Testcross performance and combining ability of early-medium maturing quality protein maize inbred lines in Eastern and Southern Africa

Analysis of variance and mean performances

Analysis of variance for each environment showed significant differences among genotypes for grain yield and most agronomic and protein quality traits (Supplementary Table S1). The combined analysis of variance across environments showed significant (P ≤ 0.01) environment, genotype (new QPM hybrids and commercial checks), hybrid, and hybrid × environment interaction (HEI) effects for all traits, except HEI effects for protein content and tryptophan concentration and protein quality index (Table 3). Significant differences observed among the genotypes and hybrids for all measured traits across environments demonstrated the existence of adequate genetic variation among the genotypes for the studied traits. The existence of genetic variations among QPM hybrids for grain yield and other agronomic traits under different environments were reported by several investigators^{8,30,31,32,35}. Significant variations among QPM genotypes for kernel endosperm modification, protein and tryptophan concentration were also reported before^29,31,34. The significant HEI observed in the current study for grain yield and all other agronomic traits justified inconsistent performance of the hybrids across the test environments and the need for testing genotypes across environments to select stable hybrids. Previous studies also showed significant HEI for grain yield and other agronomic traits in QPM hybrids^{8,30,31,34,35}. Non-significant HEI for protein and tryptophan contents and protein quality index suggested that the expression of these traits was not affected under different environmental conditions. This indicates that testing genotypes in fewer environments would be adequate to phenotype these traits. In agreement with the current study, non-significant HEI was reported earlier for protein quality traits^29,31.

Mean grain yields ranged from 5.25 t ha^–1 at Chisumbanje to 6.85 t ha^-1 at Bako. Across environments, grain yield for the entries ranged from 2.47 to 7.66 t ha^-1 with a mean of 6.18 t ha^-1 (Supplementary Table S1). The highest-yielding hybrids across environments were H80 (7.66 t ha^-1), H72 (7.63 t ha^-1), H78 (7.56 t ha^-1), H104 (7.42 t ha^-1) and H16 (7.41 t ha^-1) (Table 4). These hybrids showed 19–21% and 9–12% grain yield advantage over the QPM (ZS261) and non-QPM (SC627) commercial checks, respectively. Across environments, 28% and 68% of the QPM hybrids had higher grain yield than the commercial non-QPM (SC627) and QPM (ZS261) hybrids, respectively, indicating the genetic progress made in QPM breeding for high yield with enhanced protein quality in ESA. Previous studies also demonstrated the comparative yield advantage of QPM genotypes to conventional maize genotypes adapted to ESA^6,8,30. Days to anthesis for the hybrids ranged from 67 to 82, with a mean of 75 days, while days to silking varied from 68 to 83 with a mean of 77 days (Table 3, Supplementary Table S1). Among the high-yielding hybrids (Table 4), H11 and H49 mature earlier than all the other hybrids, including the commercial checks. Hybrids with high grain yield and early flowering characteristics are important in areas with short rainy seasons and environments affected by terminal drought. Plant height ranged from 211 to 271 cm, with a mean of 243 cm, and ear height ranged from 96 to 161 cm, with a mean of 135 cm. Hybrids H28, H34 and H11 had shorter plant and ear heights (Table 4). Mean number of ears per plant was 1.12 with a range of 0.92 to 1.60 (Table 3, Supplementary Table S1). Among the selected hybrids, H28, H32, H34, H80 and H104 showed higher number of ears per plant (Table 4).

Mean grain protein content was 99 g kg⁻¹, with a range of 84–118 g kg⁻¹, whereas tryptophan concentration ranged from 0.45 to 0.99 g kg⁻¹, with a mean of 0.77 g kg⁻¹. Among the top-yielding 25 hybrids, H76 (111 g kg⁻¹), H104 (106 g kg⁻¹) and H56 (106 g kg⁻¹) had higher protein levels, while H32 (0.94 g kg⁻¹), H56 (0.92 g kg⁻¹) and H48 (0.89 g kg⁻¹) had higher grain tryptophan concentrations. These values were higher than the levels of protein (105 g kg⁻¹) and tryptophan (0.77 g kg⁻¹) concentrations of the commercial QPM check hybrid (ZS261). The mean protein quality index was 0.78%, with a range of 0.44 to 1.01%, whereas endosperm modification ranged from 1.3 to 3.8, with a mean of 2.3. Hybrids H32 (0.95%), H48 (0.93%) and H106 (0.93%) had higher quality index. Hybrids H80 (1.5), H104 (1.6), and H79 (1.8) showed the most desirable level of endosperm modification (Table 4). According to Vivek et al.³⁷, standard QPM genotypes should have an endosperm modification score close to 2.0, a quality index of at least 0.80% and 0.75 g kg⁻¹ tryptophan concentration, and 80 g kg⁻¹ protein content in whole grain. Most QPM hybrids evaluated in this study had above the recommended levels of quality traits than the commercial QPM check hybrid (ZS261), indicating QPM breeding progresses in improving protein quality without grain yield penalty.

Estimated H² across environments was high for grain yield (74%), days to anthesis (90%), days to silking (89%), plant height (87%), ear height (88%), protein content (72%), tryptophan concentration (87%), quality index (93%) and endosperm modification (70%). Moderate H² was recorded for number of ears per plant (61%).

GGE-biplot analysis

The “which-won-where” pattern of the multi-environment polygon view of Fig. 1 depicted which genotype performed best in which environment (Fig. 1). PC1 and PC2 explained 47.5% and 20.6% of the variation for grain yield, indicating that the biplot accounted for 68.1% of the total variation related to genotype and genotype by environment interaction. The biplot was sub-divided into six sectors and H73, H72, H16, H48, H17 and H26 constituted vertex genotypes in each sector. The winning hybrids were H72, H80 and H104 at Ambo, Gwebi and Mpongwe; H73 at Bako; and H16 and H48 at Glendale. Chisumbanje didn’t discriminate among the hybrids; hence, it was less informative. Two corner hybrids, G17 and G26, were disposed of far from all the test locations on the GGE-biplot, signifying their inferior yield performance at all the environments. The GGE analysis delineated the test environments into two mega-environments (Fig. 1). The first comprised of Ambo, Bako, Chisumbanje, Gwebi and Mpongwe and the second comprised of only a single environment, Glendale. Within the first mega environment Ambo, Chisumbanje, Gwebi and Mpongwe are positively correlated because they are placed in less than 90° in the GGE-biplot. Different authors used AMMI and GGE bi-plot models to know how many mega environments exist in a specific target environment^32,51,52,53. Mean and stability of the top 27 hybrids (25 high-yielding testcrosses and two commercial checks) were visualized by drawing an average environment coordination (AEC) view graph, represented by a small circle (Fig. 2). The thick line perpendicular to AEC ordinate separated genotypes with yield less than the average (to the left side line) from those with grain yield greater than the mean (to the right side line). Accordingly, H72, H80, H50, H104, H16, H34 and H48 had higher grain yield across environments. Their projections onto the AEC ordinate measured the stability of the hybrids. Among the highest-yielding hybrids, H50, H80, H72 and H104 were the most stable hybrids that with shorter projections onto the AEC ordinate. High-yielding and stable hybrids identified in this study could be recommended for on-farm testing and commercial production in ESA after fulfilling the requirements for varietal release. However, hybrids like H72 and H50, with very low tryptophan concentration and protein quality index, cannot be advanced further as QPM hybrid. Previous research findings in Southern Africa⁶ and Eastern and Southern Africa^8,34 also reported stable and high-yielding QPM hybrids.

Shows the ‘which-won-where’ view of genotype main effect plus genotype by environment interaction (GGE) biplot constructed based on environment-centered singular-value partitioning for grain yield of 108 genotypes tested across six environments. The codes of genotypes are stated in Supplementary Table S1.

Shows the ‘mean versus stability’ view of the genotype main effect plus genotype by environment interaction (GGE) biplot constructed based on grain yield data of 25 top-yielding quality protein maize hybrids and two commercial checks evaluated across six environments. The codes of genotypes are stated in Supplementary Table S1.

Combining ability analysis

Mean squares attributable to line GCA and tester GCA were significant for most traits, except tester GCA for grain yield and ears per plant (Table 3). Similarly, SCA mean squares were significant for most traits, except kernel modification, tryptophan concentration, and quality index, indicating the importance of both additive and non-additive genetic effects in the inheritance of these traits. Thus, effective selection of these traits for further improvement could be feasible through hybridization, recurrent selection and back-cross breeding. The present findings are consistent with previous studies on QPM genotypes^30,32,34,35. The analysis also revealed highly significant line GCA × environment interactions for most traits, except for kernel modification, protein content, and quality index (Table 3). Significant tester GCA × environment interaction was observed for most traits, except for protein content, tryptophan concentration and quality index. Significant GCA × environment interactions for most measured traits indicate that combining abilities of inbred lines and testers varied across test environments. This implies the need to test the combining abilities of the inbred lines and testers across environments prior to the selection of stable parental genotypes⁵⁴. Significant GCA × environment interaction was reported previously in QPM genotypes across environments for grain yield and agronomic traits^21,31,32,34. In contrast, Njeri et al.³⁵ reported non-significant GCA × E interaction for grain yield across optimally managed environments. This study showed non-significant GCA × environment interaction effects for protein content, tryptophan concentration and protein quality index that would enable selection of QPM inbred lines with stable GCA effects across test environments.

The mean square due to SCA × environment interaction were significant only for ears per plant (Table 3). The existence of non-significant SCA × environment interaction for most studied traits indicated consistent expression of SCA effects for these traits in different environments. Likewise, Machida et al.²⁹ and Nepir et al.³¹ reported non-significant SCA interaction with the test environments for agronomic and protein quality traits. Contrary to the current findings, significant SCA × environment interaction was reported for endosperm modification and protein quality traits by Wegary et al.³⁴ and Abakemal et al.⁵⁵.

GCA sum of squares (both line and tester GCAs) as a percentage of the hybrid sum of squares were larger than SCA sum of squares for all studied traits. The contribution of GCA sum squares ranged from 51% (grain yield) to 90% (quality index), while the SCA sum of squares ranged from 10 to 49% (Table 3). The greater contribution of the GCA sum of squares among the hybrids for most traits observed in this study implies the preponderance of additive genetic effects for these traits in the set of QPM inbred lines studied. Therefore, progeny performance can adequately be predicted based on parental performances^34,56. Early generation testing and selecting potential single-cross hybrids through prediction from GCA effects alone could be feasible⁵⁴. The preponderance of additive effect for agronomic and protein quality traits in QPM germplasm have also been reported in previous studies^31,34,35,55.

Similar contributions of both GCA and SCA (51 vs. 49%) to the hybrid sum of squares for grain yield suggested that both additive and non-additive genetic effects are almost equally important. In such a scenario, breeding programs should exploit both components by evaluating parents for GCA and testing the resulting hybrids in target environments^55,57. Contrary to the current results, Wegary et al.³⁴ reported the importance of non-additive gene effects in the inheritance of grain yield in QPM germplasm.

Estimates of combining ability effects

The 27 QPM inbred lines and four testers used in this study depicted considerable variations in GCA effects for most studied traits (Table 5), indicating the existence of sizable diversity in the genetic constitution of the inbred lines. For grain yield, L13, L19, L20, L21 and L27 showed highly significant positive GCA effects, indicating that these inbred lines could be useful sources of favorable alleles for higher grain yield. These inbred lines also have greater potential to be used as testers in the breeding program. L1, L2, L3, L9 and L17 showed highly significant negative GCA effects for days to anthesis and silking and for plant and ear height. Such inbred lines could be utilized in early maturing and short-statured QPM hybrid development by considering the yield potential and other desirable attributes. L7, L8 and L16 had highly significant positive GCA effects for number of ears per plant and can be sources of favorable alleles for enhancing prolificacy in QPM germplasm. For kernel endosperm modification, L21, L23 and L24 showed significant negative GCA effects, signifying their value in developing QPM varieties with well-modified kernel endosperm. QPM inbred lines exhibiting modified endosperm phenotype could be used as o2 donor parents for the conversion of non-QPM inbred lines to QPM counterparts^30,37. L2, L10, L11 and L20 had desirable GCA effects for protein content, indicating that these inbred lines contain a higher frequency of favorable alleles to elevate protein content in the hybrids. About one-third of the inbred lines studied showed significant and positive GCA effects for tryptophan concentration and protein quality index. L8, L12 and L18 led to these traits’ most desirable GCA effects.

None of the testers showed significant GCA effects for grain yield (Table 5). However, this study identified testers with desirable GCA effects and, hence, favorable additive effects for days to anthesis and silking (T3), plant and ear height (T4), endosperm modification (T2 and T3), protein content (T2), tryptophan concentration and protein quality index (T4). The presence of significant and desirable inbred line and tester GCA effects for most of the studied traits indicated the breeding value of the parents attributable to additive genetic effects that enable breeders to predict progeny performance based on  parental performances. Highly variable SCA effects that ranged from negative to positive were observed among the line-by-tester cross combinations for grain yield, days to anthesis and silking, plant and ear height, ears per plant, and protein content (Supplementary Table S2). This indicated that specific crosses performed better or poorer than what could be expected based on the GCA effects of respective parental inbred lines and/or testers. This can be witnessed by the fact that none of the cross combinations with best SCA effects for grain yield, which were also among the high-yielding hybrids across environments, viz. L4 × T4 (H16), L7 × T4 (H28) and L17 × T1 (H65), contain parents with high GCA effects for the same trait.