Patterns of Molecular Diversity in Current and Previously Developed Hybrid Parents of Pearl Millet [ Pennisetum glaucum ( L . ) R . Br . ]

ICRISAT’s pearl millet (Pennisetum glaucum (L.) R. Br.) breeding program at Patancheru, India, has developed genetically diverse hybrid parents since 1980s. The present study investigated genetic diversity pattern between two groups of parents in this program, bred till 2004 and developed during 2004-2010. Combined analysis of 379 hybrid parents (current 166 parents and 213 previously developed hybrid parents) carried out using a set of highly polymorphic 28 SSRs detected 12.7 alleles per locus. An average of 8.5 and 8.7 SSR alleles per locus were found in previously developed and current parents, respectively, indicating marginal improvement in the levels of genetic diversity of hybrid parents in this program. Distance matrix differentiated these current and previously developed hybrid parents into 2 separate clusters, indicating infusion of new genetic variability over time as reflected by development of more genotype-specific alleles. Also, the seed and restorer parents were found clearly separated from each other in both the sets with few crossovers, indicating existence of two diverse and broad-based pools in hybrid parents of pearl millet. Restorer parents (R-lines) were found more diverse than seed parents (B-lines), as higher average gene diversity was detected among R-lines (0.70) than B-lines (0.56), though variation between Band R-lines was found reduced in newly developed lines to 9.22% from 16.98% in previously developed lines. Results suggested that newly developed lines were as much divergent when compared with previously developed lines, indicating that current ICRISAT pearl millet breeding program was moving towards development of diverse new hybrid parental lines. The study suggested use of trait-specific donors in Band R-lines separately to maintain sufficient genetic distance between seed and restorer breeding lines. It was pointed out to cross parents having higher genetic distance within the seed (B-lines) and restorer (R-lines) breeding programs to derive diverse and productive hybrid parental lines in future.


Introduction
Pearl millet (Pennisetum glaucum (L.) R. Br.) is cultivated on about 30 million ha in more than 30 countries with the majority of this area in Asia (>10 million ha), Africa (about 18 million ha), and >2 million ha in the Americas.It is important as a high temperature and drought tolerant grain crop in Africa and Asia, particularly on low-fertility soils with limited water and nutrient holding capacity and is equally important as a hot season forage crop in semi-arid and arid parts of the world [1].In Americas, it is used primarily as temporary summer pasture or as mulch in no-till soybean production systems [2].India is the largest producer of this crop with 9.6 million ha area and 10 million tons of grain production [3].The productivity of pearl millet in India rose from 305 kg•ha −1 during pre-hybrid era (1951)(1952)(1953)(1954)(1955) to 1040 kg•ha −1 during present hybrid dominated era (2008)(2009)(2010)(2011)(2012)(2013), and registered 227% improvement [4].Hybrid technology contributed significantly to this increase in grain yield performance in India, and currently about 70% of the pearl millet area in India is sown to hybrids.Based on the expected significant contribution of hybrids to pearl millet productivity enhancement in India, in the 1980s ICRISAT-Patancheru re-aligned its research strategy to better complement those of the Indian national program and private seed industry, and so focused on genetic diversification of hybrid parents.A recently conducted study found that private seed companies developed 103 hybrids in India during 2000-2010, and 62 (60%) were directly or indirectly based on ICRISAT bred hybrid parents [5].
Genetic diversification of hybrid parents has been the key to developing the diverse range of hybrids required to adapt under different agro-ecologies in India.Hence, ICRISAT-Patancheru's pearl millet hybrid breeding program utilizes new genetic stocks regularly in the crossing program to widen the genetic diversity of hybrid parents.Hence, in the late 1990s and early 2000 years, there was an increase in the use of already designated hybrid parents and newly developed composites in ICRISAT pearl millet hybrid parent's development crossing program to derive new wave of genetically diverse inbreds.In this breeding process, it is important to monitor the levels of genetic diversity among these newly developed promising hybrid parents, and SSRs being environment independent multi-allelic genomic tools offer considerable advantages for this.Hence, the present study was undertaken to assess the genetic diversity in 166 newly developed hybrid parents (yet to be designated 88 B-lines and 78 R-lines, drawn from ICRISAT-Patancheru's hybrid breeding program in 2010) using a subset of highly polymorphic 28 SSR loci as suggested by previous studies [6]- [11] and results were compared with those for 213 previously developed hybrid parents (98 B-lines and 113 R-lines, developed during 1981 to 2004) to provide information on trends for genetic diversity in this breeding program over time.

Plant Material
Two sets of hybrid parents, the first set (set-I of 213 hybrid parental lines comprising of 98 seed parents and 115 restorer parents, designated between 1984 to 2004 at ICRISAT-Patancheru), and a second set (set-II) of 166 current hybrid parents comprising of 88 seed parents (numbered from B-1 to B-88) and 78 restorer parents (numbered from R-89 to R-166) bred by the hybrid parent breeding program of ICRISAT-Patancheru and available in 2010, were used in the present study.The set-I parents were characterized for morphological traits [12] [13] and using molecular markers [14].Tift 23D 2 B 1 , used as a reference genotype, was bred at the Coastal Plain Experiment Station, Tifton, GA, by introducing the d 2 dwarfing gene into the genetic background of an elite line Tift 23B 1 , which is a maintainer of the A 1 cytoplasmic-nuclear male sterility (CMS) system [15].

DNA Extraction
The B-and R-lines along with Tift 23D 2 B 1 were grown in small plastic pots in a greenhouse.Approximately 30 mg of leaf tissue from 20 -25 seedlings (14 d old) were collected and bulked after removing the leaf tips and midribs.The harvested leaf samples were immediately collected in 96-well plate that consisted of 95 lines and one control (Tift 23D 2 B 1 ).Genomic DNA was extracted following the standard protocol [16].Quantification of DNA, quality check and normalization up to 5 ng•μl −1 were done on agarose gel (1.2%, containing ethidium bromide).

Molecular Markers
A set of 23 genomic-and five expressed sequence tag-simple sequence repeat (SSR) primer pairs were selected (based on the high level of PIC detected in earlier studies), to genotype the B-and R-lines of both the sets (set-I and set-II) using an ABI 3700 DNA fragment analyzer (Applied Biosystems).Twenty-five of these 28 SSRs mapped on pearl millet genome: 4 to 7 SSR loci on linkage groups (LGs) 1, 2, 3, and 7; and 2 SSR loci each on LGs 5 and 6 [6] [17].

Polymerase Chain Reaction (PCR) and Genotyping
Polymerase chain reactions (PCRs) were performed in 5 μl volumes in 384-well PCR plates (AB Gene Rochester).Each PCR reaction mixture contained 5 ng of genomic DNA, 2 pmol/μl of each primer, 25 mM MgCl 2 , 2 mM of each deoxyribonucleotide triphosphate, 10× reaction buffer, and 0.2 U Amplitaq Gold Polymerase (Applied Bio-systems, India).After one denaturing step of 15 min at 94˚C, a touchdown amplification program was performed on GeneAmp 9700 thermal cycler (Applied Biosystems).This profile consisted of a denaturing step of 25 sec at 94˚C and an extension step of 30 sec at 72˚C.The initial annealing step was 20 sec at 64˚C for one cycle and subsequently the temperature was reduced by 1˚C for every cycle until a final temperature of 55˚C was reached.The annealing temperature of 55˚C was maintained for the last 35 cycles of the amplification, followed by the final extension of 72˚C for 7 min.PCR products were pooled post-PCR, which contained 1 μl of each of dye-labeled (FAM, VIC, NED and PET) product, 7 μl of formamide, 0.3 μl of the LIZ-labeled [500 (−250)] size standard, and 4.2 μl of distilled water.The DNA fragments were size-separated on ABI 3700 automatic DNA sequencer (Perkin-Elmer/Applied Biosystems).GeneScan 3.1 [18] was used to size peak patterns by using the internal LIZ-labeled (500 [−250]) size standard and Genotyper 3.1 [19] was used for allele calling.Final bins for individual marker were assigned with the help of marker repeat length using AlleloBin 2.0 program [20] and the binned data was used for further data analysis.

Data Analysis
Genotypic data for the 28 common SSR markers was analyzed collectively for all the 379 lines (213 lines of set-I and 166 lines of set-II), and also separately for set-I and set-II lines.PIC; allelic richness as determined by total number of alleles, alleles per locus, and occurrence of common, most frequent, rare, and unique alleles; gene diversity; and heterozygosity were estimated using the PowerMarker V3.0 software [21].Unique alleles are those that are present in one line but absent in other lines.Rare alleles are defined as those whose frequency is ≤1% in the investigated materials.Common alleles are defined as those whose frequency is 1% to 20% while the most frequent alleles are those whose frequency is >20%.This classification of alleles was done following method used by other workers [22] [23].
AMOVA was performed [24] for both the sets individually and also pooled analysis to estimate the variance components among and within B-and R-line groups.The F value-the fixation index (or Wright's F statistic) (Fst)-from the AMOVA analysis provided a measure of genetic differentiation of the sub-groups.A simple matching allele frequency-based distance matrix was used in the DARwin-5.0program [25] to construct a tree diagram to examine the genetic structure and diversity among B-and R-lines.The grouping of B-and R-lines into clusters and sub-clusters was done at 5% dissimilarity level.

Results
In the combined set of 379 hybrid parents, comprising of both previously designated (hereafter, referred to as set-I) as well as current (hereafter, referred as set-II) pearl millet hybrid parents, 355 alleles were detected by 28 SSR loci, with an average of 12.68 alleles per locus.The number of alleles per locus varied from 4 (Xpsmp2202, Xpsmp2246 & Xipes3048) to 40 (Xpsmp2218), with ten to twenty alleles at 15 of the 28 SSR loci (Table 1).Marker Xpsmp2218 had up to 40 alleles while Xpsmp2068, Xpsmp2079 and Xpsmp2218 amplified 23 to 40 alleles per locus.The allele size range across the loci and lines varied from 4 bp (Xpsmp2201) to 121 bp (Xpsmp2079).Polymorphic information content, (PIC) ranged from 0.26 (Xpsmp2222) to 0.93 (Xpsmp2218) with an average of 0.67.Of the 28 SSRs used, 25 were moderately polymorphic (PIC > 0.44), of which 12 were highly polymorphic (PIC > 0.70) with PIC values ranging from 0.70 to 0.93.Furthermore, the average PIC values for B-and R-lines were 0.56 and 0.70, respectively.Gene diversity, defined as the probability that two randomly chosen alleles from the population are different, varied from 0.28 (Xpsmp2222) to 0.93 (Xpsmp2218) with an average of 0.70.Restorer lines (R-lines) had higher average gene diversity (0.73) than B-lines (0.60).The level of heterozygosity in SSRs across B-and R-lines ranged from 0.01 to 0.07 (excluding for Xpsmp2077 which had high value of 0.12) and averaging 0.04, with >0.05 in eight SSRs and >0.10 in one SSR.The R-lines had higher average heterozygosity (0.05) than B-lines (0.03).Moreover, ICMB 96666, B-23, B-24, B-57, B-65, B-6, B-45 and B-46 among B-lines; and ICMP 451, IPC 337, IPC 492, IPC 962, IPC 997, IPC 1018, IPC 1027, IPC 1078, IPC 1307, IPC 1503, IPC 1617, R-118, R-125, R-128, R-131, R-144, R-148, R-155, R-159, R-160, and R-166 among R-lines were genetically more variable than other lines, with heterozygosity at four to eight SSR loci detected each in the B-lines and in the R-lines (data not presented).Allele size for the internal control (Tift 23D 2 B 1 ) were uniform and reproducible for each of the markers indicating the accuracy of the protocol and reproducibility of allelic data for a given primer across assays.
The combined Analysis of Molecular Variance (AMOVA) for all the lines showed significant differences between set-I and set-II, which accounted for 19.20% of the total genetic variation (Table 3).Also, significant differences were observed between B-and R-line groups within each set, which accounted for 11.55% of the total genetic variation.The within-group variation, both in the B-lines and R-lines, was much larger and accounted for 69.24% of the total observed genetic variation.Set-wise AMOVA showed higher variation between B-and R-line groups in set-I (16.98%) than in set-II (9.92%), as detected by 28 SSR loci.The fixation index (Fst) between B-and R-lines for the individual markers varied from 0.138 (Xpsmp2222) to 0.456 (Xpsmp2218) and was significant for all 28 SSR primer pairs (information not presented).Some of the markers such as Xipes3088, Xpsmp2209, Xpsmp2203, Xpsmp2068, Xpsmp2089, Xpsmp2079 and Xpsmp2218 had Fst values greater than 0.40, contributed most to the variation between B-and R-line groups.
Mean for number of alleles per locus, gene diversity, heterozygosity and PIC for set-I using the set of 28 SSRs was 8.50, 0.63 (0.50 for B-lines and 0.63 for R-lines), 0.03 (0.02 for B-lines and 0.04 for R-lines), and 0.59 (0.46 for B-lines and 0.60 for R-lines) respectively (Table 4).In set-II of current pearl millet hybrid parents, 28 SSR loci detected a total of 244 alleles among 88 B-and 78 R-lines, with an average of 8.7 alleles per locus, and PIC ranged from 0.17 (Xpsmp2203) to 0.91 (Xpsmp2089) with an average of 0.57; the average PIC values for B-and R-lines were 0.46 and 0.62, respectively; gene diversity varied from 0.19 (Xpsmp2204) to 0.91 (Xpsmp2089) with an average of 0.61; restorer lines (R-lines) had higher average gene diversity (0.66) than B-lines (0.50); and the R-lines had higher average heterozygosity (0.05) than B-lines (0.04).Of the 244 alleles detected in B-and R-lines, more alleles were observed in R-lines (222 alleles in 78 lines averaging 2.85 alleles per line) than the B-lines (163 alleles in 88 lines averaging 1.85 alleles per line).
The neighbor-joining (NJ) tree based on a simple matching dissimilarity matrix of both the sets clearly differentiated set-I hybrid parents from set-II hybrid parents (Figure 1) into two separate clusters.11 B-lines (11/186) were found in the R-lines cluster, while 30 R-lines (30/193) were found in the B-lines cluster.The dissimilarity matrix for 166 set-II parents clearly differentiated the majority of B-and R-lines into two separate clusters (Figure 2).Tree diagrams were also constructed separately for set-II B-and R-lines, which further had four sub-clusters each for B-and R-lines (Figure 3).However, 7 B-lines grouped within the R-line cluster and eighteen R-lines grouped with the B-line cluster for this set of lines.Nine of the 18 R-lines found in the B-line cluster had presence of seed parents (B-lines) in their parentage, of which 7 had the ICRISAT High Head Volume B-Composite (HHVBC) or its derived progeny (Table 5).Of the 7 B-lines found clustered with R-lines, 2 had ICTP 8202 in their parentage and one was derived from Nigerian Composite D2 Dwarf (NCD2) [26].programs to maximize heterosis for higher grain yields.Thus, the ICRISAT-Patancheru pearl millet hybrid parent breeding program utilizes different kinds of germplasm and breeding materials in its crossing program to diversify the genetic base of the hybrid parents it develops.In the present study, the combined clustering analysis of 379 hybrid parents (213 previously designated parents of set-I and 166 newly developed parents of set-II) based on Polymerase Chain Reaction (PCR) products detected by 28 SSR primer pairs, showed 12.68 alleles per locus.This value was higher than reported earlier on pearl millet hybrid parents using SSR markers.For instance, Kapila et al. [10] detected 6.26 alleles per primer pair among 72 inbred lines (primarily B-lines bred at ICRISAT-Patancheru); and Sumathi et al. [11] reported 2.76 alleles per primer pair among 42 inbred lines bred This established that the levels of genetic diversity are consistent in this breeding program over time.The dissimilarity matrix-based tree clearly differentiated set-I from set-II lines, indicating that new variability is being continuously infused into the ICRISAT-Patancheru's pearl millet breeding program.New genetic stocks are regularly used in the pearl millet breeding program at ICRISAT-Patancheru to increase the genetic diversity of hybrid parents in the development of B-and R-lines.In the late 1990s and early 2000 years, there was increase in the use of already designated hybrid parents and newly developed composites in crossing program to derive new wave of genetically diverse second cycle inbreds.This was reflected from the parentages of lines under study, as new composites and improved populations, like MC 94, AIMP 92901, ESRC II, JBV 3, HHVBC, Raj 171, SDMV 90031, GB 8735, SDMV 95045, ICMS 7704, and RCB were found in the parentage of 48 of the R-lines in set-II (and none of those in set-I).Moreover, once the hybrid parents are developed, some of the promising ones amongst these first cycle parents are involved in the crossing programs to initiate another cycle of inbreeding and selection to derive new inbreds, as was reflected from set-II lines where 42 of the 88 B-lines assessed had 19 B-lines designated between 1992 to 2004 in their parentage.This approach of crossing elite inbreds is also practiced in maize to derive new inbreds [32] [33].Contrary to our findings, reduction in the genetic diversity of new inbreds was found in comparison to the historical maize inbreds, in a study conducted on 8 new and 32 historical inbreds using 83 SSR markers [34].

Within-Line Genetic Variability in the Hybrid Parents
The levels of heterozygosity detected for most of the SSR primer pairs in B-and R-lines of both the sets was within acceptable limits (<0.05), while it was high up to 0.12 in case of some primer pairs.A few SSR primer pairs detected heterozygosity in greater numbers of B-and R-lines than others.For instance, SSR locus Xpsmp2077 detected heterozygosity in 10 B-lines and 22 R-lines in set II.Likewise, some of the B-and R-lines were more heterozygous than others.For example, B-line (B-23) showed heterozygosity for eight SSR primer pairs while R-lines such as R-144 and R-160 were found heterozygous for 7 to 8 SSR primer pairs.The probable reason for high heterozygosity in some inbreds in comparison to others may be either due to 1) residual heterozygosity, 2) mutation and/or mutational bias at specific SSR loci, or 3) duplicate loci [35] [36].Some level of within-line genetic variability has been reported earlier in advanced generation hybrid parents of pearl millet [37] [38].Similar to these results, within-line variability has also been found in long-time bred maize inbreds for SSR markers [39] [40].

Genetic Diversity between Seed and Restorer Parents and Its Linkage with Parentage
ICRISAT follows a trait-based breeding approach to develop a phenotypically diverse range of hybrid parents to meet the diverse needs of various agro-ecologies, with high grain yield as a common desired trait.The traits used as selection criteria are based on regional preferences, which include various maturity types, plant height (grain vs. dual-purpose), tillering ability, panicle traits (length, thickness, and compactness), and seed traits (color and seed size).Seed parents (B-lines) are generally bred for short height (<100 cm) and for high grain yield component traits, while R-lines are generally bred for taller height (150 -180 cm), more tillers, relatively smaller seed size, and profuse pollen production [41].This unique differential trait requirement of B-and R-lines is met by involving quite diverse parents in each of these B-and R-line development programs.This distinctness in parentage and ideotypes was reflected in marker-based clustering pattern of set I and set II lines which clearly separated B-lines from R-lines in two clusters.This was also evidenced from combined AMOVA of both the sets, which indicated significant variation between B-and R-line groups.The grouping of B-and R-lines into two separate clusters was well detected by markers under study, as 27 of the 28 markers had significant Fst.This marker-based clustering pattern for these recently bred 166 set-II B-and R-lines, which separated most of the B-and R-lines into 2 separate clusters, was also reported earlier for ICRISAT-bred designated Band R-line hybrid parents [14], indicating existence of two diverse and broad-based pools in hybrid parents of pearl millet.Genetic variation assessed for 166 newly developed hybrid parental inbreds (88 B-lines and 78 R-lines) using 28 SSR primer pairs detected an average gene diversity of 0.61, indicating sufficient polymorphism to characterize this set of breeding lines for their genetic diversity.The clustering pattern of newly developed set-II Band R-lines further revealed that the B-and R-lines each grouped into four sub-clusters within their respective cluster, and sharing of a common parent can contribute to this clustering of inbreds in common sub-cluster.Se-venteen of the 22 B-lines in cluster B-I had 843B in their parentage, while 7 of the 15 lines in cluster B-II had involvement of both 843B and 81B in their parentage.Seed parental lines sharing 843B in their parentage were scattered in all the four sub-clusters, with the maximum number being found in cluster B-I, which further supported earlier observation about clustering of most of the B-lines with 843B in their parentage in one large cluster [10] [14].Also, 14 of 24 B-lines in cluster B-III had involvement of progeny from a single composite HHVBC, which was used as a source of thick-panicles in the crossing programs.Earlier, such clustering of thick-panicled B-lines derived from HHVBC, like ICMB 00555, ICMB 01222 and ICMB 01333 was observed in pearl millet [13].Likewise, 4 of the 8 R-lines in cluster R-II had lines derived from ICRISAT's Smut Resistant Composite (SRC) in their parentage, while 13 of the 19 R-lines in cluster R-IV had either progeny from MRC (Mandore Restorer Composite), Rajasthan Composite Bajra (RCB) or Rajasthan Inbred Bajra (RIB), with each of these three parental composites having been constituted using breeding lines from the same arid parts of western India.

Pattern of Genetic Diversity in Hybrid Parents and Breeding Implications
Most of the 379 ICRISAT-Patancheru bred pearl millet B-and R-lines were found in their respective groups based on the clustering patterns detected by 28 SSR primer pairs, as 94% of the B-lines and 84% of the R-lines were found in their respective B-and R-line clusters, indicating existence of two diverse and broad-based pools, one each representing seed and restorer parents.However, the variation between B-lines and R-lines has decreased over time as evident from decrease of variation between B-and R-lines from 16.98% in set-I to 9.92% in set-II.This narrowing down of variability between B-and R-lines might be due to involvement of certain common parents, both in seed parent (B-line) and restorer parent (R-line) breeding programs.Though, mostly a specific genetic stock is used either in the B-line or R-line crossing program based on its plant ideotype and fertility/sterility reaction, occasionally a particular line owing to its very specific trait is used simultaneously in development of both B-and R-lines.For instance, thick-panicle progenies derived from a B-composite (HHVBC) should be used only in the B-line program due to their closeness to B-plant ideotype, and also for having involvement of B-lines in its constitution.But owing to the shortage of donors for thick-panicle trait in R-line backgrounds, this line has been used occasionally in the R-line crossing program to introgress thick-panicle in R-lines.This approach has the risk of narrowing down the variability between B-and R-lines, as is reflected from the clustering of 23% of set-II R-lines (18 of 78 R-lines) with B-line clusters.Nine of these 18 R-lines had one or more B-lines in their parentage, of which 7 had involvement of a progeny from seed parent composite HHVBC.Also, of the 7 B-lines found clustered with R-lines, 2 had of progenies of a population ICTP 8202 in their parentage, a population with several restorer plant ideotype traits.Hence, efforts must be made to search for trait-specific donors separately in the B-and R-specific backgrounds for use in future crossing purposes, rather than continuing to use the same genetic stocks for a common trait in both of B-and R-line breeding programs.The separation of most of B-and R-lines into two separate clusters, in both previously designated and newly developed hybrid parents suggested that these two broad pools are continuously separated from each other, and B × B and R × R crosses should be made between and within set-I and set-II of B-and R-lines to generate diverse range of new recombinants for B-and R-line development.
The present study identified 15 B-and 37 R-lines from both the sets of lines with unique alleles (1 to 2 alleles) not present in other lines that can be used as genetic tags to supplement distinctness, uniformity, and stability (DUS) tests, thus enabling ICRISAT to protect these from possible infringement by multiple users who have received these lines for use in their hybrid breeding programs.Furthermore, presence of line-specific alleles may have association with some distinct trait(s) of the particular line, which merits further investigation.Moreover, the higher number of rare and unique alleles found among both B-and R-lines in this study suggests that these lines have been derived from a very diverse genetic base (which is in agreement with the diverse pedigrees of the lines included in this study).

Conclusion
The present study clearly differentiated newly developed hybrid parents bred by the pearl millet hybrid parents breeding program at ICRISAT-Patancheru from the previously designated hybrid parents bred at this location, indicating continuous involvement of new genetic stocks by the ICRISAT-Patancheru pearl millet hybrid parents breeding program, while maintaining consistent levels of genetic diversity among hybrid parents over time.

Figure 1 .
Figure 1.Unweighted neighbor-joining tree based on a simple matching dissimilarity matrix for allele sizes, detected by 28 simple sequence repeat primer pairs across 379 pearl millet hybrid parental lines (213 set-I lines and 166 set-II lines).Accessions are identified as "B" for seed parental lines and "R" as restorer lines.Suffix 1 and 2 with B-and R-lines represent set-I and set-II lines, respectively.B-lines are shown in blue and R-lines in red color.

Figure 2 .
Figure 2. Unweighted neighbor-joining tree based on a simple matching dissimilarity matrix for allele sizes detected by 28 simple sequence repeat primer pairs across 88 seed parental lines (B-lines; B-1 to B-88) and 78 restorer lines (R-lines; R-89 to R-166) of pearl millet set-II.(EEBC) in their parentage; EEBC is an extra-early-maturing and day-length insensitive maintainer composite developed using iniadi germplasm[28].In cluster B-II, 7 lines had both 843B and 81B[29] in their parentage.Eleven lines, derived either directly from HHVBC or having its progeny in their parentage, were found very close in cluster B-III, along with 3 other such lines in this sub-cluster.Sub-cluster B-IV, with maximum number of 27 lines, had lines with mixed parentage.For instance, it had 8 lines with 843B, 13 lines with HHVBC progenies, and 5 lines with EEBC progenies in their parentages.Likewise, the 78 R-lines formed 4 sub-clusters (R-I, R-II, R-III and R-IV) comprised of 12, 8, 39, and 19 lines (Figure3(b)).Four of the 8 lines in cluster R-II had lines derived from the ICRISAT Smut Resistant Composite (SRC) in their parentage.Of the 19 lines in cluster R-IV, 8 lines had progenies of MRC (Mandore Restorer Composite), and 5 lines had Rajasthan Composite Bajra (RCB) or Rajasthan Inbred Bajra (RIB) progenies in their parentage.Clusters R-I and R-III had lines with mixed pedigrees.ICRISAT-bred B-lines designated between 1992 and 2004 were found in the parentage of 42 of the 88 B-lines in set-II.New pearl millet composites and improved breeding populations, like the ICRISAT Medium Composite 94 (MC 94), released composite variety AIMP 92901 (based on the ICRISAT Bold-Seeded Early Composite,

1 .Figure 3 .
Figure 3. Un-weighted neighbor-joining tree based on a simple matching dissimilarity matrix for allele sizes detected by 28 simple sequence repeat primer pairs in set-II lines.(a) Across 88 B-lines (B-1 to B-88); and (b) Across 77 R-lines (R-89 to R-166).

Table 1 .
Allelic composition, polymorphic information content (PIC), gene diversity, and observed heterozygosity of the 28 simple sequence repeat (SSR) loci in combined set of pearl millet (186 B-lines and 193 R-lines).

Table 2 .
Genotype-specific alleles present in 15 seed parents (B-lines) and 37 restorer parents (R-lines) of pearl millet (from both the sets).

Table 3 .
Analysis of molecular variance (AMOVA) of set-I, set-II and for combined sets using 28 SSRs.

Table 4 .
Genetic diversity parameters of set-I and set-II lines using same set of the 28 simple sequence repeat (SSR) loci in pearl millet.