Yield stability and agronomic performances of provitamin A maize (Zea mays L.) genotypes in South-East of DR Congo

1 Introduction

Maize (Zea mays L.) is one of the largest strategic crops in the Democratic Republic of the Congo in general and in the former Katanga Province in particular, where it ranks first among cereal crops [1]. Despite the large planted areas of this crop, its production remains low and cannot meet the needs of the population. This leads to the import of maize from southern African countries to meet the growing local demand [2]. The increase in its demand is linked to the direct use in human food of almost all of its production [3], placing maize cultivation in the second position after cassava among food crops in the former Katanga province. Hyman et al. [4] and Zuma et al. [5] have shown that the rate of malnutrition (hidden hunger) is high in areas where maize is the dominant diet. In these areas the white maize varieties which are commonly grown and consumed are deficient in essential nutrients such as provitamin A [6]. This would result in the existence of hidden hunger in the former Katanga province. In addition, this former province is a mining area known to have a high rate of malnutrition [7]. To alleviate this problem of malnutrition, the use of provitamin maize varieties has become an cost-effective solution [8], especially for the poor populations of the former Katanga province [9]. In this context, a series of provitamin A maize genotypes were developed by the BioDeV Laboratory in collaboration with the HarvesPlus program. However, like many crops, maize generally exhibits genotype × environment (G × E) interaction in terms of performance when grown in different environments [10,11], which generally complicates variety selection and/or recommendation. Genotype × environment (G × E) interaction is variation in the response of a genotype in different environments. A significant GEI results from changes in the magnitude of differences between genotypes in different environments. It is an important aspect in plant breeding programs because it constrains breeding progress in a given environment [12]. In plant breeding programs, cross-interaction is of particular interest because the best genotype in one environment may not be in another environment. Thus, understanding the interaction observed in multi-environment trials is very useful in plant breeding programs to identify high performing varieties with broad adaptability or specific adaptation [11].

Farmers grow crops in diverse environments and need stable genotypes that can withstand the effects of environmental changes. Because of its importance, plant breeders have undertaken research to establish stable genotypes and have defined a stable genotype as one that has the ability to have the same performance, regardless of any variation in the environment [12–14]. This concept of stability is useful for quality traits or disease resistance [15]. However, for this study only the biological concept will be used, which states that a genotype is considered stable if its variance between environments is small [15].

To assess the genotype × environment interaction and to recommend the productive and stable genotype in various environments, several stability analyses have been proposed among which are Eberhart and Russell joint regression method based on the mean performance, the regression coefficient and the deviation from linear regression [16] and the GGE biplot proposed by Yan et al. [17]. Since the development of the GGE biplot model, it has been widely used to study the stability and adaptability of various cereals such as sorghum [18], wheat [19], barley [20,21], rice [22] and maize [11,23–28] and it has proved to be one of the most frequently used procedures in plant breeding [17,29–32]. This method has the advantage of graphically displaying the interaction in a bidirectional dimension to facilitate the selection of stable and high-yielding genotypes [29,30,33]. The growing demand for maize in the South-East of the Democratic Republic of the Congo [3] increasingly requires yield optimization and this is where GGE biplot analysis and Eberhart and Russell joint regression method can be used to identify high-yielding varieties. Therefore, in this study, the biplot and parameters of Eberhart and Russell’s model were used to highlight the responses of maize genotypes evaluated at different locations. The aim of this study was to evaluate the newly developed provitamin A maize genotypes for stability of performance across diverse agro-ecologies in the South-East of the Democratic Republic of the Congo.

2 Materials and methods

3 Results

3.3 GGE biplot analysis

The results of the GGE biplot analyses for grain yields of the eight genotypes evaluated in six different environments in southeastern Democratic Republic of the Congo are presented in Figures 1–3. The GGE biplot distribution of genotype by environment interaction showed that CP1 and CP2 accounted for 64.27 and 26.58% of the total variation in yield, respectively, indicating that the two principal components adequately approximated the environment-centered multi-environment trials data.

Figure 1

Genotype, genotype by environment biplot showing the mean and stability of eight provitamin A maize genotypes evaluated across six environments.

Figure 2

Polygon view of genotype and genotype × environment interaction biplot of grain yield for eight provitamin A maize genotypes evaluated in six test environments.

Figure 3

Ranking of eight provitamin A maize relative to an ideal genotype.

3.7 Judging the best environments based on the GGE biplot

An ideal environment can be considered as the most discriminating and representative environment. It is located in the center of the concentric circles. In Figure 4a, L’shi21 is the closest environment to the ideal environment followed by K’zi21. On the other hand, Figure 4b shows that the longest vector was observed on Kal21 followed by L’shi21, while the shortest vector was observed on Kal22. As for the angle between the environment vector and the abscissa of the mean environment axis, Figure 4b shows that K’zi21 forms the smallest angle followed by L’shi21.

Figure 4

Visualization of “ideal environment” using GGE biplot. (a) Concentric circles show the position of ideal environments. (b) Environments closer to the ideal spot are ideal.

4 Discussion

The Democratic Republic of the Congo is one of the countries most affected by vitamin A deficiency. According to Stevens et al. [35] about 26% of children aged 6–59 months are affected by vitamin A deficiency. The use of varieties with high level content of provitamin A called provitamin A varieties is the most effective and least expensive solution for a poor population such as that of the Democratic Republic of the Congo in general and that of its South-East in particular. Recent studies have shown the impact of provitamin A varieties in improving the nutritional status of children under 6 years of age [36–38] and woman [38]. In the South-East of the DR Congo, maize is the staple crop for the majority of the population [1]. The BioDeV Laboratory has recently developed varieties rich in provitaimin A to reduce the prevalence of vitamin A deficiency in this region.

In this study, the agronomic performance of maize genotypes in different environments of the South-East of the Democratic Republic of the Congo was evaluated. The results of combined variance showed that the effect of genotypes was significant for all parameters except cob/plant and ASI. According to Makinde et al. [28], significant effect of genotypes shows the genetic difference between genotypes but also indicates the possibility of selecting successful genotypes. In case of this study, the genotype PVAH-1L can be selected because its yield (9.41 t/ha) was 5.5% higher than Sam 4 vita, the commercial variety. This genotype shows its potential to substitute the commercial variety Sam 4 vita. The remaining genotypes except PVAH-4L have similar yields to Sam 4 vita although there are differences in other parameters. These genetic differences expressed by the genotypes would be due to the crosses from which they were derived. Contrary to the results reported by Sserumaga et al. [39] on ASI and cob/plant and Greveniotis et al. [40] on number of row/cob, in this study, the impact of environment was not significant on these traits. These results are in agreement with the allegations of Nyembo et al. [2], which indicated that the degree of influence of the environment would differ from one trait to another. Some of these, including ASI, cob/plant, rows/cob and 1,000 kernels weight are less influenced by the environment than others.

In this study, the environment influenced a large number of parameters, including grain yield, confirming that the agro-climatic conditions of the test environments used differ from each other. Indeed, Kichanga is a low-lying area characterized by high temperatures during the rainy season, while Kasapa and Musompo are medium-lying areas. In addition, these sites differ from Kichanga in soil type. Kichanga is dominated by Acrisols, whereas the soils at Kasapa and Musompo are predominantly ferralitic. The major difference between the Kasapa and Musompo sites is the average temperature, which is higher at the Kasapa farm (Table 1).

Taking into account the contribution of each factor to the total variation on the sum of squares, the environment played a very important role on the phenotypic expression and performance of the genotypes for grain yield and other components such as emergence rate, plant height, cob length, grain/cob weight and plant weight. According to Fan et al. [24], In multi-environment trials, only G and the genotype–environment interaction are relevant and taken into account simultaneously. Of the two sources of variation, the genotype–environment interaction is the more difficult to control and even to predict. Thus, a significant effect of genotype–environment interaction, as observed in this study, demonstrates not only the existence of various mega-environments with different winning genotypes but also a variation in genotype performance across environments. In this context, Yan and Tinker [29] suggest that selection should be based on the average performance of the genotype in the test environments as well as its stability in these environments. According to Wang et al. [26], a genotype with high and stable performance in test environments has a wider adaptation and high value for promotion and use. In the case of this study, PVAH-1L gave the highest performance, followed by Sam 4 vita, PVAH-3L and PVAH-6L. However, PVAH7L was the most stable, followed by PVAH-4L, indicating that the most stable genotypes were not the most productive. According to Eberhart and Russell [16], the stable genotype is the one with good average, unit coefficient of regression (bi = 1) and the deviations from regression as small as possible (s ² _di = 0). In the case of this study, four genotypes (PVAH-1L, PVAH-3L, PVAH-6L and Sam 4 vita) have an overall above average yield and low regression coefficient (bi < 1.0). According to Alwala et al. [12], s ² _di may be sufficient as a stability parameter when the regression coefficient is not different. Based on this criterion, s ² _di categorized the four genotypes into two groups. Genotypes PVAH-1L and PVAH-6L had insignificant s ² _di values, indicating that these genotypes are only adapted to poor environmental conditions. On the other hand, the genotypes PVAH-3L and Sam 4 vita had significant s ² _di values, so prediction of their yield over environment would not be authentic. However, the S ² _di stability parameter indicated that the genotypes PVAH-1L > PVAH-4L > PVAH-7L > PVAH-6L were very stable (with low s ² _di); however, the yield ranks of PVAH-4L and PVAH-7L (8 and 5, respectively) and the slopes of these genotypes (1.36 and 1.12, respectively) make it difficult to select these genotypes for a wider range of environments even though they appear to be stable. Furthermore, Fernandez [13] mentioned the importance of the coefficient of determination (R²) in the selection of stable genotypes. Based on this parameter, the five most stable genotypes are PVAH-4L > PVAH-5L > PVAH-7L > PVAH-1L > PVAH-6L. However, Yan and Tinker [29] indicate that stability is only beneficial when accompanied by high productivity and only the ideal genotype can meet these criteria. In the case of this study, taking into account the coefficient of determination, the most stable genotype was not the most productive. Moreover, these most productive genotypes also have a low coefficient of variation (Table 3). Taking all stability parameters PVAH-1L and PVAH-6L are considered as stable and desirable genotype. However, no genotype has met all the conditions to be ideal. Therefore, no genotype has a broader adaptation. Even PVAH-1L and PVAH-6L have a specific adaptation to well-defined environments. Yan and Tinker [29] call the ideal genotype, the most productive and stable genotype. However, the ideal genotype only serves as a reference for varietal selection and is located in the middle of GGE biplot concentric circles, but it cannot exist in reality [29,30,32]. Based on this principle, PVAH-1L followed by PVAH-6L were close to the ideal genotype and should be considered as the best provitamin A maize genotypes and are to be recommended in South-East of DRC. On the other hand, Eberhart and Russell [16] stated that a desirable cultivar should have an average yield that is higher under favorable conditions and less fluctuating under unfavorable conditions than that of the cultivar group when tested in many environments. Thus, of all the genotypes evaluated, PVAH-1L was found to be the most desirable. Furthermore, Wang et al. [26] reported that a genotype that is well adapted to a given environment can still be recommended for that environment despite its poor performance and instability. Thus, Sam 4 vita and PVAH-3L, although less stable (Figure 1), as shown in Figure 2 perform well in Kal21 and Kal22 environments. However, Kal21 and Kal22 are environments in Kalemie characterized by low altitude. Therefore, these two genotypes are adapted to the agro-ecological conditions of Kalemie and would easily adapt to agro-ecological zones close to Kalemie. In addition, PVAH-1L and PVAH-6L performed well in K’zi21, K’zi22, L’shi21 and L’shi22. This shows that they are suitable for the Lubumbashi–Kolwezi production area known as the Katangan copper–cobalt arc. The fact that Lubumbashi and Kolwezi belong to the same mega-environment suggests the possibility of choosing one of the two environments multi-environment trials (METs), especially when seed quantity and financial resources are limited. According to Yan et al. [30], L’shi21 is the ideal test environment for the Lubumbashi–Kolwezi mega-environment. Indeed, this environment has a long vector and is the ideal test environment for the choice of a performing genotype. However, the Kal21 environment has the longest vector but with a wide angle; therefore, it cannot be used to select superior genotypes, but should be used in culling unstable genotypes. In addition, Kal22, K’zi22 and L’shi22 have short vectors and provide little or no information on genotypes and therefore should not be used as test environments (Figure 4b). However, the ideal environment must be discriminating and representative and is located in the middle of concentric circles of the GGE biplot. In the absence of this, the near ideal environment is recommended. It is located very close to the ideal environment. In the case of this study, L’shi21 is the closest to the ideal and must be considered as better (Figure 4a).

5 Conclusion

The results of this study showed that agronomic traits such as emergence rate, cob’s insertion height, day to anthesis, day to silking, cob length, grain weight per cob and grain yield were influenced by the effect of genotypes, environment and genotype by environment interaction. The GGE biplot model was effective in analyzing and visualizing genotype by environment interaction to identify the best performing and the most stable genotype and mega-environments. In addition, the joint regression analysis of Eberhart and Russell was performed to compare the selection of stable genotypes by the GGE biplot method. The Eberhart and Russell method and the CV identified PVAH-1L, PVAH-4L, PVAH-7L and PVAH-6L as the most stable genotypes, but could not reliably identify the high yielding genotypes. On the other hand, the GGE biplot reliably and graphically showed the stable and high-yielding genotypes (PVAH-1L > PVAH-6L) as well as the low-yielding genotypes (PVAH-7 > PVAH-4L). The GGE biplot analysis revealed the existence of mega-environments with their own best genotypes. The PVAH-6L and PVAH-1L genotypes are adapted to the Lubumbashi–Kolwezi mega-environment consisting of the K’zi21, K’zi22, L’shi21 and L’shi22 environments in which PVAH-6L performed better. The Sam 4 vita and PVAH-3L genotypes had the best specific adaptation in the Kal21 and Kal22 environments constituting the Kalemie environment. L’shi21 proved to be the desirable test environment for selecting provitamin A maize genotypes due to its high discriminating power and representativeness of the target environments. Thus, the GGE biplot showed that the genotypes PVAH-1L followed by PVAH-6L are to be recommended in the Lubumbashi–Kolwezi Zone, PVAH-3L and Sam 4 vita in the Kalemie Zone for crop improvement and adoption by farmers. Furthermore, of all the genotypes evaluated, PVAH-1L and PVAH-6L proved to be the most desirable. This would contribute to the improvement of the food and nutritional status of children in Democratic Republic of the Congo, particularly in the South-East of the country.