Targeted mutation breeding of horticultural plants

Mutation breeding can be enhanced by genetic selection for novel alleles. Through targeted mutation breeding, genotypes with induced or natural mutations in candidate genes are identified for cultivar development. For most horticultural plants, targeted mutation breeding may be a more economically feasible approach to trait development than through transgenic technology. Substantial progress has been made in applying targeted mutation breeding to horticulture and this review summarizes recently published work in this area. To date, at least 16 horticultural crops have been screened for natural or induced allelic diversity in over 100 candidate genes. This approach has resulted in traits of commercial use, such as longer shelf-life (tomato, melon), improved starch quality (potato), and virus-resistance (peppers, tomato). Advances in genome sequencing and genetic screening will facilitate the development of cultivars with value-added traits derived from candidate gene polymorphisms.


Introduction
Targeted mutation breeding uses genomics to increase the efficiency of mutation breeding.Mutation breeding has been an effective approach to producing horticultural varieties with improved traits (Ahloowalia et al., 2004), but it requires the phenotypic analysis of a large number of plants.An advantage of mutation breeding is that cultivars developed through this approach generally do not face the regulatory, intellectual property, and economic challenges that limit the cultivation of transgenic horticultural plants (Alston, 2004;Dobres, 2008).The unregulated status of cultivars derived from induced mutations is warranted, given that many traditionally bred traits in horticultural plants are due to natural mutations (Janick, 2004).For example, natural mutations in single genes of tomato are completely or mostly responsible for its determinate growth habit (Pnueli et al., 1998), resistance to powdery mildew (Bai et al., 2008), and yield heterosis (Krieger et al., 2010).
Through targeted mutation breeding, genotypes with induced or natural mutations in specific genes are identified and brought into a cultivar development program.With early genetic selection, greenhouse and field testing can be conducted more efficiently.Genetic analysis allows genotypes of woody horticultural plants with candidate gene variation to be identified years before an associated flower or fruit trait is expressed.In addition, genetic screening can identify plants with heterozygous recessive mutations, which are usually missed by phenotypic analysis.The target gene polymorphism can be used for marker-assisted selection during cultivar development.
There are several excellent reviews that describe the methods, the challenges, and the potential of mutation screening for crop improvement (e.g.Comai and Henikoff, 2006;Till et al., 2007;Barkley and Wang, 2008;Parry et al., 2009;Gilchrist and Haughn, 2010;Jankowicz-Cieslak et al., 2011).This review focuses on the substantial progress that has recently been made on the application of targeted mutation breeding to horticultural plants.Targeted mutation breeding requires candidate gene sequence data, screening populations, and mutation screening techniques and these topics will be covered in the context of their application to horticulture.Examples are presented of value-added traits of horticultural crops that have been developed through the targeted mutation breeding.

Sequence data of candidate genes
Molecular genetic analysis of model and crop species over the past 25 years has characterized genes involved in a wide range of traits (e.g.Bowman et al., 1989, Kodrzycki et al., 1989).Association genetics is being used to discover genes that have large effects on quantitative traits such as flowering time (Saidou et al., 2009) and oil quality (Belo et al., 2008).As a result of this research, candidate genes have been identified that play key roles in horticultural traits such as fruit quality, starch and oil composition, and disease resistance.Table 1 shows many of the candidate genes that have been targeted in horticultural crops.These genes encode regulatory proteins (e.g.AGAMOUS), key biosynthetic enzymes (e.g.polygalacturonidase), and pathogen-required proteins (e.g.translation initiation factor 4E).  TILLING, EcoTILLING Till et al., 2010, 2011 Candidate gene sequence from the target species is needed for genetic screening.Sequenced genomes of over 30 plant species are currently available, at least in draft form, and genomic analysis of more than 100 other plant species is in progress (NCBI Genome Project database; http://www.ncbi.nlm.nih.gov/genomes/leuks.cgi).To date, horticultural crops with sequenced genomes include apple, banana, cassava, citrus, coffee, cucumber, grape, papaya, peach, potato, strawberry, and tomato.In addition to genomic resources, bioinformatic tools have been developed for many crop species (Mochida and Shinozaki, 2010).Comparative genomics is giving insights that improve the discovery of orthologous genes across plant taxa (Paterson et al., 2010).Advances in sequencing, bioinformatics, and comparative genomics will make the acquisition of sequence data feasible for any horticultural species.Genomic data can be applied to cultivar development using targeted mutation breeding.

Detection of natural or induced polymorphism in candidate genes
A number of genetic screening methods have been used to identify sequence variation in plant genes.Techniques such as conformation-sensitive capillary electrophoresis (CSCE), single-strand conformation polymorphism (SSCP), and denaturing high performance liquid chromatography (dHPLC) detect polymorphisms in candidate genes based on changes in conformation of PCR-amplified sequences.CSCE, SSCP, and dHPLC have been used to find sequence variation in specific genes of tomato (Gady et al., 2009), tobacco (Julio et al., 2008), and sunflower (Fusari et al., 2010), respectively.
TILLING is a reverse genetics approach that, in its most widely used application, identifies mutations through nuclease cleavage of DNA mismatches (Till et al., 2007).TILLING has come to refer to the general process of detecting induced polymorphism, whether by nuclease cleavage, dHLPC (McCallum et al., 2000), high resolution melting (Bush and Krysan, 2010), or sequencing (Tsai et al., 2011).For clarity in this review, TILLING will refer to detection through mismatchspecific nuclease cleavage, which was used for all TILLING screens shown in Table 1.EcoTILLING is a technically similar approach that has been used to identify natural polymorphisms in target genes of melon, bean, sunflower, and peppers (Nieto et al., 2007;Galeano et al., 2009;Fusari et al., 2010;Ibiza et al., 2010).
High resolution melting (HRM) detects mutations due to changes in melting properties of PCR products with DNA mismatches.PCR and HRM analysis are conducted on the same samples in multiwell format, which facilitates throughput.HRM has been used to discover induced variability in specific tomato, wheat, and maize genes (Gady et al., 2009, Dong et al., 2009, Li et al., 2010) and natural variability in target genes of almond, barley, lettuce, olive, potato, and peach (Wu et al., 2009;Hofinger et al., 2009;Simko et al., 2009;Muleo et al., 2009;De Koeyer et al., 2009;Chen and Wilde, 2011).Improvement of mismatch detection by HRM and TILLING is an active field of study, largely due to medical diagnostic applications (e.g.Li et al., 2008;Montgomery et al., 2010).
High-throughput sequencing of PCR products from candidate genes is an alternative to DNA mismatch detection.Sequencing can be particularly useful for outcrossing, polyploid crops like potato, where high levels of allelic diversity can complicate mismatch detection (Muth et al., 2008).Next-generation sequencing has considerable potential as a mutation screening tool when strategies to distinguish mutations from sequencing errors are employed and sample pooling is used to improve cost-efficiency (Gilchrist and Haughn, 2010).Roche 454 sequencing, for example, was used to identify candidate gene mutations in pooled samples of tomato (Rigola et al., 2009) and petunia (Stuurman, 2010).The combination of sample pooling, barcoding, and bioinformatics resulted in sensitive and specific screens for induced mutations in rice and wheat genes through Illumina sequencing (Tsai et al., 2011).

Genetic screening and cultivar development
The size of the population screened for natural or induced candidate gene polymorphisms depends on genomic characteristics such as ploidy and the degree of allelic diversity.An EcoTILLING analysis of 80 banana accessions identified 870 novel alleles in 14 genes (Till et al., 2010).Variants in eIF4E that conferred virus resistance were discovered by the genetic screening of 113 melon accessions and 233 pepper accessions (Nieto et al., 2007;Ibiza et al., 2010).High-throughput sequencing of eIF4E from 92 tomato lines identified six allelic variants, although virus resistance was not reported (Rigola et al., 2009).
Because the natural allelic diversity of melon eIF4E was found to be limited, mutagenesis was proposed as a means to obtain novel eIF4E alleles (Nieto et al., 2007).Genetic diversity in melon and other plants can be induced using chemical agents such as ethyl methanesulfonate (EMS) or physical agents such as gamma rays (Mba et al., 2010).Novel alleles of melon eIF4E were identified by the TILLING of 2483 EMS-mutagenized M2 families (Gonzalez et al., 2011).EMS-induced mutations in tomato eIF4E were identified by sequencing eIF4E genes from 3008 M2 families (Rigola et al., 2009) and by the TILLING of 4759 M3 families (Piron et al., 2010).The sizes of these screening populations are typical for horticultural plants with EMS-induced mutations (Table 2).EMS primarily causes single nucleotide polymorphisms (SNPs), which can alter encoded proteins through premature termination, missplicing, and codon changes.Table 3 shows the frequency of mutation types found in a large number of EMS-mutagenized Arabidopsis genes (Greene et al. 2003), as well as in the horticultural plants for which similar data are available.The distribution of mutation types in EMS-mutagenized tomato, pea, and melon genes was found to be similar to that of Arabidopsis, but with higher nonsynonymous:synonymous ratios (Dalmais et al., 2008;Dahmani-Mardas et al., 2010;Piron et al., 2010).Bioinformatic tools such as SIFT and PARSESNP have been developed to predict the severity of the nonsynomonous mutations (Ng and Taylor 2003;Taylor and Greene, 2003).In addition to coding sequence changes, polymorphisms in transcription and translation initiation signals were found to alter gene expression (Zhao et al., 2009;Knoll et al., 2011).
In several studies, novel phenotypes from induced or natural mutations in specific genes were confirmed and the types of mutations responsible for tested phenotypes are shown in Table 3. SNPs that result in truncation or missplicing have severe effects and comprise nearly one-third of the cases leading to novel phenotypes.
Plants with synonymous mutations are generally not tested further, although it has recently been observed that synonymous mutations can also affect the level and structure of the encoded proteins (Saunders and Deane, 2010;Delker and Quint, 2011).The deliberate introduction of unpreferred synonomous codons was found to alter levels of the encoded protein due to changes in translation (Carlini and Stephen, 2003) or miRNA regulation (Mallory et al., 2004).Synonymous substitutions can alter protein function through folding changes that result from ribosomal stalling (Tsai et al., 2008), although the frequency of this result is unknown.
Following the selection of lines with candidate gene polymorphisms, cultivar development follows practices that have been successful for traditional mutation breeding.This can involve one or more of the following: breeding to reduce background mutations, combining variant homeologous genes of polyploid plants, and tissue culture to eliminate chimerism or manipulate ploidy (McCallum et al., 2008;Dong et al., 2009;Muth et al., 2008).
Cultivar development can be streamlined by marker-assisted selection based on the candidate gene polymorphism.

Horticultural traits developed through targeted mutation breeding
Table 2 shows several examples of horticultural plants with improved traits from induced mutations in candidate genes.In particular, progress has been made in developing value-added traits involving fruit quality, storage product modification, and disease resistance.For example, fruit with improved shelf life have been developed by identifying induced mutations in the polygalacturonidase (PG) genes of tomato and melon (McCallum et al., 2008;Dahmami-Mardas et al., 2010).Both species were self-fertilized to obtain homozygous PG mutations.Prior to selfing, tomato lines with missense PG mutations were backcrossed twice to reduce background mutations.Delayed ripening in tomato was also obtained by TILLING for induced variation in genes for expansin 1 (Colbert et al., 2011), βgalactosidase 4 (Hurst et al., 2011), and two ethylene receptors (Okabe et al., 2011).Candidate genes for other fruit traits such as color and sugar content are being investigated in EMS-mutagenized populations of tomato (Minoia et al., 2010;Gady et al., 2009;Gady et al., 2011) and melon (Gonzalez et al., 2011).
Plant products such as starch and oil have been modified by mutations in genes for key biosynthetic enzymes.High-amylopectin starch was produced in potato lines with waxy gene mutations (Muth et al., 2010), in an approach similar to that used for wheat (e.g.Dong et al., 2009).Like wheat, potato is polyploid and a combination of tissue culture and breeding was required to develop plants homozygous for the variant alleles of waxy homeologs.Waxy mutations for starch modification are also being investigated in gamma-irradiated cassava (Tofino et al., 2009).Changes in seed oil composition have obtained from mutations induced in the gene encoding Δ12 fatty acid desaturase (FAD2).TILLING for fad2 variants that increase oleic acid has been successful in soybean (Dierking and Bilyeu, 2009) and novel fad2 alleles are being pursued in peanut (Knoll et al., 2011) and sunflower (Sabetta et al., 2011).
The identification of variant eIF4E alleles that conferred virus resistance was mentioned above.Natural allelic variation found in eIF4E genes provided melon with resistance to melon necrotic spot virus (Nieto et al., 2007) and peppers with resistance to potato virus Y and tobacco etch virus (Ibiza et al., 2010;Jeong et al., 2011).An EMSinduced splice site mutation in tomato eIF4E conferred resistance to strains of potato virus Y and pepper mottle virus (Piron et al., 2010).EMSinduced mutations were found in melon eIF4E that are predicted by SIFT analysis to affect protein function (Gonzalez et al., 2011).The new eIF4E alleles can be used as a genetic resource for potyvirus resistance in breeding programs.Natural SNPs in other virus resistance loci were identified in lettuce and common bean for application in marker-assisted selection (Simko et al., 2009;Galeano et al., 2009).Fungal resistance has also  2010).In addition to eIF4E, Mlo, and Prm6, there are other susceptibility gene candidates that may provide resistance when their pathogen-required function is disrupted (Pavan et al., 2010).

Advances in targeted mutation breeding
Continued progress in genome sequencing and genetic screening will facilitate the development of traits derived from candidate gene variation.Capillary electrophoresis, for example, has increased the throughput of TILLING and SSCP (e.g.Suzuki et al., 2008;Julio et al., 2008), and supported commercial applications (Loeffler et al., 2011).Increasing the sensitivity of mismatch detection would permit more accurate analysis of large sample pools.Improvements developed for medical diagnostics, such as COLD-PCR (Li et al., 2008) and QMC-PCR (Fadhil et al., 2010), enable the detection of a target allele at 1-3% of wild-type background.COLD-PCR, which preferentially amplifies PCR products with DNA mismatches, enhanced the HRM detection of peach SNPs in pools of up to 18 genotypes (Figure 1).The discovery of allelic diversity through nextgeneration sequencing has become feasible for horticultural crops with the development of appropriate sequencing strategies (Stuurman, 2010;Tsai et al., 2011) and bioinformatic tools (Rigola et al., 2009;Missirian et al., 2011).Many of the challenges involved in detecting variant genotypes by high-throughput sequencing of pooled samples are also being addressed by medical research (e.g.Lo et al., 2010).High-throughput sequencing and HRM analysis have been combined for effective detection of SNPs in genes of crops with complex genomes (Han et al., 2011;Oliver et al., 2011).
An alternative to screening for randomly mutated alleles is the use of gene-specific mutagenesis directed by zinc-finger nucleases, meganucleases, or oligonucleotides.Zinc-finger nucleases (ZFNs) and meganucleases cause doublestranded breaks at target sites that result in mutations due to the imprecision of non-homologous end joining (Marton et al., 2010;Gao et al., 2010).Oligonucleotide-mediated mutagenesis causes sitespecific changes through the DNA mismatch repair (Schopke et al., 2009).Plants such as petunia, maize, and canola have been modified by genespecific mutagenesis, but this has been restricted almost exclusively to genes conferring selectable phenotypes.To identify lines with ZFN-targeted mutations in an Arabidopsis gene without a selectable phenotype, genetic screening was carried out by mismatch-specific nuclease cleavage (Osakabe et al., 2010).Arabidopsis lines with genespecific mutations were found at frequencies exceeding 1:400 genotypes, but lower frequencies would be expected in non-model systems.
With the availability of candidate gene sequence data and genetic screening methods, novel alleles have been found that occur naturally or are induced randomly or gene-specifically.To date, at least 16 horticultural crops have been screened for variation in over 100 candidate genes (Table 1).A high-amylopectin potato was the first cultivar developed by targeted mutation breeding to be in commercial production (Anonymous, 2009).Arcadia Biosciences has applied for patents on tomatoes with extended shelf life due to independent mutations in three different genes (McCallum et al., 2008;Hurst et al., 2011;Colbert et al., 2011).The harvest of genetic diversity, as described by Waugh et al. (2006), is taking place in horticultural plants.

Figure 1 .
Figure 1.COLD-PCR increases the sensitivity of detection of variant peach alleles by HRM.Lines 16 and 30 had heterozygous SNPs in PpTFL1 and PpAG, respectively.Sample pools (labeled X1, Y1, etc.) contained six lines and pools were combined so that six, twelve, or eighteen lines were analyzed by HRM.Pools X3 and Y6 contained lines 16 and 30, respectively, and only these lines had target gene SNPs.The percentage of the SNP-bearing allele in the sample analyzed is shown in parentheses.Three replicates per pool were analyzed and line colors indicate grouping by LightCycler480 analysis software.Panels A and C show the results for standard PCR/HRM of PpAG and PpTFL1, respectively.Panels B and D show increased HRM sensitivity for PpAG and PpTFL1, respectively, following COLD-PCR.Figure from Chen and Wilde (2011).

Table 1 .
Horticultural plants screened for natural and induced variation in candidate genes.

Table 2 .
Results from genetic screening of EMS-mutagenized horticultural plants.
a In several of these studies, phenotypic changes from induced SNPs were not tested (NT).In cases where novel traits from induced SNPs were characterized, only one example is given.

Table 3 .
Frequency of mutation types.
Similar to an EcoTILLING approach used for barley(Mejhede et al., 2006), variation in grape orthologs of Mlo and Prm6 is being investigated as a means of conferring powdery mildew resistance (Cadle-Davidson,