HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (5) Citing Articles via Google Scholar Google Scholar Articles by Narro, L. Articles by Salazar, F. Search for Related Content PubMed Articles by Narro, L. Articles by Salazar, F. Agricola Articles by Narro, L. Articles by Salazar, F. Related Collections Germplasm Enhancement Plant and Soil Interactions Maize Crop Genetics

CROP BREEDING, GENETICS & CYTOLOGY

Using Line x Tester Interaction for the Formation of Yellow Maize Synthetics Tolerant to Acid Soils

^a International Maize and Wheat Improvement Center (CIMMYT), Apdo. Aéreo 67-13, Cali, Colombia
^b CIMMYT, Apdo. Postal 6-641, 06600 Mexico D.F., Mexico

^* Corresponding author (s.pandey{at}cgiar.org).

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Breeders need more information on selecting testers to identify lines for formation of synthetics and need more user-friendly methods to study general combining ability (GCA) and specific combining ability (SCA) of genotypes. Forty-three lines were crossed to two narrow-based (S3 line and S3 x S3 hybrid) and two broad-based testers, which were open-pollinated varieties (OPVs). Topcrosses (line x tester, L x T) were evaluated in eight acid soil environments. The additive main effect and multiplicative interaction (AMMI) and the site regression (SREG) models were used to study the GCA of the testers and lines and the SCA of the L x T interaction. The SREG biplot contains the effect of the lines plus the L x T interaction and displays both GCA and SCA, whereas the AMMI biplot contains the effect of the L x T interaction and displays only the SCA. One synthetic was developed by recombining six lines identified as superior by each tester. The four synthetics were evaluated in 12 acid and nine nonacid soil environments. The synthetic developed with the S3 line as tester yielded the highest and the one developed with an OPV as tester yielded the lowest. The AMMI and SREG models seem to provide an effective tool to visualize and study GCA and SCA of genotypes.

Abbreviations: AMMI, additive main effect and multiplicative interaction • AT, average tester • GCA, general combining ability • OPV, open-pollinated variety • SCA, specific combining ability • SREG, site regression

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
MAIZE is planted on 140 million hectares worldwide, of which 97 million hectares are in developing countries. The area planted with OPVs in developing countries is 54 million hectares, and only 25% of that is planted with improved OPVs (The Maize Program, 1999). Mean grain yield in developing countries is <3.0 Mg ha^-1, averaging 1.45 Mg ha^-1 for Africa, 2.23 Mg ha^-1 for Latin America, and 2.94 Mg ha^-1 for Asia. In contrast, average grain yield in the USA is 8.3 Mg ha^-1, where most of the area is planted to single cross hybrids (CIMMYT, 1999). Since hybrid seed production and distribution is limited in many developing countries, as is the capacity of the farmers to purchase fresh seed for each planting, adoption of improved maize OPVs and synthetics would be an important tool to increase maize grain yield in these countries. Seed production and adoption of OPVs and synthetics is easier and cheaper than that of hybrids; also, farmer-to-farmer seed movement is only possible with OPVs, a practice used by many farmers in developing countries (The Maize Program, 1999).

Hayes and Garber (1919) originally suggested the use of maize synthetics. Sprague and Jenkins (1943) suggested the role of synthetics as reservoirs of desirable germplasm. For instance, Iowa Stiff Stalk Synthetic is considered a source population of inbreds with above-average performance for combining ability (Hallauer and Miranda, 1988). Lonnquist (1961) defined a synthetic as an open-pollinated population formed by intercrossing lines and maintained subsequently by mass selection. In this paper, synthetics are defined as OPVs formed by recombining lines with high GCA and maintained from one generation to the next by open pollination.

The use of testers in a maize recurrent selection program has been well documented (Jenkins and Brunson, 1932; Matzinger, 1953; Rawlings and Thompson, 1962; Allison and Curnow, 1966, Hallauer, 1975; Hallauer and Miranda, 1988; Russell et al., 1992, Menz et al., 1999). These authors concluded that choice of a suitable tester should be based on simplicity in its use, its ability to classify the relative merit of lines, maximize genetic gain, and enhance the expected mean yield of a population generated using selected cultivars. However, it is difficult to identify testers having all these characteristics because, initially in a breeding program, only OPVs are available. The use of the parental variety as a tester results in some improvement of the mean performance of the population (Rawlings and Thompson, 1962). Allison and Curnow (1966) suggested use of low-yielding varieties as testers. The use of a single-cross as a tester has been reported by Horner et al. (1976). The use of an inbred as tester in a recurrent selection program was suggested by Russell and Eberhart (1975) and it has been widely used by breeders (Walejko and Russell, 1977; Darrah, 1985; Horner et al., 1989).

There is a limited amount of information available on the use of testers in generating improved synthetics. Lonnquist (1949) used the OPV Krug as tester and developed two contrasting synthetics (high and low yielding) by selecting the highest and lowest yielding S₁ lines based on the topcross performance of 36 S₁ lines. Highly significant grain yield differences were obtained for high- and low-yielding synthetics as compared with the original unselected Krug cultivar. Castellanos et al. (1998) studied 21 maize inbreds and seven testers (five single crosses, one synthetic, and one inbred line) to identify the best tester and concluded that the single cross was the best alternative in a breeding program oriented to generate superior three-way and double-cross hybrids.

Multiplicative models such as the AMMI (Gauch, 1988; Zobel et al., 1988; Crossa, 1990) and the SREG model (Crossa and Cornelius, 1997) have been used for studying genotype x environment interaction. These models can be also utilized for studying pattern of response of lines when crossed with testers, that is, L x T interaction. Recently, Yan and Hunt (2002), using the SREG model for analyzing a diallel data set, defined an ideal tester as one that highly discriminates among the lines and is a good representative of all testers. The authors used a graphical representation of the SREG model (biplot) to show the GCA of lines and testers as well as the SCA of L x T interaction. However, since the SREG model contains in its multiplicative components both the main effects of the lines as well as the L x T interaction, visualization of the SCA of each line with each tester is often not clear and is imprecise. The use of AMMI and its interaction multiplicative component may be useful for obtaining a clearer and more precise prediction of the SCA of the L x T interaction than that obtained by the SREG model.

Given the wide use of OPVs of maize in developing countries, selection of a tester that can identify superior inbreds for use in a synthetic development program will continue to be an important research activity. Objectives of this research, therefore, were to identify type (narrow vs. broad genetic base) of tester that would be more suitable for selecting lines for formation of synthetics and to examine use of multiplicative models (AMMI and SREG) to study GCA and SCA of genotypes.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Experimental Data
Line x Tester (Topcross)
Forty-three yellow inbreds (Table 1), 16 from CIMMYT Population SA3 (Granados et al., 1995), 17 from CIMMYT Population SA4, and 10 from CIMMYT Population SA5 (Pandey et al., 1995), selected for their tolerance to acid soils and good agronomic characters, were available at CIMMYT South American Regional Maize Program based in Cali, Colombia. Selection criteria to identify acid-soil-tolerant lines was grain yield of inbreds planted at three (S. Quilichao, Villavicencio 1, and Villavicencio 2) acid soil environments in Colombia. Lines are identified by the population name from which they were derived followed by a sequence number; for example, 3-1 means that Line 1 was derived from Population SA3. SA3 and SA4 are yellow heterotic populations under improvement for tolerance to acid soils. SA3 and SA5 are two yellow populations belonging to the same heterotic group (Pandey et al., 1994).

Line x tester (topcrosses) were developed by using a mixture of pollen of at least 100 plants when the tester was an OPV and 40 plants when the tester was a line or a hybrid. At harvest, all ears from each topcross were bulk-shelled and a sample of this seed was used for trial evaluation. Topcrosses were evaluated during 1996 in eight acid soil environments, namely Villavicencio1 96A, Villavicencio2 96A, Villavicencio1 96B, Villavicencio2 96B, Quilichao 96A, Quilichao 96B, Carimagua 96B, and Alto Mayo 96B (Table 2). Entries were arranged in a randomized complete block design with two replications per site in a split-plot arrangement where the 43 lines were included as main plots and the four testers as subplots. Lines and testers were considered as fixed effects and sites as random effects. Plant density was 53 000 plants ha^-1 (0.75 m between rows and 0.50 between hills, two plants per hill). Lime was used to decrease Al saturation to appropriate levels (Table 2). Row length was 2.5 m with 10 plants per row, fertilized with 90 kg N ha^-1, 60 kg K₂O ha^-1, and enough P₂O₅ added to increase P in the soil to levels given for each environment (Table 2).

Synthetics
On the basis of their topcross performance, six lines were selected for each tester to form the following synthetics: CIMCALI97A SA4 (T1S), CIMCALI97A SA3-1 (T2S), CIMCALI97A SA3-2 (T3S), and CIMCALI97A SA3-3 (T4S) for testers T1, T2, T3, and T4, respectively. Synthetics were developed using a diallel mating scheme. Twenty-two seeds were planted for each line and plant-to-plant crosses were made between each paired row (each paired row included two different lines). For each tester, a total of 15 two-parent crosses were obtained and five ears were selected for each cross. Ears were shelled individually and equal numbers of seeds bulked to plant 15 5-m-long rows (22 plants per row). Plant-to-plant crosses were made among plants across the 15 rows. At harvest, six ears from each row were selected and shelled individually. Equal numbers of seeds from each of the 90 selected ears was bulk-shelled to represent the F₂ seed for a synthetic.

The four synthetics were evaluated as part of a 5 x 4 lattice design, along with the four testers and other genotypes to make up a total of 20 entries, with four replications per environment. The trial was planted in 21 environments (12 acid and nine nonacid soil environments) during 1998 to 2000. Acid soil environments were Villavicencio1 98A (E1), Ayapel 98B (E5), Villavicencio2 98B (E6), Sete Lagoas 98B (E7), Suwan 99A (E9), Villavicencio1 99A (E10), Matazul 99A (E12), Villavicencio1 99B (E13), Villavicencio2 99B (E14), Assam 00A (E17), Villavicencio1 00A (E19), and Matazul 00A (E20). Nonacid soil environments were Palmira 98B (E2), Buga 98B (E3), Cerrito 98B (E4), Pajonal 98B (E8), Palmira 99A (E11), Entre Rios 00B (E15), Capitan Miranda 99B (E16), Palmira 00A (E18), and Turipana 00A (E21). Entries were planted in two-row plots 5 m long with a density of 53 000 plants ha^-1. Fertilization was similar to that used in the topcross evaluation. Data were recorded in the same way as for the topcross trials.

Statistical Analyses
The AMMI and SREG Models
The AMMI and SREG models have been primarily used for analyzing multienvironment cultivars trials and for studying genotype x environment interaction. In the context of a two-way table from a line x tester experiment, the lines are considered as the rows (cultivars) and the testers as the columns (environments).

In this study, the AMMI and the SREG models were used to determining the GCA of lines and testers and the SCA for studying the L x T interaction. The AMMI model for the mean response of the ith line crossed with the jth tester (_ij) is _ij = µ + _i + _j + ^t_{k=1 kikjk} + _ij, where µ is the grand mean, _i and _j are the main effects of the lines (i = 1, 2, ..., l) and testers (j = 1, 2, ..., t), respectively. The L x T interaction multiplicative components have the following parameters: _k, the singular value of the kth multiplicative component that is ordered _{1 2} ... _t; _ik, the left singular vector of the kth component; and _jk, the right singular vector of the kth component. The _ik and _jk satisfy the ortho-normalization constraints _iikik' = _jjkjk' = 0 for k k' and _i²_ik = _j²_jk = 1. The average experimental error is _ij. The maximum number of multiplicative terms is t = min(l - 1, t - 1).

The AMMI analysis helps the interpretation of the L x T interaction because the least squares estimates of the additive main effects of lines _i = _i. - _.. and testers _j = _.j - _.. are the line and tester GCA effects, respectively, whereas the least squares estimate of the L x T interaction ^t_k=1kikjk - _ij - _i. - _.j + _.. is the estimate of the SCA effects; and the _i., _.j, and _.. correspond to the mean of the ith line, the jth tester, and the grand mean, respectively (Cornelius et al., 1996).

The SREG model is _ij = µ + _j + ^t_k=1kikjk + _ij, where the terms are as those defined in the AMMI model. The multiplicative components of the SREG model (^t_k=1kikjk) comprise the main effect of lines (_i) plus the L x T interaction term.

Rescaling the Singular Vectors for the SREG and AMMI Biplots
For the AMMI₂ and SREG₂ biplots, the singular values of PC1 and PC2, ₁ and ₂, respectively, were absorbed into the singular vectors of testers (₁ and ₂) and lines (₁ and ₂) in a way that they make an equal contribution of the first and second multiplicative components to the yield predicted values. The rescaling method used in this study will force the range of values in the singular vector for testers to be equal to the range of values in the singular vector for lines and it was applied by Yan et al. (2001) and Crossa et al. (2002) when displaying biplots of several multiplicative models.

Interpreting the Biplots of the SREG and AMMI for GCA and SCA
The biplot of the AMMI with two multiplicative components (AMMI₂) graphs the first and second L x T interaction components, PCA1 = _1i1j1 vs. PAC2 = _2i2j2, which approximates the pattern of response of the lines across different testers and with specific testers (SCA). In this study, the AMMI model and its biplot are used to display SCA of the L x T interaction. In the AMMI biplot, an angle < 90° or > 270° between a line vector and a tester vector indicates that the line combines well with that tester (positive SCA). A negligible or negative SCA is indicated if 90° < < 270°. Thus, SCA between a line and a tester is large if the angle between them is as small as possible, that is, the vectors are nearly parallel and both vectors are long.

However, in the SREG₂ model, the interpretation of the biplot is with respect to the variation for which main effects of testers (GCA for testers) and the main effect of lines and L x T interaction (GCA of lines and SCA of L x T interaction). In this model, the PC1 scores are closely associated with the line main effects (_i). Performance of a line combined with a tester can be approximated by the orthogonal projection of the line vector onto the tester vector. The cosine of the angle between two testers (or lines) vectors approximates the correlation of the two testers (or lines). An angle of zero indicates a correlation of +1; an angle of 90° (or -90°), a correlation of 0; and an angle of 180°, a correlation of -1.

Yan et al. (2000) presented standard biplots of the SREG₂ model that helped enhance its interpretation for selecting the best performing cultivars in subsets of sites by drawing a polygon connecting the most responsive cultivars (vertex of the polygon). Each side of the polygon has a perpendicular line drawn from the center of the biplot (0,0) and extended far to subdivide the biplot into sectors so that each site and cultivar is contained in only one sector. The biplot from the SREG₂ model indicates that ideal cultivars should have large PC1 (high mean yield) and near-zero PC2 (more stable) and the ideal sites should have large PC1 (high power to discriminate cultivars) and small PC2 (more representative of all sites). Crossa et al. (2002) demonstrated that the SREG₂ biplot represents the graph of the interaction variation due to noncrossover interaction (PC1) vs. the interaction variation due to crossover interaction (PC2).

Yan and Hunt (2002) used SREG₂ biplots for interpreting diallel experiments. This biplot approximates the GCA and SCA of lines and testers and also allows identification of an ideal tester. The average tester (AT) is a hypothetical tester with PC1 and PC2 equal to the average PC1 and PC2 scores across all testers. The AT is superimposed on the standard SREG₂ biplot so that all the lines can be projected onto the abscissa and ordinate of the AT such that the GCA of a line can be approximated by its projection onto the abscissa of the AT, whereas the SCA of a line can be approximated by its projection onto the ordinate of the AT. A line with a positive GCA will show its direction toward the positive direction of the abscissa of AT, whereas its SCA will be positive if its direction is the same as the positive direction of the ordinate of AT. An ideal tester will have large PC1 (high power to discriminate lines) and small PC2 (more representative of all testers) such that the best tester will be the one with the largest PC1 value and zero projection on the AT ordinate.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 
Line x Tester (Topcross) Evaluation
There were highly significant differences for most of the sources of variation (data not shown). Line x tester and line x tester x environment interaction for grain yield were highly significant, indicating different response of topcrosses across environments. Topcross performance is not only a consequence of gamete segregation and recombination but also of the environmental effect where cultivars are evaluated. Two subsets of environments did not show significant line x tester x environment interaction. Subset 1 included Villavicencio1 96A, Villavicencio2 96A, Villavicencio2 96B, and Quilichao 96B, and Subset 2 included Quilichao 96A and Alto Mayo 96B. We focus our discussion on the first of these two subsets of environments and also discuss results across all eight environments.

Analysis of the Eight Acid Soils with Significant Line x Tester x Environment Interaction
The PC1 of the SREG₂ biplot including all eight acid environments tended to separate SA4 lines (upper left quadrant) from SA3 and SA5, whereas PC2 separated SA5 (lower quadrants) from the others (Fig. 1). As pointed out by Yan and Hunt (2002) for the SREG₂ biplot, the best tester should have the longest PC1 vector of all testers and zero projection onto the AT ordinate; these two conditions allow selection of the most discriminating tester and the one that best represents all the others. On the basis of these two criteria, T3 is the best tester, followed by T1, T4, and T2 (Fig. 1). T1 and T3 combined better with lines from SA3 and SA5 and T2 and T4 with lines from SA3 and SA4.

View larger version (20K): [in this window] [in a new window]   Fig. 1. Biplot of the SREG2 for four testers (T1, T2, T3, and T4) and 43 lines (1-43) across eight acid soil environments. Lines 3-1 to 3-16 (underlined) are from population SA3, lines 4-17 to 4-19 and 4-22 to 4-35 are from population SA4, and lines 5-36 to 5-43 (bold) are from population SA5. The first digit of the line designation has been removed in the figure to avoid cluttering. The abscissa and the ordinate of the hypothetical average tester (AT) are marked with dot lines and the average value of PC1 and PC2 for AT is marked with a circle. The first and second multiplicative components are represented by PC1 and PC2, respectively.

The AMMI₂ biplot (Fig. 2) uses only the L x T interaction (SCA) for estimating the interaction parameters; therefore, it is expected to provide a precise and clear ranking of the lines for their SCA with each tester. For example, line 3-13 is near the center of the biplot so it has low SCA with T1, T2, and T3 and a positive SCA with T4 (0.22 Mg ha^-1). The separation of SA3, SA4, and SA5, based on SCA, is clear. The vertex of the polygon identifies the lines with the high SCA for each tester. For T2, line 4-24 had the highest SCA of all lines included in that sector (Fig. 2, Table 3). Line 4-25 was the best combiner with T4. T3 combined best with line 4-18 and T1 with lines 5-40 and 5-42.

View larger version (24K): [in this window] [in a new window]   Fig. 2. Biplot of the AMMI2 for four testers (T1, T2, T3, and T4) and 43 lines (1-43) across eight acid soil environments. Lines 3-1 to 3-16 (underlined) are from population SA3, lines 4-17 to 4-19 and 4-22 to 4-35 are from population SA4, and lines 5-36 to 5-43 (bold) are from population SA5. The first digit of the line designation has been removed in the figure to avoid cluttering. The first and second multiplicative components are represented by PC1 and PC2, respectively.

In summary, the SREG₂ and AMMI₂ biplots separated the lines into three groups corresponding to the populations SA3, SA4, and SA5, from which they were derived. The best two testers, T3 and T1, combined better with SA3 and SA5 than with SA4, whereas T2 and T4 combined better with SA3 and SA4. These patterns of responses were examined in subsets of environments for which no significant line x tester x environment interaction exists. It is desirable to select lines based both on GCA and SCA. However, selection of lines based only on SCA would be advisable if its GCA is negligible. For this purpose, the SREG₂ biplot should be more useful than the AMMI₂ biplot, as suggested by the results of this study. The SREG₂ would select lines that combine well with T3, requiring formation of only one synthetic.

Analysis of Four Acid Soils with No Significant Line x Tester x Environment Interaction
The SREG₂ biplot explained 79% of the line plus L x T interaction variation and showed similar patterns as those found when all eight acid soil environments where included. PC1 separated lines from SA4 (left quadrants) from those of SA3 and SA5, whereas PC2 separated SA5 from the others (Fig. 3). On the basis of criteria of discrimination power and representativeness of the testers (Yan and Hunt, 2002), T3 is the closest to the ideal tester followed by T1, T4, and T2 (Fig. 3). The five lines with the highest GCA, 3-9, 4-17, 3-13, 3-8, and 3-3, are located toward the direction of the abscissa of AT and their SCA with T3 (-0.06, 0.41, -0.19, 0.33, and 0.40 Mg ha^-1, respectively) is not in agreement with the Yan and Hunt (2002) hypothesis that the GCA of the lines should be approximated by their performance when crossed with T3 (Table 4). Lines 4-18 and 5-43 combined well with T3 showing a SCA of 0.77 and 0.82 Mg ha^-1, respectively, but both had intermediate values of GCA. Similar to when all eight sites were included, lines with the lowest GCA, such as lines 4-22, 4-27, 4-25, 4-23, and 5-40, had negative SCA with T3.

View larger version (20K): [in this window] [in a new window]   Fig. 3. Biplot of the SREG2 for four testers (T1, T2, T3, and T4) and 43 lines (1-43) across four acid soil environments. Lines 3-1 to 3-16 (underlined) are from population SA3, lines 4-17 to 4-19 and 4-22 to 4-35 are from population SA4, and lines 5-36 to 5-43 (bold) are from population SA5. The first digit of the line designation has been removed in the figure to avoid cluttering. The abscissa and the ordinate of the hypothetical average tester (AT) are marked with dot lines and the average value of PC1 and PC2 for AT is marked with a circle. The first and second multiplicative components are represented by PC1 and PC2, respectively.

View larger version (23K): [in this window] [in a new window]   Fig. 4. Biplot of the AMMI2 for four testers (T1, T2, T3, and T4) and the 43 lines across 4 acid soil environments. Lines 3-1 to 3-16 (underlined) are from population SA3, lines 4-17 to 4-19 and 4-22 to 4-35 are from population SA4, and lines 5-36 to 5-43 (bold) are from population SA5. The first digit of the line designation has been removed in the figure to avoid cluttering. The first and second multiplicative components are represented by PC1 and PC2, respectively.

Difficulties in choosing testers to identify superior lines arise mostly because of masking and interaction gene effects of testers and lines and genotype x environment interactions in the topcross performance. The topcrosses only measure the relative behavior of the lines under evaluation. If grain yield is the main selection criterion but agronomic traits are also important, it would be necessary to make a compromise between the selection of the highest yielding crosses and lines with good agronomic traits.

Synthetics Evaluation
Criteria to identify the best lines for each tester were high grain yield (among the top 15%), high and positive GCA of the lines, and lower or similar to the average days to 50% silk and ear height of their topcrosses (data not shown). Although no one line was selected by all four testers, two testers selected lines 3-13 and 4-17, which also had the highest GCA. Lines selected with T1 were 3-11, 3-13, 4-17, 4-29, 4-32, and 5-42. Average grain yield of selected topcrosses was 4.1 Mg ha^-1 with GCA ranging between 0.01 Mg ha^-1 (line 5-42) and 0.63 Mg ha^-1 (line 3-13). Average number of days to 50% silk was 64 and ear height of the selected topcrosses 70 cm.

Using T2 as tester, lines 3-2, 3-4, 3-5, 3-9, 3-13, and 4-34 were selected to form the synthetic T2S. Average grain yield of selected topcrosses was 4.2 Mg ha^-1 and the GCA ranged from 0.18 Mg ha^-1 (line 4-34) to 0.63 Mg ha^-1 (line 3-13). Average number of days to 50% silk was 64 and ear height of the selected topcrosses 69 cm. With T3 as tester, lines 3-3, 3-4, 3-7, 3-14, 4-17, and 4-18 were selected. The GCA values ranged between 0.11 Mg ha^-1 (line 3-14) and 0.53 Mg ha^-1 (line 4-17). The average grain yield of selected topcrosses was 3.9 Mg ha^-1, the number of days to 50% silk was 63, and ear height 61 cm. Lines 3-2, 3-3, 3-9, 3-13, 4-17, and 4-25 were selected with T4. GCA of selected lines ranged from 0.44 Mg ha^-1 (line 3-3) to 0.63 Mg ha^-1 (line 3-13). The average grain yield of selected topcrosses was 4.5 Mg ha^-1, the number of days to 50% silk was 63, and ear height 72 cm.

Performance of synthetics differed with environment and with the trait under consideration. In nonacid soils, grain yield, plant and ear height, root lodging, and grain texture were higher than in acid soils (Table 5). There was no difference among environments for days to 50% silk and grain texture. Stalk lodging was higher in the acid than in the nonacid soils.

In nonacid soils, the highest yielding synthetic was T3S (4.79 Mg ha^-1) followed by T1S (4.52 Mg ha^-1), T4S (4.45 Mg ha^-1), and TS2 (4.12 Mg ha^-1). Grain texture was the most contrasting trait between the highest and lowest yielding synthetics, averaging 3.1 and 1.9, respectively. One explanation for this difference could be that the T3, a semident inbred line, combined better with flint lines (four of them derived from SA3) used to develop the synthetic T3S.

SREG₂ analysis for the 12 acid soil environments showed that the two principal components accounted for 94% of the synthetics x environment interaction sum of squares (data not shown). Consequently, a biplot including PC1 and PC2 would provide a succinct description of the data. The biplot could also be used for comparing different synthetics in one environment, for comparing the performance of synthetics in different environments, and for mega-environment identification. We used biplot to compare T3S and T1S that had the best performance in acid and nonacid soils (Fig. 5). Synthetic T3S was the best performer in acid soil environments E1, E6, E10, E12, E14, and E19, whereas T1S was the best yield performer in E7, E22 and E20. In E5, T4S was the best synthetic.

View larger version (19K): [in this window] [in a new window]   Fig. 5. Biplot of the SREG2 for four synthetics (T1S, T2S, T3S, and T4S) and 12 acid soil environments (E1, E5, E6, E7, E9, E10, E12, E13, E14, E17, E19, and E20). The first and second multiplicative components are represented by PC1 and PC2, respectively.

View larger version (17K): [in this window] [in a new window]   Fig. 6. Biplot of the SREG2 for four synthetics (T1S, T2S, T3S, and T4S) and nine nonacid soil environments (E2, E3, E4, E8, E11, E15, E16, E18, E21). The first and second multiplicative components are represented by PC1 and PC2,

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

Synthetics generated with lines identified by each tester were evaluated across 12 acid and nine nonacid soil environments. T3S and T1S were the highest-yielding synthetics in acid (2.63 and 2.59 Mg ha^-1, respectively) and nonacid (4.79 and 4.52 Mg ha^-1, respectively) soil environments. The SREG₂ biplot showed that for acid soils, T3S had higher yield in E1, E6, E10, E12, E14, and E19, whereas T1S was the best yielder in E7, E22 and E20. In both acid and nonacid environments, synthetics developed based on the topcross performance of lines with narrow-based tester (T3), an S3 line, was superior to the others. While the results of this study are inconclusive as to the type (narrow vs. broad genetic base) of tester to use in selecting lines for making synthetics, AMMI and SREG models seem to provide an effective tool to visualize and study GCA and SCA of genotypes.

	ACKNOWLEDGMENTS

 
The authors express their sincere appreciation to National Maize Program Coordinators and maize researchers from Bolivia, Brazil, Colombia, India, Paraguay, and Peru and to their colleagues from CIMMYT (Mexico) and Thailand for their collaboration in conducting the trials. Thanks are due to Dr. Weikai Yan for useful suggestions and for running some analyses using the GGE biplot Software.

Received for publication June 4, 2002.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

 

• Allison, J.C.S., and R.W. Curnow. 1966. On the choice of tester parent for the breeding of synthetic varieties of maize (Zea mays L.). Crop Sci. 6:641–644.
• Castellanos, J.S., A.R. Hallauer, and H.S. Cordova. 1998. Relative performance of testers to identify elite lines of corn (Zea mays L.). Maydica 43:217–226.
• CIMMYT. 1999. CIMMYT 1997/98 World Maize Facts and Trends; Maize Production in Drought-Stressed environments: Technical Options and Research Resource Allocation. Mexico D.F.
• Crossa, J. 1990. Statistical analyses of multilocation trials. Adv. in Agron. 44:55–85.
• Crossa, J., and P.L. Cornelius. 1997. Sites regression and shifted multiplicative model clustering of cultivar trial sites under heterogeneity of error variances. Crop Sci. 37:405–415.
• Crossa, J., P.L. Cornelius, and W. Yan. 2002. Biplot of linear-bilinear models for studying crossover genotype x environment interaction. Crop Sci. 42:619–633.[Abstract/Free Full Text]
• Cornelius, P.L., J. Crossa, and M.S. Seyedsadr. 1996. Statistical tests and estimators of multiplicative models for genotype-by-environment interaction. p. 199–234. In M.S. Kang and H.G. Gauch (ed.). Genotype-by-environment interaction. CRC Press, Boca Raton, FL.
• Darrah, L.L. 1985. Evaluation of population improvement in the Kenya maize breeding methods study. P. 160–175. In To feed ourselves. Proc. First Eastern, Central and Southern Africa Regional Workshop. Lusaka, Zambia. CIMMYT, Mexico, D.F.
• Gauch, H.G. 1988. Model selection and validation for yield trials with interaction. Biometrics 44:705–715.[Web of Science]
• Granados, G., S. Pandey, and H. Ceballos. 1995. Registration of acid soil tolerant maize populations SA-3 and SA-8. Crop Sci. 35:1236.[Free Full Text]
• Hallauer, A.R. 1975. Relation of gene action and type of tester in maize breeding procedures. Proc. Annu. Corn Sorghum Res. Conf. 30:150–165.
• Hallauer, A.R., and J.B. Miranda. 1988. Quantitative genetics in maize breeding. Iowa State Univ. Press. Ames, USA.
• Hayes, H.K., and R.J. Garber. 1919. Synthetic production of high protein corn in relation to breeding. Agron. J. 11:309–318.[Free Full Text]
• Hoegemeyer, T.C., and A.R. Hallauer. 1976. Selection among and within full-sib families to develop single-crosses of maize. Crop Sci. 16:76–81.[Abstract/Free Full Text]
• Horner, E.S., W.H. Lundy, M.C. Lutrick, and W.H. Chapman. 1973. Comparison of three methods of recurrent selection in maize. Crop Sci. 13:485–489.[Abstract/Free Full Text]
• Horner, E.S., M.C. Lutrick, W.H. Chapman, and F.G. Martin. 1976. Effect of recurrent selection for combining ability with a single-cross tester in maize. Crop Sci. 16:5–8.
• Horner, E.S., E. Magloire, and J.A. Morera. 1989. Comparison of selection for S2 progeny vs. testcross performance for population improvement in maize. Crop Sci. 29:868–874.[Abstract/Free Full Text]
• Hull, H.F. 1945. Recurrent selection for specific combining ability in corn. J. Am. Soc. Agron. 37:134–145.
• Jenkins, M.T. 1935. The effect of inbreeding and of selection within inbred lines of maize upon the hybrids made after successive generations of selfing. Iowa State J. Sci. 3:429–450.
• Jenkins, M.T., and A.M. Brunson. 1932. Methods of testing inbred lines of maize in crossbred combinations. J. Am. Soc. Agron. 24:523–530.
• Lonnquist, J.H. 1949. The development and performance of synthetic varieties of corn. Agron. J. 41:153–156.[Free Full Text]
• Lonnquist, J.H. 1961. Progress from recurrent selection procedure for the improvement of corn populations. Nebraska Agric. Exp. Stn. Res. Bull. 197.
• Matzinger, D.F. 1953. Comparison of three types of testers for the evaluation of inbred lines of corn. Agron. J. 45:493–495.[Free Full Text]
• Menz, M.A., A.R. Hallauer, and W.A. Russell. 1999. Comparative response of two reciprocal recurrent selection methods in BS21 and BS22 maize populations. Crop Sci. 39:89–97.[Abstract/Free Full Text]
• Pandey, S., H. Ceballos, R. Magnavaca, A.F.C. Bahía Filho, J. Duque-Vargas, and L.E. Vinasco. 1994. Genetics of tolerance to soil acidity in tropical maize. Crop Sci. 34:1511–1514.[Abstract/Free Full Text]
• Pandey, S., H. Ceballos, and G. Granados. 1995. Registration of four tropical maize populations with acid soil tolerance: SA-4, SA-5, SA-6, and SA-7. Crop Sci. 35:1230–1231.[Free Full Text]
• Rawlings, J.O., and D.L. Thompson. 1962. Performace level as criterion for the choice of maize testers. Crop Sci. 2:217–220.[Free Full Text]
• Russell, W.A., and S.A. Eberhart. 1975. Hybrid performance of selected maize lines from reciprocal recurrent selection and testcross selection programs. Crop Sci. 15:1–4.
• Russell, W.A., D.J. Blackburn, and K.R. Lamkey. 1992. Evaluation of modified reciprocal recurrent selection procedure for maize improvement. Maydica 37:61–67.
• Sprague, G.F., and M.T. Jenkins. 1943. A comparison of synthetic varieties, multiple crosses, and double crosses in corn. J. Am. Soc. Agron. 35:137–147.
• The Maize Program. 1999. Development, maintenance, and seed multiplication of open-pollinated maize varieties- 2nd edition. Mexico, D.F.: CIMMYT.
• Walejko, R.N., and W.A. Russell. 1977. Evaluation of recurrent selection for specific combining ability in two open-pollinated maize cultivars. Crop Sci. 17:647–651.[Abstract/Free Full Text]
• Yan, W., L.A. Hunt, Q. Sheng, and Z. Szlavnies. 2000. Cultivar evaluation and mega-environment investigation based on the GCE biplot. Crop Sci. 40:597–605.[Abstract/Free Full Text]
• Yan, W., P.L. Cornelius, J. Crossa, and L.A. Hunt. 2001. Two Types of GGE Biplots for Analyzing Multi-Environment Trial Data. Crop Sci. 41:656–663.[Abstract/Free Full Text]
• Yan, W., and L.A. Hunt. 2002. Biplot of Diallel Data. Crop Sci. 42:21–30.[Abstract/Free Full Text]
• Zobel, R.W., M.J. Wright, and H.G. Gauch. 1988. Statistical analysis of a yield trial. Agron. J. 80:388–393.[Abstract/Free Full Text]

This article has been cited by other articles:

				W. Yan, M. S. Kang, B. Ma, S. Woods, and P. L. Cornelius GGE Biplot vs. AMMI Analysis of Genotype-by-Environment Data Crop Sci., March 1, 2007; 47(2): 643 - 653. [Abstract] [Full Text] [PDF]

				H. G. Gauch Jr. Statistical Analysis of Yield Trials by AMMI and GGE Crop Sci., May 18, 2006; 46(4): 1488 - 1500. [Abstract] [Full Text] [PDF]

				A. J. de la Vega and S. C. Chapman Multivariate Analyses to Display Interactions between Environment and General or Specific Combining Ability in Hybrid Crops Crop Sci., February 24, 2006; 46(2): 957 - 967. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (5) Citing Articles via Google Scholar Google Scholar Articles by Narro, L. Articles by Salazar, F. Search for Related Content PubMed Articles by Narro, L. Articles by Salazar, F. Agricola Articles by Narro, L. Articles by Salazar, F. Related Collections Germplasm Enhancement Plant and Soil Interactions Maize Crop Genetics