Functional genomics to study stress responses in crop legumes: progress and prospects

Additional keywords: abiotic and biotic stresses, expression profiling, stress tolerance, transcriptomics.

Fabaceae, the legume family, comprising more than 650 genera and 20 000 species, is the third largest family of higher plants and the second most important family among crop plants after Poaceae (the grass family). Legumes account for ~27% of crop production in agriculture worldwide based on area harvested and total production (Graham and Vance 2003). Crop legumes complement cereals, the primary source of carbohydrates in the human diet, in terms of amino acid composition, and provide around one-third (20–40%) of all dietary protein (Zhu et al. 2005). Legumes produce secondary metabolic compounds that can protect the plant against pathogens and pests. The economic importance of legumes and some of their salient biological features including symbiotic nitrogen fixation, the prevention of erosion, suppression of weeds and adding organic matter to the soil, provide ample justification for a significant investment in genomics based improvement of this important crop family.

Some legumes constitute an important component of the human diet in developing countries; include soybean (Glycine max), peanut or groundnut (Arachis hypogaea), chickpea (Cicer arietinum), cowpea (Vigna unguiculata), common bean (Phaseolus vulgaris), pigeonpea (Cajanus cajan), pea (Pisum sativum), lentil (Lens culinaris), faba bean (Vicia faba), mungbean (Vigna radiata) and lupin (Lupinus luteus). Despite having an important role in food security, the majority of these legume crops demonstrate low productivity due to biotic (e.g. bacteria, fungi, nematodes, viruses and insects) and abiotic (e.g. drought, salinity, heat and waterlogging) stresses. For example, in peanut and chickpea, drought is an important abiotic stress constrain and major biotic stresses include anthracnose, angular leaf spot, bean rust, bacterial blight in common bean, Ascochyta blight and Fusarium wilt in chickpea. Thus, it is necessary to enhance our understanding of specific aspects of defence/stress responses to improve crop productivity. Towards this aim, emerging genomics technology can be applied to interrogate the basis of stress response and identify candidate genes or key loci controlling stress tolerance or resistance. Subsequently these genes can be used in genetic modification or molecular breeding programs to develop improved varieties with enhanced resistance/tolerance to stress.

Due to their small genome sizes, their simple genetic system and amenability to forward and reverse genetic analyses, two legume species, namely Medicago (Medicago truncatula) and Lotus (Lotus japonicus) have been used for extensive molecular studies in the past two decades (Handberg and Stougaard 1992; Cook 1999). Recently, the legume community has adopted next generation sequencing (NGS) and high-throughput genotyping technologies to undertake functional genomics studies in the crop legumes. As a result, a vast amount of genomic resources have been developed that enable isolation and characterisation of key genes involved in legume stress response. Once candidate genes are identified, it is important to validate their function before their application in crop improvement strategies (Valliyodan and Nguyen 2006). The successful application of biotechnological tools to alleviate the biotic/abiotic constraints of crop legumes will require both biological knowledge of the target species and the underlying mechanisms of crop stress response.

In view of the above considerations, this article summarises and presents a critical appraisal of the development/availability of genomic resources and their use for the identification, isolation and validation of candidate genes conferring resistance/tolerance to biotic/abiotic stress. Finally, an overview has been presented on the integration of various functional genomics approaches towards the genetic improvement of leguminous crops.

A major aim of genomic studies in plants is the identification of genes and pathways that affect crop production. Genome sequencing is fundamental to understand the genomic composition and gene repertoire of an organism; however, because of the high costs associated with sequencing a genome, initially only the genomes of model legumes were sequenced. An alternative approach to genome sequencing is targeted expressed gene sequencing. Therefore, in many crop legumes, efforts focussed on the development of cDNA libraries, the generation of expressed sequence tags (EST), gene expression analysis, and the in silico mining of functional information from EST datasets. An overview of functional genomics approaches for crop improvement is illustrated in Fig. 1.

Fig. 1.  An overview of functional genomics approaches for legume improvement. The genetic resources developed provide the basis for gene expression analysis (SSH libraries, Affymetrix array hybridisations, RNA-seq) and significantly contribute to the identification and characterisation of candidate gene(s). The use of functional genomic approaches will enhance the efficiency of functional validation of the identified candidate genes and lead to the development of improved legume varieties through molecular breeding.

In the absence of genome sequence data, EST collections produced by Sanger sequencing have proven extremely useful for many plant studies. EST databases provide basic sequence depositories for gene discovery and assist in comparative mapping. They also facilitate the identification of candidate genes for agronomic traits (Young and Bharti 2012). In legume species, extensive efforts have generated an abundance of ESTs from a range of tissues, including from plants challenged by stress. Today, more than 3 million legume ESTs are available, predominantly from soybean (1.5 million, Vodkin et al. 2004) followed by the model legumes M. truncatula (280 000, Cheung et al. 2006) and L. japonicus (242 000, Asamizu et al. 2004). Among crop legumes, cowpea contributed around 200 000 ESTs (Muchero et al. 2009), and common bean 114 139 (Blair et al. 2011). In the case of chickpea, cDNA libraries have been generated from plants under drought and salinity stress (Varshney et al. 2009a). In the case of pigeonpea, Fusarium wilt and sterility mosaic disease (SMD), responsive ESTs were generated (Raju et al. 2010). Sanger ESTs generated from stress-responsive tissues from selected key studies have been summarised in Table 1.

Table 1.  Stress responsive ESTs generated in crop legumes by Sanger sequencing

EST libraries have also been constructed using the suppression subtractive hybridisation (SSH) technique, and utilising this approach, ribosomal protein genes related to cold and salt stresses were cloned from soybean (Kim et al. 2004). In a different study, 372 high quality salt stress-responsive ESTs were generated from soybean SSH libraries (Li et al. 2012). In the case of chickpea, 477 drought-responsive ESTs were generated from root tissues (Buhariwalla et al. 2005). Deokar et al. (2011) also generated 3062 unigenes from SSH libraries of root and shoot tissues of contrasting drought-responsive genotypes in chickpea. In pigeonpea, 182 unique ESTs were generated from drought-stressed and unstressed pigeonpea seedlings using SSH (Qiao et al. 2012). Although the method can be technically demanding and labour intensive, the establishment of SSH libraries is a rewarding approach for the identification of candidate genes for a given stress.

Due to the availability of high-throughput and cost-effective NGS platforms such as the Illumina HiSeq (Illumina Inc, San Diego, CA, USA), GAIIx, MiSeq; Roche 454/FLX (454 Life Sciences, Branford, CT, USA); ABI SOLiD (Applied Biosystems, Carlsbad, CA, USA); and the Invitrogen Ion Proton (Invitrogen, Carlsbad, CA, USA), the sequencing of transcriptomes and genomes has become more efficient and economical (Varshney et al. 2009b; Edwards et al. 2013). The expansion of third generation sequencing technologies such as those of Pacific Biosciences (PacBio, Menlo Park, CA, USA), Oxford Nanopore Technologies (Oxford, UK) is expected to accelerate the large scale generation of genomic resources (Munroe and Harris 2010; Thudi et al. 2012). Several NGS platforms have already generated a vast set of transcript reads from a range of developing and stress-responsive tissues from a range of different crop legumes (Table 2).

Table 2.  ESTs generated with next generation sequencing (NGS) technologies

The application of NGS technology has led to the production of transcriptome assemblies for chickpea (Hiremath et al. 2011; Garg et al. 2011a), pigeonpea (Dubey et al. 2011; Kudapa et al. 2012), peanut (Zhang et al. 2012), pea (Franssen et al. 2011) and lupin (Parra-González et al. 2012). In addition, the combination of reads generated by NGS platforms and Sanger ESTs has improved transcriptome assemblies, particularly in the context of contig length (Hiremath et al. 2011). For example, a transcriptome assembly based on FLX/454 sequencing together with Sanger ESTs comprised 103 215 tentative unique sequences (TUSs) with an average contig length of 459 bp (Hiremath et al. 2011). By analysing sequencing data from FLX/454 (~7 million reads), Illumina (~100 million tags) and Sanger (~150 000 ESTs) platforms, an improved assembly with 46 369 unigenes with an average contig length of 965 bp could be achieved (H Kudapa, S Azam, AG Sharpe, B Taran, R Li, B Deonovic, C Cameron, AD Farmer, RK Varshney, unpubl. data). In an effort to improve transcriptome assembly, researchers have compared the performance of different assemblers including CAP3, MIRA, TGICL, and Velvet, either alone or in combination (Garg et al. 2011b; Kudapa et al. 2012). The National Center for Genome Resources (NCGR) in cooperation with the USA Department of Agriculture (USDA)-supported Legume Information System (LIS, http://www.comparative-legumes.org, accessed 9 June 2013) offers a comprehensive collection of transcriptome assemblies for several legumes.

Whole-genome sequencing is fundamental to understand the genetic composition of an organism. The two model legume species M. truncatula and L. japonicus were first selected for sequencing in depth. Draft genome assemblies were published for Medicago (Young et al. 2011) and Lotus (Sato et al. 2008). The Medicago assembly captured ~94% of expressed genes, whereas the Lotus assembly represented 91% of the gene space (Sato et al. 2008).

Among crop legumes, the assembly of a soybean reference genome was fundamental and improved our current understanding of legume genomes generally. About 969.6 Mb of the 1115 Mb genome was assembled after generating eight times the whole-genome shotgun (WGS) data using Sanger sequencing (Schmutz et al. 2010). Recently, draft genome sequences of pigeonpea and chickpea have been reported, representing 73% (Varshney et al. 2012) and 74% (Varshney et al. 2013) of the respective genomes. A common bean genome sequence has also become available on the Phytozome data portal (http://www.phytozome.net/commonbean.php, accessed 13 June 2013). Collaborative projects involving ICRISAT are underway to generate the tetraploid (US-led initiative) and the diploid genomes (China-led initiative) of peanut. Efforts are also underway to assemble the genomes of mungbean (S Ha-Lee, pers. comm.), lentil, lupin and pea (D Edwards, pers. comm.). Details of the released genome sequences and the total number of genes identified from crop legume species are summarised in Table 3. We note that the genome sequencing projects revealed several genes involved in stress response. For example, 111 drought responsive genes in pigeonpea (Varshney et al. 2012) and 187 disease resistance genes in chickpea (Varshney et al. 2013) were identified from the genome sequence analysis. The stress responsive genes identified from genome sequencing projects could be of great value for dissecting candidate genes for important stresses in respective crop legumes.

Table 3.  Available draft genome sequences of crop legumes

In addition to the generation of reference genome sequences (Imelfort et al. 2009a; Edwards and Batley 2010); NGS can facilitate the re-sequencing of genomes to identify genomic variation such as SNPs (Imelfort et al. 2009b; Lorenc et al. 2012). In the case of soybean, 17 wild and 14 cultivated soybean genomes have been re-sequenced to date (Lam et al. 2010). This study revealed patterns of genetic variation between wild and cultivated soybeans and identified greater allelic diversity in wild soybeans, and a set of 205 614 SNPs have been identified for use in QTL mapping and association studies. In the model legume Medicago, the genomes of 26 diverse accessions were sequenced to identify and characterise sequence polymorphisms and linkage disequilibrium (LD). More than 3 million SNPs were detected and this study suggested that M. truncatula demonstrated greater diversity and less LD when compared with soybean (Branca et al. 2011). Ninety chickpea genomes have been re-sequenced, revealing 4.4 million variants (SNPs and INDELs). In addition, genetic diversity and phylogenetic analyses from this study highlighted the mixed use of desi and kabuli genotypes in the history of chickpea breeding (Varshney et al. 2013).

Genome re-sequencing is a powerful approach for detecting new alleles and haplotypes that can be used in genome-wide association studies (GWAS). As re-sequencing costs are continuously decreasing, re-sequencing-based allele discovery is expected to become more popular (Tuskan et al. 2011).

For the discovery of differentially expressed genes or candidate genes for a particular trait, the following two approaches have been used as follows.

Various microarray technologies have been found useful to unravel key biological processes. Generally, microarray platforms fall into a bewildering variety of architectures that are now mostly superseded by sequencing based approaches. In the context of this review, cDNA arrays and oligonucleotide-based chips are considered in some detail (Sreenivasulu et al. 2002). Complementary DNAs or oligonucleotides representing non-redundant sets of ESTs were immobilised on either nylon membranes or glass slides and effectively applied for gene expression analysis, especially in species for which only limited genome information was available. Several cDNA-based arrays have been applied in crop legumes. For example, in soybean, high density arrays based on 27 513 cDNA inserts that represented a low redundancy ‘unigene set’ were applied for a wide range of samples from various developmental stages, including disease-challenged and stress-exposed tissues (Vodkin et al. 2004). These arrays were also used to infer global gene expression patterns in mutant iso-lines. This study identified a set of candidate genes that respond to different stresses including drought, heat, flooding, herbicide application and various pathogens. In chickpea, 8098 probes corresponding to 2013 unigenes were immobilised onto a microarray and global gene expression profiles examined in roots during vascular wilt (Ashraf et al. 2009). Potential innate immune-responsive candidate genes involved in a complex regulatory network could be identified. A 768-feature microarray comprising chickpea cDNAs (559), grasspea cDNAs (156), lentil resistance gene analogues (RGAs) (41) and controls (12) were designed by Mantri et al. (2010) to explore abiotic stress-responsive transcripts in chickpea. The authors identified differentially expressed genes in contrasting genotypes (tolerant/ sensitive for drought, salinity and cold stresses). Furthermore, genes coding for various regulatory and functional proteins are now known, and the complex mechanisms of multi-gene control in abiotic stress responses partly deciphered. As a further spin-off, common genes expressed under different stresses suggested an activation of distinct gene batteries as a general phenomenon in stress responses (Cheong et al. 2002). Another platform for hybridisation of RNA and comparison of gene expression across tissues/genotypes is the Affymetrix gene chip. For legumes, the Affymetrix Array GeneChip platform is available for Medicago (http://www.affymetrix.com/estore/browse/products.jsp?productId=131472#1_1, accessed 29 May 2013) and soybean (http://www.affymetrix.com/estore/browse/products.jsp?productId=131507#1_1, accessed 29 May 2013). Several studies in model legumes have employed these GeneChip Genome Arrays for research into developmental or metabolic pathways (Pang et al. 2009; Verdier et al. 2013). In the case of soybean, the GeneChip Soybean Genome Array permitted the characterisation of genome-wide expression patterns, and identified drought-responsive candidate genes. GeneChip Soybean Genome arrays were also applied to other legume species. For instance, the response to root knot nematode infection of resistant cowpea genotype CB46 and a susceptible near-isogenic line (null-Rk) were investigated at 3 and 9 days post inoculation (Das et al. 2010). Furthermore, GeneChip Soybean Genome Arrays identified single feature polymorphisms (SFPs) in cowpea (Das et al. 2008) and pigeonpea (Saxena et al. 2011).

Serial analysis of gene expression (SAGE) and its multiple variants allow to quantify global gene expression. In the original SAGE method, mRNA is oligo (dT)-trapped and reverse transcribed into cDNA, then a small sequence (‘tag’) is extracted from a defined position of each cDNA molecule. These small tags are ligated to form a long concatemeric chain that is cloned into a vector. Subsequently, these concatemers are sequenced. Although SAGE proved its usefulness, the size of the SAGE tag is too short to unequivocally identify the gene of origin. To overcome this problem, a variant of SAGE, called SuperSAGE was developed by Matsumura et al. (2005). If combined with one of the next-generation sequencing platforms, it is more precisely called deepSuperSAGE. This approach uses the type III-endonuclease EcoP15I of phage P1 to cut 26 bp long sequence tags from each transcript’s cDNA, expanding the tag-size and thereby the security of annotation. Quantification of a particular tag by automated counting provides the expression level of the corresponding transcript and also unravels novel expressed regions of the genome. By using SuperSAGE, Kahl et al. (2007) investigated salt- and drought- stress transcriptomes of chickpea and lentil by analysing 360 000 transcripts representing 40 000 unique mRNAs, and identified 3000 transcripts responding to these stresses. In another deepSuperSAGE application, 80 238 tags representing 17 493 unique transcripts from drought-stressed and non-stressed control roots in chickpea have been identified (Molina et al. 2008, 2011).

Massively parallel signature sequencing (MPSS) is yet another powerful technique for transcription profiling on a genome-wide scale (Brenner et al. 2000), though it is no longer competing with the NGS platforms. In this method, tagged PCR products produced from cDNA are amplified so that for each mRNA molecule ~100 000 of PCR products with a unique tag are produced. These tags are used to attach the PCR products to microbeads that avoid separate cDNA isolation, template processing and robotic procedures. Subsequently, after several rounds of ligation-based sequence determination using the type IIs restriction endonuclease BbvI, a sequence signature of ~16–20 bp is identified from each bead; routinely, 17 bp of high quality sequence is obtained. This procedure is performed in parallel, and ~1 million sequence signatures are obtained per experiment. However, because of its cost, the full potential of MPSS in the global expression profiling of the abiotic and biotic stress response is yet to be realised.

Although Sanger ESTs have also been used for digital gene expression (DGE) studies (Sreenivasulu et al. 2002), NGS platforms have greatly expanded genome wide sequence based gene expression analysis. The sequencing of RNA populations at an unprecedented depth and the quantification of the transcripts, though problematic, can be achieved through RNA-seq. The efficiency of Illumina-based DGE for the identification of differentially expressed genes was demonstrated by Hoen et al. (2008) by comparing RNA-seq with five different microarray platforms. This study concluded that deep sequencing provides a major advance in robustness, comparability and richness of expression profiling data. In soybean, DGE tag profiling was used to compare the transcriptional profiles between wild-type (CS) and a mutant isoline (CG) (Hunt et al. 2011; Wang et al. 2012). About 85 000 unique tags representing over 4.7 million DGE tags were generated (each from CS and CG) and applied to extend predicted gene models for the soybean genome. The datasets showed highly expressed genes as well as differentially expressed genes between young shoot tips CS and CG lines that encode proteins related to: ribosomes (70 different tags), protein biosynthesis/metabolism (35 tags), photosynthesis (34 tags), others (29 tags) and histones (28 tags) (Hunt et al. 2011). In the case of chickpea, Hiremath et al. (2011) observed 2974 TUSs with significant expression changes, of which 2823 could be associated with gene ontology annotations. Furthermore, expression patterns of many genes suggested their role in various pathways of secondary metabolism. In a different study, a wide range of expression levels were observed by mapping all reads onto a non-redundant set of chickpea transcripts, where the number of reads corresponding to each transcript ranged from 14 (0.16 reads per million, rpm) to 270 894 (3137 rpm), with an average of 1617 (18.7 rpm) (Garg et al. 2011a). This report identified 250 transcripts with root-, and 217 transcripts with shoot-specific expression. In the case of pigeonpea, significant differential expression was observed for 6673 to 11 518 TUSs for specific parental combinations (tolerant/sensitive for FW and SMD) (Dubey et al. 2011), and candidate FW- and SMD- responsive genes identified, which represent starting points to analyse biotic stress.

Of all the various methods of DGE, RNA-Seq is the most advanced technique for quantifying gene expression in crop legumes, especially when transcriptome assemblies and genome sequences become available. Most recently, a broad repertoire of greatly advanced techniques from proteomics, metabolomics and phenomics complement the aforementioned gene discovery suites.

Gene cloning is an approach for isolating candidate genes that are functionally related to the trait of interest. A forward genetics approach for the identification of genes controlling a trait is positional cloning. Positional cloning per se may not conclusively identify target genes associated with a particular phenotype. It provides, however, useful genetic information that often requires support at the transcriptome (mRNA), proteome (proteins) and metabolome (metabolites) levels. The ultimate proof for the causative linkage of a gene in the target region linked to the trait of interest is complementation analysis (Langridge and Fleury 2011). In map-based cloning (MBC, a variant of positional cloning) the chromosomal location of the gene is identified through genetic mapping using molecular markers. Thanks to advances in sequencing and the genome-wide identification of sequence polymorphisms (e.g. SNPs), MBC became more accessible, and has been conducted in a range of crop species such as rice (Vij and Tyagi 2007) as well as in some legumes including soybean (Watanabe et al. 2009).

With ever faster and more and more accurate DNA sequencing technologies and the availability of large scale genomic resources (molecular markers) map based cloning limitations could be overcome. The published genome sequence assemblies for legumes such as soybean, chickpea, pigeonpea and common bean as well as the advanced DNA polymorphism detection will eventually make MBC of genes from crop legumes routine. The majority of the MBC projects in crop legumes have been applied in soybean. For example, the soybean phytochrome A gene (GmPhyA3) which modulates flowering time has been cloned using an MBC approach (Watanabe et al. 2009). In a different study, a candidate gene Ln controlling leaflet and seed number per pod was cloned with a combination of MBC and association study (Fang et al. 2013). The cloning and characterisation of Fusarium wilt resistance genes using MBC was demonstrated for chickpea (Huettel et al. 2002; Sharma and Muehlbauer 2007). Efforts are under way to clone genes from within a drought tolerance QTL region in chickpea (Thudi 2013). Some examples of MBC in crop legumes are depicted in Table 4.

Table 4.  Some genes cloned in crop legumes

Meng et al. (2007), using a comparative genomics approach, isolated and cloned GmNAC1 to GmNAC6 genes in soybean that encode cold stress-responsive factors. In common bean, Torres-Franklin et al. (2008) cloned the gene glutathione reductase (dtGR), a drought stress-responsive gene. Other stress-related genes have been identified and cloned with the above approaches in crop legumes including soybean, chickpea and cowpea (Sharma and Muehlbauer 2007; França et al. 2008; Fang et al. 2013). Some examples are shown in Table 4.

After discovering trait associated genes by any of above mentioned approaches, the next step is their functional validation. Several approaches such as overexpression, RNAi, virus induced gene silencing (VIGS) and TILLING have been applied for this purpose.

One of the most reliable methods of validation of isolated/cloned genes is to generate transgenics and assess the expression of the respective trait and this approach has been applied to validate the function of stress-responsive genes. In this procedure, success depends on the incorporation of the stress-responsive gene into the genome, and its expression. The lack of routine transformation protocols with high efficiency has been a constraint in crop legumes, mainly due to poor regeneration ability (especially via callus) and lack of compatible gene delivery methods. Many techniques (e.g. electroporation of intact tissues, silicon carbide whiskers) have been tested for gene delivery to the plant cell, and Agrobacterium-mediated and particle bombardment have been extensively employed for genetic transformation in several other crop plants. However, Agrobacterium-mediated transformation has low efficiency in grain legumes (Chandra and Pental 2003).

Furthermore, different pathways of regeneration vary in their amenability to different gene delivery techniques. Regeneration of plant tissues in vitro is through two pathways: ‘organogenesis’ and ‘embryogenesis’. Shoot buds are organised by concerted meristematic activity of several cells in organogenesis, whereas a single cell or a small cluster of cells undergo differentiation to produce somatic embryos similar to zygotic embryos in embryogenesis. However, the most prevalent mode of regeneration is via direct organogenesis in crop legumes and has been found to be most responsive in several crop legumes such as soybean (Kaneda et al. 1997) and pea (Jackson and Hobbs 1990).

Particle-gun mediated transformation has been used in some legume crops to generate transgenics although at low frequencies. Nevertheless good protocols with higher efficiency are already available in several legume crops, including chickpea (e.g. Acharjee et al. 2010), cowpea (e.g. Citadin et al. 2013), pigeonpea (e.g. Sharma et al. 2006), peanut (e.g. Bhatnagar-Mathur et al. 2007) and common bean (e.g. Aragão et al. 2013). Overall, the development of transgenics in some crop legumes, such as mungbean and lentil still remains a challenge. In addition, precise evaluation of the transgenic plant under stress conditions and understanding the physiological effect of the inserted gene(s) at the whole-plant level is also necessary in understanding overexpression studies (Bhatnagar-Mathur et al. 2009).

The majority of studies involving the overexpression of biotic or abiotic stress-responsive genes from either model or crop species were conducted in model plants such as Arabidopsis or Medicago, and only recently in some crop legumes. For the engineering of biotic stress resistance, the gene encoding α-amylase inhibitor αAI-1, a bruchid resistance factor from common bean, was overexpressed in other grain legumes including chickpea, pea, azuki bean and cowpea. The αAI-1 gene present in transgenic chickpea and cowpea under the control of a cotyledon-specific promoter provided resistance to several important bruchid pest species (Lüthi et al. 2010). A comparison of the post-translational modifications of αAI expressed in transgenic peas and chickpeas damaged by bruchids, with the processed forms of the same protein from several beans revealed microheterogeneity, with variations in the frequency of addition and variable processing of glycans, and in the C-terminal exopeptidase activity.

In the case of abiotic stress, the overexpression of Arabidopsis thaliana vacuolar H⁺-PPase (AVP1) in M. sativa lead to enhanced salt and drought tolerance. The trangenic plants accumulated more Na⁺, K⁺ and Ca²⁺ in their leaves and roots, and retained more water in the leaves during drought stress as compared with the wild-type plants (Bao et al. 2009). Similarly, overexpression of the pyrroline-5-carboxylate synthetase gene (P5CS) from Arabidopsis enhanced salt tolerance in chickpea (Ghanti et al. 2011). Very recently, Hanafy et al. (2013) reported enhanced tolerance to drought and salt stress in transgenic faba bean due to the heterologous expression of the PR10a gene, which encodes a pathogenesis related (PR) protein from potato.

In brief, overexpression of the candidate genes not only provides reliable validation of gene function, but may also lead to the development of improved lines in targeted crop species. Those improved, transgenic lines can be taken to greenhouse or field and exploited for enhancing yield and improving food or feed quality subject to biosafety procedures.

RNAi and VIGS are important approaches to validate the functions of candidate genes. In both of these methods, genes belonging to gene families are blocked or expressed across several tissues and developmental stages where antisense technologies fail to perform. The RNAi approach is sequence specific and can be targeted and controlled in tissue specific and time dependent manner. It is a popular approach for validating the function of candidate genes which have been identified on the basis of sequence similarity or through genetic mapping. This technology can be applied as an initial screen and subsequently validated by other methods (Small 2007).

RNAi-induced gene silencing is well established in model legumes and soybean. For example, in Medicago, RNAi was used to interfere with the RNA encoding PIN (auxin export facilitator) proteins responsible for nodule development. Reduced expression levels of root-specific PIN proteins produced plants with a reduced number of nodules, demonstrating the important role of PIN proteins in nodule development (Huo et al. 2006). In soybean, RNAi was employed to silence the gene encoding myo-inositol-1-phosphate (GmMIPS1), which plays an important role in regulating cellular metabolism and controlling growth. Seed development was not possible in lines in which the GmMIPS1 gene was silenced, demonstrating the correlation between GmMIPS1 gene expression and seed development (Nunes et al. 2006). Through RNAi-induced gene silencing is well established, VIGS may be a better approach in the long-term, due to its persistence during vegetative and in vitro propagation which, in turn, allows the generation of genotypically identical silenced plants.

VIGS has already been employed in soybean and pea. In soybean, a bean pod mottle virus (BPMV)-based system was used to identify genes participating in basal, resistance gene-mediated, and systemic immunity (Kachroo and Ghabrial 2012). In another study, the apple latent spherical virus (ALSV) vector was used to study gene function in the reproductive and early growth stages (emergence and cotyledon) in addition to the vegetative stages (Yamagishi and Yoshikawa 2009). The pea early-browning virus (PEBV) has been developed as a VIGS vector and used for functional analysis of several genes involved in the symbiosis. This study identified genes involved in symbiosis at the early and late growth stages of the plant (Grønlund et al. 2010). In summary, VIGS can be used as a forward or reverse genetics tool to validate the function of candidate gene(s) in transgenic plants as well as to characterise germplasm lines with differential expression of a gene with a desirable trait (Senthil-Kumar and Mysore 2011). However, the lack of appropriate vectors (specific for crop legumes), and efficient method for virus vector delivery may be limiting factors for the extensive applications of VIGS in crop legumes.

Validation of genes through genetic transformation, RNAi or VIGS is a time consuming process in legumes, mainly due to lack of efficient transformation systems in legumes. This situation has promoted the application of TILLING to study gene function. In TILLING, candidate genes are screened across a mutant population (with point mutation), and line(s) with the mutation for the target gene are identified (McCallum et al. 2000). If the identified line exhibits the expected phenotype for the candidate gene, the function of the candidate gene is supported. The TILLING approach could be preferred over RNAi for irreversibly reducing or eliminating the target genes in commercial crop plants since it avoids genetic transformation and increases stability of the phenotype (Barkley and Wang 2008).

TILLING populations have been developed for several legumes. For example, in the model legumes, Medicago (12 000 M2 plants) (Rogers et al. 2009) and Lotus (4904 M2 lines) (Perry et al. 2009) mutant populations were developed for use in reverse genetics. In the case of crop legumes, over 3000 M3 lines were developed in common bean and evaluated with root nodulation tests by Porch et al. (2009). In peanut a TILLING population of 10 000 lines has been established, and a subset of this population investigated for allergenicity (Tadege et al. 2009). In chickpea, a TILLING population of ~3500 lines has been developed and is being used to identify candidate genes for drought tolerance (M. Thudi, pers. comm.). The use of NGS technologies for TILLING may increase the application of TILLING in crop legumes.

EcoTILLING is a variant of TILLING, except that its objective is to discover naturally occurring polymorphisms as opposed to experimentally induced mutations. Single nucleotide polymorphisms (SNPs), small insertions and deletions, and variations in microsatellite repeat number can be efficiently detected using the EcoTILLING technique. For example, in legumes this method has been used to develop molecular markers for cyst nematode candidate resistance genes in soybean (Liu et al. 2012). In mungbean, it has been proven to be a valuable method for detecting polymorphisms in a collection that was previously shown to have limited diversity (Barkley and Wang 2008).

In recent years, significant progress has been made in developing genomic resources including genome sequences and transcriptome assemblies for a handful of crop legumes. These genomic tools may help to identify key factors involved in legume stress response. The availability of the complete genome or draft genome sequences of legume species and advances in sequencing and bioinformatics will accelerate gene discovery, particularly tolerance/resistance genes to abiotic/biotic stress in legumes (Edwards 2007). Though bioinformatics can predict potential gene function, it is important to validate gene function at the plant level, which can be done with overexpression, VIGS, RNAi, or TILLING, to name few. The outcome of experiments with these techniques can be complemented with metabolomics and proteomics. Once identified and validated, candidate genes enable the researcher to convert them into markers, to mine for superior alleles in germplasm collections, and to use them for a production of transgenics. Furthermore, identified and isolated candidate genes from one species can be channelled into comparative genomics studies of related species.