| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Formerly Dep. of Crop Sciences, University of Illinois, 1101 S. Goodwin Ave., Urbana IL, 61801
b Dep. of Horticulture and Crop Science, California Polytechnic State University, San Luis Obispo, CA 93407
c I.I.C.A., P.O. Box 70, Belize City, Belize, Central America
d Seminis Vegetable Seeds, Inc., 37437 State Hwy. 16, Woodland, CA 95695
e Betaseeds, Inc., 898 Center St. W., Kimberly, ID 83341
f Pioneer Hi-Bred International, Inc., 216 West 2nd S., Brookings, SD 57006
g Dep. of Chemical Engineering, University of Delaware, Newark, DE 19716
h Dep. of Crop Sciences, University of Illinois, 1101 S. Goodwin Ave., Urbana, IL 61801
* Corresponding author (mewassom{at}sio.midco.net).
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
0.01) and with yield in TCs (rp = 0.59**). Oil was negatively correlated with mass in BC1S1s (rp = –0.29**) and with yield in TCs (rp = –0.30**). Oil was negatively correlated with starch in BC1S1s (rp = –0.75**) and TCs (rp = –0.66**). A genetic map with length = 1486 cM was created with 110 markers. Multiple regression models with QTL detected by composite interval mapping (CIM) explained 46.9, 45.2, 44.3, and 17.7% of phenotypic variance for oil, protein, starch, and mass, respectively, in BC1S1s and 17.5, 22.9, 40.1, and 28.7% for oil, protein, starch, and yield, respectively, in TCs. A 22 cM-interval on chromosome 6 in BC1S1s included oil, protein, and starch QTL, including a QTL explaining 36.7% of the BC1S1 phenotypic variation for oil. No yield QTL were detected in this region. Introgression of this QTL into breeding lines might increase oil while maintaining yield.
Abbreviations: BC, backcross; BC1S1s, backcross1-derived S1 lines • CIM, composite interval mapping • H2, broad-sense heritability • HOC, high-oil corn or maize • IHO, Illinois High Oil • ILO, Illinois Low Oil • ILO(EM), ILO-early maturity • LOD, likelihood of odds • PCR, polymerase chain reaction • 
, proportion of genotypic variance explained • QTL, quantitative trait loci • Radj2, coefficient of determination with adjustment for the number of terms in a model • RFLP, restriction fragment length polymorphism • rg, genotypic correlation • rp, phenotypic correlation • SSR, simple sequence repeat • TCs, topcross hybrids
|
ACKNOWLEDGMENTS |
|---|
|
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Received for publication April 12, 2007.
a Formerly Dep. of Crop Sciences, University of Illinois, 1101 S. Goodwin Ave., Urbana IL, 61801
b Dep. of Horticulture and Crop Science, California Polytechnic State University, San Luis Obispo, CA 93407
c I.I.C.A., P.O. Box 70, Belize City, Belize, Central America
d Seminis Vegetable Seeds, Inc., 37437 State Hwy. 16, Woodland, CA 95695
e Betaseeds, Inc., 898 Center St. W., Kimberly, ID 83341
f Pioneer Hi-Bred International, Inc., 216 West 2nd S., Brookings, SD 57006
g Dep. of Chemical Engineering, University of Delaware, Newark, DE 19716
h Dep. of Crop Sciences, University of Illinois, 1101 S. Goodwin Ave., Urbana, IL 61801
* Corresponding author (mewassom{at}sio.midco.net).
Illinois long-term selection strains of maize (Zea mays L.) have been useful for identifying genomic regions controlling kernel oil, protein, and starch concentrations. To identify kernel trait quantitative trait loci (QTL) in a genetic background more relevant to practical breeding, 150 BC1-derived S1 lines (BC1S1s) were produced from Illinois High Oil and recurrent parent B73. Oil, protein, and starch were measured in BC1S1s and in Mo17-topcross hybrids (TCs). Kernel mass of BC1S1s and grain yield of TCs were also determined. Starch was positively correlated with mass in BC1S1s (rp = 0.67**, 0.01) and with yield in TCs (rp = 0.59**). Oil was negatively correlated with mass in BC1S1s (rp = –0.29**) and with yield in TCs (rp = –0.30**). Oil was negatively correlated with starch in BC1S1s (rp = –0.75**) and TCs (rp = –0.66**). A genetic map with length = 1486 cM was created with 110 markers. Multiple regression models with QTL detected by composite interval mapping (CIM) explained 46.9, 45.2, 44.3, and 17.7% of phenotypic variance for oil, protein, starch, and mass, respectively, in BC1S1s and 17.5, 22.9, 40.1, and 28.7% for oil, protein, starch, and yield, respectively, in TCs. A 22 cM-interval on chromosome 6 in BC1S1s included oil, protein, and starch QTL, including a QTL explaining 36.7% of the BC1S1 phenotypic variation for oil. No yield QTL were detected in this region. Introgression of this QTL into breeding lines might increase oil while maintaining yield.
Abbreviations: BC, backcross; BC1S1s, backcross1-derived S1 lines • CIM, composite interval mapping • H2, broad-sense heritability • HOC, high-oil corn or maize • IHO, Illinois High Oil • ILO, Illinois Low Oil • ILO(EM), ILO-early maturity • LOD, likelihood of odds • PCR, polymerase chain reaction • , proportion of genotypic variance explained • QTL, quantitative trait loci • Radj2, coefficient of determination with adjustment for the number of terms in a model • RFLP, restriction fragment length polymorphism • rg, genotypic correlation • rp, phenotypic correlation • SSR, simple sequence repeat • TCs, topcross hybrids
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
There are practical uses for specialty maize with levels and/or forms of oil, protein, or starch tailored to specific products (Hallauer, 2001). One specialty maize useful in livestock feed is high-oil corn (HOC) (Lambert, 2001). Two advantages of HOC in livestock feed are increased energy density and improved amino acid balance. Because the energy density of lipids is about 2.25 times greater than that of starches or proteins, HOC might increase the growth and productivity of livestock (O'Quinn et al., 2000). Protein quality is improved because the embryo of HOC tends to be larger relative to the endosperm where lysine- and tryptophan-deficient zein proteins are prevalent (Lambert, 2001).
Selection for high oil was accompanied by decreases in kernel mass in IHO (Dudley et al., 1974), in a synthetic population (Mi
evi
and Alexander, 1989), and in Reid's Yellow Dent variety (Miller et al., 1981). Phenotypic correlations (rp) of grain yield and kernel mass of hybrids derived from nine Illinois selection strains were positive (rp = 0.62**), whereas correlations of yield and oil were negative (rp = –0.49**) (Dudley et al., 1977). The tendency for kernel mass, a positive factor for yield (Ottaviano and Camussi, 1981), to decrease as oil content increases might impair the development of high-yielding HOC hybrids. Nevertheless, yield reductions were not significant in a population derived from Reid's Yellow Dent after seven cycles of recurrent selection had increased kernel oil from 40 to 90 mg g–1 (Miller et al., 1981). Yields of commercial HOC hybrids have been too low, however, to generate widespread interest.
Previous investigators detected genomic regions related to kernel composition in IHO x ILO-early maturity [ILO(EM)] (Berke and Rocheford, 1995) and Illinois High Protein (IHP) x Illinois Low Protein (ILP) populations (Goldman et al., 1993, 1994). The extreme divergence of the parents used to create these populations ensured genetic variation for kernel composition traits, but these populations are not representative of modern maize hybrids. To advance the kernel trait QTL studies to a more practical level, we analyzed kernel traits in a backcross population derived from a donor parent with high kernel oil and a recurrent parent with average kernel composition and agronomic traits useful in modern maize breeding. The objectives were to characterize the population for variation in oil, protein, starch, kernel mass, and yield; and to detect, determine, and model the effects of QTL associated with these traits.
|
MATERIALS AND METHODS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Field Technique
Field trials were conducted at the University of Illinois Crop Sciences Research and Education Center at Urbana, IL. Experiments were arranged in an -incomplete block design, with 32 blocks of 5 plots in each block. The plot entries included 150 BC1S1 lines, IHO90 population (random plants of the IHO cycle 90 population), and B73; or 150 Mo17 TCs and checks. Extra rows were filled with IHO90 population and B73 in the BC1S1 experiments and unrelated hybrid checks in the TC experiments. Fill-in rows were not included in the statistical analyses except to calculate means for comparison purposes. The BC1S1s were grown in 1993, 1994, 1996, 2000, and 2001 in one-row nursery plots, 4.6 m long and 0.76 m apart with two replications. Plots were thinned to 15 plants row–1 (equivalent to 43,000 plants ha–1) and plants were self-pollinated by hand using shoot and tassel bags. The low population density facilitated hand pollinations. At maturity, ears were individually harvested, kernels were removed from the centers of individual well-pollinated ears, and balanced-bulk samples were analyzed for determination of oil, protein, and starch concentrations. Kernel mass was measured only in 1993, 1994, and 2000. Testcrosses were grown in two-row plots 5.3 m long and 0.76 m apart in 1995 and 1996 with three replications. Plots were thinned to 23 plants row–1 (56,500 plants ha–1) with open pollination. At maturity, plots were harvested mechanically, and grain mass and moisture were measured for yield determination. A random bulk sample of grain was used for measuring concentrations of kernel oil, protein, and starch by near-infrared light spectroscopy (Dudley and Lambert, 1992) using a DICKEY-john GAC III near-infrared analyzer (DICKEY-john Corporation, Auburn, IL). Kernel mass was determined from a 300-kernel sample of uniformly dried grain.
Statistical Analysis
Statistical analyses of traits were performed using SAS software (SAS, 1999; SAS Institute, Cary, NC) and PLABSTAT, a computer program for analysis of plant breeding experiments (Utz, 2005; University of Hohenheim, Stuttgart, Germany). Lines, years, and replications were treated as random variables. Homogeneity of variance was tested by the SAS GLM procedure using the hovtest option and the Bartlett and Levene-Welch tests. Variance for oil in BC1S1s was not homogenous among lines and variances for protein and kernel mass were not homogenous among years. In TCs, variances of protein, starch, and yield were not homogenous between years. Log or arc sin conversions did not correct the non-homogeneity of variance of any traits, so we proceeded with the analyses on the basic data with the understanding that the sensitivity of some tests might be reduced.
The PLABSTAT LATTICE and ANOVA programs were used to calculate analysis of variance, correlation, estimation of variance components, and broad-sense heritability (H2). A PLABSTAT LATTICE analysis was performed within years and ANOVA over years, treating years and lines as random variables. The PLABSTAT LATTICE analysis estimated variance components within years using an intra-block analysis, and PLABSTAT ANOVA estimated variance components in the combined analysis over years. Variance components were also obtained with the SAS VARCOMP procedure using the MIVQUE0 method to verify the PLABSTAT estimates. Treatment means were adjusted for block effects (Cochran and Cox, 1957). Heritability and 90% confidence intervals were calculated by PLABSTAT on an entry-mean basis using mean squares from the ANOVA as described by Knapp et al. (1985). Because the backcross design does not enable separation of additive and dominance variance, only broad sense heritability was determined. The PLABSTAT ANOVA program used block-adjusted line means by year to calculate phenotypic Pearson correlation coefficients. Genotypic correlations (rg) were calculated by PLABSTAT ANOVA in an analysis of covariance as described by Baker (1986).
Molecular Marker and QTL Analysis
DNA was extracted from a bulk sample of fresh leaf tissue from 20 to 30 seedlings of each BC1S1 line by a cetyl trimethylammonium bromide (CTAB) method (Doyle and Doyle, 1990). Molecular marker profiles of the BC1S1s were produced using restriction fragment length polymorphism (RFLP) (Goldman et al., 1994) and polymerase chain reaction (PCR) of simple sequence repeat (SSR) DNA fragments (Senior et al., 1996). The RFLP and SSR markers are described in the Maize Genetics and Genomics Database (MaizeGDB) (2007) except markers Cfbbi58 and Cqrak57, which are fatty acid desaturase-6 cDNA clones (obtained from Pioneer Hi-Bred International, Inc., Johnston, IA) used for RFLP probes (Mikkilineni and Rocheford, 2003). Primers for SSR markers were obtained from Research Genetics, Inc. (ResGen, Inc., Huntsville, AL).
A linkage map was constructed using the JoinMap Version 3 computer program (Van Ooijen and Voorrips, 2001; Plant Research International, Wagenignen, the Netherlands). The map was constructed using Haldane's mapping function and 110 molecular markers (38 RFLP and 72 SSR). Joinmap tested markers for segregation distortion with a 
2 test. Six markers (bnlg1137, bnlg2244, npi410, bnlg292, umc1101, and php20608a) that were mapped to a single block on chromosome 4 had a significantly (p = 0.01) higher frequency (0.57–0.62) of the heterozygous marker class than the frequency (0.38–0.43) of the homozygous B73 marker class. Because these were the only markers representing a large part of chromosome 4 they were retained and p = 0.005 was used as the cutoff for markers with segregation distortion. Eight other markers representing diverse segments of chromosomes 3, 5, 6, 7, 9, and 10 also had significant segregation distortion. The initial assembly was conducted at likelihood of odds (LOD) 5.5 and clusters were combined to form 10 primary linkage groups. A map was produced with length = 1486 cM, average distance between markers = 14.9 cM, and 96.8% of the genome within 20 cM of a marker. Most markers were mapped to regions similar to the MaizeGDB maps, except phi053, umc76, and ACCase. On the basis of flanking markers on our map and the bin locations of these markers on the Maize GDB maps, we estimated their bin locations to be: phi053, 1.09, ACCase, 9.03, and umc76, 9.03.
Composite interval mapping of QTL (Haley and Knott, 1992) was performed with the PLABQTL computer program (Utz and Melchinger, 2006; University of Hohenheim, Stuttgart, Germany.) The CIM model is yi = µ + bxi + 
bkxik + eI; with yi the phenotypic value of backcross line i for trait y, xi the coded genotype variable at a putative QTL for i, µ the mean, b the effect of the QTL on the trait, and ei, the residual variation in i (Zeng et al., 1999). Models for regression of kernel traits on QTL included additive and additive x additive digenic epistatic effects. The initial detection of QTL with CIM was performed with a LOD threshold of 2.5.
The proportion of phenotypic variance explained by all QTL in the models with adjustment for the number of terms in the multiple regression models, the adjusted R2 (Radj2), was calculated as described by Hospital et al. (1997): Radj2 = R2 – [z/(n – z – 1)](1 – R2). R2 is the coefficient of determination, z is the number of QTL and interaction terms, and n is the number of individuals. The proportion of genotypic variance explained by all QTL in the models, , was calculated according to Utz et al. (2000): =Radj2/H2. PLABQTL determined partial R2 for each term in the regression model as the change in the regression model R2 with that term removed from the model, calculated as: R2% = [R2(full model) – R2(reduced model)]/(1 – R2(reduced model) x 100% (PLABQTL Frequently Asked Questions, 2000). Note that the denominator in the formula will be different for each partial R2 calculated. Therefore, the partial R2 values will not sum up to the full model R2. Effects were declared significant if 0.05.
The QTL included in the multiple regression models were limited to those detected with LOD thresholds equivalent to an = 0.05 genome-wide error rate, as described by Cassady et al. (2001). The LOD thresholds were determined by testing 1000 permutations of the data. The LOD thresholds at = 0.05 for each trait in the BC1S1 analyses were oil, 3.11; protein, 3.34; starch, 3.62; and kernel mass, 3.05. The LOD thresholds at = 0.05 for each trait in the TC analyses were oil, 2.82; protein, 3.07; starch, 3.30; and yield, 2.84.
Cofactors for CIM were selected by PLABQTL in stepwise multiple regression. Cofactors were added to the regression model with F to enter = 3.5. Cross validation with the data set randomly divided among lines in 200 five-fold detection and validation runs was employed to ensure that bias in the data did not lead to false identification of QTL (Utz et al., 2000). Cofactor sets were empirically evaluated and modified to maximize the Radj2.
Some lines were excluded from the analyses because PLABQTL, on the basis of the studentized residual ( > 3.5) or the ANDREWS-PREGIBON statistic second factor ( < 0.5) (Draper and John, 1981), indicated they had an extreme influence on the QTL analysis. Thus, there were 148, 149, 147, 150, and 149 lines included in the QTL and statistical analyses for oil, protein, starch, kernel mass, and yield, respectively.
The sign of QTL effects were defined relative to the IHO90 allele. If the sign of a QTL effect is positive, the IHO90 allele was associated with higher levels of the trait, and if a QTL effect is negative the B73 allele was associated with higher levels of the trait.
|
RESULTS AND DISCUSSION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
|
Phenotypic and genotypic correlations for BC1S1s are shown in Table 3
and for TCs in Table 4
. All phenotypic correlations were significant for both BC1S1s and TCs. The strongest correlation in BC1S1s was between oil and starch [rp = –0.75**, rg 
–1.00 (PLABSTAT calculated rg = –1.20], indicating that most of the genetic control of oil and starch involves closely linked or pleiotropic genes. There was a close positive correlation between kernel mass and starch (rp = 0.67**, rg = 0.70) and a close negative correlation between protein and starch (rp = –0.63**, rg = –0.83). The weakest correlations in BC1S1s were between kernel mass and oil (rp = –0.29**, rg = –0.48) and kernel mass and protein (rp = –0.19**, rg = –0.44).
|
|
Ranks of BC1S1 lines for oil and yield had no apparent relationship and the Spearman rank correlation for oil and yield was equal to –0.15. Most of the top 10 lines for yield were in the lower half of the rank for oil except two ranked 22 and 50. Similarly, most of the top ten lines for oil were in the lower half of the rank for yield excerpt four ranked at 47, 53, 61, and 71. This suggests that it might be possible to increase both yield and oil concentration in this population.
Molecular Marker and QTL Analysis
The multiple regression model for oil in BC1S1s has seven terms, five QTL, plus two digenic epistatic interactions (Table 5
). High oil in the BC1S1s was favored by the IHO90 allele at all five QTL. A QTL on chromosome 6 at position 64 was the strongest QTL for oil in BC1S1s and, with other effects fixed, explained 36.7% of the variation for oil.
|
|
The multiple regression model for protein in TCs includes five QTL plus one digenic epistatic interaction term. The two most influential QTL were on chromosome 1 at position 48 (partial R2 = 7.2%) and on chromosome 5 at position 60 (partial R2 = 6.7%). The IHO90 alleles favored high protein for all the QTL except the QTL at position 76 on chromosome 7.
The multiple regression model for starch in BC1S1s includes nine QTL and no interaction terms. The only QTL in the model for which high starch was favored by an IHO90 allele is on chromosome 3 at position 62. The QTL explaining the most variation for starch in BC1S1s are on chromosome 5 at position 96 and chromosome 8 at position 90, with partial R2 = 20.5 and 18%, respectively.
The multiple regression model for starch in TCs includes six QTL and three digenic epistatic interactions. The QTL explaining the most variation for starch in TCs were on chromosome 1 at position 40 (partial R2 = 14.9%) and on chromosome 5 at position 164 (partial R2 = 7.2%). Higher starch was favored by B73 alleles at all QTL.
The multiple regression model for kernel mass in BC1S1s included two QTL. High kernel mass was favored by an IHO90 allele for a QTL on chromosome 2 and a B73 allele for a QTL on chromosome 5.
The multiple regression model for TC yield included three QTL with no interactions, and yield was favored by B73 alleles at all three QTL. The model for TC yield explained more variation than the model for BC1S1 kernel mass, as the TC yield Radj2 was 28.7% and was 55.1%. Two yield QTL, on chromosome 5 at position 162 and chromosome 8 at position 92, were near QTL for other traits that might have affected yield, including a TC starch QTL on chromosome 5 at position 164 and QTL for oil, protein, and starch in BC1S1s on chromosome 8 at positions 88, 98, and 90, respectively.
Some QTL (LOD 
2.5) were at approximately the same chromosome positions in BC1S1s and TCs (Tables 5 and 6). These include oil QTL in BC1S1s and TCs, respectively, at positions 14 and 6 on chromosome 3, and positions 64 and 74 on chromosome 6; starch QTL at positions 28 and 18 on chromosome 4, positions 96 and 100 on chromosome 5, positions 164 and 166 (the QTL at 166 was not included in the model because the LOD was too low) on chromosome 5; and protein QTL at positions 118 and 134 (not included in the model) on chromosome 8, and at position 166 on chromosome 8 in both BC1S1s and TCs (the TC QTL was not included in the model). Because QTL detection depends on phenotypic variation among lines, estimates of QTL positions might vary because of progeny type (BC1S1 or TC), environment, genotype x environment interaction, and experimental error. Therefore, some of these pairs of BC1S1 and TC QTL might represent the same QTL.
Some QTL for different traits were mapped to the same or nearby positions (Tables 5 and 6, Fig. 1 ), and have directional effects consistent with the sign of the correlations between the traits, suggesting that some of these QTL indicate pleiotropic genes. We grouped the QTL for different traits together as possible common QTL if the sign of the effects for the different traits were in accord with the correlations between traits. The length of the chromosome intervals where common QTL have been grouped ranges from 1 to 22 cM, but most are less than 15 cM. For example, on chromosome 1 in TCs the starch QTL mapped at position 40 had a negative effect and the protein QTL at position 48 had a positive effect. Because starch and protein were negatively correlated, the two QTL might represent a single pleiotropic gene. The protein and starch QTL on chromosome 1 within the interval 40 to 48, on chromosome 4 within the interval 18 to 28, and on chromosome 6 within the interval 8 to 20 all favored starch accumulation with B73 alleles and protein accumulation with IHO90 alleles. The directions of these QTL effects were consistent with the sign of the starch-protein correlation and the influence expected from IHO90 and B73. The QTL on chromosome 5 at position 162 for yield and at position 164 for starch in TCs might be a common QTL, since they have the same sign, and yield was positively correlated with starch. The QTL on chromosome 6 at position 8 for TC starch and at position 20 for TC protein are opposite in sign and might be a common QTL, since protein and starch were negatively correlated.
|
Some of the QTL we detected are in regions that were shown to be important for kernel traits in other reports. Much of the variation for oil was explained by a QTL on chromosome 6 at position 64 flanked by markers umc65 and nc009 in BC1S1s, and a QTL at position 74 flanked by markers nc009 and umc21 in TCs. Berke and Rocheford (1995) reported that in IHO x ILO(EM), with molecular markers analyzed as single factor ANOVAs, umc65 had a significant effect on oil. Their experiment did not include nc009, the other marker flanking our chromosome 6 oil QTL. Goldman et al. (1994), analyzing IHP x ILP, did not detect a significant effect on oil from marker umc21, one of the markers flanking the strongest QTL for oil in our TCs, but they did find other significant markers on chromosome 6, including bnl5.47 of bin 6.05 (Maize GDB, 2007). They did not have any markers representing bin 6.04, where umc65 is located on the bins map. Séne et al. (2001) identified QTL for kernel traits in a flint x dent population. Some of their QTL were mapped to locations similar to our QTL. These include a starch QTL in bin 1.03 (we detected a TC starch QTL in 1.03), an interaction for oil involving a QTL in 1.08 (we have a TC oil QTL in 1.08), an oil QTL in 4.05, an interaction for starch involving a QTL in 4.05 (we detected a BC1S1 starch QTL in 4.05), a protein QTL and an interaction for kernel mass involving QTL in 6.04 (we detected BC1S1 QTL for oil, starch, and protein in 6.04), an interaction involving an oil QTL in 6.05 (we detected a TC oil QTL in 6.05), a kernel mass QTL and a protein interaction QTL in 9.07 (we detected a BC1S1 protein QTL in 9.07).
The most influential QTL for oil was mapped on chromosome 6 at position 64 in BC1S1s, and at position 74 in TCs. Because the 1-LOD support intervals of these QTL overlap (58 to 70 for the BC1S1s and 68 to 78 for the TCs) and both have a positive effect, these might be the same QTL. Because our sample of 150 BC1S1s is small, the accuracy of our QTL analysis might be less than optimal, and our estimate for the magnitude of this QTL's influence on oil (36.7% in BC1S1s) might be high. Nevertheless, the analysis clearly shows this QTL strongly influences oil concentration. Furthermore, Berke and Rocheford's (1995) multiple regression including umc65 (a flanking marker for our chromosome 6 position 64 oil QTL in BC1S1s) explained 52% of the phenotypic and 58% of the genotypic variation for oil in IHO x ILO(EM). And Séne et al. (2001) located QTL for oil, protein, and kernel mass to this region in their flint x dent population. Therefore, there is strong evidence from different populations that this region of chromosome 6 has an important QTL for oil.
Opposing effects from oil and yield QTL were expected (Lambert, 2001). But correlations of oil and yield or oil and kernel mass were weaker than most correlations. And of the three QTL detected for yield, all were mapped 50 cM or more from a TC oil QTL. Even at LOD 1.5, which corresponds to a genome-wide error rate greater than 0.30 in permutation analysis, there were no oil QTL in TCs detected less than 50 cM from a QTL for yield. There was, however, a BC1S1 oil QTL on chromosome 8 at position 88, which is near a TC yield QTL on chromosome 8 at position 92. The QTL with the greatest effects for oil were detected on chromosome 6, but no yield QTL were mapped to chromosome 6, even at LOD 1.5. Other researchers, however, did find yield QTL on chromosome 6L in other populations (Ajmone Marsan et al., 2001; Austin et al., 2000; Ho et al., 2002). Further testing is needed to determine whether introgression of the region of chromosome 6 flanked by markers umc65, nc009, and umc21 of IHO90 into agronomically desirable lines might aid development of high-oil breeding lines without unacceptable reductions in yield.
|
CONCLUSIONS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
This research was supported by grants from the Midwest Plant Biotechnology Consortium, The Consortium for Plant Biotechnology Research, The North Central Biotechnology Program, and the Illinois-Missouri Biotechnology Alliance. Matching support was provided at various times from Cargill Seeds, Pioneer Hi-Bred International, and Monsanto, and from the Department of Crop Sciences and University of Illinois Agricultural Experiment Station. We acknowledge the technical help of Jerry Chandler, Craig Anderson, and Don Roberts, and the numerous undergraduate field and lab helpers.
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.
Received for publication April 12, 2007.
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
evi, D., and D.E. Alexander. 1989. Twenty-four cycles of phenotypic recurrent selection for percent oil in maize. I. Per se and test-cross performance. Crop Sci. 29:320–324.This article has been cited by other articles:
|
|
 |
|
  H.-W. Wang, B.-Y. Wu, T.-M. Song, and S.-J. Chen Effects of Long-Term Selection for Kernel Oil Concentration in KYHO, a High-Oil Maize Population Crop Sci., March 17, 2009; 49(2): 459 - 466. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. J. Wassom, V. Mikkelineni, M. O. Bohn, and T. R. Rocheford QTL for Fatty Acid Composition of Maize Kernel Oil in Illinois High Oil x B73 Backcross-Derived Lines Crop Sci., January 16, 2008; 48(1): 69 - 78. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Vadose Zone Journal | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
