Detection of Non-Additive Gene Action within Elite Maize Populations Evaluated in Contrasting Environments under Rainforest Ecology in Nigeria

Proper understanding of the mode of gene action in open-pollinated varieties (OPVs) maize parents helps breeder in the choice of appropriate breeding method to improve its genetic value. The objectives of the present study were to determine gene action controlling grain yield and other agronomic traits of late maturing elite OPVs and classify the varieties into heterotic groups. Ninety-one hybrids generated from 14 elite OPV parents using diallel mating design were evaluated with their parents plus three commercial checks under marginal rainfall, drought, and optimal environments in Nigeria from 2017 to 2018. The experiment was laid out in a 9 × 12 alpha lattice with three replications in each environment. Results showed that there were significant mean squares for grain yield and most agronomic traits. Significant general combining ability (GCA) and specific combining ability (SCA) mean squares for some of the traits indicated that additive and non-additive genetic gene actions were important in the inheritance of yield and those traits among this set of genotypes. However, non-additive genetic effects were more important than additive for grain yield and other agronomic traits in all research environments. Only TZL Comp-3 C3 DT had significant and positive GCA effects (0.336*) while three hybrids had significant and positive SCA for grain yield across research environments. Because of the preponderance of non-additive gene action over additive gene action, distinct heterotic groups could not be identified but four tester groups were identified by HSGCA (Heterotic grouping based on Specific and General Combining Ability) and three groups by HGCAMT (Heterotic grouping based on the GCA of Multiple Traits). Results of grouping were not related to the endosperm colour but grouping based on HGCAMT was related to the Open Access Received: 09 July 2020 Accepted: 01 December 2020 Published: 28 January 2021 Copyright © 2021 by the author(s). Licensee Hapres, London, United Kingdom. This is an open access article distributed under the terms and conditions of Creative Commons Attribution 4.0 International License. Crop Breeding, Genetics and Genomics 2 of 24 pedigree of the varieties. In conclusion, this study has demonstrated preponderance of non-additive gene action over the additive gene action for all measured traits. The presence of the non-additive gene action present in the studied materials can enhance identification of outstanding varietal hybrids and population testers that can serve as base genetic materials for future maize improvement through reciprocal recurrent selection program in SSA.


INTRODUCTION
Maize is a major food security crop that supports millions of people in sub-Saharan Africa (SSA) and the rest of the developing world. When compared to developed countries, the low maize yield in SSA (1.5-2.0 t·ha −1 ) is principally attributed to production constraints, comprising of several abiotic stress factors and low adaptation of exotic germplasm to target environments in the major maize production agro-ecological regions of the SSA [1,2]. Several maize breeding programs use openpollinated varieties (OPVs) as a source of genetic variability and favourable alleles of target traits. Though there might be cases of low yield, OPVs may be economically viable and sustainable, reveal phenotypic plasticity, and present wide adaptability to growing environments [3]. The conventional maize hybrid seed is relatively too expensive for a smallholder rural farmer in SSA, and requires more input for its production. Thus, considering elite OPVs as interim solution to boost maize productivity in rural areas of SSA is justifiable. Semagn et al. [4] pointed out that OPVs continue to occupy a large proportion of the maize production area in SSA because they represent the quickest and the easiest way to get improved genetic materials to resource-poor farmers at an affordable price. Improvement in yield and yield components of OPVs may be achieved through crosses, which enhances the exploitation of the intervarietal heterotic effects, allelic complementarity, as well as gene actions and effects [5]. Even though farmers in this region are now been discouraged from cultivating OPVs because of their poor yield, studies have shown that hybrids formed from them exhibit outstanding heterotic potential, in terms of productivity across stressed and non-stressed environments when the heterotic patterns among the varieties are exploited [6][7][8].
Breeders make thousands of crosses and evaluate grain yield in resulting F1 plants in replicated field experiments of lines from known or unknown sources. By classifying these lines into heterotic groups, the development and evaluation of crosses that should be discarded would be avoided, thereby, allowing the exploitation of maximum heterosis by crossing lines belonging to different heterotic clusters [9]. Lines extracted Crop Breed Genet Genom. 2021;3(1):e210003. https://doi.org/10.20900/cbgg20210003 Crop Breeding, Genetics and Genomics 3 of 24 from complementary populations developed from the parents of the opposite heterotic clusters usually show high combining ability with lines from the complementary population. Thus, the success of a hybrid program partly depends on the availability of information regarding the heterotic groupings of the parental lines. According to Badu-Apraku and Fakorede [10], grouping of maize inbred lines into appropriate heterotic groups defines the potential usefulness of such inbreds in a breeding program since it allows a proper understanding of the genetic relationships that exist within the inbred lines and enhances their efficient utilization in the development of hybrids, synthetic varieties, and subsequent formation of heterotic populations.
Several researchers have used the SCA effects of grain yield to classify maize genotypes into heterotic groups [11][12][13][14]. However, SCA effects for grain yield have been found to be influenced often by the interaction between two maize genotypes and the environment. Most times, this has led to the classification of the same genotypes into different heterotic groups in different studies [15,16]. Fan et al. [15] proposed the use of heterotic grouping based on Specific and General Combining Ability (HSGCA) method by combining both SCA and GCA effects of grain yield as a more appropriate method for assigning genotypes into heterotic groups. However, Badu-Apraku et al. [16] indicated that heterotic grouping of inbred lines based on one trait (grain yield) poses a challenge due to the complexity of this trait. Grain yield is controlled by several genes (polygenic), influenced by other traits and has low heritability under stress conditions. Bolaños and Edmeades [17] reported that selection for grain yield under drought conditions is inefficient due to the decline in the estimate of heritability of grain yield under environmental stress. Hence, heterotic grouping based on the GCA of multiple traits (HGCAMT) was proposed [16]. This method is based on measured multiple traits that are related to grain yield of genotypes with significant GCA effects across environments. Classification based on the GCA effects of multiple traits should be a better, realistic and more predictable approach for heterotic grouping of genotypes since GCA measures the additive gene effects of each trait. Badu-Apraku et al. [16,18] have successfully used this method to group early and extra-early inbred lines into heterotic clusters.
There has been a research gap in the improvement and utilization of elite OP maize varieties as an alternative and sustainable source of maize seed for smallholder farmers in SSA. With the dynamism associated with farmers' environment, worsened by climate change, the availability of maize varieties with stable yields across contrasting environments would be a welcome development in this part of the world.
Hence, the objectives of the present study were to determine gene action controlling grain yield and other agronomic traits in late maturing elite OPVs and use the two heterotic grouping methods (HSGCA and HGCAMT) to classify the varieties into heterotic groups under rainforest agro-ecological conditions of Nigeria.

Field Evaluations of Crosses and Stress Management
The 14 parental varieties, their 91variety hybrids plus 3 commercial checks were evaluated in two locations under each of marginal rainfall, drought and optimal growing conditions in 2018 ( Table 2). The checks were made up of 2 improved OPVs sourced from IITA and a local variety commonly grown by small-scale farmers in the locations. The growing conditions, which constituted six environments, were based on the time of planting and the total amount of rainfall. Under the marginal rainfall condition, the trials were set out at the onset of rainfall when the rain frequency is erratic and soil moisture is sub-optimal for maize cultivation. Under the optimal growing condition, the trials were established during the main planting season of maize with optimum amount of rainfall. Drought was achieved by planting towards the end of the rainy season (precisely third week in September), when flowering of the genetic materials was targeted to coincide with drought spell. The National Root

Crops
Research Institute agrometeorological unit (https://nrcri.gov.ng/index.php/agro-meteorology/) provided meteorological data for the location at Umudike, while the Micrometeorology Unit, Department of Physics, OAU provided that of the Ile-Ife location, which has its weather station at the experimental site. The experiment was laid out in a randomized incomplete block design (9 × 12 alpha lattice) with 3 replications in each environment. Experimental units consisted of 5 m two-row plots, with a spacing of 0.75 m. The intra row spacing was 0.50 m in all trials. Three seeds were planted, and the seedlings were later thinned down to two per hill at about 2 weeks after emergence to have a final plant population density of about 53,333 plants ha −1 . A common 2-row border was used at the beginning and end of each block to remove bias that may be created as a result of undue advantage to entries that fell at the borders or extremes of the blocks.

Field Measurements
Data were recorded under all the growing conditions on days to 50% silking (DS) and days to anthesis (DA) as the number of days from planting to when 50% of plants in a plot had emerged silks and had shed pollen, respectively. The anthesis-silking interval (ASI) was calculated as the difference between DS and DA. Ear (EHT) and plant heights (PHT) were measured as the distance from the base of the plant to the height of the node bearing the upper ear and the first tassel branch, respectively. Stalk lodging (the percentage of plants broken at or below the highest ear node) and root lodging (the percentage of plants leaning more than 30° from the vertical) were also recorded. Plant aspect (PASP) was an assessment of overall plant architecture and appeal (plant and ear heights, uniformity of plant height, cob size, disease and insect damage and lodging) and was recorded on a scale of 1 to 5 where 1 = excellent plant type and 5 = poor plant type. Ear aspect (EASP) was based on freedom from disease and insect damage, ear size, uniformity of ears and will be recorded on a scale of 1 to 5, where 1 = clean, uniform, large, and well-filled ears and 5 = rotten, variable, small, and partially or poorly filled ears. Ear number per plant (EPP) was estimated by dividing the total number of ears harvested in a plot by the number of plants in that plot. Grain yield was estimated from the ear weight and converted to kg·ha −1 . A shelling percentage of 80% was assumed for all genotypes and the grain yield was adjusted to 15% moisture [18] using the formula: where γ = grain yield (kg·ha −1 ), ϵ = ear weight (kg·m −2 ), n = moisture at harvest, ᵠ = plot area (m 2 ), 85 = 100 − 15 (a factor for 15% moisture content adjustment).

Statistical Analysis
Separate analysis of variances (ANOVAs) were performed on the data collected across locations for each research condition (marginal, drought and optimal growing) with PROC GLM in SAS using a RANDOM statement with the TEST option [19]. Subsequently, combined ANOVA was performed across the test environments for all the data collected. Environments, replicates and block were treated as random factors while entries were considered as fixed effects. The least significant difference (LSD) was used in separating means. The statistical model used for the ANOVA is as follows: where Yijkg is the observed measurement for the gth genotype grown in the environment i, in the block k, in replicate j; u is the grand mean; Ei is the main effect of environment; Rj(i) is the effect of replicate nested within environment; Bk(ij) is the effect of block nested within replicate j by environment i; Gg is the effect of genotypes; GEgi is the interaction effect between genotype and environment, and εijkg the error term. The combining ability analysis was carried out based on diallel Griffing's method 4 approach The GCA and SCA effects of the parents and the crosses, respectively, as well as their mean squares in each and across the environments were estimated following Griffing's method 4 model I [20]. Data for the parental varieties were excluded from the analysis due to poor and irregular germination of the seeds of the parental varieties during evaluation. The DIALLEL-SAS program developed by Zhang et al. [21] and adapted to SAS software version 9.4 was utilized. The t-test was used in testing both the GCA and SCA effects for significance. Their standard errors were estimated as the square root of the GCA and SCA variances [20].
The relative importance of GCA and SCA was investigated using both ratio of sum of squares of GCA to SCA and Baker's approach [22] using the following formula: where σ 2 gca = GCA variance components and σ 2 sca = SCA variance component. The closer the ratio is to one (1), the greater the predictability of a specific hybrid's performance based on GCA alone. The proportions of the additive and non-additive gene actions were estimated from the GCA and SCA effects respectively. Similarly, proportion of GCA to SCA was computed on the ratio of sum of squares of GCA to the genotypic sum of squares by dividing sum of squares of GCA by genotypic sum of squares. As proposed by Fan et al. [15], the HSGCA method was used to assign the OPVs into heterotic groups using GCA for grain yield of the parents and SCA for grain yield of their crosses, thus, HSGCA = cross mean (Xij) -tester mean (Xi) = GCA + SCA (4) where Xij is the mean yield of the cross between the ith tester and the jth parent, Xi is the mean yield of the ith tester across jth parents. The HSGCA estimates were then subjected to Ward's minimum variance cluster analysis using the JMP software, version 14 (SAS Institute Inc., Cary, NC, USA).
The statistical model used by the HGCAMT method to assign the parents into heterotic groups is represented as: Where Y is HGCAMT, the genetic value measuring the relationship among Heterotic grouping by the HGCAMT was carried out by standardizing the GCA effects (mean of zero and standard deviation of 1) of 10 considered traits. The standardization was done to minimize the effects of different scales of the traits. The standardized GCA effects were subsequently subjected to Ward's minimum variance cluster analysis using the software JMP version 14 (SAS Institute Inc., Cary, NC, USA). Dendrograms were subsequently constructed for the groupings based on HSGCA and HGCAMT.

Gene Action Controlling Inheritance of Yield and Other Agronomic Traits in Late Maturing Elite Maize Varieties
Results from the combined ANOVAs of the genotypes evaluated across the test environments indicated significant entry (G) and environment (E) effects for all traits except G for flowering traits and ear rot ( Table 3). The entry mean squares, partitioned into its components, showed that GCA mean squares were significant for grain yield, emergence percentage (EMERG), DA and PHT while SCA was significant for all the traits except ear rot (EROT) across the test environments. The GCA × E interaction effect was shown to be significant for EASP, EMERG, ASI, and EROT. For all the traits measured across the test environments, no significant SCA × E interaction effects were recorded ( Table 3).
The observed relative importance of GCA over SCA across the test environments, as given by the Baker ratio, was low for all the traits because the values are closer to zero than 1 (   Figure 1). The proportions of the non-additive genetic variance was greater than the additive genetic components in all environments for grain yield and other measured traits (Figure 1) Positive and significant GCA effects for grain yield and emergence were observed for TZC3. However, its GCA effect for DA was negative and significant. Positive GCA effects were further observed for STRY2, TZLC1, DTSY14, DTSY2 and WDTS2 across environments with respect to grain yield ( Table 4). The parents with negative GCA effects for PASP and EASP across the environments are STRY2, TZC3, WDTS2, and PVA2. Parents TZC3 and DTSY14 had significant and positive GCA effects for EMERG while PVA4 had a highly significant but negative GCA effect for EMERG. For DA, TZC3, TZLC1, and TZC4C2 had significant and negative GCA effects while STRY2 and PVA4 had significant but positive GCA effects. For PHT, all the parents had significant either positive or negative GCA effects except TZC3, DTSY14, PVA7 and WDTS1 whose effects were not significant (Table 4)

Groups across Environments
Since there was preponderance of SCA over GCA for grain yield and most traits under the study environments, it implies that distinct heterotic groups cannot be identified [23]. However, significant GCA and SCA mean squares from the results of the diallel crosses justify identification of tester groups among this set of parent. In the dendrograms constructed in   In summary, the tester groups by the different grouping methods are presented in Table 5. DTSY2 and DTSY14 were grouped together by HGCAMT; the case was different with HSGCA method. In addition, the HGCAMT method grouped WDTS1 and WDTS2 together across the test environments. However, they were grouped in different clusters by HSGCA method. Correlation among traits measured revealed that grain yield had highly significant correlations with plant aspect (r = −0.62 **), ear aspect (r = −0.51 **), ears per plant (r = 0.53 **), and plant height (r = −0.73 **). Similarly, yield was also significantly correlated (p < 0.05) with days to anthesis (r = −0.36 *), days to silk (r = −0.43 **), and number of ear rot (r = 0.47 **) but the strength of correlation coefficients were weaker (Table 6). Table 6. Correlation matrix between grain yield (GY) and other agronomic traits considered across environments.

DISCUSSION
The presence of significant means squares for grain yield and some agronomic traits indicated that there was substantial genetic variability among the genotypes upon which genetic improvement program can be based. When genotypic variability is partitioned into its components, the result revealed significant GCA and SCA mean squares for most important traits under each and across all test environments indicating that both additive and non-additive gene actions are important in the inheritance of most traits among the set of genotypes been studied. In addition, the result also implies that wide scope for improvement exists for some of the measured traits using hybridization, recurrent selection and backcrossing methods to develop varietal hybrids, synthetics as well as population.
It was observed that both additive and non-additive gene actions were involved in the inheritance of all the considered traits. However, there was the preponderance of SCA over GCA mean squares for grain yield and other traits measured in all the contrasting environments with the relative importance of GCA to SCA effects for grain yield and ear aspect increasing from stress to non-stress environments while most of the other traits decreased from stress to non-stress environments. In the same vein, the low Baker's ratios also indicate the predominance of non-additive gene action over additive gene action; hence, a low predictability of progenies performance from parents' GCA effects. The performance of progeny in this set of crosses was better in specific combinations and therefore could not be predicted for a wide range of crosses. The result obtained suggests that non-additive gene action prevailed over the additive counterpart for these traits [18] and that the major component accounting for the differences among the OPVs evaluated was the SCA. The result is in agreement with the report of Adewale et al. [24] and Badu-Apraku et al. [18] who also reported preponderance of non-additive gene action over the additive gene action among early maturing maize inbred lines.
The present results are also in agreement with the findings of other researchers [25][26][27][28] who reported the preponderance of non-additive gene action for the expression and inheritance of yield and other traits measured in maize. The results of this study is in partial agreement with the findings of Wegary et al. [29] who reported SCA effects to be more important under optimal conditions and GCA effects, more important under drought for grain yield. The variations in the results reported in the two studies may be attributed to the different backgrounds of the materials used and might have contained some genes with different modes of action [16]. This current research suggested that under marginal, drought, and optimal growing conditions, the gene actions controlling grain yield and other measured traits were similar in this set of elite OPVs. In the work of Badu-Apraku et al. [18], similar gene actions (additive in this case) were responsible for controlling grain yield and most measured traits in the early maturing QPM inbred lines used. However, since additive gene action is significant, it can be deduced that progress can be made from can serve as an initial gene pool for subsequent improvement and development of synthetic varieties and hybrids that are high yielding for the SSA region [33]. The positive and significant GCA effects (grain yield) observed for TZL Comp-3 C3 DT (TZC3) across research environments indicated that this variety possesses favourable alleles for grain yield and would contribute high yields to its progenies. Thus, it could serve as a tester for improving grain yield in a population. Similarly, the variety also possessed desirable GCA for emergence percentage and days to anthesis.
The combining ability estimate of a genotype across research environments is a measure of the performance and stability of that genotype in a hybrid combination or in a population development. Genotypes with outstanding GCA and SCA across research environments are suitable for hybrid and population development for the region of SSA [33]. Parent DTSY14 had significant negative GCA effects for ear aspect across the environments. This is an indication that this parent would contribute favourable alleles to ear aspect and indirectly to grain yield in their progenies since ear aspect is always closely associated with grain yield. Parents TZLC1, TZC3 and TZC4C2 showed significant and negative GCA effects for days to anthesis across research conditions. These results suggest that these OPVs will contribute favourable alleles to their progenies for earliness under contrasting environments. Mhike et al. [34] and Halilu et al. [35]  that the non-additive gene action was preponderant over the additive counterpart for grain yield and other traits; it is expedient to say that the grouping method that include SCA (non-additive effect) will be more efficient than the one that exclude the same effect in its classification. Thus, classification based on HGSCA is preferable in this study, which identified 4 groups. The grouping of the PVA varieties into separate clusters by the two methods indicates that there is wide variability among the provitamin A varieties and there is possibility of improving the agronomic performance of the provitamin A varieties.
In the grouping of early maturing QPM inbred lines using HSGCA and HGCAMT methods by Badu-Apraku et al. [18], HSGCA method also had one group more than HGCAMT method. The grouping was not related to the endosperm colour of the OPVs as the groups were consistently composed of OPVs from both endosperm colour types. Similar result was reported for 28 early maturing inbreds classified into heterotic groups based on combining ability by Akinwale et al. [33]. The classification of DTSY2 and DTSY14, WDTS1 and WDTS2 together in the same group across test environments by the HGCAMT method indicated that the grouping of the OPVs was based mainly on their pedigree and to a lesser degree on the reaction of the OPVs to the stress environments. This outcome is in line with the findings of several other authors [14,16,33].
Based on the analysis of relationship among traits, plant height, plant and ear aspects and ears per plant were identified as important secondary traits for indirect selection for grain yield across the studied conditions due to their significant correlation with it. This result is in agreement with findings of [40,41].

CONCLUSION
There was wide genetic variability among the varietal parents used for this study. Although, both additive and non-additive gene actions were significant in the control of grain yield and other agronomic traits, nonadditive gene action was preponderant over additive gene action for all traits. The OPV, TZC3 (TZL Comp -3 C3 DT) was identified as the best in terms of good general combining ability effects for grain yield and other traits across research environments. The favourable alleles from this parent should be harnessed for the development of high yielding and drought tolerant open pollinated varieties that can serve the rural maize farmers of the sub region in the face of climate change.
Four tester groups were identified among the 14 varieties from which population crosses could be made, which will serve as base population from where superior inbreds could be extracted.

DATA AVAILABILITY STATEMENT
Some of the dataset generated from this study are in the manuscript. The rest of the data is available from the authors upon request.

AUTHOR CONTRIBUTIONS
Richard Olutayo Akinwale conceptualized, laid out the field, supervised, collected data and guided in the data analysis and contribute to the writing of the paper.
Chinedu Emmanuel Eze carried out the fieldwork, field layout, data collection and analysis. He wrote the first draft of the article. Diakaridia Traore contributed to the supervision of the research work.
Abebe Menkir supplied the genetic materials used and contributed to the methodology and carried out the editing of the manuscript before submission.

CONFLICTS OF INTEREST
The authors declare no conflicting interest.