| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Genetics Department, North Carolina State University, Gardner Hall, Raleigh, NC 27695
b USDA-ARS Plant Genetics Research Unit and Department of Agronomy, University of Missouri-Columbia, Curtis Hall, Columbia, MO 65211
* Corresponding author (DarrahL{at}missouri.edu)
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Abbreviations: ECB, European corn borer • EH, ear height • LOD, log-odds ratio • MoSCSSS-High, MoSCSSS(H24-HRP)C10 • MoSCSSS-Low, MoSCSSS(H25-LRP)C11 • MoSQB-Low, MoSQB(S10)C6 • PH, plant height • QTL, quantitative trait locus • R2, percent of phenotypic variation explained • RadjE, rind penetrometer resistance adjusted for ear height • RPR, rind penetrometer resistance • SSR, simple sequence repeat • TL, tunnel length
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Thompson (1964) found that lower internodes had greater stalk strength than higher internodes. Chesang-Chumo (1993) reported that ear height (EH) was decreased by 22.5% over five cycles of selection for high RPR in MoSCSSS. The correlation between RPR and EH is of interest in terms of direct vs. indirect selection for stalk strength. That is, to what extent does selection for high RPR increase stalk strength per se, or does decreasing EH result in increased stalk strength?
Little research has been conducted to investigate the nature of QTL for correlated traits. Paterson et al. (1991) found a chromosomal region affecting both mass per fruit and soluble solids in tomato (Lycopersicon esculentum Mill). This region contained the Self-pruning Protein (sp) locus which was shown to affect both traits. Schön et al. (1993) adjusted tunnel length (TL), a measure of European corn borer (ECB, Ostrinia nubilalis Hübner) damage, for plant height (PH). They found that four of the seven original TL QTL remained significant. One of the three QTL that lost significance after adjusting for PH overlapped a PH QTL. The authors determined that that particular QTL was not responsible for a physiological mechanism for resistance to ECB, but rather, was limiting TL since PH determines the total length of tunnels possible.
Several statistical programs have been developed to map QTL for multiple traits. In the multiple trait analysis method proposed by Jiang and Zeng (1995), multiple traits are mapped simultaneously, taking advantage of the correlation between traits by considering the correlated trait data as repeated measurements of a single trait. In doing this, parameter estimation is improved and statistical power is increased. Because it was our goal to reduce correlated trait effects on QTL detection to obtain a truer genetic description of the trait of interest, RPR, we decided analyze correlated traits in a way nearly opposite of the method of Jiang and Zeng. We adjusted RPR means for differences in EH by using analysis of covariance procedures (Steele and Torrie, 1960, p. 305).
The objective of this study was to determine the genetic relationship between RPR and the correlated trait EH. This was accomplished by utilizing QTL analysis to characterize and compare QTL for RPR, EH, and RPR after adjusting for EH.
|
MATERIALS AND METHODS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Population 1 was formed by crossing an individual MoSQB-Low plant (female) with an individual MoSCSSS-High1 plant. Population 2 was formed by crossing an individual MoSCSSS-High2 plant (female) with an individual plant from MoSCSSS(H25-LRP) (hereafter referred to as MoSCSSS-Low). Population 3 was formed by crossing Mo47 (female) with an individual MoSCSSS-High3 plant. Population 4 was formed by crossing B73 (female) with Mo47. For all populations, F1 plants were self pollinated yielding F2 individuals, which were self pollinated to produce F2:3 families. Populations 1, 2, 3, and 4 included 282, 291, 291,and 244 F2:3 families, respectively. The F2:3 families were sib pollinated to increase seed for phenotypic data collection.
Populations 1 and 2 were designed specifically to map stalk rind strength QTL since both parents were selected for high and low stalk strength phenotypes. Population 3 was designed to map QTL for RPR and resistance to second-generation ECB with Mo47 as the source of resistance.
Phenotypic and Genotypic Data Collection
Phenotypic data used in this study were collected as described in Flint-Garcia et al. (2003)(this issue). Locations used for evaluation trials in this study included Agronomy Research Center near Columbia, MO, on Mexico silt loam; a site near Tipton, MO, in Cooper County on Clafork and Crestmeade silt loam; and a site managed by the Illinois Crop Improvement Association near Juana Diaz, Puerto Rico, on San Antón sandy clay loam. Briefly, Population 1 was planted as three 10x10 triple lattices at each of three locations: Agronomy Research Center in 1999, and Tipton and Agronomy Research Center in 2000. Populations 2 and 3 were planted as three 10x10 triple lattices per population at each of four locations: two locations in Puerto Rico in 1999, and Tipton and Agronomy Research Center in 2000. Data collected included RPR, PH, and EH. Rind penetrometer resistance was determined for 10 competitive plants plot-1 with a modified electronic rind penetrometer (Sibale et al., 1992). About two weeks after flowering, plants were probed in the middle of the flat side of the internode below the primary ear node. Ear and plant heights were obtained by measuring the distance from the ground to the primary ear node and the collar of the flag leaf, respectively, for 10 competitive plants plot-1.
Populations 1, 2, and 3 were genotyped with 89, 77, and 86 SSR markers, respectively, according to the procedures described in Flint-Garcia et al. (2003)(this issue]).
Data Analysis
Year–location combinations were treated as independent environments. Rind penetrometer resistance was adjusted for EH at each environment generating a new trait, RadjE, as follows: RadjEi = RPRi - bl
where bl is the linear regression coefficient between RPR and EH. Environment means and broad-sense heritability estimates were calculated (Flint-Garcia et al., 2003, this issue). Family means across environments were used to compute phenotypic correlation coefficients with SAS PROC CORR (SAS Institute, Inc., 1998).
MAPMAKER/EXP version 3.0b was used to construct linkage maps (Lander et al., 1987; Lincoln et al., 1992). QTL Cartographer version 1.14d (Basten et al., 1994; Basten et al., 2000) was employed for QTL analysis of family means across environments as described in Flint-Garcia et al. (2003)(this issue). For each trait-population combination, experiment-wise log-odds ratio (LOD) thresholds at P = 0.05 were determined by analyzing 1000 permutations of the data (Churchill and Doerge, 1994). A one-LOD drop from the peak position was used as a confidence interval for QTL location. Overlapping confidence intervals were used to determine QTL in common among traits.
The statistical program EPISTACY was used to test for the presence of epistatic interactions between marker pairs at P < 0.001 (Holland, 1998). To build multilocus models, markers nearest to single-effect QTL and those involved in epistatic interactions were subjected to stepwise regression at P < 0.05 by SAS PROC REG (SAS Institute, Inc., 1998). Markers were added to the model in order of increasing significance (forward regression in), and were removed if their significance while in the model exceeded 0.05 (backward regression out).
|
RESULTS AND DISCUSSION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
|
There are three main causes for correlation between traits: pleiotropy, linkage, and environmental influence (Aastveit and Aastveit, 1993). If EH had a pleiotropic effect and was responsible for differences in RPR, then one should make selections for high RPR on the basis of low EH since measuring EH is easier and faster than measuring RPR. If RPR and EH are correlated because of linkage, that is, high RPR alleles at a given locus are linked with low EH alleles, then one would expect some proportion of individuals in a population to have high RPR and high EH, and vice versa, because of recombination between loci. It would also be possible to break the correlation between EH and RPR through recombination between RPR and EH loci. Environmental correlation results when two traits are influenced by the same difference in environment; for example, drought usually causes decreased PH and decreased yield. Of course, it is probable that there is simultaneous action of any or all of these causes on correlation.
We decided to analyze the relationship between stalk strength and EH to obtain a more accurate estimate of RPR effects. We adjusted RPR means for differences in EH by using analysis of covariance procedures (Steele and Torrie, 1960, p. 305).
The inheritance of RPR is rather complex (Flint-Garcia et al., 2003, this issue) (summarized in Table 2 and Fig. 1). Many QTL (8, 10, and 8 QTL for Populations 1, 2, and 3) with small effects were detected with an average partial R2 across populations of 5.5%, and only one QTL in Population 1 with a partial R2 greater than 10%. Four, two, and zero significant epistatic interactions were detected in Populations 1, 2, and 3. The amount of variation explained by multi-locus models (total R2) ranged from 33.4 to 48.4%.
|
|
On the basis of overlapping confidence intervals, there were zero, two, and two QTL in common between RPR and EH in Populations 1, 2, and 3, respectively (Fig. 1). It was surprising that there were so few QTL in common between RPR and EH as the correlation between them was significant in all three populations (Table 1). However, there were three, one, and one EH QTL linked to (within 50 centimorgans) RPR QTL for Populations 1, 2, and 3, respectively. Nevertheless, out of a total of 26 RPR QTL, there were only nine RPR QTL that were either linked to or overlapped EH QTL.
The results for the QTL analysis of RadjE (Table 2, Fig. 1) are similar to those of the original RPR data, that is, there were a large number of QTL (6, 7, and 10) with small-to-moderate effects. The average partial R2 was 7.5%, and there were four QTL with partial R2s greater than 10% across populations. Four, one, and two epistatic interactions were significant in the three populations, respectively. After adjustment for EH, multi-locus models for RadjE accounted for 32.0, 41.7, and 49.6% of the phenotypic variation.
Fifteen of the original 26 RPR QTL remained significant as QTL for RadjE across populations: 4 of 8, 6 of 10, and 5 of 8 QTL in Populations 1, 2, and 3, respectively. Two interactions in Population 1 and one interaction in Population 2 remained significant after adjustment for EH. Adjustment for EH caused several changes in LOD score and/or partial R2 for the QTL in common between RPR and RadjE. Decreases in LOD score ranged from 0.4 to 3.1, and increases ranged from 0.3 to 12.3. Overall, increases in LOD score were more significant than decreases. Associated with these changes in LOD scores were changes in partial R2s for the individual QTL: decreases ranged from 0.6 to 2.2%, and increases ranged from 0.4 to 13.2% across populations. There were six decreases in R2 vs. nine increases. Again, increases in R2 were greater overall than decreases in R2. The increases in LOD score and R2 for RadjE vs. RPR can be interpreted such that RadjE QTL represent stalk strength loci per se, and removing the correlated effects of EH enhanced the estimation of genetic effects attributed to stalk strength per se.
Interestingly, after adjusting RPR for EH, more alleles that increased stalk strength originated from MoSCSSS-High for Population 1 (data not shown). The original RPR results for Population 1, that the majority of the positive RPR alleles originated from MoSQB-Low (Flint-Garcia et al., 2003, this issue), were an artifact of the MoSCSSS-High parent being taller and having higher ear heights. This particular relationship between RPR and PH/EH in MoSCSSS-High is opposite of what is usually observed, and, as a result, high RPR alleles originated from MoSQB-Low based on its lower EH. Adjustment of RPR for EH removed the correlated effects of EH such that the remaining QTL were true RPR QTL. These results clearly indicate that correlated loci from EH have an effect on RPR. Populations 2 and 3 remained the same with respect to the source of high RPR alleles as compared to the original RPR data.
There was a loss of 11 RPR QTL across all three populations after adjustment for EH. In Population 1, two of the four QTL lost were linked to EH QTL. A third lost QTL was the largest RPR QTL in Population 1, with a LOD score of 8.8 and a partial R2 of 12.9%. In Population 2, two of the four QTL lost overlapped the two largest EH QTL. The remaining six original RPR QTL lost from the populations were not linked to EH QTL detected in this study, and had relatively small effects on RPR, with partial R2s ranging from 2.5 to 7.5% and LOD scores ranging from 3.8 to 6.5. Some of the QTL lost after adjusting for EH likely represent the correlated QTL. However, not all of the QTL lost after adjustment will be correlated QTL because of sampling size for the QTL analyses. In experimental populations with less than 500 families, only a subset of QTL with small effects will be detected because of insufficient statistical power (Beavis, 1998). Several of RPR QTL detected in this study had LOD scores just above the significance threshold. These loci may not be significant after adjusting for EH simply because of the random sample of genotypes obtained.
There was a gain of eight RadjE QTL across populations when compared with the original RPR analysis: two, one, and five in the three populations. Two novel QTL in Populations 1 and 2 were highly significant with LOD scores of 6.3 and 6.5, and partial R2 values of 8.2 and 7.3%. The remaining novel RadjE QTL had small effects ranging from 2.5 to 4.8% and LOD scores ranging from 3.6 to 5.4. It is likely that these QTL were approaching the level of significance in the original RPR analysis, but were masked by EH. Removal of the confounding effects of EH caused by the correlation between RPR and EH allowed the QTL LOD score to exceed the significance threshold. Again, small population size may have caused only a subset of QTL to be detected (Beavis, 1998). The QTL that were just below the significance threshold in the RPR analysis may become significant simply because of sampling error within the population.
The amount of phenotypic variation explained by multi-locus models did not change substantially from RPR to RadjE for any of the populations (Table 2). This is not surprising when one considers the total amount of change resulting from the adjustment for EH. In Population 1, four QTL were lost, two QTL were gained, and two QTL were significantly increased in their effects, resulting in a slight reduction in total R2 from 33.4 to 32.0%. In Population 2, four QTL were lost, one was gained, and two were significantly increased in their effects, resulting in slight a decrease in total R2 from 44.7 to 41.7%. In Population 3, three QTL were lost, five were gained, and one QTL was significantly increased in its effect, resulting in a change in total R2 from 48.4 to 49.6%. A balance in changes caused the total R2 to remain the same; lost QTL caused a decrease in total R2, while the novel QTL increased total R2. While the RPR QTL represent a directly measured field trait, the RadjE QTL should be a better genetic estimate of the physiological trait of stalk strength independent of position of the ear on the stalk.
Interestingly, though, there was still a significant negative correlation between RadjE and EH in Populations 2 and 3, r = -0.31 (P < 0.01) and r = -0.15 (P < 0.05), respectively. The adjustment made to RPR did not remove all EH effects. This was surprising because of the lack of QTL in common between RadjE and EH for these populations. In Population 2, where the absolute value of the correlation was higher, there were no QTL in common between RadjE and EH, while in Population 3, where the correlation was lower, there were three QTL in common between RadjE and EH.
These results demonstrate that EH has a significant effect on RPR as reflected in the negative correlation. When the effects of EH were reduced, 11 of the 26 original RPR QTL lost their significance and eight gained significance. On the other hand, RPR is not merely determined by the ear's position on the plant. The majority of RPR QTL, 15 of 26, remained significant after adjustment for EH, and 11 EH QTL were detected that weren't linked to or overlapped RPR QTL. Moreover, the original RPR variable was still highly correlated with the adjusted variable, RadjE (Table 1).
There are many examples of phenotypic correlation between agronomic traits in maize, i.e., EH and flowering date (Beavis et al., 1994; Veldboom and Lee, 1996), etc. Photoperiod sensitivity may be used to explain partially the correlation between these two traits; however, EH and flowering date likely have distinct developmental and physiological pathways. Ultimately, dissecting the genetic basis of complex traits to their underlying physiological and biochemical bases is essential to enhance predictive modifications of crop plants (McMullen et al., 1998; Beavis et al., 1991). Knowledge of the proper physiological and biochemical pathways underlying the trait is also necessary to isolate the correct candidate genes for QTL when candidate genes for a correlated trait are located nearby.
The results of our study confirm the importance of examining correlated traits when conducting QTL experiments. Before initiating a QTL experiment, correlated traits which might influence measurement of the trait of interest should be determined. Then, one must ask if it is necessary to adjust for the correlated trait to improve measurements. To determine whether the correlated trait needs to be further studied, the variability of the correlated trait needs to be examined. A sample situation where analysis of correlated traits might be especially relevant is the stage of plant development during infestation–inoculation in entomology–pathology studies, particularly when a mixture of adapted and unadapted varieties is studied.
|
ACKNOWLEDGMENTS |
|---|
|
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Received for publication November 13, 2001.
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. A. Flint-Garcia, C. Jampatong, L. L. Darrah, and M. D. McMullen Quantitative Trait Locus Analysis of Stalk Strength in Four Maize Populations Crop Sci., January 1, 2003; 43(1): 13 - 22. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Vadose Zone Journal | |||
| Journal of Plant Registrations | Soil Science Society of America Journal | ||||
| Journal of Natural Resources and Life Sciences Education |
Journal of Environmental Quality |
||||
