Genetic diversity of sweet yam “ Dioscorea dumetorum “ ( Kunth ) Pax revealed by morphological traits in two agro-ecological zones of Cameroon

Genetic diversity of 43 Dioscorea dumetorum accessions and two of Dioscorea cayenensis Lam as outgroup species was evaluated for qualitative and quantitative traits in two different agro-climatic zones Baham (05°20.040' N/010°22.572’ E and 1634±3 masl, agro-ecological zone III with rainfall of 1500 to 2000 mm, and mean annual temperature of 19°C) and Ekona (04°12.773'N/009°19.425'E and 445±3 masl, agroecological zone IV with rainfall of 2500 to 4000 mm and mean annual temperature of 26°C). Forty one (41) morphological characters according to International Plant Genetic Resources Institute (IPGRI) descriptor, including nine quantitative and 32 qualitative characters were analyzed for multivariate analysis using cluster analysis, principal component analysis (PCA), and analysis of variance. For quantitative characters, cluster analysis revealed three major clusters. The first three components with eigen values > 1 contributed 83% of the variability. The PCA results indicate that traits which largely contributed to the variability within and between the accessions were stem length, leaf length, internode number, leaf width, harvest index, leaf number and internode length. Traits such as harvest index, internode number and stem length showed high heritability and high genetic advance. For qualitative traits, cluster and principal component analysis revealed two major clusters in Baham as well as in Ekona. The traits such as spines on stem base, spines on stem above base, spines length contributed to the variability within and between the accessions. However, the limits of morphological characters in the study of diversity have been detected.


INTRODUCTION
Yams constitute a staple food crop for over 100 million people in the humid and sub humid tropics (Mignouna et al., 2003).Yams (Dioscorea spp.) belong to the family of Dioscoreaceae and are widely distributed throughout the humid and sub humid tropics with about 600 described species (Coursey, 1976).Among the eight yam species *Corresponding author.E-mail address: mahbousomo@yahoo.fr.Tel: +237 99 55 0111, +237 78 34 28 84.
Author(s) agree that this article remains permanently open access under the terms of the Creative Commons Attribution License 4.0 International License commonly grown and consumed in West Africa, Dioscorea dumetorum is the most nutritious (Sefa-Deheh and Afoakwa, 2002).D. dumetorum, which belongs to the family Dioscoreaceae, originated in tropical Africa and occurs in both wild and cultivated forms but its cultivation is still restricted in West and Central Africa (Degras, 1993).Cultivation of D. dumetorum, reported especially in West Africa, is widespread in Western Cameroon (Treche and Delpeuch, 1979).Tubers of D. dumetorum are rich in protein (9.6%), fairly balanced in essential amino acids (chemical score of 0.94) and its starch is easily digestible (Mbome Lape and Trèche, 1994;Afoakwa and Sefa-Dedeh, 2001).The use of D. dumetorum as famine food is therefore recommended with caution.D. dumetorum is not only used for human consumption but also for pharmaceutical purposes.Thus, a novel bio-active compound dioscoretine in D. dumetorum has been detected (Iwu et al., 1990), which is acceptable pharmaceutically and which can be used advantageously as a hypoglycemic agent to reduce the blood glucose level in situations of acute stress such as experienced by patients with hyperthermia, trauma, sepsis and burns and those undergoing general anesthesia (Sonibare et al., 2010).In the practice of folk medicine in West Africa, D. dumetorum has been used by herbalists and traditional medical practitioners for the treatment of diabetes and as a topical anesthetic while others have used it as an arrow poison and as bait for monkeys (Corley et al., 1985).
Agronomically, D. dumetorum is high-yielding, with yields of 10 and 40 tones/hectare recorded under traditional farming conditions and in agricultural stations, respectively (Lyonga and Ayuk-takem, 1979;Ngong-Nasah et al., 1980;Treche, 1989).It is one of the only two species in which high level of resistance to two major nematode parasites of yam has been reported (Kwoseh, 2000).Unlike the other yams, staking is not necessary to maintain yield (Agbor-Egbe and Treche, 1995) and the tubers grow near the surface of the soil.This saves labour and allows for mechanization of harvest (Sefa-Dedeh and Afoakwa, 2002).Despite these qualities, in general little information is available on D. dumetorum diversity but in the case of Cameroon no information is available.Genetic diversity is the prerequisite for any plant breeding program.
Furthermore, the key to success for any genetic improvement programme lies in the availability and measure of genetic variability of desired traits.Morphological characterization is an important first step in description and classification of crop germplasm because a breeding program mainly depends upon the magnitude of genetic variability.However, morphological traits have been widely used in the study of yam germplasm and other yam species, specifically to determine the relationships between the various species and landraces, as well as to develop identification keys (Dansi et al., 1999;Mignouna et al., 2002;Norman et al., 2011;Tamiru et al., 2011).Such studies have not been conducted on D. dumetorum species.The aim of this study was to assess the genetic diversity of D. dumetorum in two different agro-ecological zones of Cameroon in order to show the influence of environmental conditions on morphological characters.

Plant materials
A total of 43 accessions of D. dumetorum with 2 accessions of Dioscorea cayenensis as out-group species are considered in this study (Table 1).42 are collected from various localities in the main yam growing regions (West, South-West and Nord-west) in Cameroon, 3 from Nigeria.Two accessions of D. cayenensis were chosen as out-group species.The accessions were planted in the same period (April 2013) in two experimental fields, one in the West region at ""Ferme Ecole Boukue"" (FEBO) in Baham and the other in the South-West region at ""Institue de Recherche Agricole pour le Developpement"" (IRAD) of Ekona.There was no control over environmental conditions.FEBO is located at latitude 05°20.040'N, longitude 010°22.572E and an elevation of 1634±3 m asl.FEBO belongs to agro-ecological zone III (Western highlands) characterized by rainfall of 1500 to 2000 mm, with mean annual temperature of 19°C (Anonymous, 2008) whereas, IRAD of Ekona is located at latitude 04°12.773'N,longitude 009°19.425'E,and elevation of 445±3 m and belongs to agro-ecological zone IV (Humid forest with monomodal rainfall) with rainfall of 2500 to 4000 mm and mean annual temperature of 26°C (Figure 1).

Field planting
The tubers were planted in the field at 1 m spacing in each row, and 1 m between rows.The tubers were planted in an arranged randomized complete block with a replicate plot of ten plants per accession.Each plant was supported by an independent stake of Phyllostachys to induce good canopy development (Figure 2).

Morphological characterization
The morphological description was done according to International Plant Genetic Resources Institute (IPGRI) and International Institute of Tropical Agriculture (IITA) descriptors for yam (IPGRI/IITA, 1997).These descriptors for yam have been previously used for characterizing cultivated yam (Dansi et al., 1999;Mignouna et al., 2002;Tamiru et al., 2011).

Data collection and analysis
Data was recorded on different types of qualitative and quantitative morphological characters with varying scales of measurements (Table 2).The characters were measured on all the plants in each replication, and mean values from the two replications were used for analysis.Data was analyzed for simple statistic (mean, standard deviation, and variance).Variance components (genotypic δ 2 g and phenotypic δ 2 p) were estimated using Wricke and Weber (1986) formula.Phenotypic (PCV) and genotypic (GCV) coefficient of variation were evaluated as suggested by Singh and Chaudhry (1985).Heritability (broad sense h 2 ) and genetic advance as percentage of mean were calculated as per Hanson et al. (1956) and Johnson et al. (1955), respectively.Cluster analysis was performed using single linkage with the Euclidean distance.The analyses were made using the computer program IBM/SPSS  analysis.Principal component analysis was used for variables reduction.Multivariate analyses are a valid system to deal with germplasm collections (Falcinelli et al., 2008).

RESULTS
The estimates of genotypic and phenotypic variance, broad sense heritability and genetic advance in percentages of means and genotypic and phenotypic coefficients of variation are given in Table 3. Variance analysis for characters revealed highly significant differences among genotypes for all the traits evaluated.Genetic advance as percentage of the mean was highest for harvest index (86.84%,P=0.05), internode number (59.03%, P=0.05), and stem length (53.45%,P=0.05) (Table 3).Broad sense heritability estimates was high for all the characters.The magnitude of phenotypic coefficient of variation (PCV) values for all the traits was higher than the corresponding genotypic coefficient of variation.quantitative traits (Table 4).The first principal component (PC1) was more related to stem length, leaf length, internode number and leaf width.The second principal component (PC2), explaining 18.49% of total variation was associated with harvest index, leaf number and internode length.The third principal component (PC3), explaining 14% of total variation was correlated with day to emergence and stem diameter.
Based on the average distance of k-means, quantitative traits accessions were grouped into three.Group I consisted of five accessions, group II of 10, and group III of 30 accessions.Cluster I was constituted by one accession from the North-West (30), three from South-West (1, 24, 26) and one accession of D. cayenensis (12).Cluster II, included three accessions from the West region (8, 15, and 20), two accessions from Nigeria (32, 34), four from South-West region (13, 37, 43, and 45) and one accession from out-group species (5).Cluster II included all the D. dumetorum accessions form West, North-West, South-West regions and Nigeria.Mean values along S.D. for each cluster (Table 5) showed that accessions in Cluster II were highest mean performance for most of the characters studied.Accessions of this cluster gave high yielding and medium number of days to emergence, whereas accessions in cluster I was high yielding but more days to emergence.The first two principal components were plotted to observe relationships between the clusters (Figure 3).Cluster I and III were clearly separated, whereas cluster II was not clearly separated.

Morphology diversity assessed by cluster for qualitative characters
Cluster analysis using single linkage of 32 morphological qualitative characters gave two major groups in Baham (Figure 4) and two major groups in Ekona (Figure 5).Cluster I in Baham was constituted of 34 accessions.It included all the accessions from West and North-West region with six accessions from South-West region (1, 13, 23, 24, 43 and 37) and three from Nigeria (32, 33 and 34).This cluster is characterized by low spines on and   13, 23, 36, 37 and 43).As in Baham, this cluster was characterized by low spines on and above stem base.Cluster II contained nine accessions from 22,24,26,38,40,41,42 and 45) and three from Nigeria (32, 33 and 34).They were characterized by many spines on and above stem base.It contained the only accession which formed flowers and fruits (32).

Morphology diversity assessed by principal component for qualitative characters
The result of principal component analysis based on 25 and 26 qualitative characters of 43 accessions of D. dumetorum in Baham and Ekona, respectively, are presented in Tables 6 and 7.In Baham six principal components are observed with Eigen-values greater than 1.0 contrary to Ekona with seven principal components.
In Baham, the first six principal components explained 71.39% of the variation, while the first principal component (PC1) alone explained 21.60% of the total variation.The PC1 had high loading for tendency of tuber branched, tuber shape, leaf density, number of tuber per seedbed, flowering.The second component (PC2) explaining about 16% of total variation was correlated with spines on stem base, spines on stem above base, spines length and plant vigor.The third component (PC3) explaining about 11% of total variation was mainly associated with characters such as roots on tuber surface, place of roots on the tuber surface, and cracks on the tuber surface.
In Ekona, the first seven principal components explained 77.34% of the variation, while the first principal component (PC1) alone explained about 22% of the total variation.The PC1 had high loading for spines on stem base, spines on stem abovebase, spines length and tuber length.The second component (PC2) explaining about 17% of total variation was correlated with roots on tuber surface, and place of roots on the tuber surface.The third component (PC3) explaining about 12% of total variation was mainly associated with characters such as tuber shape, petiole length and leaf density.
To assess the scores of individual accessions PC1 and PC2 were plotted (Figure 6).In Baham, the biggest set which represented the cluster I in cluster analysis occupied the low corner of the plot with relatively the highest positive and negative values for PC1.Contrary to cluster analysis, PCA analysis detected one sub-group within cluster I.This sub-group was constituted by nine accessions from West (8, 9, 10, 15, and 20), South-West (37 and 43) regions and Nigeria (32 and 33).It constituted the flowering accessions.The second set represented the cluster II which occupied the top corner of the plot with relatively high score for PC2.In Ekona, the biggest set which represented the cluster I in cluster analysis occupied the right corner of the plot with relatively the highest positive and negative values for PC2.
The second set represented the cluster II, and occupied the left corner of the plot with relatively high positive score for PC1.

DISCUSSION
The key to success of any genetic improvement programme lies in the availability of genetic variability of the desired traits (Heller, 1996).Morphological charac-terization is an important first step in description and classification of crop germplasm because a breeding program mainly depends upon the magnitude of genetic variability (Ghafoor et al., 2001).

Quantitative characters
For all the quantitative characters, the magnitude of     ) values for all the traits was near the corresponding genotypic coefficient of variation.This indicated a highly significant effect of genotype on phenotypic expression with little effect of the environment (Manju and Sreelathakumary, 2002).High GCV along with high heritability and high genetic advance will provide better information than single parameters alone (Saha et al., 1990;Baye et al., 2005).

I II
Hence, traits such as harvest index, internode number and stem length can be improved in the breeding program of D. dumetorum.However, all the traits of this study showed high heritability, and according to Manju and Sreelathakumary (2002), high heritability estimates indicate the presence of large number of fixable variable additive factors and hence these traits may be improved by selection.The PCA revealed that stem length, leaf length, internode number, leaf width, harvest index, leaf number and internode length were positive, indicating that these traits contributed at their peak towards diversity and can be used in the classification of D. dumetorum accession.

Qualitative character
Cluster analysis using single linkage of 32 morphological qualitative characters gave two major groups of D. dumetorum accessions in Ekona as well as in Baham with a few differences between the members of these groups, that is, accession 32 formed flowers and fruits in Ekona but has not flowered in Baham.This difference is probably due to environmental effects, in fact plants respond and adapt differently to pressure differences imposed by distinctive agro-climatic zones (Singh et al., 1998).There was no clear separation between accessions and area of collection.The same result has been reported by Sonibare et al. (2010), resulting fromthe natural migration of populations.Cluster analysis has also clearly separated D. dumetorum accessions from D. cayenensis accessions in the two areas of cultivation.This observation is explained by the fact that both species belong to two different sections; lasiophyton and Enantiophylum for D. dumetorum and D. cayenensis, respectively.The section lasiophyton is characterized by the fact that the vines twine to the right, that is, in a clockwise direction when viewed from the ground upwards, whereas species in section Enantiophylum twine to the left (Onwueme and Winston, 1994).
The PCA analysis is a way of identifying patterns in the data and expressing the data in such a way as to highlight their similarities and differences (Winterota et al., 2008).Traits such as the tendency of the tuber to branch, tuber shape, leaf density, number of tubers per seedbed, flowering, spines on stem base, spines on stem above base, spines length and plant vigor were positive for PC1 and PC2, thus contributing to the maximum diversity of D. dumetorum accessions, whereas in Ekona, we had traits such as spines on stem base, spines on stem above base, spines length and tuber length, roots on tuber surface and place of roots on the tuber surface.This difference is related to the fact that plants react differently to varying environmental conditions.That is why morphological and agronomic features cannot be relied upon due to the inconsistency in identification (Muthamia et al., 2013).
The results were consistent between cluster and PCA analysis, but the PC detected in Baham shows that the subgroup of D. dumetorum was constituted by the flowering accessions.Thus, grouping accessions by multivariate methods is practical value in the breeding programme, Ghafoor et al. (2001) have already remarked this in blackgram (Vigna mungo L.).

Conclusion
This study shows the relatively high genetic diversity of D. dumetorum accessions.This diversity could effectively be used in the breeding programme of D. dumetorum accessions.Although, morphological traits allowed the

Figure 1 .
Figure 1.Map of the sample collection areas including the two experimental sites (Baham and Ekona) and some locations where accessions have been collected.

Figure 3 .
Figure 3. Scattered diagram of 43 accessions of D. dumetorum and 2 accessions of D. cayenensis for two PCs.The digits 1, 2 and 3 represent the cluster number.

Figure 4 .
Figure 4. Relationship of 43 accessions of D. dumetorum and 2 accessions of D. cayenensis based on single linkage clustering of 32 qualitative characters in Baham.

Figure 5 .
Figure 5. Relationship of 43 accessions of D. dumetorum and 2 accessions of D. cayenensis based on single linkage clustering of 32 qualitative characters in Ekona.

Figure 6 .
Figure 6.Plot of the first (PC1) and the second (PC2) principal components for 43 accessions of D. dumetorum based on qualitative characters.A: Baham; B: Ekona.The numbers represent the accessions.
D. dumetorum in different clusters, their expression is subjected to agro-climatic variations and thus provide limited genetic information.This is why, molecular markers are recommended with caution to reveal the full extent of existing biodiversity of D. dumetorum in Cameroon.

Table 1 .
Accessions of D. dumetorum used in this study.

Accessions code Local name Area of collection Altitude (±3 masl.)
Bazou The highest values of genetic advance as percentage of mean were for harvest index (86.84%),internode number (59.03%) and stem length (53.45%).The first three components with eigen values > 1 contributed 83.26% of the variability, while the first component (PC1) alone accounted for 60% of the total of the variation amongst 45 accessions evaluated for nine Baham

Table 2 .
Morphological descriptors used for characterization.
*State not specified in IPGRI/IITA, ** Characters used for clustering and principal component analysis.

Table 5 .
Mean and S.D. for three clusters on agronomic traits.
above stem base.Cluster II contained nine accessions(21, 22, 26, 36, 38, 40, 41, 42 and 45).All the accessions of this cluster were from South-West region.They were characterized by many spines on and above stem base, except accession 36 with low spines on stem base.In Ekona, Cluster I was constituted by 31 accessions.It included all the accessions from West and North-West regions with six accessions from South-West region (1,

Table 7 .
Principal components (PCs) for qualitative characters in 43 accessions of D. dumetorum in Ekona.