Pigeonpea [Cajanus cajan (L.) Millspaugh] is an important food legume predominantly cultivated in the tropical and subtropical regions of Asia and Africa. It is a diploid (2n = 22), often cross-pollinated crop with a genome size of 858 Mbp [1]. Pigeonpea plays an important role in food and nutritional security because it is a rich source of protein, minerals and vitamins. Pigeonpea seeds are mainly consumed as split pea soups or 'dal' but a significant proportion is also eaten as green pea vegetable and as wholegrain preparations. In addition, pigeonpea leaves, seed husks and pods are used as animal feed, whereas the stem and branches are used as firewood. The world acreage of pigeonpea is 4.67 Mha with an annual production of 3.30 Mt. India is the largest producer and consumer of pigeonpea (local names 'arhar' and 'toor') with an annual production of 2.31 Mt, followed by Myanmar (0.60 Mt), Malawi (0.16 Mt) and Kenya (0.10 Mt) [2].

Knowledge of the genetic basis of yield, resistance to diseases and insect pests and abiotic stress tolerance are important factors for deciding the breeding strategies for genetic improvement of pigeonpea. However, in comparison to other economically important crops, relatively less effort has been invested in understanding the genetics of important agronomic traits of pigeonpea. Although there are ongoing efforts for pigeonpea improvement through conventional breeding, including hybrid technology, molecular breeding has a greater potential to accelerate the utilization of genetic resources in pigeonpea, especially among land races and related germplasm lines [3–8]. The availability of molecular markers that are tightly linked to important agronomic traits is a prerequisite for undertaking molecular breeding in plants. However, the genetic basis of most agronomic traits in pigeonpea has been worked out using conventional biometric techniques that have inherent limitations. The molecular basis of traits remains entirely unexplored and to date no molecular linkage map has been reported for pigeonpea [9, 10]. This can be attributed to: (i) the low level of DNA polymorphism within the primary (cultivated) gene pool assessed by means of RAPD, RFLP, AFLP and recently by diversity array technologies (DArT) [11–15]; and (ii) a paucity of molecular markers available for genetic analysis in pigeonpea [16–20].

Simple sequence repeat (SSR) markers have the advantage of high abundance, random distribution within the genome, high polymorphism information content and co-dominant inheritance. However, genomic SSR markers developed from SSR-enriched genomic libraries or random genomic sequences are derived primarily from inter-genic DNA regions, and therefore have uncertain linkage to the transcribed regions of the genome. In contrast, genic-SSRs specifically target the transcribed region of the genome and have increased potential for linkage to loci that contribute to agronomic phenotypes. As a consequence, when polymorphic genic-SSRs are identified in high value breeding lines they can have considerable utility for marker assisted selection (MAS) [21]. Genic-SSR markers can also facilitate better cross-genome comparisons because they target protein-coding regions that are more likely to be conserved between related species [22]. Expressed sequence tags (ESTs) based on Sanger's sequencing technology have become increasingly abundant in public DNA databases and are being used for genetic analyses, comparative mapping, DNA fingerprinting, diversity analysis and evolutionary studies [23, 24]; but only a limited number of pigeonpea Sanger ESTs are available in the public database [9].

We report the development of a large expressed sequence dataset based on 454 GS-FLX pyrosequencing of cDNA pools from two popular cultivars of pigeonpea, which are parents of the reference mapping population. We mined and validated a large set of genic-SSR markers and describe their application for understanding the genetic relationship among selected pigeonpea cultivars and wild Cajanus species. The dataset was also useful for in silico mining of polymorphic genic-SSR loci for the creation of an EST-based intra-species reference genetic map.

Conventional breeding of pigeonpea has continued entirely without the aid of molecular methods and made limited use of germplasm resources, resulting in a very narrow genetic base in the domesticated species. As a consequence, pigeonpea genetic improvement programs have made relatively little progress in addressing the primary constraints to crop production, which include a range of abiotic (e.g. drought, salinity and water-logging) and biotic (e.g. Fusarium wilt, sterility mosaic disease and pod boring insect Helicoverpa armigera) stresses. With the advent of next generation sequencing technologies several crop legumes have recently been subjected to intensive analyses, making marker-assisted breeding a reality [30]. Margulis et al. [31] demonstrated a 100-fold sequencing capability with 454 GS-FLX pyrosequencing but with relatively lower accuracy at the homopolymer positions than Sanger-based capillary electrophoresis sequencing. We used 454 GS-FLX pyrosequencing to develop an extensive collection of expressed sequence reads from two parental pigeonpea cultivars and mined and validated a comprehensive set of genic-SSR markers.

A total of 1.696 million high quality sequence reads were assembled to generate 43,324 TSA unigene contigs, which together represented a large fraction of the pigeonpea transcriptome and helped develop a comprehensive set of genic-SSR markers. Application of a subset of these markers in pigeonpea was sufficient to assess the genetic diversity among cultivars and position domesticated accessions relative to related species and genera. These markers represent a significant addition to the limited set of genic-SSR markers available in pigeonpea [16–20].

Only 81.8% of the SSR-containing unigene sequences showed significant hits in the NCBI non-redundant protein database. This may be due to: (i) EST fragments sequenced directly instead of after cloning; this leads to truly random sequencing of all the expressed genes that may facilitate the discovery of new rare transcripts as evidenced by Emrich et al. [32]; or (ii) unique contigs being part of a consensus sequence representing 3'-UTRs; C-termini or 3' sequences which are often less conserved than other transcript regions [33].

The deep transcriptome sequence data allowed the discovery of a set of 3,771 perfect SSR loci of ≥10 bp length in pigeonpea. About 7.6% of the pigeonpea TSA unigene contigs possessed at least one SSR- similar to previously reported SSR prevalence in the ESTs of wheat (7.41%), higher than grapes (2.5%), barley (2.8%) and flax (3.5%), but lower than coffee (18.5%) [22, 34–37]. The genic-SSR frequency also depended on the parameters used in exploring SSR markers, e.g. the repeat length and number of repeat unit thresholds. The abundance of genic-SSR (kbp/SSR) in pigeonpea was 8.4 compared to 3.4 in rice, 5.4 in wheat, 7.4 in soybean, 11.1 in tomato, 14.0 in Arabidopsis, and 20.0 in cotton [37, 38]. Differences in genic-SSR abundance could be partly due to the size of the EST unigene assembly dataset and use of different search criteria and data mining tools [21]. Frequency distribution of EST-SSR motifs in our study was not comparable with earlier work on pigeonpea by Raju et al. [20] because we developed TSA unigene contigs using FLX-454 sequencing, instead of Sanger sequencing. The number of SSRs identified in the present study was 3,771 from 43,324 unigene sequences, whereas Raju et al. described 3,583 SSRs from only 5,508 unigenes. The main reason for the overestimation of SSR frequency by Raju et al. is the inclusion of compound SSRs and homopolymers which are the most frequent repeats.

Dinucleotide SSR loci were most frequent in the pigeonpea TSA contigs analyzed here, representing 60.41% of the SSR loci identified, i.e. about double that of the trinucleotide SSR loci (34.52%), the second most abundant motifs. This was in agreement with the genic-SSR distribution reported in peach, pumpkin, spruce, coffee and kiwifruit, where dinucleotide repeats are most frequent [39–42]. However, this is in contrast to a number of earlier reports showing trinucleotides as the most abundant class of SSR loci in ESTs [22, 35, 43–47]. A possible explanation for the high frequency of dinucleotide SSR loci in pigeonpea TSA is that these include large amounts of information representing UTRs due to deep transcriptome sequencing. Yu et al. [48] reported 19% of dinucleotide repeats in the coding region and 81% in the 5'- and 3'-UTRs, whereas 74% of the trinucleotide repeats were in the coding regions and only 26% in UTRs. Among the 550 validated SSR loci with n≥18 bp in the present study, only 97 (17.6%) were dinucleotide repeats and 56 were in the UTR. Most of the dinucleotide SSR loci showed a smaller size range of 10-12 bp (Additional file 2). Our study also showed that the overall proportion of polymorphic SSR markers was much higher in UTRs compared to the coding region- there were 40 polymorphic SSR markers in UTRs (11 in 5'-UTR and 29 SSR in 3'-UTR), whereas only 31 were polymorphic in the coding region, despite 61.7% of all amplified SSR markers being located in the coding region. This is due to the tendency of sequence conservation in the coding regions.

Only 80% of the 772 tested SSR primers amplified the target pigeonpea genomic DNA. The success rate is comparable to barley, where 67-70% of the primers amplified [43, 34], but higher than sugarcane (48.5%) and lower than flax (92.2%) [35, 49]. A possible explanation for the lack of amplification could be flanking primers extending across a splice site with a large intron or chimeric cDNA contigs [34]. Although the majority (519 numbers) of the designed SSR markers amplified a single expected product size at the annealing temperature of 55°C, we optimized the annealing temperature of 31 additional primers to maximize the availability of genic-SSR markers for pigeonpea.

Generally genic-SSR markers show a lower level of polymorphism than genomic-SSR markers [22, 50–52], but in this study the use of type I genic-SSR markers showed a high level of polymorphism. Previous diversity studies with pigeonpea species using genomic-SSR markers reported an average of 3.1-4.9 alleles per locus with average PIC values of 0.41-0.52 [15, 17–19]. An earlier study with genic-SSR markers in pigeonpea reported the average number of alleles per marker as 4 and an average PIC value of 0.40 [20]. We observed a higher average of 6.25 alleles per locus and an average PIC value of 0.63 by using type I genic-SSR markers. The possible reasons are: (i) choice of 20 highly polymorphic SSR markers for the diversity assessment on 30 genotypes after initial testing of 550 markers on eight varieties; (ii) higher depth of coverage generated by the 454 GS-FLX sequencing technology that produced larger sequence contigs including UTRs which are more polymorphic; (iii) use of a diverse genotype set including interspecific derivatives and wild Cajanus species for diversity assessment. Contrary to other plant species where dinucleotide repeats showed high polymorphism [33, 43, 46], hexanucleotide repeats were highly polymorphic (38.57%) in pigeonpea genic-SSR markers, followed by pentanucleotides (29.14%) and trinucleotides (15.25%). Larger repeats have been linked to a higher degree of polymorphism in earlier studies [25, 53]; we also found the maximum polymorphism with 40-50 bp SSR length on agarose gel electrophoresis.

On the basis of SSR polymorphism, cluster analysis and earlier diversity studies involving RFLP, AFLP, RAPD, SSR and DArT markers, it is concluded that genetic diversity in the pigeonpea gene pool is very low [11–20]. The genic-SSR markers reported here open up new opportunities to assess the genotypic diversity in the pigeonpea germplasm. Most of the earlier reported SSR markers in pigeonpea are of genomic origin except for 84 genic-SSR markers reported recently [20]. This study is the first report a comprehensive set of genic-SSR markers for pigeonpea.

Wild species of crop plants are placed in different gene pools based on their crossability with the cultivated species. Closely related and easily crossable species are placed in the primary or secondary gene pools, whereas species which are distantly related and are incompatible with the cultivated species, are placed in a tertiary gene pool. Species in the primary and secondary gene pools can be readily utilized for varietal improvement. In this study, the "unweighted pair group method with arithmetic mean" (UPGMA)-based cluster analysis grouped the genotypes according to their gene pool. The variability within the C. cajan cultivars which easily cross-hybridize among themselves formed the primary gene pool (Cluster I), whereas four species that have poor crossability with the C. cajan formed the secondary gene pool (Cluster II). Sub cluster Ib (C. platycarpus 1 and 2) and sub-sub cluster Ia₂ (R. aurea) included species from the tertiary gene pool even though they were closely related to C. cajan based on the marker analysis. The varieties of C. cajan showed different levels of similarity, e.g. PCMF 40 and PCMF 43-7, both inter-specific derivatives belonging to the short maturity group, shared 52% similarity between them and 40% similarity with PCMF 39-1 having similar pedigree. Likewise, short duration variety pair (PS-971/PS-956) and long duration variety pair (Pusa-9/Kudarat) were closer to each other than varieties belonging to different maturity groups. Genotypes of secondary and tertiary gene pools clustered separately into two sub-clusters. R. aurea and C. platycarpus belong to the tertiary gene pool due to poor crossability with the cultivated pigeonpea, but they showed genetic similarities with C. cajan. These results are also supported by Raju et al. [20] who used 15 EST-SSRs to study the genetic diversity of 32 cultivars and eight accessions of two Cajanus species C. platycarpus and C. scarabaeoides. Earlier, a close relationship was reported between C. cajanifolius and C. cajan using genomic SSR markers [17, 19], but our study based on genic-SSR markers showed that C. cajanifolius is more distant to C. cajan compared to C. platycarpus. UPAS 120, TTB 7, Pusa Dwarf and Bahar genotypes belonging to different maturity groups were part of a single cluster. Genotypes of secondary and tertiary gene pools clustered separately in two sub-clusters, but R. bracteata which belongs to the tertiary gene pool based on the crossability criteria clustered with genotypes of the secondary gene pool. The closeness between Cajanus and Rhynchosia is also supported by morphological and genetic evidence, i.e. the presence of strophiole, an important characteristic used to distinguish between the genera. Seeds of Cajanus and Rhynchosia are generally described without strophioles. Various species of Rhynchosia, even though genetically closer to Cajanas, fail to produce hybrids because of reproductive barriers, and therefore Rhynchosia and Cajanus are classified as separate genera. High resolution mapping of these genotypes using a large number of genomic markers for diversity analysis may provide different results because genic-SSRs represent the transcribed portion of the genome, while the repetitive heterochromatin portion of the genome plays a major role in the evolution of species [54].

This is the first report of development and validation of a comprehensive set of genic-SSR markers in pigeonpea by deep transcriptome sequencing using next generation sequencing technology. A set of 2,877 genic-SSR markers was developed, and 550 SSR markers from this were validated for robust amplification in eight pigeonpea varieties, that will be useful for diversity analysis as well as mapping and tagging of genes and quantitative trait loci for economically important traits in pigeonpea.

A dataset of 43,324 TSA unigene contigs derived from 1.69 million 454 GS-FLX sequence reads of two pigeonpea varieties was produced. A comprehensive set of 2,877 genic-SSR markers was developed and 550 of these were validated for amplification and polymorphism, which will be useful for the development of molecular maps based on genic markers. Of the 550, 20 highly polymorphic markers identified all the individuals of a set of 30 genotypes including cultivars and wild species. Due to conservation of genic sequences these markers have a higher chance of transferability across species, compared to genomic SSR markers which show high polymorphism but are less conserved between species. A combination of these genic-SSR markers, single nucleotide polymorphism markers being mined from the TSA contigs assembled in this study and genomic SSR markers developed in other laboratories will be a powerful resource for molecular taxonomic studies and construction of a reference molecular map of the pigeonpea genome. Since genic-SSR markers belong to the gene-rich regions of the genome, some of these can be exploited for use in marker-assisted breeding of pigeonpea. Therefore, the set of genic-SSR markers developed here is a promising genomic resource.