| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a CIMMYT, Int. Apdo. Postal 6-641, 06600 México, D.F., Mexico; A.S.I. Saad, ARC, Wad Medani, Sudan
b CSIRO, Plant Industry, GPO 1600, Canberra, ACT, 2601, Australia. M.P. Reynolds secondary address: ACPFG, Adelaide, Australia
* Corresponding author (m.reynolds{at}cgiar.org).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and MethodsResultsDiscussionREFERENCES |
|---|
Abbreviations: ANT, days to anthesis • BM, dry aboveground biomass at maturity • BMA, biomass shortly after anthesis • CHL, flag leaf chlorophyll shortly after anthesis • CHO, soluble carbohydrate content of stems shortly after anthesis • CID, carbon isotope discrimination of well-watered leaves • COND, stomatal conductance (flag leaves shortly after anthesis) • CT, canopy temperature • CTV/CTG, canopy temperature during vegetative/grainfilling stages • GEI, genotype by year interaction • HI, harvest index • LI, light interception • nBM, dry aboveground biomass at maturity normalized by maturity date • nBMA, biomass shortly after anthesis normalized by anthesis date • NDVI, normalized difference vegetative index (used to estimate relative biomass before heading) • OA, osmotic adjustment between well-watered and drought-stressed leaves • PCA, principal component analysis • RARSc, ratio analysis of reflectance spectra to estimate carotenoid pigments in the canopy • RUE, radiation use efficiency • TE, transpiration efficiency • TGW, thousand grain weight • WI, water index (spectral reflectance index associated with water content of the canopy) • WU, apparent water use based on gravimetric soil measurement • WUE, water use efficiency • WUEa, apparent water use efficiency (biomass/WU) • YLD, yield
|
ACKNOWLEDGMENTS |
|---|
Received for publication April 10, 2007.
a CIMMYT, Int. Apdo. Postal 6-641, 06600 México, D.F., Mexico; A.S.I. Saad, ARC, Wad Medani, Sudan
b CSIRO, Plant Industry, GPO 1600, Canberra, ACT, 2601, Australia. M.P. Reynolds secondary address: ACPFG, Adelaide, Australia
* Corresponding author (m.reynolds{at}cgiar.org).
While genetic resources provide an invaluable gene pool for crop breeding, the majority of accessions in germplasm collections remain uncharacterized and their potential to improve stress adaptation is not quantified. A selection of 25 elite genetic resources for wheat (Triticum aestivum L.) were characterized for agronomic and physiological trait expression in drought- and heat-stressed environments. Under drought, the physiological traits best associated with yield were canopy temperature, associated with water uptake, and carbon isotope discrimination, associated with transpiration efficiency. Under heat stress stomatal conductance, leaf chlorophyll content, and canopy temperature (associated with radiation use efficiency in this environment) were well correlated with yield. Theoretical yield gains based on extrapolating the best trait expression to the highest yielding backgrounds were also estimated. Under drought, the best expression of canopy temperature and carbon isotope discrimination suggested potential yield gains of approximately 10 and 9% above the best yielding cultivars, respectively; under heat stress, canopy temperature and remobilization of stem carbohydrates suggested potential yield gains of approximately 7 and 9%, respectively. Other physiological trait expression was associated with potential yield gains to varying degrees. When considering agronomic traits, the best expression of harvest index suggested yield gains of approximately 14 and 24% in drought and hot environments, respectively, while the combined best expression of both harvest index and final aboveground biomass suggested yield gains of 30 and 34%, respectively. Principal component analysis indicated that many of the physiological traits that were associated with yield and biomass were not strongly associated with each other, suggesting potential cumulative gene action for yield if traits were combined. When comparing trait expression across drought and hot environments, several physiological traits (e.g., canopy temperature) showed closer association with each other than did performance traits, supporting the idea that such stress-adaptive traits have generic value across stresses.
Abbreviations: ANT, days to anthesis • BM, dry aboveground biomass at maturity • BMA, biomass shortly after anthesis • CHL, flag leaf chlorophyll shortly after anthesis • CHO, soluble carbohydrate content of stems shortly after anthesis • CID, carbon isotope discrimination of well-watered leaves • COND, stomatal conductance (flag leaves shortly after anthesis) • CT, canopy temperature • CTV/CTG, canopy temperature during vegetative/grainfilling stages • GEI, genotype by year interaction • HI, harvest index • LI, light interception • nBM, dry aboveground biomass at maturity normalized by maturity date • nBMA, biomass shortly after anthesis normalized by anthesis date • NDVI, normalized difference vegetative index (used to estimate relative biomass before heading) • OA, osmotic adjustment between well-watered and drought-stressed leaves • PCA, principal component analysis • RARSc, ratio analysis of reflectance spectra to estimate carotenoid pigments in the canopy • RUE, radiation use efficiency • TE, transpiration efficiency • TGW, thousand grain weight • WI, water index (spectral reflectance index associated with water content of the canopy) • WU, apparent water use based on gravimetric soil measurement • WUE, water use efficiency • WUEa, apparent water use efficiency (biomass/WU) • YLD, yield
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and MethodsResultsDiscussionREFERENCES |
|---|
While the revolution in genetics has already provided useful markers associated with genetically simple traits in wheat (Triticum aestivum L.), mostly for biotic stress resistance (Bonnet et al., 2005; William et al., 2007), cost-effective biotechnological interventions have yet to be developed for genetically more complex characteristics such as adaptation to abiotic stress (Snape, 2004; Varshney et al., 2005). Nonetheless, polygenic physiological traits are used as proxy genetic markers in crop improvement (Bolaños and Edmeades, 1996; Araus et al., 2002; Condon et al., 2002, 2004; Rebetzke et al., 2002; Richards et al., 2002; Reynolds and Trethowan, 2007) and there is renewed interest in physiological methods. Specifically, there is a manifest need for information on the phenotypic responses of crop genotypes to environment to (i) complement molecular mapping exercises (Snape, 2004); (ii) permit more focused experimental design in genomics experiments through understanding the way physiological trait expression interacts with phenology and environment (Dingkuhn et al., 2005); and (iii) accelerate the identification of candidate genes that will impact on crop productivity when used in transformation (Langridge et al., 2006).
However, both the application of genetic technologies and the generation of high quality phenotypic data sets on large populations in different environments are costly. Since a great number of physiological traits have been shown to be associated with potential crop improvement under abiotic stress (Blum, 1988; Boyer, 1996; Araus et al., 2002; Bruce et al., 2002; Chaves et al., 2003; Wang et al., 2003; Condon et al., 2004; Richards, 2006; Foulkes et al., 2007; Reynolds et al., 2007), there is an imperative to focus on those likely to provide the largest return on investment. One approach to achieving this is to measure phenotypic expression of stress-adaptive traits among elite genetic resources and estimate a theoretical impact associated with the best trait expression in high yielding backgrounds (Reynolds and Condon, 2007). Related to the issue of choosing worthwhile candidate traits is the question of whether "generic" or "unified" stress-adaptive traits exist that may be associated with genetic gains under different abiotic stresses (Yang et al., 2002; Plaut et al., 2004; Shaheen and Hood-Nowotny, 2005; Langridge et al., 2006). Identification of traits and their genes that confer both drought and heat adaptation, for example, represents a potentially larger return on research investment since outcomes would have application in a broader context.
While there is currently a strong emphasis on investigating traits associated with the response of cellular metabolism to abiotic stress (Bruce et al., 2002; Chaves et al., 2003; Wang et al., 2003), such work is generally not conducted in the environment in which crops grow, making extrapolation to crop improvement tenuous (Passioura, 2004). For these reasons, the International Maize and Wheat Improvement Centre's (CIMMYT) Wheat Program developed a general conceptual model of drought adaptation in wheat that encompassed a broad range of traits for which evidence exists in the literature of potential application in breeding (Reynolds et al., 2005). The model describes four main groups of traits relating to
The conceptual model is presented (Fig. 1a
) indicating traits in relation to the main drivers of yield under drought, namely water use (WU), WUE, and harvest index (HI) as shown in Eq. [1] (Passioura, 1977):
 | [1] |
 | [2] |
|
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and MethodsResultsDiscussionREFERENCES |
|---|
|
Relative estimates of flag leaf chlorophyll shortly after anthesis (CHL) were estimated using a Minolta SPAD meter (SPAD 502, Minolta, Spectrum Technologies Inc., Plainfield, IL). Early estimates of relative biomass were measured remotely using the spectral reflectance index normalized difference vegetative index (NDVI) as described by Gutiérrez-Rodríguez et al. (2004). For the drought environment this was measured once during crop establishment approximately 20 d after emergence and for the hot, irrigated environment it was measured once during the early boot stage. Two other spectral reflectance indices were measured during boot stage; the water index (WI), which is affected by water content of the canopy and shows correlation with aboveground biomass as described by Babar et al. (2006b), and an index which uses ratio analysis of reflectance spectra to estimate carotenoid pigments (RARSc) in the canopy (Chapelle et al., 1992). Canopy temperature was measured with an infrared thermometer three times during vegetative or boot stage (CTV) and again during grainfilling (CTG), as described elsewhere (Reynolds et al., 1998). Stomatal conductance (COND) was measured in the hot, irrigated environment only, on three flag leaves per plot shortly after anthesis using a steady state porometer (Leaf Porometer, Decagon Devices, Pullman, WA).
Apparent water use (WU) was determined by measuring gravimetric water content at the following profile depths: 0 to 30 cm, 30 to 60 cm, 60 to 90 cm, and 90 to 120 cm. This was determined at six random locations per experiment at crop emergence, and on two spots of every plot after harvest. When calculating available water, it was assumed that, in addition to the water applied as rain and irrigation water, additional fluxes associated with evaporation of water from the soil surface after crop emergence and gains associated with heavy dew (an almost daily phenomenon in this environment) cancelled each other out. This is not necessarily a good assumption (Reynolds and Condon, 2007), although it is necessary given the lack of information on these fluxes or genotype interactions. Values of permanent wilting point of 18 to 20% volumetric soil water content, depending on soil profile, and a soil bulk density of 1.3 were assumed, based on previous samplings at the same locality (K.D. Sayre, personal communication, 2007). Apparent water use efficiency was calculated using apparent total water extracted for each genotype and final biomass. The value is apparent as it does not consider plant uptake or drainage beyond 120 cm, direct losses from the soil surface especially during crop establishment, or influxes from dew.
Carbon isotope discrimination was measured on leaves sampled from irrigated plots approximately 25 d after emergence using mass spectrometry to provide relative estimates of constitutive transpiration efficiency of genotypes as described by Condon et al. (2002). Osmotic adjustment (OA) was estimated in soil-potted plants in a glasshouse during the summer of 2005 in El Batan, Mexico. Genotypes were grown in shared pots such that the genotypes of a statistical subblock (from the lattice designs) were grown together to ensure common soil water potential. Two leaves were sampled for each genotype from both well-watered and drought-stressed treatments. Average soil water potentials, estimated from predawn leaf water potentials, were –0.6 MPa for well-watered treatments and –1.7 MPa for drought treatments. Leaves were sampled early in the morning, 12 h after rewatering to cause leaf hydration on plants that were covered with plastic bags overnight and to ensure that leaves were not subjected to desiccation. Tissue was quickly dried of all surface moisture and placed in a 2-mL propylene tube, sealed with the lid and placed in a deep freeze to rupture the cells. After this, a drop of cell sap was extracted using a glass rod and placed on the sampling cuvette of a vapor pressure osmometer (Wescor model 5500, Logan, UT). Values of osmotic potential obtained from the leaf sap (in mmol kg–1) were then expressed in megapascals by dividing by the conversion factor 0.02525. Osmotic adjustment was calculated as the difference in osmotic potential between drought-stressed and well-watered treatments, averaged across treatments.
After physiological maturity was reached, yield (YLD) was measured by machine-harvesting a bordered area of 4.8 m2. Before that, a random subsample of 100 spike-bearing culms was removed from each plot, dried, weighed, and threshed, so that HI could be estimated. Using these data and an estimate of individual kernel weight (thousand grain weight [TGW]), grains per square meter and final aboveground biomass (BM) were calculated.
Germplasm
Genotypes selected for this study came from several sources of germplasm, including advanced lines and commonly used parents from CIMMYT's spring bread wheat breeding program for dry areas (Trethowan and Reynolds, 2007), as well as genetic resources such as landraces (Reynolds et al., 2007) and lines derived from products of interspecific hybridization (Trethowan et al., 2003; Reynolds et al., 2007). The lines used had been evaluated under drought and/or heat stress in previous field cycles and individuals were identified for high expression of one or more of the physiological or agronomic traits (Tables 2
and 3
).
|
|
The proportion of the total sums of squares accounted for the GEI for each trait was obtained from a combined analysis of variance conducted with the GLM procedure from SAS (2004), with all the effects, years, replications within years, blocks within years and replications, genotypes and GEI being considered as fixed effects. The multiple regression was conducted with the REG procedure from SAS (2004) using the stepwise selection procedure. Broad sense heritability (h2) was estimated for each trait over the 3 yr as
 | [3] |
2 = error variance, g2 = genotypic variance, and ge2 = GEI variance. Similarly the genetic correlation can be estimated as
 | [4] |
g = genetic correlation between the traits X and Y, gxy = genetic covariance between the traits, and gx2 and gy2 are the genetic variances for traits X and Y, respectively. The genetic covariance was estimated using the statistical property of the sum of two random variables, which states: (x+y)2 = x2 + y2 + 2xy which can be rearranged and written as
 | [5] |
For both broad sense heritability and genetic correlation, all the variance components were estimated using the MIXED procedure from SAS (2004) considering all the terms in the model (environments, replications within environments, blocks within replications and environments, genotypes and GEI) as random effects.
Since the traits were measured in different units, we performed the PCA based on the correlation matrix using the PRINCOMP procedure from SAS (2004), and then graphed the first two eigenvectors.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and MethodsResultsDiscussionREFERENCES |
|---|
The range of expression for agronomic and physiological traits under both heat- and drought-stressed environments is presented as a comparison of the two genotypes showing the best and worst expression of traits, respectively, using the main effects of genotypes across years (Table 3). To demonstrate the degree to which stress-adaptive traits are expressed in the highest yielding genotypes in comparison to the most favorable expression observed across all genotypes, expression of each trait averaged across the two best yielding genotypes is also presented.
The rationale for grouping traits in Tables 3 to 5
was as follows. For the drought environment, water uptake included canopy temperature (CT), which has been associated with water uptake under moisture stress (Reynolds et al., 2007). NDVI, an indicator of early ground cover, is implicated in differential loss of soil water to the atmosphere (Richards et al., 2002). Water use efficiency included anthesis date, as earliness can permit escape from intense terminal stress. RARSc is associated with photoprotective pigments (Araus et al., 2001) that would be anticipated to increase transpiration efficiency. Apparent WUE (i.e., WUEa = biomass/apparent water uptake) is assumed to be associated with actual WUE. Carbon isotope discrimination is directly associated with transpiration efficiency (Condon et al., 2002). Harvest index included CHO and BMA as their contribution to yield and was considered in the context of partitioning of soluble carbohydrates to grain filling.
|
|
It is clear that traits within each of these drivers may have independent genetic mechanisms. For example, crop water use is influenced by early ground cover as well as root depth. Water use efficiency is influenced by intrinsic transpiration efficiency, but adaptive mechanisms, such as osmotic adjustment or photoprotective mechanisms, may also be influential. Harvest index is a function of constitutive factors relating to Rht genes, but can also be influenced by the proportion of stem mass in the form of soluble stem carbohydrates that are remobilized to grains. On the other hand expression of facultative traits may show genetic overlap between trait groups, a good example being HI which can interact with other drivers of yield (Eq. [1]) during grain filling (Richards et al., 2002).
Theoretical Yield Gains Associated with Stress-Adaptive Traits
Two approaches were taken to estimate theoretical yield gains (Tables 4 and 5). For the traits where equivalent mass could be estimated directly (BM, HI, BMA, CHO, WUE, WU, and TE), an arithmetic approach was adopted as described previously (Reynolds and Condon, 2007). However, since this material represented a significant range in phenology for all of the theoretical calculations described (Tables 4 and 5), values of biomass at anthesis and final biomass were normalized to the number of days between emergence and postanthesis harvest dates and emergence and physiological maturity date, respectively. Thus, in each environment, the difference in normalized final biomass (nBM) expression between the average value for three best yielding lines and three lines expressing the best nBM was multiplied by the HI value of the three best yielding lines to give an estimate of additional yield associated with best expression of nBM in the best yielding backgrounds (i.e., since yield = HI x BM). Similarly the difference in HI between the average value for three best yielding lines and three lines expressing the best HI was multiplied by the nBM of the three best yielding lines to give an estimate of additional yield associated with best expression of HI in the best yielding backgrounds.
Theoretical yield gains associated with expression of WUE, WU, and TE were calculated as follows: the difference in WUE expression between the average value for three best yielding lines and three lines expressing the best WUE were multiplied by the total amount of water available for the cycle and by the HI value of the three best yielding lines. Similarly, the difference in WU expression between the average value for three best yielding lines and three lines expressing best WU were multiplied by the WUE and by HI of the three best yielding. In the case of TE, theoretical values of biomass and yield were calculated for the best three yielding lines with a WUE corresponding to the proportional increases in TE associated with the average value of the three lines showing the lowest values of CID compared to the actual CID values of the best yielding lines. Expression of normalized biomass at flowering (nBMA) and CHO were considered in the context of additional yield that might be associated with soluble stem carbohydrates available for remobilization to grains. Therefore, the difference in nBMA expression between the average value for three best yielding lines and three lines expressing the best nBMA was multiplied by two coefficients, one representing the proportion of biomass as stems (shortly after anthesis) and the other representing the proportion of stem dry weight in the form of soluble carbohydrates (shortly after anthesis) in both cases using the average values of the three best yielding lines. Similarly the difference in CHO expression between the average value for three best yielding lines and three lines expressing the best CHO was multiplied by nBMA of the three best yielding lines and by the coefficient representing the proportion of biomass as stems (shortly after anthesis) using the average values of the three best yielding lines. For yield gains associated with soluble stem carbohydrates, estimated availability was discounted by 30% to account for probable respiratory losses during remobilization (Bugbee and Salisbury, 1988). The analysis described was performed for all cycles and theoretical yield gains are presented for drought-stressed (Table 4) and heat-stressed environments (Table 5). In those calculations (i.e., presented in Tables 4 and 5) the three best yielding lines were compared with the three lines showing best trait expression, as opposed to the two best lines (as presented in Table 3), to give a more conservative and, therefore, more robust estimate of potential genetic gains associated with trait expression.
For traits where theoretical yield gains could not be estimated arithmetically, linear models were used based on regression of biomass on the trait of interest. In these cases, theoretical yield gains were estimated as the difference in trait expression between the average value for three best yielding lines and three lines showing the best trait expression multiplied by the regression coefficient of the trait with nBMA and HI of the best three yielding lines. Since OA showed negative association with yield under drought in this germplasm, it was not possible to estimate any positive impact.
It is clear from these analyses that the best expression of all traits considered in this study was associated with potential theoretical yield gains in the majority of environments. However, the degree to which they may contribute to yield improvement varies considerably and will be discussed subsequently.
Association among Stress-Adaptive Traits
Considering the drought environment, PCA across all three crop cycles indicated that the physiological traits associated with yield were CTV, CTG, NDVI, RARSc, days to anthesis (ANT), and CID (Fig. 2
). There was a strong (negative) association between WU and CTG, indicating the causal link between water extraction from the soil and cool canopies in grain filling. However, this link was not apparent at preheading and may have been confounded by differences in water losses associated with evaporation from the soil surface, for which there seems to be evidence of genotypic effects (Reynolds and Condon, 2007). Both kernel weight and osmotic adjustment (not shown) were weakly and negatively associated with yield, suggesting that their expression was not contributing to drought adaptation in this germplasm. Yield was predictably strongly associated with grains per square meter, BM, and WUE. Multiple regression of drought-adaptive traits with both yield and BM in each of three cycles included the CT traits in five out of six models (Table 6
). In total, CTG and RARSc were selected four times, NDVI three times, CTV and WU twice, and CHO and ANT once each. The trait CID was included in both the yield and BM models for main effects of genotypes across years along with WU and CTV. These results suggest that all three drivers of yield under drought (Eq. [1]) were contributing to yield, with water use as indicated by CT, NDVI, and WU making the strongest contribution (not including traits autocorrelated with yield, i.e., biomass, TGW, WUEa, HI).
|
|
|
|
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and MethodsResultsDiscussionREFERENCES |
|---|
A number of physiological traits have been reported to be associated with yield in both heat- and drought-stressed field environments. These traits include remobilization of stem carbohydrates (Blum, 1998; Asseng and van Herwaarden, 2003) canopy temperature (Blum et al., 1989; Reynolds et al., 1998), ground cover (Rawson, 1986; Richards et al., 2002), and chlorophyll retention or stay-green (van Herwaarden et al., 1998; Reynolds et al., 2000). Therefore, this data set also provided an opportunity to test whether stress-adaptive traits that apparently have common value across environments showed similar phenotypic expression under drought- and heat-stressed conditions.
Theoretical Yield Gains Associated with Stress-Adaptive Traits
Theoretical yield gains were calculated individually for all three crop cycles in both drought and heat environments, as well as using main effects of genotypes across years (Tables 4 and 5). Conclusions focus mainly on the main effects across years as these are considered more robust, however, individual year effects are presented in the tables to give an idea of the range of possible outcomes. Due to the inevitability of interactions of genes or traits with genetic background, the reliability of such estimates is clearly somewhat subjective. This is especially the case for values predicted using linear regression where values of the coefficient of determination (r2) are not high or where deviations from linearity may occur at the high yield end of the relationship. Furthermore, their validity is also subject to the assumption that relationship between traits and yield in this collection of unrelated fixed lines is representative of any relationship that may occur in populations of sister lines to which selection pressure would be applied in breeding, which itself would depend on the genetic background of the individual parents. As an example, genetic gains associated with increased expression of OA could not be estimated at all because of their negative association with yield in this germplasm. This is despite the fact that the trait has been previously reported to be associated with genetic gains under drought in different environments (Morgan and Condon, 1986). The theoretical yield gains calculated with mass differentials and differences in water use are probably more reliable, since they are based on extrapolation from the best trait expression to the best yielding lines. This approach has the merit of being clear in its assumptions and no attempt was made to extrapolate beyond the effect of a single trait at a time. Simulation approaches, while representing a valuable attempt to model such processes quantitatively, have the drawback of compounding assumptions, since, when comparing contrasting genotypes, relationships between many of the traits are not sufficiently well understood to be accurately quantified. Precise frameworks dealing with interactions resulting from epistasis, pleiotropy, and gene by environment interactions are required for creating reliable empirical models to predict phenotypic responses (Cooper et al., 2005).
The main reason for adopting the approach presented was to attempt to compare the relative value of many traits (from the literature) in a single controlled experiment with the view to identifying the most promising ones for more detailed physiological and genetic analysis. From the point of view of systematically applying physiological or genetic markers for these traits in a large breeding program, data would be required from representative set of environments and genetic backgrounds, ideally in populations with controlled genetic structure. Examples of these kinds of studies in wheat can be found in restricted numbers of genetic backgrounds and environments for traits, including osmotic adjustment (Morgan and Condon, 1986), stem carbohydrates (Blum, 1998; Asseng and van Herwaarden, 2003), canopy temperature (Blum et al., 1989, Reynolds et al., 1998; Olivares-Villegas et al., 2007), carbon isotope discrimination (Condon et al., 2002; Rebetzke et al., 2002, 2006), early ground cover (Richards et al., 2002), and spectral reflectance indices (Royo et al., 2003; Gutiérrez-Rodríguez et al., 2004; Babar et al., 2006a, 2006b).
When comparing the estimated theoretical value of each trait under drought over years or using the combined analysis, HI appeared to offer the most potential for yield improvement, though this is almost certainly represents the combined effect of a number of genetically more simple traits (Richards et al., 2002). Canopy temperature traits showed promise, suggesting theoretical yield gains in the region of 10%. This observation agrees with results of PCA and regression analysis, showing that traits associated with water use were making the strongest contribution to yield. Ability to extract water at depth is a characteristic that has been identified in selected Mexican landraces and the trait was strongly associated with cooler canopies (Reynolds et al., 2007). The result is consistent with studies in random inbred lines showing a high and consistent association of CT with yield under drought in the same environment (Olivares-Villegas et al., 2007). The trait CID also appeared to be promising suggesting theoretical yield gains of up to 9% from increased WUE; however, an assumption with this calculation is that genetic differences in TE are treated as constitutive, while there is some evidence that differences in TE may converge at low soil water potential (Condon and Richards, 1992). Nonetheless, the result is consistent with the successful application of CID as a selection criterion in breeding for WUE of wheat in Australia (Condon et al., 2002, 2004; Rebetzke et al., 2002, 2006). Best expression of stem carbohydrates shortly after anthesis (for remobilization during grainfilling) suggested significant though relatively smaller potential yield gains in this germplasm set (Table 4), as predicted by modeling studies (Asseng and van Herwaarden, 2003). Best expression of the spectral reflectance index RARSc—used to estimate carotenoid pigments in the canopy (Chapelle et al., 1992)—suggested small potential yield gains presumably associated with photoprotection and, therefore, improved WUE. It is not credible to infer what the combined effect of these traits might be in a common genetic background for reasons already stated. However, when considering the effect of the more integrative agronomic traits, best expression of biomass (normalized by maturity date) multiplied by the best expression of HI suggested potential yield increases of 30% (Table 4).
Under the heat-stressed environment, best expression of physiological traits in all three drivers of yield showed theoretical potential to improve yield including CHO associated with partitioning, NDVI associated with light interception, and CT associated with RUE, with values in the range of 7 to 9% (Table 5). The results are consistent with previous studies indicating the value of CT (Reynolds et al., 1998) and CHO (Blum, 1998) in the hot environment. Considering integrative agronomic traits, the effect of the best expression of biomass (normalized by maturity date) multiplied by the best expression of HI suggested that, if these two traits could be combined, yield increases would average 34% (Table 5).
Association among Stress-Adaptive Traits
In general, the spread of the vectors for PCA for drought (Fig. 2) suggested that many of the traits that were associated with yield and BM under drought were not necessarily associated with each other, indicating the possibility of cumulative gene action if traits were to be combined in a common genetic background. This conclusion was backed by multiple regression, where CT traits, presumably reflecting the ability to access water at depth, were complemented by traits that would be expected to show a degree of genetic independence as indicated by PCA (Fig. 2). These traits included CID and RARSc and, to a lesser extent, NDVI and CHO.
In general the spread of the vectors for PCA for heat (Fig. 3) also suggested that the traits associated with yield and BM under heat were not necessarily associated with each other. However, multiple regression analysis identified mainly stomatal aperture traits, CT, and leaf conductance presumably reflecting the ability to meet evaporative demand. Additional traits identified by regression analysis as potentially cumulative were either CHL or WI (the two were associated with each other).
Comparison of Stress-Adaptive Trait Expression under Drought and Heat Stress
When comparing expression of individual traits in both the drought- and heat-stressed environments using PCA, the physiological traits generally showed closer association with each other than did the performance traits (Fig. 4). This is perhaps not surprising, as genetically more complex traits would be expected to show larger GEI. This supports the idea that some traits that are beneficial in drought conditions may also be valuable under heat-stressed conditions (Reynolds and Trethowan, 2007). The latter presented a conceptual model that included groups of traits, of which some were considered to be of generic value across these stresses (Fig. 1). For example, rapid ground cover (estimated by spectral reflectance index NDVI) avoids wasteful evaporation of soil water under pre-anthesis drought stress (Loss and Siddique, 1994; Richards et al., 2002), while, under hot, irrigated conditions, it increases light interception and, therefore, assimilation capacity, thereby reducing the risk of reduced tillering normally associated with accelerated development rate at high temperature (Rawson, 1986). High expression of stem carbohydrates provide an extra source of assimilates for grain growth when either drought or heat stress is experienced during grain filling stage (Blum, 1998; Asseng and van Herwaarden, 2003). Root systems that permit increased access to soil water are indicated by a relatively cool canopy (Reynolds et al., 2007) and have obvious benefit under drought (Blum et al., 1989; Olivares-Villegas et al., 2007), but can also permit plants to more easily match the high evaporative demand associated with hot, low relative humidity environments, resulting in higher leaf gas exchange rates and heat escape through evaporative cooling (Reynolds et al., 1998). The fact that these traits, NDVI, CHO, and CT, were closely associated across drought- and heat-stressed conditions (Fig. 4) endorses the idea of their generic stress-adaptive nature. Molecular marker studies in the wheat recombinant inbred lines population mentioned earlier (Olivares-Villegas et al., 2007) have already identified common quantitative trait loci for CT across the drought- and heat-stressed environments used in the current study (R.S. Pinto et al., unpublished data, 2007), strongly supporting the notion of some common genetic bases for CT expression under these two stresses.
Implications for Breeding
The vast majority of accessions in germplasm collections remain uncharacterized in terms of their potential to improve yield under abiotic stress. Results presented from the current study support the following interventions in breeding and prebreeding: (i) investment in phenotyping to identify elite sources of traits among genetic resources and to estimate potential yield gains associated with trait expression; (ii) simultaneous evaluation of candidate traits in controlled phenotyping studies to define those which are potentially complementary; (iii) adoption of a crossing strategy where parents with potentially complementary traits are hybridized with the specific objective of realizing cumulative gene action (Reynolds et al., 2005; Reynolds and Trethowan, 2007); (iv) and determination of which traits may have generic value across different stress environments as these represent a potentially larger return on research investment.
Although it cannot be predicted with any degree of certainty that a specific combination of traits will be cumulative due to the complex nature of gene action, these data nonetheless show that a relatively large proportion of the phenotypic variation in performance under drought and heat can be explained by a small number of traits (Table 6). Furthermore, estimates of broad sense heritability for these traits and their genetic correlation with yield (Table 3) indicated that, in most cases, they would be amenable to reliable quantification in parents and verification of expression in segregating progeny.
The analysis also provides evidence that certain traits, for example canopy establishment and cooler canopies, may be considered generic across drought- and heat-stressed environments. It is perhaps fortuitous that both traits are easy to measure, even on a large scale, using spectral reflectance techniques (Araus et al., 2001, 2002; Royo et al., 2003; Babar et al., 2006b) and infrared thermometry (Blum et al., 1989, Reynolds et al., 1998; Olivares-Villegas et al., 2007), respectively.
The authors acknowledge the Australian Grains Research and Development Corporation (GRDC) for financially supporting research in Mexico. They also acknowledge Julian Pietragalla and Araceli Torres for statistical analysis and Eugenio Perez and Jose Luis Barrios for technical assistance.
Received for publication April 10, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and MethodsResultsDiscussionREFERENCES |
|---|
