Genetic variability , correlation and path analysis of yield and grain quality traits in bread wheat ( Tritium aestivum L . ) genotypes at Axum , Northern Ethiopia

Forty-nine bread wheat genotypes were tested at Axum, Northern Ethiopia in 2016/17, with the objective of assessing the extent of genetic variation, correlation and path analysis of wheat genotypes in yield and grain quality traits using 7 x 7 triple lattice design. Data were collected for 17 agronomic and grain quality characters. For each of the test entries, samples of 500 g grains were taken from each plot for quality analysis. The NIR spectrophotometer (NIR Infratec 1241 Grain analyzer, Sweden) was used to analyze wheat samples for their protein, wet gluten, zeleny sedimentation volume and starch content based on dry weight basis. Data were subjected to analysis of variance which revealed significant differences among the genotypes for all the characters. The genotypic coefficient of variation (GCV) ranged from 1.63 (for starch content) to13.30% (for grain yield). The broad sense heritability (H 2 ) ranged from 15.89 (for number of tillers) to 97.16% (for days to heading), while genetic advance as percent of mean (GAM) from 2.01 (for starch content) to 19.63% (for days to heading). The GCV and phenotypic coefficient of variation (PCV) differences were low in magnitude for days to heading and days to maturity, and H 2 values were coupled with moderate to high GAM. This suggests selection based on phenotype of genotypes could be effective to improve these characters. Grain yield was positively and significantly correlated with biological yield (0.72), harvest index (0.65), plant height (0.51), thousand kernel weight (0.31), hectoliter weight (0.37) and starch content (0.32), of which biomass yield (0.85) and harvest index (0.70) had the highest positive direct effect on grain yield. Thus, selection for higher mean values of biomass yield and harvest index could be considered simultaneously for selection of higher grain yield.


INTRODUCTION
Wheat is one of the most important export and strategic cereal crops in the world and in Ethiopia in terms of production and utilization (Suresh, 2013).It is the second most important staple food crop of the world; it provides more calories in human diet than any other crop worldwide.It accounts for nearly 30% of global cereal *Corresponding author.E-mail: brehortic@gmail.com.
Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License production, covering an area of 222.42 million hectares with total production of 725.12 million tons (FAO, 2015).Given its predominance in human diets, cultivated wheat has to meet the specific quality criteria for the manufacture of a wide range of food products derived from it.
Wheat is one of the most important small cereal crops in Ethiopia, which ranks fourth both in area coverage (1,663,845.63 hectares) and in total annual production (4,231,588.716 tons).The productivity of the crop remains low (2.54 t ha -1 ) (CSA, 2015) in the country as compared to the world average yield (3.19 t ha -1 ) (FAO, 2013).The low yield per hectare is attributed to many factors, such as unavailability of quality seed for varieties that are high yielding as well as adapted to wide range of agro-ecologies of the country.Hence, the first step in the development of varieties is assessing the genetic variability of available genotypes for the characters of interest (Rahman et al., 2016).High genetic advancement coupled with high heritability estimates offers the most suitable condition for selection (Johnson et al., 1955).The presence of variability, heritability and genetic advance in different yield related characters of bread wheat has been reported by Desalegn and Chauhan (2016), Kifle et al. (2016) and Rahman et al. (2016).However, no variability studies have been conducted in the study area.Moreover, the variability studies in the region were not on moisture stress tolerant bread wheat genotypes.In addition, genetic information is limited to grain quality traits in bread wheat genotypes evaluated in the country.Considering the importance of such information, this research was initiated with the objective of assessing genetic variability for yield and grain quality traits, and determining the association among the yield components of bread wheat genotypes.

MATERIALS AND METHODS
Field experiment was conducted at Axum Agricultural Research Center (AxARC), Northern Ethiopia during 2016-2017.The experimental site is located at latitude 13°15'40.2''N, and 38°34'45.8''E longitudes with an altitude of 2148 m above sea level.It is characterized by uni-modal rainfall pattern concentrated in one season from July to August with total annual rain fall of 500 to 782.8 mm per annum.The mean minimum and maximum temperatures ranged from 12.6 to 25.51°C, respectively.The soil type of the site is clay type with pH ranging from7.5 to 8.3.A total of 49 bread wheat genotypes introduced from ICARDA-CIMMYT (Table 1) were included in the study.The experiment was laid down in 7x7 triple lattice design.Each genotype was planted in a plot consisting of six rows of 2.5 m long and 1.2 m width; a total of 3 m 2 with spacing of 20 cm between rows.The distances between plots, blocks and replications were 0.5, 0.5 and 1.5 m, respectively.A seed rate of 150 kg ha -1 and fertilizer rate of 100-100 kg ha -1 N-P2O5 in the forms of Urea and DAP (di -ammonium phosphate) were used.
For each of the test entries, samples of 500 g grains were taken from each plot for quality analysis.The NIR spectrophotometer (NIR Infratec 1241 Grain analyzer, Sweden) was used to analyze wheat samples for their protein, wet gluten, zeleny sedimentation, starch content and moisture content based on dry weight basis.While, hectoliter weight was estimated using grain analyzer computer 2100.

Data collected
Data were collected both from plot and plant basis.The four central rows were used for data collection based on plots, such as days to 50% heading, days to physiological maturity, grain yield, biomass yield and harvest index.Ten randomly selected plants from the four central rows of each plot were used for data collection on plant basis and the averages of the ten plants in each experimental plot were used for statistical analysis for traits such as plant height, productive tillers per plant, number of kernels per spike, number of spike lets per spike and spike length.

Data for grain quality traits
For each of the test entries, samples of 500 g were taken from each plot for quality analysis and the NIR spectrophotometer (NIR Infratec 1241 Grain analyzer, Sweden) was used to analyze wheat samples.

Statistical analysis
The mean values of the genotypes were subjected to analysis of variance based on triple lattice design.Analysis of variance was done using Proc lattice and Proc GLM procedures of SAS version 9.1.3 (SAS Institute Inc, 2004) after testing the ANOVA assumptions.Mean separations were estimated using Duncan's multiple range (DMRT) test at 5% probability levels.

Estimation of variance components and association among characters
The phenotypic and genotypic coefficients of variation were estimated according to the methods suggested by Burton and De Vane (1953).
Where, σ 2 p = phenotypic variance and ̅ = mean of the characters evaluated.

GCV=
√ ̅ x 100 Where, σ 2 g = genotypic variance, ̅ = mean of the characters evaluated.Broad sense heritability was computed for each character based on the formula developed by Allard (1960) as: H 2= x 100 The genetic advance (GA) for selection intensity (K) at 5% was calculated by the formula suggested by Allard (1960) as: Where, GA = Expected genetic advance, σp = the phenotypic standard deviation, H 2 = broad sense heritability, K= selection differential (K=2.06 at 5% selection intensity).

Correlation coefficient
Estimation of genotypic and phenotypic correlation coefficients was done based on the procedure of Dabholkar (1992).

Path coefficient analysis
Path coefficient analysis which refers to the estimation of direct and indirect effects of the yield attributing characters on grain yield was calculated based on the method used by Dewey and Lu (1959) as follows: The residual effect, which determines how best the causal factors account for the variability of the dependent factor yield, was computed using the formula: Where, p 2 R is the residual effect; p ij rij = the product of direct effect of any variable and its correlation coefficient with yield.

RESULTS AND DISCUSSION
The mean values for 17 characters of 49 bread wheat genotypes are presented in Appendix Table 1.
Genotypes had in between 49 to 73.33 days to heading and 87 to 118 days to maturity with a mean of 57.99 and 101.83 days, respectively.The result showed a wide range of variations for days to heading and maturity.Grain yield ranged from 2.37 to 5.44 t ha -1 with a mean of 3.95 t ha -1 .Maximum grain yield was obtained from the genotypes ETBW9016 (5.44 t ha -1 ), ETBW8480 (5.37 t ha -1 ), ETBW8475 (4.64 t ha -1 ) and ETBW8486 (4.56 t ha -1 ).Grain protein content ranged from 11.93% for the check variety King bird to 15.43% for ETBW8489 with a mean value of 13.79%.
Mean squares of 17 characters from analysis of variance (ANOVA) are presented in Table 2.The analysis of variance showed highly significant (P<0.01)differences among genotypes for all the characters except number of effective tillers per plant and harvest index in which genotypes had significant differences (P<0.05).Significant genetic variation among genotypes for various characters suggested that the genotypes were genetically diverse and could be a good opportunity for breeders to select genotypes for trait of interest.Several researchers reported significant differences among bread wheat genotypes studied (Kifle et al., 2016;Kumar et al., 2016;Tesfaye et al., 2016;Birhanu et al., 2016).

Estimation of variability components
The estimated phenotypic coefficient of variation (PCV) and genotypic (GCV) coefficients of variations are presented in Table 3.The GCV ranged from 1.26% for starch content to 13.30% for grain yield and PCV from 1.63% for starch content to 21.88% for number of productive tillers per plant.The GCV and PCV values were categorized as low (<10%), moderate (10 to 20%) and high (>20%) as indicated by Deshmukh et al. (1986).Accordingly, moderate GCV and PCV was observed for grain yield (13.30 and 18.21%) and biomass yield (10.64 and 15.81%), respectively.This indicated that the genotype could be reflected by the phenotype and the effectiveness of selection based on the phenotypic performance for these characters.Report of Birhanu et al. (2016) is in line with the occurrence of GCV and PCV media in this study.
The PCV value was high for number of productive tillers, while medium PCV values were observed for harvest index, kernels per spike, thousand seed weight and Zeleny sedimentation value.The lowest GCV and PCV were recorded for days to heading, days to maturity, grain filling period, plant height, number of spikelets per spike, hectoliter weight, grain protein content, wet gluten content and starch content.The result indicates the environmental factors had more influence on the expression of these characters than the genetic factors, suggesting the limited scope for improvement of these characters by direct selection of high performing genotypes.This is in agreement with reports of Naik et al. (2015) and Rahman et al. (2016).

Estimation of heritability and expected genetic advance
The heritability estimates ranged from 15.89% for number of productive tillers per plant to 97.16% for days to heading.According to Singh (1990), for a character with high heritability (≥80%), selection is fairly easy, because there would be a close correspondence between genotype and phenotype due to a relatively smaller contribution of environment to phenotype.High heritability was estimated for days to heading (97.16%) and days to maturity (87.81%).This implies the variation observed was mainly under genetic control and was less influenced by the environment and the possibility of progress from selection.The obtained results are in agreement with results reported by Tesfaye et al. (2016).Moderate heritability values (40-80%) were computed for grain filling period, plant height, kernels per spike, spike lets per spike, spike length, biomass yield, grain yield, thousand kernel weight, hectoliter weight, grain protein content, wet gluten content and Zeleny sedimentation value.Low heritability (<40) estimated for number of effective tillers per plant and harvest index indicated that  selection for these characters would not be effective due to the predominant effects of nonadditive genes.In consonance with the current result, Desalegn and Chauhan (2016) (Johnson et al., 1955).High heritability is coupled with moderate genetic advance as percent of mean observed for days to heading and days to maturity.This indicates that most likely the heritability of these characters is due to additive gene effects, and selection might be effective for these characters (Salman et al., 2014;Rahman et al., 2016).

Correlation of grain yield with other characters
Grain yield had positive and highly significant (P<0.01)genotypic correlation with biomass yield (0.65) and harvest index (0.53) (Table 4).Grain yield also exhibited positive and significant (P<0.05)genotypic correlation with plant height (0.51), thousand kernel weight (0.31), hectoliter weight (0.37) and starch content (0.32).The positive association of these characters with grain yield might be due to the higher assimilation of photosynthesis as biomass because of the increased plant height and the more photosynthesis partitioned to kernels that increased their weight and thereby harvest index.This suggested that improvement of biomass yield would result in a substantial increment on grain yield that could be used in selection of genotypes for high grain yield at optimum condition.
According to Kearsey and Pooni (1996), the positive correlation of these characters with grain yield resulted from the presence of strong coupling linkage of genes or the characters may be the result of pleiotropic genes that control these characters in the same direction.They further suggested that the presence of such genes effects leads to the improvement of yield as seen in these characters.The positive and significant association of grain yield with biological yield and harvest index had been reported by Kifle et al. (2016), Kumar et al. (2016) and Ebrahimnejad and Rameeh (2016).The work of Surma et al. (2012) showed positive and significant correlation of grain yield with thousand kernel weight, hectoliter weight and starch content.In contrast to the current study result, Singh (2014) reported the presence of negative correlation between grain yield and plant height.
Grain yield was negatively and significantly correlated with grain protein content (-0.38).It also had negative and non-significant association with wet gluten content and Zeleny sedimentation value.The low yielding ability of the high protein genotypes is usually explained by the high energy needed for protein production as compared to starch production (Monaghan et al., 2001).But under ideal environment, assimilates are used more for grain yield than protein content.This indicated the importance of considering harvest index as it contributed more to the grain yield.However, different hypotheses dealing with the cause of this negative correlation have been also proposed, mainly related to genetic incompatibility (linkage, pleotropy) (Iqbal et al., 2007).Therefore, care should be given while selecting genotypes for grain yield and grain protein content.The results obtained in this study are in agreement with the findings of Surma (2012), in which grain yield was negatively correlated with protein content, wet gluten and Zeleny sedimentation value.Days to maturity had significant and negative association with number of productive tillers (-0.42 and -0.17), harvest index (-0.34 and -0.19), thousand kernel weight (-0.35 and -0.23), hectoliter weight (-0.49 and -0.27) and grain protein content (-0.30 and -0.27) both at genotypic and phenotypic levels (Table 4).The negative association of grain protein content with maturity suggested that early maturity and high protein content can be readily achieved simultaneously.

Genotypic path analysis
Biomass yield (0.85) followed by harvest index (0.70) exerted the highest positive direct effect on grain yield, while plant height had negligible positive direct effect, though it exhibited significant and positive association with grain yield (Table 5).The result indicated that the positive and significant correlation of biomass yield and harvest index with grain yield at genotypic level was due to the direct effect of these characters on grain yield.However, the positive association of plant height with grain yield was due to the indirect effect of this character on yield through other characters such as biomass yield, grain filling period and days to heading.The maximum positive genotypic direct effect of biomass yield and harvest index on grain yield was reported by many authors (Obsa, 2014;Dargicho et al., 2015;Alemu et al., 2016).
The genotypic correlation coefficients of thousand kernel weight, hectoliter weight and starch content were significant and positive with grain yield; however, these characters had low and negligible negative direct effect on grain yield.This implies that the indirect effects of these characters on grain yield through other characters could be the cause for significant and positive correlation.For instance, the indirect positive effect of thousand kernel weight via harvest index (0.27), hectoliter weight via harvest index (0.25) and starch content via biomass yield (0.29) on grain yield were high.This shows the importance of considering harvest index and biomass yield when selection of wheat genotypes for higher grain yield is desired.In agreement with the current study results, similar results were reported by Ermias (2005), Senayt (2007) and Adhiena (2015).Grain protein content exerted negative direct effect on grain yield, consequently, selection of genotypes for high performance of grain protein content might not be effective when the breeding objective is selection of genotypes for high grain yield.Singh (2014) reported negative direct effect of grain protein content on grain yield.

Conclusion
The study indicated the presence of wide genetic variation among the wheat genotypes which can be exploited to develop high yielding varieties with desirable grain quality and early maturity in the study area and similar agroecologies, where terminal moisture stress is the major constraint of wheat production.Moderate GCV coupled with moderate PCV (10 to 20%) was observed for grain yield and biomass yield, indicating the effectiveness of selection based on the phenotypic performance of the genotypes.High heritability (>80%) coupled with moderate genetic advance as percent of mean (10 to 20%) was observed for days to heading and days to maturity.This implies that the variation observed was mainly under genetic control and the possibility of progress from selection.In general, in the context of plant breeding, traits that exhibited good GCV, H 2 and GAM would be useful as a base for selection; hence days to heading, days to maturity, grain yield and biomass yield were identified as the major contributors.Grain yield had positive and highly significant correlation with biomass yield and harvest index, and also significantly correlated with plant height, thousand kernel weight, hectoliter weight and starch content both at genotypic and phenotypic level.This suggested that, grain yield potential can be effectively improved by obtaining maximum expression of these characters.However, grain yield had negative and significant correlation with grain protein content, and protein content exerted negative direct  effect.This implies simultaneous improvement of these two characters is difficult, thus care should be given during selection of these two traits.The highest positive direct effect on grain yield was exerted by biomass yield followed by harvest index both.Therefore, selection for high mean values of biomass yield and harvest index could be considered as the simultaneous selection of genotypes for high gain yield.Generally, it is recommended to further evaluate high yielding genotypes with high grain protein content and early maturing once more at similar agro-ecologies to develop varieties.Beside this, genetic information is limited for grain quality characteristics in bread wheat genotypes in the country (Ethiopia).Hence, due attention should be given to grain quality and yield performance of bread wheat genotypes to exploit genetic potential of the crop via selection or hybridization.

Table 1 .
Genotypes used in the study.

Table 2 .
Mean squares from analysis of variance for the 17 characters of 49 bread wheat genotypes.

Table 3 .
Phenotypic and genotypic variances and coefficients of variations, heritability in broad sense and genetic advance for 17 characters of 49 bread wheat genotypes

Table 4 .
Estimation of genotypic (above diagonal) and phenotypic (below diagonal) correlation coefficient for 17 morphological and quality traits in 49 bread wheat advanced lines.

Table 5 .
Estimates of direct (bold and diagonal) and indirect effect (off diagonal) of different traits on grain yield at genotypic level in 49 bread wheat genotypes at Laelay-Maichew in 2016.