Phenotype characterization and diversity assessment of mango (Mangifera indica L.) cultivars in Ethiopia

Little efforts have been made on mango genetic resource assessment in Ethiopia though it is one of the major fruit crops. This study was conducted to assess the diversity of 69 mango cultivars of different growing regions of the country based on 44 phenotypic descriptors. The results of both univariate and multivariate analysis of variance computed for quantitative data, and results from descriptive statistics for qualitative characters indicated the presence of phenotypic variation among the cultivars. Further analysis of Principal Component Analysis (PCA) indicated the first four components explained more than 75% of the total variation in which most fruit, seed and leaf characters contributed much to the observed variation. The cultivars were grouped into 13 clusters by Unweighted Pair Group Method with Arithmetic Means clustering method from the Euclidean distances estimated from phenotypic characters. The three clusters (II, X, and XIII) constructed each by one cultivar while others encompass more than one irrespective of their geographic regions. This indicated the presence of diversity among cultivars in Ethiopia which can be exploited for further improvement, use, and conservation of mango genetic resources. 
 
 Key words: Cluster, Euclidean distances, genetic resources, principal component.


INTRODUCTION
Mango (Mangifera indica L.) is amongst the most widely grown tropical and subtropical fruits of the world (Rajwana et al., 2011).The origin of cultivated mango is believed to be eastern India, Assam-Burma region; and South East Asia is believed to be the center of diversity for Mangifera genus (Begum et al., 2014;Kaur et al., 2014).More than 1000 varieties of M. indica L. have been identified all over the world (Rymbai et al., 2014).It is thought to have been introduced to East Africa by the Persians in the 10th century A.D. and the crop started growing in West Africa in the 16th century A.D. (Janick, 2005;Rey et al., 2006).
World mango production is spread over 100 countries that produce over 38.67 million tons of fruit annually (Mitra, 2016).India is the largest producer in the world (18.0 million tons per year), while the leading producer in Africa is Kenya (582,907 ton per year) (FAO, 2015).Mango is the second most important fruit crop in Ethiopia, *Corresponding author.E-mail: tedrosneguse@yahoo.co.uk.
Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License after banana.It constituted 16.01% of 92,362.36ha of land under fruit crops and 14.76% of 6,797,42.83tons of produced fruit (CSA, 2015).Moreover, both its area of coverage and production increased by 208.4 and 247%, respectively from 2003208.4 and 247%, respectively from to 2013208.4 and 247%, respectively from (Dessaleg et al., 2014)).Western and eastern Ethiopia are among the major growing regions that account for 28 and 23% of the wholesale market share of Addis Ababa (capital city of Ethiopia), respectively (Ssemwanga et al., 2008).
Despite the crop"s potential and increasing production trend in Ethiopia, it is hampered by various biotic and abiotic factors (Dessalegn et al., 2014).The majority of farmer cultivars are raised from seedlings arising from a natural cross population and consequently, the trees are of mixed origin and difficult to identify (Bezu et al., 2014).Moreover, growers are replacing the existing locally adapted landraces with recently introduced commercial varieties.Although, the landraces could be having desirable traits such as high and stable yield, low management requirements, low susceptibility to pests and high drought tolerance (Govindaraj et al., 2015;Sennhenn et al., 2013).Very little information has been documented on Ethiopian local mango varieties.Such information is important and could be utilized in the conservation of the valuable genetic resource.There is also much confusion and uncertainty concerning the identity of local mango cultivars due to the usage of different local names for the same varieties.
Characterization of varieties is a necessary requirement for crop improvement, use, and conservation of plant genetic resources (Khan et al., 2015;Krishna and Singh, 2007;Rajwana et al., 2011).Phenotype characterization is the first step before biochemical and molecular markers due to its simplicity, low-cost requirement, standardized, repeatable method and availability of published descriptors for most major crops (Khan et al., 2015;Mohamed and Ahmed, 2015;Ravani and Joshi, 2013).In the last decade, various phenotype markers have been successfully applied in determining the intra cultivar diversity of mango in different parts of the world (Ahmed and Mohamed, 2015;Mohamed and Ahmed, 2015;Preisigke et al., 2013;Rajwana et al., 2011;Rymbai et al., 2014).However, such vital studies have not been conducted on mango in Ethiopia that is on the verge of extinction.Therefore, the objective of this research was to assess the phenotype variation and thereby to estimate the diversity of cultivars across major growing regions of Ethiopia.

Description of experimental sites
Four major mango growing districts were selected: Babile (eastern Hararghe Zone, Oromia Regional State); Erer-Woldia and Sofi (Harari People's National Regional State) from east and Asosa (Benishangul Gumuz Regional State) from western Ethiopia.In addition, mango collections conserved at Melkasa Agricultural Research Center located in central Ethiopia were also included in the study (Figure 1).

Sampling of experimental materials
Survey of mango cultivars was carried out on 113 farmers purposively selected from the four growing districts and Mango orchards of Melkassa Agricultural Research center in 2016.The study sites were selected based on their extensive mango production, as was inferred through consultation with the districts agricultural officers from the Ministry of Agriculture and key informants from the Ethiopian Agricultural Research Institute.Of the germplasm surveyed, 69 cultivars (53 from farmer"s field and 16 from Melkassa Agricultural research center) of unknown genetic origin were selected considering their local naming, geographic locations, accessibility, age and distinct features of the trees.The geographic location of each of the sampled trees was recorded using a global positioning system (GPS) along with location information and local names (Table 1).

Phenotype characterization
A total of 44 characters, 15 quantitative and 29 qualitative traits were evaluated according to IPGRI (2006).The qualitative data were collected on the farm while the quantitative data was recorded from randomly collected healthy and undamaged ten leaves and 20 ripe fruits of each sampled tree in Horticulture Laboratory of the Haramaya University, Ethiopia.The collected sample leaves and fruits of each tree were randomly assigned into three replications to conduct an analysis of variance for completely randomized design for a better estimate of error variances (Gomez and Gomez, 1984).

Data analysis
Descriptive statistics (mean, percentage, standard deviation) and Chi-square test of qualitative characters were performed using the Statistical Package for Social Scientists (SPSS) Version 21.0.Variations among the cultivars for all quantitative traits except trunk circumference and crown diameter that were recorded from the single tree were computed using the analysis of variance (ANOVA) for completely randomized design.Moreover, multivariate analysis of variance (MANOVA) was done to test if the combined dependent variables (quantitative traits) were significantly affected by cultivars using SAS software version 9.1.Standardization of data was conducted following the z-score transformation method (Ramette, 2007).Principal Component Analysis (PCA) was done to identify traits that explain the phenotype variability best.Then, clustering of cultivars following the Unweighted Pair-Group Method with Arithmetic Mean (UPGMA) (Sneath and Sokal, 1973), where the distance between two clusters is the average distance between all inter-cluster pairs, was made using GENES version 2015.05 program (Cruz, 2013).

Variation of mango cultivars in quantitative traits
The mango cultivars had a wide range of circumference of tree ranging from 59 to 370 cm with a coefficient of variation (CV) of 44.7%.The crown diameter of the tree also ranged from 3.5 to 20 m with a CV of 39.7%.The univariate analysis of variance computed for 13 quantitative traits revealed a highly significant difference (p < 0.01) among cultivars (Table 2).In addition, the multivariate analysis of variance as predicted, based on the Wilk"s lambda criterion, showed the combined quantitative traits were also significantly (p < 0.001) affected by the cultivars (Table 3).This suggested the cultivars varied for all the quantitative traits that could be used for further breeding work.Similarly, authors from India (Singh et al., 2012;Bajpai et al., 2016) and Kenya (Toili et al., 2016;Gitahi et al., 2016) also reported the presence of significant differences among mango cultivars they studied considering similar quantitative traits.

Tree and leaf characters
The cultivars were largely non-grafted seedlings, irregular (alternate) bearing behavior, tree height range from medium to very tall group, broadly pyramidal to semi-circular crown shape, and spreading growth habit (Table 4).The alternate bearing which is dependent on agronomic practices (Saxena et al., 2014), environmental conditions and genetic makeup (Kaur et al., 2014), is a common phenomenon of mango.Most cultivated mango trees are between 3 and 10 m in height when fully matured depending on the way of pruning (Balley, 2006).However, mango trees can reach a height of 40 m or more (Mukherjee and Litz, 2009) while grafted ones are usually shorter (Khan et al., 2015).Tree canopies vary in genotypes, propagation method, the density of plantation, and prevailing agro-climatic conditions (Khan et al., 2015).Intermediate foliage density, oblong leaf blade shape, semi-erect to horizontal leaf attitude in relation to branch, a medium category in the angle of secondary veins to the midrib, acuminate apex, acute base shape, and mild leaf fragrance were observed in the majority of the cultivars (Table 4).These characters are among the important attributes that could be utilized for classification of the cultivars (Sharma et al., 2016)  The number of degrees of freedom in the model.b The number of degrees of freedom associated with the model errors.
characterization of the cultivars they studied.

Fruit, stone and seed characters
The predominant fruit shape of the cultivars was oblong followed by roundish, obtuse fruit apex, absent fruit stalk cavity, absent to slightly neck prominence and perceptible beak type.The majority of the cultivars had orange, greenish yellow to yellow skin color and orange to yellow pulp color when ripe.The cultivars fruit attractiveness was from average to good though there were excellent attractive cultivars (26.1%).Most had low to intermediate fiber in fruit pulp, very juicy, intermediate aroma, and very good to excellent eating quality (Table 5).This indicated the potential of cultivars for the table as well as processing purpose if further studied (Jha et al., 2010;Vijayanand et al., 2015).Most cultivars in Shendi, Sudan also reported oblong fruit shape followed by round and obtuse fruit apex (Ahmed and Mohamed, 2015).A study by Kheshin et al. (2016) on some "Sukkary" mango genotypes in Egypt also revealed a roundish fruit shape, smooth and waxy yellow skin, and obtuse shape of fruit apex.The predominant fruit shape of mangos in eastern Kenya was roundish and yellow-orange, orange-red and red colors of the fruit skin (Gitahi et al., 2016).Elevated vein level with the surface of stone (85.5%) and parallel stone venation (85.5%) was recorded from the majority (89.9%) of cultivars.Moreover, 88.4and 62.3% of cultivars seed was reniform shape and monoembryony, respectively (Table 5).The difference in the cultivars embryony type is the most important trait that affects propagation methods (Kuhn et al., 2017).It could be associated with their origin where, most Indian cultivars are mono-embryonic, while generally cultivars from Indonesia, Thailand and the Philippines are reported polyembryonic (Damodaran et al., 2012;Griesbach, 2003).

Principal component analysis
Principal component analysis (PCA) for quantitative traits   showed the first four components with Eigenvalues greater than one explained 75.89% of the total variation (Table 6).The first principal component (PC-1) accounted for 31.87% of the total variation, included fruit length, fruit diameter, stone length, stone width, seed length, and seed width.The second component (PC-2) explained 19.65% of the total variation and was associated with fruit weight, pulp content, stone, and seed weight.The third component (PC-3) that explained 13.80% of the total variation was mainly associated with leaf length, leaf width, and petiole length; and the fourth component (PC-4) accounted for 10.57% of the total variation correlated with trunk circumference and crown diameter.Moreover, the distribution of the cultivars based on the first two components (Figure 2) also showed the phenotypic variation among the cultivars and how widely dispersed they are along the axis.The aforementioned characters which contributed most to the observed variations were also reported by Krishnapillai and Wijeratnam (2016) and Majumder et al. (2013).Hence, it indicated to give greater emphasis on those traits that had a significant contribution to the observed variation for the future breeding program.

Genetic distances of cultivars
The genetic distance of cultivars estimated by Euclidean distance (ED) varied from 2.3 to 9.7 with a mean and a standard deviation (SD) of 5.9 and 1.1, respectively (Table 7).Majority of pairs of cultivars (53.1%) had Euclidean distances less than the overall mean Euclidean distance.Whereas 42.6 and 4.2% of pairs of mango cultivars had Euclidean distances of 5.9 to 7.0 and >7.0 (mean +SD), respectively (Figure 3).The result indicated that considerable mango cultivars of different geographic regions were genetically diverse.The higher the ED of a pair of cultivars indicated the differences of genotypes with more number of genes (alleles) while the lower ED of a pair of cultivars suggested the differences of genotypes with few genes (alleles) (Bhandari et al., 2017).
The mean Euclidean distance result revealed, the most distant cultivar was HA14 (7.4) followed by AS03 (7.1), HA10 (7.0), BA12 (7.0), and ER16 (7.0).The closest cultivars to others were ML13 (5.9), ML03 (5.9), HA01 (5.9), and AS05 (5.9) (Table 7).The growing regions of the most distant cultivars were from eastern (Harewe, Babile, and Erer Districts) and western (Assosa district) Ethiopia.Whereas, the closest cultivars were from central (Melkassa Research center), eastern (Harewe District), and western (Asosa District) Ethiopia.This suggested the existence of genetic diversity among the cultivars based on their geographic regions.Though, there were cultivars that closely related irrespective of their geographic regions.It is evident that the geographic distance has a contribution to the genetic distances of genotypes (Rao and Hodgkin, 2002).However, the influence of geographical distance on genetic divergence could be suppressed by another factor (s) like genetic drift and natural selection, that could result in the difference in genetic diversity of genotypes of the same location or vice versa (Bhandari et al., 2017;Majumder et al., 2013).

Clustering of mango cultivars
The 69 mango cultivars were grouped into 13 clusters (Table 8 and Figure 4), with mean Euclidean distances of 5.9 with 1.14 and 19.38% standard deviation and coefficient of variation, respectively.Three clusters (II, X, and XIII) constructed each by one cultivar, while others encompass more than one cultivars.Clusters X and XIII had significantly higher mean Euclidean distances than other clusters.The six clusters (I, II, IV, VII, VIII and XI) consisted of 29 cultivars had mean Euclidean distances greater than the overall mean distance of cultivars.Cultivars in Cluster X and XIII were the most divergent of all.While cultivars in the remaining clusters could have the closely related attribute.The formation of the solitary cluster might be due to intensive natural or human selection for diverse adaptive complexes and specific fruit quality in the growing region.The cross-pollination of mango cultivars could also result in specific gene recombination and selection made by growers for propagation might also lead to phenotypic diversity amongst the studied cultivars like most of the existing cultivars in different parts of the world originated (Singh et al., 2016).Although information is lacking on genetic divergence of mango cultivars in Ethiopia, similar study in neighbouring countries such as Kenya (Gitahi et al., 2016;Sennhenn et al., 2013;Toili et al., 2016), Sudan (Ahmed and Mohamed, 2015;Mohamed and Ahmed, 2015) and Egypt (Kheshin et al., 2016) confirmed the existence of phenotype diversity.Hence, the investigated result in Ethiopia is useful for efficient utilization and conservation of the cultivars (Majumder et al., 2013).

Distinguishing characters of clusters
The three solitary clusters (II, X and XIII) were distinguished from others by more than one characters.Cluster II had oblanceolate leaf blade shape, green with red blush ripe fruit skin color, orange ripe fruit pulp color and high fiber in fruit pulp while, Cluster X had dropping  tree growth habit, medium height of matured tree, obovoid fruit shape, green skin colour of ripe fruit, absence of fiber in fruit pulp and intermediate pulp aroma.Cluster XIII had acute leaf base shape, the yellow skin color of ripe fruit and low fiber in fruit pulp.Clusters VI, VIII and XI each had also distinguishing qualitative traits from others.Cluster VI distinguished by obtuse leaf apex shape of cultivars, while Cluster VIII established from grafted seedlings, short height of matured trees, very good eating quality and parallel pattern of fruit stone venation, mean values for seed length and trunk circumference were lower than the minimum values of cultivars.Cluster XI had the medium depth of fruit stalk cavity, had mean values greater than the maximum mean values of cultivars for fruit length, stone width, and length as well as for seed width and length.Cultivars in cluster I for stone and seed width; cluster V for trunk circumference, and cluster X for petiole length, leaf blade length and width with mean values lower than the minimum values of cultivars.Whereas cultivars in cluster XIII had mean values for fruit diameter, stone width, seed length, and width was greater than the minimum values of cultivars.Mango cultivars grouped in the rest of clusters had mean values of the characters within the ranges of minimum and maximum overall mean values of cultivars and had similarity for two or more qualitative traits.
Grouping of crop genotypes in different clusters is helpful to identify parental lines for breeding or further development of varieties through selection (Karanjalker and Begane, 2016;Majumder et al., 2013).Hence, the resulted cultivar clusters with distinguished characters can be used in mango improvement programs of the country.Likewise, several findings reported on clustering of mango cultivars with their distinguishing traits (Ahmed and Mohamed, 2015;Gitahi et al., 2016;Kheshin et al., 2016;Mohamed and Ahmed, 2015;Sennhenn et al., 2013).

Conclusion
The assessment of phenotypic characters of the studied mango cultivars in Ethiopia revealed the existence of significant phenotype variations.The quantitative characters significantly contributed to the total variation of the cultivars but with varying degree of contribution.The observed range of genetic distances and clustering of cultivars indicated the presence of considerable diversity among cultivars and the existence of cultivars with distinguished characters that can be used for the mango improvement program of the country.However, the results of the study need to be supported by further diversity assessment using molecular markers data, since phenotypic characters are less reliable due to the high influence of environmental factors.

Figure 1 .
Figure 1.Geographic locations of districts in east, central and western Ethiopia where mango cultivars were sampled.

Figure 3 .
Figure 3. Distribution of 2346 pairs of mango cultivars in respect to Euclidean distance (ED).

Figure 3 .
Figure 3. Distribution of 2346 pairs of mango cultivars in respect to Euclidean distance (ED).

Figure 4 .
Figure 4. Dendrogram depicting dissimilarity of 69 mango cultivars from the east, central and western Ethiopia obtained by Unweighted Pair-Group Method with Arithmetic Mean (UPGMA) clustering method, based on the Euclidean distances from 44 characters.

Table 1 .
The lists of 69 mango cultivars collected from three geographical regions of Ethiopia.

Table 2 .
Descriptive statistics and univariate analysis of variance of 13 quantitative traits of mango cultivars from Ethiopia.SE=standard error and CV (%) = coefficient of variation in percent.a Analysis of variance was not computed since the data were collected from a single tree.

Table 3 .
Multivariate analysis of variance (MANOVA) test criteria and F approximations for the hypothesis of no overall cultivars effect on the overall quantitative traits.

Table 4 .
Summary of tree and leaf qualitative phenotype characters of the 69 mango cultivars from Ethiopia.
a Characters according to IPGRI (2006); b Numbers in brackets indicate the percentage of cultivars per class of trait.c Chi-squared test to indicate significant differences between phenotypic classes; *** significant at 0.1%.

Table 5 .
Summary of fruit, stone and seed qualitative phenotype characters of the 69 mango cultivars from Ethiopia.
a Characters according to IPGRI (2006); b Numbers in brackets indicate the percentage of cultivars per class of trait.c Chi-squared test to indicate significance differences of phenotypic classes; *, ** and *** significant at 0.05, 0.01 and 0.001%, respectively.

Table 6 .
Principal component loadings of 15 quantitative traits in 69 cultivars of mango in Ethiopia.
Values in bold indicate the variables that contributed most to the specific principal component and the squared cosine is the largest.

Table 7 .
Range and mean Euclidean distances of 69 mango cultivars in Ethiopia.Cultivars with the initial letter M (Melkassa district) and A (Assosa district) were from central and western Ethiopia, respectively, while cultivars with initial letter H, E, and B were from eastern Ethiopia Sofi, Erer and Babile districts, respectively.SD: = Standard deviation; CV (%): coefficient of variation.

Table 8 .
Range and mean Euclidean distances of 13 clusters of mango cultivars in Ethiopia.