The experimental material comprised of 94 pigeonpea genotypes, which included 59 varieties, released during 1975–2015, 23 breeding lines/accessions, four exotic lines, four landraces, three fertility restorers, and two CMS lines (Table 1). These genotypes are relevant to pigeonpea breeding objectives owing to their suitability to different agro-climatic zones and having several traits of interests such as high yield and disease resistance etc (Supplementary Table 1). A subset of these genotypes (40) was screened against Fusarium wilt (FW) variant 2 during 2014–15 in the wilt sick field at IIPR, Kanpur (Supplementary Table 2). The 40 genotypes were selected because of their breeding value (extensively used by the breeders) for the improvement of resistance to FW. Few genotypes, e.g., IPA 8F, IPA 9F, and IPA 16F were registered with the National Bureau of Plant Genetic Resources (NBPGR), New Delhi as donors for FW resistance (Singh et al., 2011). Similarly, the genotypes like BDN 2, C11, MAL 13, and ICP 8863 are being used as differential set to characterize virulence of the wilt pathogen. The experiment was conducted in randomized block design (RBD) with two replications. The FW incidence was recorded on plants from November to February as the weather conditions during this period favor wilt incidence. Percent incidence was calculated using the formula: % Disease incidence=(Number of infected plants/Total                                 number of plants screened)×100 Arcsine transformation was used to normalize the data.

Genomic DNA was extracted from 94 pigeonpea genotypes for diversity analysis. Initially eight genotypes (Type 7, ICP 8863, PA 163A, ICPA 2089, AK 261322R, AK 261354R, AK 250189R, AK 250173R) were selected to assess the amplification status of SSR markers. For hybrid purity testing, DNA was extracted from 10 individuals each of three CMS hybrids viz. IPAH 16-06, IPAH 16-07, and IPH 15-03. To perform genetic linkage analysis, DNA from 102 individuals of a backcross population [ICPL 88039A × (ICPL 88039A × AK 250189R)] was also extracted. Also, to gain insights about within genotype variability in landraces DNA was isolated from 15 individuals each of six genotypes (two varieties and four landraces). Genomic DNA was isolated from young leaves according to Cuc et al. (2008). The quantity and quality of DNA was estimated through electrophoresis using 0.8% agarose gel (Sigma-Aldrich, St. Louis, USA).

For amplification of genomic DNA, a reaction mixture of 10 μl volume was prepared using 5.9 μl of sterilized distilled water, 1.0 μl template DNA (25 ng), 0.5 μl of forward and 0.5 μl of reverse primer (5 μM), 1.0 μl 10 × PCR buffer (10 mM Tris-HCl, 50 mM KCl, pH 8.3), 1.00 μl dNTP mix (0.2 mM each of dATP, dGTP, dCTP, and dTTP), and 0.1 μl Taq polymerase (5 U/μl) (Thermo Scientific, Mumbai, India). The amplification was carried out in G-40402 thermo cycler (G-STORM, Somerset, UK) in a touchdown PCR profile, in which the following condition was set: Initial denaturation at 94°C for 5 min followed by 10 cycles of touchdown 55–45°C, 20 s at 94°C, annealing for 20 s at 55°C (the annealing temperature for each cycle being reduced by 1°C per cycle) and extension for 30 s at 72°C. This was accompanied by 40 cycle of denaturation at 94°C for 30 s, annealing at 50°C for 30 s, elongation at 72°C for 45 s, and 10 min of final extension at 72°C. Amplified products were resolved in 3% agarose gel using 0.5 × TBE running buffer and images were analyzed in Quantity one software (Bio-Rad, CA 94547, USA).

The genetic diversity parameters such as major allele frequency, polymorphic information content (PIC), heterozygosity and alleles per locus were computed using PowerMarker v. 3.25 (Liu and Muse, 2005). Nei's gene diversity (h) and Shannon information index (I) were estimated in the sample using POPGENE v. 1.32. The model-based Bayesian approach was used to infer population structure with STRUCTURE v. 2.3.4 (Pritchard et al., 2000). The project was run with the admixture model and correlated allele frequency using burn in period of 20,000 and 200,000 Markov chain Monte Carlo (MCMC) replications. Five independent runs were performed with each K value ranging from 1 to 10. Evanno's Δk value was calculated by using STRUCTURE HARVESTER program by processing the STRUCTURE results (Evanno et al., 2005). Concerning distance based clustering, DARwin v. 6.0.13 (Perrier and Jacquemoud-Collet, 2006) was employed to generate genetic distance (GD) matrix, which was then used to create dendrogram using unweighted neighbor joining. Factorial analysis was also performed with GD matrix created using DARwin software. We performed analysis of molecular variance (AMOVA) among and within subpopulations (assigned by STRUCTURE), implemented in GenAlEx (Peakall and Smouse, 2012). GenAlEx was also employed to cluster the genetic variation by performing principal coordinate analysis (PCoA).

The MTA for FW resistance was detected using TASSEL v. 2.1 (Bradbury et al., 2007) as described by Iqbal and Rahman (2017). The general linear model (GLM) and mixed linear model (MLM) based on Q matrix and Q+K matrix, respectively were employed to discover SSR markers associated with FW resistance. The structure analysis generated the Q matrix, while the relative kinship matrix was calculated by TASSEL software. Significant MTAs were declared at P ≤ 0.05 with corresponding R² indicating the phenotypic variation explained by the MTA.

In parallel, a “coarse” linkage map was developed for the backcross population using IciMapping v. 4.1 with a minimum logarithm of odds (LOD) value of 2.5 (Meng et al., 2015). Segregation data was assembled for 75 SSR markers. By using 1% acetocarmine solution, segregation data on pollen fertility and sterility were reported earlier in this backcross population (Bohra et al., 2017). Kosambi mapping function was used to calculate genetic distances. Linkage map was drawn with MapChart v. 2.5 (Voorrips, 2002).

A set of 23,410 primer pairs was designed out of 309,052 SSRs identified in pigeonpea genome (Varshney et al., 2012). In the current analysis, 421 SSRs were selected based on length of the SSR tract (Class I or hypervariable) for amplification in pigeonpea lines. These 421 SSR markers were assayed on eight pigeonpea genotypes viz. Type 7, ICP 8863, PA 163A, ICPA 2089, AK 261322R, AK 261354R, AK 250189R, AK 250173R that are parents of different mapping populations (F₂ and backcross) segregating for important traits such as Fusarium wilt and fertility restoration. Of the total 421 markers, 401 provided scorable amplicons and 217 markers generated polymorphic fragments. Twenty SSR markers failed to show amplification in any of the eight genotypes (Supplementary Table 3). In perfect SSR category (as defined by Weber, 1990), tetra- (NNNN), and penta-nucleotide (NNNNN) repeats showed 60% polymorphism followed by tri (NNN) and hexa-nucleotide (NNNNNN) repeats (Table 2).

The polymorphism information content (PIC) of these 217 SSRs was in the range of 0.11–0.71 with an average 0.38 (Table 3). A total of 570 alleles were detected and the number of alleles per marker ranged from 2 to 6 with an average of 3. Similarly, the range of gene diversity lied between 0.12 and 0.75 with a mean of 0.44. The highest PIC value was shown by the marker CcGM20721, while the marker CcGM21506 generated maximum number of alleles across eight genotypes. We also attempted to find out a relation between the length of the SSR tract and the PIC value with the dataset comprising only perfect SSRs. However, a weak positive relationship (r = 0.1427 and r² = 0.0204) could be inferred from the dataset.

In the case of pairwise polymorphism among parental lines of mapping populations a total 179 SSR markers showed polymorphism at least within one crossing combination (Supplementary Table 3). The number of polymorphic markers for parental combinations ranged from 39 (PA 163A × AK 261322R) to 89 (Type 7 × ICP 8863).

Keeping in view the high quality marker profiling patterns and PIC values, a subset of 60 markers was selected from the 217 polymorphic markers for analyzing 94 genotypes. As a result, a total of 233 alleles were obtained across the genotypes. The average values for PIC and number of alleles per marker were 0.50 and 3.88, respectively. A representative image illustrating the SSR fingerprints of the 94 genotypes using the marker CcGM22990 has been shown in Figure 1. Though only four landraces were used in the current analysis, we analyzed 15 plants of these landraces and two varieties (IPA 203 and Narendra Arhar 1) using CcGM18684 to evaluate the within-genotype variability. No within genotype variability and heterozygosity was observed in the varieties. By contrast, the three landraces showed two alleles each and variable level of heterozygosity [0.06 (Banda Palera and Allahabad Local) and 0.2 (JBP-13)] (Supplementary Figure 1).

To examine the genetic variations identified through 60 SSR markers across 94 genotypes, model- as well as distance-based approaches were used. In model based clustering, five independent runs were performed using STRUCTURE programme with K values ranging from 1 to 10. The maximum delta K (ad-hoc quantity) was reached at K = 2 suggesting presence of two subpopulations in the collection (Figure 2). Of the total 94 genotypes, 28 belonged to subpopulation 1, while remaining 66 corresponded to subpopulation 2 in STRUCTURE analysis. With a few exceptions, the subpopulation 2 contained short- and medium-duration pigeonpea, whereas the genotypes with longer maturity duration were in subpopulation 1. The mean Nei's gene diversity was found to be 0.50 and 0.56 for subpopulation 1 and 2, respectively. Similarly, mean values for Shannon information index for subpopulation 1 and 2 were 0.91 and 1.04, respectively. Overall, the average values of Nei's gene diversity and Shannon information index were 0.58 and 1.089, respectively.

A neighbor-joining tree was constructed based on genetic distance matrix, which suggested existence of four clusters (Figure 3). The factorial analysis was also undertaken to offer an overall representation of the diversity in the studied panel. Interestingly, a close agreement was observed between the results arising from STRUCTURE and factorial analysis. The subpopulation 2 of STRUCTURE was contained primarily in quadrants II and III of factorial analysis (Figure 4). On the other hand, quadrant I contained genotypes that were assigned to subpopulation 1 in STRUCTURE analysis. Similarly, two clusters were also obtained in PCoA with the PC1 and PC2 explaining 10.47 and 7.98% of the variance, respectively (Figure 5). Further, the total genetic variation was partitioned by using AMOVA based on PhiPT-values, which accounted majority of the genetic variation to within groups (89%) in comparison to the variation observed between the groups (11%) (Table 4).

The marker genotyping data on 59 varieties released in India (a subset of above mentioned 94 genotypes) during 1975 and 2015 were also examined to fathom trend of genetic diversity according to both decadal period and zone. Four subgroups could be created each for decadal and zonal study. A pairwise comparison among different groups suggested the maximum GD between the varietal groups pertaining to decade 1 (1975–1985) and decade 4 (2005–2015), while the least GD was shown between the groups decade 1 (1975–1985) and decade 2 (1986–1995). On the other hand, the least GD was found between CZ and SZ groups in contrast to the highest GD between NWPZ and NEPZ (Table 5). The data pertaining to means of gene diversity and number of alleles among these groups are shown in Table 6. Concerning decadal trend, the estimates of gene diversity and average number of alleles per locus did not show large difference. In a similar manner, average values for gene diversity and allele count remain almost similar among the four groups obtained based on their suitability to different zones.

CMS technology has emerged as a promising means to deliver noticeable yield gains in pigeonpea. The CMS-based hybrid breeding involves three parental genotypes: male sterile (A) line, its isogenic line (B) line and fertility restorer (R) line. Ten SSR markers were tested in three CMS hybrids (IPAH 16-06, IPAH 16-07, and IPH 15-03) and respective parents i.e., A, B, and R lines. These hybrids are being tested in evaluation trails at different locations in India under all India coordinated research projects on pigeonpea (AICRPP) and consortium research platform on hybrid technology (CRPHT) schemes. The SSR markers that generated monomorphic fragments between A and B lines and polymorphic fragments between the A and R lines were selected for testing the genetic purity of the respective hybrids. Ten random samples were taken from each hybrid for SSR analysis. As a result, sets of SSRs that were found promising for the molecular characterization of the three CMS hybrids included three markers (CcGM16529, CcGM16633, CcGM16772) for IPAH 16-06; three markers (CcGM16772, CcGM17150, CcGM23176) for IPAH 16-07 and four markers (CcGM18291, CcGM17648, CcGM12217, CcGM16417) for IPH 15-03 (Figure 6).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.