QTL mapping for yield components in a tropical maize population using
microsatellite markers
PRISCILLA KAREN SABADIN1, CLA´ UDIO LOPES DE SOUZA JU´NIOR1, ANETE PEREIRA DE SOUZA2and ANTONIO AUGUSTO FRANCO GARCIA1
1
Departamento de Gene´tica, Escola Superior de Agricultura ‘‘Luiz de Queiroz’’, Universidade de Sa˜o Paulo (ESALQ/USP), Departamento de Gene´tica, Piracicaba, SP, Brazil
2
Instituto de Biologia, Departamento de Gene´tica e Evoluc¸a˜o, Universidade Estadual de Campinas, Cidade Universita´ria ‘‘Zeferino Vaz’’, Campinas, SP, Brazil
Sabadin, P. K., Souza, Jr. C. L., Souza, A. P. and Garcia, A. A. F. 2008. QTL mapping for yield components in a tropical maize population using microsatellite markers. * Hereditas 145: 194203. Lund, Sweden. eISSN 1601-5223. Received March 3, 2008. Accepted June 19, 2008
QTL mapping provides usefull information for breeding programs since it allows the estimation of genomic locations and genetic effects of chromossomal regions related to the expression of quantitative traits. The objective of this study was to map QTL related to several agronomic important traits associated with grain yield: ear weight (EW), prolificacy (PROL), ear number (NE), ear length (EL) and diameter (ED), number of rows on the ear (NRE) and number of kernels per row on the ear (NKPR). Four hundred F2:3 tropical maize progenies were evaluated in five environments in Piracicaba, Sao Paulo, Brazil. The genetic map was previously estimated and had 117 microssatelite loci with average distance of 14 cM. Data was analysed using Composite Interval Mapping for each trait. Thirty six QTL were mapped and related to the expression of EW (2), PROL (3), NE (2), EL (5), ED (5), NRE (10), NKPR (5). Few QTL were mapped since there was high G E interaction. Traits EW, PROL and EN showed high genetic correlation with grain yield and several QTL mapped to similar genomic regions, which could cause the observed correlation. However, further analysis using apropriate statistical models are required to separate linked versus pleiotropic QTL. Five QTL (named Ew1, Ne1, Ed3, Nre3 and Nre10) had high genetic effects, explaining from 10.8% (Nre3) to 16.9% (Nre10) of the phenotypic variance, and could be considered in further studies.
Antonio Augusto Franco Garcia, Departamento de Gene´tica, Escola Superior de Agricultura ‘‘Luiz de Queiroz’’, Universidade de Sa˜o Paulo (ESALQ/USP), CEP 13400-970, CP 83, Piracicaba, SP, Brazil. E-mail: [email protected].
Maize (Zea mays L.) is among the crops of greatest worldwide economic importance. It has been used as human and animal food, as well as raw material in the high-technology industry (WATSON1988; PENY1988).
In terms of breeding, it is one of the most studied species and has been used as a model in many situations. Among the various traits normally consid-ered in breeding programs, yield is generally the most important one; nevertheless it usually has a low heritability (ROBINSON et al. 1949; MALVAR et al.
1996; AUSTIN and LEE 1998). Consequently, studies
involving yield-related traits become important, since they can be used in indirect selection. Some yield components normally considered are prolificacy, ear length and diameter, number of row kernels on the ear and number of kernels per row (JUGENHEIMER1976).
The efficiency of indirect selection depends on the correlation between traits, the evaluation conditions, and the heritability of the trait being evaluated (HALLAUER and MIRANDA FILHO 1988). Several
studies of this kind have been conducted and, in general, results showed that prolificacy allows greater response to indirect selection for grain yield, as it
generally has a high heritability coefficient and a high correlation with yield (HALLAUER and SEARS
1969; COORS and MARDONES 1989; HOLTHAUS and
LAMKEY 1995; MAITA and COORS 1996; L OPEZ-REYNOSO and HALLAUER 1998; LEON and COORS
2002).
Due to the great availability of molecular markers nowadays, it has been possible to map QTL (quanti-tative trait loci) associated with several traits of interest, including yield components (AGRAMA
and MOUSSA 1996a, 1996b; BOHNet al. 1996, 1997;
RIBAUTet al. 1997; GROH et al. 1998; KHAIRALLAH
et al. 1998; LU¨ BBERSTEDTet al. 1998; AGRAMAet al.
1999; AUSTIN et al. 2001; CARDINAL et al. 2001;
FLINT-GARCIA et al. 2003; SIBOV et al. 2003b;
MANGOLIN et al. 2004; LIMA et al. 2006). Most of
this research, however, has been conducted using germplasm from temperate climates. In previous investigations, SIBOV et al. (2003a, 2003b) mapped
QTL for grain yield and other traits (plant and ear height) in a tropical maize population. Four QTL were identified for grain yield, located on chromo-somes 2 (one QTL), 7 (one QTL) and 8 (two QTL),
together explaining 32.73% of phenotypic variance. These were the first QTL mapping results published for tropical maize.
QTL mapping provides estimates for number, position, effects, and interactions between QTL. Therefore, it allows a better knowledge about the genetic architecture of quantitative traits. For yield components, such knowledge should help in designing breeding strategies to use those traits in breeding programs. For example, it could be possible to know whether QTL for correlated traits are close linked in the genome, or whether the QTL have pleiotropic effects. Thus, it would be possible to understand precisely the causes of genetic correlation and then correctly define breeding schemes. RIBAUT et al.
(1997) suggested using information from QTL mapped for yield components in order to do mar-ker-assisted selection (MAS) aiming to increase grain yield. However, results for MAS based on QTL mapping results are not common in the literature, either for corn (STUBERand SISCO1991; OPENSHAW
and FRASCAROLI1997; YOUSEFand JUVIK2001) or
other crops (AHMADIet al. 2001; FALEIROet al. 2004;
LIANGet al. 2004; TWARDOWSKAet al. 2005). Several
authors have mentioned that mapped QTL can occasionally be used in marker-assisted selection (MAS) (BEAVIS 1994; LU¨ BBERSTEDT et al. 1997; RIBAUT et al. 1997; AUSTIN and LEE 1998;
AJMORE-MARSAN et al. 2001; SIBOV et al. 2003b;
LIMAet al. 2006; LI et al. 2007), but results are still
scarce.
In spite of the promising results obtained for grain yield, plant and ear height, SIBOVet al. (2003b) did not
map QTL for yield components in that study. In order to have a better knowledge about the genetic archi-tecture of quantitative traits, yield-component QTL should also be taken into consideration. The objective of this investigation was to detect and estimate the effects of QTL for several yield components in the same tropical corn population previously studied. Traits studied were: ear weight, prolificacy, number of ears, ear length and diameter, number of rows on the ear, and number of kernels per row on the ear.
MATERIAL AND METHODS Experimental design
The phenotypic and molecular data used in this work are presented in greater detail in SIBOVet al. (2003a,
2003b) and their results for grain yield were also included on this paper for comparisons. In short, the segregating population was obtained by crossing the inbreeding lines from different heterotic groups, L-08-05F and L-14-4B, which have contrasting behavior for
grain yield. These lines come from the tropical populations IG-1 and BR-106, respectively. The F1 progeny was obtained by crossing the two inbred lines, and four F1 plants were selfed to generate 400 F2 plants, from which 400 F2:3 progenies were obtained. These were sown in rows containing 20 plants to increase seed of each line for the experimental evalua-tions (SIBOV et al. 2003b). The progenies were
evaluated during the 1999 and 2000 cropping seasons, in the city of Piracicaba, State of Sa˜o Paulo, Brazil. In 1999, the phenotypic evaluations were made in two locations, and in 2000 three locations were used. Each location year combination was considered as one environment in the subsequent statistical analyses, totaling five environments. The 400 progenies were divided into four sets containing 100 progenies and each set was evaluated in a 10 10 lattice design, with two replicates. Plots size consisted of single rows, 0.8 m apart and 4 m long. The final plant density was 10 plants m2 with a total of 20 plants per row, corresponding to a planting density of 62 500 plants ha1in all environments.
The traits evaluated were: grain yield (GY), as the total weight (g) of hand harvested, shelled grain adjusted to 150 g kg1 grain moisture; ear weight (EW) as the total weight of ears of the plot, expressed in Mg ha1; prolificacy (PROL): number of ears plant1, obtained dividing the number of ears of the plot by the stand (number of plants) of the plot; total number of ears plot1 (NE); ear length (EL): mean length of five ears randomly taken on the plots, in cm; ear diameter (ED): mean diameter of five ears randomly taken on the plots, in cm; number of rows on the ear (NRE): mean number of rows measured on five ears randomly taken on the plots; and number of kernels per row on the ear (NKPR): mean number of kernels per row measured on five ears randomly taken on the plots. Traits GY, EW, PROL and NE were corrected for the stand obtained in the experiment using stand as a covariate; this did not change the values for EW significantly. Since EL, ED, NRE and NKPR were obtained based on a fixed number of ears (five per plot), this correction for stand was not applied. Traits EW, PROL and NE were evaluated in the five environments considered. Traits EL, ED, NRE and NKPR were evaluated in four environments.
Phenotypic analyses
Analyses of variance for each lattice and trait were made and then were grouped for the 400 progenies, since they can be considered as a sample of the same reference population. Later, a joint analysis was run across environments, providing estimates of the inter-action between progenies and environments. All
assumptions for the analysis of variance linear model were verified (normality and homocedasticity, homo-geneity of variances, presence of outliers, etc.). The statistical model adopted was considered to be ran-dom, having the mean as the only fixed effect. Based on the expectations of the mean squares, variance estimates were obtained for progenies, interaction genotype environments, and phenotypic variance among means, as well as for the heritability coefficient in the broad sense (BURDICK and GRAYBILL 1992).
Confidence intervals were estimated for heritability according to KNAPP et al. (1990). Estimates for the
genetic and phenotypic correlation between all traits were obtained according to KEMPTHORNE(1966), also
based on the expectations of mean squares. The adjusted means across all environments for each trait and progeny were calculated, and then used for QTL mappings. Those analyses were performed using the proc glm and proc mixed modules of SAS software (SASINSTITUTE2001).
QTL mapping
The linkage map used has 117 microsatellite loci, as previously described in SIBOV et al. (2003a, 2003b).
Shortly, the genetic map was constructed using the MAPMAKER/EXP ver. 3.0 software program (LANDER et al. 1987), where a LOD score higher
than 3 was used to declare linkage in the two-point (pairwise) analysis. Multipoint analyses were used to determine the order and distance between markers. The genetic map has an approximate size of 1634 cM, with an average of 14 cM between adjacent markers. Composite interval mapping (CIM) (ZENG1994) was
used to map QTL for each trait. CIM is based on the hypothesis that a QTL affecting a trait is located within an interval flanked by two markers adjacent to the QTL. Markers linked to putative QTL outside the interval under consideration are taken as model covariables, in order to control their variation. The hypothesis of QTL presence is tested using the like-lihood ratio (LR) statistics (converted to LOD), calculated at every 1 cM. The cofactors were obtained using stepwise multiple regression, with a 5% level of entering-removing probability. Because multiple tests were conducted, the critical genomic level (threshold) to reject H0(absence of QTL) was obtained based on 1000 permutations (CHURCHILL and DOERGE 1994).
All analyses were made using model 6 of the Zmapqtl module in QTLCartographer ver. 1.17 (BASTENet al.
2003), with a window size of 10 cM. The proportion of phenotypic variance (R2) that is explained by all QTL mapped for a given trait was obtained by fitting a model with all QTL simultaneously, using the Win-dows version of QTL Cartographer (WANG et al.
2005). The sign of the additive effects was used to identify the line from which the favorable alleles originated. Positive or negative signs indicate that the alleles that increase a given trait come from line L-14-4B or line L-08-05F, respectively.
RESULTS Phenotypic data
Regarding the yield components evaluated (Table 1), it can be noted that the variance estimates have good accuracy, since their confidence intervals are relatively small. High values were observed for the heritability estimates (broad sense, means of F2:3 progenies), ranging from 91.0% (NRE) to 70.5% (PROL), as well as intermediate value to 68.0% (EW). These values occurred because the traits were evaluated with good precision in five environments with two replicates, resulting in reduced experimental error. This is also confirmed by the low coefficient of variation values, which were relatively low, ranging from 4.3% (ED) to 14.4% (NE). The only exception was EW, which is the most complex trait under consideration and had a CV of 23.6%, but still with an acceptable value for field experiments.
The magnitudes of the variance estimates for the progenies environments interaction were, in general, similar to the variance estimates between progenies, with the exception of EL, ED and NRE, and were all significantly different from zero. EL, ED and NRE traits showed variances between progenies higher than the variances for the genotype environment interac-tion, demonstrating less interaction; however, the interactions were also significantly different from zero. Since the progenies environments interaction was generally high and the QTL were mapped based on the adjusted means across environments, it is expected that only QTL with high average effects would be detected.
In general, the genetic (rG) and phenotypic correla-tion coefficients had very close values (Table 2) and ranged from very low to very high. High genetic correlations were verified between the traits EW and PROL (rG0.90), EW and NE (rG0.90), PROL and NE (rG0.88). Intermediate genetic correlations were observed between NKPR and the traits EW (rG 0.52), PROL (rG0.52), NE (rG0.54), EL (rG 0.53), and NRE and ED (rG0.57). Low genetic correlations occurred, for example, between PROL and traits EL (rG0.27), ED (rG0.38). No signifi-cant correlations we observed for PROL and NRE (rG0), EW and NRE (rG0.04), NE and NRE (rG0), and NKPR and NRE (rG0.04). Grain yield (GY), previously studied by SIBOVet al. (2003b), was
included to verify its relations with yield components, had high genetic correlation values with traits EW (rG0.92), PROL (rG0.90) and NE (rG0.88); intermediate correlation values with NKPR (rG 0.51), and low correlation values with traits EL (rG0.26) and ED (rG0.43), and non significant correlation between GY and NRE (rG0.05). The high genetic correlation values between GY and EW, PROL and NE are indications of great dependence between these traits, indicating that many of the QTL mapped for these traits will be linked or will have pleiotropic effects.
QTL mapping
Based on the permutations, the threshold values for LOD were similar to all traits, with values of 3.90 for EW, 3.85 for PROL, 3.66 for NE, 3.63 for EL, 3.71
for ED, 3.74 for NRE and 3.62 for NKPR. The number of cofactors obtained using stepwise multiple regression was 17 for EW, 19 for PROL, 13 for NE, 10 for EL, 16 for ED, 17 for NRE and 18 for NKPR. Because the number of progenies evaluated was relatively high (n 400), in none of the cases there was a risk of over fitting the model with too many parameters.
Thirty-six QTL responsible for the expression of traits EW, PROL, NE, EL, ED, NRE and NKPR were mapped (Fig. 1, Table 3). As expected, this number is relatively small when each trait is considered sepa-rately, since the adjusted means across five environ-ments were used. QTL were mapped on all chromosomes, except on chromosome 6.
Only two QTL were mapped for EW on chromo-somes 2 (Ew1) and 7 (Ew2), which has the highest Table 1. Parameters estimated: phenotypic variances between means (/s2¯
F), variance between progenies (/s 2 P), variance of the progeniesenvironments interaction (/s2PxE), residual variance (/s2e), and heritability coefficient
(broad sense) for progeny means (/h2¯
P), with corresponding 95% confidence intervals (CI).
Parameter Traitsa EW PROL NE EL ED NRE NKPR /s2¯ F 23.85 1.85 6.94 61.20 34.00 96.20 4.74 CI 20.85;27.55 1.65;2.19 6.01;8.01 53.51;70.7 30.00;39.00 84.09;111.10 4.14;5.47 /s2P 16.21 1.33 4.87 48.94 28.50 87.53 3.59 CI 13.31;20.18 1.10;1.65 4.03;6.02 41.38;58.60 24.3;33.9 75.53;102.67 3.01;4.36 /s2PxE 18.19 1.55 5.58 19.98 7.70 8.00 1.62 CI 15.52;21.60 1.35;1.78 4.89;5.58 16.12;25.43 5.90;10.30 5.42;13.00 1.27;2.17 /s2e 36.40 2.27 8.42 52.94 26.70 48.45 5.39 CI 34.01;39.04 2.12;2.44 7.87;9.04 49.90;57.30 24.80;29.0 44.92;52.40 5.00;5.83 /h2¯ P c 68.0 70.5 70.5 80.0 83.6 91.0 75.8 CI 62.7;74.9 65.6;74.8 65.6;74.8 76.5;82.9 80.8;86.1 89.5;92.4 71.6;79.4 CV(%)b,c 23.6 14.2 14.4 5.3 4.3 6.1 8.1 Meanc 14.5 1.1 20.1 13.8 3.8 11.3 28.6 a
EW ear weight (Mg ha1) 101; PROL prolificacy 102; NE number of ears plot1; EL ear length (cm) 102; EL ear length (cm) 102; ED ear diameter (cm) 103; NRE number of row kernels on the ear 102; NKPR number of kernels per row on the ear.bCoefficient of variation.cThe heritability coefficient, CV, and mean were not multiplied by a power of 10.
Table 2. Phenotypic (below the diagonal) and genetic (above the diagonal) correlation coefficient estimates relative to all traits evaluated.
Parametera GY* EW PROL NE EL ED NRE NKPR
GY* 1.00 0.92 0.90 0.88 0.26 0.43 0.05 0.51 EW 0.91 1.00 0.90 0.90 0.29 0.43 0.04 0.52 PROL 0.88 0.87 1.00 0.88 0.27 0.38 0.00 0.52 NE 0.77 0.88 0.45 1.00 0.28 0.38 0.00 0.54 EL 0.24 0.26 0.29 0.25 1.00 0.13 0.14 0.53 ED 0.39 0.38 0.35 0.34 0.05 1.00 0.57 0.19 NRE 0.07 0.06 0.02 0.02 0.12 0.55 1.00 0.04 NKPR 0.43 0.44 0.43 0.45 0.54 0.22 0.05 1.00 a
Acronyms: see Table 1.
correlation with grain yield. The additive effects (a) showed only negative values, indicating that the source of favorable alleles was inbred line L08-05F. QTL Ew1 individually explained 10.8% of phenotypic variance (one of the highest values). Altogether, the QTL identified for this trait explained 21.2% of total phenotypic variance. The QTL on chromosome 2 closely located to QTL Gy2 identified by SIBOVet al.
(2003b); this could explain the high correlation between these traits. Three QTL were identified for PROL, distributed along chromosomes 2 (Prol1), 7 (Prol2) and 8 (Prol3), explaining 29.5% of total phenotypic variance. QTL Gy2 was also mapped close to the Prol1. The additive effect values ranged from 0.052 to 0.059, and the dominance effects (d) ranged from 0.030 to 0.053. The Prol1 and Prol3 QTL explained each about 9.5% of the phenotypic variance. As to NE, 2 QTL were mapped on chromosomes 2 (Ne1) (close to Gy2) and 7 (Ne2) (close to Prol2). The additive effects only had negative values. Together, both QTL explained 21.8% of total phenotypic variance.
As to EL and ED, the number of QTL mapped (5) was slightly higher than those found for EW, PROL and NE. In EL, QTL are located on chromosomes 1 (El1), 3 (El2), 4 (El3) and 5 (El4 and El5), and no chromosomal region showed a QTL concentration. The additive effects ranged from 0.204 cm to 0.328 cm, and dominance effects ranged from 0.051 cm to 0.240 cm. The proportion of phenotypic variance explained by these QTL individually ranged from 3.1% to 7.9%. Jointly, the QTL explained 29.5% of phenotypic variance. For ED, five QTL were mapped on chromosomes 5 (Ed1 and Ed2), 8 (Ed3, close to Gy8a and Gy8b, and Ed4) and 10 (Ed5), together explaining 30.3% of phenotypic variance. Individually, the QTL explained from 3.3% to 13.8% of phenotypic variance. There were no QTL concentration regions, although QTL Ed3 and Ed4 were located on the same chromosomes, but far apart from each other. The QTL for EL and ED located on chromosome 5 occurred at different intervals.
NRE had the highest number of QTL mapped (10), jointly explaining 60% of phenotypic variance. These
Chromosome 1 Position (cM) LOD 0 50 100 150 200 250 0 5 10 15 20 25 30 EW PROL NE EL ED NRE NKPR Chromosome 2 0 50Gy2 100 150 200 Chromosome 3 0 50 100 150 200 Chromosome 4 0 50 100 150 200 0 5 10 15 20 25 30 LOD Chromosome 5 0 50 100 150 200 Chromosome 7 0 50 100 Gy7 150 Chromosome 8 0 50 100 150 0 5 10 15 20 25 30 LOD Gy8a Gy8b Chromosome 9 0 20 40 60 80 100 Chromosome 10 0 20 40 60 80 100
Fig. 1. LOD values indicating the location of QTL detected by CIM. Triangles (') indicate the position of markers, while inverted triangles (%) indicate the position of mapped QTL, with each color indicating a different trait. Because the threshold LODs were very similar, we chose to graphically present just a straight line at the position with LOD 3.70. The QTL for grain yield were previously mapped by SIBOv et al. (2003a) and are indicated as Gy2, Gy7, Gy8a and Gy8b. Only chromosomes with at least one mapped QTL were represented. Legend for the traits: see Table 1.
Table 3. QTL mapped for several traits (yield components) with corresponding additive (a) and dominance genetic effects (d), using the CIM model.
Chrom/QTL Interval/Binb Position (cM) LOD Genetic effect
/R2fc a LOD d LOD EW (Mg ha1)a 2/Ew1 umc1845-bnlg0166/2.03 55.9 6.3 0.360 3.9 0.328 1.8 10.8 7/Ew2 dupssr13-umc1154/7.04 124.2 4.7 0.303 3.7 0.219 0.6 7.7 /R2Td 21.2 PROLa 2/Prol1 umc1845-bnlg0166/2.03 54.9 5.1 0.052 3.2 0.053 1.9 9.7 7/Prol2 dupssr13-umc1154/7.04 122.2 4.5 0.049 4.1 0.021 0.1 6.9 8/Prol3 bnlg1176-bnlg1607/8.05 78.7 4.8 0.059 4.2 0.030 0.2 9.6 /R2T 29.5 NEa 2/Ne1 umc1845-bnlg0166/2.03 52.9 7.8 1.128 4.4 0.931 1.8 10.8 7/Ne2 dupssr13-umc1154/7.04 123.2 4.5 0.915 4.1 0.412 0.2 6.4 /R2T 21.8 EL (cm)a 1/El1 umc1630-bnlg1338/1.11 235.9 5.0 0.269 4.9 0.043 0.0 5.3 3/El2 umc1659-umc1320/3.07 150.0 7.1 0.328 6.7 0.091 0.0 7.9 4/El3 bnlg2291-umc1989/4.07 128.4 4.1 0.311 4.1 0.051 0.1 7.1 5/El4 bnlg1006-umc1587/5.00 39.0 5.7 0.204 1.6 0.240 1.9 5.1 5/El5 mmc0081-umc1019/5.05 111.0 3.7 0.191 3.5 0.049 0.1 3.1 /R2T 29.5 ED (cm)a 5/Ed1 umc1056-bnlg1902/5.03 76.1 7.6 0.071 7.5 0.004 0.1 6.5 5/Ed2 phi0128-umc1156/5.07 184.2 4.0 0.051 2.5 0.042 0.5 4.5 8/Ed3 phi0115-bnlg1176/8.03 50.9 16.4 0.103 15.0 0.033 0.4 13.8 8/Ed4 dupssr14-umc1893/8.09 155.5 4.6 0.047 4.0 0.020 0.5 3.3 10/Ed5 bnlg0640-bnlg1526/10.03 59.2 10.0 0.083 9.3 0.020 0.4 7.6 /R2T 30.3 NREa 1/Nre1 umc1021-bnlg2238/1.03-1.04 64.3 8.8 0.383 8.7 0.015 0.0 5.5 1/Nre2 umc2025-umc1601/1.05 103.5 7.3 0.324 7.2 0.037 0.0 4.3 2/Nre3 bnlg0125-umc1845/2.02 31.2 17.7 0.508 16.6 0.126 0.0 10.6 2/Nre4 umc1230-umc1394/2.09 197.3 4.4 0.226 4.4 0.007 0.1 2.4 3/Nre5 umc1659-umc1320/3.07 155.0 5.1 0.245 4.8 0.070 0.5 2.9 7/Nre6 bnlg1094-bnlg0434/7.02-7.03 53.2 8.5 0.324 5.8 0.282 1.8 7.1 8/Nre7 phi0115-bnlg1176/8.03 54.9 12.5 0.427 12.5 0.015 0.1 8.2 9/Nre8 bnlg0430-umc1107/9.03 21.2 5.1 0.315 4.3 0.107 0.5 3.4 9/Nre9 umc1107-bnlg1012/9.04 33.1 5.3 0.241 3.5 0.146 0.9 2.8 10/Nre10 bnlg0640-bnlg1526/10.03 60.2 27.6 0.602 27.2 0.003 0.0 16.9 /R2T 60.0 NKPRa 1/Nkpr1 umc1073-umc1021/1.03 55.1 5.6 0.742 4.2 0.439 7.9 5.3 1/Nkpr2 umc1601-bnlg1598/1.05 112.0 7.8 0.847 3.9 0.676 1.3 7.6 5/Nkpr3 bnlg1879-umc1056/5.03 60.0 3.8 0.578 2.7 0.419 0.9 3.9 5/Nkpr4 bnlg1902-umc1221/5.03 101.3 4.1 0.643 3.8 0.213 0.1 4.2 10/Nkpr5 bnlg1526-umc1930/10.04 89.0 6.9 0.505 2.5 0.701 4.2 5.5 /R2T 27.1 a
EW: ear weight; PROL: prolificacy; NE: number of ears plot1; EL: ear length; ED: ear diameter; NRE: number of rows on the ear; NKPR: number of kernels per row on the ear.
b
The Bin position is designated by the X.Y code, where X is the linkage group containing the Bin and Y is the Bin position within the linkage group (GARDINERet al. 1993).
c
Coefficient of phenotypic determination, %. d
QTL were distributed along chromosomes 1 (Nre1 and Nre2), 2 (Nre3, close to Gy2, and Nre4), 3 (Nre5, close to El2), 7 (Nre6, close to Gy7), 8 (Nre7, close to Gy8a and Ed3, and on an interval adjacent to Gy8b and Prol3), 9 (Nre8, on an interval adjacent to Nre9) and 10 (Nre10, close to Ed5). In the case of QTL located on chromosomes 1 and 2, they were relatively distant from each other. However, the QTL on chromosome 9 are on adjacent bins, being only 11.9 cM apart from each other. There was a predominance of positive signs both for additive and dominance effects. Finally, five QTL were mapped for NKPR, distributed on chromosomes 1 (Nkpr1 and Nkpr2, on intervals adjacent to Nre1 and Nre2, respectively), 5 (Nkpr3, on an interval adjacent to Ed1, and Nkpr4), and 10 (Nkpr5, interval adjacent to QTL Ed5 and Nre10). The QTL on chromosome 1 are distant from each other (more than 50 cM), but the QTL on chromosome 5 are relatively close (adjacent bins); the distance between them was 41.3 cM. There was a predominance of negative signs for the additive effects, and positive signs for dominance. The values for the proportion of phenotypic variance explained by these QTL indivi-dually ranged from 3.9% to 7.6%, whereas jointly the detected QTL explained 27.1% of phenotypic var-iance.
DISCUSSION
Most of the economic important traits in maize are quantitative. Because of the environmental effects, difficulties arise to correctly determine the genotype of individuals. In order to work around this problem, sophisticated quantitative methods have been devel-oped, including QTL mapping, by means of statistical associations between markers and regions along the chromosomes, seeking to provide a better understand-ing about the genetic architecture of quantitative traits.
In this study, a total of 36 QTL were mapped, considering all traits evaluated. This number can be considered relatively low for each trait individually when compared with other studies reported in the literature (BEAVISet al. 1994; AUSTINand LEE1996a,
1996b; LUet al. 2003; MELCHINGERet al. 2004; LIMA
et al. 2006), but it is important to mention that results obtained with tropical maize in tropical environments are scarce (BOHNet al. 1996, 1997; RIBAUTet al. 1997;
GROH et al. 1998; KHAIRALLAH et al. 1998;
MANGOLIN et al. 2004; SIBOV et al. 2003b; LIMA
et al. 2006). A small number of QTL were mapped for EW, PROL and NE (the traits with the highest correlation with GY), since them are the most affected by environmental effects, having intermediate values
for heritability and the highest coefficient of variation values observed in this study. For NRE (10 QTL mapped, jointly explaining 60% of phenotypic varia-tion), however, the environmental effects are smaller, resulting in a high heritability coefficient value and low CV. According to FLINT-GARCIA et al. (2005),
high heritability coefficient values indicate that varia-tion for the trait is mostly due to variavaria-tions in the genotypes of individuals, which seems to be the case for NRE.
The use of adjusted means across environments also decreased the number of QTL identified, as only the ones with high average effects were mapped. Previous studies have demonstrated that the use of adjusted means in F2:3populations are more effective to detect QTL (VELDBOOMand LEE1996), but it is worthwhile
to point out that interaction effects in tropical environments are generally much higher than in temperate areas. In this respect, the progenies environments interaction reported here was significant for all traits, indicating a differentiated performance of the progenies in the five environments evaluated. Consequently, because the edaphic-climatic conditions in tropical climates are quite variable, there is great genotype environment interaction, and it is relatively harder to find genotypes with high performance across several environments, consequently making it difficult to locate QTL with high mean effects (RIBAUTet al.
1997). This possibly justifies the small number of QTL mapped in the present work.
It is not always simple to compare the QTL mapping results obtained here with different studies from the literature. For example, differences in the genetic design, genetic maps, markers used and statistical methods make comparisons difficult. There-fore, QTL mapped in this study were compared with other results mainly based on bin matching. For GY, EW, PROL and NE, other authors also detected QTL between bins 2.03 and 2.04 (BEAVIS et al. 1994;
RIBAUT et al. 1997; MELCHINGER et al. 1998; LU
et al. 2003) and between 7.04 and 7.05 (BEAVIS et al.
1994; AUSTIN and LEE 1996b; RIBAUT et al. 1997;
MELCHINGER et al. 1998; YAN et al. 2006). For
PROL, matches were observed with QTL in bin 1.07 (Prol1), with those of VELDBOOMand LEE(1996), and
in bin 7.04 (Prol2) with those of RIBAUTet al. (2007);
in addition, the QTL mapped in the interval adjacent to bin 2.03 was also identified by LIMAet al. (2006).
For ED, there was a match for the position of QTL De3, located in bin 8.03 by VELDBOOM and LEE
(1994). With regard to EL, AUSTIN and LEE (1998)
also mapped QTL on chromosome 3, although in a different interval, and for NRE, QTL were also mapped on chromosomes 2 (VELDBOOM and LEE
1996) and 3 (AUSTIN and LEE 1996b) (adjacent
intervals). For NKPR, QTL were mapped by other authors on chromosomes 1 (BEAVIS et al. 1994;
AUSTIN and LEE 1996b) and 10 (AUSTIN and LEE
1996b).
It can be noted that QTL were observed on all maize chromosomes, except on chromosome 6, where no significant associations were observed between marker and phenotypes for any of the traits. This result can also be observed in many other studies in the literature (STUBER et al. 1992; YAN et al. 2006;
RIBAUTet al. 2007). Therefore, it can be stated that the
QTL mapped in our investigation have good consis-tency with the literature, although there are some differences since we dealt with tropical material. With the ever increasing availability of information on the maize genome, in a near future it will be possible to identify which genes of these regions are associated with QTL, and a better understand of the genetic basis of these traits will be possible.
For some of the traits considered here there were very close regions where the QTL were mapped. These coincidences are observed due to the high genetic correlation that exists in some cases. However, in general, the number of QTL mapped in close genomic regions was small. For the traits EW, PROL and NE, QTL were found in two similar genomic regions in the umc1845-bnlg0166 interval on chromosome 2 (Ew1, Prol1 and Ne1) and in the dupssr13-umc1154 interval on chromosome 7 (Ew2, Prol2 and Ne2). Since these traits are highly correlated, there is evidence of the potential presence of pleiotropy and/or genetic linkage phenomena between these QTL, which could have caused the observed correlation, since they are the most affected by environmental effects. These QTL had additive values with the same sign. In addition, the R2values were relatively high for the QTL located on those chromosomes. The highest values occurred on chromosome 2: 10.8% for Ew1, 9.7% for Prol1 and 10.8% for Ne1.
For ED and NRE, two intervals, phi0115-bnlg1176 (Ed3 and Nre7) and bnlg0640-bnlg1526 (Ed5 and Nre10), located on chromosomes 8 and 10, respec-tively, had QTL on very close positions. These QTL had relatively high R2values, and the additive effects had the same sign for both QTL, which is in accordance with the positive genetic correlation shown by those traits. Some traits, such as EL and NRE, had just one region (umc1659-umc1320) with close QTL, but there was not high genetic correlation between one another and R2had low values. The other traits did not have regions with more than one QTL mapped for each trait on them. It is important to point out that the presence of different QTL for each trait in some
intervals could be the cause of the observed correla-tions, which therefore could be due to the presence of pleiotropic effects or linkage between these QTL. However, additional tests using more elaborate mod-els, like the one presented by JIANGand ZENG(1995),
are necessary.
Among all mapped QTL, five of them (Ew1, Ne1, Ed3, Nre3 and Nre10) individually explained more than 10% of phenotypic variation, suggesting they can be considered for further studies, especially when considering that they have high performance on the average of the environments. The variation not ex-plained by the QTL in this study may have occurred by several reasons, also noted by SIBOV et al. (2003b):
QTL in genome regions that have not been mapped; QTL with little, unidentified effect; epistatic interac-tions between QTL, and the sample size of the populations.
The CIM model used in the present work allows locating and estimating QTL effects for several traits, and was successfully employed in the present case. However, its limitation consists in the fact that it do not allow to dissect the causes of genetic correlation between traits (pleiotropy and linkage), resulting in interpretations based solely on subjective comparisons of the positions of mapped QTL, without any statistical tests. One alternative would be to use the method proposed by JIANG and ZENG(1995), which
allows statistical testing for the presence of pleiotropy vs close linkage, allowing the genetic correlation between traits to be studied in greater detail.
Acknowledgements All researches are recipients of a research fellowship from Conselho Nacional de Desenvolvi-mento Cientı´fico e Tecnolo´gico (CNPq). The authors would like to thank Fundac¸a˜o de Amparo a` Pesquisa do Estado de Sao Paulo (FAPESP, 99/11479-6 and 99/12143-1) for finan-cial support during the development of this research.
REFERENCES
Agrama, H. A. and Moussa, M. E. 1996a. Identification of RAPD markers tightly linked to the dwarf mosaic vı´rus resistance gene in maize. Maydica 41: 205210. Agrama, H. A. and Moussa, M. E. 1996b. Mapping QTLs in
breeding for drought tolerance in maize (Zea mays L.). Euphytica 91: 8997.
Agrama, H. A., Zakaria, A. G., Said, F. B. et al. 1999. Identification of quantitative trait loci for nitrogen use efficiency in maize. Mol. Breed. 5: 187195.
Ahmadi, N., Albar, L., Pressoir, G. et al. 2001. Genetics basis and mapping of resistence to Rice yellow motle virus III. Analysis of QTL efficiency in introgressed progenies confirmed the hypothesis of complementary epistasis between two resistence QTLs. Theor. Appl. Genet. 103: 10841092.
Ajmore-Marsan, P., Gorni, C., Chitto`, A. et al. 2001. Identification of QTLs for grain yield and grain-related traits of maize (Zea mays L.) using an AFLP map,
different testers and cofator analysis. Theor. Appl. Genet. 102: 230243.
Austin, D. F. and Lee, M. 1996a. Genetic resolution and verification of quantitative trait loci for flowering and plant height with recombinant inbred lines of maize. Genome 39: 957968.
Austin, D. F. and Lee, M. 1996b. Comparative mapping in F2:3 and F6:7 generations of quantitative trait loci for grain yield and yield components in maize. Theor. Appl. Genet. 92: 817826.
Austin, D. F. and Lee, M. 1998. Detection of quantitative loci for grain yield and yield components in maize across generations in stress and nostress environments. Crop Sci 38: 12961308.
Austin, D. F., Lee, M. and Veldboom, L. R. 2001. Genetic mapping in maize with hybrid progeny across testers across testers and generations: plant height ad flowering. Theor. Appl. Genet. 102: 163176.
Basten, C. J., Weir, B. S. and Zeng, Z. B. 2003. QTL Cartographer: ver.6. Dept of Statistics, North Carolina State Univ., Raleigh, NC.
Beavis, W. D. 1994. The power and the deceit of QTL experiments: lessons from comparative QTL studies. Proc. 49th Annu. Corn Sorghum Res. Conf. ASTA,
Washington, pp. 250266.
Beavis, W. D., Smith, O. S., Grant, D. et al. 1994. Identification of quantitative trait loci using a small sample of topcrossed and F4 progeny from maize. Crop Sci 34: 882896.
Bohn, M., Khairallah, M. M., Gonza´lez-de-Leo´n, D. et al. 1996. QTL mapping in tropical maize: I. Genomic regions affecting leaf feeding resistance to sugarcane borer and other traits. Crop Sci 36: 13521361. Bohn, M., Khairallah, M. M., Jiang, C. et al. 1997. QTL
mapping in tropical maize: II. Comparison of genomic regions for resistance to Diatraea spp. Crop Sci 37: 18921902.
Burdick, R. K. and Graybill, F. A. 1992. Confidence intervals on variance components. Marcel Dekker. Cardinal, A. J., Lee, M., Sharopova, N. et al. 2001. Genetic
mapping and analysis of quantitative trait loci for resistance to stalk tunneling by the European corn borer in maize. Crop Sci 41: 835845.
Churchill, G. A. and Doerge, R. W. 1994. Empirical thresh-old values for quantitative trait mapping. Genetics 138: 963971.
Coors, J. G. and Mardones, M. C. 1989. Twelve cycles of mass selection for prolificacy in maize I. Direct and correlated responses. Crop Sci 29: 262266.
Faleiro, F. G., Ragagnin, V. A., Moreira, M. A. et al. 2004. Use of molecular markers to accelerate the breeding of common beans resistant to rust and anthracnose. Euphytica 138: 213218.
Flint-Garcia, S. A., Darrah, L. L., Mcmullen, M. D. et al. 2003. Phenotypic versus marker-assisted seletion for stalk strength and second-generation European corn borer resitance in maize. Theor. Appl. Genet. 107: 13311336. Flint-Garcia, S. A., Thuillet, A. C., Yu, J. M. et al. 2005. Maize association population: a high-resolution plata-form for quantitative trait locus dissection. Plant J. 44: 10541064.
Gardiner, J. M., Coe, E. H., Melia-Hancock, S. et al. 1993. Development of a core RFLP map in maize using an immortalized-F2population. Genetics 134: 917930.
Groh, S., Gonza´lez-de-Leo´n, D., Khairallah, M. M. et al. 1998. QTL mapping in tropical maize: III. Genomic regions for resistance to Diatraea spp. and associated traits in two RIL populations. Crop Sci 38: 1062 1072.
Hallauer, A. R. and Sears, J. H. 1969. Mass selection for grain yield in two varieties of maize. Crop Sci 9: 4750. Hallauer, A. R. and Miranda-Filho, J. B. 1988. Quantitative genetics in maize breeding (2nd ed.). Iowa State Univ. Holthaus, J. F. and Lamkey, K. R. 1995. Response to selection and changes in genetic parameters for 13 plant and ear traits in two maize recurrent selection programs. Maydica 40: 357370.
Jiang, C. and Zeng, Z. B. 1995. Multiple trait analysis of genetic mapping for quantitative trait loci. Genetics 140: 11111127.
Jugenheimer, R. W. 1976. Corn improvement, seed produc-tion and uses. Wiley.
Kempthorne, O. 1966. An introdution to genetic statistics. Wiley.
Khairallah, M. M., Bohn, M., Jiang, C. et al. 1998. Molecular mapping of QTL for southwestern corn borer resistance, plant height and flowering in tropical maize. Plant Breeding 117: 309318.
Knapp, S. J., Brindges, W. C. and Birkes, D. 1990. Mapping quantitative trait loci using molecular marker linkage maps. Theor. Appl. Genet. 79: 583592.
Lander, E. S., Green, P., Abrahanson, J. et al. 1987. MAPMAKER: an interactive computer package for constructing primary genetic linkage maps of experimen-tal and natural populations. Genomics 1: 174181. Leon, N. and Coors, J. G. 2002. Twenty-four cycles of mass
selection for prolificacy in the Golden Glow maize population. Crop Sci 42: 325333.
Li, Y., Dong, Y., Niu, S. et al. 2007. The genetic relationship among plant-height traits found using multiple-trait QTL mapping of a dent corn and popcorn cross. Genome 50: 357364.
Liang, F., Deng, Q., Wang, Y. et al. 2004. Molecular markers-assisted selection for yield-enhancing genes in the progeny of ‘‘9311’’ X O. rufipogon using SSR. Euphytica 139: 159165.
Lima, M. L. A., de Souza Jr., C. L., Bento, D. A. et al. 2006. Mapping QTL for grain yield and plant traits in a tropical maize population. Mol. Breed. 17: 227239.
Lopez-Reynoso, J., de, J. and Hallauer, A. R. 1998. Twenty-seven cycles of divergent mass selection for ear length in maize. Crop Sci 38: 10991107.
Lu, H., Romero-Severson, J. and Bernardo, R. 2003. Genetic basis of heterosis explored by simple sequence repeat markers in a random-mated maize population. Theor. Appl. Genet. 107: 494502.
Lu¨bberstedt, T., Melchinger, A. E., Scho¨n, C. C. et al. 1997. Mapping in testcrosses of European Flint lines of maize: I. Comparison of different testers for forage yield traits. Crop Sci 37: 921931.
Lu¨bberstedt, T., Melchinger, A. E., Fahr, A. E. et al. 1998. QTL mapping in testcrosses of flint lines of maize: III. Comparison across populations for forage traits. Crop Sci 38: 12781289.
Maita, R. and Coors, J. G. 1996. Twenty cycles of biparental mass selection for prolificacy in the open-pollinated maize population Golden-Glow. Crop Sci 36: 1527 1532.
Malvar, R. A., Orda´s, A., Revilla, P. et al. 1996. Estimates of genetic variance in two Spanish populations of maize. Crop Sci 36: 291295.
Mangolin, C. A., de Souza Jr, C. L., Garcia, A. A. F. et al. 2004. Mapping QTLs for kernel oil content in a tropical maize population. Euphytica 137: 251259.
Melchinger, A. E., Utz, H. and Scho¨n, C. C. 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 QTL effects. Genetics 149: 343383. Melchinger, A. E., Utz, H. F. and Scho¨n, C. C. 2004. QTL
analysis of complex traits with cross validation, boot-strapping and other biometric methods. Euphytica 137: 111.
Openshaw, S. and Frascaroli, E. 1997. QTL detection and marker-assisted selection for complex traits in maize. Proc. 52th Annu. Corn Sorghum Res. Conf. ASTA, Washington.
Peny, T. W. 1988. Corn as a livestock feed. In: Sprague, G. F. and Dudley, J. W. (eds), Corn and corn improvement. Madison: ASA, CSSA, SSSA, pp. 941963.
Ribaut, J. M., Jiang, C., Gonza´lez-de-Le´on, D. et al. 1997. Identification of quantitative trait loci under drought condition in tropical maize. 2. Yield components and marker-assisted selection strategies. Theor. Appl. Genet. 94: 887896.
Ribaut, J. M., Fracheboud, Y., Monneveux, F. et al. 2007. Quantitative trait loci for yield and correlated traits under high and low soil nitrogen conditions in tropical maize. Mol. Breed. 20: 1529.
Robinson, H. F., Comstock, R. E. and Harvey, P. H. 1949. Estimates of heritability and the degree of dominance in corn. Agron. J. 41: 353359.
SAS Institute 2001. SAS/STAT user?s guide ver. 8.2. SAS Inst.
Sibov, S. T., de Souza Jr, C. L., Garcia, A. A. F. et al. 2003a. Molecular mapping in tropical maize (Zea mays L.) using microsatelites markers. 1. Map construction and localiza-tion of loci showing distorted segregalocaliza-tion. Hereditas 139: 96106.
Sibov, T. S., de Souza, Jr, C. L., Garcia, A. A. F. et al. 2003b. Molecular mapping in tropical maize (Zea mays L.) using
microsatellite markers. 2. Quantitative trait loci (QTL) for grain yield, plant height, ear height and grain moisture. Hereditas 139: 107115.
Stuber, C. W. and Sisco, P. 1991. Marker-facilitated transfer of QTL alleles between elite inbred lines and responses in hybrids. Proc. 46th Annu. Corn Sorghum Res. Conf. ASTA, Chicago, pp. 250266.
Stuber, C. W., Lincoln, S. E., Wolff, D. W. et al. 1992. Identification of genetic factors contributing to heterosis in a hybrid from two elite maize inbred lines using molecular markers. Genetics 132: 823839.
Twardowska, M., Masojc, P. and Milczarski, P. 2005. Pyramiding genes affecting sprouting resistance in rye by means of markers assisted selection. Euphytica 143: 257260.
Veldboom, L. R. and Lee, M. 1994. Molecular-marker-facilitated studies of orphological traits in maize. II. Determination of QTLs for grain yield and yield compo-nents. Theor. Appl. Genet. 89: 415458.
Veldboom, L. R. and Lee, M. 1996. Genetic mapping of quantitative trait loci in maize in stress and nonstress environments: I. Grain yield and yield components. Crop Sci 36: 13101319.
Wang, S., Basten, C. J. and Zeng, Z. B. 2005. Windows QTL Cartographer: ver. 2.5. Dept of Statistics, North Carolina State Univ., Raleigh, NC.
Watson, S. A. 1988. Corn marketing, processing and utilization. In: Sprague, G. F. and Dudley, J. W. (eds), Corn and corn improvement. Madison: ASA, CSSA, SSSA, pp. 881940.
Yan, J. B., Tang, H., Huang, Y. Q. et al. 2006. Quantitative trait loci mapping and epistatic analysis for grain yield and yield components using molecular markers with an elite maize hybrid. Euphytica 149: 121131.
Yousef, G. G. and Juvik, J. A. 2001. Comparison of phenotypic and marker-assisted selection for quantitative traits in sweet corn. Crop Sci 41: 645655.
Zeng, Z. B. 1994. Precision mapping of quantitative trait loci. Genetics 136: 14571468.