Genomic Tools in Pea Breeding Programs: Status and Perspectives

Pea (Pisum sativum L.) is an annual cool-season legume and one of the oldest domesticated crops. Dry pea seeds contain 22–25% protein, complex starch and fiber constituents, and a rich array of vitamins, minerals, and phytochemicals which make them a valuable source for human consumption and livestock feed. Dry pea ranks third to common bean and chickpea as the most widely grown pulse in the world with more than 11 million tons produced in 2013. Pea breeding has achieved great success since the time of Mendel's experiments in the mid-1800s. However, several traits still require significant improvement for better yield stability in a larger growing area. Key breeding objectives in pea include improving biotic and abiotic stress resistance and enhancing yield components and seed quality. Taking advantage of the diversity present in the pea genepool, many mapping populations have been constructed in the last decades and efforts have been deployed to identify loci involved in the control of target traits and further introgress them into elite breeding materials. Pea now benefits from next-generation sequencing and high-throughput genotyping technologies that are paving the way for genome-wide association studies and genomic selection approaches. This review covers the significant development and deployment of genomic tools for pea breeding in recent years. Future prospects are discussed especially in light of current progress toward deciphering the pea genome.


INTRODUCTION
Thanks to significant technological breakthroughs, pea genetics is rapidly evolving from conventional to large-scale molecular-assisted approaches to uncover the molecular bases of important traits and enhance breeding. Several reviews have been published recently that dealt with pea systematics, seed quality, and breeding (Burstin et al., 2011;Dahl et al., 2012;Bohra et al., 2014;Arnoldi et al., 2015;Smýkal et al., 2015;Varshney et al., 2015;Warkentin et al., 2015). The present review focuses on the genomic toolkit that was developed recently in pea thanks in part to next-generation sequencing technologies. This includes transcriptome, genotyping, and mapping resources that will pave the way to renewed pea breeding programs.

ECONOMIC IMPORTANCE, NUTRITIVE VALUE, GROWING REGIONS
Pea (Pisum sativum L.) is a major cool-season pulse crop and an essential component of sustainable cropping systems (Nemecek et al., 2008;Duc et al., 2010;Jensen et al., 2012). Significant agroecological services linked with its ability to develop symbiotic nitrogen fixation as well as its role as a break crop for pest and pathogen pressure reduction have been described (Nemecek and Kägi, 2007;Hayer et al., 2010;Macwilliam et al., 2014). In 2013, the vegetable pea production amounted to 17.43 Mt worldwide (FAOSTAT) 1 and dry pea represented the third most important pulse crop production after common bean and chickpea with 11.16 Mt produced worldwide (FAOSTAT).
Pea seeds are an important source of proteins and provide an exceptionally varied nutrient profile (for a review, Burstin et al., 2011): major constituents are starch (from 18.6 to 54.1%) and proteins (15.8-32.1%), followed by fibers (5.9-12.7%), sucrose (1.3-2.1%), and oil (0.6-5.5%). Seeds also contain minerals, vitamins, and micro-nutrients such as polyphenolics, saponins, α-galactosides, and phytic acids whose health-promoting effects are being tested (Bastianelli et al., 1998;Mitchell et al., 2009;Dahl et al., 2012;Marles et al., 2013;Arnoldi et al., 2015). Peas enter in human nutrition in a wide diversity of forms: fresh seedlings, immature pods, and seeds provide a green vegetable, and whole or ground dry seeds are cooked in various dishes. High quality starch, protein, or oligoside isolates are being extracted from dry pea seeds and whole seed structural and functional characteristics have been assessed for food improvement (Brummer et al., 2015). Because dry seeds contain little anti-nutritional factors, they are also introduced as a protein source mainly in monogastric diets without affecting growth and production traits (Stein et al., 2006;Laudadio et al., 2012;Dotas et al., 2014). Pea hay is used as fodder in ruminant diets (Bastida Garcia et al., 2011).
Pea is mainly cultivated in temperate regions of the world on well-drained and fertile soils. However, being distributed over all continents, the pea production area is characterized by a large range of pedo-climatic conditions. Indeed, China is the largest producer of vegetable peas (10.60 Mt, FAOSTAT) followed by India (4 Mt). Canada is the main producer of dry peas (3.85 Mt) followed by China (1.6 Mt), the Russian Federation (1.35 Mt), USA (0.71 Mt), India (0.60 Mt), France (0.50 Mt), and Ethiopia (0.40 Mt).

VARIABILITY OF GERMPLASM USED FOR BREEDING
Pea was domesticated by Neolithic farmers in the Fertile Crescent some 10,000 years ago (Willcox et al., 2009;Weiss and Zohary, 2011;Smýkal, et al., 2014). Pea then spread rapidly toward south-west Asia, the Mediterranean basin, and Europe (Zohary, 1999). Probably linked with their large range of cultivation and the diversity of their use as food, feed, or fodder, pea landraces and varieties now exhibit an incredible diversity of forms and growing types, adapted to diverse environments, 1 FAOSTAT. http://faostat3.fao.org/. 10/06/2015. cropping systems, and end-uses (Burstin et al., 2015). This vast diversity of cultivated forms is the major reservoir for present crop improvement. Different types of pea varieties have been developed for vegetable pea production, varying at major genes controlling seed and plant traits. For example, wrinkled seeds are associated with significant changes of seed composition, linked with starch synthesis modification (Wang et al., 2003). Various types of dry peas are also available that differ by their cotyledon color, plant architecture, or flowering time. In addition to this cultivated reservoir of diversity, wild peas can be crossed with cultivated peas. Ben-Ze'Ev and Zohary (1973) showed that chromosomal rearrangements among accessions from the different Pisum species and subspecies could cause partial sterility in hybrids. Recently, Bogdanova et al. (2014) have identified a nucleo-cytoplasmic incompatibility between a P. sativum elatius accession and cultivated peas. However, within the Pisum genus, wild P. fulvum, wild subspecies P. sativum elatius, and P. sativum humile as well as P. abyssinicum, a taxon cultivated in Ethiopia, are in most cases inter-crossable with P. sativum sativum as long as the cultivated pea is used as female donor (Ben-Ze'Ev and Zohary, 1973;Ochatt et al., 2004). Different authors have thus used P. fulvum as well as wild P. sativum subspecies as a source of alleles for important breeding traits, such as resistance to various fungal diseases (Barilli et al., 2010;Fondevilla et al., 2011;Jha et al., 2012) or to Bruchus pisorum L. (Clement et al., 2009).

TARGET TRAITS AND ACHIEVEMENTS THROUGH CONVENTIONAL BREEDING
Continued grain yield improvement is necessary for pea to remain an attractive option compared to cereals and oilseeds in crop rotations. Improving yield involves addressing many biotic and abiotic stresses, using a large set of strategies including diverse germplasm as parents, making many crosses, selecting for major gene traits under conditions conducive to selection, and yield testing of a large number of breeding lines. These stresses are specific to each growing region and/or growing type. However, fungal diseases are the major biotic stress in most cases, followed by various insects, viruses, and parasitic plants such as broomrape. Drought and heat stress at flowering are the main abiotic stresses, while frost, salinity, and early season flooding are diversely important according to growing types. Important achievements were obtained in pea cultivars through conventional breeding over the past 20 years. Yield gains of approximately 2% per year have been achieved . Lodging resistance has been improved through deployment of the afila gene for semi-leafless type (Kujala, 1953;Goldenberg, 1965) and secondarily through selection for increased stem strength (Banniza et al., 2005). Powdery mildew resistance based on the single recessive gene er-1 (Harland, 1948) has been widely deployed. Partial resistance to the Ascochyta blight complex has been achieved through pyramiding of genes with minor effects (Kraft et al., 1998). Resistance to pea weevil (Bruchus pisorum L.) identified in the secondary gene pool (P. fulvum) (Clement et al., 2002) was transferred into cultivated pea through backcrossing (Clement et al., 2009;Aryamanesh et al., 2012). Cultivars adapted to winter sowing have been developed and deployed in Europe and north-west USA giving the potential for better yields because of a longer growing season, higher biomass production, and earlier maturity to avoid late season drought and heat stress (Hanocq et al., 2009). The introgression of the Hr allele which delays flower initiation until after the main winter freezing periods have passed (Lejeune-Hénaut et al., 2008) permitted to obtain some cultivars with notably improved winter hardiness. Field pea production for whole seed food markets requires appropriate seed visual quality. Quantitative inheritance, transgressive segregation, and moderately high heritability were observed for seed color, shape, and surface dimpling (Ubayasena et al., 2011) allowing for good progress in breeding. Seed protein concentration has been maintained in pea cultivars, even though overall seed yield has increased (Jha et al., 2013).
In some cases, achievements are at an earlier stage of deployment. Useful germplasms such as sources of resistance to various biotic and abiotic stresses have been identified and are currently being evaluated and introgressed. Research in France and USA has led to the identification and introgression of useful variation for resistance to Aphanomyces root rot; partial resistance controlled by several quantitative trait loci (QTLs) is being deployed (Pilet-Nayel et al., 2002McGee et al., 2012). Improved stress tolerance has been identified in landrace accessions for toxicity to boron (Bagheri et al., 1994), salinity (Leonforte et al., 2013a), iron deficiency (Kabir et al., 2012), and for heat tolerance during flowering (Petkova et al., 2009). Selection for major gene resistance to pea seedborne mosaic virus and potyviruses is now incorporated into breeding strategies (van Leur et al., 2007). Diversity in pea seed micronutrient concentration (Ray et al., 2014) and an approach to improving iron bioavailability for humans  have been described.
Future targets in pea breeding include (i) the optimization of pea interactions with Rhizobia, Mycorrhiza and other beneficial microorganisms in view of crop resilience against stresses, (ii) the adaptation of plant morphology and phenology to novel cropping systems, and (iii) the adaptation of seed composition to novel end-use application possibilities.

AVAILABLE GENOMIC RESOURCES IN PEA
The pea genome is organized in 7 pairs of chromosomes (2n = 2x = 14). Its haploid size is estimated at 4.45 Gb (Dolezel et al., 1998;Dolezel and Greilhuber, 2010;Praca-Fontes et al., 2014). It is largely dominated by mobile elements, mainly of the Ty3/gypsy family (Macas et al., 2007). This large genome size and high transposable element content have undoubtedly contributed to delay the development and availability of genomic tools in pea. Recently, several national and international programs have developed diverse valuable genomic resources by taking advantage of cutting-edge sequencing and genotyping technologies. These programs indicate the determination of the pea community to make rapid progress toward targeted and efficient molecular breeding exploiting the rich diversity of pea germplasm and its wild relatives. An international consortium has been initiated in order to generate the full-sequence of the pea genome (Madoui et al., 2015).

THE USE OF GENOMIC TOOLS IN MARKER-TRAIT ASSOCIATION STUDIES AND BREEDING PROGRAMS
Genetic maps have proven to be useful to uncover the molecular bases of monogenic characters such as Mendel's characters (see Ellis et al., 2011 for review) and also to decipher the determinism of complex agronomically-important traits ( Table 1). QTLs responsible for the genetic control of yieldrelated traits, seed protein content, aerial and root architecture, and biotic and abiotic stress resistance have been detected under multiple environmental conditions and located on different maps ( Table 1). In addition to QTL mapping analyses in biparental populations, association analyses have emerged as a complementary approach to dissect quantitative traits in pea by exploiting natural genetic diversity and ancestral recombination events characterizing germplasm collections. Diverse sets of cultivars with distinct geographic origins were used to determine associations between genetic markers and seed mineral nutrient concentration (Kwon et al., 2012;Cheng et al., 2015;Diapari et al., 2015), seed low-carbohydrate content (Cheng et al., 2015), seed lipid content (Ahmad et al., 2015), yield-related traits (Kwon et al., 2012), disease/pest resistance, and morphological traits such as flower color and seed coat color. Genome wide association mapping and subsequent allele mining could build on genomic resources reviewed herein and the phenotyping data available for diverse germplasm collections, as reviewed in Warkentin et al. (2015). Specific markers linked to major genes were developed for use in breeding, especially for trypsin inhibitors in pea seeds (Page et al., 2002;Duc et al., 2004), flowering (Weller and Ortega, 2015), lodging resistance (Zhang et al., 2006) and resistance to diseases such as powdery mildew (Ghafoor and McPhee, 2012;Reddy et al., 2015), pea enation and seed borne mosaic virus Jain et al., 2013), fusarium wilt (McClendon et al., 2002), Ascochyta blight (Jha et al., 2015), and rust (Barilli et al., 2010) (Supplementary Table 1). Some other resistance, flowering or seed composition genes were reviewed by Warkentin et al. (2015). Marker-assisted selection was conducted in early generation (F 2 ) breeding populations using markers linked in coupling to two major QTLs controlling lodging resistance and was demonstrated more efficient than phenotypic selection (Zhang et al., 2006). Recently, markerassisted backcrossing (MABC) was successfully used to introgress one to three of the seven main Aphanomyces root rot resistance QTLs (Hamon et al., 2013) into several recipient agronomic lines (Lavaud et al., 2015). Evaluation for resistance of the subsequent 157 BC5/6 Near Isogenic Lines (NILs) validated the effect of the major and some minor QTLs in controlled conditions and showed QTL x genetic background interactions. A MABC strategy was also used to introgress three frost tolerance QTLs among the main four QTLs identified by Lejeune-Hénaut et al. (2008). Field evaluations of 125 QTL-NILs validated the effect of these QTLs in the spring-type genetic background Eden (Hascoët et al., 2014). So far, marker-assisted construction of QTL-NILs has mainly allowed QTL effects to be validated. The rational use of these genetic regions in breeding strategies can now be considered in order to combine favorable alleles at complementary QTLs to improve multiple stress resistance in agronomic material. A list of the markers that should be useful for pea breeding is provided in Supplementary Table 1. In parallel to strategies considering the combination of individual QTLs, genomic selection seems a promising approach in pea, as first suggested by the prediction of the date of beginning of flowering and 1000 seed weight using a subset of 331 SNP markers genotyped in a reference collection of 372 pea accessions (Burstin et al., 2015). Increasing the marker coverage of the genome by using the newly-developed GenoPea 13.2K SNP Array (Tayeh et al., 2015a) further improved prediction accuracies (Tayeh et al., 2015b).

PERSPECTIVES
The gradually-developed genomic tools for pea now represent a rich resource for innovative strategies in both basic research and applied breeding. The large set of bi-parental interconnected populations segregating for diverse important agronomic traits, the individual and consensus genetic maps, the dense arrays of genetic markers, the high-throughput SNP genotyping tools, the BAC libraries, the TILLING population, and the whole-genome and transcriptome sequences from a large group of accessions should enhance significant advances in pea breeding in the next few years and foster the use of more diverse genetic resources for pea improvement. The genome sequence when released will further advance the pea genomic breeding revolution.

ACKNOWLEDGMENTS
The present work of NT, GA, IL, MP, JB was done in the frame of the French National Research Agency (ANR) Project Investissements d'Avenir PeaMUST under the grant number ANR-11-BTBR-0002.