Identifying heterotic groups and testers for hybrid development in early maturing yellow maize (Zea mays) for sub‐Saharan Africa

Abstract Identification of heterotic groups and efficient testers, which are important prerequisites for the development of outstanding hybrids, has been a major challenge to its success, especially for early and extra‐early germplasm. This study was carried out to (a) identify the most efficient heterotic grouping method for classifying a set of inbred lines and (b) determine the efficiency of testers in classifying inbred lines into heterotic groups. A total of 205 hybrids obtained by crossing 41 inbred lines with five standard testers were evaluated together with five hybrid checks under drought, low soil nitrogen (N), Striga‐infested and optimal environments in Nigeria between 2014 and 2016. The heterotic group's specific and general combining ability (HSGCA) method was more effective in classifying the inbred lines into heterotic groups. Testers TZEI 17 and TZEI 23 were the most efficient across environments and could be invaluable for classifying other lines into heterotic groups and assessing combining ability of maize inbreds. In addition, these testers and heterotic groups represent an invaluable resource for development of outstanding hybrids in sub‐Saharan Africa (SSA).

countries has been very slow, although it has now been increasing gradually in SSA.
An important requirement for the development of high-yielding commercial hybrids is the availability of information on the heterotic groups and patterns of the available inbred lines in a breeding programme (Barata & Carena, 2006;Fan et al., 2018). Globally, inbred lines and the derived commercial hybrids resulting from them must be stress tolerant. Stresses common to all SSA countries include drought and low N and infestation by parasitic weeds such as Striga hermonthica, especially in WCA. Classification of inbred lines into heterotic groups under the different environmental conditions has been initiated by several researchers in SSA (Agbaje, Badu-Apraku, & Fakorede, 2008;Badu-Apraku, Fakorede, et al., 2015;Badu-Apraku et al., 2013;Menkir, Badu-Apraku, Thé, & Adepoju, 2003) but limited success has been achieved. Taking a cue from the experience and the standards attained over a long time by the USA maize breeders, heterotic grouping of inbred lines has to be a continuous exercise in maize breeding programmes. Heterotic groups and testers for the newly developed or introduced set of inbred lines in a breeding program must be determined (Barata & Carena, 2006;Fan et al., 2018;Fan, Tan, Yang, & Chen, 2004). Additionally, it is important for any successful or effective breeding programme aimed at developing outstanding drought, low N and Striga resistant hybrids to assess the heterotic groups under individual stress, non-stress (optimum), and across environments.
The International Institute of Tropical Agriculture (IITA), Ibadan-Nigeria, in collaboration with the International Maize and Wheat Improvement Center (CIMMYT) in Kenya and with national programmes, has for years been developing and releasing OPVs in all maturity groups and hybrids in the late and intermediate groups but it is only recently that a few stress-tolerant/resistant early and extra-early hybrids have been developed and released in some WCA countries. However, maize researchers have consistently observed that the performance of early (90-95 days to physiological maturity) and extra-early (80-85 days to physiological maturity) yellow maize OPVs and hybrids lag grossly behind their white counterparts in SSA because for a long time research emphasis on maize had been on the white endosperm maize. However, recently, there has been increased research emphasis on the use of yellow endosperm maize because of the high demand for the poultry industry and for human consumption to address vitamin A deficiencies. The yellow maize is preferred for poultry feed because it imparts the yellow colour to the egg yolk and contributes to human nutrition. Therefore, it is imperative to identify the heterotic groups of the newly developed or introduced early maturing yellow maize inbred lines so that they could be successfully used for the development of high-yielding hybrids and synthetic varieties for commercialization in SSA. Hybrid maize varieties are preferred by progressive farmers in SSA due to the high yield compared to OPVs (Ayinde, Fola, & Ibrahim, 2011;Correjado & Magulama, 2008) and uniformity in growth and other agronomically desirable characteristics.
Results of the attempts by maize breeders to identify the most efficient heterotic grouping methods have not been consistent. For example, Fan, Miles, Takahashi, andYao (2009) andBadu-Apraku, Fakorede, et al. (2015) compared the specific combining ability (SCA) of several lines using molecular markers and the heterotic group's specific and general combining ability (HSGCA) methods and found the HSGCA method to be the most efficient based on the breeding efficiency (the average of the proportion of total inter-heterotic group hybrids that is due to superior high-yielding inter-heterotic group hybrids plus the proportion of total low-yielding intra-heterotic group hybrids that is due to the low-yielding intra-heterotic group hybrids). In contrast, Badu-Apraku, Annor, et al. (2015), Badu-Apraku, Fakorede, Talabi, et al. (2016) reported the heterotic grouping based on molecular markers as the most efficient in a study involving early maturing quality protein maize (QPM) inbred lines.
The contrasting results reported have been attributed to the differences in the genetic materials used (Badu-Apraku, Fakorede, Gedil, et al., 2016). It is therefore of utmost importance to classify the newly developed early maturing maize inbred lines using the most efficient grouping method to identify the best set of new inbred lines for effective use in maize breeding programmes.
Another important requirement of any successful maize breeding programme is the availability of efficient testers, which could effectively discriminate and classify inbred lines into appropriate heterotic groups for the development of high-yielding hybrids and synthetic varieties. An effective tester should be able to rank inbred lines correctly for performance in hybrid combinations and increase the differences between testcrosses for efficient discrimination (Rawlings & Thompson, 1962). Several early maturing yellow endosperm inbred lines [TZEI 10 (A), TZEI 17 (B), TZE 23 (C), TZEI 129 (D) and ENT 13 (E)] have been identified as standard testers in the IITA Maize Improvement Program (MIP). It is therefore important to authenticate the efficiency of these early inbred lines as testers which could be effectively utilized for grouping other inbred lines and for the development of productive hybrids and synthetic varieties. The objectives of the present study were to (a) identify the most efficient of the two heterotic grouping methods, HSGCA, and genetic distance from SNP markers for SSA agro-ecological conditions and (b) determine the efficiency of testers in classifying selected tropical inbred lines into heterotic groups.   Note: (TZEI 11 × TZEI 8) S 7 inb 18-1/3-1/2-1/1 = TZEI 415 was developed from a cross between TZEI 11 × TZEI 8, taken through seven cycles of inbreeding (S 7 ) and was the 18th line selected after the first cycle (S 1 ) of inbreeding. This was followed by several cycles of repeated selfing and selection at the different stages of inbreeding.

| Experimental materials and generation of testcrosses
(three of the testers were identified from a set of inbreds derived from

| Field experiments
The 205 (Table S1). The low N fields at both locations were depleted of N by growing maize continuously at high density for several years and removing the biomass after each harvest. Soil samples taken from 0 to 15 cm depth before field preparation were analysed for nitrogen (N), phosphorus (P) and potassium (K) contents at the IITA analytical services laboratory, Ibadan, Nigeria, following the Kjeldahl digestion and colorimetric methods (Bremner & Mulvaney, 1982). The Mokwa soil contained 0.033% of N, 4.11 mg/kg of P and 0.14 cmol/kg of K while that of Ile-Ife had 0.081% of N, 4.04 mg/kg of P and 0.23 cmol/kg of K. Nitrogen fertilizer (Urea) was applied at two WAP following thinning to bring the total available N to 30 kg/ha as indicated by the soil tests. Single superphosphate and muriate of potash fertilizers were applied to obtain 60 kg/ha each of P and K.
Also, the hybrids were evaluated under optimal growing conditions during the 2015 and 2016 rainy seasons at Mokwa, Ikenne, Abuja (9°16'N, 7°20'E, 300 m asl, 1,500 mm annual rainfall) and during the 2015 rainy season at Ile-Ife (Table S1). The compound fertilizer, NPK (15:15:15), was applied to all the optimal trials at two WAP to provide 60 kg/ha each of N, P and K and top-dressed at four WAP with 60 kg N/ha. The low N, drought and optimal fields were kept weed-free by the application of atrazine and gramozone as preemergence and postemergence herbicides at 5 L/ha each of primextra and paraquat and later by manual weeding.
The hybrids were evaluated under artificial Striga infestation at Mokwa and Abuja in 2015 and 2016 (Table S1). Ethylene gas was injected into the soil at two weeks before planting to stimulate suicidal germination of existing Striga seeds. The infestation with Striga was carried out using the method of Kim (1991

| Genotyping by sequencing
The enzyme ApeKI was used for digestion and creating a genotyping by sequencing (GBS) library with unique barcodes for each inbred line as described by Elshire et al. (2011). The reads obtained from the GBS library were called in the GBS pipeline Tassel version 3.0.147 which is an extension to the Java program TASSEL (Bradbury et al., 2007). The sequences were aligned to the maize reference genome B73 RefGen v1 after filtering using the Burrows-Wheeler alignment tool (BWA) (Schnable, Ware, Fulton, Stein, & Wei, 2009). The procedure produced 51,009 SNPs covering all the ten chromosomes of the maize genome, out of which 3,508 SNP loci, having a minor allele frequency of more than 5% and no missing values, were selected using TASSEL version 5.0, and employed for analyzing the genetic diversity of the inbred lines in the present study. The pair-wise Rogers (1972) genetic distances were estimated among the inbred lines using PowerMarker version 3.25 (Liu & Muse, 2005).

| Field phenotyping
Data were recorded on all plots for days to silking, and days to anthesis, anthesis-silking interval (ASI), plant and ear heights, root lodging, stalk lodging, ear aspect, ear rot and ears per plant. In addition, plant aspect was recorded on the drought, low N and optimal plots, stay-green characteristic on the low N and drought plots as described by Badu-Apraku, Fakorede, Gedil, et al. (2016) and

Striga damage and number of emerged Striga plants on only the
Striga-infested plots. Under optimal and Striga-infested environments, a shelling percentage of 80% was assumed for each plot.
Grain yield was obtained from the ear weight and converted to kg/ha by adjusting the moisture content to 15%. For experiments conducted under low N and drought conditions, harvested ears from each plot were shelled to determine the percentage grain moisture. Grain yield in kg/ha adjusted to 15% moisture content was then computed from the shelled grain weight. The 80% shelling percentage was assumed for entries only under Striga infestation and optimal conditions because grain filling is usually normal under such conditions.

| Statistical analysis
Analysis of variance (ANOVA) was performed across environments

(location-year combinations) with PROC GLM in the Statistical
Analysis System (SAS) using a random statement with test option (SAS Institute, 2011). In the combined ANOVA, genotypes were considered as a fixed factor while environment, replication within environment and incomplete blocks within replication by environment were regarded as random factors.
A line x tester analysis of variance was used to determine the sta- To identify the most efficient inbred tester across the four contrasting research conditions, data on grain yield mean values across the four research conditions adjusted for block and replication effects were subjected to genotype main effect plus genotype x environment interaction (GGE) biplot analysis (testers were used in the analysis in place of environments) as described by Yan and Hunt (2002).

| RE SULTS
Analysis of variance of grain yield and other traits across contrasting environments.
Significant (p < .05) mean squares of environments (E), hybrid (G) and hybrid × environment interactions (GEI) were observed for grain yield and most measured agronomic traits across environments ( Table 2). Partitioning of the hybrid components of variation into GCA of line (GCA-line) and GCA of tester (GCA-tester) and SCA mean squares exhibited significant gains for GCA-line, GCA-tester and SCA for grain yield and most measured agronomic traits across environments. The GCA-line × E and GCA-tester x E interaction mean squares were also significant for most measured traits (Table 2). In contrast, the mean squares of the SCA × E interactions were not significantly different for most measured traits (

| Efficiency of testers based on discriminating ability across environments
The efficiency of a tester was assessed by the average environment (tester) coordination (AEC) view of the GGE biplot (Figure 1) as described by Yan (2014) and Akinwale, Badu-Apraku, Fakorede, and Vroh-Bi (2014).
The thick single-arrow line is referred to as the average environment (tester) coordinate abscissa (AEC abscissa) while the double-headed arrow line is called the AEC ordinate. The efficiency is determined by the relationship among the testers and the length of the tester vector. In the biplot display, the cosine of the angle between any two tester vectors indicates the correlation coefficient between the testers. The smaller the angle between any two testers, the more closely related the testers are in classifying inbred lines into heterotic groups. In addition, in the biplot display, the rays connecting the tester label to the biplot origin are described as tester vectors and the vector length of a tester approximates the standard deviation, which measures the magnitude (discriminating power) of its ability to assess the grain yield of the crosses. Testers with shorter vectors provide little or no information about the entries evaluated compared to those with longer vectors. Based on these criteria, the ranking based on discriminating ability of the testers was as follows: TZE 23 (C)> TZEI 17 (B)> TZEI 10 (A)> ENT 13 (E)> TZEI 129 (D). Strong positive correlations (similarity) were found among the testers B, A, E and D ( Figure 1) whereas tester C was found to be unique from all the other testers. Tester C was therefore identified as the most efficient early maturing yellow inbred tester across the four research conditions. Tester B was found to be the most efficient among the testers which were found to be similar, that is A, B D and E. Hence, tester B represented the four testers. Testers C and B were therefore the most efficient among the five testers based on the discriminating ability.

| Heterotic grouping of inbred lines based on HSGCA method
Considering the fact that testers A, B, D and E were similar while tester C was unique, the HSGCA heterotic grouping method classified the 41 inbred lines into three heterotic groups across the research environments (Table 3). Twenty-two of the inbred lines were placed in heterotic group 1 (heterotic group of testers A, B, D and E), 15 in the heterotic group 2 (heterotic group of tester C) and four in the heterotic group 3 (heterotic group of inbred lines which were not grouped by any of the five testers) (Table 3).

| Heterotic grouping of inbred lines based on the SNP marker method
The heterotic grouping method based on the SNP markers also classified the inbred lines into three heterotic groups (Table 3)

| D ISCUSS I ON
The failure of the HSGCA and SNP-marker-based grouping methods to classify some inbreds into the heterotic groups of the five testers suggested that those inbred lines belonged to heterotic groups other than those of the five testers. (2017) who reported that the HSGCA was the most efficient for classifying inbred lines under drought, low N and optimal environments.
The differences observed in the present study and those of the earlier workers could be attributed to genetic differences in the set of inbred lines used in the present study. The classification of the inbred lines into three heterotic groups or clusters (based on the most efficient method, HSGCA) indicated that there was a broad genetic diversity among the set of inbred lines used in the present study.
Another important requirement of any successful breeding programme is the availability of efficient testers which could effectively discriminate and classify inbred lines into appropriate heterotic groups and for the development of high-yielding hybrids and synthetic varieties. The efficient testers identified in the present study could be utilized for cost-effective classification of other early-maturing tropical yellow inbred lines into heterotic groups, assess the combining ability and identify superior hybrid combinations under drought, low N, Striga, optimal and across the research environments.

| CON CLUS IONS
The HSGCA method had a higher breeding efficiency than the SNP-based grouping method across the research environments indicating that it was more effective in classifying the inbred lines into heterotic groups. Maximum heterosis could therefore be exploited if inbred lines with significant and positive GCA effects for grain yield and classified into opposing heterotic groups by the HSGCA method are crossed for hybrid or synthetic variety development. Testers TZEI 17 (B) and TZEI 23 (C) were identified as the most efficient across contrasting environments and would be invaluable for classifying other inbred lines into heterotic groups, assessing the combining ability and developing superior multiple stress tolerant early maturing yellow hybrids for use in the tropics including SSA.