| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a The School of Land and Food Sciences, The Univ. of Queensland, St Lucia QLD 4072 Australia
b CSIRO Plant Industry, Queensland Biosciences Precinct, 306 Carmody Rd St Lucia QLD 4067 Australia
c CIMMYT, Apdo. Postal 6-641, 06600 Mexico, D.F. Mexico
* Corresponding author (scott.chapman{at}csiro.au)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Abbreviations: BLUE, best linear unbiased estimator • BLUP, best linear unbiased predictor • CIANO, Centro de Investigaciones Agricolas del Noroeste • CIMMYT, International Maize and Wheat Improvement Center • DET, drought environment trial • G x E, genotype x environment • IAT, International Adaptation Trial • RIL, recombinant inbred line
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
However, the adoption of semidwarf wheats was not confined to the high input high yielding wheat production regions. Despite reports that semidwarf cultivars may have a yield penalty in less favorable environments where trial mean yields are generally below 2 Mg ha–1 (Keyes and Sorrells, 1989; Richards, 1992), Heisey et al. (1999) reported that 95% of all cultivars released in developing countries, including both favorable and unfavorable environments, contain one of these height reducing (Rht) alleles. Since most spring wheat programs utilize semidwarf backgrounds, it is rare for tall lines to be released. Consequently, tall lines developed before the widespread adoption of semidwarfs are generally susceptible to diseases that have arisen since their release and comparison with more recently developed, disease resistant semidwarfs is misleading.
Gale and Youssefian (1985) and Flintham et al. (1997) provide literature summaries highlighting the varying conclusions of studies into semidwarf grain yield performance compared with talls. The summarized studies encompass comparisons of the Rht alleles in winter and spring wheats, in bread and durum wheats, and under drought and heat stressed environments. With the exception of Laing and Fischer (1977), who compared nonisogenic dwarf and tall lines from the sixth and seventh International Spring Wheat Yield Nurseries grown in 44 global locations, all other studies were confined to specific regions; including winter wheats in northwest USA, Germany, the UK, Canada, and Hungary and spring wheat in Australia. Worland et al. (1994) showed the widespread distribution of the Rht-B1b and Rht-D1b alleles in European winter wheats. A recent study of a spring wheat population, grown in the central Great Plains of the USA, showed tall lines yielded the same or better than semidwarf isolines in four irrigated or nonirrigated environments (Butler et al., 2005). However, the performance of Rht isolines has not previously been studied, in either spring or winter wheats, over a wide range of global locations.
Singh et al. (2001) developed a set of near-isogenic lines in 10 modern CIMMYT (International Maize and Wheat Improvement Center) bread wheat and six durum wheat backgrounds. The lines are near-isogenic for Rht-B1b, except for Pavon 76 which carries the Rht-D1b allele. They assessed the effect of the Rht-B1b and Rht-D1b dwarfing alleles in six environments with varying levels of drought stress, and concluded there was no yield penalty for the semidwarfs even in the lowest yielding trial (2.5 Mg ha–1), and that in the more frequently irrigated treatments the semidwarfs yielded significantly more than the talls. Trethowan et al. (2001) studied the same germplasm and found that while the presence of a height reducing allele decreased both the coleoptile length and plant height, there was a significant genetic background effect for these traits among the different isolines relative to any height reducing alleles.
The International Adaptation Trial (IAT) is an investigative spring wheat yield trial distributed globally by CIMMYT. It was designed to improve knowledge of adaptive patterns in global spring wheat growing locations and genotype x environment (G x E) interaction using genotype performance data. In a trial like the IAT, we are interested in understanding some of the environmental factors associated with G x E. Determining appropriate environmental variates to directly characterize environments depends on the availability of climatic, soil, and crop management data. Cooper and Fox (1996) suggested using "probe" genotypes as an indirect approach to characterizing environments. When the contrasting performance of two (or more) genotypes is attributed to a specific environmental factor, e.g., water stress, then comparing these probe genotypes in a multienvironment trial can aid in environmental classification. A limitation of this approach is the dependence on the specifically chosen probe genotypes; a different set of genotypes potentially results in a different type of environmental characterization. The Rht near-isogenic pairs developed by Singh et al. (2001) were ideal probe genotypes to indicate environmental effects as they are near-identical lines with the exception of one allele (or allelic region) that has a major effect on plant height. In contrast to trials where "old" tall lines were used, these tall isolines have the same disease resistance attributes of the modern semidwarfs.
The objective of this research was to compare near-isogenic reduced height lines in different genetic backgrounds across a wide range of environments. Yield and height data from the globally distributed IAT and the managed drought environment trials, reported in Singh et al. (2001), were analyzed.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
Singh et al. (2001) estimated the average effects of Rht alleles over genetic backgrounds. Here we reanalyzed the original yield and height data to compare near-isogenic pairs derived from different genetic backgrounds in environments of varying, and known, water stress. The year by treatment factors were treated as one factor called "environment" with six levels (2 yr x three irrigation regimes). We use the notation "1IR97" to represent the environment with one irrigation harvested in 1997.
International Adaptation Trial (IAT)
The IAT was distributed globally by CIMMYT in 2000–2003 to representative spring wheat production regions. The 60 bread and 20 durum wheat cultivars, primarily of either CIMMYT or Australian origin, were chosen for their drought adaptation characteristics, their utility as probe genotypes to identify soil borne problems (abiotic and biotic), and for differential agronomic traits allowing environmental characterization. In general, bread and durum wheat lines were grown in separate two-replicate 
-lattice designs. All participants in the IAT were requested to apply fungicide and employ local agronomic management practices. Summary data and further information are available from http://www.cimmyt.org/GIS/iat_wheat (verified 27 September 2005).
Grain yield (Mg ha–1) and plant height (cm) data from 152 trials (99 locations) conducted between 2000 and 2003, were available. The dataset was filtered for low disease incidence and good data quality (nonzero trial genetic variance); yield data from 81 trials and height data from 55 trials was retained for analysis (Table 2). Twelve of the 81 trials were managed environment trials grown at CIANO, CIMMYT's arid–irrigated research station in northwestern Mexico, 26 were grown in Asia, and the remainder distributed among Europe, North and South America, the Middle East, and Africa (Table 2). Rainfall and temperature data were not always reported with the trials. In these cases we determined the seasonal rainfall from the nearest meteorological station (up to 100-km distance and<100-m altitude) available in the Global Summary of Day database (www.ncdc.noaa.gov/oa/ncdc.html; verified 13 Dec. 2005).
|
In the first stage, different experimental designs between the DET and IAT datasets required different mixed models. For the DETs, replicates and blocks were considered as random effects and genetic background and allelic type as fixed effects. All 80 lines in the IAT were required for the individual trial spatially adjusted models. Replicates and blocks were considered as random effects and genotypes as fixed effects. A fixed term for the Rht effect within each near-isogenic pair was included to test the significance of each allelic contrast per genetic background in each trial. For those IATs grown in CIANO, Mexico, spatial information was available and the best spatial models fitted. The best spatial models involved fitting the appropriate experimental design and modeling the residual variation with autoregressive row and column terms. Significant spatial effects (such as row trends) were then identified following Cullis et al. (1998). Observations with absolute standardized residuals greater than a value of 3 were identified as outliers, set to missing and the analysis repeated.
The best linear unbiased estimates (BLUEs) obtained from the individual trial analyses and their weights were used in the subsequent across trial analyses to calculate the adjusted means (BLUPs– best linear unbiased predictors) for use in pattern analyses. The weights, wij, for the ith genotype in the jth environment were computed as in Smith et al. (2001a),
 |
is the average of the sj2's. In the IAT dataset, only the BLUEs and weights for the Rht near-isoline subset were retained for the second stage analysis.
The standard second stage across trial mixed model for multienvironment trials is
 | [1] |
ijk are the residuals containing the plot errors. These are provided by the weights, wij, calculated from the trial error variances, sj2, in the first stage.
For these datasets, it was of interest to partition the genotype and genotype x environment variation into genetic background (bi), allele type (ak) and genetic background by allele type (b:aik), and their subsequent interactions with environment (ej). Thus the following model was fitted:
 | [2] |
ijkl are the residuals as before. To estimate the allele type effect for each genetic background, within each environment the genetic background (bi), genetic background by allele type (b:aik), and environment by genetic background by allele type (e:b:aijk) terms were considered as random. For the purposes of calculating variance components the above model was rerun for yield and plant height with ak (and therefore e:ajk) treated as random effects (Table 3).
|
Regression Analyses
For both datasets, diffYIELD was regressed against trial mean yield of the dwarf lines for each genetic background to examine the relationship between allelic difference and grain yield.
Coefficients of determination (R2) values for linear regressions in the larger IAT dataset were low, at 30%. The observed values in Fig. 1b suggested that there was a level for each background below which there was no relationship between trial mean yield and diffYIELD. Hence, segmented (broken-stick) regression was used to describe this relationship (Seber and Wild, 1989). The hinge (or knot) point for each genetic background was the trial mean yield (between 2 and 4 Mg ha–1 in steps of 0.5) that minimized the residual mean square (maximizing the R2) of the segmented regression. A hinge point represents the trial mean yield where there was a change in the rate of response of diffYIELD to trial mean yield of the semidwarf lines. A 95% confidence interval was calculated.
|
Biplots, from the principal components analysis, indicate the allelic and genetic discrimination among environments. The center point of a biplot approximates "average" yield in all environments, while the cosine of an angle between any two vectors is the genetic correlation between two environments, i.e., adjacent vectors represent highly correlated environments. A point represents the modeled observation for a genotype and the size of the effect of a genotype in an environment is estimated by the length of the environment vector from the center point to the point of intersection with a perpendicular line from the genotype point.
For the IAT dataset, the clustering methods described in DeLacy et al. (1996) were used to classify the large number of trials (81) into three environmental groups. The clustering analysis was conducted using the squared Euclidean distance as the proximity measure and Ward's method as the fusion criterion. Environmental variables (e.g., in-crop rainfall) were used to describe the classes obtained at the three group level.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Drought Environment Trials
The semidwarf b alleles, at Rht-B1 and Rht-D1, in mid- to high-yielding environments (trial mean yield >3 Mg ha–1) generally outperformed the tall a alleles for yield. Regardless of genetic background, diffYIELD was greater than the zero reference line in most cases (Fig. 1a). In the lowest yielding trial (1IR97) there was no significant difference in performance between the semidwarf and tall types within genetic backgrounds, although the genetic background main effect in this (and all other DETs) was significant (data not shown). In other environments, the magnitude of the response varied between backgrounds. For example, in the highest yielding trial (environment 6IR98) the smallest yield difference in bread wheats was for the Kauz-derived pair (0.52 Mg ha–1) and in durum the Lavanco-derived pair (0.51 Mg ha–1). The largest differences were in the Pavon- and Galvez-derived bread wheat pairs (1.71 and 1.70 Mg ha–1) (Fig. 1a). Although nets were used to support the plants, Singh et al. (2001) reported that some tall wheats lodged in this trial. Averaged across all environments, Pavon- and Galvez-derived tall lines were 117 and 123 cm tall whereas the Kauz-derived tall was one of the shortest "tall" lines at 108 cm and therefore less sensitive to lodging (Table 1).
In the 1997–1998 season, greater rainfall increased yield in the low irrigation treatments 1IR98 (3.95 Mg ha–1) and 2IR98 (5.02 Mg ha–1) compared with 1996–1997 (2.12 and 4.31 Mg ha–1). Notably, several tall durums yielded more than the short durums in 2IR98. Simple linear regression coefficients of diffYIELD regressed against trial mean yield showed the different responses of the Rht alleles in the various genetic backgrounds. Pavon and Galvez-derived pairs have larger responses (slopes of 0.27 and 0.26 respectively) than the Kauz-derived near-isogenic lines (slope of 0.10). As presented in Singh et al. (2001), the trial mean yields were significantly positively correlated with the number of irrigations (r = 0.88; n = 6, P < 0.05).
The effect of the Rht-B1b and Rht-D1b alleles on yield in the drought environment trials is evident in the biplot, Fig. 2a . Correlations of the first and second principal component loadings (horizontal and vertical biplot axes) with trial mean yield were 0.88 and –0.92, and 0.86 and –0.85 with trial mean height; (n = 6, P < 0.05). The first two principal components explained 79% of the genotype x environment interaction. Generally, the semidwarf lines (dark circles) yielded more in both drought and irrigated environments than talls (light circles), indicated by the semidwarf lines lying to the right of the tall lines and near to the higher yielding environments (largest diamonds).
|
The presence of the dwarfing alleles affected yield performance depending on genetic background and environment. In general, the durum wheat semidwarf lines were more closely associated with the irrigated environments (lower right-hand quadrat of the biplot) than the bread wheat semidwarfs. The durum wheats, with the exception of the Aconchi-derived isogenic pair, had lines (dotted) joining allelic pairs that were almost perpendicular to the average drought environment vector (vertical axis). This indicated no allelic effect within these pairs in drought environments. The semidwarf-tall lines for the Galvez and Pavon-derived isogenic pairs were parallel to each other with the Galvez-derived pair yielding more in drought environments than the Pavon-derived (it is further away from the center of the biplot), i.e., the Galvez-derived lines generally performed better under drought compared with the Pavon-derived lines. The 1BL.1RS bread wheats (Table 1) tended to have a shorter distance between isolines (solid lines) indicating a smaller allelic response to changing environments than non-1BL.1RS bread wheats (long dashed lines), e.g., compare the lines representing the Kauz- and Seri-derived pairs with those of Pavon- and Galvez-derivation.
These results indicate that the magnitude of the Rht allelic difference for yield and height may depend on the genetic backgrounds in which the near-isolines were developed, and that allelic type discriminates among environments and irrigation regimes.
International Adaptation Trial
A summary of the allelic effect on performance (yield and height) within different genetic backgrounds across the diverse IAT environments showed considerable differences in yield and height response to the absence/presence of the Rht-B1b (or Rht-D1b) allele (Table 1). Lines of Pavon and Galvez parentage were tallest, regardless of allele type. Their average semidwarf heights (88 and 91 cm, respectively) were 5 to 14 cm greater than the other semidwarfs. Their %diffYIELD was triple (18 and 21%) all other lines except Nesser derivatives (11%), while their %diffHEIGHT was on the median (–24%); this suggested the allelic difference in these backgrounds had a greater impact on yield than height, relative to isolines derived from other genetic backgrounds. In contrast, the durum wheats Yavaros- and Aconchi-derived near-isolines had the largest %diffHEIGHT, in absolute value, (31 and 36%) and the smallest %diffYIELD (7 and 6%). They were the shortest semidwarfs and their tall near-isolines had middle-range heights. The Kauz-derived pair also had a low %diffYIELD but contrasts with the durum and other bread wheats by having the smallest %diffHEIGHT, in absolute value, (16%) and the lowest number of trials where the semidwarf outyielded the tall (13%). %diffYIELD for the Nesser-derived pair was closer to the Kauz-derived pair overall, but the %diffHEIGHT was the same as the Pavon- and Galvez-derived lines.
In the IAT dataset, the yield difference (diffYIELD) generally increased with increasing yield of the semidwarf (Fig. 1b). The proportion of trials where the semidwarf (Rht-B1b or Rht-D1b) yielded statistically significantly more than the tall (Rht-B1a or Rht-D1a) line was largest for those developed in the Nesser, Pavon, and Galvez backgrounds (26–34%) while for the Kauz-derived pair and the durum wheats, Yavaros- and Aconchi-derived pairs, this proportion was much smaller (13–18%) (Table 1). The proportion of trials where the semidwarf yielded significantly less than the tall line was less than 11% for derived pairs from all backgrounds. In no trials did the semidwarf derived lines from Pavon and Nesser yield significantly less than their tall counterparts.
Figure 2b shows trials clustering on the yield performance of bread versus durum wheats and tall versus semidwarf, with semidwarf lines associated with higher yielding trials (larger diamonds). The first two principal components explain 65% of the genotype by environment variability; allele type (dwarf versus tall) clearly segregates along the first principal component (horizontal axis) and bread versus durum wheat along the second principal component (vertical axis). The correlations of trial mean yield with the first and second principal component loadings were 0.60 and –0.37 (n = 81, P < 0.05) and with height were 0.40 (n = 55, P < 0.05) and –0.07 (n = 55, P > 0.05), respectively.
The mean yields of the three environment groups on the basis of cluster analysis were 2.8, 4.3, and 4.8 Mg ha–1. The group with a mean yield of 2.8 Mg ha–1 was classified as stressed and 15 of the 17 trials in this group had <220 mm of in-crop rainfall (mean 105 mm), although 11 of these trials received some supplementary irrigation of an unknown amount (Table 2). Two of the stressed environment vectors are drought managed treatments at CIANO, Mexico, with negligible in-crop rainfall (Table 2). The two higher yielding groups were considered nonstressed (mean yield about 4.5 Mg ha–1) and had a mean rainfall 180 mm with the low rainfall trials receiving irrigation. Some of the lower yielding trials in Fig. 2b (small diamonds) may also have been considered as stressed environments if the environmental data were available.
The pattern analysis summarizes the mixed model and regression analyses. Tall and semidwarf lines of the bread wheats were similarly adapted to stressed environments in the biplot (Fig. 2b), i.e., a perpendicular line from each isoline (within a genetic background) intersects the stressed environment vectors at a similar distance from the biplot center. The patterns of genetic background dispersal over the biplot of Fig. 2b were consistent with those presented in Fig. 2a where the environments were better defined. Near-isolines derived from Galvez and Pavon had the largest allelic yield response (longest distance between near-isolines) and Kauz-derived the smallest. Yavaros- and Aconchi-derived near-isolines performed similarly to each other, with Aconchi being more broadly adapted.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The removal of the Rht-B1b (or Rht-D1b) allele had a greater effect on yield and height in the 1BL.1BS bread wheats than the 1BL.1RS bread wheats (see Table 1), with the difference being more obvious in the IAT than the DET dataset. For example, the %diffYIELD and %diffHEIGHT in the IAT for Pavon and Galvez-derived near-isogenic pairs were similar (18 and 21%, and –24% respectively) and much larger than the same effects in the 1BL.1RS Kauz-derived pair (7% and –16%). In dryland and irrigated environments in northern Australia, several 1BL.1RS lines (Genaro and Seri among them) have been shown to yield 15 to 20% more than local 1BL.1BS high yielding lines (including Hartog) (Cooper et al., 1994). However, Peake et al. (1996), showed that among recombinant inbred lines (RILs) from Hartog/Seri and Hartog/Genaro crosses, that there was no significant difference in yield averaged across environments between RILs that were homozygous for 1BL.1BS compared with RILs homozygous for 1BL.1RS. Therefore the differences we saw between the 1BL.1BS and 1BL.1RS lines may be a result of using only one isogenic pair derived from each genetic background.
Although reports vary depending on environmental conditions and genetic backgrounds, it has been frequently noted that semidwarfs perform better than their tall counterparts in high yielding environments, while in lower yielding environments their yield response was similar or sometimes reversed (Flintham et al. (1997) and Butler et al., 2005). These results were confirmed here in both the smaller, structured DETs and the larger, climatically and geographically diverse trials of the IAT. The diffYIELD increased in the DET linearly with increasing trial mean yield (Fig. 1a). However, the lowest yielding DET (1IR97) had a trial mean yield of 2.5 Mg ha–1, which was high compared with the lower yielding trials in the IAT, where 13 had trial mean yields less than 2.5 Mg ha–1, with the lowest trial mean yield of 0.69 Mg ha–1 in Kenya. The hinge points (vertical dashed lines in Fig. 1b) for the segmented regression of each background ranged from 2.5 to 3.5 Mg ha–1. Where the trial mean yield was above the average hinge point of 3 Mg ha–1, the semidwarfs yielded more than the talls in 54% of trials; where the trial mean yield was less than this point, only 24% of the trials had semidwarfs outyielding the talls. The talls were superior to the semidwarfs in only 5% of trials. Despite these general trends, there were still some high yielding IATs (>6 Mg ha–1), where the tall lines in all genetic backgrounds were as high yielding as their semidwarf counterparts, and in the case of the durum wheats, Aconchi and Yavaros, higher yielding.
Biplots produced from pattern analysis of the two datasets (Fig. 2a and b) showed that genetic background and allele type varied with environment. The semidwarf bread wheats all had similar adaptation across both stressed (irrigated) and nonstressed (nonirrigated) environments; they tended to cluster in the top right hand quadrant of both biplots, at the point of broadest adaptation (high yielding across a range of environments). Semidwarf durums performed well in nonstressed environments, while the variation among the tall isolines was more pronounced. The tall lines derived from Galvez and Pavon were least adapted to all environments, whereas the tall lines derived from Kauz and Nesser were associated with lower yielding trials than their semidwarf pair. In particular, the Kauz-derived tall was equally or better adapted than the Kauz-derived semidwarf to the 17 stress environments in Fig. 2b.
The rapid global spread of the Rht-B1b and Rht-D1b dwarfing alleles during the Green Revolution allowed wheat breeders to improve wheat with less risk of lodging. Hence, improvements in yield, quality, and disease resistance have, in general, been made in the presence of one of these dwarfing alleles. However, tall varieties may be desirable for reasons other than high yield, including high biomass and longer straw lengths. The longer coleoptiles and larger root systems of tall cultivars compared with the semidwarfs likely contributes to their relative adaptation to dry environments, where deep sowing ensures contact with available soil moisture (Rebetzke and Richards, 2000; Trethowan et al., 2001). Poor emergence from depth because of short coleoptiles is a known problem with Rht-B1b and Rht-D1b alleles. This has lead to investigation of alternative Rht genes (Rebetzke and Richards, 2000) and/or additional genes that increase coleoptile length in Rht-B1b or Rht-D1b backgrounds while maintaining semidwarf height.
Richards (1992) and Flintham et al. (1997) suggest that the optimal height for wheat to obtain maximum yield is between 70 and 100 cm. Certainly the semidwarf lines were superior in most environments. However, other studies have proposed breeding "tall dwarfs" for adaptation to dry environments, e.g., Börner et al. (1993), Budak et al. (1995), and Law et al. (1978). In our experiments, the comparatively short statured Kauz-derived tall yielded only 7% less than the semidwarf, although it was 16% taller. In no environment below 2.5 Mg ha–1 was the Kauz-derived semidwarf higher yielding than the tall. This supports the arguments of Rebetzke and Richards (2000) and Trethowan et al. (2001) that there may be a place for "short talls" in low yielding environments.
|
ACKNOWLEDGMENTS |
|---|
Received for publication May 8, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 | |||
for semidwarf and tall isolines in the Drought Environment Trial and International Adaptation Trial (IAT). The %diff is (semidwarf meanTRAIT – tall meanTRAIT)/semidwarf meanTRAIT x 100. For the IAT a summary of the % trials where the semidwarf isoline yields more, less, or the same as the tall is provided, as are the average days to flowering from sowing. Yavaros, Aconchi, Focha, Nehama, Lavanco, and Bichena are durum wheats; all others are bread wheats.
