Morphological characterization of shea tree ( Vitellaria paradoxa subsp . paradoxa ) populations in the region of Mandoul in Chad

1 Faculté des Sciences Exactes et Appliquées. Département de Biologie, Université de N'Djaména, B.P 1027, Tchad. 2 Laboratoire de Biotechnologies Végétales, Département de Biologie Végétale, Faculté des Sciences et Techniques, Université Cheikh Anta Diop, BP 5005 Dakar-Fann, Dakar, Sénégal. 3 Laboratoire de Botanique et Biodiversité, Département de Biologie Végétale, Faculté des Sciences et Techniques, Université Cheikh Anta Diop, BP 5005 Dakar-Fann, Dakar, Sénégal.


INTRODUCTION
The semi-domesticated shea butter tree (Vitellaria paradoxa (C.F.Gaertner) syn.Butyrospermum parkii (Kotschy), Butyrospermum paradoxum (C.F.Gaertner) Hepper, Family Sapotaceae) is wildly distributed in the Sudano-Sahalian region from Senegal to Uganda (Hall et al., 1996;Hemsley, 1968;Salle et al., 1991).Presently two subspecies have been identified.V. paradoxa subsp.paradoxa is found in West and Central Africa (Hall et al., 1996;Salle et al., 1991;Sanou et al., 2005;Fontaine et al., 2004;Allal et al., 2008;Nyarko et al., 2012;Kelly et al., 2004), while V. paradoxa subsp.nilotica is common in East Africa (Soudan, Ethiopia, Uganda and Republic Demotratic of Congo) (Gwali et al., 2012;Okullo et al., 2004;Byakagaba et al., 2011;Okiror et al., 2012).The tree shape is influenced by various environmental factors and they are well identified by farmers according to the folk classification.During the wet season, the tree produces fruits edible by both human and animals.The fruits contain 1 to 3 large solitary seeds, rich in fat and oil used in a variety of purposes such as cooking (Abbiw, 1990), *Corresponding author.E-mail: diaga.diouf@ucad.edu.sn.medicinal, hair and skin ointments and as a base for industrial manufacture of confectioneries (Cidell and Alberts, 2006).The oil is also used in traditional and social rituals such as marriages, funerals, coronations and rainmaking (Ferris et al., 2004;Gwali et al., 2012;Hall et al., 1996;Moore, 2008).The wood of the shea butter tree is used for charcoal, furniture and construction, and the latex for glue making (Lovett and Haq, 2000a).In addition, the trees are used in agroforestry systems that play an important role in the adaptation to climate change such as contribution to soil fertility (Rao et al., 2007).For these reasons, shea butter tree generates significant incomes for households.
Because of its economic importance, genotype studies were performed based on morphological characters leading to the identification of several phenotypes including the domesticated V. paradoxa (Chevalier, 1943(Chevalier, , 1948;;Nafan et al., 2007;Ruyssen, 1957;Sanou et al., 2005Sanou et al., , 2006;;Ugese et al., 2010).In 1943, Chevalier identified eight varieties based on fruit and leaf variation (cuneata, ferruginea, floccosa, mangifolia, nilotica, parvifolia, poissoni and serotina).In 1957, the taxonomy was revisited by Ruyssen using tree shapes and sizes, fruits, nuts and leaves leading to the description of V. mangifolium as a subspecies containing two varieties (viridis and rubifolia).Further, using fruit morphology, nut color, crown shape and habitat types, phenotype variation was noticed for the shea tree in Cameroon (Nafan et al., 2007;Lamien et al., 2007).This variation was in agreement with the folk classification distinguishing ethno-varieties which was used by West African farmers to select and preserve shea tree (Gwali et al., 2011;Lovett and Haq, 2000b).Gwali et al. (2012) used morphological characters of 176 trees representing 44 ethno-varieties in Uganda to establish the congruence between the morphological variation and folk classification.Their results showed a good congruence with folk classification when they combined the qualitative traits as perceived by farmers.Recently, Mbaiguinam et al. (2007) performed studies in the population of shea tree from Mandoul region using chemical characteristics and concluded that there was no significant difference of fatty acids content within varieties.In addition, they reported that the shea butter profile was between those from Cameroon and Uganda.Nevertheless, substantial darkness points have to be addressed particularly in Chad where the level of the morphological diversity of the tree is still understood.
The objectives of this study were to discriminate the morphotypes of the shea trees growing in the region of Mandoul located in Southern Chad using leaf, fruit and nut characteristics.

Study area and experimental design
The present study was conducted in three sites (Matekaga, Djekota et al. 185 Kemkian and Kol) located in the region of Mandoul in southern Chad, where the mean annual precipitation was 1,200 mm (Figure 1a, b).The rainy season lasts from May to October with mean temperatures of 22°C.The dry season lasts from November to April with average temperature of 32°C.In the area of our experiment, the soil was sandy, lateritic or ferralitic.Each site was swept by 4 or 8 transects crossing in its center using a GPS (Garmin, city and country).Along each transect, the quadrants were separated by 100 m.These sites were chosen because of the high density of their shea tree.It should be noted that Kemkian means the village of shea tree in the local language.

Plant material
Two hundred and forty (240) shea trees distributed in six ethnovarieties according to folk classification were investigated.The folk classification of the accessions was consistent because in the area where the materials were collected, people spoke the same language.The sampling method consisted of selecting 104 trees in Matekaga, 64 in Kemkian and 72 in Kol.

Data collection
Ten adult fresh leafy twigs and mature fruits without parasites were collected randomly from each tree.The length and the width of the leaves were measured using a vernier caliper (Shanghai, China).The total length of the leaf consisted of the length of petiole and that of lamina.The length and the width of fruit and the diameter of the nuts collected from fruits were measured for each accession (Figure 2).For accurate measurement, a mean value was calculated from ten organs.In addition, the mean value of length and width of each organ within site were calculated.The mature fruits were collected in May and September 2007.These data were used as raw material, subjected to principal component analysis (PCA) and analysis of molecular variance (AMOVA).

Multivariate analysis
PCA was performed using statistical package ADE-4 coupled with a hierarchical cluster ascendant (HCA) to group the accessions based on their similarities.Leaves, fruits and nuts were considered as variable but the 240 accessions were projected in a plane including the two first axes.To perform an ascending hierarchical clustering of the accessions, the coordinates of the individuals on the factorial axes as similarity matrix, the squared Euclidean distance and the Ward's method were used.The dendrogram was generated using the R (version R-2.9.0, ADE4 package) software (R Development Core Team, 2011).

Morphological character analyses
The mean values of leaf, fruit and nut parameters of the samples collected from each site allowed to estimate their variation.The parameters measured from the leaves showed that the lamina length ranged from 15.8 to 23.6 cm for the trees growing in Kemkian, 15.5 to 24.2 cm in Kol, 16.5 to 26.3 cm in Matekaga.The percentage of leaves for which the length was more than 20 cm was 81% in Kemkian, 76% in Kol and 69% in Matekaga.The width of the lamina ranged from 3.2 to 5 cm in Matekaga,   3 to 5 cm in Kemkian, 3 to 5.4 cm in Kol.The length of the petiole ranged from 5.8 to 11.9 cm for the leaves collected in Matekaga, from 5.7 to 11.2 cm in Kemkian and 5.7 to 10.2 cm in Kol.
Fruit peduncle length variability was also reported in this study.It ranged from 1 to 3.1 cm in Kol, 1 to 3.3 cm in Kemkian and 1 to 2.7 cm in Matekaga.The fruit length varied from 2.5 to 5.5 cm in Matekaga, 2.6 to 5 cm in Kemkian and from 2.6 to 5.5 cm in Kol.Assessing the fruit width, our study founded that it varied from 2.3 to 4.3 cm in Matekaga, 2.4 to 4.4 cm in Kemkian and 2.4 to 4.3 cm in Kol.The nut length varied from 1.9 to 3.6 cm in Matekaga, 1.9 to 3.3 cm in Kemkian and 1.9 to 3.8 cm in Kol.Finally, the measure performed on the nut width showed that their values ranged from 1.5 to 2.8 cm in Matekaga, 1.4 to 2.4 cm in Kemkian and 1.5 to 2.6 cm in Kol.

Statistical analysis of morphological data
PCA showed that the two principal axes explained 72.95% of the variance observed.The first axis expressed 46.54% of the total variance (data not shown).The variables, nut length, fruit width, nut width and fruit length, contributed to 86.96, 83.59, 78.53 and 77.97%, respectively.The second axis expressed 26.41% of the total variance.The lamina length and petiole length represented 97.4 and 82.87% of the variance, respectively.The third axis explained 10.66% of the total variance where lamina width explained 65.71% of this value.Finally, the fourth axis expressed 8.11% of the variance where the peduncle length explained 54.84% of this value and is associated with the nut width which contributed to 5.34% of the variance.
The correlation matrix showed that lamina length and petiole length were correlated with 92%.Fruit width was correlated with peduncle length and fruit length with 52 and 83%, respectively.Nut length was correlated with peduncle length, fruit length and fruit width with 55, 78 and 77%, respectively.In contrast, nut width was correlated Djekota et al. 187 with fruit length, fruit width and nut length with 68, 78 and 87%, respectively (Table 1).In addition, there was significant variation of the mean of the peduncle length between the samples collected in Matekaga, Kemkian and Kol.The mean of the fruit length was similar between Kol and Matekaga but it was significantly different with the ones found in Kol (Figure 3A). Figure 3B showed that nut characteristics (means of length and width) were not different between Matekaga and Kol.These characterristics were significantly different with the ones collected on the nuts from Kol.No significant difference was observed between the mean of the peduncle length in Matekaga, Kemkian and Kol.Similar results were observed for the means of the lamina length and lamina width taken individually (Figure 3C).

Morphological variation within sites
Dendrograms performed using leaf, fruit or nut parameters separately or in pair combinations failed to discriminate accurately the samples collected from each site (data not shown).In contrast, a combination of leaf, fruit and nut parameters allowed a good resolution between individual samples within each site.In Figure 4A, samples collected in Kemkian were divided into four main groups.In the group I, two sub-groups were observed.
The first sub-group encompassed A121 and A141 which were clustering together while A110 and A162 were sister of A163.In the same sub-group, A120 and A142 were clustering together as A109 and A130 did.The genetic relationship among the individuals forming the second sub-group is also well resolved.The group II was also sub-divided into two sub-groups which are well resolved.
In group III, the clustering was very clear except for A139 and A160 which erre linked with a short branch.The group IV was sub-divided into several numbers of subgroups.
The material collected in Kol was divided into four groups (Figure 4B).The first group showed a high coefficient of similarity among accessions and two main sub-groups as the second group did.The third group encompassed two main sub-groups including several subdivisions each, while the fourth group was also well structured.
In Matekaga, based on the dendrogram, the biological material was divided into four groups (Figure 4C).The first group was divided in two main sub-groups which were well resolved.The second group encompassed several sub-groups as the third but the fourth was more diversified.

Trait variation between sites
The dendrogram in Figure 5

DISCUSSION
Understanding population genetic structure is relevant to phytogenetic resources management because it is the first step before implementing any selection process.Phytogenetic resources management was applied to a wide range of economically important plants including shea tree.In its area of distribution particularly in West Africa, shea tree resource management has been mainly based on folk classification for centuries aiming at conservation, domestication and selection of superior individuals (Lovett and Haq, 2000b).On the other hand, in Chad, few studies were reported aiming to enhance our understanding of shea tree genetic variation (Mbaiguinam et al., 2007).

Morphological variations of the shea tree
In this study, a variation in lamina length was observed within and between sites.The smallest lamina length (15.5 cm) was found in Kol, while the longest (26.3 cm) in Matekaga.In addition, the biggest lamina width was found in Kol with 5.4 cm.This morphological variation suggested that a single morphotype was not growing in these areas.Variation of the length of petiole was also observed within site and between sites.The longest petiole was reported in the population from Matekaga while the smallest was observed in Kol.These findings are agreement with the results of Nyarko et al. (2012) who reported petiole length variation in shea tree from Ghana.
In the same manner, as the variation of the peduncle length within sites and between sites, the longest peduncle length (3.3 cm) was observed in Kemkian while the longest fruit was found in Matekaga and Kol.The morphological parameters collected from fruits and nuts showed variation within sites but significant differences were not observed between sites.The variations of the parameters from Chad observed in this study were close to those estimated from shea trees in Mali, Ghana, Guinea, Sudano-Sahelain and Uganda.These findings suggested the same amplitude of morphological variation between Central Africa, East and West accessions (Gwali et al., 2012;Sanou et al., 2006;Nyarko et al., 2012).Variation of fruit morphological characters has been reported for the tropical species such as Balanites aegyptiaca and Tamarindus indica (Soloviev et al., 2004).These variations can be explained by natural and/or human selection, gene flow mediated from genetic drift (Irwin, 2000;Tremblay et al., 2010;Darwin, 1869;Vaughan et al., 2007;Abasse et al., 2011).In addition, rainfall regimes and soil characteristics might be involved in the morphological variations as it was reported in West African provenances (Sanou et al., 2006).

Statistical analysis
Statistical analysis using morphological characters showed high variability of V. paradoxa subsp.paradoxa growing in the region of Mandoul located in the South of Chad.Individually, these morphological characters were allowed classifying the samples in different morphotypes.Similar results are reported by several authors (Lovett and Haq, 2000a;Nafan et al., 2007).In this study, significant correlations between lamina length and petiole length or between fruit and nut characteristics were observed and it is in agreement with results obtained in provenances from Mali, Côte d'Ivoire and Ghana (Sanou et al., 2006;Lovett and Haq, 2000a;Nafan et al., 2007).Therefore, four main characters as fruit length, fruit width, nut length and nut width were useful to discriminate morphotypes.This assertion confirms the work of Chevalier (1943) who used morphological characters (leave and fruit) to identify eight varieties (cuneata, ferruginea, floccosa, Mangifolia, nilotica, parvifolia, poissonietserotina) within V. paradoxa subsp.paradoxa.

Genetic relationship between shea trees
HCA showed that the use of a single morphological character was not efficient to differentiate the accessions but combining leaf, fruit and nut parameters allowed a good resolution.In this study, each site showed four groups as the dendrogram including all the sites did.These findings were not congruent with the folk classification which identified 6 varieties in the same sites as that of the present study (Mbaiguinam et al., 2007).This incongruence might result from allogamous nature of shea tree which induces phenotypic variation.Phenotypic variation can be influenced by environmental factors or result from genetic variation (Tremblay et al., 2010).Therefore, it is difficult to identify shea tree based on morphological characters alone.The grouping of A11/A30 both from Matekaga and A98 (Matekaga)/A185 (Kol) for example suggested a hybridization by insects pollination or diverse forms of gene flow within or between sites.Hybrids can be problematic for butter quality production because previous studies showed that the morphotypes growing in this area do not produce the same amount of chemical compound (Mbaiguinam et al., 2007).On the other hand, hybridization can be beneficial because high genetic variation induces variability in the population.

Conclusion
Using morphological characters, our study pointed out a high variation of V. paradoxa subsp.paradoxa populations within and between sites in the region of Mandoul in Southern Chad.A high resolution of the variation was obtained when several morphological characters were combined but a lack of congruence with the folk classification was noticed.The present study can be extended to others Chadian regions were the shea trees are endemic for comparing the local knowledge and for better identification of the morphotypes growing in Chad.Molecular approach can also be used to test if the morphological variation resulted from the DNA polymorphism.
Figure 1.Map of Chad and region of Mandoul showing the localization of the sites where the samples were collected (Ministery of Interior, 2009).
contained the 240 individuals growing in the three sites (Kemkian, Kol and A: Peduncle and fruit characters variation between sites.B: Nut characters variation between sites.C: Leave characters variation between sites.

Figure 3 .Figure 4 .
Figure 3. Variation of morphological characters of shea butter tree between sites.The characters affacted by the same letter are not statistically differents.