Diversity analysis of sweet potato ( Ipomoea batatas [ L . ] Lam ) germplasm from Burkina Faso using morphological and simple sequence repeats markers

Collecting and characterizing plant material has been basic for crop improvement, and diversity has long been seen as vital for rational management and use of crops. Thirty (30) morphological characters and thirty (30) simple sequence repeat (SSR) markers were used to assess the diversity among 112 sweet potato (Ipomoea batatas [L.] Lam) cultivars in Burkina Faso and to develop a core collection. Eight morphological characters were able to differentiate the 112 accessions and to identify 11 duplicates while 28 SSR markers were more informative in discriminating the accessions and to identify five duplicates. The diversity assessment using the two approaches revealed high diversity with a coefficient of 0.73 using the phenotypic data, while moderate diversity with a coefficient of 0.49 was obtained using the SSR markers. These results show no correlation between the two approaches (with dissimilarity index of 0.95). A core collection was constituted using the SSR based data while the eight discriminative phenotypic descriptors will be used in the identification of cultivars.


INTRODUCTION
Sweet potato (Ipomoea batatas [L.] Lam), a hexaploid crop (2n = 6X = 90) is one of the most economically important crops in the world.In Burkina Faso, the major production areas are near the borders with Mali, Ghana, Togo and Benin suggesting that important exchanges of planting material has occurred between these neighbouring countries.Cultivar names differ from one location to another, therefore placing limitations on accurate identification on locally available sweet potato germplasm that is vital to the rational management and use of the crop.Collection, characterization and maintenance of local germplasm are the bases of varietal improvement (Mok and Schmiediche, 1998).
Morphological characterization has been used extensively on various crop plants diversity assessments in many places of the world (Bos et al., 2000;Kaplan, 2001;Lacroix et al., 2005;Li et al., 2009;K'Opondo, 2011).Despite the environmental influences on plant morphology, this direct inexpensive and easy to use method of estimations was perceived as the strongest determinant of the agronomic value and taxonomic classification of plants (Li et al., 2009) and the first step in the assessment of plant diversity.On sweet potato, this tool has been used successfully to analyse genetic diversity necessary for the germplasm conservation, to reduce accession number by identification and elimination of duplicates and to enhance crop breeding (Huaman, 1992;Mok and Schmiediche, 1998;Tairo et al., 2008;Li et al., 2009;Karuri et al., 2009;Yada et al., 2010a).
According to La Bonte (2002), when trait expression is environmentally unstable or difficult to evaluate, molecular markers become more useful than traditional phenotypic evaluations.During the last decade a lot of molecular information has been accumulated and used for genetic diversity assessment on sweet potato germplasm (Jarret et al., 1992;Kowyama et al., 1992;Jarret and Austin, 1994;Bruckner, 2004;Tseng et al., 2002;Hu et al., 2003;He et al., 2006;He et al., 2007;Soegianto et al., 2011).The most widely used molecular marker procedures for population genetic analysis of both animals and plants during the past few years are the simple sequence repeat (SSR) markers or microsatellites (Shih et al., 2002;Veasey et al., 2008;Zhang et al, 2001;Karuri et al., 2010;Yada et al, 2010b;Li et al., 2009) (Weising et al., 1995).These markers are highly polymorphic, co-dominant, and can easily be detected on high-resolution gels.
Limited success has been achieved with morphological diversity analysis alone (Yada et al., 2010a).Therefore, to optimize the characterization efficiency, morphological characterization has now been combined with molecular techniques.SSR markers have been used in combination with morphological descriptors to analyse genetic diversity in sweet potato germplasm and useful core collections have been developed using this combination (Li et al., 2009;Karuri et al., 2010).
The objective of this research was to quantify the diversity in sweet potato germplasm collected in Burkina Faso using morphological descriptors and SSR molecular markers.

Collection of plant materials
One hundred and forty-four (140) sweet potato accessions (Table 1) were collected from December 2008 to January 2009 and January 2010 from the main production areas located in the Cascades, Western, Central-West, Southern, Central-South, Central-East and Eastern regions of Burkina Faso using the method described by Huaman (1991).One hundred and seven (107) accessions survived and were maintained at the INERA research station of Kamboinse located in the centre of the country in the Soudanian zone characterized by an annual rainfall ranged from 600 to 1100 mm.Three varieties introduced from the International Potato Center (CIP) East Africa CIP-440001 (known as Resisto), CIP-199062-1 and TIB-440060, one from China (TN-Leo) and Tiebele-2 an orange fleshed sweet potato of unknown origin were added and used as control.

The experiment
The 112 accessions were grown at the INERA station of Kamboinsé during the rainy season, from July to October 2009.Based on the records of the first year, the experiment was replicated from July to October 2010 and the materials were planted in groups of relatedness to allow further morphological comparisons between those accessions which were morphologically alike.Planting was done on ridges of 3 m long with distance between ridges of 1 m.On each ridge, 11 cuttings were planted at a spacing of 30 cm.The fields were maintained by frequent weeding.NPK (14-23-14) fertilizer was applied 21 days after planting when the cuttings were well established.Additional watering was done by irrigation to complement rainfall.

Data collection
Morphological data were collected 60 days after planting based on the average of three measurements from the middle portion of the main stem as recommended by Huaman (1992).Qualitative characters were scored using a scale of 0 to 9. The following variables were scored: Plant growth characteristics: plant type (PT), ground cover (GC); mature vine characteristics: vine internode diameter (VID), vine internode length (VIL), predominant vine colour (PVC), secondary vine colour (SVC), vine tip pubescence (VTP); mature leaf characteristics: general outline of leaf (GOL), leaf lobes number (LLN), leaf lobes type (LLT), mature leaf size (MLS), abaxial leaf vein pigmentation (ALVP), petiole length (PL), petiole pigmentation (PP), shape of central leaf lobe (SCLL), mature leaf colour (MLC), immature leaf colour (ILC); flowering habit (FH); Storage root characteristics: predominant skin colour (PSC), intensity of predominant skin colour (IPSC), secondary skin colour (SSC), predominant flesh colour (PFC), secondary flesh colour (SFC), distribution of secondary flesh colour (DSFC), storage root formation (SRF), storage root shape (SRS), latex production in storage roots (LPSR), oxidation in storage roots (OSR), storage root surface defects (SRSD), storage root cortex thickness (SRCT).Measurements were done on three plants chosen randomly from the 11 plants per plot and averaged for the variable.

Data analysis
The computer program Genstat 14 th edition was used to analyse the morphological data.Stepwise discriminant analysis was performed to select a subset of variables that best discriminate among the classes.The Wilks' Lamda criterion was used to measure the variable contribution to the discriminatory power of the model as described by Daulfrey (1976); least contribution leads to removal of the variable.
The significant level of retaining or adding a discriminative variable was 0.15.Subsequently, principal component analysis was applied to examine the structure of the correlations between variables.The null hypothesis that any r ij was equal to zero was tested by computing the ratio of the explained variance to the unexplained variance.The eigenvalues and eigenvectors of the correlation matrix were derived, and the eigenvectors scaled by the square root of the corresponding eigenvalues to produce the matrix of component loadings.The eigenvalues and their associated eigenvectors, the correlation matrix are used to reduce the number of variables in the statistical analyses (Daulfrey, 1976).
A graphical display of the genetic relationships was also computed by principal coordinate analysis using the Rogers Tanimoto dissimilarity index of DARwin5.0.158 software.Cluster analyses were performed to group observations together using the method of Euclidian distance.Data points with the smaller distances between them were grouped together.A dendrogram was plotted from these computed clusters as a graphical relationship among accessions.From the dendrogram duplicates, samples were identified as a result of complete similarity between accessions.

Leaf sampling procedure
Leaf sampling was done as recommend by the DNALandmarks, a Canadian biotechnology laboratory, where the molecular work was done.Using 96-wells blocks, two leaf discs of 5 mm diameter were harvested from young leaves of each accession using a whole paper punch and put into a specific well position.The block was then placed inside a plastic bag with 50 g of silica gel and kept for 24 h to dry.

DNA extraction and SSR amplification
DNA extraction and amplification were done using an internal protocol at DNALandmarks laboratory in Canada.After extraction, the quality of the DNA was tested on 1% agarose gel.The DNA samples were then diluted to 25 ng/ul.The diluted DNA samples were then used for polymerase chain reaction (PCR) amplification with 30 SSR markers which sequences were provided by the International Potato Center (Table 2).PCR reactions were performed following an internal protocol of DNALandmarks with minor modifications (Ghislain et al., 2009).Forward primers were tailed with a M13 primer and the M13 primer (CACGACGTTGTAAAACGAC) labelled with one of the four fluorescence dyes (6FAM, PET, NED or VIC) for multiplexed PCR products detection using the ABI3730xl apparatus.The PCR conditions consisted of an initial denaturation at 95°C for 15 min, annealing at 60°C for 1 min and 72°C followed by 35 cycles of 94°C for 1 min, annealing at 60°C for 1 min, and 72°C for 1 min.This was followed by a final extension step of 20 min at 72°C and a halt at 4°C.The allele sizes were scored using GeneMapper software.Multiple peaks were detected due to the polyploidy nature of sweet potato.Any peak with the peak height greater than one sixth of the highest peak was scored.Allele size was calculated by subtracting 19 (M13 primer length) from the peak size.The raw data were provided for the further analysis.Failed samples were repeated one to two times.

Data analysis
The polymorphic information content (PIC) that is the importance of each SSR marker in distinguishing between accessions was determined (Weir, 1996) as: Where, P i is the frequency of the i th allele.
Each SSR fragment was treated as binary matrix in which band presence was coded as present or absent by 1 and 0, respectively.Based on the binary matrix, Jaccard'sdissimilarity index was computed as follows.A graphical display of the genetic relationships was also computed by principal coordinate analysis.Subsequently,

Discriminant analysis
Eight morphological traits with sufficient discriminative power to differentiate the accessions were identified based on their significant p-value forWilk's Lambda (P < 0.0001) and p-values for the average squared canonical correlations (P <0.0001) (Table 3).These were: PFC, LLN and LLT (Figure 3), MLS, VTP, SRSD, PP and SRF.The correlation matrices from Table 4 shows that these eight descriptors were not correlated with one another; this therefore indicates that using them will not create redundancy in the measurements.The F values revealed that the PFC and the LLN, respectively, with 299.81 and 36.90 had the greatest discriminating power associated with highly significant F values.Among the 22 variables discarded were the PSC commonly used by farmers to identify cultivars; the FH very important in breeding and other visible traits such as PT, MLC, ILC, GOL, and PVC.

Principal component analysis
Four principal components (PC) were identified which accounted for 67.22% of the total variation among the accessions (Table 5).The first PC accounted for 23.08% whereas the second, the third and the forth PC axes accounted respectively for 18.08, 13.32 and 12.73%.The first PC with reference to its high loadings (Table 6) was positively associated with traits such as leaf lobe number and predominant flesh colour.The second PC was associated with storage root characteristics (predominant flesh colour, storage root surface defects); the third with leaf characteristics (mature leaf size and petiole pigmentation) as well as with storage root formation, while the forth was associated with traits related to stems (leaf lobe type, petiole pigmentation and vine tip pubescence).

Cluster analysis
From the hierarchical cluster analysis, leaf lobe number, leaf lobe type, petiole pigmentation, vine tip pubescence, predominant flesh colour, storage root formation, storage surface defectand storage root surface defect showed a high polymorphism of 0.75 within the 112 sweet potato accessions (Figure 1).The accessions were grouped into eleven (11) clusters based on their average linkage and the Euclidean test.Clusters IV, VIII, IX and XI can be considered as outliers as they contained only one accession each, BF90, BF120, BF81 and BF137, respectively.Cluster I consisted of 37 accessions, cluster II had 11 accessions, cluster III had 10 accessions, cluster V of 6 accessions, cluster VI and VII had 21 accessions each, whereas cluster X had two accessions.Cluster II and cluster III were entirely constituted by orange fleshed accessions mostly with three leaf lobes, while the other clusters did not show any distinguishable relationship or pattern.The three East African OFSPs:Resisto (CIP 440001) belonged to cluster II while CIP-199062-1 and TIB-440060 belonged to cluster III.Cluster I was associated mostly with accessions with yellow flesh and a leaf lobe number of nine except for BF16 and BF42 which had 13 and 11 leaf lobes, respectively.Cluster V was constituted by accessions with white flesh and seven leaf lobes, cluster VI had individuals characterized by white flesh and one leaf lobe while cluster VII had white flesh with a very divergent number of leaf lobes ranging from one to   five with most of the accessions having five leaf lobes.The Rogers-Tanimoto pairwise dissimilarity coefficients computed as single and modality data using DARwin 5.0.158revealed a dissimilarity index ranging from 0 to 1 with an average value of 0.73 (Figure 2) suggesting a very high diversity among these 112 accessions.Most of accessions had dissimilarity indices ranging from 0.75 to 0.875 explaining 72.51% of the total frequency of dissimilarity with a maximum pair-wise dissimilarity of 1.

Identification of duplicates
From the hierarchical cluster analysis (Figure 4), duplicates were identified.Accessions BF1 and BF3 from two close villages in the central region were identical.
Accession BF13 from the central south was identical to accession BF62 from the Eastern region; two accessions BF78 and BF67 from the "Hauts-Bassins" region were identical as well as accessionsBF129 from the "Hauts-Bassins" and BF87 from the "Cascades".BF80 and BF68 from the Hauts-Bassins were also identical as were BF65 and BF63 from the same region.BF10 and BF18 from the central south and BF61 from the Eastern region were also identical.BF116 and BF114 from the Central west were morphologically identical as were BF52 and BF47 from the Central South.

Number of alleles detected
Among the 30 SSR markers, 27 were detected between one to six alleles while the remaining three markers detected between seven to eight alleles.The SSR marker IbO2 detected one to six alleles from 61 samples, seven alleles from 48 samples and eight alleles from three samples.The markers IBS139 and IBS199 detected one to six alleles each from 111 samples and seven alleles from one sample.The samples BF32 and BF99 showed between seven to eight alleles by the three SSR markers.

Polymorphic information content (PIC)
The thirty SSR markers revealed the usefulness of a marker in distinguishing between accessions with PIC values ranging from 0.00 for IbU20 to 0.881 for IbO2with an average of 0.727 (Table 7).Except for two SSR markers that had PIC values lower than 0.50 (IbU20 with 0 and IBS156 with 0.283), twenty eight (28) markers had high power of polymorphism (PIC>0.50).The high PIC values observed in this study indicated that the twenty eight SSR markers used were informative.

Genetic dissimilarity analyses and identification of duplicates
The frequency of pair-wise dissimilarity coefficients of the 112 sweet potato accessions based on the Jaccard's coefficient is shown in Figure 3.These SSR-basedpairwise dissimilarity coefficients ranged from 0 to 0.69 with a mean of 0.49 suggesting a relatively moderate diversity among the 112 sweet potato accessions.Most of the dissimilarity coefficients were between 0.52 and 0.69 explaining 82.35% of the total frequency.Nine accessions were identified with a pair-wise dissimilarity of 0 and therefore were considered as duplicates.This observation is confirmed by the dendrogram (Figure 4) generated using the unweighted pair group method (UPGMA).Thus, BF61 and BF94 with yellow flesh which were collected from"Cascades" and the Central-East region, respectively, were genetically identical; BF17 and BF18, two yellow fleshed accessions collected in two different communities in the Bazega province (Central-South region), were identical.BF38, BF19 and BF9, three orange fleshed accessions from the Bazega province, were also identical and different from the OFSP introduced from CIP-Eastern Africa.BF74 and BF68, with white flesh from the Kenedougou province, constituted a unique accession.After removing the duplicates, the initial number of 112 accessions was reduced to 107.These 107 sweet potato accessions will constitute a national core collection of sweet potato germplasm.

Comparison between morphological and SSR data
Using the morphological characters, the 112 accessions were grouped into 11 clusters with dissimilarity indices ranging from 0 1 with a mean of 0.73 suggesting a very high genetic diversity among the accessions.The use of the morphological data reduced the number of accessions from 112 to 101.Conversely, using the SSRbased analysis, 7 clusters were obtained.The dissimilarity indices ranged from 0 to 0.69 with a mean of 0.49, therefore, showing a relatively moderate diversity among the 112 accessions.The accession numbers were reduced from 112 to 107 using SSR markers.The accessions BF87 and BF88; BF63 and BF65; BF114 and BF116 identified as group of duplicates by morphological descriptors were closely related (nested on the dendrogram) using the SSR markers.Except for the groups of duplicates BF47 and BF52; BF67 and BF78 that were seen far away by the SSR markers, the other morphologically duplicates accessions belonged to the same molecular cluster.
In the other side, the duplicates identified using the SSR marker procedure BF17 and BF18 belonged to the same morphological cluster, as did BF94 and BF61.The duplicates BF74 and BF68 were seen morphologically far away, while BF9, BF19 and BF38 identified as the same accessions by the molecular procedure were found nested closely on the morphological dendrogram.The consensus between the morphological and the molecular based treeswas performed by using the strict rule consensus method consisting of simple counts of the frequency of occurrence of clusters in the set of trees (Perrier Perrier and Jacquemoud-Collet, 2006).It was observed that between the two trees, 4.7% of the clusters were in agreement.This weak consensus between the two trees suggested that there was no correlation between the morphological and the molecular data.
The Quartet tree distance estimate used as a measure of dissimilarity between the two trees was 0.95 demonstrating the absence of correlation between the two approaches used in the genetic diversity estimation.

DISCUSSION
The high diversity (mean of 0.73) detected within the 112 accessions regarding dissimilarity coefficient values suggests that the sweet potato accessions used in the current work would be a good source of selection for sweet potato breeding materials.Diversity studies have been done on sweet potato using morphological descriptors in various parts of the world and similarities or differences have been ascribed to sample size, number and type of descriptors used, the origin of accessions and the method of analysis.Using forty morphological descriptors in Uganda on 1256 accessions, 20 discriminatory descriptors were identified (Yada et al., 2010a).These 20 descriptors contained seven of the eight descriptors identified in this present study.Predo-minant skin color, commonly used in identification of cultivars in farmers' fields in Burkina Faso was not useful in differentiation among the accessions.Contrary to the results of this present work, Yada et al. (2010a) found this descriptor as discriminatory.In Kenya, Karuri et al. (2010) identified two descriptors (general outline of leaf, and, the shape of central leaf lobe) that differentiated among 89 accessions and separated them into two clusters.Karuri et al. (2010) found in agreement with the results of the current work, that flower habit was not significantly discriminative.High diversity index was also observed in a population of sweet potato in Kenya (Karuri et al., 2010), Uganda and India (Vimala and Hariprakash, 2011) using morphological traits.However, Tairo et al. (2008) observed lowdiversity of 0.52 among 280 sweet potato accessions in Tanzania.
Considering that SSR-based data are more accurate than the morphological data, the moderate diversity obtained in this study suggests that high priority should be given to further collect and/or introduce divergent materials, since variation in the collections is needed for a successful breeding program.Results from similar studies using SSR markers in sweet potato diversity analysis have been reported and most of the differences in results have been ascribed to sample size, the number of SSR markers used and the source of materials.Moderate genetic diversity values have been reported in Uganda (Yada et al., 2010b) among 192 accessions using 10 SSR markers; Gichuru et al. (2006) also reported low diversity in East African sweet potato cultivars while Soegianto et al., (2011) in Java reported similarity ranging from 15 to 78% between Indonesian accessions.Considering Eastern Africa as the second zone of diversity of sweet potato after the Central America (Villordon et al., 2007), one would expect a high diversity.The reason for the low diversity has been attributed to narrow geographic zone of collection of the cultivars.High SSR-based diversity has been noticed by Veasyet al. (2008) in Brazil, in Taiwan by Shih et al. (2002) and in China by Li et al. (2009) where the Jaccard's coefficient of similarity ranging from 0.400 to 0.938 was observed.
The weak agreement between the morphological based tree and the SSR based tree was also confirmed by different duplicates identified by each of these approaches.The findings of the present study are in agreement with those of Karuri et al. (2010) in Kenya who compared morphological and SSR-based evaluation of diversity.
A low correlation of -0.05 was observed between the two data sets.Further studies have reported low correlation between morphological and molecular markers in many crops (Koehler-Santos et al., 2003;Ferriol et al., 2004;Bushehri et al., 2005).The suggested reasons were that it could be a result of the independent nature of morphological and molecular variations.According to Vieira et al. (2007), this low correlation could also be due to the fact that a large portion of variation detected by molecular markers is non-adaptive as compared with phenotypic characters, which are influenced by the environment.The core collection obtained using the SSR markers' approach will be used for breeding purposes but the identified eight phenotypic characters will be used for the physical identification of the cultivars within the core collection.

Conclusion
Findings of the present study reveal that sweet potato germplasm in Burkina Faso presented moderate to high diversity based on molecular and phenotypic assessment approaches.The results obtained will serve as a guide for the basis germplasm management and improvement in the Burkina Faso and in the Sahelian zone of West Africa.However, further diversity is needed that can be achieved through introduction or more collection.The power of eight morphological descriptors and 28 SSR markers in the differentiation of cultivars was identified and could be useful in subsequent studies.Despite the poor correlation between morphological and molecular markers, both techniques can be use defectively in sweet potato characterization.The constitution of core collection will be done based on the SSR based data, but the eight phenotypic characters will be useful in distinguishing the cultivars in the field.

Figure 1 .
Figure 1.Dendrogram of the 112 sweet potato accessions revealed by average linkage cluster analysis based on the eight discriminant phenotypic characters.IX

Table 1 .
List of accessions collected in Burkina Fasoand the varieties introduced used for the characterisation.

Table 2 .
The 30 SSR primers used for the genotyping of the 112 sweet potatoaccessions.

Table 3 .
Selected morphological characters by The STEPDISC procedure.
Number of observation =109, Variables in the analysis=30, Class level=3, Significance level to enter=0.15,Significance level to stay=0.15.

Table 4 .
Correlation matrix for the 8 morphological traits used to distinguish the 112 sweet potato accessions.

Table 5 .
Eigenvalues of the correlation matrix.

Table 6 .
Eigenvectors from the eight principal component axes used to classified the 112 sweet potato accessions.

Table 7 .
Markers, number of alleles per locus, total number of alleles and PIC for 30 SSR.