Back to the wilds: Tapping evolutionary adaptations for resilient crops through systematic hybridization with crop wild relatives^†

All domesticated species, both plants and animals, are impacted in unintended, often negative ways during domestication and breeding (e.g., Ladizinsky, 1985; Spillane and Gepts, 2001; Hyten et al., 2006; Taberlet et al., 2008; Gross and Olsen, 2010; Meyer et al., 2012; Olsen and Wendel, 2013). In particular, many crops lack genetic diversity and possess properties that reduce fitness in the natural environment. This problem derives both from demographic processes (e.g., genetic drift, population bottlenecks) and from changes in the nature of selection during breeding and cultivation that elevate the frequency of alleles with unique value in the agricultural environment and that permit the persistence of deleterious alleles (e.g., through selection trade-offs and selection relaxation) (Olsen and Wendel, 2013). The combination of the loss of adaptive alleles through drift and fixation of deleterious alleles through altered selection necessarily constrains our ability to expand the cultivation of domesticated species into environments beyond those in which domestication occurred, e.g., into more extreme climates, into marginal soils, into degraded agricultural landscapes, or into “sustainable” systems with reduced agricultural inputs. As part of this special issue, “Speaking of Food,” we argue that there is a need for systematic efforts to introgress broad subsets of wild relative diversity into our crop plants to incorporate the range of useful adaptations for disease resistance, abiotic stress tolerance, and other agronomic challenges that are required in order to increase the resiliency and productivity of agriculture in the 21st century. Here we review the ecological and evolutionary literature on the effects of hybridization to show the capacity of hybridization to generate phenotypic novelty, then detail examples of hybridization of crop wild relatives with domesticated plants.

Wild species have an important role to play in meeting the challenges for 21st century agriculture, which must become increasingly efficient to meet humankind's demand for a more plentiful and nutritious food supply (e.g., Tanksley and McCouch, 1997; Pimentel et al., 1997; Haussmann et al., 2004; Maxted and Kell, 2009; Tester and Langridge, 2010; Ford-Lloyd et al., 2011; McCouch et al., 2013). Such challenges are particularly acute in the developing world, where extreme climatic conditions, marginal soils, and reduced inputs limit productivity, create increased risk, and diminish livelihoods through reduced income and malnutrition. Yet the impact of a properly implemented and well-used resource of wild germplasm would extend beyond the developing world. Many of the crop phenotypes important to cultivation in the developing world (e.g., tolerance to heat and drought, reduced dependence on inputs [e.g., nitrogen, phosphate, pesticides, water], and increased seed nutrient density) are also key to meeting the global demand for crops that incorporate traits for climate-resilience, increased sustainability, and increased nutritional value.

The potential for genetic gains from use of crop wild relatives is well documented (e.g., Pimentel et al., 1997; Tanksley and McCouch, 1997; Maxted and Kell, 2009). Nevertheless, crop wild relatives have been used sparingly and typically in an ad hoc manner in many crop breeding programs (Hajjar and Hodgkin, 2007; Maxted and Kell, 2009; Brumlop et al., 2013). Impediments to the systematic use of wild material in crop improvement programs include the often poor agronomic performance of crop–wild hybrids and their immediate backcrosses, and the labor intensive process of constructing large-scale, representative populations that are suitable for phenotypic assessment. For perennial crop species, which can take many years to reach reproductive age, such repeated backcrossing is prohibitively time consuming. Moreover, for many crops, the use of wild germplasm is further constrained by the limited state of international germplasm collections. Compounding the problem, many crop wild relatives are at risk of extinction from habitat loss, habitat fragmentation, changing landuse and management practices, climate change, and introgression from agricultural relatives (Ford-Lloyd et al., 2011).

The fact that crop wild relatives are under-used in crop improvement programs presents an opportunity. One can restructure germplasm resources, essentially de novo, guided by appropriate ecological and population genetic theory; when properly implemented, such collections would represent a diversity of source habitats and encompass the breadth of segregating genetic variation and adaptations characteristic of the target species. Genomics, phenotyping, and computational approaches can subsequently be used to infer natural adaptations in situ, for example, based on knowledge of population structure, allele frequency, and recombination history, combined with knowledge about selective constraints in individual populations. Such analyses can motivate targeted phenotyping activities and ultimately nominate candidate genes for adaptative traits, leading to increased understanding of the autecology of crop wild relatives. In parallel to the analysis of gene function in situ, purpose-built populations that are hybrids between crops and their wild relatives provide powerful tools for trait dissection, and as such they become the vehicle by which the genetic (genomic) basis of valuable agronomic traits can be understood. Examples of such populations include nested association mapping (NAM) (e.g., Yu et al., 2008; McMullen et al., 2009) and multiparent advanced generation intercross (MAGIC) (e.g., Cavanagh et al., 2008) panels, to which the logic of association genetics can be applied (e.g., Huang and Han, 2013; Korte and Farlow, 2013), as well as advanced backcross introgression lines (Tanksley and McCouch, 1997) that capture genome intervals and their adaptive traits from wild relatives within the essential crop genome. The value of combining ecology, population genetics, genomics, and phenotyping is well documented in model species, such as Arabidopsis, Drosophila, mice and maize (e.g., Yu et al., 2008; Ayroles et al., 2009; McMullen et al., 2009; Atwell et al., 2010, Tian et al., 2011; Flint and Eskin, 2012; MacKay et al., 2012; Korte and Farlow, 2013), but has not been used widely in support of crop species and their wild relatives (e.g., Huang and Han, 2013). To make the most of this approach, however, we must understand more about the complex effects of hybridization. To that end, we review examples of hybridization in several species, both intentionally produced and naturally occurring, to illustrate the gains that are possible.

THE IMPORTANCE AND PREVALENCE OF HYBRIDIZATION

One of the most reviewed and most debated sources of variation in sexually reproducing organisms is hybridization, or reproduction among members of genetically distinct groups (i.e., populations within species, or distinct but closely related species) (Ellstrand and Schierenbeck, 2000; Mallet, 2005; Soltis and Soltis, 2009; Abbott et al., 2013; Schumer et al., 2014). Hybridization of crops and their wild relatives has long been an important source of variation in breeding, despite its ad hoc application. We argue there is a need for systematic efforts to introgress a broad range of wild relative diversity into our crop plants, with the goal of creating a genetic toolbox from which natural adaptations for traits such as disease resistance, tolerance to climatic extremes (especially temperature and moisture), and productivity in otherwise marginal soils can be identified and deployed. First, we summarize the extensive literature that illustrates the potential of hybridization and introgression to generate phenotypic novelty, including both plant and animal examples. We then document that most examples of intentional introgression from wild relatives into cultivated species have focused on a narrow range of traits and a limited range of the variation present in crop wild relatives. Finally, we argue that a growing understanding of both the genetic architecture of domestication and the genomic consequences of hybridization makes it feasible to systematically introgress substantial amounts of the diversity present in wild relatives into cultivated genetic backgrounds. From systematic introgressions, it is feasible to quickly recover both wild relative stress tolerance and cultivated agronomic traits of interest through advance generation backcrosses (Tanksley and McCouch, 1997) and nested association mapping or multiparent advance generation intercross populations (e.g., Cavanagh et al., 2008; Yu et al., 2008; McMullen et al., 2009).

Hybridization occurs between individuals with varying levels of genetic differentiation and via multiple mechanisms, and it is therefore not surprising that such interbreeding events can have drastically different consequences (e.g., Barton and Hewitt, 1985; Mallet, 2005; Soltis and Soltis, 2009; Abbott et al., 2013). From reinforcement of isolating mechanisms, to low levels of genetic introgression between slightly divergent populations, to the formation of distinct hybrid species, hybridization is thought of as both a creative and a restrictive force in evolution (e.g., Anderson and Stebbins, 1954; Barton and Hewitt, 1985; Mallet, 2007; Genner and Turner, 2012; Abbott et al., 2013; Schumer et al., 2014). It is the potential for the production of novelty that makes hybridization such an intriguing—and potentially useful—phenomenon.

In some cases, hybridization can lead to saltational evolution (Mallet, 2007). In plants, this process often occurs via polyploidization, wherein an individual is produced that has the complete genomes of both parental species (Soltis and Soltis, 2009). While few hybrid animal species arise in this fashion, in plants this mode of saltational evolution is a common mechanism of hybrid speciation (Wood et al., 2009; Otto and Whitton, 2000). Thus, historical polyploid events are known to have played an important role in the evolution of angiosperms (Cui et al., 2006), with subsequent diploidization to the modern genomes. Similar but more recent polyploid events underlie the evolution of several modern crop species (see below and Appendix S1 [see Supplemental Data with the online version of this article]). In other instances, hybridization does not result in genome duplication, but leads to repeated rounds of natural backcrossing and selection, resulting in the introduction of genome segments that contain novel adaptive traits. Segmental introgressions have been important, for example, in the adaptation of highland maize varieties based on gene flow from highland-adapted wild species (Hufford et al., 2013).

Regardless of whether hybridization results in a new lineage, gene flow between divergent lineages, or fusion of lineages, it can generate multilocus genotypes that are not present in either parent, leading to offspring with particular traits that exceed those of either parental population. In fact, this effect of transgressive segregation is the rule rather than the exception (Rieseberg et al., 1999). Furthermore, the very nature of hybridization may predispose hybrid lineages to have novel traits through the restructuring of genetic interactions and by altering predispositions for reproductive isolation (Seehausen, 2013). Although hybrids tend to occur in small numbers and may often be maladapted, strong selection and genetic drift can lead to colonization of new niche space. Given these facts, it is no surprise that hybrid vigor has been cited as an impetus for the evolution of invasiveness (e.g., Ellstrand and Schierenbeck, 2000) and for adaptive radiations, during which phenotypic novelty emerges at a rapid rate (e.g., Seehausen, 2004; East African Great Lake cichlid fish, Joyce et al., 2011; Genner and Turner, 2012; Keller et al., 2013; Hawaiian silverswords, Baldwin, 1997; Barrier et al., 1999), and it is clear why hybridization is such a powerful tool for the improvement of crops.

Examples of hybridization are widespread in plants and have become increasingly common in animals (e.g., Mallet, 2005, 2007; Abbott et al., 2013). Although the rate of hybridization among related species tends to be low, the number of species that hybridize is relatively high (Mallet, 2005). An estimated 10% of all animal species and 25% of all plant species undergo hybridization (Mallet, 2005), and genome-wide scans of an increasingly large number of organisms reveal that their genomes are subject to introgression (e.g., Baack and Rieseberg, 2007; Arnold and Martin, 2010; Green et al., 2010; Abbott et al., 2013).

Hybridization is known to be common in some groups of animals—for example, 75% of the ducks of the British Isles (Gillham and Gillham, 1996; Mallet, 2005) and over 25% of all tit species (Paridae) can hybridize (Harrap and Quinn, 1996). In recent years, the prevalence of introgressive hybridization in animals has become even more apparent, with examples arising from across the animal kingdom, including both recent and historical hybridization in the primate family (e.g., Pastorini et al., 2009; Green et al., 2010; Zinner et al., 2011). Interestingly, an estimated 20% of the Neandertal genome survived in modern humans through hybridization and introgression, with 1–3% admixture on an individual basis (Vernot and Akey, 2014). In the past decade, there have emerged several examples of homoploid hybrid speciation in animals, where hybridization between two species has led to a third with a distinct morphology or niche (Mallet, 2007). These hybridization events can lead to important novel phenotypes, such as the emergence of the specialized feeding forms of tephritid fly Rhagoletes mendax ×zephyria on invasive honeysuckle plants (Lonicera spp.) arising from hybridization between parental species specialized on blueberry (R. mendax) and snowberry (R. zephyria) (Schwarz et al., 2005). As is the case in plants, introgression between domesticated animals and their wild relatives is known to occur (e.g., pigs and wild boars; Goedbloed et al., 2013), and these wild relatives are considered valuable genetic resources for livestock improvement efforts (Taberlet et al., 2008). While this paper aims to explore the capacity for hybridization to generate phenotypic novelty in crop plants, examples from the animal kingdom demonstrate the widespread prevalence of hybridization and suggest that livestock may also benefit from similar genomically based breeding programs.

Naturally occurring hybridization in plants has been known since the time of Linneaus (e.g., Gustafsson, 1979) and featured prominently in On the Origin of Species (chapter 8) and perhaps in Darwin's conception of species (e.g., Kottler, 1978). Ever since, hybridization has been an often discussed, reviewed, and debated topic in plant evolution (e.g., Stebbins, 1950; Abbott, 1992; Arnold, 1992; Rieseberg, 1995, 1997; Rhymer and Simberloff, 1996; Levin et al., 1996; Ellstrand and Schierenbeck, 2000; Barton, 2001; Seehausen, 2004; Soltis and Soltis, 2009; Arnold and Martin, 2010; Abbott et al., 2013; Schumer et al., 2014). Cases of natural hybridization in plants, both homoploid and polyploid, give us insight into the range of phenotypic and ecological effects hybridization can have, and can act as models for harnessing the power of hybridization in agriculture.

A striking and well-known example of homoploid hybrid speciation is that of Helianthus sunflowers of the United States. Rieseberg (1991) identified three natural homoploid hybrid species, H. anomalus, H. deserticola, and H. paradoxicus, as the offspring of H. annuus and H. petiolaris. These hybrid Helianthus are exemplars of transgressive segregation, highlighting the expanded potential of hybrid species, in this case through colonization of extreme habitats where neither parental species can survive (sand dune, desert floor, and salt flats, respectively) (Rieseberg et al., 2007). Additionally, these species show that repeated hybridization events between the same parental species can have vastly different outcomes. As Arnold et al. (2012) discussed, work investigating another homoploid hybrid complex, the Louisiana iris has demonstrated the variability of hybrid fitness, which is dependent on both genotype and environment. These lessons from natural hybrids bear particular relevance to crop improvement efforts that aim to produce crops adapted to changing climatic conditions.

AGRICULTURE AND HYBRIDIZATION

In agriculture, hybridization between crops and wild relatives has long been a major research focus. One goal is to introgress adaptive traits from wild relatives into cultivated forms as part of breeding programs, which we detail below. A second focus has been on crops of hybrid origin. A number of crops have formed by way of hybridization, including many of polyploid origin (e.g., Nagaharu, 1935; Udall and Wendel, 2006; Vaughan et al., 2007; Appendix S1). Meyer et al. (2012) cite 37 of 203 crops as having a change in ploidy as part of their domestication, or about 15%, similar to estimates of speciation events across angiosperms involving shifts in ploidy (Wood et al., 2009). Some notable examples of polyploid crops include wheat (Peng et al., 2011), bananas (Heslop-Harrison and Schwarzacher, 2007), strawberries (Folta and Davis, 2006), and vegetable and oilseed brassicas (Prakash et al., 2011). Some more recently developed crops involve hybridization events that have occurred far beyond the region of domestication and rather recently, such as the formation of grapefruit in Barbados in the 18th century as a homoploid hybrid of the sweet orange (Citrus sinensis) and shaddock (C. maxima) (Kumamoto et al., 1987). A third area of intense recent interest has been in the escape of transgenes from genetically modified crops into weedy wild relatives. This field has been growing and has been reviewed several times (Pilson and Prendeville, 2004; Armstrong et al., 2005; Ellstrand et al., 2013). Although we do not aim to thoroughly review this topic, we mention it due to its importance and the way it complements intentional introgression from wild into cultivated backgrounds. For example, studies by Mercer and colleagues of crop introgression into wild (weedy) populations found shifts in growth rate and flowering time that likely impact the ability of transgenes to persist in populations (Mercer et al., 2006a, b, 2007). Vacher et al. (2011) found similar results from introgression of genetically modified crops into populations of weedy Brassica crop relatives. Snow and colleagues showed that the transgenes introgressed into weedy relatives can persist in weedy populations (e.g., Snow et al., 2010). Importantly, our ability to detect crop–wild hybridization at both a fine scale and broad scope is improving as the costs of sequencing decline. For example, Hufford et al. (2013) found widespread genomic signatures of crop and wild alleles moving quite frequently between cultivated maize and wild teosinte in southern Mexico. Such gene flow is likely more common than previously appreciated in crops that are grown in proximity to wild relatives, even those that primarily self-pollinate.

HYBRIDIZATION OF CROPS AND THEIR WILD RELATIVES TO CONFER ADAPTIVE TRAITS

Early examples of targeted introgression can be traced to the work of Vavilov (1926, 1951). Since that time, crop wild relatives have been used to confer adaptive traits in a variety of crops, with the most widespread use occurring in a limited number of annual crops such as wheat, rice, barley, cassava, potato, and tomato. Maxted and Kell (2009) report 291 articles that identify and attempt to transfer useful traits from 185 wild relative taxa to 29 crop species. More than 50% of these traits are for disease and pest resistance, with traits for abiotic stress tolerance accounting for an additional 10–15%. Yield improvement also accounts for perhaps another 20%, although this can be hard to differentiate from other categories. Another review by Brumlop et al. (2013) of 104 molecular assisted breeding papers published from 1995–2012 found that approximately 74% of these studies were focused on introgression of traits that confer disease resistance, with the rest focused on traits involved in abiotic stress tolerance, improved yield, and growth habit. Although generations of breeders have performed crop–wild crosses across a large number of taxa and involving thousands of individual crosses, these crosses are still limited in comparison to the range of variation present in wild relatives of all of our cultivated plant species. Records of pedigrees for many crops show that most breeding programs have had a narrowing of the crop genetic basis over the past few decades as breeders tend to reuse a few favored parents to establish new elite varieties (e.g., Kumar et al., 2003). Efforts to improve plant health and production through the use of introgression from crop wild relatives have substantial economic value, which was estimated at nearly $115 billion globally over 15 yr ago (Pimentel et al., 1997); we can only assume that this value has risen since that time. Given the many challenges posed by climate change, the scale of usage of crop wild relatives must be increased dramatically to keep up with changing conditions.

A noteworthy example of targeted introgression of a wild relative comes from common bean, Phaseolus vulgaris (reviewed by Acosta-Gallegos et al., 2007). Breeders have successfully introgressed genes conferring resistance to insects (e.g., bruchid beetle seed predators, Apion pod weevils) and pathogens (e.g., Fusarium), as well as higher nitrogen, iron, and calcium seed content from existing collections of wild Phaseolus (reviewed in Acosta-Gallegos et al., 2007). These efforts have contributed to both higher yields and improved nutritional quality and have also lessened the environmental impact of crop production by facilitating reduced pesticide, herbicide, and fertilizer use. While these gains are significant, the collection of wild relatives in Phaseolus was not built systematically or with an intention of building a resource that reflects the complete range of habitats in which wild Phaseolus thrives. Without such a systematic search, it may be difficult to find the full range of alleles for particularly valuable traits, like acrelin insecticidal proteins in Phaseolus, which occur at low frequencies in natural populations of crop wild relatives (Acosta-Gallegos et al., 1998). Furthermore, Acosta-Gallegos et al. (2007) argued persuasively that the wild diversity can be better used if converted or incorporated (sensu Simmonds, 1993) into the common bean breeding pool. This goal is most easily accomplished with marker assisted breeding or genomic selection (e.g., Nakaya and Isobe, 2012) and an understanding of the genetic basis of domestication traits (in Phaseolus, these traits include pod shattering, growth habit, and photoperiod insensitivity) that must be recovered in crop–wild crosses to create a new cultigen that is suitable for future agricultural conditions.

The example of common beans extends to other crops critical to food security in the developing world, such as the 19 crop species for which the Consultative Group on International Agricultural Research (CGIAR) coordinates breeding efforts. Hajjar and Hodgkin (2007) reviewed the use of crop wild relatives by CGIAR in 16 of their mandate crops. Through both an examination of literature and interviews with CGIAR breeders and germplasm managers, the authors extensively surveyed breeder usage in the CGIAR system and uncovered patterns not obvious from published literature alone. As in the literature-based reviews by Maxted and Kell (2009) and Brumlop et al. (2013), they found that over 80% of usage has been for disease and pest resistance. However, in 13 of the 16 mandate crops some traits besides resistance have been successfully transferred from crop wild relatives, representing a rise in the usage of wild relatives in breeding since an earlier review (Prescott-Allen and Prescott-Allen, 1986). This trend toward greater usage of wild relatives is consistent with the broader breeding community (Dulloo et al., 2013). However, these reviews illustrate that the majority of examples of crop–wild hybridization in breeding have been ad hoc in their usage of wild germplasm. None of these efforts have screened existing wild relatives for more than a few traits, and none have used crop wild relative collections that were systematically built to represent the range of adaptations found in natural populations.

OBSTACLES TO THE USAGE OF CROP WILD RELATIVES

In addition to highlighting the potential benefits of crop wild relatives, the described studies also discussed obstacles that limit the use of crop wild relatives in breeding programs, including their poor agronomic performance (Haussmann et al., 2004). Poor performance can take many forms. For example, crop wild relatives often lack important domestication traits, such as shattering pods or shifted germination timing (e.g., Acosta-Gallegos et al., 2007), or broader environmental adaptations (e.g., Haussmann et al., 2004). In some crops, such as chickpea, phenological differences make the temperate wild relative unsuited to subtropical or tropical conditions (e.g., Abbo et al., 2003; Berger et al., 2006), and the same issues are at play for tropical crops grown in temperate regions, such as maize and common bean (reviewed by Jung and Müller [2009], Buckler et al. [2009], and Acosta-Gallegos et al. [2007]). It can be difficult to remove such undesirable traits from crop–wild hybrid lines. For attempts to introgress a targeted trait with a fairly simple genetic basis, such as a resistance gene, backcrossing can be time-consuming and difficult. Even after three generations, regions of a wild chromosome spanning many centimorgans may remain around an average selected gene (Stam and Zeven, 1981; Welz and Geiger, 2000; Haussmann et al., 2004). Linked regions that negatively influence agronomic performance, pleiotropy, and other complications make the task harder (e.g., Xu et al., 2006). Loci associated with domestication are similar to barrier loci (sensu Abbott et al., 2013), reducing gene flow between populations; these loci are central to the genetics of speciation and may reduce the fitness of hybrids. Although molecular-assisted breeding and increasingly genomic selection can be of great assistance if the candidate gene is known, these techniques remains time consuming (e.g., Young, 1999; Varshney et al., 2005; Xu and Crouch, 2008; Kumar et al., 2011).

Another obstacle in the use of wild relatives is their poor representation in international germplasm collections. Maxted and Kell (2009) estimated that only 2–6% of international germplasm collections are of crop wild relatives, with landraces and varieties making up the vast majority of accessions for most crops. Although collections for a few crops and their wild relatives are large, wild relatives of many crops have been poorly collected or have been almost ignored, and some such as faba bean even lack well-identified wild relatives (e.g., Kaur et al., 2014). Two striking examples with which we are familiar are grain legumes of considerable importance in the semiarid tropics and many temperate areas: chickpea and peanut. Berger et al. (2003) estimated that for the immediate wild ancestor of chickpea, Cicer reticulatum, the existing international collections of more than 150 named accessions stem from only 18 independent accessions; the large number of accessions counted in these collections appears to derive from proliferation of these original 18 accessions as distinct lineages. This practice grossly inflates the adequacy of the collection, because most accessions are redundant. For peanut, an allotetraploid with an A and a B genome, there is only a single individual available of the B genome parent, Arachis ipaensis in the USDA and ICRISAT collections, despite over 40 collecting trips organized by USDA and other collectors (Holbrook and Stalker, 2003). Moreover, peanut appears to derive from a single hybridization event, creating an unusually strong genetic bottleneck in the crop. To combat this genetic deficiency, synthetic allotetraploid hybrids have been created from related A genome accessions and the sole B genome representative (Fonceka et al., 2012). Many more Arachis species are poorly collected and at high risk of extinction (e.g., Jarvis et al., 2003). Yet even for well-collected crops like Phaseolus, collections of wild relatives are likely not geographically exhaustive; gap analyses still indicate regions and taxa that are underrepresented (e.g., Ramirez-Villegas et al., 2010). Assessments of the adequacy of current wild collections have commonly been based on the number of wild accessions in germplasm repositories. However, this measure often overestimates diversity in the collections because initially collected samples are generally assigned additional accession identifiers during distribution and evaluation. Reliance on numerical coverage in collections has also shifted the focus of future collection efforts to taxa with lower numeric or limited geographic representation while overlooking the inadequacies of current redundant collections. Furthermore, nearly all older collections of wild relatives have incomplete passport information and most have all of the seeds from a particular geographic location bulked into a single bag, making it difficult to impossible to determine patterns of within and among population variation in crop wild relatives (Greene and Hart, 1999). In addition to the inadequacies of many ex situ germplasm collections, many crop wild relatives occur in geopolitically unstable areas where collection has long been complicated, and where in situ conservation is at best challenging.

An additional obstacle to the use of wild relatives is the unpredictability of both a wild individual's phenotype under agronomic conditions and the phenotype of crop–wild hybrids. Phenotypes of wild individuals are often assessed in agricultural settings, a largely uninformative practice when the overall wild phenotype is specifically adapted for fitness in wild but not cultivated settings. For instance, when plant phenology (e.g., flowering time and/or vernalization) differ substantially between wild and cultivated material then phenotypic comparisons are problematic and it may be first necessary to “correct” the timing of development to cultivated set points before initiating phenotypic assessment. Further complicating the issue, genotype–environment interactions can make the phenotype expressed under agricultural conditions different from what it would be under natural conditions. Predicting the phenotypes of crop–wild hybrids also remains complicated. In very few cases, even for model organisms such as Drosophila or Arabidopsis, do we understand the genotype–phenotype map well enough to fully predict phenotypes of crosses or advanced introgressions. However, an important first step toward building this capacity for crop–wild hybrids is understanding the major loci that have been under strong artificial selection during domestication. For an increasing number of crops, major domestication loci have been identified (e.g., reviews by Doebley et al. [2006], Gross and Olsen [2010], Meyer et al. [2012]). In advanced backcross lines, breeders can recover crop alleles of the major domestication loci, speeding the recovery of the essential crop phenotype and retaining adaptive variation from the wild relatives.

A PROPOSAL FOR FUTURE WORK

We propose a multistep framework for utilizing naturally occurring variation in wild relatives of crops (Fig. 1). It is increasingly possible to digitize genotype–environment interactions in wild progenitor populations and from there predict the effect of wild alleles in cultivated backgrounds. We propose five steps to better use crop wild relatives.