (Wang, Pan, et al., 2018). TaGW2 CRISPR mutant lines in three different genetic backgrounds were used to evaluate the dosage effect of each homeolog. The homeolog with highest expression level exhibited the greatest impact on the grain morphology phenotypes (Wang, Simmonds, et al., 2018; Zhang, Li, et al., 2018). Wang, Pan, et al. (2019) used CRISPR/Cas9 to generate a bbdd double mutant of GRAIN WEIGHT AND WIDTH 7 (TaGW7-A, -B, and -D) with shorter, wider grains and increased TGW. Triple mutants of GIBBERELLIC ACID-STIMULATED REGULATOR 7 (TaGASR7) also exhibited significant increases in TGW compared to wild type in two genetic backgrounds (Zhang et al., 2016). Biallelic knockout of the D homeolog of CYTOKININ OXIDASE/DEHYDROGENASE 2 (TaCKX2-D1) increased grain number and grain weight per spike; the cumulative effect on grain yield was not reported (Zhang, Hua, et al., 2019).

Increasing leaflet number per leaf to elevate leaf-to-stem ratio represents an analogous approach for yield increase in a forage crop. CRISPR/Cas9-mediated disruption of PALMATE-LIKE PENTAFOLIATA1 (PALM1) produced tetra-allelic palm1 alfalfa lines with pentafoliate leaves compared to the trifoliate leaves of wild type alfalfa, although effects on forage yield and quality were not evaluated (Chen et al., 2020). Similarly, increasing tiller number may improve biomass yields in biofuel feedstocks. Knockout of both switchgrass (Panicum virgatum) TEOSINTE BRANCHED1 (PvTb1-A and -B) homeologs generated plants with comparable tiller diameter and height and more than twice the tiller number of the nonedited control (Liu et al., 2018). Consequences of these modifications on biomass yield were not reported so far.

Other changes in plant architecture mediated by growth hormones, including modifications to stature and branch number, may positively impact yield. Plant height decreased and numbers of primary branches and siliques, as well as yield per plant, increased when the strigolactone biosynthesis gene MORE AXILLARY GROWTH 1 (BnaMAX1-A -C) was edited in canola (Zheng et al., 2020). Reductions in leaf inclination angle may enhance yield under high-density plantings. CRISPR mutants of the A subgenome homeolog of BREVIPEDICELLUS (BnA03.BP) in canola and all three homeologs of SQUAMOSA PROMOTER BINDING-LIKE 8 (TaSPL8-A, -B, -D) in wheat reduced branch and leaf angles (Fan et al., 2021; Liu, Cao, et al., 2019). Wheat spl8 mutant lines produced a significantly higher number of spikes compared to wild type under dense planting conditions, but the authors did not provide comparisons of grain yield per plot between wild type and edited lines (Liu, Cao, et al., 2019).

Targeted mutagenesis of the wheat ABNORMAL CYTOKININ RESPONSE1 REPRESSOR1 (TaARE1-A, -B, -D) was undertaken with the aim of improving nitrogen use efficiency. Edited lines displayed superior root growth and higher chlorophyll content under nitrogen starvation, along with delayed senescence and improved grain yield under field conditions. However, the triple mutant showed some defects in spike fertility (Zhang, Zhang, Li, et al., 2021).

End-use quality

Targeted mutagenesis has been applied most extensively for improvement of end-use quality traits in polyploid crops. SSNs have been used to modify the quantity and quality of macro- and micronutrients in harvested plant material from several polyploid species. For example, editing the hom(e)ologs of FATTY ACID DESATURASE in peanut (Wen et al., 2018), canola (Huang, Cui, et al., 2020; Okuzaki et al., 2018), cotton (Chen, Fu, et al., 2021), and allohexaploid camelina (Jiang et al., 2017; Lee et al., 2021; Morineau et al., 2017) elevated the oleic acid content of seed oils. This resulted in higher oxidative stability and nutritional value of the extracted oils. CRISPR-induced mutations in the flavonoid biosynthesis gene TRANSPARENT TESTA8 (TT8) and GDSL lipases SEED FATTY ACID REDUCER4 and 5 (BnSFAR4/5) in canola elevated seed oil content without compromising plant performance (Karunarathna et al., 2020; Zhai et al., 2020).

Grain protein content and composition are key determinants of processing, food quality, and allergenicity of cereal crops. Mutagenesis of wheat PUROINDOLINE B (TaPinb-D) increased grain hardness index over four times that of wild type. This effectively converted a soft wheat to a hard wheat (Zhang, Zhang, Gao, et al., 2021). Single (bb), double (bbdd), and triple (aabbdd) mutants of TaGW2 showed enhanced grain and flour protein content and sodium dodecyl sulfate (SDS) sedimentation values compared to wild type (Zhang, Li, et al., 2018). Gene families encoding γ-gliadin and α-gliadin proteins are associated with celiac disease and have been targeted for knockout in polyploid cereals. Sánchez-León et al. (2018) targeted conserved regions of the α-gliadin gene family and obtained bread and durum wheat lines with strong reductions in allergenicity. Jouanin et al. (2019) used six sgRNA sequences to target both α- and γ-gliadins and obtained lines with modified gliadin content. Allergenic α-amylase trypsin inhibitor subunits WTAI-CM3 and WTAI-CM16 were targeted in durum wheat, resulting in multiple CM3-free mutant lines (Camerlengo et al., 2020).

Modifying the amylose or amylopectin content of plant starch results in different end-use properties. High-amylose foods have a lower glycemic index and offer health benefits, while high-amylopectin starches have desirable processing characteristics. The impact of mutating amylopectin biosynthesis gene STARCH-BRANCHING ENZYME (SBE) on amylose content has been explored in polyploid grain and tuber crops. Triple knockout of STARCH-BRANCHING ENZYME IIa (TaSBEIIa-A, -B, -D) in wheat increased amylose content but impaired spike and grain traits (Li, Jiao, et al., 2021). Editing all four alleles of STARCH-BRANCHING ENZYME 3 (StSBE3) in potato reduced the amylose content (Takeuchi et al., 2021). Editing of GRANULE-BOUND STARCH SYNTHASE (GBSS), which directs synthesis of amylose, decreased amylose biosynthesis in tetra-allelic potato mutants (Andersson et al., 2017, 2018; Kusano et al., 2018). In allohexaploid sweet potato (Ipomoea batatas), CRISPR-mediated knockout of IbSBEII raised amylose content while targeted mutagenesis of IbGBSSI resulted in lower amylose (Wang, Wu, et al., 2019). CRISPR/Cas has also been used to increase micronutrient content. Up to sixfold increases in β-carotene were observed in triploid banana when all homologs of LYCOPENE EPSILON-CYCLASE (MaLCYε) were knocked out (Kaur et al., 2020).

Eliminating expression of genes influencing synthesis of undesirable or harmful phytochemicals is another avenue for improving crop quality. Targeted mutagenesis of all polyphenol oxidase (PPO) alleles, which cause postharvest browning, resulted in 69% and 86% reductions in PPO activity in potato and wheat, respectively (González et al., 2020; Zhang, Zhang, Gao, et al., 2021). STEROL SIDE CHAIN REDUCTASE 2 (StSSR2) and 16α-HYDROXYLASE (St16DOX), which are implicated in biosynthesis of harmful steroidal glycoalkaloids (SGAs) in potato, were successfully targeted using TALENs (Sawai et al., 2014; Yasumoto et al., 2019, 2020) and CRISPR/Cas9 technology (Nakayasu et al., 2018; Zheng et al., 2021) to reduce or eliminate SGA content. Using TALENs, Clasen et al. (2016) knocked out all four alleles of VACUOLAR INVERTASE in potato (StVInv) to produce potato chips with 73% less acrylamide than wild type. Editing of all six alleles of ASPARAGINE SYNTHETASE 2 (TaASN2) in wheat reduced acrylamide precursor levels in grain by 90% (Raffan et al., 2021). In cotton, tetra-allelic knockout of PIGMENT GLAND FORMATION (GhPGF) mediated by LbCas12a eliminated formation of glands containing the poisonous metabolite gossypol.

In species used for molecular farming, the β-1,2-xylose and core α-1,3-fucose residues found in the N-glycans of plants can impact the efficacy and immunoreactivity of recombinant proteins when applied in mammals. TALEN- and CRISPR-mediated multiplex mutagenesis of β(1,2)-XYLOSYLTRANSFERASE (XylT) and α(1,3)-FUCOSYLTRANSFERASE (FucT) in Nicotiana benthamiana (Jansing et al., 2019; Li et al., 2016) and tobacco cell lines (Hanania et al., 2017; Mercx et al., 2017) produced mutants lacking xylose and fucose in both endogenous and recombinant proteins. Similarly, Kriechbaum et al. (2020) targeted β-GALACTOSIDASE 1 in Nicotiana benthamiana (NtBGAL1) and obtained mutant plants with increased human-type β1,4-galactosylation of recombinant proteins.

Lignin in cell walls inhibits the sugar release from lignocellulosic biomass and limits biofuel yields in feedstocks, making targeted mutagenesis of lignin biosynthesis genes a viable strategy for enhancing saccharification efficiency and biofuel production. CAFFEIC ACID O-METHYL TRANSFERASE (COMT) catalyzes the methylation of hydroxylated monomeric lignin precursors, which influences the final lignin composition and structure. Targeting the COMT gene in highly polyploid sugarcane (x = 10–13) with a single TALEN pair knocked out 107 of the 109 total COMT copies/alleles. This resulted in altered lignin monomer composition, reduced lignin content, and dramatically improved saccharification efficiency and ethanol production without negatively impacting field performance (Jung & Altpeter, 2016; Kannan et al., 2018; Ko et al., 2018). Mutants with up to 30% reductions in total lignin and significantly improved saccharification efficiency from the lignocellulosic biomass of tetraploid switchgrass were also reported after editing 4-coumarate:coenzyme A ligase 1(Pv4CL1), an enzyme involved in monolignol biosynthesis (Park, Yoo, et al., 2017).

IDENTIFICATION OF GENOMIC TARGETS AND ASSESSMENT OF MULTIALLELIC EDITING IN POLYPLOID CROPS

Gene editing depends upon knowledge of the DNA sequence for a gene of interest. The complexities of polyploid plant genomes present challenges for the identification of target genes (Figure 2). Reference genomes enable direct selection of conserved gRNA sequences and PCR primers for characterization of target loci in nonreference cultivars. Published genomes are available for most of the important polyploid crop species, although not all references are chromosome-level assemblies with haplotype resolution (Table 2).

FIGURE 2. Addressing challenges for genome editing in polyploids: Identification of chromosomal targets and selection of conserved gRNAs. Applying genome-editing techniques in a polyploid species offers several unique challenges compared to diploid systems, including the identification of candidate gene targets and design of highly conserved gRNA sequences. Factors contributing to this challenge include a lag in sequence resource development due to genomic complexity, subgenome dominance leading to biased expression of hom(e)ologs, and difficulty in gRNA selection due to limited sequence conservation across multiple alleles. These obstacles can be addressed through translational genomics approaches, the use of multiomics data to establish precise genotype–phenotype associations, and the use of Cas nucleases with diverse or relaxed PAM requirements

TABLE 2 Recent progress in assembly of polyploid crop reference genomes

Species	Ploidy	Reference genome length(s)	Assembly level(s)	Reference(s)
		(Mb)
Actinidia chinensis	Autotetraploid (2n = 4x = 116)	653	Chromosome	Wu et al. (2019)
Arachis hypogaea	Allotetraploid (2n = 4x = 40)	2700; 2540	Chromosome	Bertioli et al. (2019); Zhuang et al. (2019)
Avena sativa	Allohexaploid (2n = 6x = 42)	11,000	Chromosome	Kamal et al. (2022)
Brassica napus	Allotetraploid (2n = 4x = 38)	922	Chromosome	Chen, Tong, et al. (2021)
Camelina sativa	Allohexaploid (2n = 6x = 40)	641	Chromosome	Kagale et al. (2014)
Dendrocalamus latiflorus Munro	Allohexaploid (2n = 6x = 72)	2737	Chromosome	Zheng et al. (2022)
Fragaria x ananassa	Allooctaploid (2n = 8x = 56)	813	Chromosome	Edger et al. (2019)
Gossypium hirsutum	Allotetraploid (2n = 4x = 52)	2290	Chromosome	Huang, Wu, et al. (2020)
Ipomoea batatas	Allohexaploid (2n = 6x = 90)	836	Contig	Yang et al. (2017)
Medicago sativa	Autotetraploid (2n = 4x = 32)	2738; 816	Chromosome	Chen et al. (2020); Shen et al. (2020)
Nicotiana benthamiana	Allotetraploid (2n = 4x = 38)	2600	Scaffold	Bombarely et al. (2012)
Nicotiana tabacum	Allotetraploid (2n = 4x = 48)	4695	Scaffold	Edwards et al. (2017)
Saccharum spp. hybrid	Variable allopolyploid (2n = 10–13x = 100–130)	382	Contig	Garsmeur et al. (2018)
Solanum tuberosum	Autotetraploid (2n = 4x = 48)	1991–2717; 2670	Chromosome	Hoopes et al. (2022); Wang et al. (2022)
Triticum aestivum	Allohexaploid (2n = 6x = 42)	14,300–14,900	Chromosome and scaffold	Walkowiak et al. (2020)
Triticum durum	Allotetraploid (2n = 4x = 28)	10,450	Chromosome	Maccaferri et al. (2019)
Vaccinium corymbosum	Autotetraploid (2n = 4x = 48)	1680	Chromosome	Colle et al. (2019)

Desired outcomes of targeted mutagenesis in polyploids may involve a range of genotypes from knockouts in single or few to all copies/alleles. The expanding catalog of multiomics data is therefore critical for addressing expression bias and neo- or subfunctionalization of polyploid homologs. Fine mapping of QTLs combined with whole transcriptome analysis can be used to identify causal loss-of-function mutations, which are then reconstructed in the desired genetic background through targeted mutagenesis. Notable examples of this approach include editing of the TaHRC-B and TaSPL8-A, -B, and -D homeoalleles in wheat to confer resistance to Fusarium head blight and upright leaf angle, respectively (Liu, Cao, et al., 2019; Su et al., 2019). Differentially expressed genes (DEGs) have been selected as knockout targets to confirm candidate genes. Exposure of canola and wheat lines to disease or mycotoxin followed by RNA-seq differential expression analysis was used to select candidate genes for developing resistance to white mold and Fusarium graminearum, respectively (Brauer et al., 2020; Sun et al., 2018).

Alternatively, analysis of copy/allele-specific expression may be used for identification of candidate genes with biochemical activity in a tissue of interest. For example, quantitative real-time polymerase chain reaction (qrtPCR) analysis of three copies of 4-coumarate:coenzyme A ligase in switchgrass identified Pv4CL1 as preferentially expressed in stem tissues that exhibit the highest degree of lignification (Park, Yoo, et al., 2017). In the absence of genomic data, sequences of candidate targets from closely related species can be used as a query in homology searches of transcriptome databases.

The molecular characterization of edits in polyploid genomes can also be challenging (Figure 3). Medium-throughput assays involving polymerase chain reaction (PCR) and/or electrophoresis of PCR amplicons for surveying putative mutants for edits can reduce labor and costs associated with genotyping polyploid plant species. Such analyses involve PCR amplification of the genomic target sequence spanning the gRNA(s) and gel or capillary electrophoresis (CE) for visualization of longer or shorter insertions or deletions (indels), respectively. Cleaved amplified polymorphic sequences (CAPS or PCR-RE) assays (Kusano et al., 2018; Wang et al., 2014), T7E1 mismatch assays (Tripathi et al., 2019; Zhang, Hua, et al., 2019), heteroduplex mobility assays (HMA) (Yasumoto et al., 2020), CE (Jung & Altpeter, 2016; Okada et al., 2019), and high-resolution melting (HRM) curve analysis (Camerlengo et al., 2020; Jansing et al., 2019) have been used for identification of indels (Tables S1 and S2). Selection of the appropriate screening technique is dependent upon the number and sequence of gRNAs delivered. Co-delivery of gRNA pairs that may create larger deletions enables identification of mutant lines by gel electrophoresis of target PCR amplicons (Kieu et al., 2021). A CAPS assay relies on restriction recognition sequence overlap with the site of DNA cleavage (e.g., 3 bp upstream of PAM for SpCas9 nuclease). This requirement may restrict the selection of suitable gRNAs due to limited sequence conservation across multiple alleles. In such cases, T7E1 mismatch assays, HMA, CE, or HRM analysis can be used to screen events before more detailed molecular characterization. The single-nucleotide resolution of CE allows identification of indels in a pool of PCR amplicons without requiring a restriction enzyme digest. CE is also semiquantitative in predicting the co-editing of copies/alleles and does not require cloning of PCR amplicons (Jung & Altpeter, 2016). Acid and SDS polyacrylamide gel electrophoresis have also been used for qualitative assessment of protein composition to identify mutant lines (Jouanin et al., 2019; Sánchez-León et al., 2018).

FIGURE 3. Addressing challenges for genome editing in polyploids: Molecular characterization of targeted mutations. Due to the high sequencing depth and longer read lengths required to discriminate the number of edited copies/alleles, labor and cost requirements can be higher when genotyping polyploid mutant lines. PCR-based assays such as CAPS, T7E1, capillary electrophoresis (CE), and high-resolution melting (HRM) analysis provide low-cost, medium-throughput screening methods for identifying putative edited lines prior to deep sequencing. Platforms for genotyping edited polyploid plants by sequencing range from high-throughput, short-read platforms (Illumina) to low-throughput mid-length reads (Sanger) to high-throughput long-read sequencing technologies (Pacific Biosciences SMRT, Oxford Nanopore). Illumina provides high-depth analysis of target regions. CasAnalyzer and HI-TOM software use Illumina sequencing output to calculate the frequency of mutated reads. Long-read sequencing platforms enable precise evaluation of the number of co-mutated copies/alleles by accessing the native single-nucleotide polymorphism (SNP) pattern of each mutated copy/allele

DNA sequencing provides direct evidence of targeted mutagenesis. Sanger sequencing and Illumina high-throughput sequencing have been the most frequently used technologies for characterizing the type and extent of mutations in polyploid crops (Tables S1 and S2). Computational tools such as Tracking of Indels by Decomposition (TIDE) (Brinkman et al., 2014) and Inference of CRISPR Edits (ICE) (Conant et al., 2022) enable quantitative assessment of the extent and type of targeted mutations using Sanger sequencing data for crops with low to moderate level of ploidy. HI-TOM (Liu, Wang, et al., 2019) and Cas-analyzer (Park, Lim, et al., 2017) software provide high-throughput tracking of targeted mutations using next-generation sequencing data.

In cases where quantification of the number of co-edited copies/alleles in a line is required, Sanger sequencing of a large number of cloned target amplicons has been a suitable technique. Sanger sequencing provides a longer read length, which enables more accurate discrimination of copies/allelic variants based on diagnostic allele-specific SNPs, as well as assessment of editing at multiple gRNA targets within a single copy/allele (Eid et al., 2021; González et al., 2020; Li et al., 2017). Third-generation long-read sequencing technologies including Pacific Biosciences Single Molecule Real-Time (SMRT) and Oxford Nanopore enable sequencing of amplicons on the order of several kilobase in length, facilitating the quantification of numbers of mutated copies/alleles in polyploid crops (Curtin et al., 2021).

FACTORS IMPACTING MULTIALLELIC EDITING EFFICIENCY IN POLYPLOIDS

Poor mutagenesis efficiency is a major challenge in polyploid species that often require the simultaneous editing of multiple homoeologous/paralogous gene copies to produce the desired phenotype. Increasing the activity and precision of gene editing technologies depends on a combination of multiple factors and experimental conditions, including but not limited to gRNA design, number of gRNAs used, the type of nuclease used, and the expression of gene editing components (Figure 4). The following section will outline approaches for assay optimization that have been shown to increase multiallelic gene editing efficiency in polyploid crops.

FIGURE 4. Addressing challenges for genome editing in polyploids: refining editing reagents for elevating co-editing of multiple copies/alleles to confer a phenotypic response. In polyploid species, a larger number of functionally redundant copies/alleles must be co-edited to generate a desired phenotype. In vivo methods such as protoplast transformation, transient transformation (Agro-infiltration), and/or visual markers (PDS, MgCh) are useful for assessing the efficacy of gene-editing reagents. Refinement of gRNA and nuclease expression cassettes can significantly improve mutagenesis efficiency. These include transcriptional regulatory factors, such as promoters and introns, as well as elements that influence the translation or activity of the nuclease. Delivery of multiple gRNAs to a single gene target can increase the likelihood of knockouts. Multiplexing strategies range from discrete gRNA transcriptional units (monocistronic) to transcription of multiple gRNAs from a single promoter (polycistronic). Elevated culture incubation temperatures after delivery of CRISPR/Cas reagents have been shown to increase the catalytic activity of the nuclease; optimal culture conditions vary according to the nuclease and species used

Evaluation of editing reagents prior to crop transformation

Cas-mediated DNA cleavage depends on the formation of the gRNA–DNA duplex. The Cas/gRNA complex negotiates the DNA strand for an appropriate PAM sequence, after which target recognition occurs if strong complementarity between the gRNA seed region and target DNA is found. The formation of the RNA–DNA hybrid facilitates conformational changes within the Cas protein that triggers the nuclease activity (Globyte et al., 2019; Jiang & Doudna, 2017; Sternberg et al., 2015; Wu et al., 2014). The careful design, evaluation, and selection of gRNAs are therefore essential for a successful experiment.

Because generation of transgenic lines typically requires several months, endonuclease cleavage assays can characterize the in vitro or in vivo activity of multiple gRNA designs prior to stable transformation. Experimental validation can identify gRNAs with poor activity before practical application (Abe et al., 2019; Howells et al., 2018; Morineau et al., 2017; Okada et al., 2019) and allow the selection of functioning designs (Gao et al., 2017; Ludman et al., 2017; Ye et al., 2020; Zhang et al., 2016, 2017). There are several computational approaches for gRNA design and on/off-target activity prediction (Cui et al., 2018; Wilson et al., 2018); online tools that have been used for prediction of sgRNA efficiency in polyploid crops are listed in Tables S1 and S2. However, in silico prediction tools cannot guarantee the efficacy of a particular gRNA design in living cells. For example, Okada et al. (2019) compared two in silico on-target activity prediction tools, sgRNA Designer (Doench et al., 2014) and WU-CRISPR (Wong et al., 2015), in wheat and found that the high algorithm score provided by sgRNA Designer for one gRNA was not predictive of its in vivo activity.

Guide RNA efficiency can also be evaluated prior to stable transformation in vivo using protoplasts or transient co-expression assays. Abe et al. (2019) designed two gRNAs to target the TaQsd1 gene in wheat, with one of the gRNAs exhibiting low activity in the in vitro transient assay; subsequently, this gRNA also failed to induce mutations, while three out of eight T0 mutants were recovered with multiallelic mutations at the other target site. After testing the activity of several designs in wheat protoplasts and selecting the most efficient sgRNA, Zhang, Ge, et al. (2018) observed edits in all five regenerated T0s, with three containing multiallelic mutations for the target gene TaEDR1. Ludman et al. (2017) also tested the efficiency of the Cas9-sgRNA targeting construct in transient agro-infiltration assays in tobacco leaves before stable transformation and recovered nine mutants out of nine T0 transformants. Another group tested the functionality of their dual gRNA/Cas9 editing constructs in an in vivo transient assay in fruits of a diploid strawberry cultivar and, after stable transformation, obtained five independent octoploid strawberry plants with multiallelic mutations (Martín-Pizarro et al., 2019). By transiently co-expressing candidate gRNAs with Cas9 in Nicotiana benthamiana leaves and selecting the three most efficient, Jansing et al. (2019) surpassed 70% editing efficiency in the T0. Gao et al. (2017) used a high-throughput sgRNA validation method via transient expression in cotton cotyledons and attained a mutagenesis frequency of over 80% after stable transformation. Furthermore, visual marker systems such as mutagenesis of the carotenoid biosynthesis gene PHYTOENE DESATURASE (PDS) or chlorophyll biosynthesis gene MG-PROTOPORPHYRIN IX CHELATASE (MgCH) can provide rapid readouts of editing efficiency following stable transformation and have been applied for proof-of-concept in polyploid species (Eid et al., 2021; Ma et al., 2020; Maher et al., 2020; Ye et al., 2020; Zhang, Wang, et al., 2021).

Alternatively, gRNAs can be deployed that successfully edited their target sites in previous experiments. For example, Li et al. (2019) designed the crRNA used for Cpf1 targeting of the cotton gene GhCLA1 based on a gRNA used by Wang, Zhang, et al. (2018) and reached an 87% editing efficiency at T0, with a very high number of plants exhibiting multiallelic edits. In strawberry, Wilson et al. (2019) successfully used a target sequence for FvPDS gene editing that corresponded to a previously used site in fellow Rosaceae member apple that was tested in an in vitro cleavage assay (Nishitani et al., 2016).

Number of gRNAs delivered

While frameshift mutations can be achieved using a single gRNA, targeting a single gene with multiple gRNAs can generate large deletions between target sites and increase the probability of generating multiallelic loss-of-function mutations. Generally, two approaches have been applied for expressing multiple gRNAs from one construct: separate gRNA cassettes may be designed using individual RNA polymerase promoters (Gao et al., 2015; Shan et al., 2013), or they can be processed from a single long transcript (Hanania et al., 2017; Mercx et al., 2017; Ordon et al., 2017).

Tripathi et al. (2019) achieved a mutation rate of 95% using three gRNAs simultaneously targeting the eBSV gene in triploid plantain. In potato, similar editing efficiencies were achieved using two and three gRNAs targeting the GBSSI gene; however, the three-guide system was more successful in generating tetra-allelic mutations (Kusano et al., 2018). When the editing efficiency of single and dual gRNA vectors targeting IbGBSSI and IbSBEII was compared in sweet potato, a higher mutation efficiency was observed for the dual gRNA system (Wang, Wu, et al., 2019). In canola, Zaman et al. (2019) used three gRNAs to target all five homeologs of pod shattering-associated JAGGED genes and found that 24% of the transgenic lines produced carried multiallelic edits.

A polycistronic system employs a transcriptional unit that is processed into discrete gRNAs by enzymes that cut RNA. Such polycistronic mRNAs can consist of individual gRNAs separated by tRNAs, which are cleaved by RNases P and Z to release the individual gRNAs from the primary transcript (Xie et al., 2015). A different type of linker system, which is recognized by CRISPR/Cas Subtype Ypest (Csy)-type endoribonuclease 4 (Csy4), can also be used to express multiple gRNAs from the same transcriptional unit (Haurwitz et al., 2012).

The Csy4 and tRNA polycistronic expression systems have been reported to be almost twice as effective as monocistronic designs in inducing mutations in several plant species (Čermák et al., 2017). When comparing the tRNA–gRNA arrays with previously used single gRNA-CRISPR/Cas9 constructs in alfalfa, another group found that multiplexing led to a 30-fold increase in genome-editing efficiency (Wolabu et al., 2020). However, Wang, Pan, et al. (2018) reported that the editing efficiency of individual gRNAs from gRNA-multiplexing transcription units were comparable to that of single expression constructs deployed in wheat. Ribozyme processing is another approach for generating multiplexed edits and offers the advantage of expression of Cas and multiple sgRNAs from one RNA polymerase II (Pol II) promoter (Tang et al., 2016). The ribozyme processing system surpassed the editing efficiencies of both monocistronic and tRNA-processing designs in one study in wheat (Li, Zhang, et al., 2021).

Simultaneous co-expression of multiple gRNAs also enables multiplex genome editing, although reduced mutagenesis and co-editing efficiencies have been observed when multiple genes (and their corresponding copies/alleles) were co-targeted in polyploid species. Jansing et al. (2019) designed three and four gRNAs to target NbXylT and NbFucT in tobacco, respectively. When individual genes were targeted with the polycistronic system, the authors observed over 80% editing efficiency for each target. The polycistronic construct carrying all seven gRNAs also induced 73% mutagenesis efficiency; however, none of the transgenic plants carried mutations in all six target sites simultaneously (Jansing et al., 2019). Ordon et al. (2017) designed an sgRNA “transcription unit” that can express up to eight individual guides from a gRNA array designed to generate large deletions in tobacco genes; however, only a single pair of gRNAs generated mutations. Significant differences in editing efficiency were also reported between two gRNAs targeting StSSR2 expressed from the same polycistronic construct in potato (Zheng et al., 2021). The number of genes targeted in an experiment can also be increased by other approaches, such as the pooled-sgRNA assembly method optimized by Ramadan et al. (2021) that enabled the knockout of 112 development-related genes in cotton.

Use of species-specific promoters for optimized expression of gRNAs

A high abundance of gRNA and Cas9 transcripts can increase gene editing efficiency. Therefore, the promoters driving expression of gene editing components have a significant impact on attaining high mutation frequencies. The U3 and U6 RNA polymerase III (Pol III) promoters are small nuclear RNA promoters and are the most used for gRNA expression in plants. Studies from diploid species report increased mutagenesis success by using endogenous promoters to drive gene editing components (Huang et al., 2021; Ordon et al., 2020; Ren, Liu, et al., 2021; Tang et al., 2016; Wang et al., 2016; Yan et al., 2015; Zhang, Zhang, et al., 2022).

The use of native promoter sequences has also enabled higher editing efficiency and extent of co-editing in polyploids (Table 3). When four different Pol III promoters (TaU3, TaU6, Oryza sativa OsU3, and OsU6) were compared in wheat protoplasts, the highest mutation rate was achieved using native TaU3 promoter-driven optimized gRNA scaffolds (Li, Wang, et al., 2020). Andersson et al. (2017) found that the same guide sequence driven by an endogenous promoter compared to AtU6 generated a twofold increase in mutation frequency in tetraploid potato. Johansen et al. (2019) also observed improved CRISPR/Cas9 editing efficacies when using the potato U6 promoters to control expression of the CRISPR component. When an OsU6 promoter was replaced with the native MaU6 promoter in banana, a fourfold increase in MaPDS editing efficiency was observed (Zhang, Wu, et al., 2022). Wang, Zhang, et al. (2018) used a modified CRISPR vector containing the native promoter GhU6.9, facilitating highly efficient genome editing in tetraploid cotton: 75% of the regenerated plants exhibited the visual phenotype associated with disruptions in the GhCLA1 gene, and sequencing revealed a gene editing efficiency between 66.7% and 100% at the four target sites. In a follow-up study, Chen, Fu, et al. (2021) deployed the same vector system and identified mutations in GhFAD2 homeologs in 76% of the recovered transgenic cotton lines. Ramadan et al. (2021) also used this vector to optimize their powerful pooled-gRNA assembly method.

TABLE 3 Studies using native promoters for gRNA expression in polyploid plants

Species	Promoter	Reference(s)
Camelina sativa	CsU3 RNA polymerase III	Morineau et al. (2017)
	CsU6 RNA polymerase III	Morineau et al. (2017)
Fragaria × ananassa	FvU6 RNA polymerase III	Wilson et al. (2019)
Gossypium hirsutum	GhU6.9 RNA polymerase III	Wang, Zhang, et al. (2018)
	GhU6.7 RNA polymerase III	Chen, Fu, et al. (2021); Li et al. (2019); Ramadan et al. (2021)
Musa acuminata	MaU6c RNA polymerase III	Zhang, Wu, et al. (2022)
Solanum tuberosum	StU6 RNA polymerase III	Andersson et al. (2017); Johansen et al. (2019)
Triticum aestivum	TaU6 RNA polymerase III	Cui et al. (2019); Hamada et al. (2018); Howells et al. (2018); Jouanin et al. (2019); Kim et al. (2021); Li, Wang, et al. (2020); Liu, Luo, et al. (2021); Lv et al. (2020); Okada et al. (2019); Rey et al. (2018); Sánchez-León et al. (2018); Shan et al. (2014); Su et al. (2019); Wang, Pan, et al. (2019); Zhang et al. (2017); Zhang, Hua, et al. (2019); Zhang, Zhang, Li, et al. (2021)
Triticum durum	TaU6 RNA polymerase III	Sánchez-León et al. (2018); Zhang et al. (2016)

The use of Pol III promoters from diploid relatives to drive gRNA expression may not always translate successfully to the polyploid context. In strawberry, the U6III promoter from wild diploid Fragaria vesca proved to be superior to the Arabidopsis U6-26 (AtU6-26) promoter in inducing mutations in the diploid “Hawaii 4” cultivar, while in the octoploid “Calypso” cultivar the AtU6-26 promoter was more effective, highlighting the need to evaluate alternative promoters for boosting sgRNA expression in polyploid species (Wilson et al., 2019).

Optimization of Cas expression and activity

A number of CRISPR/Cas toolkits have been made available for genome editing in plants that express the nuclease under different promoters, use various Cas orthologues and variants, employ species-specific codon optimization, and contain additional features (e.g., enhancers, introns, nuclear localization signals [NLSs]) that can directly affect Cas editing efficiency (Čermák et al., 2017; Hahn et al., 2020; Lowder et al., 2015; Ren, Sretenovic, et al., 2021; Xing et al., 2014).

In dicots, the most used promoter for Cas expression is the Cauliflower mosaic virus 35S promoter (CaMV 35S), while in monocots the maize Ubiquitin (ZmUbi) promoter is the most prevalent, followed by CaMV 35S and rice Ubiquitin (OsUbi) promoters. Besides constitutive expression, cell type-specific expression of Cas is also possible. Egg cell-specific Cas9 expression has been achieved in autopolyploid Arabidopsis and hexaploid camelina using the AtEC1.2 promoter (Lee et al., 2021; Ryder et al., 2017), while expression in pollen was accomplished using the GhPLIMP2b and GhMYB24 promoters in cotton (Lei et al., 2021) and three pollen-specific promoters (ZmPRF3, ZmEXPB1, and ZmEXPB2) in wheat (Kelliher et al., 2019). Endogenous promoter sequences have also been used to drive Cas9 expression in polyploid wheat and banana (Brauer et al., 2020; Zhang, Wu, et al., 2022).

Several approaches have led to increased translation and nuclease activity of the Cas protein in polyploid crops. Kusano et al. (2018) established efficient CRISPR/Cas9-mediated editing in potato by employing a translational enhancer from rice OsMac3 mRNA 5′ untranslated region (UTR) called dMac3. When the dMac3 enhancer was cloned upstream of the Cas9 open reading frame, the ratio of mutants increased 2.5-fold compared to enhancer-less or 5′-UTR of alcohol dehydrogenase (ADH)-enhanced Cas9. Co-editing of multiple alleles was identified in 54% of potato transformants containing the dMac3–Cas9 system, while only 7% of transformants containing the ADH enhancer carried mutations in multiple target alleles (Kusano et al., 2018). Michalski et al. (2021) augmented the Cas9 gene with monocot codon-optimized human THREE PRIME REPAIR EXONUCLEASE 2 (TREX2) sequence to improve nuclease activity in hexaploid triticale (x Triticosecale; 2n = 6x = 42) protoplasts and observed increases in mutagenesis efficiency of up to 19-fold for gRNAs expressed from TREX2-enhanced vectors.

Codon preference, NLSs, and introns in the coding sequence of the Cas gene can also impact genome editing efficiency. While the majority of polyploid plant genome editing studies have used plant codon-optimized Cas9, some studies have employed species-specific codon composition to attempt to improve Cas9 translational efficiency in polyploids. Wang, Pan, et al. (2018) compared the gene editing efficiency of wheat and maize codon-optimized Cas9 in wheat and found it to be similar for all gRNA targets, which they explained with their similar codon usage. Zhang, Wu, et al. (2022) reported that use of banana codon-optimized Cas9 increased MaPDS editing efficiency to 1.5 times higher than a previously used Poaceae codon-optimized Cas9.

The reported plant CRISPR vector designs have an NLS fused to the N-, C-, or both the N- and C-termini of the Cas9 to facilitate protein transport into the nucleus. Grützner et al. (2021) demonstrated use of Cas9 containing two NLSs increased gene knockout efficiency by 14% compared to Cas9 with a single NLS in Arabidopsis. When the subcellular localization of the two Cas9 versions was compared in tetraploid Nicotiana benthamiana, the 2xNLS Cas9 was more prevalent in the nucleus, suggesting a single NLS is insufficient for efficient transport of Cas9 to the nucleus. Introns have also been reported to have a positive effect on gene expression in transgenic plants, although the mechanism behind intron-mediated enhancement is not well understood (Laxa, 2017). Many studies have sought to increase Cas9 gene expression using promoters followed by introns from the 5′-UTR of genes with strong constitutive expression. Examples from polyploid species include the use of CaMV 35S promoter with the maize HEAT SHOCK PROTEIN 70 (HSP70) intron (Eid et al., 2021), rice ACTIN promoter with 5′ intron (Jouanin et al., 2019), or maize Ubiquitin1 promoter with the first intron (Budhagatapalli et al., 2020).

Other studies of gene editing in polyploid crops have incorporated introns in the Cas9 coding sequence to enhance editing efficiency. For example, Singh et al. (2018) used maize codon-optimized Cas9 with potato ST-LS1 intron to edit wheat TaMLO genes, and Wilson et al. (2019) used a Cas9 coding sequence with potato IV2 intron to generate edited strawberry plants. Grützner et al. (2021) achieved their highest editing efficiency in Nicotiana benthamiana using a maize codon-optimized Cas9 containing 13 Arabidopsis introns, while none of the primary transformants of the intron-less Cas9 gene displayed the knockout mutant phenotype.

Finally, exposure of CRISPR/Cas-transformed cells to elevated temperatures has been shown to improve editing efficiencies through enhanced Cas nuclease activity. Heat treatments have boosted the rate of CRISPR/Cas-mediated targeted mutagenesis in polyploid cotton (Li, Liang, et al., 2021), sugarcane (Eid et al., 2021), tobacco (Blomme et al., 2022), and wheat (Milner et al., 2020; Tanaka et al., 2022).

PROSPECTS FOR PRODUCTION OF TRANSGENE-FREE EDITED POLYPLOID PLANTS

Another significant challenge for generating edited polyploid cultivars is obtaining transgene-free mutant lines. Removal of the transgenic editing components from the genome of the edited crops is required in several regulatory frameworks to obtain nonregulated status (Tsuda et al., 2019). Asexual reproduction and clonal propagation in polyploid species preclude transgene removal through segregation. Vegetative propagation of polyploid crop varieties is typically required due to their high level of heterozygosity. Favorable allelic combinations found in elite genetic backgrounds are lost if plants are self-pollinated or crossed to nonmodified parental lines for removal of transgenes in recombinant lines by Mendelian segregation. A list of published reports of transgene-free edited polyploid lines can be found in Table 4.

TABLE 4 Generation of T0 transgene-free polyploid plants with targeted mutations

Species	Nuclease used (form)	Delivery method	Gene target(s)	Transgene-free mutagenesis frequencya (%)	Co-editing range in T0b (%)	Reference
Brassica napus	SpCas9 (DNA)	Pollen from canola double haploid inducer	FATTY ACID DESATURASE 2 (FAD2)	6 (2/34)	<100	Li, Sang, et al. (2021)
	LbCas12a (RNP)	PEG-mediated transfection	Not specified	40 (12/30)	≤50–100	Sidorov et al. (2022)
Nicotiana benthamiana	SpCas9 (RNA)	Potato virus X (PVX)	TOBAMOVIRUS MULTIPLICATION 1 (TOM1)	35 (6/17)–50 (7/14)	≤50–100	Ariga et al. (2020)
	SpCas9 (RNA)	Sonchus yellow net rhabdovirus (SYNV)	PHYTOENE DESATURASE (PDS)	93 (28/30)	≤50–100	Ma et al. (2020)
Solanum tuberosum	TALEN (DNA)	PEG-mediated transfection	ACETOLACTATE SYNTHASE (ALS)	10 (2/20)	<100	Nicolia et al. (2015)
	TALEN (DNA)	PEG-mediated transfection	VACUOLAR INVERTASE (VInv)	1 (7/600)	25–100	Clasen et al. (2016)
	SpCas9 (DNA)	PEG-mediated transfection	GRANULE-BOUND STARCH SYNTHASE (GBSS)	>1 (>26/2051)	25–100	Andersson et al. (2017)
	SpCas9 (RNP)	PEG-mediated transfection	GBSS	1 (4/420)–100 (3/3)	25–100	Andersson et al. (2018)
	SpCas9 (RNP)	PEG-mediated transfection	POLYPHENOL OXIDASE 2 (PPO2)	27 (17/64)–39 (11/28)	25–100	González et al. (2020)
Triticum aestivum	SpCas9 (DNA)	Particle bombardment	Total of six different genes	1 (14/1280)–8 (66/800)	17–100	Zhang et al. (2016)
	SpCas9 (RNA)	Particle bombardment	GRAIN WIDTH AND WEIGHT 2 (GW2)	1 (17/1600)	≤33–100	Zhang et al. (2016)
	SpCas9 (RNP)	Particle bombardment	GW2	4 (28/640)	17–33	Liang et al. (2017)
	SpCas9 (DNA)	Particle bombardment	GIBBERELLIC ACID-STIMULATED REGULATOR 7 (GASR7)	5 (11/210)	<100	Hamada et al. (2018)
	SpCas9 (DNA)	Pollen from maize haploid inducer	GRASSY TILLER1 (GT1)	0 (0/61)–1.8 (NR)	33	Kelliher et al. (2019)
	SpCas9 (DNA)	Pollen from maize haploid inducer	BRASSINOSTEROID-INSENSITIVE 1 (BRI1), SEMI-DWARF 1 (SD1)	4 (3/83) - 25 (1/4)	33–67	Budhagatapalli et al. (2020)
Triticum durum	SpCas9 (DNA)	Particle bombardment	GASR7	0.8 (6/800)–0.9 (7/800)	50–100	Zhang et al. (2016)
	SpCas9 (DNA)	Pollen from maize haploid inducer	BRI1, SD1	10 (1/10)–33 (1/3)	50	Budhagatapalli et al. (2020)

Abbreviations: DNA, transiently expressed DNA; LbCas12a, Cas12a from Lachnospiraceae bacterium ND2006; NR, not reported; RNA, transiently expressed RNA; RNP, preassembled Cas-gRNA ribonucleoprotein; SpCas9, Cas9 from Streptococcus pyogenes; TALEN, transcription activator-like effector nuclease.

a

Calculated as (total number transgene-free mutant plants / total number explants, protoplast regenerants, or doubled haploid progeny) × 100%; values in parentheses indicate the number of transgene-free mutants and number of explants, protoplast regenerants, or doubled haploid progeny; presented as range when multiple experiments were conducted; greater/less than symbols indicate analysis was not comprehensive.

b

Calculated as (number edited copies/alleles / total number copies/alleles) × 100% and presented as a range from lowest to highest observed.

Transgene removal is possible through inclusion of site-specific recombination systems such as flippase/flippase recognition target (FLP/FRT) and Cre-lox in vector designs, in which editing components (e.g., sgRNAs, Cas), selectable markers, and an inducibly expressed recombinase (e.g., Cre) are flanked by recombination sites (e.g., lox). This allows for deletion of the region that is flanked by recombination sites after targeted mutagenesis has been achieved in stably transformed cells, for example, prior to plantlet regeneration from callus. Cre/loxP- mediated transgene removal has been accomplished in polyploid strawberry (Schaart et al., 2004) and potato (Kondrák et al., 2006). However, hypermethylation of genes that are flanked by lox sites has been observed, leading to transgene silencing and prevention of transgene deletion (Liu, Long, et al., 2021). Truncation can occur at the ends of transgene cassettes and inactivate recombination sites, as was observed for FRT sequences flanking Flp-CRISPR/Cas9 cassettes that failed to excise in grapevine (Vitis vinifera L.) (Dalla Costa et al., 2020). Furthermore, residual recombination sites following transgene excision are capable of silencing endogenous genes and may still trigger the requirement for regulatory approval.

In addition, complex polyploid genomes convolute whole-genome sequencing to confirm the absence of transgenes or transgene fragments. Therefore, strategies for nonintegrative delivery of gene-editing reagents are strongly preferred over strategies that require transgene removal following their stable integration into the genome. Transient expression of TALENs and Cas9 DNA has generated multiallelic edits in both wheat (Hamada et al., 2018; Zhang et al., 2016) and potato (Nicolia et al., 2015). However, full genome sequencing is required to conclude the absence of DNA integration with the transient DNA expression approach. Direct delivery of genome-editing tools as RNA, as demonstrated by Zhang et al. (2016) in wheat, or as preassembled RNP complexes are therefore the preferred approaches to produce transgene-free edited plants in polyploids. RNP delivery can be approached by way of various strategies originally developed for direct gene transfer including protoplast transformation via polyethylene glycol (PEG), electroporation or lipofection, biolistic transformation, or nanoparticle-mediated transfer (Gong et al., 2021; Metje-Sprink et al., 2019).

In two separate studies, PEG-mediated Cas9-RNP delivery followed by whole plant regeneration from potato protoplasts produced transgene-free tetra-allelic mutants in up to 3% and 28.6% of regenerated shoots (Andersson et al., 2018; González et al., 2020), while 20% of shoots regenerated from canola protoplasts transfected with LbCas12a-RNP exhibited biallelic mutations in the target gene (Sidorov et al., 2022). Use of 40% PEG solution in transfection generally resulted in elevated mutagenesis efficiencies compared to 25% PEG (Andersson et al., 2018; González et al., 2020). Lipofection is an alternative protoplast transformation method that deploys cationic lipids to neutralize the strong negative charge of RNP molecules and facilitates their uptake into negatively charged cellular membranes. Liu, Rudis, et al. (2020) used cationic lipid-mediated delivery of Cas9-RNPs to BY2 tobacco cells and achieved 6% mutagenesis frequency when targeting the RFP reporter gene. However, protoplast regeneration is challenging and the long tissue culture phase is prone to cause somaclonal variation. Among polyploid crops, protoplast regeneration protocols have been described only for banana, canola, cotton, potato, and tobacco (Eeckhaut et al., 2013; Reed & Bargmann, 2021; Yue et al., 2021).

Cationic, cell-penetrating peptides (CPPs) have also been utilized for nonintegrative delivery of nucleic acids and/or proteins to plant calls. Bilichak et al. (2020) used complexes of CPPs and purified ZFN protein to transfect and edit INOSITOL PENTAKISPHOSPHATE KINASE 1 (TaIPK1-A, -B) in wheat microspores and haploid embryos, but only attained mutagenesis in the A subgenome targets. Furthermore, current microspore culture protocols are inefficient and genotype dependent, limiting wider applications of this strategy.

Bombardment of explant tissues with RNP-coated gold particles has been explored in hexaploid bread wheat, which lacks a protoplast regeneration protocol. One notable problem with biolistic delivery of RNPs is that the lack of a selectable marker necessitates laborious screening of numerous plants regenerated in the absence of selective agents. Nevertheless, Liang et al. (2017) successfully edited the B1 and D1 homeoalleles of TaGW2 using Cas9-RNP but the efficiency of co-editing was lower than using transient expression of gene-editing reagents, and the authors failed to obtain edited T0 lines with mutations at all four targeted TaGW2 alleles. However, RNP off-targeting was lower than that in the transient system, which is consistent with other reports of reduced or abolished off-target activity of RNPs in plants (González et al., 2020; Kim et al., 2017).

Because of the impermanent nature of the RNP molecule, modifying the amounts of Cas and sgRNA seems an intuitive approach to ensuring a sufficient level of mutagenesis is achieved within target cells. However, adjusting the amount and/or ratio of transfected Cas and sgRNA has not led to consistent improvements in editing efficiency. Decreasing the molar amounts of CRISPR editing components delivered to human cells resulted in lower off-target activity but came at the cost of reduced on-target activity (Hsu et al., 2013), while increasing the amount of Cas9 or sgRNA delivered to apple and grape protoplasts failed to improve mutagenesis efficiency (Malnoy et al., 2016).

Developing more efficient methods of delivering RNPs should improve their mutagenesis efficiency. Single-walled carbon nanotubes (SWCNTs), mesoporous silica nanoparticles (MSNs), and polymeric nanomaterials offer promising alternative nucleic acid and/or protein delivery methods due to their protective capabilities, large payload capacities, and ability to carry out controlled spatiotemporal release of molecular cargo (Sanzari et al., 2019). For example, when infiltrating GFP DNA-SWCNT conjugates into abraded leaves of polyploid wheat, cotton, and Nicotiana benthamiana, Demirer et al. (2019) demonstrated strong expression of the recombinant protein while avoiding integration of foreign DNA into the nuclear genome. Using MSN technology, Torney et al. (2007) verified controlled release of chemical and DNA cargo using particle bombardment of maize and tobacco explants, while Martin-Ortigosa et al. (2014) delivered functional Cre protein to transgenic maize cells, also by particle bombardment, to remove selectable marker and reporter genes. Although gene editing with RNP-loaded MSNs has not yet been reported in plants, RNPs were loaded in lipid-coated MSNs, delivered, and efficiently released to edit human cell lines (Noureddine et al., 2020), demonstrating sufficient cargo capacity.

The potential of viral nonintegrative delivery of genome-editing reagents has received attention from the plant science community in recent years. In allotetraploid Nicotiana benthamiana, the Tobacco rattle virus, Pea early browning virus, Potato virus X (PVX), and Apple latent spherical virus have been used for nonintegrative sgRNA delivery in transgenic plants expressing Cas9 to generate multiallelic edits (Ali, Abul-Faraj, Li, et al., 2015; Ali, Abul-Faraj, Piatek, et al., 2015; Ali et al., 2018; Luo et al., 2021; Uranga, Aragonés, et al., 2021). A similar approach employing Barley stripe mosaic virus has produced multiallelic edits in infiltrated wheat tissues (Hu et al., 2019) as well as homozygous edited wheat plants (Li, Hu, et al., 2021). Interestingly, Ellison et al. (2020) used fusions of sgRNAs and mobile RNA sequences to access meristem cells and generate germline edits in tetraploid Nicotiana benthamiana plants expressing Cas9.

The shortcomings of the above approaches are the limited cargo capacities of the viral genomes, which in most cases only allows for delivery of the sgRNAs to transgenic plants. Other researchers have addressed this issue through the use of viruses with greater cargo capacities, including sonchus yellow net rhabdovirus (Ma et al., 2020) and PVX (Ariga et al., 2020) to attain multiallelic mutagenesis in Nicotiana benthamiana. Another study used two compatible viral vectors, tobacco etch virus and PVX, to deliver Cas12a and sgRNA, respectively, for multiplex editing (Uranga, Vazquez-Vilar, et al., 2021). Alternatively, smaller-sized nucleases, including a split SaCas9 and I-SceI homing endonuclease, have been delivered using tomato mosaic virus to generate multicopy mutations in Nicotiana benthamiana (Chujo et al., 2017; Kaya et al., 2017).

Another nonintegrative delivery strategy is the use of pollen from haploid inducer (HI) lines carrying the gRNA and Cas transgenes. In this case, gene-editing components are transcribed from the chromosomes of the male (pollen) genome after fertilization and prior to their elimination. The resulting sterile haploid plants can be treated with mitotic inhibitors to restore their fertility and obtain homozygous, fertile doubled haploid (DH) lines from seed. The initial publication of HI-mediated editing performed in wheat reported that 0.7% (2/292) and 0% (0/61) of recovered haploids had edits in the target gene, which were limited to a single subgenome; while use of a pollen-specific promoter to drive Cas9 improved mutagenesis efficiency, no abd haploid mutants were obtained (Kelliher et al., 2019).

Budhagatapalli et al. (2020) also used pollen from a transgenic maize HI to edit tetraploid durum and hexaploid bread wheat and recovered only chimeric durum plants and single subgenome (AABBdd) bread wheat mutants. A transgenic maternal double haploid inducer (DHI) canola line was used to generate transgene-free edited DH lines with multiallelic mutations, but yet again all recovered plants were chimeras (Li, Sang, et al., 2021). Fine-tuning spatiotemporal expression and catalytic activity of the gene editing reagents will be critical for improving the efficiency of HI/DHI-mediated approaches to targeted mutagenesis in polyploids. Furthermore, HI-mediated editing is best suited for polyploid species in which inbred or F1 hybrid cultivar development is desired.

FUTURE PERSPECTIVES AND CONCLUSIONS

For nearly a decade, targeted mutagenesis mediated by SSNs has been revolutionizing crop improvement, including polyploids. Gene knockouts enabled by meganucleases, ZFNs, TALENs, and CRISPR/Cas have improved disease resistance, yield, and end-use characteristics in a number of economically important polyploid crops. Generation of targeted nucleotide substitutions and in-frame mutations that replace inferior alleles with superior alleles has also been accomplished in polyploids (Hegde et al., 2021; Oz et al., 2021; Ran et al., 2018; Wang, Wan, et al., 2021). These precision editing approaches offer great opportunities for improved or new gene function and will become increasingly achievable with the implementation of base editing (Cheng et al., 2021; Zhang, Liu, et al., 2019; Zong et al., 2018) and prime editing (Lin et al., 2020) technologies in polyploids.

Another promising development in recent years has been the engineering of CRISPR enzymes with relaxed PAM requirements, including xCas9, which can recognize 5′-NG-3′, 5′-GAA-3′, and 5′-GAT-3′ PAM sequences (Hu et al., 2018), Cas9-NG, which recognizes 5′-NG-3′ PAM sites (Nishimasu et al., 2018), and SpRY Cas9, which recognizes an NRN or NYN PAM (Ren, Sretenovic, et al., 2021; Walton et al., 2020). Both xCas9 and Cas9-NG have been deployed for gene knockout in wheat (Wang, Tian, et al., 2021). Other engineered variants of SpCas9, including “enhanced specificity” SpCas9 (eSpCas9) (Slaymaker et al., 2016) and high-fidelity1 SpCas9 (Cas9-HF1) (Kleinstiver et al., 2016), were selected for lower off-target activity; their applicability for multiallelic gene editing was confirmed in potato protoplasts (Nadakuduti et al., 2019). Optimization of the efficiency and on-target activity of novel Cas enzymes should help to address the challenge of identifying highly conserved gRNA sequences in polyploids.

Targeting of cis-regulatory elements, as opposed to protein coding sequence, is an emerging strategy that has been used to generate continuous trait variation at single loci in diploids (Rodríguez-Leal et al., 2017; Wang, Aguirre, et al., 2021) and could be used to further fine-tune continuous trait variation in polyploids. However, because greater sequence variation is found in noncoding sequences of different genotypes than in coding sequences, co-editing of multiple copies of regulatory elements is more challenging in polyploids than in diploids.

The development of Cas nucleases with relaxed PAM requirements like Cas9-NG or SpRY Cas9 in combination with sgRNA libraries spanning the entire open reading frame has enabled directed evolution of novel desirable gene variants. Delivery of Cas enzymes and pooled sgRNAs to plant cells followed by strategic application of selection pressure has produced rice plants with resistance to splicing inhibitors as well as ACCase and ALS inhibitor herbicides (Butt et al., 2019; Kuang et al., 2020; Li, Zhang, et al., 2020). In the future, polyploid plant species could provide a superior platform to diploids for directed evolution. The presence of multiple functionally redundant allelic variants in polyploid genomes may provide greater protection from adverse effects of sgRNA off-targeting as well as increased opportunities for accumulating mutations resulting in neo- or subfunctionalization of native target genes, as occurs in the natural evolution of polyploid genomes.

Because recalcitrance to genetic transformation continues to hinder applications of genome-editing tools for crop improvement and functional genomics, emerging technologies that bypass tissue culture and/or enable species- and genotype-independent plant transformation are potential game-changers for plant science research. The use of morphogenic regulators, including overexpression of WUSCHEL, BABY BOOM, LEAFY COTYLEDON1, LEAFY COTYLEDON2, as well as a fusion protein of GROWTH REGULATING FACTOR 4 and GRF-INTERACTING FACTOR1, has improved embryogenesis response and transformation efficiencies as well as expanded the number of transformable genotypes in plants, including the polyploid species canola, wheat, triticale, and sugarcane (Boutilier et al., 2002; Debernardi et al., 2020; Lotan et al., 1998; Lowe et al., 2016; Stone et al., 2001; Zuo et al., 2002). Developmental regulators have also been used to regenerate tetra-allelic edited lines by de novo meristem induction on whole Nicotiana benthamiana plants (Maher et al., 2020), although there are no published reports yet for this technique in monocots. The continuing innovations in plant transformation technologies will help to sustain future efforts in genome editing for polyploid crop improvement.

In summary, given the central importance of polyploids in world agriculture, rapid technological advancements in their breeding will be required to tackle burgeoning threats to global food security. Continuing breakthroughs in sequencing technologies, phenomics, computational biology, artificial intelligence, and speed breeding techniques will further the understanding of the genetic control of critical agricultural traits. For improving quantitative traits, phenotypic prediction holds great potential. Genome editing is the most promising technology for improving and stacking qualitative traits controlled by a single or small number of genes. Targeted mutagenesis is particularly promising for polyploid systems that impede conventional trait introgression methods.

The biological predilection for imprecise repair of DNA breaks has limited the vast majority of genome-editing applications to targeted mutagenesis. In this review, we have surveyed the considerable progress made in targeted mutagenesis of polyploid crops, including numerous examples of trait development and protocol optimizations. While optimized base and prime editing platforms, along with more efficient HDR-mediated gene knock-in, will enable production of important gain-of-function mutations, targeted mutagenesis should remain a vital plant breeding and reverse genetics technique in polyploid crops, with targeting of cis-regulatory elements and directed evolution expanding its capabilities beyond the generation of null mutations.

ACKNOWLEDGMENTS

This research was funded by the DOE Center for Advanced Bioenergy and Bioproducts Innovation (U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research) under award number DE-SC0018420. Any opinions, findings and conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the U.S. Department of Energy. This work was also supported by the USDA National Institute of Food and Agriculture, Hatch project 1020425. We also thank Kaitlyn Vondracek for fact and reference checking prior manuscript submission.

AUTHOR CONTRIBUTIONS

David May curated the data, performed investigation and visualization, and wrote the original draft. Katalin Paldi wrote section 7 of the original draft and curated data for supplemental tables. Fredy Altpeter conceptualized the idea of the study, acquired funding, designed the methodology, performed supervision, and reviewed and edited the manuscript.