Genetic diversity of qualitative traits of barley ( Hordeum Vulgare L . ) landrace populations collected from Gamo Highlands of Ethiopia

Barley (Hordeum vulgare L.) has great adaptability to a wide range of environments. To determine genetic diversity in barley landraces, a total of 43 landrace populations were randomly sampled from the farmers’ field on plant basis and characterized for eight qualitative traits; namely, kernel row number, spike density, lemma awn barb, glume color, lemma type, length of rachila hair, kernel covering and lemma/kernel color. Morphological diversity was determined by the Shannon-Weaver index (H’). Overall barley landrace populations showed an average diversity index of 0.59, implying large diversity for the populations. Selection for adaptation to different altitude classes appears to be the main factor that has determined the observed variation, along with population-size effects. The result showed that barley landraces from Gamo highlands, Ethiopia are constituted by highly variable landraces that have large within-population diversity. These landraces are also shown to be locally adapted, with the major driving force that has shaped their population structure being consistent with selection for adaptation along an altitudinal gradient. Overall, this study highlights the potential of such landraces as a source of useful genes that can be exploited in crop improvement programmes.


INTRODUCTION
Barley (Hordeum vulgare L.) is one of the world's most ancient food crops.It has been an important cereal crop since 8,000 to 10,000 years ago in the area of the Middle East known as "the Fertile Crescent" (Giles and von Bothmer, 1985;von Bothmer and Jacobsen, 1985).It is ranked third among the major cereal crops on the basis of production tonnage after wheat and rice (FAOSTAT, 2013).In order of importance, barley is used for animal feed, brewing malts and human consumption (Hayes et al., 2002).Ethiopia is recognized as a major Vavilovian gene center.Earlier introduction from Mediterranean countries and centuries of natural and artificial selections on native crops has resulted in tremendous genetic diversity.Among the major cereal crops in which valuable genetic diversity observed were tetraploid wheat and barley are the most prominent crops for which the country is recognized as the secondary center of diversity (Demissie and Giorgis, 1991).
The characterization of genetic variability between and within populations is important for determining the rate of adaptive evolution and response to traditional crop improvement (Hunter, 1996).Genetic diversity is a raw material for evolution, thus enabling populations of species to survive, evolve and adapt to resist long-term changes in the environment.Differences within and between population can be of a strategic value to conservation as they provide a clear justification for protecting the valuable genes across their entire geographic range and all the subspecies of major populations.
The barley landraces exhibit variation both between and within populations.This within population diversity of these barley landraces might allow them to cope with environmental stresses, which is very important for achieving yield stability (Zhu et al., 2000).The knowledge of the population structure of Ethiopian barley landraces, together with a deeper understanding of the nature and extent of their variation, is an important prerequisite for the efficient conservation and use of the existing plant materials.Several studies have reported a high level of genetic diversity in barley populations from Ethiopia, such as those based on morphological traits (Engels, 1994;Demissie and Bjornstad, 1996) and on biochemical data (Bekele, 1983a).
Thus, the study was conducted to assess the extent of morphological diversity in Gamo highlands of Ethiopia in relation to districts and altitudes.This characterization was carried out based on spike morphological traits, to classify landrace genotypes into morphologically similar clusters, and to identify the major traits responsible for most of genetic variation among the collected landrace populations.

Plant collection
The barley landrace populations were collected in 2012 in the Gamo highlands of Ethiopia.A total of 43 populations/farmer varieties were collected during October and November, within three districts (Chencha, Dita and Bonke) in Meher growing season (Figure 1) considering three altitude classes randomly chosen based on collection data: 2000-2395, 2395-2683 and 2683-3000 m.a.s.l.
The three districts were selected based on their barley crop production potential, long rainy season which covers from March to September.Agro-ecological condition varies from district to district; Dita district, has an altitude ranging from 1800-3500 m a.s.l. and the annual rainfall ranges between 801 to 1600 mm.The rainfall is of a bi-modal pattern, giving rise to two distinct seasons, the short rainy season from March to May (peak in April).It has three types of soil based on their colors: red (85%), black (13%) and brown (2%) and their texture is clay loam.The minimum and maximum temperature is 10.1 and 27.5°C, respectively.Chencha district has an altitude range from 1800 to 3500 m.a.s.l.The area has bimodal rainfall.The annual rainfall ranges between 1201 and 1600 mm.Bonke district is highly degraded due to extensive tree cutting for fuel wood.This has resulted to continuous top soil loss due to erosion.Rainfall in the area is erratic and unreliable and ranges between 801 and 1600 mm.It has an altitude range of 801-3500 m.a.s.l.(Daniel, 1977).
The geographical position (latitude, longitude and altitude) of each sampling site was determined using a Geographical Positioning System (GPS).Each of the surveyed farmers' field size of barley was measured and found to have average range approximately 0.12 -0.25 hectare.The name of each landrace population was recorded during sampling (as informed by the farmers).
Twenty spikes per population were randomly sampled from each site.Accordingly, a total of 860 spikes/ears were used to measure the phenotypic diversity using qualitative morphological traits that are easily seen by the eye, expressed in all environments and governed by few genes.Data on the following eight morphological traits from the matured spikes were recorded.These included kernel row number, spike density, lemma awn barbs, glume color, lemma type, length of rachilla hair, kernel covering and lemma/kernel color (Table 1).The phenotypic character states of each qualitative trait were recorded using Bioversity International Barley Descriptor (1994).Color traits were recorded using the Eagle Spirit Ministry Color Chart which was developed by Kohe't (1996).

Statistical analysis
Data were analyzed for eight qualitative traits (Table 1) that have shown their usefulness in earlier studies (Engels, 1994;Demissie and Bjornstad, 1996;Tanto et al., 2009) and proved to be consistent over the years in characterization work and their scoring turned out to be reliable.
Polymorphism is measured by the percentage of polymorphic loci (PPL), Nei's genetic diversity (H), for single population or groups of populations (by landrace name, altitude class or district), within population diversity (HS), coefficient of gene differentiation (GST) (Nei, 1973) and Shannon information index of diversity (H') (Lewontin, 1972), were all calculated by using Pop Gene ver.1.32 (Yeh et al., 2000) software.The haploid option of the software was used for analysis in accordance with the assumption of Ferguson et al. (1998).
The actual sample sizes per altitudinal class and districts are presented in Table 2.For the comparative analyses of the diversity indices, the altitude classes were arbitrarily grouped into three altitudinal classes.
The Shannon diversity index (H') was also calculated as group of populations for districts, altitude classes and populations/farmer varieties by using the Shannon-Weaver equation as presented by Poole (1974)

Gegnaw and Hadado 665
Where n represents the number of phenotypic classes of a given character, and Pi the proportion of the total number of accessions consisting of the i th class.Each value of H' was divided by ln n (where ln n is a natural logarism of (n) in order to keep the values of H' between zero and one (Goodwin et al., 1992).
A one-way analysis of variance (ANOVA) for a normalized diversity index (H') was carried out for each trait using districts and altitude tude classes as classifying variables.These were treated as fixed effects and populations (farm fields) as random using SPSS Statistics 17 software.
SPSS statistics 17 software, was further used to analyze the total morphological data variation, according to two nested models of analysis of variance (ANOVA): (1) among populations, among individuals within populations (where the 'within populations' component is accounted for by the within individual barley fields); and (2) among 'groups', among populations within groups, and among individuals within populations (where the 'groups' were the landrace names/ farmer varieties, the altitude classes, the districts and the 'populations' were represented by individual barley fields).
To study the patterns of diversity among the sampled populations, a dendrogram was obtained on the basis of Nei's unbiased pair wise distance matrix between populations (1973).This showed the relationships among all the 43 landrace populations.For this analysis, the Pop Gene ver.1.32 (Yeh et al., 2000) software was used.

Morphological diversity of the whole population
A total of 90 different combinations of traits (morphotypes) were detected, of which 25 accounted for 75% frequency of the total collection (647 out of 860). Figure 2 shows the percentages of the morphotypes belonging to the 25 most frequent morphotypes ranging from 1.8-5.7%.

Morphological diversity indices derived from qualitative traits
Estimates of Shannon diversity index (H') analyzed for individual traits are presented in Table 3.These estimates ranged from 0.03 (low polymorphic) for lemma awn barb to 0.91 (highly polymorphic) for spike density.

Phenotypic diversity index (H') across the three districts
Unlike the individual population estimates, the district wise comparison displayed a higher magnitude of differences.It ranged from average H' 0.52 ± 0.11 for Dita to 0.59± 0.11 for Chencha.All the three districts showed medium to high (H' > 0.52 ± 0.11) level of polymorphism.Samples from Chencha were the most diverse (H' = 0.59 ± 0.11) followed by Bonke ( H' = 0.56 ± 0.11) and Dita (H' = 0.52 ± 0.11) (Table 4).The analysis of variance of H' for indivi-dual traits was performed and much of the variation was due to variations within populations rather than districts (Table 5).

Phenotypic diversity indices across altitude classes
Like district wise estimates, Shannon diversity index (H') values pooled over all traits for altitudinal classes showed less variation (Table 6).For characters like glume color, kernel covering and lemma color, an increase of the diversity index was observed with increasing altitude from 2,000 to 2,683 m.a.s.l and after this altitude class, a decrease occurs but not statistically significant.When the diversity for single traits was considered, the diversity only in kernel row number increased with an increasing altitude.No comprehensible associations with altitude classes were seen for the traits: length of rachila hair, lemma type and spike density.The above results indicated that the mean diversity indices indeed vary with altitude and that the indices are relatively highest for medium altitudes between 2,395 and 2,683 m.a.s.l.The analysis of variance of diversity (H') by considering individual traits were also performed and much of the variation was due to variations within altitude rather than among altitudes.However, length of rachila hair showed significant variation among altitude classes (P < 0.05) (Table 7).

Differentiation comparison between districts and altitudes
The divergence estimates between the districts, altitude classes and populations are given in Table 8.When consisidering the genetic diversity differences between the three districts, an overall medium differentiation level (13.0%) was found; the differentiation was significantly different from zero (p < 0.001).The differentiation between districts was significant for six of the eight traits considered (kernel row number, spike density, lemma awn barb, length of rachilla hair, kernel covering and lemma color; 75%) with values ranging from 3.4% (lemma  awn barb) to a maximum of 39.36% (kernel row number).Kernel row number and length of rachilla hair were the only traits for which the differentiation value was higher between the districts than among the altitude classes.The level of differentiation between the altitude classes (15.6%; p < 0.001) surmount that seen among the three districts.
The differentiation among the landrace fields was relatively high (31.2%;p < 0.001) as compared to that of among the altitude classes and districts.Thus, it is remarkable that the largest amount of total morphological variance was seen among individuals within populations (68.8%).

Cluster analysis
Cluster analysis was used to examine the aggregation patterns for all 43 barley landrace populations (Figure 3).
However, four major groups of populations accounted for 38 of the 43 landrace populations (88%).These four clusters have different compositions in terms of districts and altitude classes (Table 9).The number of landrace populations per cluster varied from 15 landraces in cluster II to 6 landraces in cluster I and IV.About 23% of landraces from Bonke grouped in cluster I and all of them have predominance of lemma teeth and covered grains since the majority of the landraces were from altitude class I. Furthermore, cluster I was the only cluster without landraces from altitude class II.Cluster II included landraces from all districts with the highest percentage from Chencha (53%) which is a district contributing the highest percentage of landraces collected from altitude classes I, II and III and most of them have brown glume colors, lax and dense spike types, and

Population diversity
Shannon diversity index estimated from all data pulled together varied from 0.03 to 0.91 with an overall mean of 0.59 (Table 3).This implied the existence of considerable variability in the studied barley population.There was also variability in barley population from studies of 51 Ethiopian barley accessions and it was found that the Shannon diversity varied from 0.00 to 0.62 for barley populations, and the diversity index estimated for an overall population ranged from 0.29 to 0.92 with an overall mean of 0.71 (Demissie and Bjornastad, 1996).This data implies that over all, Ethiopia has similar barley genetic diversity taking only one zone Gamo goffa as compared to the whole country.

Diversity based on districts
This study showed morphological variation for districts and altitude classes based on qualitative characters, which indicated that, the structure of morphological variation in Ethiopian barley landraces was influenced mostly by natural selection factors.So the degree of variation for characters differed with districts and altitudes from where the landrace populations originated.presence of high level of phenotypic diversity in Ethiopian barleys was also reported by different authors (Negassa, 1985b;Vavilov, 1926;Asfaw, 1989;Engels, 1994;Tolbert et al., 1979).However, the estimate presented by Tolbert et al. (1979) is fairly low (mean H' = 0.51) when compared with Negassa's (mean H' = 0.68), Engels' (mean H' = 0.70), and the present study, the average diversity index of 0.59 was recorded for Gamo gofa zone alone (Table 4).This inconsistence might be due to inadequate sample size and/or dissimilar and different numbers of characters analyzed by others (Negassa, 1985b;Engels, 1994).The diversity index value in the present study is significantly different from the highest diversity value, 0.70, for Ethiopia as reported by Engels (1994).This difference in average H' estimates observed by Engels (1994) and the present study could be due to: 1) the different number of populations studied; 2) the different sampling procedures and strategies; or 3) the different types of characters used for analysis.
The pooled indices over characters within districts are relatively different.Dita had the lowest values which might be explained by the possibly greater degree of selection pressure exerted in the extremely marginal production conditions in Gamo gofa zone as compared to the other two districts, for instance the lowest rainfall and high temperature availability in Dita district.Another possible factor could be the relatively less number of landrace populations collected from this district based on the availability of landraces.However, Chencha district showed highest average diversity index value as compared to other two districts.This is probably due to the presence of high rainfall and/or low temperature which is important for barley in such types of climatic conditions.

Phenotypic diversity indices over altitudes
The observed variations were mainly due to traits within populations and altitude classes.The estimates of H' for each of the characters and the altitude classes as well as the mean H' per altitude class are presented in Table 6.Despite different levels of contributions of the traits to variation, the average H' tends to decrease for samples above 2683 m.a.s.l.(Table 6).The highest polymorphism is concentrated in areas between 2395 to 2683 m.a.s.l.This range includes the major barley-growing areas in the Gamo gofa zone.These phenomena, taken together, are indicative of high genetic diversity and abundance of barley in the medium altitude class.Our results are in agreement with those reported by Engels (1994) that a maximum diversity occurred in a medium altitudinal class.In the present study, "glume color", "kernel covering" and "lemma color" showed such pattern of differences.For these characters, the phenotypic expression does depend on altitude class.The result indicated that for these characters genetic diversity increased with an increasing altitude and then decreased with decreasing altitudes, as a result of natural selection at the lower and upper extreme conditions.Or in other words, the diversity indices decreased towards lower as well as higher altitudes.

Differentiation comparison between altitudes and districts
Analysis of variance was conducted for districts, altitude classes and populations with the assumption that there Gegnaw and Hadado 671 existed high variation between populations, followed by altitude classes and the districts.The result showed high variation between populations.Percentage (%) of morphological variation, which described the level of heterozygosity was low between districts (13%) and altitude classes (15%) and there was high genetic differentiation among populations (31%).Generally, high differentiation within population resulted in less gene flow among population but high gene flow within population due to some percent of cross pollination in the species coupled with planting of mixed barley populations by farmers.
A comparison of the present work with an earlier study (Tanto et al., 2009) showed a similar pattern of differentiation value among the districts, altitude classes and populations, but differentiation among populations (31%) is relatively higher as compared to differentiation among populations (25%) as studied by Tanto et al. (2009).This was probably due to sample and population size differences in the two cases.

Hierarchical clustering
The differentiation among the populations can be clustered into different coherent groups depending upon the origin/district or altitudinal classes.In the present study, the differentiation among population appeared to be weak on the basis of district/origin.Similarly, others like Demissie et al. (1998), Ould Med Mahmouda andHamza (2009) and Tiegist et al. (2010) reported lack of geographical differentiation which failed to indicate clear pat tern of division among barley accessions based on geographic origin.These results may reflect the impact of the seed exchange between farmers in small geographic areas which is likely to limit highlighting favorable genes due to local adaptation.
Grouping accessions into morphologically similar cluster of different groups is useful for selecting parents for crossing.However, clustering of collections based on the morphological traits under study revealed no distinct/origin grouping patterns because same or adjacent districts appeared in different clusters.

Conclusion
This study, presented the results of a morphological characterization of a collection of barley from Gamo highlands of Ethiopia.
The overall diversity index of Gamo highlands of Ethiopia barley collections used in this study and similar results from other studies made earlier, support the conclusion that Ethiopia is an important centre of genetic diversity for barley.This diversity is not evenly distributed over the barley producing districts in the study area.However, there was decrease in diversity from the medium altitude towards the lower and higher altitudes.Therefore, morphological traits which are under direct influence of both human and natural selections are strongly associated with altitude.This implies that to capture the most diverse genotypes one should concentrate on the medium altitudes between 2,395 and 2,683 m.a.s.l., which correspond with the best growing conditions for barley.The concentration of some morphological traits at high or low altitudes and in different sites could result from farmers selection activity based on their selection criteria to the prevailing climatic and edaphic conditions, and because of presume stronger natural selection pressure towards the extremes of the altitudinal distribution range; for instance one might expect certain desirable genotypes for abiotic stress tolerance, such as frost or drought resistance to be found at higher frequencies in specific areas.
Although genetic differentiation was less both on districts and altitudinal classes, altitudinal differentiation relatively greater than districts.Therefore, altitude difference was relatively more discriminative than origin / districts.This is probably due to the presence of low seed flow among altitude classes as compared to among districts and due to natural selection and adaptation to specific altitudinal classes.

Recommendation
The above conclusions were derived from results of studies conducted by using morphological traits.Therefore, the following recommendations are forwarded for future work.First to make this study more comprehensive, molecular marker techniques should be applied to confirm the morphological results obtained; second to study quantitative traits under field evaluation is required after seed increase by planting the genotypes on ear-to-row.Third using morphological, molecular and field evaluation, one can be able to reach the level of variety development.
The conservation of such locally common variation is important, since it may represent genotypes adapted to specific environments.Thus, the patterns of variation described in this study may be useful for researchers in designing studies on barley germplasm elsewhere.

Figure 1 .
Figure 1.Map of Southern Nations Nationalities and Peoples Region showing the site of Gamo highlands.

Figure 3 .
Figure 3. Unweighted pair group method of arithmetic mean (UPGMA) dendrogram based on Nei (1973) genetic distance, showing the relationships among all barley landrace populations.
and it was in line with Pop Gene ver.1.32 results.

Table 1 .
States (character)used for each of the eight traits considered.

Table 2 .
Distribution of the landraces/farmer varieties collected across districts and altitude classes.

Table 4 .
Estimates of the diversity indices (H') for the three districts and mean diversity (H') and the standard error for the overall characters.

Table 5 .
Analysis of variance for districts of origin using 8 qualitative traits.

Table 6 .
Estimates of the diversity indices (H') for the three altitude classes, eight characters and their standard errors.

Table 7 .
Analysis of variance for altitudinal origin using eight loci.

Table 8 .
Percentage (%) of the total morphological variation and significance levels for each spike trait considered.

Table 9 .
Distribution of 38 barley landraces over four clusters by three districts and altitude classes.