Over 300 million tonnes of potatoes are produced worldwide pa (http://faostat3.fao.org/home/index.html retrieved 2 April, 2013). A third of potato production takes place in developing countries, and over 1 billion people have potato as their staple diet. Potatoes grow in a wide range of habitats from sea level to 4000 m altitude and from 47˚S to 65˚N latitude (Hijmans, 2001). The average yield potential of potato varies markedly with local growing conditions and ranges from about 13 t/ha in Africa and Asia up to 50 t/ha in Western Europe and the USA. Potatoes are second only to soybean for amount of protein/ha, with the major storage protein being patatin, one of the most nutritionally balanced plant proteins known (Liedl et al., 1987). The tubers are a globally important dietary source of starch, protein, antioxidants and vitamins, serving the plant as both a storage organ and a vegetative propagation system (Burlingame et al., 2009). The importance of potato as a crop internationally, especially within the developing world, was highlighted by the UN in 2008, which was the Year of the Potato (www.potato2008.org/en/aboutiyp/index.html, retrieved 2 April 2013).
Potato is a member of the Solanaceae, a large plant family with more than 3000 species. It has one of the richest genetic resources of any cultivated plant, with about 190 wild and primitive species in the section Petota of the genus Solanum (Jacobs et al., 2011; Spooner and Hijmans, 2001). The tuber-bearing Solanum species are very widely distributed in the Americas. Many wild species, which can be crossed directly with cultivated potato, feature a broad spectrum of resistances to pests and diseases, tolerances to abiotic stress such as frost and drought and many other valuable traits (Plaisted and Hoopes, 1989). The commercial potato has had a narrow genetic base because of the limited foundation stock introduced from the South American center of origin to Europe. As a vegetatively propagated, highly heterozygous, polyploid crop, potato cultivars are often limited in reproductive fertility, making potato breeding a challenging endeavour. Classical improvement schemes comprise relatively long breeding cycles and are dictated by genetic complexity and the sensitivity of potato to inbreeding depression. Applications of plant biotechnology over the past five decades have helped facilitate interspecies crosses and to augment and broaden the cultivated gene pool.
Plant biotechnology has traditionally encompassed the application of cell and tissue culture for crop improvement. From the mid-1980s, the development and application of transgenic plants was the dominant research activity associated with plant biotechnology. More recently research has expanded to include the application of genomics technologies. Members of the Solanaceae family, especially tobacco and petunia (Conner and Meredith, 1989; Conner et al., 2009), have been the subject of many of the developments in plant biotechnology, primarily due to their high propensity for growth and development in cell culture. Similarly, many elite potato genotypes are highly responsive in cell culture and provide opportunities for applications of biotechnology to potato improvement. The recent publication of the full genome sequence of potato (The Potato Genome Sequencing Consortium, 2011) has presented new opportunities for potato biotechnology.
Herein, we begin with a historical account of traditional tissue culture approaches and applications of cell culture, then review the past developments and new approaches for gene transfer via transformation. We also review the applications of genetic markers and genomics for potato breeding. Our intention is to illustrate these diverse approaches with key examples, rather than provide an exhaustive survey of the scientific literature.
Historical perspectives of potato biotechnology
Biotechnology in the context of potato breeding
From the perspective of global food security, potato breeding has been highly successful over the past century. Conventional breeding methods for potato involve the hybridization of parental clones followed by selection among large seedling populations for superior individuals with the desired combination of traits (Plaisted et al., 1994). Single-plant-derived clones are propagated vegetatively and are evaluated for relevant agronomic and quality attributes during repeated seasons of selection. This general approach to potato breeding has resulted in the development of many elite clones, which have become highly successful potato cultivars.
Despite intensive breeding efforts, the yield potential of potato cultivars has remained relatively constant over the last century. When grown using the same modern field management practices, cultivars developed in the nineteenth century exhibited the same yield potential as those bred in the USA during 1932–1991 (Douches et al., 1996). The main objectives of most potato breeding programmes have involved the improvement in specific processing attributes and resistance to pests and diseases, while maintaining or improving traits such as tuber colour, shape, quality and yield. Over the past 50–60 years, achieving many of these goals has been greatly assisted by the transfer of genes from related Solanum species. Improvement in valuable traits such as resistance to pests, disease and frost (Estrada, 1980), and quality attributes such as starch content, protein level and low levels of reducing sugars have been greatly assisted by introgressions from more than 25 wild Solanum species (Ross, 1986). The value of historical interspecific hybridization to potato breeding is well exemplified by the fact that more than half of all potato cultivars have introgressions from Solanum demissum, with some cultivars containing introgressions from more than one wild species (Ross, 1986).
One of the major difficulties associated with conventional potato breeding is the autotetraploid status (2n = 4x = 48) of virtually all commercial potato cultivars. The resulting tetrasomic inheritance adds considerable complexity to potato breeding, especially in conjunction with the associated high heterozygosity (Conner et al., 1997), making the accumulation of desired alleles in a breeding context very difficult. To recover individuals for evaluation as potential new cultivars, exceptionally large populations of potato seedlings need to be screened. Initial selection for many desirable characters can be inefficient and/or time-consuming, and potato breeders often have to screen up to a million seedlings to identify a single clonal line that survives through to the release of a successful cultivar (Plaisted et al., 1994).
Over the past 60 years, plant biotechnology has contributed valuable solutions to some of the difficulties associated with conventional potato breeding. These have related primarily to applications of cell and tissue culture. Modern tools of plant transformation and DNA markers offer new opportunities for both gaining improved efficiencies in selection of elite clones and enhancing genetic gain over time.
Pathogen elimination in seed potatoes
Potatoes are susceptible to a range of pests and diseases that have large negative impacts on yield and tuber quality. Seedborne diseases of potato are numerous, and diseased seed tubers lose quality during storage between growing seasons (Struik and Wiersema, 1999). For this reason, a major focus of potato biotechnology in the mid-20th century was on pathogen elimination for the production of high-health potato seed. An early example of pathogen elimination includes the work of Kassanis (1949) who showed that heat treatment could be used to eliminate virus infection in potato tubers. The age of biotechnology in potato began in 1956, where heat treatment coupled with in vitro culture of shoot tips successfully eliminated potato virus Y (Thomson, 1956). The observation that virus is often absent in meristematic regions led to successful refinement of meristem culture to generate virus-free potato plants (Kassanis, 1957). Many programmes and companies have been established worldwide to provide pathogen-free seed potatoes for farm supply (Struik and Wiersema, 1999).
The generation of haploid potatoes was demonstrated through the use of Solanum phureja (2n = 2x = 24) as a pollen donor onto S. tuberosum (Peloquin and Hougas, 1958), and haploids are now relatively easy to generate (Liu and Douches, 1993). Due to the complex nature of genetic interactions among genes in a tetraploid species, haploid potatoes are desirable for the study of genetic interactions in a less complex background (Peloquin and Hougas, 1960). A breeding scheme involving the scaling up and down of potato chromosome sets was first proposed by Chase (1963) and referred to as analytical breeding. The first step involves the generation of maternal haploids from tetraploid cultivars through parthenogenesis by crossing to a diploid species. The resulting plants, often referred to as dihaploids in potato, are crossed with other diploid stocks for breeding at the diploid level. The final step involves returning to the tetraploid level by unilateral (or bilateral) sexual polyploidization using 4x−2x (or 2x−2x) crosses and relying on the occurrence of unreduced 2n gametes in the diploid parent. Such cycling between ploidy levels is achieved relatively easily in potato and provides a means to simplify the introgression of traits from new sources of genetic diversity (Ortiz, 1998). The use of haploid potatoes was pivotal in the potato genome sequencing project (The Potato Genome Sequencing Consortium, 2011; see 'Genomics' section, this review).
The endosperm balance number (EBN) plays an important role in the speciation of tetraploid from diploid Solanum species (Hawkes and Jackson, 1992). Upon hybridization, the EBN should be in a 2 : 1 maternal to paternal ratio for normal endosperm development and successful seed production (Johnston et al., 1990). The hybridization barriers between disomic (2 EBN) and tetrasomic (4 EBN) tuberous Solanum species can be overcome by double pollination and rescue of aborting embryos via tissue culture (e.g. Iwanaga et al., 1991; Singsit and Hanneman, 1991; Watanabe et al., 1995). Embryo culture has also been valuable for circumvention of other forms of interspecific incompatibility. For example, resistance to potato leaf roll virus was successfully introgressed from Solanum etuberosum to S. tuberosum via embryo culture (Chavez et al., 1988). An extreme example of the use of embryo culture to aid the recovery of wide hybrids involves the successful hybridization of the disomic hexaploid Solanum nigrum (black nightshade) as a female parent with tetraploid potato (Eijlander and Stiekema, 1994).
During the history of regenerating plants from cell cultures, the perceptions of the clonal integrity of the resulting plants have changed. The historical dogma was that all plants regenerated from somatic tissue are identical to the parent plant. However, the contrary view was popularized by Larkin and Scowcroft (1981), who described the high frequency of phenotypically variant plants observed among those regenerated from cell culture and coined the term ‘somaclonal variation’ for the phenomenon. For the ‘optimists’, somaclonal variation was seen as a new approach for generating novel variation in plants, especially in clonally propagated crops such as potato. For the ‘pessimists’, it was seen as an inherent curse for other applications of cell culture for crop improvement.
The observation that phenotypic changes arise during the cell culture and regeneration phase of potato tissue culture was communicated more widely following the recovery of a vast array of novel phenotypes after regeneration of plants from leaf protoplasts of ‘Russet Burbank’ and similar cultivars (Shepard et al., 1980). This finding was quickly given support by other studies that involved growing potato somaclones in the field (Larkin and Scowcroft, 1981; Potter and Jones, 1991; Secor and Shepard, 1981). Explanations that account for the observed phenotypic changes among somaclonal potato lines involve physiological, epigenetic or genetic changes associated with the cell culture and shoot regeneration phase of plant transformation (Evans, 1989; Karp, 1991). As not all variants have a genetic basis, lines exhibiting phenotypic changes need to be grown over several field seasons to ensure stability of performance. Stable phenotypic changes of either heritable and/or epigenetic origin may arise through ploidy changes, chromosomal aberrations, gene amplification, activation of transposable elements, DNA methylation changes or point mutations (Phillips et al., 1994; Seibt et al., 2012; Veilleux and Johnson, 1998) and can also occur during the long-term propagation of potato from internodal stem cuttings (Dann and Wilson, 2011). Phenotypic variation among plants regenerated from cell cultures is often correlated with changes in chromosome number and/or structural chromosome aberrations, and such changes have been frequently observed in regenerated potato plants (Gill et al., 1987; Karp et al., 1982). Such cytological changes are usually accompanied by poor agronomic performance.
At the time, it was widely believed that these variants arising in cell culture could be applied immediately as single trait improvements to existing elite cultivars already well established in the market place. Considerable effort was devoted to recovering useful variants in a wide range of cultivars, and variants with improvements in specific useful traits were reported. Examples include improved resistance to Alternaria solani (Matern et al., 1978), Verticillium dahlia (Sebastiani et al., 1994), late blight Phytophthora infestans (Shepard et al., 1980) and more recently tuber morphology (Thieme and Griess, 2005). However, it has been recognized over time that the recovery of a somaclonal line exhibiting beneficial traits without other simultaneously arising negative attributes is very rare. Nowadays, the phenomenon of somaclonal variation is widely seen as an inherent negative feature of regeneration from cell culture and considered as something to be avoided. Strategies to minimize the impact of somaclonal variation on plant performance are routinely implemented during other applications of cell culture for potato improvement.
By the early 1980s, the routine isolation and culture of protoplasts was possible for many plants using cell wall degrading enzymes, coupled with maintaining the appropriate osmotic balance until cell walls had redeveloped (Conner and Meredith, 1989). Upon the mixing of protoplasts from different species, somatic fusion can be stimulated by chemical or electrical treatments prior to the re-synthesis of cell walls. The regeneration of somatic hybrid plants from these cells is possibly provided the two parental species are closely related, even if they cannot be sexually hybridized. Such somatic hybrid plants offer new sources of germplasm for the introgression of traits into crop plants, although this is often very challenging due to the poor fertility of the initial somatic hybrids.
Somatic hybridization has provided some new opportunities for introgression of novel sources of disease and pest resistance into cultivated potato from accessions of taxa possessing sexual reproductive barriers with potato. Resistances to diseases caused by leaf roll virus, potato virus Y, early and late blight, soft rot, Columbia root-knot nematode and Colorado potato beetle have been introduced through somatic fusion of potato protoplasts with protoplasts of wild relatives, including Solanum palustre (formerly S. brevidens) and Solanum bulbocastanum (Bradshaw et al., 2006; Brown et al., 2006; Helgeson et al., 1986, 1998; Tek et al., 2004; Thieme et al., 2010). However, despite these hybrids being good sources of resistance to pathogens, as well as abiotic stress, they often produce small mis-shapen tubers that are far from suitable for agricultural production. Multiple cycles of backcrosses are required for these plants to be useful in agriculture (Tek et al., 2004).
Somatic cell selection
The application of somatic cell selection to potato improvement has mainly focused on selection for disease resistance in potato. This has involved the exposure of large populations of cultured potato cells to pathotoxins or culture filtrates of pathogens, followed by the selection of rare surviving cells. An excellent example of the application of somatic cell selection for potato improvement involves the development of clones with resistance to common scab disease (Wilson et al., 2009, 2010). Common scab is caused by Streptomyces scabies and related species and results in economic loss through deep-pitted surface lesions on tubers that severely reduce processing quality (Wilson et al., 1999). All strains and species of Streptomyces that induce common scab produce a phytotoxin known as thaxtomin A (King et al., 1991). Thaxtomin A can induce disease symptoms in the absence of Streptomyces (Lawrence et al., 1990), and increased resistance to thaxtomin A in cultured potato seedlings has been correlated with decreased incidence of scab lesions on tubers (Hiltunen et al., 2006). Somatic cell selection for resistance to thaxtomin A in potato cell cultures enabled the recovery of plants with improved resistance to common scab (Tegg et al., 2013; Wilson et al., 2009, 2010). This disease resistance was achieved in elite clones and retained in field evaluations (Tegg et al., 2013; Wilson et al., 2010).
Other examples of using somatic cell selection in potato include the development of resistance to Fusarium species (Behnke, 1980; Wenzel and Foroughi-Wehr, 1990) and salt tolerance (Queiros et al., 2007), illustrating that this strategy may be useful for future selection of resistance to a number of biotic and abiotic stresses in potato, providing appropriate selection pressures are available.
Gene transfer to potatoes by transformation
Development of potato transformation systems
Potato was one of the first crops for which transgenic plants were regenerated (An et al., 1986; Shahin and Simpson, 1986), and Agrobacterium-mediated gene transfer protocols were quickly adapted for important cultivars throughout the world (de Block, 1988; Conner et al., 1991; Ishida et al., 1989; Newell et al., 1991; Sheerman and Bevan, 1988; Stiekema et al., 1988; Wenzler et al., 1989). This progress was motivated by the advantages that transformation offered for the genetic improvement in potatoes relative to the genetic limitations associated with traditional potato breeding.
Transformation offers a highly effective means of adding single genes to existing elite potato clones with no, or very minimal, disturbances to their genetic background (Conner et al., 1997). This is virtually impossible via traditional breeding as, due to the high heterozygosity in the tetraploid potato genome, the genetic integrity of potato clones is lost upon sexual reproduction as a result of allele segregation. As a consequence, potato transformation represents the only effective way to produce isogenic lines of specific genotypes/cultivars (Conner and Christey, 1994).
Potato transformation has also been achieved by direct DNA uptake (Féher et al., 1991; Romano et al., 2003; Valkov et al., 2011), and success has recently been reported for the generation of transgenic potato using vir gene–mediated gene transfer from Rhizobium species (Wendt et al., 2011) and Ensifer adhaerens (Wendt et al., 2012). However, Agrobacterium-mediated gene transfer is the preferred approach and is routinely performed in laboratories worldwide.
The source of explant tissue for potato transformation is usually derived from in vitro plants, maintained with pathogen-free (virus-free) status. This eliminates the need for surface sterilization of the plant tissue, which saves time, reduces the possibility of contamination and minimizes stress on the plant tissue from the chemical treatment (Conner et al., 1991). Leaves from in vitro plants are commonly used because they are readily available and easy to use (de Block, 1988; Conner et al., 1991; Wenzler et al., 1989). However, stem internode segments (Heeres et al., 2002; Newell et al., 1991) or minituber segments (Ishida et al., 1989; Sheerman and Bevan, 1988; Stiekema et al., 1988) have been equally successful. The preferred choice of explant is often cultivar-dependent or based on preference of individual laboratories.
The most common marker gene used for the selection of transformation events in potato is the nptII gene conferring kanamycin resistance. This marker gene is highly effective and is used almost exclusively for potato transformation. A few studies have reported success with other marker genes such as methotrexate resistance (Jacobs et al., 1994), hygromycin resistance (Meiyalaghan et al., 2010) and imidazoline (Storck et al., 2011). The relative efficiency of recovery of transgenic potato lines using identical vectors with alternative selectable marker genes was summarized as follows: kanamycin resistance > hygromycin resistance > phosphinothricin resistance > phleomycin resistance > methotrexate resistance (Barrell and Conner, 2006; Barrell et al., 2002). Another option is marker-free transformation, in which low frequencies of transformed plants are recovered by PCR-screening of plants regenerated without selection after co-cultivation with Agrobacterium. This approach has allowed the recovery of transgenic potato lines at a frequency of <0.2% or 4.5% when using Agrobacterium strains LBA4404 and AGL0, respectively (de Vetten et al., 2003). In a similar manner, frequencies of 0.5–1.0% of potato regenerants were transgenic when using this approach with Agrobacterium strain EHA105 (Meiyalaghan et al., 2005). However, selection for transformation events in tissue culture has two advantages: it concurrently selects against the recovery of chimeric plants with both transgenic and nontransgenic tissue and ensures a bias towards recovering transformation events occurring in nonrepetitive genome regions determining phenotypic performance.
Features of potato transformation
Although transformation provides an effective approach to transfer genes into elite potato cultivars without compromising existing attributes, in reality it is not quite that simple. The insertion and expression of transgenes, as well as the tissue culture processes associated with gene transfer, can all impose important constraints that limit the performance of the resulting transgenic line (Conner, 2007). Plant transformation is highly unpredictable with respect to the nature of transgene integration, the magnitude, specificity and stability of transgene expression as well as the frequency of off-types observed within populations of independently derived transformed plants (Conner and Christey, 1994). Potatoes are no exception (Conner et al., 1997). The selection of a large population of independently transformed lines is important, followed by screening the population to identify lines with the desired expression of the transgene, while maintaining the phenotypic attributes and yield performance (i.e. the absence of somaclonal variation) of the parental cultivar. In this respect, plant improvement via transformation is very similar to traditional breeding.
As in other species, the integration of transgenes into the potato genome may occur in a complete, truncated or a rearranged manner. Furthermore, it may occur as single insertion events or as tandem repeats at one or more integration sites (Conner et al., 1991; Porsch et al., 1998; Sheerman and Bevan, 1988; Theuns et al., 2002). The preferred transformation event involves the integration of a single intact construct into the potato genome (Conner et al., 1997). Plants initially selected following transformation are ‘heterozygous’ for the insertion event and are essentially equivalent to an ‘addition/deletion’ mutation in the hemizygous state. As the genetic integrity of potato clones is lost upon sexual reproduction, transformed potato lines are most likely maintained as clones from the primary transformants initially regenerated from tissue culture, through to commercial release. For tetraploid potato cultivars, this means that the transgenes will be represented in the simplex hemizygous state (A—) (Conner et al., 1997).
The mapping of DNA sequences that flank T-DNA insertion events in potato has established that integration occurs throughout the potato genome along the length of all chromosomes (El-Kharbotly et al., 1996). These T-DNA insertions were approximately evenly split between the high copy/repetitive DNA and the low/single copy DNA (Jacobs et al., 1994). Furthermore, following transformation with a promoterless β-glucuronidase gene, approximately half of all transgenic potato lines had activated expression of this gene (Lindsey et al., 1993), suggesting that integration occurred within existing genes. Given that the transformation process selects for integration events that express genes on the T-DNA regions, it is not surprising for there to be a bias towards integration into single copy genome regions that are transcribed.
The expression of transgenes within populations of transformed potato plants is commonly reported to be highly variable (Barrell and Conner, 2009; Jacobs et al., 2009; Meiyalaghan et al., 2006b; Wenzler et al., 1989). This is generally attributed to ‘position effects’ resulting from the ‘random integration’ of insertion events and is consistent with reports in other plant species (Conner and Christey, 1994). Furthermore, even when specific promoters are used for transcriptional control of gene expression, ‘position effects’ can override the intended regulatory control of the promoter (Meiyalaghan et al., 2006a). Sometimes only a small proportion (<10%) of the transformed lines exhibit the desired magnitude and the required spatial and temporal pattern of expression of the transferred gene.
The appearance of phenotypically abnormal plants among a population of transgenic plants is not uncommon (Conner and Christey, 1994; Veilleux and Johnson, 1998) and has been frequently reported in potato, especially following field evaluation (Conner, 2007). Marked phenotypic changes in plant appearance and/or significantly reduced tuber yield in many transgenic potato lines have been reported when large numbers of independently derived transgenic lines have been evaluated (Belknap et al., 1994; Conner et al., 1994; Dale and McPartlan, 1992; Jongedijk et al., 1993). Such off-types are initially apparent as atypical shoot morphologies, leading to stunted growth and markedly reduced yield and/or plant death (Belknap et al., 1994; Conner et al., 1994; Dale and McPartlan, 1992). Typical foliage off-types include short internodes, pronounced rolling of apical leaves, small cupped leaves, wrinkled leaf surface, smaller rugose leaflets and twisted petioles, whereas typical tuber off-types include a large number of small tubers, small and elongated tubers, deep eyes, knobbly tubers caused by secondary growths, loss of skin colour and russetting (Belknap et al., 1994; Conner et al., 1994; Davidson et al., 2002b). In general terms, the appearance of aberrant foliage is usually indicative of reduced tuber yield; however, transgenic lines with phenotypically normal foliage can exhibit disappointing tuber performance upon harvest (Conner, 2007).
The frequency of off-types among populations of transgenic potatoes has been recorded as 15–80% depending on the potato cultivar (Belknap et al., 1994; Dale and McPartlan, 1992; Davidson et al., 2002a,b; Heeres et al., 2002; Jongedijk et al., 1993). When these off-type appearances were first reported, it was important to develop an understanding for the basis of these unexpected phenotypes in transgenic potatoes. While many of the phenotypic changes in foliage were similar to symptoms of viral infection, the appearance of off-types was quickly established to be independent of the presence of viruses (Conner et al., 1994). The observed off-types were consistent when transgenic lines were replanted in the second year from tubers harvested in the first year (Belknap et al., 1994; Conner et al., 1994), thereby eliminating transient responses to tissue culture as being the cause. In some instances, unusual phenotypes might be expected to arise from the expression of specific transgenes or as unexpected pleiotropic effects of transgene expression. The recovery of some transgenic lines with high transgene expression coupled with a phenotypic appearance identical to the ‘parental’ cultivar suggests that the appearance of off-types is unlikely to result from direct or indirect effects of transgene expression (Conner et al., 1994; Dale and McPartlan, 1992). Insertional mutagenesis arising from gene disruption as a consequence of random integration of transgenes into the potato genome generally requires the transgene to be in a homozygous state. This is unlikely during potato transformation as the transformants regenerated from tissue culture are usually autotetraploids in the simplex hemizygous state (A—) that are then clonally propagated (see above). Therefore, insertional mutagenic events from gene knockouts are not expected to be observed in potato and will be only evident in very rare instances of gene activation (Conner and Jacobs, 1999).
The appearance of off-types among populations of transgenic lines that are unrelated to the expression of the transgenes is usually attributed to the occurrence of somaclonal variation resulting from epigenetic or genetic changes during the cell culture and shoot regeneration phase of plant transformation (Conner and Christey, 1994; Veilleux and Johnson, 1998). Early studies reporting off-type transgenic potatoes reached the same conclusions (Belknap et al., 1994; Conner et al., 1994; Dale and McPartlan, 1992; Jongedijk et al., 1993). This was largely based on the presumption that such off-types were to be expected, given the well-documented somaclonal variation occurring among potato plants regenerated from cell culture (Potter and Jones, 1991; Secor and Shepard, 1981; Shepard et al., 1980). As a consequence, the spectrum of off-type appearances is expected to vary markedly among different independently derived transgenic lines, a feature that is indeed commonly observed (Conner, 2007).
The phenotypic variation observed between plants assumed to have regenerated from the same single cell and the repeated reports of higher frequencies of phenotypic variation after longer periods in cell culture provides circumstantial evidence that off-types can arise during tissue culture. Unequivocal confirmation of this conjecture requires tracing the origin of single somatic cells that have been established in culture with a molecular marker. This was recently achieved using transgene integration as a unique molecular marker to trace the clonal origin of single potato cells in tissue culture (Barrell and Conner, 2011). Following Agrobacterium-mediated T-DNA transfer, multiple shoots were regenerated from transformed potato cell colonies and Southern analysis used to confirm their derivation from a single transformed cell. Analysis of phenotypic variation demonstrated marked differences between these multiple regeneration events, the origin of which must have occurred after T-DNA insertion and consequently during the tissue culture phase. This result unequivocally demonstrates that the origin of off-types observed among populations of transgenic potatoes occurs during tissue culture and is independent of transgene insertion and expression (Barrell and Conner, 2011).
When developing large populations of independently derived transgenic lines, the usual practice is to select the first shoot to regenerate from each transformation (Davidson et al., 2002a; Jacobs et al., 2009). This is based on a widely held view that the frequency of somaclonal variation increases with longer periods in cell culture and the resulting assumption that a shoot taking a longer time to regenerate has a greater chance of producing off-types (Birch, 1997; Karp, 1991; Veilleux and Johnson, 1998). A recently proposed new strategy to facilitate the recovery of phenotypically normal transgenic potato lines involves the regeneration and evaluation of multiple shoots from each transformation event (Barrell and Conner, 2011; Meiyalaghan et al., 2011). The key underpinning observation is that the first shoots recovered from each transformation event do not necessarily exhibit less somaclonal variation than later regeneration events. This highlights the value of regenerating and evaluating multiple shoots from each transformation event to facilitate the recovery of phenotypically normal transgenic lines; otherwise valuable lines with phenotypically normal appearance may be discarded.
Traits conferred by genetic engineering
The use of genetic engineering approaches has allowed the successful transfer of numerous transgenes into elite potato cultivars by either direct DNA uptake or Agrobacterium-mediated transformation. These transgenes have conferred a wide range of traits, including pest and disease resistances; abiotic stress resistance; quality attributes for improved processing, nutrition and appearance; and novel products for biopharming. Traits successfully transferred to potato by genetic engineering are summarized in Table S1. Although this is not an exhaustive list, it clearly highlights the diversity of traits successfully transferred to potato by genetic transformation and illustrates the immense potential of transformation for the genetic improvement in potato.
Genetic engineering case study: Bacillus thuringiensis toxin (Bt)-containing GM potatoes
In this section, we present a case study of one example of the most successful class of genetic transformation that is the focus of considerable ongoing research: insect-resistant, Bacillus thuringiensis (Bt) cry gene-transformed potatoes. Despite the relatively limited acreage of GM potato grown commercially, insect-resistant potatoes account for 13 of the 14 approvals for the release of GM potatoes (James, 2010).
Two of the most damaging insect pests of potato are the Colorado potato beetle (CPB) [Leptinotarsa decemlineata (Say)] and potato tuber moth (PTM) [Phthorimaea operculella (Zeller)]. CPB can completely destroy a plant by defoliating it, and the insect has been observed to develop resistance rapidly to a wide variety of insecticides (Harris and Svec, 1981). PTM affects potato crops in temperate as well as tropical climates (Fenemore, 1988) and can cause damage in the field, as well as in storage. The loss while in storage can be particularly severe in warmer climates, and losses of 100% have been reported (Lagnaoui et al., 2001).
A common method for biological control of insects is the use of Bacillus thuringiensis (Bt), a gram-positive soil bacterium, as a biopesticide (Lambert and Peferoen, 1992; Schnepf et al., 1998). During the stationary phase of its growth cycle, Bt generates parasporal inclusions which are composed of proteins known as crystal (CRY) proteins (Bravo et al., 2007). The proteins act on a limited spectrum of insect species, such as Lepidoptera (including PTM), Diptera, Coleoptera (including CPB) and Hymenoptera (van Frankenhuyzen, 2009).
Since the first descriptions of transgenic plants containing Bt genes (Höfte et al., 1986) (Vaeck et al., 1987), strategies to improve Bt transgene efficacy have been employed. Native cry genes from Bacillus thuringiensis are often AT-rich at the DNA sequence level, which is rarely found in plant exons. Strategies involving changes in codon usage and therefore mRNA stability and/or translational efficiency, but not amino acid sequence, have been demonstrated to enhance the expression of cry genes in transgenic plants (Perlak et al., 1991). Potatoes transformed with a cry3A gene modified in this way were shown to be resistant to CPB (Perlak et al., 1993). Potato plants expressing the Cry3A toxin were released by Monsanto under the name ‘NewLeaf’ (Hoy, 1999). They were commercially available in the USA from 1996 to 2000 and provided good CPB control, but were later discontinued following perceived concerns from consumers, marketing issues and the introduction of a novel insecticide that controls both beetles and aphids (Shelton et al., 2002). To engineer resistance to CPB in potato, Cry3B has also been shown to be effective (Arpaia et al., 2000).
A class of genes that are particularly active against Lepidopteran insects, such as PTM, are the Bt genes that belong to the cry1 class. Many such cry genes, when expressed in potato plants, have been shown to be highly effective in controlling PTM. Among them are cry1Ab2 (Chakrabarti et al., 2000), cry1Ab5 (Cañedo et al., 1999; Rico et al., 1998), cry1Ab6 (Jansens et al., 1995) cry1Ac1 (Ebora et al., 1994), cry1Ac9 (Davidson et al., 2002a, 2004; Meiyalaghan et al., 2006b), cry1Ia1 (Douches et al., 2011; Li et al., 1999), cry1Ba1 and cry1Ca5 (Meiyalaghan et al., 2006b). In addition to the cry1 class, success has been achieved with the cry9Aa2 gene (Jacobs et al., 2009; Meiyalaghan et al., 2005, 2006a).
Chimeric Bt cry genes have been developed to act on a wider spectrum of insect species. A hybrid Bt gene comprising domains I and II from cry1Ba and domain II of cry1Ia was developed, and transgenic potato plants were resistant to insect pests from two different orders—Coleoptera (CPB) and Lepidoptera (PTM and European corn borer) (Naimov et al., 2003).
Concerns have been raised about the development of resistance by the target pest to Bt-transgenic crops (Ferré and Van Rie, 2002; Gassmann, 2012; Gould, 1998; Roush and Shelton, 1997; Tabashnik et al., 1994). Changes in toxin binding sites in insects is the most commonly occurring resistance mechanism against CRY proteins (Ferré and Van Rie, 2002) and have occurred in Cry3Aa-resistant CPB (Loseva et al., 2002). It is therefore important to manage the release of the newly developed transgenic crops simultaneously with strategies to manage the development of resistance to CRY proteins.
Many approaches are possible for the introduction of multiple genes into a plant genome. These approaches are generically referred to as ‘transgene pyramiding’ (Berger, 2000; Halpin, 2005) and have the goal of preventing or delaying the development of insect resistance to transgenic crops. The simplest approach is by sexual crossing of transgenic lines to achieve the desired pyramided loci for multiple traits such as pest and disease resistance. However, sexual hybridization to pyramid transgenes is unsuitable in clonal crops such as potatoes, and it is necessary to pyramid transgenes in potato either by a simultaneous transformation strategy with multiple genes or by a sequential re-transformation strategy into an existing transgenic line using different selectable marker genes (Meiyalaghan et al., 2010). Pyramiding of the cry1Ac9 and cry9Aa2 genes has been achieved in potato, and although no lines expressing both cry genes exhibited any greater resistance to PTM larvae over that previously observed for the individual genes (Meiyalaghan et al., 2010), it is anticipated that these lines will permit more durable resistance by delaying the opportunities for PTM adaptation to the individual cry genes.
In summary, transformation of potatoes with a range of Bt-based transgenes has proven to be a highly successful approach to controlling insect pests of potato.
Intragenics and cisgenics
Despite the rapid global adoption of GM technology in agricultural crops (James, 2010) including potato, many concerns have been raised about the use of GM crops in agricultural production (Conner et al., 2003; Nap et al., 2003; Ryffel, 2010; Singh et al., 2006). One of the main underlying sources of concern involves the transfer of genes across very wide taxonomic boundaries, for example the insertion of bacterial genes into plant genomes (Lammerts van Bueren et al., 2007; Macer et al., 1991; Nuffield Council on Bioethics, 1999).
However, the transfer of genes between plants of the same species does not raise the same ethical concerns as the transfer of genes between species (Holme et al., 2013; Jochemsen, 2008). To help address concerns about GM technology, transformation vectors that consist of plant-derived DNA have been developed for Agrobacterium-mediated transformation. Agrobacterium-mediated plant transformation is well defined (and reviewed elsewhere, e.g. Gelvin (2003); Zupan and Zambryski (1995)). Two strategies that have been developed to achieve the transfer of plant DNA within a species are cisgenic (Schouten et al., 2006) and intragenic approaches (Barrell et al., 2010; Conner et al., 2007). Agrobacterium is used to transfer the intragenic or cisgenic genes and T-DNA into the plant host. The main difference between the two strategies lies in the use of bacterial DNA-derived T-DNA to host the gene being transferred (cisgenics), versus the use of plant-derived sequences that function as T-DNA and harbour plant-derived genes of interest (intragenics).
Any gene from a wild relative that confers a trait of interest has the potential to improve elite potato germplasm through cisgenics/intragenics. For example, resistance to late blight (P. infestans) has been reported in a number of wild relatives of potato. Genes conferring resistance to late blight have been identified from S. bulbocastanum (Song et al., 2003; van der Vossen et al., 2005) and S. demissum (Huang et al., 2005). Park et al. (2009) propose these resistance (R) genes as ideal candidates for the generation of cisgenic potatoes.
Gene silencing and transformation
Since the discovery of gene silencing in plants (Voinnet and Baulcombe, 1997), RNAi has been used down-regulate genes in potato to analyse gene function (Gonzalez-Schain et al., 2012; Hofius et al., 2004; Roumeliotis et al., 2013) as well as targeting specific traits of interest, for example reduced cold-induced sweetening and increased carotenoid content (Bhaskar et al., 2010; Chen et al., 2008; Giuliano et al., 2006). The specific down-regulation of genes with the aim of improving potato traits using RNAi has also been achieved using intragenics/cisgenics. Reduced acrylamide content on cooking was achieved by limiting asparagine formation with a tuber-specific RNAi construct against the potato asparagine synthetase-1 gene (Chawla et al., 2012).
As more functional analyses of genes are carried out, the specific regulation of genes through RNAi type approaches will expand the possibilities for molecular breeding in potato.
Genetics and genomics
Molecular technologies have huge potential for speeding up the process of conventional plant breeding. Identification of naturally existing allelic variation at the molecular level can provide a powerful tool to accelerate the process of breeding for improved cultivars. The association of a genotype with a phenotype of interest allows genetically elite plants to be identified early in plant growth before the phenotype can be observed. The ability to identify elite plants and discard nonelite plants saves both time and money in the process of plant breeding. In this section, we review the various approaches and technologies that have been used to characterize loci, (candidate) genes and alleles in potato, associating phenotype with genotype. We also discuss current genomics-based technologies and their potential for application in marker-assisted breeding of potato.
Traditionally, linkage mapping has been the most commonly used way to correlate natural variation in phenotype with genotype (Myles et al., 2009). Genetic mapping in cultivated potato has been hindered by its complex genetics. Most cultivars and breeding lines are autotetraploid and carry a high genetic load (Milbourne et al., 2007). Tools such as (homozygous) mutant lines, recombinant inbred lines and near-isogenic lines are not available in potato. Diploid lines derived from tetraploid S. tuberosum, and diploid (wild) species of potato have been used over recent decades to unravel the genetics of traits. In potato, this has been performed predominantly in segregating diploid F1 mapping populations, established using bi-parental crosses of heterozygous lines. Early examples of diploid linkage maps include Bonierbale et al. (1988) who took advantage of the high level of synteny between potato and tomato and used tomato restriction fragment length polymorphism (RFLP) markers for map construction; Gebhardt et al. (1989) who used potato RFLP markers; and Jacobs et al. (1995) who combined molecular (RFLP) markers with morphological traits in one genetic map.
RFLP-based genetic linkage maps were soon followed by potato maps containing amplified fragment length polymorphism (AFLP, Vos et al., 1995) markers (van Eck et al., 1995). Randomly amplified polymorphic DNA (RAPD) markers (Williams et al., 1990) combined with bulked segregant analysis (BSA; Michelmore et al., 1991) have been used to identify DNA sequences linked to major gene traits of interest in potatoes (e.g. Jacobs et al., 1996). However, RAPD markers have a low reproducibility and have been superseded by more reliable markers such as SSR (Simple Sequence Repeats, or microsatellites). These markers were first reported in potato by Provan et al. (1996) and have been used since for locating SSR markers on linkage groups using segregating mapping populations, as well as for identifying and fingerprinting potato germplasm accessions and cultivars (Feingold et al., 2005; Ghislain et al., 2004; Milbourne et al., 1998; Reid et al., 2011). Studies using Diversity Array Technology (DArT; Jaccoud et al., 2001) markers in potato have been published only recently; Śliwka et al. (2012) used DArT markers to map Rpi-mch1 a late blight resistance gene.
High-resolution melting analysis (HRM) is a technology that discriminates amplicons of alleles with different haplotypes (one or multiple SNPs) and/or can be used to detect mutations (Wittwer et al., 2003). It makes use of the small differences in melting temperatures of amplicons of double-stranded DNA molecules with one or multiple SNPs. It can be performed as a simple closed-tube assay, on DNA amplicons post-PCR without the need for separation or processing of the samples. HRM mapping relies on prior knowledge of allele sequences and haplotype variation. Although developing HRM assays capable of distinguishing all alleles in tetraploid potato can be time-consuming, HRM has been demonstrated as an efficient genotyping system. De Koeyer et al. (2010) developed HRM assays for five molecular markers/candidate genes for genotyping and variant scanning in diploid, as well as tetraploid potato and demonstrated that HRM-based candidate gene analysis efficiently provides information on allele dosage and discriminates different haplotypes. HRM technology was also successfully applied to examine the effect of allele dosage of the zeaxanthin epoxidase gene (Zep1, a recessive gene) on total carotenoid content in yellow-fleshed tetraploid potato germplasm (McCord et al., 2012).
A more direct approach to understanding the genetic inheritance of a trait of interest is the use of candidate genes. Rather than using anonymous molecular markers, genes thought to play a role in the trait of interest are investigated as candidate markers for the trait. This requires prior knowledge of the biochemical or physiological processes and pathways involved in determining a phenotype, as well as the gene sequences underpinning these processes and pathways, and can result in ‘perfect markers’ where no recombination occurs between the marker and the trait. Chen et al. (2001) created a molecular function map for carbohydrate metabolism and transport comprising genes involved in carbohydrate metabolism.
Quantitative traits refer to traits that are often attributed to more than one gene controlling or influencing the observed phenotype. Quantitative trait loci (QTL) are regions of the genome that contain gene(s) influencing the phenotypic expression of the trait. In potato, progeny lines from a biparental cross segregating for the trait of interest are assessed, and markers of choice are used to genotype individuals in the population. Numerous examples exist for QTL mapping in potato in both diploid and tetraploid populations. Early QTL mapping studies include using RFLP and RAPD markers to determine QTL for chip colour and tuber dormancy (Freyre et al., 1994), and RFLP markers to identify QTLs for resistance to Phytophthora infestans (Leonards-Schippers et al., 1994). More recently, QTL analysis in combination with a candidate gene approach was successfully used by Werij et al. (2012) to analyse the genetic basis of various tuber quality traits in a diploid mapping population.
Association mapping, also known as linkage disequilibrium mapping, identifies loci involved in the inheritance of complex traits by determining whether a statistically significant association exists between the genotype at a locus and the phenotype (Mackay and Powell, 2007). Association genetics is not limited to bi-parental crosses and can be applied in collections of breeding lines and cultivars. This is an advantage of association mapping over (QTL) linkage mapping, where the generation of segregating populations with large numbers of progeny for analysis is required. In addition, a larger pool of alleles across a genetically diverse range of lines can be assessed. Examples in potato include associations between tuber quality traits and AFLP markers (D'hoop et al., 2008), associations between candidate gene alleles and cold-induced sweetening of potato tubers (Baldwin et al., 2011) and the association of various combinations of candidate gene alleles and their positive and/or negative effects on tuber quality (Li et al., 2013). In recent years, as larger numbers of molecular markers have become available (e.g. Single Nucleotide Polymorphisms, SNPs), association mapping across the entire genome has become feasible (Genome-wide Association Studies, GWAS; reviewed by Morrell et al. (2012).
Marker-assisted selection (MAS) and potato breeding
The relative lack of implementation of molecular markers in tetraploid potato breeding programmes compared with some other crops is mainly due to the high level of natural allelic variation in potato, caused by the autotetraploid nature of cultivated potato and its tetrasomic inheritance (Luo et al., 2001). This high level of allelic variation hampers the ability to transfer markers across from mapping populations to breeding lines, and hence, marker validation in breeding germplasm is critically important (Milczarek et al., 2011). Because mapping in tetraploid potato is far more complicated than in diploid potato, it has frequently been restricted to regions of the genome containing a trait of interest (examples include Meyer et al. (1998); Bradshaw et al. (2008) and Bryan et al. (2004)). In addition, computer programs to assist in the genetic mapping of traits in an autotetraploid species need to eliminate many markers and/or marker alleles from analyses as the complete allelic composition and dosage remain elusive (Hackett and Luo, 2003). An example of the successful development of marker-assisted breeding in tetraploid potato is described by Moloney et al. (2010) who selected for resistance to the potato cyst nematode Globodera pallida. Gebhardt et al. (2006) developed potato clones with multiple pathogen resistance traits by applying PCR-based markers to combine Ryadg (resistance to PVY), Gro1 (resistance to the nematode Globodera rostochiensis) and Rx1 (resistance to potato virus X), or Sen1 (resistance to potato wart, Synchytrium endobioticum).
The genome sequence of potato (The Potato Genome Sequencing Consortium, 2011) was determined using a homozygous doubled monoploid (DM1-3 518 R44 or ‘DM’; (Paz and Veilleux, 1999), as well as a heterozygous diploid line (RH89-039-16 or ‘RH’; van Os et al. (2006)). The elucidation of the reference potato genome, including the annotation of around 39,000 protein-coding genes, has opened up opportunities to rapidly identify candidate genes in regions associated with a trait of interest. For example, the identification of both the StSP6A gene for tuber initiation (Navarro et al., 2011) and the StCDF1 gene responsible for plant maturity phenotype (Kloosterman et al., 2013) was greatly aided by the genome sequence. The genome sequence also provides a catalogue of candidate resistance genes in the potato genome, radically enhancing our ability for rapid discovery and introgression of R-genes in potato (Jupe et al., 2012; Lozano et al., 2012). The targeted re-sequencing of the many wild species of potato that harbour resistances to the major pests and pathogens of potato should enable identification of a wide array of valuable resistances for breeders.
A reference genetic map using a mapping population (‘DMDD’) derived from the doubled monoploid ‘DM’ was developed to assist with the anchoring of the genome sequence scaffolds to chromosomal positions (The Potato Genome Sequencing Consortium, 2011). This genetic map contains SNP markers as well as SSR and DArT markers. Genome sequences were anchored to the 12 linkage groups using a combination of in silico and genetic mapping data.
The single nucleotide polymorphism (SNP) frequency is very high in the potato genome (The Potato Genome Sequencing Consortium, 2011). A SNP chip based on the ‘DM’ genome sequence has been developed and contains 8,303 SNPs, including many targeted to candidate genes (Hamilton et al., 2011). The positions of these SNPs on the ‘DM’ genome are known, which allows for rapid identification of genomic regions of interest. The first genetic maps based on diploid biparental populations using the SNP chip include over 4,400 markers and refined the anchoring data of the potato genome sequence (Felcher et al., 2012). The high frequency of SNPs in the potato genome implies that selection for favourable alleles based on one single SNP is unreliable because it may not in all cases be indicative of the desired phenotype. Effective selection is more likely to be based on haplotype selection, targeting a combination of several SNPs in one gene. This requires knowledge of the various alleles and involves (re)sequencing of all possible alleles and/or genome resequencing of lines to ensure all allelic/haplotype variation is represented. A major hurdle in the analysis of SNP chip data in tetraploid potato is the difficulty of scoring heterozygous allele dosage (Douches et al., 2012; Hirsch et al., 2013; Voorrips et al., 2011). Software such as GenomeStudio (Illumina) for analysis of SNP data was originally developed for diploid species and is currently unable to differentiate the simplex (AAAB, ABBB) and duplex (AABB) heterozygous genotypes. Development of experimental and computational methods for haplotype estimation in polyploid species is an important goal. In addition to natural allelic variation, presence/absence variation (PAV, visible as ‘null alleles’) was found to be very common in potato (The Potato Genome Sequencing Consortium, 2011), and this will present additional challenges for application of marker-assisted selection. In addition, gene copy number variation has been shown recently to vary markedly between different potato cultivars, further highlighting the complex nature of potato genetics (Iovene et al., 2013). As well as being powerful tools for gene discovery in their own right, genomewide assays will provide immediate benefit to plant breeders by enabling the development of robustly unique marker haplotypes spanning QTL regions, which will be useful in both introgression breeding and whole-genome approaches such as genomic selection (Morrell et al., 2012). The availability of a genomewide marker set polymorphic in elite germplasm will make it possible to genotype increasing numbers of cultivars and breeding clones and will be a valuable tool for advancing whole-genome selection in potato breeding.
In addition to the potato genome sequence, RNA sequence data from 32 ‘DM’ and 16 ‘RH’ libraries representing all major tissue types, developmental stages and responses to abiotic and biotic stresses were generated (The Potato Genome Sequencing Consortium, 2011). These provide a valuable resource that determines the expression profiles of genes of interest in different tissues, stages of growth and in response to different growing conditions. For example, Massa et al. (2011) identified both tissue-specific gene expression profiles (including tuber-specific expression) and genes with condition-restricted expression.
With the reduction in cost of sequencing and the concomitant increase in data output per experiment, genotyping-by-sequencing (GBS Elshire et al., 2011) is becoming feasible for species with a high level of diversity. Reduced representation can be achieved by sequencing libraries digested with methylation-sensitive restriction enzymes, so that gene-rich regions are targeted. Alternatively, targeted resequencing of preselected genome regions can be achieved through sequence capture approaches as recently described by (Uitdewilligen et al., 2013).
The utility of the potato genome sequence for genomics-assisted breeding strategies will probably be realised in two distinct phases. Firstly, the sequence will be a powerful tool for gene discovery and placing any gene sequence into its genetic and genomic context. In the longer term, the identification of genes responsible for key agronomic traits coupled with the description of their allelic variation and their effect on the phenotype of the plant will afford breeders precision in grouping complementary alleles that will maximize the effect on the phenotype of the resulting breeding line and ultimately the cultivar.
The ability to mine an entire genome sequence is the ultimate tool for molecular breeding strategies. Many of the traits of interest to plant breeders are quantitative in nature. Even with the availability of the genome sequence, SNP chips and GBS, the lead time for the comprehensive genetic dissection of these traits may be several years. The overwhelming amount of sequence data available to us in the near future will overshadow the amount of accurate and reliable phenotypic data necessary to advance the potato breeding efforts for traits of interest. Phenotyping has already become the limiting factor in the exploration of the genetic potential of potato.