| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Department of Plant Pathology and Department of Crop Science, North Carolina State University, Raleigh, NC 27695-7620
b Department of Crop Science, North Carolina State University, Raleigh, NC 27695-7620
c USDA-ARS, Plant Science Research Unit, Department of Plant Pathology, North Carolina State University, Raleigh, NC 27695-7616
d Department of Crop Sciences, University of Illinois, Urbana-Champaign, IL 61801
e Department of Plant Pathology, North Carolina State University, Raleigh, NC 27695-7567
f USDA-ARS National Center for Agricultural Utilization Research, 1815 N University St., Peoria, IL, 61604
g Pioneer Génétique, Pacé, France
h USDA-ARS, Plant Science Research Unit, Department of Crop Science, North Carolina State University, Raleigh, NC 27695-7620
* Corresponding author (james_holland{at}ncsu.edu)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSQTL ANALYSESDISCUSSIONREFERENCES |
|---|
Abbreviations: ELISA, enzyme-linked immunosorbant assay • GEFR, (GE440 x FR1064) x FR1064 • GEI, genotype x environment interaction • NCB, NC300 x B104 • QTL, quantitative trait locus • RIL, Recombinant inbred lines
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSQTL ANALYSESDISCUSSIONREFERENCES |
|---|
Quantitative genetic variation exists for resistance to Fusarium ear rot and fumonisin contamination among genotypes (Clements et al., 2004; Robertson et al., 2006) and both disease traits have moderate to high entry mean heritabilities (Robertson et al., 2006; Table 1). Therefore, we expect that phenotypic selection for improved resistance should be effective. However, practical difficulties complicate phenotypic selection for these traits. For example, although many diseases can be evaluated on juvenile plants or before flowering, ear rots and mycotoxin concentrations can only be evaluated on grain produced by mature plants. Also, whereas most polycyclic leaf diseases in maize can be reliably induced either by planting in a disease-prone location or by dropping sorghum grains colonized with fungi into juvenile plant whorls (Carson et al., 2004), two inoculations into ears with calibrated concentrations of liquid spore suspensions are often needed to obtain consistent Fusarium ear rot ratings (Clements et al., 2003a). Similarly, whereas relatively rapid and simple visual ratings can be used to accurately assess resistance to many plant diseases, evaluation of fumonisin contamination resistance requires time consuming and expensive toxin assays on precisely ground and weighed grain samples. Finally, whereas some diseases can be reliably evaluated in greenhouses and off-season nurseries, Fusarium ear rot and fumonisin contamination appear to be strongly affected by environmental factors, restricting their evaluation to target environments that are conducive to disease and mycotoxin accumulation (Shelby et al., 1994). Given the difficulties of phenotypically evaluating Fusarium ear rot and mycotoxin contamination, the availability of PCR-based DNA markers linked to genes with at least moderate effects on resistance could permit more efficient selection strategies based on marker genotyping (Holland, 2004; Robertson et al., 2005).
|
In addition to providing the opportunity to conduct marker-assisted selection, QTL mapping provides a powerful method to understand the genetic relationships between correlated traits. Although phenotypic correlations between Fusarium ear rot and fumonisin concentration tend to be low to moderate (Clements et al., 2003b; Clements et al., 2004) and fumonisin can accumulate to high levels in symptomless or only slightly rotted grain (Bacon and Hinton, 1996; Munkvold and Desjardins, 1997; Clements et al., 2004), Robertson et al. (2006) reported high genotypic correlations between Fusarium ear rot and fumonisin concentration (Table 1). These high genotypic correlations suggest that although Fusarium ear rot and fumonisin concentration are not highly phenotypically correlated, genotypes with greater potential to resist ear rot also tend to have greater potential to avoid contamination by fumonisin. Further, it suggests that the genetic components of resistance are largely the same for the two traits (Robertson et al., 2006).
The objectives of this project were to identify QTLs for resistance to Fusarium ear rot and fumonisin contamination in two maize populations, to evaluate the consistency of QTL positions across populations and environments and to evaluate the extent to which QTLs for ear rot resistance had pleiotropic effects on fumonisin contamination.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSQTL ANALYSESDISCUSSIONREFERENCES |
|---|
GEFR Population Evaluation
In 2002, the GEFR population was grown at two locations, Mt. Olive, NC, and Urbana, IL. The experiment was planted in a randomized complete block design with two replications at each location. Plots consisted of single-rows containing individual BC1F1:2 families. In 2003, the GEFR population was grown at two locations, Clayton and Plymouth, NC, with single-row plots of each BC1F1:2 family replicated twice at each location. The treatment design in 2003 was a replication within sets design, wherein each BC1F1:2 family was randomly assigned to one of three sets. Each set also contained the two parent lines and seven check lines (B73, Mo17, K55, NC300, NC390, NC402, and Va35). The experimental design for each set was an 8 x10 
-lattice with two replications at each location. Sets were planted in adjacent plots at each location. At all locations, plots were 3.05 m in length with 0.91 m between plots, and were overplanted and thinned to approximately 20 plants.
NCB Population Evaluation
In 2002, the NCB population was grown at Clayton, NC, with single row plots of each RIL in a randomized complete block design replicated twice. Each complete block also contained the two parent lines. In the summer of 2003, the population was grown at two locations, Clayton and Plymouth, NC. The NCB population plus the two parent lines and the check inbred B73 were evaluated using a 12 x 12 lattice design with two replications at each location. At both locations, plots were 3.05 m in length with 0.91 m between plots. Plots were overplanted and thinned to approximately 20 plants.
Artificial Inoculation Techniques
To represent the heterogeneity of pathogen species in the natural environment, we inoculated ears with three isolates each of F. verticillioides and F. proliferatum (Clements et al., 2003a). Fusarium verticillioides isolates ISU95082, ISU94445, and ISU94040 and F. proliferatum isolates 310, 37–2, and 19 were removed from glycerol storage at –80°C and cultured separately on potato dextrose agar. Inoculum was prepared by washing conidia from the cultures with 0.05% (v/v) Triton X-100 (Fisher Biotech, Fairlawn, NJ) and diluting the resulting conidial suspension of the six isolates to a final concentration to approximately 1 x106 conidia mL–1 in water.
At each inoculation, the primary ear of each plant was inoculated with 10 mL of inoculum. To reduce the chance of escapes, two inoculations, 1 wk apart, were performed on primary ears of all plants in each plot, except for the NCB population in 2002, where only 12 plants were inoculated per plot. Silk channel inoculation (injecting inoculum into the silk channel while the silks are fresh) was performed approximately 10 d after mean silking date for each experiment within each location, and direct ear inoculation (injecting inoculum directly through the husk onto the ear) was performed 7 d later (Clements et al., 2003a).
Phenotypic Data Collection
When all plants had reached physiological maturity, primary ears were hand-harvested and air-dried to approximately 140 g kg–1 moisture. After drying, ears were rated by visually identifying the percentage of kernels exhibiting visible Fusarium ear rot symptoms on each inoculated ear. Ear rot ratings were estimated in 5% increments. The grain from all inoculated ears within a plot was then shelled, bulked, and ground to a fine particle size with a Romer mill (Series II Mill, Romer Labs Inc., Union, MO). The fumonisin concentration was obtained from 25 g of ground grain from each plot, which was evaluated in triplicate by enzyme linked immunosorbant assay (ELISA) (Clements et al., 2003a). One family in the GEFR population and four lines in the NCB population were mostly barren in all environments, preventing measurement of ear rot or fumonisin concentration in those lines.
Genotyping and Linkage Map Construction
For the GEFR population, young leaves from eight plants per family and per parent line were sampled for DNA extraction, and for the NCB population, four plants per RIL and parental line were sampled for extraction. In the NCB population, 142 of the 143 lines were genotyped. DNA extractions were performed with DNeasy Plant Mini Kit system (Qiagen, Inc., Valencia, CA). Simple sequence repeat (SSR) primers were selected for screening in both populations with consensus map information from the Maize Genetics and Genomics Database (http://www.maizegdb.org/; verified 11 April 2006) to reference location of marker loci. Parental lines in both populations were screened with SSR primers to identify polymorphisms (Senior et al., 1998). Polymerase chain reaction products were separated by electrophoresis in 4% (w/v) SFR agarose gels (Amresco, Solon, OH) and visualized by staining gels with 0.05% (w/v) ethidium bromide and exposing them to ultraviolet light.
Locus orders and recombination frequencies were estimated using multipoint mapping in MapmakerEXP (Whitehead Institute, Cambridge, MA). Recombination frequencies were then transformed to centimorgans by Haldane's mapping function in MapmakerEXP. Segregation distortion was tested with chi-square tests by Windows QTL Cartographer version 2.5 (Wang et al., 2005).
Statistical Analyses
Details of the statistical analysis to obtain line means for subsequent QTL analysis were given by Robertson et al. (2006). Briefly, fumonisin data on the GEFR population were natural log transformed and fumonisin data on the NCB population were square root transformed to remove heterogeneity of variance. Genotype least square means across environments and across the two common environments were estimated for each trait within each population by a single mixed models analysis in PROC MIXED of SAS version 8.2 (Littell et al., 1996; SAS Institute, 1999). For these analyses, genotypes were considered fixed effects and all other effects (environments, replications, incomplete blocks, sets, and genotype x environment interactions) were considered random. Heritabilities across environments and across the two common environments were estimated for each trait each population on an entry-mean basis from the univariate mixed model analyses (Holland et al., 2003). For these analyses, all effects were considered random.
QTL Detection and Estimation
Composite interval mapping (CIM) was implemented in Windows QTL Cartographer version 2.5 to provide initial models for the multiple interval mapping procedure (MIM). For CIM, the forward regression method was used for cofactor selection and a 10-cM window size was used for the genome scans. The threshold LOD score to declare significance at = 0.05 was estimated empirically with 300 permutations (Churchill and Doerge, 1994). QTL analysis was performed for each trait within the two common environments and across all environments.
QTL peaks from the CIM analysis (with a LOD threshold value of 2.0 and minimum 5-cM minimum between QTLs) were used as the initial models for MIM in Windows QTL Cartographer 2.5. MIM models were created and tested in an iterative, stepwise fashion, searching for new QTLs to add to the current model, and testing their significance after each search cycle. New models were accepted if they decreased the Bayesian Information Criterion (BIC) (Piepho and Gauch, 2001.) The BIC favors models with higher likelihoods, but includes a penalty for each additional parameter added to the model, to help prevent overfitting the models. After no additional QTLs could be added to a model according to the BIC, each pair of QTLs in the model was tested for epistatic interactions. Epistatic interactions were maintained in the model if they decreased the BIC. While deriving the model using multiple interval mapping, we were mindful not to overfit the model such that the proportion of total variation due to QTLs exceeded the heritability, which can occur even when BIC is used as a model selection criterion. If this occurred, the QTLs associated with the smallest amount of variation was dropped from the model and the remaining QTL effects were re-estimated. This process continued until we obtained a model that explained a proportion of the phenotypic variance that was equal to or less than the heritable proportion of phenotypic variance.
When no additional QTL main effects could be added to the model, the model with minimum BIC was chosen and QTL effects were simultaneously estimated using the "summary" option. The proportion of phenoptyic variation explained by each QTL was estimated from the simultaneous fitting of QTL positions in the final model by Windows QTL Cartographer. Genetic variability explained by QTLs mapped for each trait was calculated as the total phenotypic variation explained by QTLs combined divided by the entry mean heritability of the trait.
QTL main and epistatic effects for fumonisin concentration were estimated by natural log transformed (GEFR population) or square root transformed (NCB population) data. To report QTL effects in original units, we back-transformed the population mean and marginal genotypic effects for each QTL allele class. Additive and epistatic effects were estimated from the back-transformed genotypic mean estimates following Holland (2001). QTLs for maturity were not estimated because Robertson et al. (2006) found that in both the GEFR and NCB populations the genotypic correlation between fumonisin concentration and silk date was not significant and the genotypic correlations between ear rot and silk date were low.
Multivariate QTL Mapping
To investigate the effects of QTLs on ear rot and fumonisin contamination simultaneously, we conducted multivariate multiple regression analysis on the two traits, using marker loci as regressor variables. For each population, the initial regression analysis model contained one marker locus nearest each QTL peak from the final MIM models for both traits. Multivariate analysis was conducted using the MANOVA option of PROC GLM in SAS version 8.2 (SAS Institute, 1999). A single degree of freedom contrast was defined for each marker locus or epistatic interaction. The contrast with the largest Type III P-value of Wilks' Lambda statistic (which tests the null hypothesis of no marker effect on the two traits simultaneously) was dropped from the model. The resulting new model was tested in the same way, and this process continued iteratively until all marker loci had significant (P = 0.05) effects according to the Wilks' Lambda statistic.
The genetic covariance between the two traits associated with each marker locus in the final model was estimated by the method of moments, using the Type III mean cross products for each marker contrast obtained from the Proc GLM analysis. To obtain unbiased estimates of the covariances, the coefficients of the covariance components were based on the equations of Charcosset and Gallais (1996). These estimates were used to calculate the proportion of total genotypic covariance between the two traits due to each marker locus.
QTL x Environment Interactions
QTLs detected in the final MIM models for each trait and population were tested for QTL x environment interactions. A single marker locus closest to each QTL identified by MIM was tested for environmental interactions by PROC GLM in SAS version 8.2. The model used for all analyses was:
 |
ijk = residual variation effect. Loci with significant (P < 0.05) locus x environment interaction effects were then analyzed by single factor analysis in SAS PROC GLM within each individual environment to estimate environment-specific QTL effects. |
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSQTL ANALYSESDISCUSSIONREFERENCES |
|---|
|
|
|
QTL ANALYSES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSQTL ANALYSESDISCUSSIONREFERENCES |
|---|
|
|
Over all environments, in the GEFR population, nine fumonisin contamination resistance QTLs were identified, as well as one epistatic interaction between QTLs (Table 3). The combined QTL and epistasis model explained 67% of the phenotypic variation and 89% of the genotypic variation. In the NCB population, six fumonisin contamination resistance QTLs and three epistatic interactions between QTLs were identified (Table 3). These QTLs explained 81% of the phenotypic variation and 92% of the genotypic variation.
In the GEFR population, two fumonisin contamination resistance QTLs were identified that each explained over 10% of the phenotypic variation, one on chromosome one near marker phi001 that explained 10% of the variation and one on chromosome five near marker umc2111 that explained 13% of the phenotypic variation (Table 3). In the NCB population, three fumonisin QTLs were identified that each explained more than 10% of the phenotypic variation (Table 3). These major QTLs were located on chromosome 1 near bnlg1884 (associated with 17% of the phenotypic variation), on chromosome 6 near umc2059 (associated with 16% of the phenotypic variation), and on chromosome 8 near umc1034 (associated with 11% of the phenotypic variation).
In the GEFR population, eight QTLs explained a combined 65% of the covariance between ear rot and fumonisin concentration (Table 4, Fig. 1 ). Two QTLs were unique to ear rot, as they were detected with multiple interval mapping of ear rot alone but were not detected in the multivariate marker analysis (Fig. 1). One nonpleiotropic ear rot QTL was on chromosome 1 near bnlg1347 (associated with 6% of the phenotypic variation), and the other was on chromosome two near bnlg1017 (associated with 4% of the phenotypic variation) (Table 2). One QTL was unique to fumonisin contamination (Fig. 1), and it was located on chromosome nine near dupssr6 but explained only 0.1% of the phenotypic variation (Table 3). However, if the true effect of this QTL were really that small, it could not be reliably estimated with our population size (Beavis, 1998).
|
|
QTLs detected by MIM on the basis of genotypic means across environments were tested for interactions with environments. Among QTLs for both traits detected in the GEFR population, only one locus, one QTL for ear rot, located near phi001 (bin 1.02), exhibited significant (P < 0.001) locus x environment interaction. This QTL was significant (P < 0.0001) in only two of the four individual environments (Mount Olive, NC, 2002 and Urbana, IL, 2002) when using single marker analysis at phi001. The QTL effects varied in magnitude, but not sign, across environments.
Among the QTLs identified on the basis of genotypic means across environments in the NCB population, only one QTL region exhibited significant locus x environment interaction. The QTL located near the marker pair umc1941-umc1524 (0.8 cM apart in bin 5.06) had significant locus x environment interactions for both ear rot (P < 0.05) and fumonisin content (P < 0.01). This QTL was significant (P < 0.05) for fumonisin content only at Plymouth, NC in 2003, but was almost significant (0.05 < P < 0.10) at Clayton, NC in 2002. Its effect on ear rot was significant (P < 0.05) at Clayton in 2002 and almost significant at Plymouth (0.05 < P < 0.10) in 2003. However, this QTL had no effect on either trait at Clayton in 2003. The QTL effects differed in magnitude, but not sign, across environments.
The GEFR and NCB maps cannot be compared directly because different sets of markers were used in them. Therefore, consistency of QTL positions between the two populations was judged by referencing chromosomal bin position of markers in the maize IBM2 Neighbors consensus map (or other consensus maps, as required; www.maizegdb.org). Furthermore, the phenotypic evaluations were conducted in different sets of environments. Therefore, we also conducted multiple interval mapping on phenotypic means across the two environments that were common to the two studies (Clayton, 2003 and Plymouth, 2003) to make cross-population comparisons.
Among the QTLs identified by means across all environments, three ear rot QTLs and two fumonisin QTLs were putatively consistent across populations (Tables 2 and 3). For ear rot resistance, three QTLs (bins 2.08/2.09, 4.03, and 5.05/5.06) were putatively consistent QTL. For fumonisin contamination resistance, two QTL regions (bins 4.02/4.03 and 5.05/5.06) were putatively consistent QTL. Two QTL regions were consistent across populations for both traits (bins 4.02/4.03 and 5.05/5.06).
QTL mapping based on phenotypic means across the two common environments (Clayton and Plymouth in 2003) identified QTLs that were consistent across populations in bins 1.04/1.05, 3.06/3.07, and 4.03 (Tables 5 and 6). One QTL identified was potentially consistent across populations for fumonisin contamination resistance (bins 1.04/1.05) (Table 6). The QTL region identified in bins 3.06/3.07 was potentially consistent for fumonisin contamination resistance across populations (and also affected ear rot resistance in the NCB population). This is partly congruent with the results of the analysis of means across all environments, in which QTLs for both ear rot and fumonisin contamination resistance in the NCB population were identified in bin 3.06 (Tables 2 and 3) and a QTL was identified in bin 3.09 for fumonisin contamination resistance in the GEFR population (15 cM from the QTLs identified in the GEFR population in the common environment analysis, Tables 3 and 6).
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSQTL ANALYSESDISCUSSIONREFERENCES |
|---|
By comparing QTLs identified between the two populations in the two common environments, we identified several QTL regions that were consistent across populations. The largest QTL for each trait was consistent with at least one QTL for either trait in the other population, suggesting that some resistance genes are common between the populations. However, the population-consistent QTLs identified explained only small amounts of phenotypic variation, further highlighting both the complexity of the resistance and the likelihood that the different resistant parental lines contain different resistance QTLs that could be combined into one line.
Both disease traits were affected by overall genotype x environment interactions (Robertson et al., 2006); therefore QTLs detected in MIM analyses of line means across environments were further analyzed for QTL x environment interactions. Only one QTL region within each population interacted significantly with environments, indicating that most of the QTLs had stable effects across environments. This is congruent with Robertson et al.'s (2006) finding that the genetic correlations of genotypic values measured in different environments were high for both traits. The two QTL regions that interacted with environments interactions seem to be affected by either a change in magnitude between locations or by QTLs not being significant in all environments. However, there was no case of a QTL that had significant positive effects in one environment and significant negative effects in a different environment. Such QTLs may, in fact, exist, but we likely would never have detected them in the original MIM analysis of genotype means across environments. In summary, the QTLs detected were generally stable across environments with some minor exceptions.
QTLs in the GEFR population evaluated across all environments explained more of the genotypic variability for both traits and more of the genotypic covariance between traits than did QTLs in the NCB population. This is probably largely due to the NCB population being smaller than the GEFR population, and hence provides lower power to detect QTLs. It may also be due in part to the difference in structure of the two populations. The NCB population consists of inbred lines which have undergone more generations of recombination, such that linkage disequilibrium between loci with equal recombination frequencies is lower in the recombinant inbred lines than in the first backcross generation. Therefore, a denser linkage map may be required in the NCB than in the GEFR population to achieve the same power to detect the same number of QTLs.
Our experiment could not determine the relative importance of additive and dominant gene action because the NCB population was composed of nearly homozygous recombinant inbred lines and the GEFR population was composed of first backcross generation-derived families (in which the dominant and additive effects of donor parent alleles are confounded). However, we did detect epistasis for fumonisin contamination resistance. Heritabilities were higher for fumonisin contamination resistance than for ear rot resistance, and the higher precision of fumonisin data may have provided greater power to detect QTLs and QTL x QTL interactions.
Robertson et al. (2006) reported high genetic correlations between ear rot and fumonisin contamination in these two populations, on the basis of this same phenotypic data set. On the basis of that result, they suggested that the genetic mechanisms for resistance to ear rot and fumonisin contamination were largely identical. However, we found fewer QTLs with effects on the two traits than would be expected given the high genetic correlations. Multivariate analysis of markers closest to the detected QTLs demonstrated that we detected only 31% of the total estimated genetic covariance in the NCB population and 65% of the total estimated genetic covariance in the GEFR population. These may be underestimates of the true genotypic covariances associated with markers, since the multivariate analyses were conducted directly at marker loci, where the effects of linked QTLs are estimated with a downward bias due to recombination (Edwards et al., 1987). This contrasts with the univariate analyses conducted with multiple interval mapping, which permitted the estimation of QTL effects at their most likely positions, which may be in intervals between markers. Nevertheless, the relatively low proportion of genotypic covariance that was associated with markers near QTLs suggests that not all QTLs were identified. It is possible that a large number of genes with small effects that will be hard to detect (i.e., the polygenic background) may explain much of the remaining genetic covariances. Greater power to detect QTLs and estimate QTL effects could be gained by increasing population size, improving ear rot phenotyping methods, increasing the number of environments for phenotyping, and increasing the number of markers used (Robertson et al., 2005).
Our results are consistent with those of Pérez-Brito et al. (2001), who identified many QTLs, each with small effects on Fusarium ear rot resistance. Pérez-Brito et al. (2001) used two mapping populations for ear rot QTL identification. In their first population, on the basis of bin position, QTL regions were identified that were consistent with QTLs in the GEFR population (bins 1.10, 2.02, 2.08, and 4.09) and two ear rot QTL regions were identified that were consistent with QTLs in the NCB population (bins 2.09 and 6.07). In their second population, they identified, on the basis of bin position, one QTL region that was consistent with a QTL identified in the GEFR population (bin 4.03), and two regions that were consistent with QTLs identified in the NCB population (bins 4.03 and 6.07), and one region that was consistent with a QTL identified across the NCB and both Pérez-Brito populations (bin 6.07). Fine mapping of QTLs in the different populations could determine the true consistency of these QTLs.
Marker assisted breeding could prove very useful when breeding for a trait, such as resistance to Fusarium ear rot or fumonisin contamination, that is very difficult to phenotype and is highly influenced by the environment, provided that the true QTLs can be accurately identified (Robertson et al., 2005). The number of loci conditioning a trait influences the potential efficiency of marker assisted breeding. Bernardo (2001) used computer simulation to determine the usefulness of marker-assisted selections assuming that all of the QTL positions affecting yield were known. He found that it is more advantageous to make selections on the basis of known QTLs if there are only a few loci controlling the trait (e.g., ten loci). With large numbers of loci controlling a trait, estimates of QTL effects become imprecise and selections based on genotypes can become less and less useful, and may even become detrimental as more and more QTLs are discovered. Therefore, in practice, marker-assisted selection is most effective when a few QTLs, each with moderate to large effects on target traits and consistent effects across breeding populations, can be identified (Holland, 2004).
In our study, we found numerous QTLs with small to moderate effects, and limited consistency across populations. Therefore, it is appears that marker-assisted selection cannot be generally recommended for these traits at this time. However, given the costs and effort required to obtain accurate fumonisin concentration data, a program focused on improving agronomically good lines, such as FR1064, with alleles at the most important resistance QTLs from resistant but agronomically poor lines, such as GE440, via marker-assisted backcrossing may be relatively economically efficient. For breeding programs in general, however, phenotypic recurrent selection or pedigree breeding approaches, targeted at improving both ear rot and fumonisin contamination resistance, while also selecting for yield and other agronomic traits may be more efficient. With sufficient resources, such traditional methods can be combined with marker-assisted breeding by screening juvenile plants for larger effect QTLs before pollinations, with the goal of increasing the proportion of favorable QTLs in the population each generation. Because of the high environmental effect on these traits, further studies may need to be performed to confirm the presence, location, and effects of ear rot and fumonisin contamination resistance QTLs and to more precisely map the QTLs to minimize linkage drag.
We hesitate to speculate about resistance mechanisms or candidate genes because each of the QTL regions spans about 20 cM, or about 1% of the maize genetic map. If this relates to 1% of all estimated 
59 000 maize genes (Messing et al., 2004), then we expect that almost 600 genes may reside within each QTL region. Relating QTLs to candidate genes and potential mechanisms will require finer-scale genetic mapping of these regions.
|
ACKNOWLEDGMENTS |
|---|
Received for publication December 2, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSQTL ANALYSESDISCUSSIONREFERENCES |
|---|
dms/fumonbg3.html (verified 11 April 2006).dms/fumongu2.html (verified 11 April 2006).This article has been cited by other articles:
|
|
 |
|
  L. A. Robertson-Hoyt, C. E. Kleinschmidt, D. G. White, G. A. Payne, C. M. Maragos, and J. B. Holland Relationships of Resistance to Fusarium Ear Rot and Fumonisin Contamination with Agronomic Performance of Maize Crop Sci., September 1, 2007; 47(5): 1770 - 1778. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Uphaus, E. Walker, M. Shankar, H. Golzar, R. Loughman, M. Francki, and H. Ohm Quantitative Trait Loci Identified for Resistance to Stagonospora Glume Blotch in Wheat in the USA and Australia Crop Sci., September 1, 2007; 47(5): 1813 - 1822. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. D. Krakowsky, M. Lee, and J. B. Holland Genotypic Correlation and Multivariate QTL Analyses for Cell Wall Components and Resistance to Stalk Tunneling by the European Corn Borer in Maize Crop Sci., March 1, 2007; 47(2): 485 - 488. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
