MATERIALS AND METHODS

The experiment was conducted at Holetta (38°E and 9°N) Agricultural Research Center (HARC) in 2010 cropping season. Holetta is one of the representatives of oil seed Brassica growing areas in the central highlands of Ethiopia with its annual rainfall of 1100 mm, altitude of 2400 m a.s.l and temperatures of 22°C (maximum) and 6°C (minimum) (Alemayehu and Mesfin, 1994). Thirty six genotypes of Ethiopian mustard including the standard checks were used in the study. The genotypes were collected by formerly Institute of Plant Genetic Resource Conservation and Research, currently Institute of Biodiversity Conservation (IBC) from diverse geographical region of the country. Genotypes by origin are described in Table 1. The experiment was carried out using 6x6 simple lattice design with two replications and with a plot size of 3x1.8 m. Each genotype was planted in a plot consisting of six rows of 3 m long with spacing of 30 cm between rows. All recommended agronomic practices (Alemayehu and Mesfin, 1994) were followed to raise good crop.

Table 1:	List of genotypes considered in the study and their origin
*Donation by foundation for Agricultural Plant Breeding S.V.P. P.O. Box 117 Wageningen, The Netherlands, -Information not available, Acc. No. = Genotype accession number

Data were collected both on plot basis and plant basis for days to flowering, days to maturity, number of primary branches per plant, number of secondary branches per plant, number of pods per plant, number of seeds per pod, plant height, number of seeds per plant, 1000 seeds weight, seed yield per plot and oil yield per plot.

Data analysis: Data were subjected to analysis of variance using the procedures outlined by Gomez and Gomez (1984). Hierarchical clustering was performed by a serious of successive mergers (agglomerative) of groups of individuals using the procedure CLUSTER (ward’s minimum) as described in SAS (1985). The most similar individuals were first grouped and these initial groups were merged according to their similarities.

Using PROC CANDISC procedure of SAS (1985), genetic divergence between clusters was calculated using the generalized Mahalanobis (1936) D2.

Principal component analysis was also done using the procedure PRINCOMP as described in SAS (1985) to clarify the relationships between two or more characters and to divide the total variance of the original characters into a limited number of uncorrelated new variables according to Wiley (1981).

Genotypic correlation coefficients (Pearson’s correlation) were computed by the genotypic values of the traits determined as the method described by Zhu (1996) but assuming the interaction component as nil (Falconer, 1981).

The path analysis technique performed according to the method suggested by Dewey and Lu (1959) using the procedure PROC CALIS of the SAS software version 9.00 (SAS, 1985).

RESULTS AND DISCUSSION

The univariate analysis of variance (not presented) showed that there were significant differences among genotypes for all traits except number of pods per plant, secondary branches per plant and number of seeds per plant. This indicates the existence of considerable genetic variability for selection and breeding.

Clustering produced a clear grouping of the 36 genotypes into seven clusters, whereby the individuals within any one cluster are more closely related than are individuals in different clusters (Fig. 1). The acceptable limit for number of cluster was made based on pseudo F and t²-statistics. Genotypes with the same geographic origin were grouped in the same cluster such as PGRC/E 21057, PGRC/E 215284, PGRC/E 208594, PGRC/E 208596, PGRC/E 208865 and PGRC/E 215562 in C5 and PGRC/E 208961 and PGRC/E 215790 in C1. This phenomenon might have resulted from their similar genetic background. On the other hand, there are also genotypes with same geographical origin but grouped in different clusters which might be due to difference in their genetic background. Besides, genotypes with different geographical origin were grouped in same cluster which might have been as a result of synchronization of selection differential applied on different components of various geographical areas. In general, the present investigation indicated that factors other than geographic diversity such as genetic drift, selection pressure, closeness in their pedigree and environment may be responsible for differential grouping of genotypes. Similarly, these situations have been reported by various authors (Teklewold et al., 2000; Verma and Sachan, 2000; Alemayehu and Becker, 2001; Iftekharuddaula et al., 2002; Jeena and Sheikh 2003; Keneni et al., 2005). On the other hand, Zhuang et al. (2011) reported that the clustering analysis done on Persian Wheat (Triticum turgidum ssp. carthlicum) accessions using EST-SSR markers suggested that most of the accessions with adjacent geographic origins had the tendency to cluster together.

Fig. 1:	Wards minimum variance dendrogram which shows the distribution of the 36 Ethiopian mustard genotypes based on morpho-agronomic traits

Solitary genotype in C2, such as PGRC/E 208507 was among donations which have been collected from Ethiopia as Brassica carinata. This Solitary genotype might have resulted from geographical barriers preventing gene flow or intensive natural and human selection for adaptive gene complexes which is in accordance to the other findings (Bhatt, 1973; Joshi and Singh, 1979; Teklewold et al., 2000). On the other hand, other donated materials were also grouped in different clusters such as C2, C3, C4, C5, C6 and C7 which may be inferred that they may genetically affiliate with their respective cluster’s member. Similarly, the two unknown genotypes such as PGRC/E 21069 and PGRC/E 213168 were grouped in the same cluster which indicates that they may be closely related in their pedigree and/or the operation of similar forces of natural or artificial selection made them to be grouped together. By the same talk, S-67 and Holetta-1 grouped together in C1 but Yellow-Dodolla was identified in C4, which indicates that these varieties may have close pedigree relationships with the genotypes in their respective clusters.

In crop improvement venture, genetically distant parents are needed for crossing/hybridization. This is because the genetic diversity between genotypes indicates differences in gene frequencies which may result for heterotic group and/or transgressive segregants. The pairwise generalized squared distance (D²) among the seven clusters is presented in Table 2. The distance among clusters ranged between 76.0 to 6328. The highest genetic distance was recorded between C2 and C4 (6328) followed by the cluster C3 and C4 (4468) and C2 and C7 (3744). The genetic divergences among other clusters were also highly significant. Parents may be selected from those clusters which had significant genetic distance for crossing in order to obtain genetic recombination and transgressive segregants in the subsequent generations. Likewise, Rohman et al. (2004) reported that high divergence between the clusters for 25 sorghum varieties which were grouped into four clusters.

Intra-class average and range of genetic divergence for traits measured is shown in Table 3. Genotypes in C3 showed earliness in days to flowering and maturity than other clusters. Similarly, high number of seeds per pod and seed yield per plot as well as high oil yield per plot and 1000 seeds weight were identified in C6 and C5, respectively.

In order to assess the patterns of variations, Principal Component Analysis (PCA) was done using 9 significant morpho-agronomic traits simultaneously (Table 4). Principal component analysis showed that 91.4% of the variation was contributed by the first five principal components. Days to flowering was the major positive contributor of the variation in the first principal component in which 40.6% of the variation revealed. Days to maturity and length of pod had relatively high positive weight. Number of seeds per pod, seed yield per plot and oil yield per plot had negative weight. It indicated that early varieties would be developed with improved seed and oil yield of the genotypes. The 21.1% additional variation in the second principal component was mainly observed through trait such as seed yield per plot and oil yield per plot.

Table 2:	Pairwise generalized squared distance (D2) among 36 genotypes of Ethiopian mustard in seven clusters based on their morpho-agronomic traits
**χ2 test at p = 0.01 significance level, C: Cluster

Table 3:	Intra-class average and range of genetic divergence in morpho-agronomic traits of the seven clusters of Ethiopian mustard genotypes
DF: Days to flowering, DM: Days to maturity, PH: Plant height, NPB: No. of primary branches per plant, LP: Length of pod, SPPD: No. of seeds per pod, SYPP: Seed yield per plot, OYPP: Oil yield per plot, TSW: 1000 seeds weight, C: Cluster, R: Range, M: Mean

Table 4:	Component scores of the first six principal components of 36 genotypes of Ethiopian mustard
DF: Days to flowering, DML: days to maturity, PH: plant height, NPB: No. of primary branches per plant, LP: length of pod, SPPD: No. of seeds per pod, SYPP: Seed yield per plot, OYPP: Oil yield per plot, TSW: 1000 seeds weight

Plant height and number of primary branches per plant had relatively high positive coefficients in this principal component. The third principal component was accounted for another additional 16.0% of the variation in which 1000 seeds weight was the major contributor. Number of primary branches per plant had the highest negative weight. Principal component 4 and 5 contributed 7.9 and 5.8% additional variations respectively. Length of pod in principal component 4, number of primary branches per plant in principal component 5 was the major contributors. Plant height in principal component 4 and days to flowering in principal component 6 had the most negative weight. In this study, most of the traits individually contributed small effects (±0.06-0.44) to the total variation and, therefore, differential grouping of genotypes was mainly attributed by the cumulative effect of the individual traits.

Genotypic correlation coefficients (Table 5) revealed that seed yield per plot showed significant positive correlation with number of seeds per pod (0.583), number of seeds per plant (0.587) and oil yield per plot (0.993) which indicates that considering number of seeds per pod and number of seeds per plant as selection criteria will be an effective way to increase both seed and oil yield.

Table 5:	Genotypic correlation coefficients among 12 agronomic traits in 36 Ethiopian mustard genotypes tested at Holetta, 2010/11
*, **Significant at p = 0.05 and p<0.01 significance level respectively. DF: Days to flowering, DM: Days to maturity, PH: Plant height, NPB: No. of primary branches per plant, NSB: No. of secondary branches per plant, NPP: No. of pods per plant, LP: Length of pod, SPPD: No. of seeds per pod, SPP: No. of seeds per plant, SYPP: Seed yield per plot, OYPP: Oil yield per plot, TSW: 1000 seeds weight

This result is in agreement with the findings of Aytac and Kinaci (2009), Engqvist and Becker (1993) and Jeromela et al. (2007) who reported positive correlation of seed yield per plot with oil yield per plot and seed yield per plant. Similarly, associations between the seed yield and the number of pods per plant, as well as number of seeds per pod and seed weight per pod was found positive (Shabana et al., 1990; Thompson, 1983).

Correlation of days to flowering with days to maturity (0.8), plant height (0.558), length of pod (0.57) was found significant and positive whereas correlation between days to flowering and number of seeds per pod (-0.519), number of seeds per plant (-0.492), seed yield per plot (-0.338) and oil yield per plot (-0.364) was negative though it was insignificant. Similarly, days to maturity showed significant positive correlation with plant height (0.493), number of primary branches per plant (0.347) and length of pod (0.435), while highly significant but negative correlation was found with number of seeds per pod (-0.577) and number of seeds per plant (-0.533). This indicated that earliness with short stature and high yielding could be achieved in efforts of variety development. Days to flowering showed significant negative correlation with 1000 seeds weight (-0.024) which implies that earliness in flowering may be achieved at the expense of 1000 seeds weight. Result of positive correlation of days to flowering with days to maturity as well as negative correlation of these two traits with seed and oil yield are in agreement with the result of Delesa (2006).

Association of plant height with length of pod (0.387) was significant and positive, while association with number of seeds per pod (-0.484) and number of seeds per plant (-0.386) was found significant but negative. Number of primary branches per plant showed significant correlation with number of secondary branches per plant (0.514) and number of pods per plant (0.377). Likewise, correlation of number of secondary branches per plant with number of pods per plant (0.674) and number of seeds per plant (0.536) was highly significant and positive. Besides, number of pods per plant showed significant positive correlation with number of seeds per plant (0.4). These results indicate that the greater the number of branches per plant, higher will be the number of pods and number of seeds per plant and thus ultimately contributing positively towards yield.

Basalma (2008), Khan and Rashid (1999), Guo et al. (1987) and Kumar et al. (1987) reported similar findings of a positive and significant correlation between number of secondary branches per plant and number of pods per plant.

In the present investigation, relationship between 1000 seeds weight and seed yield per plot was low. This agrees with the result of Basalma (2008) and Jiang and Guan (1988). Relationship of number of seeds per pod with number of seeds per plant (0.757), seed yield per plot (0.583) and oil yield per plot (0.584) was significant and positive. Similarly, number of seeds per plant showed significant positive correlation with seed yield per plot (0.587) and oil yield per plot (0.595). Most of the correlation results of this study are in agreement with findings of Sheikh et al. (1999) who reported positive correlation of all yield components with seed yield. These results also substantiated the results of principal component analysis for instance, positive correlation of umber of seeds per pod with seed and oil yield per plot and negative correlation of days to flowering, days to maturity and length of pod with number of seeds per pod, seed yield per plot and oil yield per plot had been expressed in the study of pattern of variations too.

Cause and effect interrelationships analysis (Table 6) showed that number of seeds per plant and oil yield per plot showed positive direct effect on seed yield per plot, these traits also showed positive and significant correlation with seed yield per plot. Therefore, considering these traits as selection criteria will be advantageous in improvement of Ethiopian mustard genotypes. This result is inconformity with findings of Singh and Singh (1997) and Basalma (2008). Days to maturity showed positive direct effect on seed yield per plot, while they had negative correlation which may be as a result of its negative effects via other traits. Similarly, number of primary branches per plant and length of pod showed no direct effect on seed yield per plot, while they had correlated with seed yield in which their indirect effect via other traits might be the cause. On the other hand, days to flowering had negative direct effect on seed yield per plot which had also been negatively correlated with seed yield per plot which implies consideration of this trait in breeding work for achieving earliness in days to flowering is valuable. Number of seeds per pod revealed negative direct effect on seed yield per plot but its indirect effect through oil yield per plot was high and positive. Direct effect of 1000 seeds weight on seed yield per plot was negative, which indicates selection of this trait may be ineffective in improving the seed yield per plot, however, its effect via number of seeds per plant and oil yield per plot is helpful.

Table 6:	Genotypic direct and indirect effects of agronomic traits on seed yield per plot tested at Holetta, 2010/11
Bold values show direct genotype. **Significant at p<0.01 significance level. DF: Days to flowering, DM: Days to maturity, PH: Plant height, NPB: No. of primary branches per plant, NSB: No. of secondary branches per plant, NPP: No. of pods per plant, LP: Length of pod, SPPD: No. of seeds per pod, SPP: No. of seeds per plant, OYPP: Oil yield per plot, TSW: 1000 seeds weight, rg: Genotypic correlation coefficient

CONCLUSIONS

This study has shown the existence of considerable genetic variation among the genotypes considered which may help for further selection and breeding. Parents may be selected from those clusters which had significant genetic distance for crossing in order to obtain genetic recombination and transgressive segregants in the subsequent generations. However, it is also worthy, considering genotypes within a cluster with respect to a trait of interest. Traits association and cause and effect interrelationships have revealed that number of seeds per plant is the most important yield component. However, further study across location and years needs to be done in order to corroborate the result obtained in the present investigation.

ACKNOWLEDGMENT

The study was sponsored by the Ethiopian Institute of Agricultural Research (EIAR) and Rural Capacity Building Project (RCBP).