Forty millet accessions evaluated in this study were provided by the Germplasm of the Genetic Resource Unit of the Agricultural Research Corporation, Wad Medani. The seeds were collected from 30 villages from different regions of Sudan, namely: 1) North Kordofan, 2) South Kordofan, 3) Blue Nile State, and 4) Darfur (Table 1). Two Sorghum genotypes (Tabat and Wad Ahmed) were also included in the study for comparison.

The DNA extraction was done by CTAB 2X method. Genomic DNA was extracted from fresh leaf tissues using modified CTAB method [13] . In this method the fine powdered plant materials were immediately transferred into 13 ml Falcon tubes containing 6 ml of pre-warmed lysis solution. Tubes containing the

N = North Kordofan, S = South Kordofan, B = Blue Nile State, D = Darfur.

samples were then incubated in water path at 65˚C with gentle shaking for 30 minutes and left to cool at room temperature for 5 min. Isoamyl alcohol and chloroform mixture (1:24) was added to each tube and the phases were mixed gently for 5 min. at room temperature to make a homogenous mixture. The cell debris was removed by centrifugation at 5000 rpm for 15 min at room temperature and the resulted clear aqueous phases (containing DNA) were transferred to new sterile tubes. The step of the chloroform: Isoamyl alcohol extraction was repeated twice. The nucleic acid in the aqueous phase was precipitated by adding equal volume of deep cooled isopropanol. The contents were mixed gently and collected by centrifugation at 4000 rpm for 10 min. Then the formed DNA pellet was washed twice with 70% ethanol and the ethanol was discarded after spinning with flash centrifugation the remained ethanol was removed by leaving the pellet to dry at room temperature. The pellet was dissolved in TE buffer and stored at −20˚C for further use. The extracted DNA samples were observed under UV illumination after staining with ethidium bromide and agarose gel electrophoresis.

Twenty primers were used for PCR amplification, ten primers that produced strongly amplified polymorphic bands with these test templates were selected for RAPD-PCR analysis (Table 2). The PCR reaction mixtures were prepared in 25 µl volumes containing 2.5 µl of 10X Taq buffer, 1.5 µl MgCl₂ (50 mM), 2.5 µl dNTPs (2 Mm/µl), 2 µl random primer (10 pmol/ µl), 0.5 µl Taq DNA polymerase (5 U/µl) and 1 µl of the extracted DNA (10 ng). The mixture was made up to 25 µl by addition of DDH₂O (double sterilized distilled water). RAPD/PCR reactions were optimized and initiated using an applied Biometra thermalcycler programmed to repeat the thermal profile. Setting of the PCR program is based on three steps. Step one, was an initial denaturation step at 94˚C for 5 mins. Step two, was run for 40 cycles, each starting with denaturation at 94˚C for 1 min, followed by annealing 36˚C for 1 min and ended by extension at 72˚C for 1 min. Step three, was a final extension cycle performed at 72˚C for 7 minutes .The PCR machine was adjusted to hold the product at 4˚C. The PCR product was mixed with 3 µl of loading dye and spun briefly in a micro centrifuge before loading. The PCR products and 1 kb DNA ladder were electrophoresed using 2% agarose gel at 100 volts followed by ethidium bromide then separated fragments and were visualized with ultraviolet (UV) transilluminator (Figure 1).

Each band in the RAPDs profile was treated as an independent locus with two alleles. The numbers of bands produced for each primer were scored manually for presence (1), or absence (0) and a binary matrix was generated and then used for further analysis [14] . The PhiPT (analogue of F_ST fixation index) value for genetic variability was calculated for the four regions using GenALEx v.6.1 [15] . With AMOVA, the variance components and their significance levels for variation

among regions and within regions were obtained by using RAPD data of the 40 accessions (Table 3) and (Figure 2).

A principal coordinate analysis (PCA) was conducted with GenALEx v.6.1. This multivariate approach was chosen to complement the cluster analysis information (Figure 3), because cluster analysis is more sensitive to closely related individuals whereas PCA is more informative regarding distances among major groups [16] .

Pairwise genetic distance between individuals were calculated by the percentage disagreement method (Table 4). These data were used in cluster analysis with the unweighted pair-group method using arithmetic averages (UPGMA), in which samples were grouped based on their similarity (Figure 4) with the aid of statistical software package STATISTCA-ver.6 [17] .

Phenotypic diversity and agronomic descriptors for the 40 millet accessions were also recorded based on 26 agronomical descriptors for each accession. These observed data were taken during the season 2006-2007 (Table 5).

Analysis of variance was done using the SPSS version 17 computer software [18] program in order to assort accessions according to their agronomic characters (Figure 5).

The analysis is based on RAPD phenotypes consisting of 63 bands states.

Ten primers which were informative (UBC-101, UBC-104, UBC-122, UBC-155, A-1, B-8, B-20, C-2, C-10, D-8) were selected and used to evaluate the degree of polymorphism and genetic relationship within and between all individuals under study (Table 2). The selected primers generated an appropriate amplification pattern with clear and consistent reproducible bands. The maximum

A = Plant height (cm), B = Stem diameter (cm), C = Early vigour, D = Total no. of tillers, E = no. of productive tillers, F = No. of Nodal tillers, G=Plant aspect, I = Green fodder yield potential, J = Spike shape, K = Spikelet shattering, L = Bristle length, M = Days to 50% flowering, N = Flowering range, O = Ear exertion type, P = Ear exertion distance (cm), Q = Spike length (cm), R = Spike thickness (cm), S = Seed color, T = Seed covering, U = Seed shape, V = No. of leaves, W = Senescence, X = Juice quality, Y = Stalk Juiciness, Z = 1000 Seed weight (g).

number of fragments was produced by the primer B-8 (10 bands) with 100% polymorphism while the minimum number of fragments was produced by the primer UBC-101 (4 bands) with 75% polymorphism. A total of 64 amplified fragments were distinguished across the selected primers and the statistical analysis showed 63 polymorphic bands among the 40 accessions with an average of 6.3 polymorphic bands per primer and only one monomorphic band with an average 0.1. The number of bands per primer ranged from 4 to 10 and the number of polymorphic bands per primer varied from 3 to 10. All the primers which were used in this study were 100% polymorphic, except primer UBC-101 which was 75% polymorphic (Table 2).

The analysis of molecular variance (AMOVA) for the four regions showed significant difference (P < 0.01), with 80% of the differentiation attributed to within regions and 20% attributed to among regions (Figure 2).

The PhiPT estimates were moderate to high (0.196) and significant (P = 0.01) (Table 3). The UPGMA and percent disagreement values (PDV) were used to estimate the degree of relationships between the individuals analyzed based on common amplified fragments. Based on the matrix obtained (Table 4), the percent disagreement values (PDV) for all individuals was ranging from 4% to 43%. Principal components analysis of molecular variance performed in GenAlex 6.1 was based on individual banding pattern.

The cluster analysis based on Euclidean distance (Figure 5) using agronomical characters grouped the forty pearl millet accessions into two main clusters. B32 was alone in a separate arm as the most divergent accession. Cluster one included

two accessions S17 and S22, cluster two divided into two subgroups, subgroup one includes two accessions and subgroup two divided into many sub-clusters. The evaluation of pearl millet accessions based on the 26 agronomical descriptors had shown that there is a high amount of variation among the accessions.

The genetic similarity obtained in this study ranged from 4% to 43%. These results are in agreement with other results reported by [12] who used 32 RAPD primers for 11 pearl millet accessions from different commercial agricultural areas (the Cerrados region of Brazil). They found that polymorphism among the pearl millet accessions tested, ranged from 9% to 76%. In the first cluster of the dendrogram, it is clear that millet accessions from Darfour except D33, Blue Nile and most millet accessions from South Kordofan grouped together. However, within this group, Darfour millet genotypes are clearly distinct as they fall in a separate sub-group and accessions from Blue Nile and South Kordofan fall in another sub group. All North Kordofan accessions fall in a second group containing only two accessions from South Kordofan and one accession from Darfour. The close genetic similarity which exhibited by accessions from different regions (Blue Nile and Darfur, Blue Nile and South Kordofan) might be as a result movement of seeds among these regions. The fact that it is very difficult for farmers to keep their own selected seeds and the consequent substantial seed exchanges between them is probably the main factor accounting for the genetic homogeneity which has been observed in this study. South Kordofan accessions were the most heterogeneous genotypes based on their scattered positions in the dendrogram and their highest percentage of polymorphic (88.2%). Whereas, Darfour genotypes were the most distinct millet accessions. So results from this study can help breeding programs and samples selections can become very easy, quick and useful for the millet breeders. The genetic “similarity” between the significant fraction of the South Kordofan pearl millet accessions and the Blue Nile accessions could illustrate a preferential seed flow. It is likely that South Kordofan region is a provider of seed lots to Blue Nile region that can be approved by the higher polymorphism found in South Kordofan region (88.2%), whereas Blue Nile maintains (68.7%) as shown from the regional diversity analysis. This illustrates a significant seed flow. PhiPT and F_ST are analogous standardizing measures of the degree of genetic differentiation among regions: scores for both measures range from 0 (non differentiation) to 1 (non alleles shared). The Principle Coordinate Analysis for the four Pennisetum glaucum regions showed clear scattering of the South Kordofan genotypes and all the other three regions. However, it was also clear that Darfur genotypes remained intact in a distant group. In this study, the PhiPT value obtained over all regions was (0.196) with high significance (P = 0.01), suggesting tendency to high structuring within the region. There was genetic variation of 80% within the regions and the estimated genetic variation was 20% among them. Every region still maintains its indigenous genetic material so that these farmers keep their own seeds sources. So the pattern that is depicted by the PCA analysis is in consistence with the UPGMA tree, the PCA revealed the real regions structure, Darfour grouped separately in the dendrogram and in the PCA graph. The high proportion of variation attributable to within the region (80%), in this study was consistent with the expected population structure for an out crossing species. For pearl millet, it is expected that the heterogeneity and heterozygosity at many loci within a landrace accession to be high. This is due to cross-pollinated nature of the crop. The self-pollinated species such as Hordeum spontaneum has 43% of within population variation [19] . In contrast, cross-pollinated species has been reported to have 72% - 100% of their total variation [20] [21] . Our results are in agreement with the findings of [22] , who studied genetic diversity of pearl millet in Niger by using AFLP marker. They found that most of the diversity was found at the population level. Also, a previous study done by [23] in India and by [24] in Nigeria reported 88% to 98% of the variation was found between individuals within populations. They concluded that the high diversity found within a population may be due to out-breeding reproduction system of pearl millet as well as from the seed flow observed at the regional level. Previous studies on pearl millet from Africa and India based on isozymes diversity indicated intra-population diversity of 70% - 90% of the total diversity, depending on their regions of origin [25] [26] . Generally, RAPD technique in this study was very useful. The study provided very useful knowledge about genetic variability of some Pennisetum glaucum accessions from Sudan which may allow more efficient and effective uses of resources in plant improvement programs. In this study, RAPD technique was utilized for better assessment of Pennisetum glaucum accessions relationship collected from four regions of Sudan. The genetic dissimilarity values obtained with RAPD have been introduced for measuring genetic relationships in many plant species for easiness of the method, which only requires PCR technology. This study clearly showed that it was possible to analyze the RAPD and for correlating their similarity and distance between Pennisetum glaucum accessions. From the results of RAPD profiling, it was observed that Pennisetum glaucum accessions produced very good amplification with the primers used. Molecular markers can be used to study the genetic diversity and genetic relationships among Pennisetum glaucum accessions at the DNA level. The present investigation of DNA profiling in Pennisetum glaucum species clearly demonstrates that it is possible to analyze the RAPD patterns for correlating their similarity and distance between accessions. Although accessions from Darfour had the least percentage of polymorphic fragments (60.3%) from all the other accessions, they were grouped together in both the dendrogram and the PCA analysis. This shows that, these accessions might maintain alleles (genes) that are unique to the region. North Kordofan had the second least percentage of polymorphic loci (62.9%). These accessions as well grouped together and were distant from those of Darfour. Although accessions from South Kordofan had the highest percentage of polymorphic loci (88.2%), they were dispersed with groups of North Kordofan and Blue Nile. The analyses of RAPD revealed high levels of genetic variability, even with the use of limited set of primers. This high level of polymorphism among individuals suggests that RAPD techniques can be useful for Pennisetum glaucum for the maintenance of Germplasm banks and the efficient selection of parents for breeding.

The results showed that B32 was least similar to others, and seems to be a different millet genotype. S22 was the second least similar genotype to other accessions. From the two dendrograms, the cluster based on data obtained from RAPD markers and the cluster based on the agronomical descriptors, showed no similarity between the two results. The accessions were distributed in different sites in both clusters, so the classification based on RAPD markers could not be related to agronomic descriptors about the accessions in this study. These results are in agreement with those results reported by [27] they studied complex samples in two pearl millet fields in two villages of southwestern Niger, by using both morphological and AFLP marker. Also the results are in agreement with [28] they studded diversity in 21 cotton genotypes by using RAPD marker and morphology markers, and [29] they used RAPD marker and morphological marker to study some olive trees. In all previous studies mentioned, there was no co-relation between morphology and genetic data.

The estimates of genetic relationship can be helpful for organizing germplasm, conservation of genetic resources, identification of cultivars for selection of parents for hybridization, and for predicting favorable heterotic combinations. This also helps to reduce the number of samples required for sampling of genetically variable broad range of accessions in breeding programs. Accessions with the most distinct DNA profiles are likely to contain the greatest number of novel alleles, which are likely to uncover the largest number of unique and potentially agronomically useful alleles [27] . However, using both morphological and molecular characterization, provide us with a more complete and informative characterization of the germplasm and this can make breeding programs more easy and useful.