Heterosis and character association of mid altitude adapted quality protein maize (Zea mays L.) hybrids at Bako, Western Ethiopia

Maize (Zea mays L.) is the basic staple cereal grain and currently produced on above two million hectares and the 2nd most widely grown crop in Ethiopia and stands fi rst in terms of production (9.5 million ton) and productivity 4 ton per ha. It is produced by about nine million farmers indicating its popularity and importance to the livelihoods of more than 70% of the farming households in Ethiopia [1].

population heavily dependent on maize run the risk of incurring retarded growth and health problem associated with protein defi ciency if maize protein is not complemented in the diet by other protein sources which are high in these amino acids, such as meat, legumes and dairy products [5,6].
Improving the quality of maize protein for human consumption by increasing its lysine and tryptophan through decreasing the zeins protein fraction content has therefore been a long-term goal of several maize breeding programs [7,8]. The bio-fortifi cation of conventional maize by doubling the amount of lysine and tryptophan through the identifi cation of a mutant genotype, popularly known as opaque-2 (o2), with reduced zeins protein fraction and improved agronomic performance gave rise to newly commercialized maize types called Quality Protein Maize (QPM) [9].
In most maize hybrid breeding programs, the main objective is to develop improved inbred lines that can form best hybrids upon crossing [10]. Sustainable production of QPM in Ethiopia and elsewhere is dependent not only on its nutrition benefi t but also on how much can be harvested per unit area of land.
Unless the productivity of QPM is comparable or better than the conventional maize varieties currently in use, farmers may be reluctant to produce it. This requires a rigorous inbred line development and evaluation work to identify potential parental lines for hybrid variety development.
The concept of heterosis is practically exploited to develop hybrid varieties. Heterosis may be defi ned as the increase in size, vigor, fertility, and overall productivity of a hybrid plant, over the mid parent value (average performance of the two parents) and over the performance of best parent. It is occurred when two inbred lines of out bred species are crossed, as much as when crosses are made between pure lines [11].
Although yield is usually the primary trait of interest, maturity, stand-ability, grain quality, stem quality, and resistance to major diseases and insects are all corollary traits that the maize breeder must consider for eventual usefulness of genotypes evaluated for yield [12].
Correlation is the degree to which two or more variables are related and change together [13]. Usually more than one trait is measured on progenies evaluated either for a specifi c trait in cyclical selection programs or in applied breeding programs that require a combination of traits to satisfy growers. In genetics, there are two main causes of correlation between characters, genetic and environmental. The genetic cause of correlation is chiefl y pleiotropy, though linkage is a cause of transient correlation, particularly in population derived from crosses between divergent strains [14]. It has established in classical genetics that many genes have manifold effects; i.e., some genes seem to affect traits that are unrelated. Genes that have manifold effects are pleiotropic, i.e., the same gene affects different traits in a complementary way. The existence of pleiotropic effects of genes in different classical genetic studies showed the presence of pleiotropy in different quantitatively inherited traits. Then it is possible that selection may be exerted on secondary traits that have greater heritability than the primary trait. Indirect selection will be effective if the heritability of the secondary trait is greater than that of the primary trait and the genetic correlation between them is substantial [12]. In maize, both genetic and environmental correlations have been extensively studied by various researchers and their importance with respect to a particular trait has been well documented [15,16].
Correlation coeffi cients do not give a complete picture of the causal basis of association and selection based on correlation coeffi cients without taking into consideration the interaction between the component traits could be misleading. Therefore, to design appropriate breeding strategies for improvement in yield through selection, it would be desirable to conduct both correlation and path coeffi cient analysis [17]. Path coeffi cient can be defi ned as a ratio of the standard deviation of the effect due to a given cause to the total standard deviation of the effect [18].
Therefore, this study is initiated to support the quest for better mid-altitude adapted QPM hybrid varieties in Ethiopia with the following specifi c objectives: to estimate the magnitude of heterosis in crosses derived from mid altitude QPM inbred and the association of traits with grain yield.

Description of experimental site
The experiment was conducted at Bako National Maize Research Center (BNMRC), which is located in Western Ethiopia. Bako Maize Research Center lies between 9 0 6' North latitude and 37 0 09' east longitude at an altitude of 1650 meters above sea level (m.a.s.l.) in the sub-humid agroecology of Ethiopia.

Experimental materials
Ten white-grained Quality Protein Maize (QPM) inbred lines obtained from BNMRC were crossed using diallel mating design during the main cropping season of 2014 and forty-fi ve single cross hybrids were generated. The parental inbred lines were selected based on their tryptophan and lysine content and per se performance history for grain yield and yield related traits. The inbred lines which contain good level of essential amino acids, lysine (4% in whole grain) and tryptophan (>0.8% in whole grain) were selected [19]. The parental inbred lines and the resulting hybrids (45) were organized into two separate sets of trials and tested in adjacent blocks (hybrids and inbred lines trials) at Bako trials evaluating site to avoid the unbalanced competition between the hybrids and inbred lines.

Experimental design
The 45 F 1 hybrids and the 10 inbred lines were planted following experimental design 9 x 5 alpha-lattice (0.1) for the hybrid trial Patterson and Williams, [19] and a randomized complete block design (RCBD) for the inbred line trial each with three replications. Each entry (the hybrids and parental inbred lines) was planted in a one-row plot of 5.1 m length and 0.75 m between rows and 0.3 m distance between plants in a row.

Data analysis
Mid Parent Heterosis (MPH) and Better Parent Heterosis (BPH) in percent were calculated for those parameters that showed signifi cant differences among F 1 hybrids and parental lines following the method suggested by [14]. The estimate of mid and better parent heterosis, was done only when both parents and crosses had showed signifi cance difference for the respective traits.

Association of characters
Genotypic and phenotypic correlation coeffi cients were calculated according to Al-Jibouri, et al. [22], from the analysis of variance and covariance as follow:  where  p12 is the phenotypic covariance between the two traits,  2 p1 is the phenotypic variance of the fi rst trait and  2 p2 is phenotypic variance of the second triat,  2 g12 is the genotypic covariance between the two traits,  2 g1 is the genotypic variance of the fi rst trait and  2 g2 is the genotypic variance of the second traits The phenotypic correlation coeffi cients were tested for traits of signifi cance with 'r' table for sample correlation coeffi cients at n-2 degree of freedom, as suggested by [13]; while the genotypic correlation coeffi cients were tested for their signifi cance using the formula. (1 ) 2 The 't' value, calculated using the above formula, was compared with 't' tabulated at (g-2) degree of freedom at 1% and 5% levels of signifi cance; where, r gxy is the genotypic correlation between x and y traits; g = number of genotypes, h 2 x and h 2 y are heritability for traits x and y, respectively.

Path coeffi cient
A path coeffi cient analysis was computed according to [17].
A path coeffi cient analysis is simply a standardized partial regression coeffi cient. The general formula used was: Where, rij = mutual association between the independent character (i) and dependent character (j) as measured by the correlation coeffi cients; p ij = components of direct effects of the independent character (j) as measured by the path-coeffi cients, and ∑ r ik p kj = summation of the components of indirect effects of a given independent character (i) via all other independent characters (k).
The residual effect was computed as: Where, R is residual, p ij is direct effect, and r ij is the correlation coeffi cients.

Analysis of variances
Analysis of variance conducted for both hybrid and inbred lines trials showed the existence of signifi cant differences among genotypes for all traits Table 1 below.
The signifi cant genotypic mean squares observed for most traits in both sets of trials indicated the existence of appreciable level of differences in the performances of hybrids and inbred lines for those traits. This indicates the possibility of making selection for further improvement of both sets of genotypes.  ED. In line with the present fi ndings, Berhanu [28], reported positive and signifi cant mid and better parent heterosis for ED in most F 1 hybrids studied. Habtamu, et al. [23], reported positive and signifi cant mid parent and better parent heterosis ranging from positive to negative for EL. Among 80 F 1 crosses evaluated, Gudeta [29], reported that more than 61% of the crosses had positive and signifi cant heterosis over the better parent while more than 98% of the crosses showed positive and signifi cant heterosis over the mid parent for EL. He also reported that most of the F 1 crosses had positive and signifi cant better parent and mid parent heterosis. The heterosis observed for EL and ED could be exploited in mid-altitude quality protein maize breeding program to develop desirable genotypes.
All the F 1 crosses showed highly signifi cant (P<0.01) and positive mid and better parent heterosis for number of kernels row -1 (NKPR). Mid parent heterosis were ranged from 12.39% (L8×L7) to 66.4% (L9×L1), whereas better parent heterosis ranged from 1.08% (L7×L5) to 28.26% (L9×L1). The results of present study corroborate with the fi ndings of Jehan, et al. [30], who observed high heterosis for NKPR in diallel crosses of maize inbred lines. Bayisa, et al. [31], also reported that 98% of the crosses they evaluated showed positive mid parent heterosis while, 65% of the same crosses had positive better parent heterosis for this trait. In contrast to the current study, Habtamu, et al. [23], reported low mid and better parent heterosis values for late maturity group of Maize inbred lines.
The disparity among the results of these studies might be attributed to the differences in the type and maturity group of the materials used. Values of mid and better parent heterosis for all traits are presented in Table 2

Association of characters
Association of yield and other attributes assumed special importance as basis for selection of desired strain. Genetic correlation between different characters can be often also because of either.

Genotypic and phenotypic correlation
The values of estimated genotypic and phenotypic correlation coeffi cient between pair of characters in all possible combination are presented in Table 3. It was found that the genotypic correlation coeffi cients were higher than the corresponding phenotypic correlation coeffi cient for all traits in similar direction. Similar with the present study, Assaduzzaman [34], reported the genotypic correlation coeffi cients were higher than their corresponding phenotypic correlation for all traits studied on fourteen Lablab genotypes.  In disagreement of the present fi ndings, Aminu and Izge [42], reported AD, EH and PH had exhibited negative correlation with GY and suggested that these traits were not closely associated and therefore, they may not be jointly selected. The difference results were found due to the genotypes was evaluated under drought condition. On the other hand, Berhanu [28], Bello, et al. [43] and Kinfe, et al. [44], reported that, GY had signifi cant and positive phenotypic correlation with EH, PH, EL, EPP, NKPR and ADin agreement with this study and proven the existence of direct association between the traits.
Other traits i.e., plant Aspect (PA), Ear Aspect (EA) and Anthesis Silking Interval (ASI) had strongly signifi cant negative association with grain yield at genotypic and phenotypic level, except ASI had moderately signifi cant at phenotypic level. Where, ASI, Ear Rot (ER) and Phaeosphaeria Leaf Spot (PLS) had signifi cant negative association at genotypic and phenotypic level, and Turcicum Leaf Blight (TLB) had at genotypic level. This revealed that, by decreasing these attributes, could consistently increase grain yield. The selection made to improve yield of maize genotype may be useful through decreasing these traits. In line with the current study, Hadji [45], observed GY had exhibited signifi cant and negative association with number of diseased ears. In agreement with this result, Kinfe, et al. [44], reported that, GY had signifi cant and negative association with ASI. Other research fi ndings reported by Aminu and Izge [42], exhibited ASI; bad Husk Cover (HC) and EPP had showed signifi cant negative correlation with GY at genotypic level. This showed that, these genotypes had short days to ASI, husked cobs and a smaller number of EPP had potential to give high grain yield.
However, non-signifi cant correlations were observed between GY and other traits due to masking effects of environment. This is indicating that selection for increase the level of these traits may not bring signifi cant change in GY.

Path coeffi cient
Phenotypic and genotypic path coeffi cient analysis is a proved effective means of separating direct and indirect effect of associated traits on yield. The analysis using grain yield as a dependent variable was conducted for the traits that exhibited signifi cant genotypic and phenotypic association with yield. The phenotypic and genotypic direct (bold) and indirect effects of twelve and eleven traits on grain yield were presented in Tables 4,5 below, respectively.
Days to Anthesis (AD) had a negative direct effect on Grain Yield (GY) at phenotypic level. The correlation coeffi cient between the two traits was positive and signifi cant (P <0.05). Moreover, the negative indirect contribution of AD to GY was through number of Ears Per Plant (EPP) at phenotypic level. Since correlation is positive, but the direct effect is negative, the indirect effects seem to be cause for correlation. This result agrees with some earlier fi ndings, Saleem [46], reported AD had negative direct on grain yield by the study on ten S 1 families evaluated under irrigated and drought condition. Therefore, in such situations, the indirect causal factors are to be considered simultaneously for selection. At both genotypic and phenotypic level, Anthesis Silking Interval (ASI) had showed highly signifi cant negative correlation and negative direct effect on grain yield. Ear Height (EH) had highly signifi cant positive correlation and direct effect on GY at phenotypic level; while, the genotypic correlation is positive and statistically non-signifi cant. In contrast to this fi nding, Muhammad et al. [15] reported that, EH had negative direct effect on GY by the study on eight local hybrids Maize. The different result was obtained may be due to the use of different source materials for the study. In agreement with the current fi nding, Hadji [45], observed, EH had exerted positive direct effects on GY at phenotypic level.
Other works agreed with this fi nding reported by Munawar, et al. [35], showed that the number of kernels row -1 (NKPR) had positive direct effect followed by EH for seven exotic hybrids maize sourced from different seed companies in Pakistan.
Similarly, Sreckov, et al. [47], they reported EH had positive direct effect on GY, however, it was non-signifi cant.  genotypes for resistant to CLR could be effective by considering indirect causal factors at the same time.
The other maize foliar disease was PLS which had signifi cant negative genotypic and phenotypic correlation coeffi cient with GY. In addition, the magnitude of the direct effect was negative at both genotypic and phenotypic level. The correlation coeffi cient at both genotypic and phenotypic level showed that, the genotypes with minimum disease reaction score could have a potential to boost the yield. Again, the selection for disease resistant high yielding hybrids should consider the indirect causal factors which could enhance yield and reduce disease development.
The other bad character encountered GY was Ear Rot disease (ER). It had negative and highly signifi cant (P <0.01) correlation at phenotypic level and signifi cant (P <0.05) correlation at genotypic level. At both genotypic and phenotypic level, it had negative direct effect on GY. Like other disease parameter, the selection of genotypes with a smaller number of ears susceptible to ER could help to get the varieties giving high yield and vice-versa. Therefore, the selection for ER free hybrids will be valuable if it considered other indirect causal factors in addition to the disease.
The residual effect estimation of 69.51% indicated that the causal variables explained only about 30.49% of the variability in grain yield and the remaining 69.51% of variability stays unexplored at phenotypic level. On the other hand, the residual effect of 58.84% exhibited that the fundamental variables elucidated only about 41.16% of the variability in grain yield and about 58.84% of the variability remain uninvestigated at genotypic level of path coeffi cient analysis. In contrast with the current study, Hadji [45], reported small residual effect 44% at phenotypic and 11% very small genotypic level for QPM inbred lines evaluated at the same location with the current study. Similary, Adesoji, et al. [41] and Kinfe, et al. [44], found small residual effects as compared with the current fi nding.
The reason seems to be very low variability in the present study were due to non-signifi cant correlations coeffi cient of the remaining traits with the causal factor, GY at both genotypic and phenotypic level. Besides, some other factors which have not been considered here need to be included in this analysis to account fully for the remaining variation in grain yield. It means that, the high value obtained in residual effects indicated that other factors and variables not considered in this study were of high effect on grain yield. Similar with this work, Oad, et al. [50] found the maximum (79%) residual effect for thirty varieties and advanced lines of Rice belonged to early to medium maturity collected from Philippines. In support of the current fi ndings, Yucel and Anlarsal [51] reported 78.7 percent of residual effect for twenty-two selected F 4 Chickpea genotypes obtained from ICARDA. However, it was not reported for maize, Abebe [52,53], found maximum residual effects (72.48 percent) at phenotypic level for Ethiopian Mustard.

Summary and conclusion
In this study, almost all crosses were showed positive and highly signifi cant mid and better parent heterosis for all traits assessed. In both cases, the same cross, L4×L2 and L5×L3 In general, it can be concluded that the research results suggested the importance of continuous and extensive research on quality protein maize best fi t to mid altitudes of the country to generate information that can be used to design breeding strategy.