# Correlations and QTL correspondence between line per se and testcross performance for agronomic traits in four populations of European maize.

TESTCROSS PERFORMANCE of experimental lines is the prime selection criterion in hybrid breeding of maize. An indirect improvement of TP in early selfing generations by selecting for LP is economically advantageous, with a high positive correlation between LP and TP. Experimental estimates of the genotypic correlation between LP and TP, [r.sub.g] (LP, TP), vary considerably for different crops, traits, and selfing generations, in maize, for traits showing small heterotic effects such as grain moisture, ear length, or days to flower, estimates of [r.sub.g] (LP, TP) were medium to high. However, they were generally low for the highly heterotic trait, grain yield (for review see Hallauer and Miranda. 1981; Seitz, 1989).In early studies, low values of [r.sub.g] (LP, TP) observed for grain yield in advanced selfing generations were most probably due to recessive genes with detrimental effect in homozygous state (Genter and Alexander, 1966). Overdominance, epistasis, and linkage, or the combined action of these factors may also decrease the correlation (Schnell, 1961: Smith, 1986). For heterotic traits, estimates of [r.sub.g] (LP, TP) with experimental lines from early selfing generations may be reduced because different levels of heterozygosity affect LP but not TP.

Assuming absence of linkage and epistasis, Smith (1986) demonstrated theoretically that low correlations between LP and TP can be fully explained by a model with additive and dominance effects. Thus, with biallelism and allele frequencies of 0.5 in a set of lines derived froth a population in Hardy-Weinberg equilibrium, [r.sub.g] (LP, TP) is a linear function of the proportion of QTL at which the inbred tester is homozygous for the favorable allele. As the latter increases, [r.sub.g] (LP, TP) decreases due to a reduced genotypic variance ([[sigma].sup.2.sub.g]) for TP. Thus, the ratio of [[sigma].sup.2.sub.g] for LP and TP provides a crude estimate of the proportion of dominant favorable alleles fixed in the tester.

The importance of epistatic interactions relative to the masking effect of dominant tester alleles for the reduction of [r.sub.g] (LP, TP) can be assessed from quantitative genetic parameters. Differences among testcross means and changes in the ratios of segregation variances from different testcross generations are expected in the presence of linked epistatic effects (Melchinger, 1987). Such differences are not expected to occur if the masking effect of dominant tester alleles prevails.

While estimates of [r.sub.g] (LP, TP) rely on the net effect of all QTL influencing LP and TP for a given trait, QTL analyses provide a tool to clarify the genetic basis of this correlation at the molecular level. The proportion of common QTL for LP and TP was largest for plant and ear height with an unrelated tester, and smallest for grain yield with a related tester (Austin et al., 2000). This was in accordance with the magnitude of genotypic correlations between LP and TP estimated for these traits. However, comparative QTL studies for LP and TP (Guffy et al., 1988; Beavis et al., 1994; Groh et al.; 1998; Kerns el al., 1999; Austin et al., 2000: Mechin et al., 2001) have so far not targeted the causes of the low genotypic correlations estimated in previous studies.

In this study, we evaluated four populations derived from three crosses of elite inbreds of European flint maize in different selfing generations ([F.sub.3] to [F.sub.6] lines) for both LP and TP. Our objectives were to (i) obtain reliable estimates of the correlation between LP and TP for five agronomic traits, (ii) examine possible causes for their magnitude by comparing genetic variances as well as the proportion of common QTL for LP and TP across populations and traits, and (iii) determine the gene action of QTL identified for LP and their value for the prediction of TP.

MATERIALS AND METHODS

Plant Materials

Four early maturing elite European flint lines KW1265, D146, D145, and KW1292, subsequently referred to as A, B, C, and D, were used as parents (P1 and P2) to produce four populations of 380 [F.sub.2:3] lines (A x [B.sup.I]), 120 [F.sub.4:5] lines (A x [B.sup.III]), and 131 (A x C) and 135 (C x D) [F.sub.3:4] lines. Superscripts I and III denote two different samples of the cross A x B according to the notation used in Mihaljevic et al. (2004). Testcross seed was produced in isolation plots by mating the unrelated dent inbred tester (KW5361, subsequently referred to as T2 in the notation of Schon et al., 1994), as pollinator to a random sample of 40 plants from each of the [F.sub.n] lines ([F.sub.2:3] lines in A x [B.sup.I]: [F.sub.4:5] lines in A x [B.sup.III]: [F.sub.3:4] lines in A x C and C x D) as well as to the parent lines A, B, C, and D. Lines of each population except for A x [B.sup.I] were further selfed, and the resulting [F.sub.4:6] lines of the cross A x [B.sup.III] as well as [F.sub.3:5] lines of the crosses A x C and C x D were evaluated for LP (Table 1). For A x [B.sup.I] however, seed for evaluation of LP was produced by chain crossing of 20 plants of each [F.sub.2:3] line.

Field Experiments

The lines were evaluated for LP in separate experiments in the Upper Rhine valley. The experimental design employed was a 30 x 10(A x [B.sup.I]) and 15 x 10 (A x [B.sup.III], A x C, and C x D) [alpha]-design (Patterson and Williams, 1976) with two replications and one-row plots overplanted and later thinned to obtain a final stand of 8.7 plants [m.sup.-2] in all experiments. All trials were conducted at five different environments. For populations A x [B.sup.III], A x C, and C x D, data from one environment was excluded from the combined analysis across environments due to severe drought stress (Table 1). The corresponding testcrosses evaluated in the same environment were far less affected by the unfavorable weather conditions. Each year-site combination was treated as an environment in subsequent statistical analyses.

The corresponding testcross progenies of the populations A x [B.sup.I] A x [B.sup.III] A x C, and C x D were evaluated for TP in separate experiments in the Upper Rhine valley, Lower Bavaria, and France, as described by Melchinger et al. (1998) and Mihaljevic et al. (2004). Population A x [B.sup.I] was grown in four, the remaining three populations (A x [B.sup.III], A x C, and C x D) in five environments (Table 1). Because of insufficient quantities of seeds, fewer lines were tested for TP than LP in A x [B.sup.III] A x C, and C x D. The experimental design was a 40 x 10 (A x [B.sup.I]) or a 15 X 10 [alpha]-design (A x [B.sup.III], A x C, and C x D) with two replications and two-row plots overplanted and later thinned to obtain a final stand of 8.7 plants [m.sup.-2] in the Upper Rhine valley (two environments) and 11 plants [m.sup.-2] in the other regions (three environments). All experiments were machine planted and harvested as grain trials with a combine. In the case of A x [B.sup.III], A x C, and C x D, one test environment was in common for LP and TP but none in the case of A x [B.sup.I] (Table 1).

Data were analyzed for the following traits: grain yield (Mg [ha.sup.-1]) adjusted to 155 g [kg.sup.-1] grain moisture, grain moisture (g [kg.sup.-1]) at harvest, kernel weight expressed as grams per 1000 kernels determined from four samples of 50 kernels from each plot, protein concentration in grain (g [kg.sup.-1]) measured by near-infrared reflectance spectroscopy as described by Melchinger et al. (1986), and plant height (cm) on a plot basis as the distance from the soil level to the lowest tassel branch.

RFLP Marker Genotyping and Linkage Map Construction

The procedures for RFLP assays were described by Schon et al. (1994). A total of 89 RFLP marker loci was used to genotype 344 parental [F.sub.2] plants of the 380 [F.sub.2:3] lines from cross A x [B.sup.I], and 151 RFLPs were used to genotype parental [F.sub.4] plants of 120 [F.sub.4:5] or [F.sub.4:6] lines (A x [B.sup.III]) (Table 1). A total of 104 and 122 RFLPs was mapped with 131 and 140 [F.sub.3] lines derived from cross A x C and C x D, respectively. The joint linkage map reported by Mihaljevic et al. (2004) comprising data of the four populations plus an additional population (independent sample A x [B.sup.II] of cross A x B), formed the basis of all further analyses. The joint map is available at http://www.maizegdb.org (verified 3 Sept. 2004).

Agronomic Data Analyses

Adjusted entry means and effective error mean squares derived from ANOVAs of each environment (year-site-combination) were used to calculate the combined ANOVAs and ANCOVAs for each experiment. Quantitative genetic parameters, such as variance components and heritabilities, were estimated as described by Melchinger et al. (1998). An approximative F test was used to test whether the genotypic variance ([[??].sup.2.sub.g]) for LP was larger than [[??].sup.2.sub.g] for TP. Degrees of freedom for the one-tailed F test were calculated according to Satterthwaite (1946). Corresponding F tests were also employed to compare [[??].sup.2.sub.g] from different generations (A x [B.sup.I] and A x [B.sup.III]).

Phenotypic, [r.sub.p] (LP, TP), and genotypic, [r.sub.p] (LP, TP), correlations were calculated between LP and TP using only the common lines (Table 1). Both types of correlation coefficients were calculated using the MANOVA estimators of adjusted entry means described by Liu et al. (1997). Here, the phenotypic covariance was used as an estimator of the genotypic covariance, assuming the covariance of genotype x environment interactions to be negligible. Empirical 95% confidence intervals of the correlation coefficients were estimated by 2000 bootstrap samples according to Liu et al. (1997).

QTL Analyses

QTL mapping and estimation of their effects were performed with PLABQTL (Utz and Melchinger, 1996) employing CIM by the regression approach (Haley and Knott. 1992). The additive genetic model underlying the analysis of TP was described in detail by Utz et al. (2000). For analyses of LP of the F lines, the following model was employed:

[Y.sub.j] = m + [b.sup.*.sub.l][x.sup.*.sub.ajl] + [b.sup.*.sub.2][x.sup.*.sub.djl] + [[summation over k] [b.sub.k][x.sub.jk] + [[epsilon].sub.j],

where [Y.sub.j] denotes the phenotypic trait mean of the jth [F.sub.n] line averaged across environments: m is the phenotypic trait mean of [F.sub.n] lines with genotype qq at the lth putative QTL: [b.sup.*.sub.1] and [b.sup.*.sub.2] are the additive (a) and the dominance (d, estimated only for [F.sub.2:3] lines of A x [B.sup.I]) effects as defined by Falconer and Mackay (1996, p. 112) at the putative QTL in the marker interval l with flanking markers l' and l". [x.sup.*.sub.ajl], and [x.sup.*.sub.djl] are the conditional expectations of the dummy variables [[THETA].sub.ajl] and [[THETA].sub.djl] given the observed genotypes at the flanking marker loci l' and l", where [[THETA].sub.ajl] assumes values 0, l, or 2, and [[THETA].sub.djl] assumes values 0, 0.5, or 0 if the genotype of the parental [F.sub.n] individual at the putative QTL is qq, Qq, or QQ. respectively. [[THETA].sub.djl] is 0.5 rather than 1 for heterozygotes Qq, because phenotypic traits were evaluated in A x [B.sup.I] for [F.sub.2:3] lines and not [F.sub.2] plants, which reduces the dominance effect by one half. [b.sub.K] is the partial regression coefficient of phenotype [Y.sub.j] on the kth (selected) marker: [x.sub.jk] is a dummy variable (cofactor) taking values 0, 1, or 2, depending on whether the marker genotype of the parental [F.sub.n] individual j at marker locus k is homozygous qq, heterozygous Qq, or homozygous QQ, respectively. [[epsilon].sub.j] is a residual variable for the jth [F.sub.n] line.

Cofactors were selected by stepwise regression according to Miller (1990, p. 49) with an "F-to-enter" and "F-to-delete" value of 3.5. Testing for presence of a putative QTL in an interval by a likelihood ratio (LR) test was performed by using a LOD threshold of 2.5 (= 0.217 LR). Estimates of QTL positions were obtained at the point where the LOD score assumed its maximum value in the region under consideration. For each population, the proportion of the phenotypic variance ([[??].sup.2.sub.p]) explained by a single QTL was determined as the square of the partial correlation coefficient ([R.sup.2]). Estimates of the additive effects (and dominance effects for A x [B.sup.I]) of each putative QTL for LP and their partial [R.sup.2] were obtained by fitting a model including all QTL for the respective trait simultaneously. The proportion p of the genotypic variance explained by all detected QTL was also determined from this model for each data set (DS) as [[??].sub.DS] by dividing the adjusted total [R.sup.2] ([R.sup.2.sub.adj]) by the heritability ([h.sup.2]) as described by Utz et al. (2000).

Five-fold standard cross validation implemented in PLABQTL was used to obtain asymptotically unbiased estimates of p (Utz et al., 2000). For each population, the DS comprising the entry means across environments was divided into five genotypic subsamples. Four of these were combined in an estimation set (ES) for QTL detection and estimation of genetic effects, whereas the remaining fifth subsample was used as a test set (TS) to validate the predictions gained from ES and calculate [[??].sub.TS.ES] by correlating data predicted on the basis of QTL estimates in ES with those observed in the TS. Five different cross validation runs are possible by permutating the respective subsamples. A total of 1000 replicated cross validation runs was performed with 200 randomizations for assigning genotypes to the respective subsamples. The median [[??].sub.TS.ES] was obtained from [[??].sub.TS.ES] across the 1000 runs.

Congruency of QTL for Line Per Se and Testcross Performance

We assessed congruency of QTL detected for LP and TP of a particular trait in the same population. Two approaches were used for this purpose: (i) counting the number of congruent QTL, whereby individual QTL were considered congruent if their estimated map position was within a 20-cM distance, irrespective of the sign of estimated QTL effects, and (ii) the genotypic correlation between predicted and observed testcross performance, [r.sub.g] ([M.sub.LP], [Y.sub.TP]), where [M.sub.LP] is the predicted value of a line based on the QTL positions and effects estimated from QTL for LP in a given population, and [Y.sub.TP] is the observed TP of this line (Utz et al., 2000).

RESULTS

Segregation and Linkage of RFLP Markers

The results of the RFLP analyses have been reported previously (Mihaljevic et al., 2004). The joint linkage map of the populations A x [B.sup.I], A x [B.sup.II], A x [B.sup.III], A x C. and C x D spanned a total of 1138 cM. This joint map covered about 70% of the genome from the original map of A x [B.sup.I] published by Schon et al. (1994).

Agronomic Trail Analysis for Line Per Se Performance

The means of parents P1 and P2 differed significantly (P < 0.01) for all traits in all populations except for plant height in A x [B.sup.I] and A x C, grain moisture in A x [B.sup.III] and C x D, and grain yield and kernel weight in C x D (Table 2). An orthogonal contrast between the mean performance of the parent lines (P) and the population mean of the [F.sub.n] lines ([[??].sub.n]) was highly significant (P < 0.01) for all traits in A x [B.sup.I], in A x C for grain yield, and in C x D for plant height only. For grain yield, kernel weight, and plant height, [??] was significantly smaller than [[??].sub.n] in all of these cases. In contrast, [??] vs. [[??].sub.n] was not significant for any trait in A x [B.sup.III].

Genotypic variances for LP were highly significant for all traits in all four populations (Table 3). As expected from quantitative genetic theory, the lines in A x [B.sup.I] from an early selfing generation had a significantly (P < 0.05) smaller [[??].sup.2.sub.g] than lines in A x [B.sup.III] from an advanced selfing generation for all traits except grain yield. For comparison, [[??].sup.2.sub.g] for TP was significantly smaller in A x [B.sup.I] than in A x [B.sup.III] for all traits.

Estimates of genotype x environment interaction variance ([[??].sup.2.sub.ge]) for LP were significantly greater than zero (P < 0.01) and consistently smaller than [[??].sup.2.sub.g] for all traits in all populations (Table 3). Heritabilities ([??].sup.2]) were high for all traits ranging from 0.88 to 0.95 across traits and populations.

Comparison of Line Per Se and Testcross Performance

Mihaljevic et al. (2004) reported results of testcross progeny analysis for A x [B.sup.I], A x [B.sup.III], A x C, and C x D. In all four crosses, the population mean [bar.F.sub.n] for LP was lower than [[??].sub.n] for TP for all traits except grain moisture and protein concentration (Table 2). The range of [F.sub.n] lines for LP was larger than for TP in all populations and for all traits (data not shown).

As expected, estimates of [[sigma].sup.2.sub.g] for LP were significantly greater than those for TP in all populations and for all traits. Estimates of [[sigma].sup.2.sub.ge] also were generally greater for LP than for TP, except for grain yield in A x [B.sup.III] and A x C (Table 3).

Phenotypic correlations between LP and TP, [r.sub.p] (LP, TP), were low for grain yield, but significant in all populations (Table 4). For the other traits, [r.sub.p] (LP, TP) values were intermediate (0.40 < [r.sub.p] < 0.75). Genotypic correlations between LP and TP, [r.sub.g] (LP, TP), were significant and always greater than [r.sub.p] (LP, TP) across all traits and populations. Estimates of [r.sub.g] (LP, TP) ranged from 0.28 to 0.56 for grain yield and from 0.52 to 0.87 for the other four traits (Table 4).

QTL Analyses of Line Per Se Performance

Results from QTL analyses for LP of all four populations based on the joint map are presented here for means across environments (Table 4). Detailed information on the position and magnitude of effects of individual QTL can be obtained at http://www.maizegdb.org. In the large population A x [B.sup.I], substantially more QTL were detected than in the smaller populations. The number of congruent QTL detected across the four populations was low. Most QTL found for A x [B.sup.III] were also found in A x [B.sup.I]. Only one QTL with dominant gene action was detected for grain yield in A x [B.sup.I]. The QTL results for TP were reported previously (Mihaljevic et al., 2004).

Comparison of QTL for Line Per Se and Testcross Performance

Across all five traits in A x [B.sup.I], 21 out of 44 QTL detected for LP were found within a 20-cM distance from QTL detected for TP (Table 4). The relationship between the number of common QTL for LP and TP and the total number of QTL detected for LP was lowest for grain yield. In the advanced generation of cross A x B (A x [B.sup.III]), five out of eight QTL detected for LP were common to QTL detected for TP across all five traits. Out of 24 QTL detected in A x C for LP, 10 QTL were within a 20-cM distance to QTL detected for TP for the same trait. In C x D, six out of 24 QTL detected for LP across all five traits were common to QTL detected for TP of the same traits.

Estimates of the genotypic correlation between predicted and observed testcross performance, [r.sub.g] ([M.sub.LP], [Y.sub.TP]), varied considerably across populations for all traits (Table 4). For grain yield, [r.sub.g] ([M.sub.LP], [Y.sub.TP]) was highest in C x D and lowest in A x [B.sup.I], which was unexpected considering the difference in population size. For the other traits, [r.sub.g] ([M.sub.LP], [Y.sub.TP]) was highest (0.61) for plant height [B.sup.III] A x [B.sup.I]. and lowest (0.15) for grain moisture in A x [B.sup.III].

The number of common QTL generally was not reflected in the magnitude of [r.sub.g] ([M.sub.LP], [Y.sub.TP]) (Table 4). For grain moisture and plant height in A x [B.sup.III] and kernel weight in C x D, significant correlations [r.sub.g] ([M.sub.LP], [Y.sub.TP]) were detected in spite of zero common QTL between LP and TP. The correlations [r.sub.g] (LP, TP) and [r.sub.g], ([M.sub.LP], [Y.sub.TP]) corresponded well for grain yield. This was not the case for the other four traits, where [r.sub.g] (LP, TP) was substantially higher than [r.sub.g] ([M.sub.LP], [Y.sub.TP]) except for grain moisture and plant height in C x D.

DISCUSSION

Correlations between Line Per Se and Testcross Performance

The magnitude of the genotypic correlation between LP and TP is an indicator of the prospects of simultaneously improving commercial hybrids as well as their inbred parents. In maize, a wide range of estimates for phenotypic and genotypic correlations between LP and TP was reported in the literature, depending on the trait investigated (for review see Hallauer and Miranda, 1981). In our study, genotypic correlations estimated for LP and TP across four populations derived from crosses within the European flint pool were comparable with those obtained for U.S. dent material. Lowest estimates were found for grain yield [[r.sub.g] (LP, TP) = 0.28-0.56]. As expected for traits with higher heritability and presumably mainly additive gene action, such as grain moisture, kernel weight, protein concentration, and plant height, estimates of the respective correlation were generally high [[r.sub.g] (LP, TP) > 0.7] across all four populations with only a few exceptions.

Genotypic correlations were higher than phenotypic correlations for all traits and populations. As expected from theory, when LP and TP are evaluated in different environments, the difference between the genotypic and the phenotypic correlations is a function of the heritability for LP and TP for the respective cross. For grain yield, heritability estimates for TP were smaller compared with LP mainly due to the reduced genotypic variance, slightly lower testing intensity (A x [B.sup.I]), or a higher [[??].sup.2.sub.ge] for TP than LP. The [[??].sup.2.sub.ge] of TP in A x [B.sup.III] and A x C were larger than in C x D for grain yield, although all three populations were tested in the same five environments. Thus, TP of lines from C x D seems to be more robust against environmental changes than TP of lines from A x B and A x C. For the other four traits, [[??].sup.2.sub.ge] was consistently larger for LP than for TP, resulting in similar heritability estimates despite a significant decrease in [[??].sup.2.sub.g] for TP.

The decrease in [[??].sup.2.sub.g] for TP compared with LP can be used as an indication of the strength (performance level or gene frequency) of the tester and of the expected genotypic correlation between LP and TP. For a tester, which carries dominant alleles masking the effect of the segregating alleles at many loci, [[sigma].sup.2.sub.g] for TP is decreased and correlations are expected to be lower. Smith (1986) showed that with complete dominance and a gene frequency of 0.5 in the population under study, the genotypic correlation between LP and TP is inversely proportional to the ratio of [[sigma].sup.2.sub.g] for LP and TP. For the biallelic case and an above average inbred tester from the same population, the genotypic correlation between LP and TP would be 0.5 or lower (Smith, 1986).

Considering all four populations and all traits, no significant association was found between the ratio of the two variances and [r.sub.g] (LP, TP). The ratio of [[??].sup.2.sub.g] for LP vs. TP varied from 2.0 (kernel weight in A x C) to 7.2 (grain yield in C x D). Highest variance ratios were obtained for grain yield, as expected for a trait presumably controlled by many genes with large dominance effects, but only in A x [B.sup.I] and C x D. Despite surprisingly low ratios for grain yield in A x [B.sup.III] and A x C (2.1 and 2.6, respectively), genotypic correlations in these two crosses were intermediate.

Reasons can be given for the difficulties in predicting genotypic correlations from this ratio. First, Smith (1986) had assumed the biallelic case with the tester originating from the same population as the test units. In our study, however, the inbred tester originated from the opposite dent pool and was known for its excellent combining ability for yield with the flint pool.

Second, lines in all four populations had different levels of inbreeding. Different from TP, LP of an [F.sub.3] line for a heterotic trait like grain yield is affected by the heterozygosity level of its parental [F.sub.2] plant. However, despite a wide range in heterozygosity at marker loci (28.3 to 75.4%) in the [F.sub.2] plants of population A x [B.sup.I], this parameter showed only a weak correlation ([r.sub.g] = 0.13, P < 0.05) with LP for grain yield (data not shown). These results were in accordance with the detection of only one out of nine QTL with dominant gene action for LP of grain yield.

Third, the low precision in estimating genotypic correlations (see large confidence intervals of the estimates presented in Table 4) could be a further explanation for the lack of association between the magnitude of genotypic correlations and the reduction in genotypic variance in the testcrosses. For grain yield and population A x [B.sup.III] for example, the 95% confidence interval for the estimate of [r.sub.g] (LP, TP) ranged from 0.07 to 0.87. Highest precision, that is, smallest confidence intervals, was obtained for plant height and grain moisture in population A x [B.sup.I], with the highest number of common lines tested for both LP and TP (N = 280). This is in agreement with results from Liu et al. (1997), who found that the heritability of the trait and sample size had a strong effect on the precision of estimates of genotypic correlations.

QTL Detected for Line Per Se and Testcross Performance

When comparing QTL mapping results for LP and TP across populations, with the exception of grain yield, generally fewer QTL were detected in populations A x [B.sup.III], A x C. and C x D than in A x [B.sup.I], reflecting the decreased power of QTL detection with smaller sample sizes. The same was true for the proportion of [[??].sup.2.sub.g] explained by QTL estimated from cross validation. For TP and LP similar numbers of QTL were detected in a given population for all traits except grain yield. The higher heritabilities and the slightly larger sample sizes in LP trials as compared with TP trials did not have a significant effect on the number of QTL detected. For grain yield, however, substantially fewer QTL were detected for TP of population A x [B.sup.I] than in the other populations and for LP. In addition to genetic factors, sampling could be a reason for these results. With cross validation, Utz et al. (2000) showed for TP of population A x [B.sup.I] that the number of detected QTL for grain yield can vary from zero to eight, depending on the genotypic sample used for QTL detection. In cross validation of LP data from A x [B.sup.I], the number of QTL detected for grain yield varied from 3 to 11.

Evidence for genetic factors, such as dominance and epistasis, which influence both heterosis and the correlation between LP and TP, should have been provided by the QTL analysis. It was surprising, however, that in the LP of population A x [B.sup.I], only one of the nine QTL exhibited dominant gene action for grain yield, and only one pair of marker loci had a significant additive x additive epistatic effect. In the smaller populations and across all traits, epistatic effects were rarely detected. One reason for these results could be that the level of dominance for LP detected in the segregating intrapool population may not be a valid estimate for the importance of dominant allelic interactions with the tester from the opposite gene pool. Moreover, the estimation error is high for the level of dominance of QTL effects (Falconer and Mackay, 1996) especially in [F.sub.2:3] lines with only half the dominance effect assessed compared with [F.sub.2] plants. These statistical limitations apply even more to the estimation of additive x dominance or dominance x dominance type of interaction effects. It is, therefore, not surprising that controversial results can arise from the same data depending on the statistical model used for analysis (Cockerham and Zeng, 1996). Furthermore, choosing the correct model for estimation of epistatic effects is complicated, because additive x dominance epistatic effects frequently become significant only if their corresponding main effects are dropped from the model but not if they are included.

Thus, convincing evidence for allelic or nonallelic interactions at the QTL level could not be detected in our study, neither for LP nor for TP. The investigation of epistatic effects seems promising only if few genes regulate the trait under study and pairs of candidate loci are chosen a priori.

QTL Regions Common to Line Per Se and Testcross Performance

Analogous to a high genotypic correlation between LP and TP, a high congruency of QTL identified in both types of progenies is desirable. Beavis et al. (1994) and Austin et al. (2000) found little congruency of yield QTL detected for LP and TP. In this study, more than half of the QTL regions detected were in common for LP and TP in A x [B.sup.I] for all traits except grain yield. The number of detectable common QTL may have been reduced in this study because our joint map covered only 70% of the genome covered by the reference map (Schon et al., 1994). Furthermore, considering that the power of QTL detection was smaller than 100% in both samples, and that the probability of simultaneous detection of a QTL in both progeny types is obtained by multiplication, these results meet expectations. Melchinger et al. (1998) found similar results for the congruency of QTL between two testcross series derived from A x [B.sup.I]. With the exception of grain yield, their QTL mapping results agreed between testers for a number of traits and more than half of the QTL detected with one tester were also found with the other tester. Thus, we conclude that for traits with mainly additive gene action, such as grain moisture, kernel weight, protein concentration, and plant height, QTL detected for LP should be predictive for TP.

To assess the value of QTL identified for LP in predicting TP, we calculated the genotypic correlation [r.sub.g] ([M.sub.LP], [Y.sub.TP]). Except for grain moisture and plant height in A x [B.sup.III] and kernel weight in C x D, at least one common QTL could be detected for the two types of progenies for all traits and all four populations. However, the number of detected QTL was not indicative of the magnitude of [r.sub.g] ([M.sub.LP], [Y.sub.TP]). For example, for grain yield, one common QTL was detected in all four populations, but [r.sub.g] ([M.sub.LP], [Y.sub.TP]) ranged from 0.23 to 0.47 due to the differences in partial [R.sup.2] explained by the respective QTL. On the other hand, even with zero common QTL, a correlation significantly different from zero could be observed for plant height in A x [B.sup.III] and kernel weight in C x D. This must be attributed to (i) QTL detected for LP but with effects below the detection threshold for TP or (ii) QTL linked to those detected for LP. Whether the choice of LOD threshold in QTL mapping for LP has an effect on the magnitude of [r.sub.g] ([M.sub.LP], [Y.sub.TP]) needs to be investigated. Using cross validation, Schon et al. (2004) showed that with a less conservative threshold in QTL estimation, on average, a larger proportion of the genotypic variance could be predicted in test sets.

Estimates of [r.sub.g] ([M.sub.LP], [Y.sub.TP]) were smaller than those of [r.sub.g] (LP, TP) for all traits in all populations, because [r.sub.g] ([M.sub.LP], [Y.sub.TP]) can only be predictive for the proportion of genotypic variance explained by the QTL for LP ([p.sub.TS.ES]), which was generally smaller than 50%. The magnitude of [r.sub.g] ([M.sub.LP], [Y.sub.TP]) should vary for the different traits under study and be a function of the validated genotypic variance explained by the QTL for LP. However, the experimental data only partially confirmed these expectations. A major reason could be the lack of precision in estimates of [r.sub.g] ([M.sub.LP], [Y.sub.TP]) shown by the large confidence intervals especially for the three smaller populations, which was most pronounced for grain yield.

Implications for Hybrid Maize Breeding

The magnitude of the genotypic correlation estimated for LP and TP of four different crosses were in accordance with earlier published results on U.S. dent material. Results for traits with mainly additive gene action, such as grain moisture, kernel weight, protein concentration, and plant height, were encouraging with respect to early selection for LP and indirect improvement of TP. For these traits, more than half the QTL detected for LP and TP were in common. Probably because of the limited power of QTL detection especially in the smaller populations, the proportion of [[??].sup.2.sub.g] explained by QTL for LP was medium to low, and thus resulted in a relatively low correlation between the marker-predicted and the observed TP. With sufficiently large sample sizes for QTL estimation and independent validation, it seems feasible, however, to apply marker-assisted selection based on QTL detected for LP if a substantial proportion of [[??].sup.2.sub.g] can be accounted for. For grain yield, [r.sub.g] (LP, TP) were low, though always greater than the prediction based on markers. Therefore, the application of marker-assisted selection and/or phenotypic selection for LP to improve TP must be evaluated economically. Because of statistical limitations, it was not possible to separate genetic effects such as dominance or epistatic interactions to obtain an unambiguous explanation for the low correlations between LP and TP, neither from the analysis of the phenotypic data nor from the results of the QTL analyses. Therefore, the expansion of the theoretical and simulation study performed by Smith (1986) to the multi-allelic case with different levels of dominance warrants further research.

Table 1. Dimensions of field experiments and of restriction fragment length polymorphism (RFLP) genotyping employed for evaluation of line per se performance (LP) and testcross performance (TP) in four populations (A x [B.sup.I], A x [B.sup.III], A x C, and C x D) of European maize. Population A x [B.sup.I] Experiment LP TP Generation [F.sub.2:3] [F.sub.2:3] Field experiments No. of entries 300 400 Parental lines (P1, P2) 10, 10 5, 5 [F.sub.n] lines 280 380 Common [F.sub.n] lines 280 for LP and TP No. of environments 5 4 Common environments 0 RFLP genotyping No. of genotypes 344 [F.sub.2] No. of loci 89 Population A x [B.sup.III] Experiment LP TP Generation [F.sub.4:6] [F.sub.4:5] Field experiments No. of entries 150 150 Parental lines (P1, P2) 10, 10 5, 5 [F.sub.n] lines 120 71 Common [F.sub.n] lines 65 for LP and TP No. of environments 4 5 Common environments 1 RFLP genotyping No. of genotypes 120 [F.sub.4] No. of loci 151 Population A x C Experiment LP TP Generation [F.sub.3:5] [F.sub.3:4] Field experiments No. of entries 150 150 Parental lines (P1, P2) 7, 7 5, 5 [F.sub.n] lines 131 109 Common [F.sub.n] lines 109 for LP and TP No. of environments 4 5 Common environments 1 RFLP genotyping No. of genotypes 131 [F.sub.3] No. of loci 104 Population C x D Experiment LP TP Generation [F.sub.3:5] [F.sub.3:4] Field experiments No. of entries 150 150 Parental lines (P1, P2) 5, 5 5, 5 [F.sub.n] lines 135 84 Common [F.sub.n] lines 82 for LP and TP No. of environments 4 5 Common environments 1 RFLP genotyping No. of genotypes 140 [F.sub.3] No. of loci 122 Table 2. Means of parents P1 and P2, 280 [F.sub.2:3] (A x [B.sup.I]), 120 [F.sub.4:6] (A x [B.sup.III]), 131 [F.sub.3:5] (A x C), and 135 [F.sub.3:5] (C x D) evaluated for line per se performance (LP) for five agronomic traits of European maize estimated in five or four environments. Population Generation A x [B.sup.I] A x [B.sup.III] Mg [ha.sup.-1] Grain yield P1 2.77 [+ or -] 0.22 4.03 [+ or -] 0.27 ([dagger]) P2 4.88 [+ or -] 0.22 5.79 [+ or -] 0.27 [bar.P] 3.83 [+ or -] 0.16 4.91 [+ or -] 0.19 [bar.F.sub.n] 5.70 [+ or -] 0.06 4.83 [+ or -] 0.10 g [kg.sup.-1] Grain moisture P1 358.3 [+ or -] 3.8 318.8 [+ or -] 4.6 P2 342.0 [+ or -] 3.8 310.7 [+ or -] 4.6 [bar.P] 350.2 [+ or -] 2.7 314.7 [+ or -] 3.3 [bar.F.sub.n] 333.9 [+ or -] 0.8 311.0 [+ or -] 1.9 g Kernel weight P1 264.0 [+ or -] 3.6 275.8 [+ or -] 5.8 P2 225.3 [+ or -] 3.6 228.2 [+ or -] 5.8 [bar.P] 244.7 [+ or -] 2.6 252.0 [+ or -] 4.1 [bar.F.sub.n] 264.2 [+ or -] 1.3 258.6 [+ or -] 2.2 g [kg.sup.-1] Protein concentration P1 129.7 [+ or -] 1.0 126.5 [+ or -] 1.6 P2 115.6 [+ or -] 1.0 113.6 [+ or -] 1.6 [bar.P] 122.6 [+ or -] 0.7 120.1 [+ or -] 1.2 [bar.F.sub.n] 118.0 [+ or -] 0.3 119.1 [+ or -] 0.7 cm Plant height P1 171.4 [+ or -] 2.1 180.3 [+ or -] 2.3 P2 167.3 [+ or -] 2.1 171.4 [+ or -] 2.3 [bar.P] 169.3 [+ or -] 1.5 175.9 [+ or -] 1.6 [bar.F.sub.n] 184.4 [+ or -] 0.6 176.8 [+ or -] 1.3 Population Generation A x C C x D Mg [ha.sup.-1] Grain yield P1 3.92 [+ or -] 0.23 5.27 [+ or -] 0.30 P2 5.39 [+ or -] 0.23 4.48 [+ or -] 0.30 [bar.P] 4.65 [+ or -] 0.16 4.88 [+ or -] 0.21 [bar.F.sub.n] 5.36 [+ or -] 0.08 5.35 [+ or -] 0.11 g [kg.sup.-1] Grain moisture P1 322.8 [+ or -] 3.8 343.4 [+ or -] 4.4 P2 344.4 [+ or -] 3.8 329.0 [+ or -] 4.4 [bar.P] 333.6 [+ or -] 2.7 336.2 [+ or -] 3.1 [bar.F.sub.n] 339.4 [+ or -] 1.4 330.3 [+ or -] 1.5 g Kernel weight P1 269.7 [+ or -] 5.0 192.3 [+ or -] 3.3 P2 190.6 [+ or -] 5.0 194.1 [+ or -] 3.3 [bar.P] 230.1 [+ or -] 3.5 193.2 [+ or -] 2.4 [bar.F.sub.n] 230.7 [+ or -] 1.5 194.6 [+ or -] 1.8 g [kg.sup.-1] Protein concentration P1 127.6 [+ or -] 1.5 98.39 [+ or -] 1.7 P2 96.6 [+ or -] 1.5 120.5 [+ or -] 1.7 [bar.P] 112.1 [+ or -] 1.1 109.4 [+ or -] 1.2 [bar.F.sub.n] 110.8 [+ or -] 0.6 108.7 [+ or -] 0.8 cm Plant height P1 179.5 [+ or -] 2.7 169.3 [+ or -] 2.6 P2 177.0 [+ or -] 2.7 130.9 [+ or -] 2.6 [bar.P] 178.3 [+ or -] 1.9 150.1 [+ or -] 1.8 [bar.F.sub.n] 180.1 [+ or -] 1.0 159.4 [+ or -] 1.1 ([dagger]) Standard errors are attached. Table 3. Variance components and heritabilities of 280 [F.sub.2:3] (A x [B.sub.I]), 120 [F.sub.4:6] (A x [B.sub.III]), 131 [F.sub.3:5] (A x C), and 135 [F.sub.3.5] (C x D) lines evaluated for line per se performance (LP) as well as 380 [F.sub.2:3] (A x [B.sub.I]), 71 [F.sub.4:5] (A x [B.sub.III]), 109 [F.sub.3:4] (A x C), and 84 [F.sub.3.4] (C x D) lines evaluated for testcross performance (TP) with tester T2 for five agronomic traits of European maize estimated in five or four environments. Population A x [B.sub.I] Parameter LP Mg [ha.sup.-1] Grain yield [[??].sup.2.sub.g] ([dagger]) 0.877 [+ or -] 0.081 ** ([dagger]) [[??].sup.2.sub.ge] ([dagger]) 0.170 [+ or -] 0.022 ** [[??].sup.2.sub.e] ([dagger]) 0.536 [+ or -] 0.022 [[??].sup.2] 0.91 95% C.I. on [[??].sup.2] 0.89-0.92 g [kg.sup.-1] Grain moisture [[??].sup.2.sub.g] ([dagger]) 135.11 [+ or -] 12.95 ** [[??].sup.2.sub.ge] ([dagger]) 53.07 [+ or -] 4.11 ** [[??].sup.2.sub.e] ([dagger]) 74.57 [+ or -] 3.05 [[??].sup.2] 0.88 95% C.I. on [[??].sup.2] 0.86-0.90 g Kernel weight [[??].sup.2.sub.g] ([dagger]) 415.50 [+ or -] 36.92 ** [[??].sup.2.sub.ge] ([dagger]) 45.96 [+ or -] 5.34 ** [[??].sup.2.sub.e] ([dagger]) 128.20 [+ or -] 5.24 [[??].sup.2] 0.95 95% C.I. on [[??].sup.2] 0.94-0.96 g [kg.sup.-1] Protein concentration [[??].sup.2.sub.g] ([dagger]) 27.14 [+ or -] 2.48 ** [[??].sup.2.sub.ge] ([dagger]) 3.24 [+ or -] 11.57 ** [[??].sup.2.sub.e] ([dagger]) 15.76 [+ or -] 0.64 [[??].sup.2] 0.92 95% C.I. on [[??].sup.2] 0.91-0.94 cm Plant height [[??].sup.2.sub.g] ([dagger]) 102.8 [+ or -] 9.42 ** [[??].sup.2.sub.ge] ([dagger]) 14.07 [+ or -] 2.24 ** [[??].sup.2.sub.e] ([dagger]) 60.39 [+ or -] 2.46 [[??].sup.2] 0.92 95% C.I. on [[??].sup.2] 0.90-0.93 Population A x [B.sub.I] Parameter TP Mg [ha.sup.-1] Grain yield [[??].sup.2.sub.g] ([dagger]) 0.129 [+ or -] 0.020 ** [[??].sup.2.sub.ge] ([dagger]) 0.155 [+ or -] 0.028 ** [[??].sup.2.sub.e] ([dagger]) 0.811 [+ or -] 0.032 [[??].sup.2] 0.48 95% C.I. on [[??].sup.2] 0.38-0.56 g [kg.sup.-1] Grain moisture [[??].sup.2.sub.g] ([dagger]) 59.44 [+ or -] 5.30 ** [[??].sup.2.sub.ge] ([dagger]) 20.53 [+ or -] 2.57 ** [[??].sup.2.sub.e] ([dagger]) 65.1 [+ or -] 2.56 [[??].sup.2] 0.82 95% C.I. on [[??].sup.2] 0.78-0.84 g Kernel weight [[??].sup.2.sub.g] ([dagger]) 76.85 [+ or -] 6.62 ** [[??].sup.2.sub.ge] ([dagger]) 19.03 [+ or -] 2.77 ** [[??].sup.2.sub.e] ([dagger]) 74.14 [+ or -] 2.93 [[??].sup.2] 0.85 95% C.I. on [[??].sup.2] 0.82-0.87 g [kg.sup.-1] Protein concentration [[??].sup.2.sub.g] ([dagger]) 5.10 [+ or -] 0.50 ** [[??].sup.2.sub.ge] ([dagger]) 2.01 [+ or -] 0.28 ** [[??].sup.2.sub.e] ([dagger]) 5.78 [+ or -] 0.26 [[??].sup.2] 0.76 95% C.I. on [[??].sup.2] 0.71-0.80 cm Plant height [[??].sup.2.sub.g] ([dagger]) 33.21 [+ or -] 2.64 ** [[??].sup.2.sub.ge] ([dagger]) 4.69 [+ or -] 0.97 ** [[??].sup.2.sub.e] ([dagger]) 47.71 [+ or -] 1.26 [[??].sup.2] 0.91 95% C.I. on [[??].sup.2] 0.90-0.92 Population A x [B.sub.III] Parameter LP Mg [ha.sup.-1] Grain yield [[??].sup.2.sub.g] ([dagger]) 1.034 [+ or -] 0.148 ** [[??].sup.2.sub.ge] ([dagger]) 0.274 [+ or -] 0.037 ** [[??].sup.2.sub.e] ([dagger]) 0.383 [+ or -] 0.025 [[??].sup.2] 0.90 95% C.I. on [[??].sup.2] 0.86-0.92 g [kg.sup.-1] Grain moisture [[??].sup.2.sub.g] ([dagger]) 378.10 [+ or -] 52.64 ** [[??].sup.2.sub.ge] ([dagger]) 80.24 [+ or -] 9.80 ** [[??].sup.2.sub.e] ([dagger]) 87.02 [+ or -] 5.67 [[??].sup.2] 0.92 95% C.I. on [[??].sup.2] 0.90-0.94 g Kernel weight [[??].sup.2.sub.g] ([dagger]) 555.21 [+ or -] 77.64 ** [[??].sup.2.sub.ge] ([dagger]) 127.25 [+ or -] 15.16 ** [[??].sup.2.sub.e] ([dagger]) 129.55 [+ or -] 8.44 [[??].sup.2] 0.92 95% C.I. on [[??].sup.2] 0.89-0.94 g [kg.sup.-1] Protein concentration [[??].sup.2.sub.g] ([dagger]) 50.63 [+ or -] 6.98 ** [[??].sup.2.sub.ge] ([dagger]) 10.12 [+ or -] 1.13 ** [[??].sup.2.sub.e] ([dagger]) 8.68 [+ or -] 0.54 [[??].sup.2] 0.93 95% C.I. on [[??].sup.2] 0.91-0.95 cm Plant height [[??].sup.2.sub.g] ([dagger]) 202.60 [+ or -] 27.40 ** [[??].sup.2.sub.ge] ([dagger]) 19.07 [+ or -] 3.49 ** [[??].sup.2.sub.e] ([dagger]) 45.33 [+ or -] 2.95 [[??].sup.2] 0.95 95% C.I. on [[??].sup.2] 0.93-0.96 Population A x [B.sub.III] Parameter TP Mg [ha.sup.-1] Grain yield [[??].sup.2.sub.g] ([dagger]) 0.492 [+ or -] 0.119 ** [[??].sup.2.sub.ge] ([dagger]) 0.825 [+ or -] 0.091 ** [[??].sup.2.sub.e] ([dagger]) 0.494 [+ or -] 0.029 [[??].sup.2] 0.70 95% C.I. on [[??].sup.2] 0.55-0.79 g [kg.sup.-1] Grain moisture [[??].sup.2.sub.g] ([dagger]) 76.81 [+ or -] 14.49 ** [[??].sup.2.sub.ge] ([dagger]) 23.93 [+ or -] 4.46 ** [[??].sup.2.sub.e] ([dagger]) 51.76 [+ or -] 3.02 [[??].sup.2] 0.88 95% C.I. on [[??].sup.2] 0.83-0.92 g Kernel weight [[??].sup.2.sub.g] ([dagger]) 180.73 [+ or -] 32.63 ** [[??].sup.2.sub.ge] ([dagger]) 32.31 [+ or -] 6.73 ** [[??].sup.2.sub.e] ([dagger]) 84.42 [+ or -] 4.89 [[??].sup.2] 0.92 95% C.I. on [[??].sup.2] 0.89-0.95 g [kg.sup.-1] Protein concentration [[??].sup.2.sub.g] ([dagger]) 15.39 [+ or -] 2.80 ** [[??].sup.2.sub.ge] ([dagger]) 3.35 [+ or -] 0.63 ** [[??].sup.2.sub.e] ([dagger]) 7.28 [+ or -] 0.42 [[??].sup.2] 0.92 95% C.I. on [[??].sup.2] 0.88-0.94 cm Plant height [[??].sup.2.sub.g] ([dagger]) 40.54 [+ or -] 7.57 ** [[??].sup.2.sub.ge] ([dagger]) 8.18 [+ or -] 2.22 ** [[??].sup.2.sub.e] ([dagger]) 31.63 [+ or -] 1.83 [[??].sup.2] 0.89 95% C.I. on [[??].sup.2] 0.84-0.93 Population A x C Parameter LP Mg [ha.sup.-1] Grain yield [[??].sup.2.sub.g] ([dagger]) 0.719 [+ or -] 0.098 ** [[??].sup.2.sub.ge] ([dagger]) 0.183 [+ or -] 0.025 ** [[??].sup.2.sub.e] ([dagger]) 0.284 [+ or -] 0.018 [[??].sup.2] 0.90 95% C.I. on [[??].sup.2] 0.86-0.92 g [kg.sup.-1] Grain moisture [[??].sup.2.sub.g] ([dagger]) 246.86 [+ or -] 33.2 ** [[??].sup.2.sub.ge] ([dagger]) 53.01 [+ or -] 6.96 ** [[??].sup.2.sub.e] ([dagger]) 76.15 [+ or -] 4.90 [[??].sup.2] 0.92 95% C.I. on [[??].sup.2] 0.89-0.94 g Kernel weight [[??].sup.2.sub.g] ([dagger]) 269.24 [+ or -] 37.49 ** [[??].sup.2.sub.ge] ([dagger]) 91.44 [+ or -] 10.36 ** [[??].sup.2.sub.e] ([dagger]) 94.27 [+ or -] 6.06 [[??].sup.2] 0.89 95% C.I. on [[??].sup.2] 0.85-0.91 g [kg.sup.-1] Protein concentration [[??].sup.2.sub.g] ([dagger]) 39.03 [+ or -] 5.21 ** [[??].sup.2.sub.ge] ([dagger]) 8.07 [+ or -] 0.98 ** [[??].sup.2.sub.e] ([dagger]) 9.71 [+ or -] 0.62 [[??].sup.2] 0.92 95% C.I. on [[??].sup.2] 0.90-0.94 cm Plant height [[??].sup.2.sub.g] ([dagger]) 113.8 [+ or -] 15.54 ** [[??].sup.2.sub.ge] ([dagger]) 26.42 [+ or -] 3.78 ** [[??].sup.2.sub.e] ([dagger]) 44.81 [+ or -] 2.88 [[??].sup.2] 0.90 95% C.I. on [[??].sup.2] 0.87-0.93 Population A x C Parameter TP Mg [ha.sup.-1] Grain yield [[??].sup.2.sub.g] ([dagger]) 0.271 [+ or -] 0.061 ** [[??].sup.2.sub.ge] ([dagger]) 0.619 [+ or -] 0.061 ** [[??].sup.2.sub.e] ([dagger]) 0.505 [+ or -] 0.029 [[??].sup.2] 0.61 95% C.I. on [[??].sup.2] 0.46-0.71 g [kg.sup.-1] Grain moisture [[??].sup.2.sub.g] ([dagger]) 53.31 [+ or -] 8.49 ** [[??].sup.2.sub.ge] ([dagger]) 20.32 [+ or -] 3.59 ** [[??].sup.2.sub.e] ([dagger]) 54.42 [+ or -] 3.14 [[??].sup.2] 0.85 95% C.I. on [[??].sup.2] 0.79-0.89 g Kernel weight [[??].sup.2.sub.g] ([dagger]) 135.28 [+ or -] 20.14 ** [[??].sup.2.sub.ge] ([dagger]) 22.69 [+ or -] 5.42 ** [[??].sup.2.sub.e] ([dagger]) 93.88 [+ or -] 5.30 [[??].sup.2] 0.91 95% C.I. on [[??].sup.2] 0.87-0.93 g [kg.sup.-1] Protein concentration [[??].sup.2.sub.g] ([dagger]) 10.43 [+ or -] 1.61 ** [[??].sup.2.sub.ge] ([dagger]) 2.96 [+ or -] 0.56 ** [[??].sup.2.sub.e] ([dagger]) 8.80 [+ or -] 0.50 [[??].sup.2] 0.88 95% C.I. on [[??].sup.2] 0.83-0.91 cm Plant height [[??].sup.2.sub.g] ([dagger]) 52.90 [+ or -] 7.88 ** [[??].sup.2.sub.ge] ([dagger]) 8.58 [+ or -] 2.15 ** [[??].sup.2.sub.e] ([dagger]) 37.57 [+ or -] 2.16 [[??].sup.2] 0.91 95% C.I. on [[??].sup.2] 0.87-0.93 Population C x D Parameter LP Mg [ha.sup.-1] Grain yield [[??].sup.2.sub.g] ([dagger]) 1.447 [+ or -] 0.194 ** [[??].sup.2.sub.ge] ([dagger]) 0.306 [+ or -] 0.044 ** [[??].sup.2.sub.e] ([dagger]) 0.527 [+ or -] 0.034 [[??].sup.2] 0.91 95% C.I. on [[??].sup.2] 0.88-0.93 g [kg.sup.-1] Grain moisture [[??].sup.2.sub.g] ([dagger]) 268.83 [+ or -] 36.29 ** [[??].sup.2.sub.ge] ([dagger]) 69.63 [+ or -] 8.31 ** [[??].sup.2.sub.e] ([dagger]) 82.72 [+ or -] 5.33 [[??].sup.2] 0.91 95% C.I. on [[??].sup.2] 0.88-0.93 g Kernel weight [[??].sup.2.sub.g] ([dagger]) 394.73 [+ or -] 50.92 ** [[??].sup.2.sub.ge] ([dagger]) 33.87 [+ or -] 7.10 ** [[??].sup.2.sub.e] ([dagger]) 107.24 [+ or -] 6.91 [[??].sup.2] 0.95 95% C.I. on [[??].sup.2] 0.93-0.96 g [kg.sup.-1] Protein concentration [[??].sup.2.sub.g] ([dagger]) 72.96 [+ or -] 9.51 ** [[??].sup.2.sub.ge] ([dagger]) 9.68 [+ or -] 1.50 ** [[??].sup.2.sub.e] ([dagger]) 19.38 [+ or -] 1.22 [[??].sup.2] 0.94 95% C.I. on [[??].sup.2] 0.92-0.95 cm Plant height [[??].sup.2.sub.g] ([dagger]) 152.60 [+ or -] 20.10 ** [[??].sup.2.sub.ge] ([dagger]) 20.86 [+ or -] 3.74 ** [[??].sup.2.sub.e] ([dagger]) 52.40 [+ or -] 3.36 [[??].sup.2] 0.93 95% C.I. on [[??].sup.2] 0.90-0.95 Population C x D Parameter TP Mg [ha.sup.-1] Grain yield [[??].sup.2.sub.g] ([dagger]) 0.201 [+ or -] 0.045 ** [[??].sup.2.sub.ge] ([dagger]) 0.175 [+ or -] 0.038 ** [[??].sup.2.sub.e] ([dagger]) 0.548 [+ or -] 0.032 [[??].sup.2] 0.69 95% C.I. on [[??].sup.2] 0.56-0.78 g [kg.sup.-1] Grain moisture [[??].sup.2.sub.g] ([dagger]) 46.40 [+ or -] 8.55 ** [[??].sup.2.sub.ge] ([dagger]) 18.10 [+ or -] 3.88 ** [[??].sup.2.sub.e] ([dagger]) 55.32 [+ or -] 3.18 [[??].sup.2] 0.84 95% C.I. on [[??].sup.2] 0.76-0.88 g Kernel weight [[??].sup.2.sub.g] ([dagger]) 157.26 [+ or -] 25.97 ** [[??].sup.2.sub.ge] ([dagger]) 12.45 [+ or -] 5.35 ** [[??].sup.2.sub.e] ([dagger]) 94.21 [+ or -] 5.43 [[??].sup.2] 0.93 95% C.I. on [[??].sup.2] 0.90-0.95 g [kg.sup.-1] Protein concentration [[??].sup.2.sub.g] ([dagger]) 13.85 [+ or -] 2.33 ** [[??].sup.2.sub.ge] ([dagger]) 2.39 [+ or -] 0.56 ** [[??].sup.2.sub.e] ([dagger]) 8.40 [+ or -] 0.48 [[??].sup.2] 0.91 95% C.I. on [[??].sup.2] 0.88-0.94 cm Plant height [[??].sup.2.sub.g] ([dagger]) 40.59 [+ or -] 7.02 ** [[??].sup.2.sub.ge] ([dagger]) 4.78 [+ or -] 2.30 * [[??].sup.2.sub.e] ([dagger]) 41.73 [+ or -] 1.19 [[??].sup.2] 0.89 95% C.I. on [[??].sup.2] 0.84-0.92 * Significant at the 0.05 probability level. ** Significant at the 0.01 probability level. ([dagger]) Standard errors are attached. Table 4. Phenotypic ([r.sup.p]) and genotypic ([r.sup.g]) correlations between line per se performance (LP) and testcross performance (TP), the number of quantitative trait loci (QTL) detected for LP and TP as well as the number of common QTL, the proportion of the genotypic variance ([[??].sub.TS.ES]) explained by these QTL for five agronomic traits of European maize, and the genotypic correlation between LP and TP based on estimated QTL [r.sub.g] ([M.sub.LP], [Y.sub.TP]). Population Parameter A x [B.sub.I] A x [B.sub.III] Grain yield [r.sub.p] (LP, TP) 0.19 (0.09; 0.30) 0.33 (0.07; 0.58) ([dagger]) [r.sub.g] (LP, TP) 0.28 (0.13; 0.44) 0.45 (0.07; 0.87) [r.sub.g] ([M.sub.LP], 0.23 (0.08; 0.36) 0.37 (0.00; 0.50) [Y.sub.TP]) ([double dagger]) No. of QTL (LP) 9 2 No. of QTL (TP) 2 7 No. of common QTL 1 1 [[??].sub.TS.ES] (%) 27.4 3.5 (LP) ([section]) [[??].sub.TS.ES] (%) 18.7 8.2 (TP) ([section]) Grain moisture [r.sub.p] (LP, TP) 0.62 (0.55; 0.69) 0.68 (0.54; 0.79) [r.sub.g] (LP, TP) 0.73 (0.65; 0.81) 0.84 (0.70; 0.98) [r.sub.g] ([M.sub.LP], 0.40 (0.29; 0.47) 0.15 (0.02; 0.30) [Y.sub.TP]) ([double dagger]) No. of QTL (LP) 5 1 No. of QTL (TP) 9 3 No. of common QTL 3 0 [[??].sub.TS.ES] (%) 13.5 2.1 (LP) ([section]) [[??].sub.TS.ES] (%) 33.0 3.1 (TP) ([section]) Kernel weight [r.sub.p] (LP, TP) 0.59 (0.50; 0.67) 0.72 (0.59; 0.82) [r.sub.g] (LP, TP) 0.66 (0.57; 0.75) 0.79 (0.67; 0.90) [r.sub.g] ([M.sub.LP], 0.53 (0.39; 0.63) 0.46 (0.11; 0.64) [Y.sub.TP]) ([double dagger]) No. of QTL (LP) 10 2 No. of QTL (TP) 10 3 No. of common QTL 6 2 [[??].sub.TS.ES] (%) 21.6 9.4 (LP) ([section]) [[??].sub.TS.ES] (%) 42.3 26.6 (TP) ([section]) Protein concentration [r.sub.p] (LP, TP) 0.62 (0.53; 0.69) 0.73 (0.58; 0.84) [r.sub.g] (LP, TP) 0.74 (0.64; 0.84) 0.82 (0.67; 0.92) [r.sub.g] ([M.sub.LP], 0.39 (0.27; 0.51) 0.55 (0.30; 0.65) [Y.sub.TP]) ([double dagger]) No. of QTL (LP) 7 2 No. of QTL (TP) 9 6 No. of common QTL 4 2 [[??].sub.TS.ES] (%) 22.6 7.6 (LP) ([section]) [[??].sub.TS.ES] (%) 38.9 9.8 (TP) ([section]) Plant height [r.sub.p] (LP, TP) 0.68 (0.61; 0.74) 0.70 (0.46; 0.86) [r.sub.g] (LP, TP) 0.81 (0.74; 0.87) 0.80 (0.51; 1.00) [r.sub.g] ([M.sub.LP], 0.65 (0.57; 0.72) 0.34 (0.31; 0.55) [Y.sub.TP]) ([double dagger]) No. of QTL (LP) 13 1 No. of QTL (TP) 12 1 No. of common QTL 7 0 [[??].sub.TS.ES] (%) 35.2 16.4 (LP) ([section]) [[??].sub.TS.ES] (%) 49.3 -0.3 (TP) ([section]) Population Parameter A x C C X D Grain yield [r.sub.p] (LP, TP) 0.38 (0.21; 0.54) 0.42 (0.23; 0.57) [r.sub.g] (LP, TP) 0.54 (0.30; 0.78) 0.56 (0.33; 0.75) [r.sub.g] ([M.sub.LP], 0.35 (0.03; 0.55) 0.47 (0.00; 0.67) [Y.sub.TP]) ([double dagger]) No. of QTL (LP) 3 3 No. of QTL (TP) 6 6 No. of common QTL 1 1 [[??].sub.TS.ES] (%) 12.3 3.8 (LP) ([section]) [[??].sub.TS.ES] (%) 51.8 35.9 (TP) ([section]) Grain moisture [r.sub.p] (LP, TP) 0.61 (0.49; 0.72) 0.40 (0.20; 0.56) [r.sub.g] (LP, TP) 0.74 (0.60; 0.89) 0.52 (0.27; 0.74) [r.sub.g] ([M.sub.LP], 0.34 (0.19; 0.45) 0.43 (0.21; 0.55) [Y.sub.TP]) ([double dagger]) No. of QTL (LP) 7 9 No. of QTL (TP) 7 6 No. of common QTL 1 3 [[??].sub.TS.ES] (%) 22.2 28.5 (LP) ([section]) [[??].sub.TS.ES] (%) 5.2 2.5 (TP) ([section]) Kernel weight [r.sub.p] (LP, TP) 0.64 (0.52; 0.73) 0.67 (0.52; 0.77) [r.sub.g] (LP, TP) 0.72 (0.61; 0.83) 0.71 (0.56; 0.82) [r.sub.g] ([M.sub.LP], 0.49 (0.39; 0.64) 0.26 (0.14; 0.48) [Y.sub.TP]) ([double dagger]) No. of QTL (LP) 3 2 No. of QTL (TP) 4 4 No. of common QTL 1 0 [[??].sub.TS.ES] (%) 14.9 12.2 (LP) ([section]) [[??].sub.TS.ES] (%) 13.5 13.5 (TP) ([section]) Protein concentration [r.sub.p] (LP, TP) 0.69 (0.54; 0.80) 0.72 (0.60; 0.81) [r.sub.g] (LP, TP) 0.78 (0.62; 0.90) 0.79 (0.66; 0.89) [r.sub.g] ([M.sub.LP], 0.49 (0.22; 0.66) 0.55 (0.45; 0.66) [Y.sub.TP]) ([double dagger]) No. of QTL (LP) 7 5 No. of QTL (TP) 6 4 No. of common QTL 5 1 [[??].sub.TS.ES] (%) 7.6 15.9 (LP) ([section]) [[??].sub.TS.ES] (%) 16.6 19.5 (TP) ([section]) Plant height [r.sub.p] (LP, TP) 0.75 (0.61; 0.85) 0.52 (0.36; 0.65) [r.sub.g] (LP, TP) 0.87 (0.72; 0.99) 0.60 (0.42; 0.74) [r.sub.g] ([M.sub.LP], 0.58 (0.12; 0.75) 0.55 (0.35; 0.64) [Y.sub.TP]) ([double dagger]) No. of QTL (LP) 4 5 No. of QTL (TP) 5 3 No. of common QTL 2 1 [[??].sub.TS.ES] (%) 5.0 19.3 (LP) ([section]) [[??].sub.TS.ES] (%) 22.4 12.8 (TP) ([section]) ([dagger]) Empirical 95% confidence interval. ([double dagger]) Correlation between the observed TP and predicted genotypic values on the basis of QTL positions and effects derived from LP, divided by the heritability. ([section]) Proportion of genotypic variance (p) explained in the test set (TS) by all QTL detected with five-fold cross validation in the estimation set (ES) given as median ([[??].sub.TS.ES]) across 1000 replicated cross validation runs.

Abbreviations: CIM, composite interval mapping; DS, data set; ES, estimation set; LP, line per sc performance; P1, parent one: P2, parent two; QTL, quantitative trait locus/loci; RFLP, restriction fragment length polymorphism; TP, testcross performance; TS, test set.

ACKNOWLEDGMENTS

The present study was part of EUREKA project 291) supported by grants from the German Ministry of Research and Technology (BMBF) and KWS Kleinwanzlebener Saatzucht AG, grant 0319233A. It was also supported by a grant from the Deutsche Eorschungsgemeinschaft, Grant No. ME 931/ 3-1. The RFLP assays were conducted in the lab of Prof. Dr. R.G. Herrmann. Ludwig-Maximilians-Universitat in Munich, by E. Brunklaus-Jung and J. Boppenmaier as well as A. Dally in the lab of Prof. Dr. P. Westhoff at the Heinrich-Heine-Universitat in Dusseldorf. The skilled technical assistance of F. Mauch, D. Schillinng-Gross, A. Vesting, and the staff at the Plant Breeding Research Station in Eckartsweier in conducting field trials is gratefully acknowledged.

REFERENCES

Austin, D.F., M. Lee, L.R. Veldboom, and A.R. Hallauer. 2000. Genetic mapping in maize with hybrid progeny across testers and generations: Grain yield and grain moisture. Crop Sci. 40:30-39.

Beavis, W.D., O.S. Smith, D. Grant, and R. Fincher. 1994. Identification of quantitative trait loci using a small sample of topcrossed and [F.sub.4] progeny from maize. Crop Sci. 34:882-896.

Cockerham, C.C., and Z.-B. gong. 1996. Design III with marker loci. Genetics 143:1437-1456.

Falconer, D.S., and T.F.C. Mackay. 1996. Introduction to quantitative genetics. 4th ed. Longman Group, London.

Genter, C.F., and M.W. Alexander. 1966. Development and selection of productive S, inbred lines of corn (Zea mays L.). Crop Sci. 6:429-431.

Groh, S., M.M. Khairallah, D. Gonzalez-de-Leon. M. Willcox. C. Jiang. D.A. Hoisington, and A.E. Melchinger. 1998. Comparison of QTLs mapped in RILs and their test-cross progenies of tropical maize for insect resistance and agronomic traits. Plant Breed. 117: 193-202.

Guffy, R.D., C.W. Stuber, and T. Helentjaris. 1988. Molecular markers for evaluating quantitative traits across varying genetic backgrounds and environments in maize, p. 82. In 1988 Agronomy abstracts. ASA, Madison, WI.

Haley, C.S., and S.A. Knott. 1992. A simple regression method for mapping quantitative trait loci in line crosses using flanking markers. Heredity 69:315-324.

Hallauer, A.R., and J.B. Miranda, Fo. 1981. Quantitative genetics in maize breeding, Iowa State Univ. Press, Ames, IA.

Kerns. M.R., J.W. Dudley, and G.K. Rufener II. 1999. Tester and type of progeny affect QTL detection in maize. Maydica 44:69-83.

Liu, B.H., S.J. Knapp, D. Birkes. 1997. Sampling distributions, biases, variances, and confidence intervals for genetic correlations. Theor. Appl. Genet. 94:8-19.

Mechin, V., O. Argillier, Y. Hebert, E. Guingo, L. Moreau, A. Charcosset, and Y. Barriere. 2001. Genetic analysis and QTL mapping of cell wall digestibility and lignification in silage maize. Crop Sci. 41:690-697.

Melchinger, A.E. 1987. Expectation of means and variances of testcrosses produced from F, and backcross individuals and their selfed progenies. Heredity 59:105-115.

Melchinger. A.E., G.A. Schmidt, and H.H. Geiger. 1986. Evaluation of near infrared reflectance spectroscopy for predicting grain and stover quality traits in maize. Plant Breed. 97:20-29.

Melchinger, A.E., H.F. Utz, and C.C. Schon. 1998. Quantitative trait locus (QTL) mapping using different testers and independent population samples in maize reveals low power of QTL detection and large bias in estimates of QIL effects. Genetics 149:383-403.

Mihaljevic, R., H.F. Utz, and A.E. Melchinger. 2004. Congruency of quantitative trait loci detected for agronomic traits in testcrosses of five populations of European maize. Crop Sci. 44:114-124.

Miller. A.J. 1990. Subset selection in regression. Chapman and Hall, London.

Patterson, H.D., and E.R. Williams. 1976. A new class of resolvable incomplete block designs. Biometrika 63:83-92.

Satterthwaite, F.E. 1946. An approximate distribution of estimates of variance components. Biometrics 2:110-114.

Schnell, F.W. 1961. On some aspects of reciprocal recurrent selection. Euphytica 10:24-30.

Schon, C.C., A.E. Melchinger, J. Boppenmaier, E. Brunklaus-Jung, R.G. Herrmann, and J.F. Seltzer. 1994. RFLP mapping in maize: Quantitative trait loci affecting testcross performance of elite European flint lines. Crop Sci. 34:378-389.

Schon, C.C., H.F. Utz, S. Groh, B. Truberg, S. Openshaw, and A.E. Melchinger. 2004. QTL mapping based on resampling in a vast maize testcross experiment confirms the infinitesimal model of quantitative genetics for complex traits. Genetics 167(1):485-498.

Seitz, G. 1989. Experimentelle und theoretische Untersuchungen zur Beziehung zwischen Linieneigenleistung und allgemeiner Kombinationsfahigkeit bet Silomais. Ph.D. diss. Univ. Hohenheim, Stuttgart. Germany.

Smith, O.S. 1986. Covariance between line per se and testcross performance. Crop Sci. 26:540-543.

Utz, H.F., and A.E. Melchinger. 1996. PLABQTL: A program for composite interval mapping of QTL. J. Quant. Trait Loci 2(1). Available at http://www.uni-hohenheim.dc/~ipspwww/soft.html (verified 10 Sept. 2004). Univ. Hohenheim, Germany.

Utz, H.F., A.E. Melchinger, and C.C. Schon. 2000. Bias and sampling error of the estimated proportion of genotypic variance explained by quantitative trait loci determined from experimental data in maize using cross validation and validation with independent samples. Genetics 154:1839-1849.

Renata Mihaljevic, Chris C. Schon, H. Friedrich Utz. and Albrecht E. Melchinger *

R. Mihaljevic, H.F. Utz, and A.E. Melchinger, Institute of Plant Breeding, Seed Science, and Population Genetics, Univ. of Hohenheim, 70593 Stuttgart, Germany; and C.C. Schon, State Plant Breeding Institute, Univ. of Hohenheim, 70593 Stuttgart, Germany. Received 26 Feb. 2004. * Corresponding author (melchinger@uni-hohenheim.de).

Printer friendly Cite/link Email Feedback | |

Title Annotation: | Crop Breeding, Genetics & Cytology |
---|---|

Author: | Mihaljevic, Renata; Schon, Chris C.; Utz, H. Friedrich; Melchinger, Albrecht E. |

Publication: | Crop Science |

Geographic Code: | 4E |

Date: | Jan 1, 2005 |

Words: | 11236 |

Previous Article: | Identifying isoyield environments for field pea production. |

Next Article: | QTL analysis of cotton fiber quality using multiple Gossypium hirsutum x Gossypium barbadense backcross generations. |

Topics: |