Stability Assessment of Single-Cross Maize Hybrids Using GGE-Biplot Analysis

Grain yield potential of new maize hybrid varieties across target environments contributes to the uptake of these varieties by farmers. Evaluation of single-cross hybrids developed from test crossing introgressed inbred lines bred for three distinct environments to elite tropical inbred line testers was carried out. The study’s objective was to assess grain yield stability and genotype adaptability of the single-cross hybrids across South African environments relative to adapted commercial hybrid checks. One hundred and twenty-two introgressed inbred lines developed using the pedigree breeding program were crossed to four tropical elite inbred line testers using line × tester mating design to obtain 488 experimental single cross hybrids. Subject to availability of adequate seed for evaluation, a panel of 444 experimental single-cross hybrids was evaluated using an augmented design in two experiments defined as Population A and B for the study’s convenience in South African environments. Data for grain yield (t/ha) performance for experimental single-cross hybrids and commercial check hybrids in Population A and B across environments and individual environments identified experimental single-cross hybrids that had significant comparable grain yield (t/ha) performance relative to best commercial check hybrid (PAN6Q445B) on the market. The selected experimental single-cross hybrids 225, 89, 246 and 43 (Population A) and 112 (Population B) also had a better average rank position for grain yield (t/ha) relative to best commercial check hybrid. These selected experimental single-cross hybrids had a grain yield (t/ha) advantage range of 0.9-6.7% for Population A and 7.3% for Population A and B, respectively, relative to the adapted commercial check hybrid. GGE biplot patterns for which won-where for Population A indicated that at Potchefstroom Research Station and Ukulinga Research Station experimental single-cross hybrids 127 and135 were the vertex (winning) hybrids. Cedera Research Station did not have a vertex hybrid for Population A. For Population B, experimental single-cross hybrids 112, 117 and 18 were the vertex hybrids at Cedera Research Station, Ukulinga Research Station and Potchefstroom Research Station, respectively. Experimental single-cross hybrid 257 was identified as ideal genotype for Population A, while experimental single-cross hybrid 121 in Population B was the ideal genotype. Ideal environments were also identified as Ukulinga Research Station for Population A, and Cedera Research Station for Population B. Average-environment coordination (AEC) view of the GGE biplot in Population A indicated that experimental single-cross hybrids 1 was highly stable across environments. In comparison, Population B experimental single-cross hybrid 161 was highly stable across environments. In conclusion, selected single-cross hybrids in the current study can also be advanced for further evaluation with a possibility for identifying high yielding and stable single-cross hybrids for variety registration and release in target environments in South Africa. jas.ccsenet.org Journal of Agricultural Science Vol. 13, No. 2; 2021 79


Introduction
In developing countries, particularly in Africa, maize (Zea mays L.) is a critical and strategic cereal crop grown across many of its regions. Its wide adaptability in target environments has rendered it a staple food crop across tropical, subtropical, and temperate regions of the world. In South Africa, a predominantly warm temperate environment, maize is the largest locally produced field crop with increasing food, feed, and industrial usage value for the population (Syngenta Foundation for Sustainable Agriculture, 2020). Maize is also regarded as a net earner of foreign currency for the South African economy. Therefore, South Africa maize production is a large and lucrative market for breeding programmes operating inside and outside South Africa. An indication that breeding programmes should ensure the release of stable maize hybrid varieties that perform well in the South African warm temperate environments, aiming for broad adoption by farmers.
In maize breeding, the primary objective is to develop hybrids with high yield potential and adaptability across target environments. According to Kuchanur et al. (2015), breeders should select top grain yielding genotypes associated with high grain yield stability. Mostafavi et al. (2011), andElias et al. (2016) report that targeting improved varieties to specific environments is difficult when genotype-by-environment interaction is present since yield is less predictable be interpreted based only on genotype and environment means. Genotype-by-environment interaction is here defined as the differential ranking of variety yields across target environments, resulting in the variable performance of hybrid varieties in selected target environments (Crossa et al., 2002;Jandong et al., 2011;Heidari et al., 2016). Thus, it complicates the utilization of hybrid maize varieties across target environments.
In this study, the emphasis is on identifying improved tropical introgressed maize inbred line hybrid combinations capable of maximizing maize production potential in South African warm temperate environments and farming systems, thus reducing crop failure or low incidences grain yields in unfavourable seasons. Breeding programmes have to develop improved maize varieties for the farmers with excellent agronomic performance relative to adapted commercial check varieties in the target environments. Recommendation of improved hybrid varieties in target environments requires these genotypes to be evaluated in several different but representative environments to identify consistently high-yielding and relatively stable genotypes and areas of specific adaptation (Balestre et al., 2009;Setimela et al., 2017).
A few methods have been applied in maize breeding programmes to evaluate cultivars' adaptability and stability in target environments. Two main approaches have been consistently used in several studies, namely: additive main effects and multiplicative interaction (AMMI) analysis (George & Lundy 2019;Gauch et al., 2008;Gauch, 2006;Duarte & Vencovsky, 1999); and a modification of the conventional AMMI analysis called genotype (G) and genotype-by-environment interaction (GE) (GGE-biplot) analysis (Yan & Tinker, 2006;Kaya et al., 2006;Yan et al., 2000). AMMI and GGE-biplot provide breeders with tools to measure maize hybrid varieties' response efficiently and accurately in multiple test environments (Yan et al., 2007). According to Balestre et al. (2009), AMMI analysis interprets the effects of genotypes and environments as an additive and GE interaction as a multiplicative principal component analysis. The GGE-biplot analysis groups the genotype effects, which are additive in the AMMI analysis, together with the GE interaction multiplicative effects and analyses these effects by principal components (Kaya et al., 2006). According to Yan andHunt (2001), andYan et al. (2007), GGE, biplot software is an excellent visual MET data analysis tool. Compared with conventional methods of the MET data analysis, the GGEbiplot approach has some advantages. The first advantage of the biplot is its graphical presentation of the MET data, which significantly enhances our ability to understand the data patterns. The second is that it is more interpretative. It facilitates pair-wise genotype comparisons. The third advantage of this method is that it enables the identification of possible mega-environments.
Genotype and genotype-by-environment interaction analysis were carried out in the current study on single-cross maize hybrid maize varieties to compare the grain yield potential of these genotypes across target environments relative to adapted commercial check entries. The comparison of grain yield potential of the maize genotypes at different environments or groups of environments in South African regions ensured the identification and recommendation of genotypes with higher grain yield potential in each target environment. As a breeder, the main objective is to breed for high grain yield potential. For that high grain yield potential to be highest or close to the highest, consistently in all locations within the geographical area for which variety will be released (Yan & Tinker, 2006). The study's objective was to assess grain yield stability and genotype adaptability of single-cross hybrids, including parents, developed for three distinct mega environments, using GGE biplot analysis across the South African warm temperate environments relative to adapted commercial hybrid checks.

Introgressed Inbred Lines Development
Introgressed inbred lines used to generated experimental single-cross hybrids evaluated in the current study were developed from a pedigree breeding program. A single common donor maize parental inbred line (08CED6_7_B) from South Africa was used to introgress genes from temperate germplasm into 12 elite tropical inbred lines from Zimbabwe through pedigree crosses in 2008 in South Africa. Tropical maize inbred lines used were representative of the major tropical heterotic groups, mainly N3 (derived from Salisbury white), SC (Southern Cross which was derived from an open-pollinated population grown by Mr South in Zimbabwe), and P (derived from the open-pollinated variety (OPV) Potchefstroom Pearl). The temperate maize population was one of the major temperate heterotic groups used in South Africa (TAB population). Hand crossings were made between the tropical and temperate populations to generate F 1 hybrid seed. Due to challenges in flowering synchronization (nicking) and seed availability, a total of eight populations were generated for advancement and selection at F 2 generation. Each population was independently advanced from F 3 -F 6 generation through selfing and selection of adapted segregants to produce 122 introgressed inbred lines.

Experimental Single-Cross Hybrids Development
Experimental single-cross hybrids used in the current study were generated from testcrossing 122 Introgressed inbred lines to four tropical elite inbred line testers T1, T2, T3, and T4 using line by tester mating design. The four tropical elite inbred line testers used represented maize germplasm from two tropical heterotic groups P and N. A total of 488 experimental single-cross hybrids were produced from the test crossing. Subject to availability of adequate seed for evaluation, a panel of 444 experimental single-cross hybrids were evaluated using an augmented experimental design. Due to the large number (444) of the experimental single-cross hybrids involved and for convenience of the study, the experimental single-cross hybrids were divided into two populations that were designated population A and B, with both populations related to heterotic groups P and N. Population A comprised 280 experimental single-cross hybrids including four commercial hybrid checks; temperate hybrids (PAN3Q740 and PAN6Q445B) and tropical hybrids (PAN67 and SC633) to give a total evaluating panel of 284 entries. Population B consisted of 164 experimental single-cross hybrids, including three commercial hybrid checks (PAN6611, PAN6Q445B, and SC633) to give a total evaluating panel of 167 entries. Commercial check hybrids used in both populations were single-cross hybrids that are predominantly used in the South African market.

Experimental Design and Trial Management
A total of five trials were planted in three locations in South Africa environments. Table 1 presents a summary of the locations. In population A, 284 entries (experimental single-cross hybrids and commercial hybrid checks) were randomly assigned into 20 blocks; in each block, 14 experimental single-cross hybrids and two repeating checks (PAN3Q740 and PAN67) were randomly assigned to each block. Due to limited seed, commercial check hybrid entries SC633, PAN6227, and PAN6Q445B were randomly assigned into blocks as non-repeating commercial checks. In population B, 162 entries (experimental single cross hybrids and commercial checks) were randomly assigned into 16 blocks; in each block, ten experimental single-cross hybrids were included with two repeating commercial checks (PAN6611 and PAN6Q445B). Due to limited seed, non-repeating commercial check SC633 was randomly assigned into the blocks. Population A was replicated over two sites, namely Ukulinga and Cedara Research Stations. In comparison, Population B was replicated over three locations: Ukulinga, Cedara, and Potchefstroom Research Station. An augmented experimental design was used to evaluate the trial (Lin & Poushinsky, 1983;Scott & Milliken, 1993;Spehar, 1994). Due to the limited availability of seed, all experiments across sites were each planted as single-row plots. At Ukulinga Research Station, each entry was planted to 5m length, spaced at 0.3 m in-row and 0.75 m between row spacing to achieve a total plant population density of at least 44 000 plants ha -1 . At Cedara Research Station, 5 m row-plots, in-row spacing 0.3, and row spacing of 0.9 m were used to achieve a plant stand of at least 37 000 plants ha -1 . While at Potchefstroom Research Station, 6.6 m length, spaced at 0.25 m in-row, and 1.5 m between row spacing were employed to attain a total plant population density of at least 26 000 plants ha -1 . Standard cultural management practices for growing maize were carried out at all the sites. Irrigation was only applied to achieve uniform establishment and to supplement rainfall as and when necessary. Fertilizer was applied at a rate of 120 kg Nitrogen (N), 33 kg Phosphorous (P), and 44 kg Potassium (K) at Cedara, Ukulinga, and Potchefstroom Research Stations.

Measurements
Data was collected at all the sites applying standard procedures used at International Maize and Wheat Improvement Centre (CIMMYT, 1985) for the following traits: days to anthesis and silking days were recorded when 50% of the plants were shedding pollen, and 50% of the plants had silks emerged, respectively; plant and ear height were measured before harvesting on five representative plants per plot; percentage stalk and root lodging was recorded as a percentage of plants per plot that had their stems broken and percentage of plants per plot which had their stems inclined at least 45 o , respectively; and the number of ears per plant-ear prolificacy (EPP) was calculated as the count of the number of ears plot as a fraction of the total number of plants in the plot. All plants were hand-harvested and shelled grain weight was measured. Grain weights were adjusted to 12.5% moisture content and 80% shelling percentage to calculate grain yield (t ha -1 ).

GGE Biplots
Genotype and genotype-environment interaction GGE-biplot analysis were carried out on yield data only. In future publications, we will report on the other traits measured. The GGE-biplot concept (Yan et al., 2000) was used to visualize the multi-environment trials (MET) data, as reported by Kaya et al. (2006). The GGE-biplot showed the first two principal components (PC1 and PC2) derived from subjecting environmental-centred yield data (yield evaluation due to GGE) to singular value decomposition (Yan et al., 2000). In the current study, genotype-focused scaling was used for visualizing genotypic comparison, with environment-focused scaling for environmental comparison using GGE-biplots (GenStat 14 edition, 2013). A mixed model for LSD analysis was also carried out for multi-treatment comparison using the Tukey-Kramer method (Yu, 2010) to compare experimental single-cross hybrids and commercial check hybrids grain yield (t/ha) performance at individual sites and across sites.

Grain Yield (t/ha) Performance for Top 12% High-Yielding Hybrids in Population A
Data from grain yield (t/ha) performance for top 12% top-yielding experimental single-cross hybrids and commercial check hybrids in Population A sorted according to the average rank for grain yield (t/ha) across sites and individual sites (Cedera Research Station, Potchefstroom Research Station and Ukulinga Research Centre environments) are presented in Table 2. Across the environments, data indicated that grain yield (t/ha) performance was highly significant with experimental single-cross hybrids and adapted commercial check hybrids showing significant similar grain yield (t/ha) performance. The average rank position across environments for grain yield (t/ha) showed that the top four (225, 89, 246, and 43) experimental single-cross hybrids had better grain yield (t/ha) rank position and higher grain yield (t/ha) equivalent to a range of 0.9 to 6.7% than the best adapted temperate environment commercial check hybrid (PAN6Q445B).
Individual environments, data for Cedera Research Station, Potchefstroom Research Station, and Ukulinga Research Centre environments grain yield (t/ha) were significant with experimental hybrids illustrating comparable performance relative to the adapted temperate environment commercial check hybrids (PAN6Q445B, PAN3Q740, and PAN6227). At Cedera Research Station environment, the average rank position for grain yield (t/ha) highlighted that eleven experimental single-cross hybrids (60,257,131,61,144,259,43,225,45, 1, and 92) had a better average rank position placement than the best adapted temperate commercial check hybrid (PAN6Q445B). These eleven experimental single-cross hybrids had a grain yield (t/ha) advantage that ranged from 0.4 to 12.1% relative to the best adapted temperate commercial check hybrid (PAN6Q445B). Similarly, at Ukulinga Research Centre environment, fourteen experimental single-cross hybrids (135, 257, 61, 225, 1, 138, 89, 259, 255, 245, 253, 263, 137, and 144) had high average range position for grain yield (t/ha) relative to best adapted commercial check hybrid (PAN6Q445B). A grain yield (t/ha) advantage of between 4.4 to 97.9% was noted over the best adapted commercial check (PAN6Q445B). Equally, the Potchefstroom Research Station environment exhibited a similar trend with three experimental single-cross hybrid entries (225, 89, and 246) exhibiting high average range position for grain yield (t/ha) relative to best adapted temperate commercial check hybrid (PAN6Q445B). The grain yield advantage over the best adapted commercial check hybrid (PAN6Q445B) ranges from 3.16 to 3.86%. The grain yield (t/ha) performance data for across environments and individual environments illustrate Genotype × Environment (G×E), as under various environments (Cedera Research Station, Potchefstroom Research Station, and Ukulinga Research Centre) not the same hybrid entries out yielded the best commercial check hybrids in terms of grain yield (t/ha) rank placement. This data is further supported by genotype main effects (G) and genotype × environment interaction effects model, known as GGE biplots Figures Note. Means with the same letter in the same column are not significantly different (P > 0.05), A1-Percentage grain yield (t/ha) advantage relative to best check entry.

GGE-Biplots
Genotype and genotype-by-environment interaction (GGE) biplots allow effective identification of the Genotype-by-Environment Interaction (GEI) pattern of the data. In this current study, biplots were plotted for entries in population A and B to allow visualization of, which-won-where patterns pattern for genotypes and environments, genotype-focused scaling for comparison of the genotypes with ideal genotype, environment-focused scaling for comparison of the environments relative to an ideal environment, and average environment coordination (AEC) views based on environment-focused scaling for the means performance and stability of genotypes. Subsequent GGE biplot analysis produced eight biplots for entries in Population A ( Figure  1 to 4) and Population B (Figures 5 to 8) to allow visualization of; which-won-where pattern for genotypes and environments, genotype-focused scaling for comparison of the genotypes with ideal genotype, environment-focused scaling for comparison of the environments relative to an ideal environment, and average environment coordination (AEC) views based on environment-focused scaling for the mean performance and      advantage over the best adapted commercial check hybrid (PAN6Q445B). Average grain yield (t/ha) performance across environments are good indicators of genotypic performance only in the absence of Genotype-by-Environment Interaction (Kaya et al., 2006;Balestre et al., 2009). Thus, indicating that these experimental single-cross hybrids have the potential to be advanced for further evaluation, possible registration, and release in target environments. These selected experimental single-cross hybrids also allow an opportunity to explore what makes these hybrids unique and whether they have shared desired economic traits, i.e., physiological maturity, drought tolerance, prolificacy, and good standability that is valuable in advancing this breeding program.

Grain Yield (t/ha) Performance for Top 14% High-Yielding Hybrids in Population B
Individual environments data for average rank position grain yield ( Similarly, at the Potchefstroom Research Station environment's top five experimental single-cross hybrid entries had a 9.2-29.3% grain yield advantage over the best-adapted commercial check hybrid (PAN6Q445B). This group of selected experimental single-cross hybrids for both Population A and B can be advanced for further testing as they can out-compete adapted commercial check hybrids. Thus, indicating a possibility for commercial registration and release if they are to maintain consistent performance is accompanied by desired economic traits (prolificacy, early physiological maturity, and good standability) for target environments. These experimental single-cross hybrids also provide an opportunity for the breeding program to explore if they have unique traits that can be utilized in advancing the breeding program.

GE-Biplots Patterns
Visualization of which-won-where pattern of multi-environmental trials data is essential for studying the possible existence of different mega environment (ME) in a region (Gauch, 2006;Kaya et al., 2006;Jalata et al., 2009;Alwala et al., 2010;Jandong et al., 2011;Shim et al., 2015). In this study, visualization of which-won-where pattern of multi-environment trials data for experimental single-cross hybrids and commercial check hybrids in Population A illustrated that there were six sectors and environments fell into three of them. Two of the environments, Potchefstroom Research Station and Ukulinga Research Station, had experimental single-cross hybrids 127 and 135 as the vertex hybrids. Most importantly, the vertex hybrids had higher (sometimes the highest) grain yield (t/ha) than the other hybrids (experimental single-cross hybrids and commercial check hybrids) in all environments that fell in these sectors. Similar results were also reported by (Crossa et al., 2002;Kaya et al., 2006). Cedera Research Station environment did not have a vertex hybrid, an indication that no hybrid was ideal for this environment. Population B data also had six sectors, and the environments fell into three of them. Cedera Research Station, Ukulinga Research Station, and Potchefstroom Research Station environments had experimental single-cross hybrids 112, 117, and 18 as the vertex hybrids. Thus, indicating that these were the winning hybrids for each respective environment for both Population A and B, with Ukulinga Research Station environment discriminating the genotypes more clearly as depicted by higher PC1 scores. GGE biplot analysis carried out on Population A, and B identified sites that best represent the target environment for these populations. Several similar studies (Balestre et al., 2009;Jalata, 2011;Kaya et al., 2006;Ndhlela, 2012;Yan et al., 2010) have been conducted across crops not only to identify high yielding cultivars but also to identify sites that best represent the target environments. Selected experimental single-cross hybrids for both Population A (127 and 135) and B (112, 117 and 18) in this current study can be advanced for further evaluation in the subsequent seasons with a possibility for variety registration and release their selected target environments.

Ideal Genotypes
According to Kaya et al. (2006), and Dehghani et al. (2009), yield potential and stability of genotypes are evaluated by an average environment coordination (AEC) method. In this method, an average environment is defined by the average PC1 and PC2 scores of all the environments. An ideal genotype should have the highest mean grain yield (t/ha) performance and stability across all the environments and may not exist but can be used as a reference for genotype evaluation (Yan and Tinker, 2006). In Population A, experimental single-cross hybrid 257 was close to ideal, while experimental single-cross hybrid 127 was the highly undesirable entry. In Population B, experimental single-cross hybrid 121 was defined as the ideal genotype, and in contrast, experimental single-cross hybrid 18 was defined as the highly undesirable entry. The ideal experimental single-cross hybrids in Population A (257) and B (121) can be used in future similar projects as a reference in selecting for maize genotypes that are defined as ideal genotypes.

Ideal Environment
An ideal testing environment should have the ability to discriminate genotypes in terms of the main genotypic effect during evaluation. This environment should have large PC1 scores and small PC2 scores, together with approach 100%, and may not exist in reality, but it can be used as a reference for genotype selection in multi-location trials (Kaya et al., 2006). In Population A and B, both environments were not ideal environments for these entries. However, they can be used to define the most favourable environment that can be used for evaluating high yield potential. In the current study, Ukulinga Research Station was defined as the ideal environment for Population A. In comparison, Cedera Research Station was defined as the ideal environment for Population B.

Mean Yield and Stability of the Genotypes
Average-environment coordination (AEC) view of the GGE biplot in Population A indicated that experimental hybrid entry 1 was highly stable across the three environments. While experimental single-cross hybrids 135, 281and 127 were highly unstable highly stable across the three environments. Population B, experimental single-cross hybrid 161, was highly stable across environments. In contrast, experimental single-cross hybrid 117 and 112 were highly unstable across the three environments.

Conclusion
Experimental single-cross hybrids 225, 89, 246, and 43 in Population A and 112 in Population B had better average rank position for grain yield (t/ha) performance across environments and grain yield (t/ha) grain yield (t/ha) advantage over the best adapted commercial check hybrid (PAN6Q445B) of 0.9-6.7% and 7.3% for Population A and B, respectively. These selected experimental single-cross hybrids can also be advanced for further evaluation with a possibility for identifying high yielding experimental single-cross hybrids for variety registration and release in target environments in South Africa if they even have desired economic traits. In terms of stability, experimental single-cross hybrids 1 and 161 in populations A and B, respectively, were defined as highly stable hybrids across environments. Experimental single-cross hybrids 257 (Population A) and 121 (Population B) were identified as the ideal hybrids.