| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Dep. of Agronomy, Univ. of Nebraska, Lincoln, NE 68583-0915
b Dep. of Biology, Univ. of Utah, Salt Lake City, UT 84112
c Pioneer Hi-Bred, 4405 Winsor Court, St. Joseph, MO 64506
d Dep. of Agronomy, Agric. College, GyeongSang National Univ., 900 Gaza-Dong, Chinju City, Geyong-Nam, South Korea, 660-701
e Bio Nebraska, 3820 N W 46, Lincoln, NE 68524
f Dep. of Agronomy and Plant Genetics, Univ. of Minnesota, St. Paul, MN 55108
Corresponding author (jspecht1{at}unl.edu)
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARYREFERENCES |
|---|
Abbreviations: beta, linear regression coefficient • CID, carbon isotope discrimination • ET, evapotranspiration • G x W, genotype x water • G x WL, linear portion of G x W • HI, harvest index • LG, linkage group • QTL, quantitative trait locus • RIL, recombinant inbred line • SY, seed yield • T, transpiration • TE, transpiration efficiency • VPD, vapor pressure deficit • WUE, water-use efficiency • 
13C, sample isotopic abundance ratio 13C/12C (expressed as a deviation from the standard ratio)
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARYREFERENCES |
|---|
Carter (1989) noted that while drought tolerance is a desirable trait, it is not an explicit selection criterion in most breeding programs. Instead, many breeders consider selection for wide adaptation (i.e., genotypes whose yields are high in most test environments) to be a coincident means of selecting for drought tolerance (Rosielle and Hamblin, 1981). Nevertheless, genotype x water (G x W) interaction exists (Korte et al., 1983; Kadhem et al., 1985; Specht et al., 1986), suggesting that selection for specific adaptation (i.e., genotypes with relatively high yields in targeted environments, but relatively lower yields in others) might offer a more direct means of genetically improving drought tolerance.
The term drought tolerance is used copiously in the literature, yet it does not intuitively invoke a universal definition in the minds of those who use or encounter it. In fact, its definition can range from purely mechanistic to purely empirical (Specht and Williams, 1984).
Mechanistic definitions fall into three general categories (Ludlow and Muchow, 1990; Nilsen and Orcutt, 1996), and each has an implicit selection target. (i) For drought escape, selection is aimed at those developmental and maturation traits that better calibrate water-sensitive periods of crop ontogeny with the seasonally recurrent meteorological patterns of the targeted production area (i.e., breeding for adaptation). (ii) For dehydration avoidance, selection is focused on traits that lessen evaporatory water loss from plant surfaces (e.g., Clawson et al., 1986; Zhang et al., 1992) or maintain water uptake during drought via a deeper and more extensive root system (e.g., Goldman et al., 1989). (iii) For dehydration tolerance, selection is directed at maintaining cell turgor (a driving force for plant growth) or enhancing cellular constituents that protect cytoplasmic proteins and membranes from desiccation. While these mechanistic forms of drought tolerance can improve plant survival (Ludlow and Muchow, 1990; Jones, 1993), they often do so at the expense of potential plant productivity (i.e., the yield attainable in optimum environments). As a result, breeders are wary of any mechanistic trait that lacks an explicit association with crop yield.
The empirical, or yield-based, definitions of drought tolerance fall into two categories (Specht et al., 1986). One is absolute—select the highest-yielding genotypes in environments where seasonal drought is predictably recurrent (again, breeding for adaptation). The other is relative—select genotypes with the smallest yield decline per unit of reduced seasonal rainfall. The drawback to empirical selection is that genetic progress is only sustainable if genotypic performance can be annually evaluated in a specified drought condition. Reliably recreating a specific drought condition year after year is extremely difficult.
Passioura (1977) observed that in water-limited environments, seed yield (SY) was a simple function of total seasonal transpiration (T), final water-use efficiency (WUE), and final harvest index (HI):
 | (1) |
 |
It is well-established that crop biomass accumulation is a linear function of cumulative crop transpiration (de Wit, 1958; Tanner and Sinclair, 1983; Musick et al., 1994), and that the linear coefficient relating SY to T is WUE. Some have argued that HI is independent of T (Spaeth et al., 1984), but HI does decline when water stress is extremely severe (Specht et al., 1986).
Crop WUE differs substantially over locations and years (de Wit, 1958), primarily because evaporative demand is driven by the leaf-to-air vapor pressure deficit (VPD). Thus, the actual value of WUE depends on site-specific meteorological conditions (Tanner and Sinclair, 1983):
 | (2) |
 |
Agronomic practices greatly influence T, but have little impact on WUE. For example, when Jones (1992) plotted yields produced by various agronomic treatments (e.g., planting dates, soil fertility levels, etc.) at a given test site against the seasonal crop water use of each treatment, all of the treatment data points fell on the linear yield-to-water regression trend line, indicating a common WUE. However, researchers have long recognized that genetic improvement in WUE (i.e., k) might result in better drought tolerance. Differences in WUE among crop species, first documented in the 1920s (Shantz and Piemeisel, 1927), were later attributed to C4 photosynthetic species having a WUE about twice that of C3 species (Jones, 1992). Moreover, species with a mostly carbohydrate seed biomass have a larger k than those with a mostly protein and/or oil seed biomass, due to greater photosynthetic cost per unit of protein and/or oil (Specht et al., 1999). Genetic variation in soybean WUE has been reported (Specht et al., 1986; Mian et al., 1996a, 1998).
Farquhar et al. (1982) suggested that carbon isotope discrimination (CID) might serve as a mechanistic means of evaluating genotypes within a given C3 species for differences in transpiration efficiency (TE). TE is defined as WUE on a leaf basis, and is measured as mass of CO2 fixed per unit mass of water transpired. The carbon isotopes 13C and 12C occur in atmospheric CO2 in a molar ratio of about 1:89. However, 13C is less abundant in plant material because this heavier isotope is discriminated against (most notably in C3 species) during the processes of CO2 diffusion and photosynthetic carboxylation (Farquhar et al., 1989; Ehleringer, 1991). Theoretical considerations suggest a negative linear relationship between CID and TE (designated as 
and W in the literature). This negative correlation has been confirmed in empirical studies (Hubick et al., 1988). However, many of the reported correlations of CID with yield have been weak or strongly influenced by plant maturity (Ehleringer et al., 1993; Hall et al., 1994). Soybean data relevant to these CID correlations have not, to our knowledge, been published. Soybean breeders are unlikely to use CID in cultivar development without unequivocal evidence that selection for lower CID (i.e., higher TE) will improve seed yield under drought conditions.
Specht et al. (1986) demonstrated that soybean yield was a linear function of the amount of seasonal water needed to replenish 100, 80, 60, 40, 20, or 0% of crop evapotranspiratory (ET) water loss. They observed that the seed yield response of each genotype was predominantly linear. The linear regression coefficient beta was thus a season-specific estimate of a genotype's WUE. The authors reported significantly different betas among the 16 cultivars tested, but the number of genotypes was insufficient to discern the genetic basis of beta.
Molecular markers are a powerful means for discerning the genetic basis of traits whose phenotypic variation is ostensibly quantitative and polygenic (Tanksley, 1993). With a molecular genetic map, every marker-flanked segment of the genome can be assayed to determine if variation in a segment is significantly associated with phenotypic variation. If so, those genomic segments are termed quantitative trait loci (QTLs). Marker-assisted selection of desired genomic segments can then be used to complement, or replace, phenotypic measurement and selection.
A large population of soybean recombinant inbred lines (RILs) was recently developed as a resource for the identification of QTLs governing agronomic traits (Mansur and Orf, 1995; Mansur et al., 1993, 1996; Orf et al., 1999a, 1999b). This population was used to evaluate the genetic nature of soybean yield response to water. The occurrence of a severe drought in the second year of a 2-yr field trial provided a fortuitous opportunity to evaluate RIL yield response to an exceptionally broad gradient of seasonal water amount. Our primary research objective was to discern the genetic basis of soybean yield response to water.
|
MATERIALS AND METHODS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARYREFERENCES |
|---|
Field trials were conducted at the Agricultural Research and Development Center near Mead, NE, in 1994 and 1995. The soil at the test site is a Sharpsburg silty clay loam (fine, smectitic, mesic Typic Argiudolls) with a water-holding capacity of about 0.2 cm of water per cm of soil depth. A line-source sprinkler system, identical to that described by Specht et al. (1996), was used to create a water application gradient over which genotypic response to water could be measured. The amount of water applied weekly to the fully irrigated treatment was matched to the amount needed to replenish the cumulative ET water loss since the last irrigation, minus any rainfall during that period. Crop ET was estimated daily for an irrigated crop using the Penman equation and meteorological data collected by an automated weather station located within 200 m of the test site.
Each year, the experiment included six irrigation treatments, which were 100, 80, 60, 40, 20, and 0% replenishment of the crop ET loss at weekly intervals from 1 June through 31 August. Within each irrigation treatment, 290 genotypes were tested, which included 267 RILs, two sets of the parents, and 19 high-yield check cultivars whose maturities spanned those of the 267 RILs. The RIL maturity ranged from about 90 to 120 d after planting. To achieve a timely harvest, the 290 entries were grouped into 29 maturity blocks of 10 entries each, ensuring that the genotypes in any given block matured within a day or two of each other.
The experimental design in both years was a randomized complete block, with the sprinkler pipeline bisecting the test field into two replicates. The treatment design was a split-split plot, in which the six irrigation levels constituted six main plots in each replicate. The 29 maturity blocks were randomly assigned to subplots, and the 10 entries of each maturity block were randomly assigned to 3.05-m sub-subplots, which were one row in 1994, due to limited seed amounts, but two rows in 1995. Planting occurred on 24 May 1994 and 19 May 1995, using a 76-cm row spacing, a 3-cm seeding depth, and a seeding rate of 300 000 seeds ha-1 in 1994 and 370 000 seeds ha-1 in 1995.
Seed yield, 100-seed weight, days to maturity, plant height, and lodging scores were collected from all sub-subplots as described by Specht et al. (1986). Seed yield was based on a plot combine harvest of the 3.05-m long one-row (1994) or two-row (1995) sub-subplots. Recombinant inbred line seed protein and oil content were quantified (zero seed moisture basis) with a near-infrared transmittance technique, excepting RILs with self black or brown seed coats, for which the technique is not suitable because seed coat pigmentation interferes with transmittance.
On 20 July 1995, a juvenile trifoliolate was excised from the stem apices of five random plants of each entry in each replicate of the 100% and 0% ET treatments. These leaf samples were dried and ground. The leaf 13C/12C measurements were performed at the Stable Isotope Ratio Facility for Environmental Research at the University of Utah. For details, see Ehleringer (1991) and Buchmann et al. (1998). Leaf 13C values were expressed in (negative) units of parts per mil (
):
 | (3) |
) is usually computed from leaf 13C using an atmospheric 13C adjustment:
 | (4) |
However, our interest was comparative rather than absolute, so only leaf 13C values were presented in this article.
A primary analysis of variance (ANOVA), appropriate for the specified experimental and treatment design, was conducted on each measured trait listed in Table 1. Irrigation treatments were treated as fixed effects; replicates and genotypes as random effects. If a genotype x water (G x W) interaction term was significant at 
= 0.05, its variance was partitioned into a linear component (G x WL) and a residual. A significant G x WL was considered evidence of heterogeneity among genotypes in their linear responses to water.
|
|
2a) to phenotypic variance (2p). Confidence intervals for those estimates were derived as per Knapp et al. (1985). Genotypic (rg) and phenotypic (rp) correlations between traits in the 236-RIL group were computed as per Mode and Robinson (1959). The current genetic map of this RIL population is based on 665 markers and 236 genotyped RILs (Cregan, et al., 1999). Markers were ordered into linkage groups using the software package Mapmaker/exp 3.0 (Lander and Botstein, 1989; Lincoln and Lander, 1993). Quantitative trait loci were first detected by interval analysis using the software package Mapmaker/QTL 1.1 (Lincoln and Lander, 1993; Patterson et al., 1988), but QTL effects and positions were subsequently determined by composite interval analysis using the software package QTL Cartographer (Basten et al., 2000). Using permutation (Churchill and Doerge, 1994), the experiment-based criterion for QTL significance was determined to be LOD 3.4, based on 16 independent tests of n = 1000. Quantitative trait loci with a LOD score peak of <3.4 in either year of our study were presented if these were (i) significant in the other year of our study, or (ii) significant in prior research with this same population or with related ones (Mansur et al., 1993, 1996; Orf et al., 1999a, 1999b), thereby providing data for those researchers interested in across-paper meta analyses.
Three practices common to QTL analyses were followed here. First, the mean position of a QTL on the genetic map was assumed to be the genomic location of its LOD score maximum. Second, adjacent LOD maxima for a given trait were treated as two different but linked QTLs only if confirmed as such in the composite interval analyses. Third, if the QTLs for two genotypically correlated traits had a coincident or near-coincident map position, a single pleiotropic QTL was assumed. As Hanson (1959) noted, disproof of a null (pleiotropy) hypothesis requires a confirmable recombinant in which the linkage phase has been reversed.
The QTL parameters presented in this article were those generated by QTL Cartographer (Basten et al., 2000) and include (i) the genomic map position of each LOD score peak (i.e., the QTL), (ii) the additive (a) genetic effect on the trait if a Minsoy B allele were to be substituted into an RIL homozygous for the Noir 1 A allele, (iii) percentage of the trait variance (R2) explained by the QTL conditioned on background markers, and (iv) the percentage of trait variance (TR2) explained by both the QTL and the background markers.
|
RESULTS AND DISCUSSION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARYREFERENCES |
|---|
|
|
|
= 0.10) depression in seed protein and a significant enhancement in seed oil (Table 1). High air temperatures during late seed-fill are known to elevate seed oil content (Howell and Cartter, 1958). Constrained transpiration and evaporative cooling in the <100% ET treatments probably led to warmer canopies, given the warm air temperatures of late August (Fig. 1).
As the water available for transpiration decreased, leaf 13C values increased (Table 1). Similar trends have been reported in the literature (Farquhar et al., 1989; Ehleringer et al., 1993). Based on Eq. [4], a less negative leaf 13C translates into a smaller CID, which in theory implies that the drought-stressed, low-yielding plants in the non-irrigated treatment had a higher leaf TE than the non-stressed, high-yielding plants in the fully irrigated treatment. Thus, a genotype's leaf TE increases when less water is made available for transpiration. In contrast, the genotype's WUE, as estimated by its yield beta, is a seasonal constant—the amount by which its yield is reduced per unit of reduction in water made available for seasonal transpiration.
Genotype x Water (G x W) Interaction and Genotypic Betas
In 1994, the genotype x water interaction (G x W) term was significant for 100-seed weight, maturity, and height, but not yield (Table 1). Its linear component (G x WL) was significant for just maturity and 100-seed weight, suggesting heterogeneity in the genotypic response of maturity and seed size to water. This was confirmed by the significant differences among genotypes in their betas for these two traits in 1994 (Table 2). In 1995, the G x W term was significant for all traits but 100-seed weight and seed oil (Table 1). The G x WL component was significant for all 1995 traits, as were the 1995 genotypic betas (Table 2). Thus, even though genotypes displayed near-linear yield-to-water responses, their WUEs, as estimated by yield beta, were significantly different.
Genotypic Means and Heritabilities
Table 2 displays the grand means and betas for Minsoy and Noir 1, and for the RIL and cultivar groups. Noir 1, because of its indeterminate growth habit, was substantially taller and more productive than the determinate Minsoy in both years, even though the maturity difference between them was small. For most traits, the 19 high-yielding check cultivars were, on average, agronomically superior to the 236 RILs.
Analysis of the 236-RIL data set revealed significant genotypic effects in each year for most traits (Table 2). The exceptions were yield beta, height beta, and the 100%ET/0%ET yield ratio in 1994. The G x W interaction variance was considerably smaller than the genetic variance, so most traits had high heritabilities, including yield grand means (h2 > 0.90), yield in the 0% or 100% ET treatments (h2 > 0.80), and leaf 13C value (h2 > 0.80). The heritability of yield beta, unknown before this study, when measured in a drought severe enough to optimize its expression, was moderate (h2 = 0.56), but double that of the 100%ET/0%ET yield ratio (h2 = 0.24). These G x W adjusted heritabilities are obviously biased upward, given their estimation from single location/year data.
Genotypic Correlations
Correlations between the yield betas and yield grand means of the genotypes were highly significant (r = 0.37 in 1994; r = 0.71 in 1995), indicating that genetic improvement in the latter would likely be accompanied by coordinate improvement in the former (Fig. 4
, top). This finding is consistent with the theoretical arguments of Rosielle and Hamblin (1981). They contended that selection to increase mean yield would improve yield in both favorable and unfavorable environments. Conversely, they argued that selection aimed at minimizing the yield difference between favorable and unfavorable environments (in our case, a lower yield beta), while improving yield in the latter, would almost invariably depress yield in the former. They concluded that yield depression in the favorable environment was unavoidable unless (i) the correlation of genotypic performance in the two environments was near unity, and (ii) the genetic variance for yield in the unfavorable environment was significantly greater than that in the favorable environment. Concept (i) is intrinsically possible, but reports of concept (ii) are not extant in the soybean literature (Sneller and Dombek, 1997).
|
|
A low irrigated/rainfed yield ratio is often presumed to indicate genotypic drought tolerance (Abraham Blum, personal communication, 1997). Significant genotypic variation in the 100%ET/0%ET yield ratio was detected only in 1995 (Table 2). The correlation of yield ratio with its numerator was low in 1994 (r = 0.26) and in 1995 (r = -0.27) (Table 3), which is consistent with the argument that yield in well-watered conditions contributes little to the variation in yield ratio. However, to indicate greater drought tolerance, the yield ratio must be negatively correlated with its denominator, and it was in the severe stress year of 1995 (r = -0.67), but not in the mild stress year of 1994 (r = -0.15). However, soybean breeders are unlikely to adopt yield ratio as a selection criterion for drought tolerance for several reasons. These include (i) low heritability (Table 2), (ii) an extremely leptokurtic data distribution, attributable to well-watered yields exceeding water-stressed yields by a near-constant amount among both high- and low-yielding genotypes in a season of severe stress, for example, about 2.5 in 1995 (Fig. 4, bottom right), and (iii) the possibility of unintended consequences arising from selection based on a ratio of two traits. In a theoretical extrapolation of the simple concepts first enumerated by Rosielle and Hamblin (1981), Rowe (1995) demonstrated why selection to reduce the value of a ratio between two traits (in our study, a reduction of the 100% ET yield/0% ET yield ratio) would be expected to reduce the numerator trait (i.e., 100% ET yield) far more than it would enhance the denominator trait (i.e., 0% ET yield).
Highly positive correlations between yield, maturity, and height were observed in 1994 and 1995 (Table 3). Later-maturing RILs were generally higher yielding and taller than early-maturing RILs. This yield–maturity–height relationship is well-known to breeders.
Leaf 13C is theorized to be a mechanistic reflection of genotypic TE, whereas yield beta is an empirical measure of genotypic WUE. As is evident from Eq. [1], WUE is a seasonal constant with respect to variation in the water available for seasonal T. In contrast, TE is a leaf parameter that tends to increase with decreases in water available for T. Still, on a genotypic basis, one would expect a strong positive association of TE with WUE. The genotypic correlation of leaf 13C grand mean with yield beta was positive (r = 0.26) (Table 3). Given Eq. [4], a less negative leaf 13C value translates into a smaller CID, and in theory a larger TE, so the correlation was indeed indicative of a positive association of TE with WUE as estimated by yield beta. However, the magnitude of that correlation was lower than what has been reported in the literature (Hall et al., 1994), despite the wide range in leaf 13C values among what may be the largest number of genotypes ever tested for this trait at one time (Fig. 5
, bottom). The low correlation is unlikely to persuade breeders that leaf 13C measurement is an effective alternative to empirical selection for higher yields in drought conditions. Correlations of genotypic leaf 13C with genotypic yield in 0% ET treatment (r = 0.40), or in the well-watered 100% ET treatment (r = 0.30), were moderate (Fig. 5, top), but the correlations of leaf 13C grand mean with maturity (r = 0.30), height (r = 0.54), and plant lodging (r = 0.45) were just as large or larger (Table 3). As was the case with yield beta, genotypic variance in leaf 13C (TE) was confounded with the genetic factors governing plant height and maturity.
|
|
The second-strongest QTL with an impact on yield was also a maturity locus located near the center of LG U09-C2 (Fig. 6). This QTL, which was also identified by Orf et al. (1999b), accounted for phenotypic variances of 13 to 15% for maturity, 7 to 13% for yield, 5 to 15% for height, and 5% for lodging. In contrast to the U11-M Minsoy allele, which advanced maturity by about 3 to 4 d, the U09-C2 Minsoy allele delayed maturity by about 2 to 3 d. The repulsion-phase allelic mating with respect to these two maturity loci explained the similar parental maturities (Table 2) and transgressive variation in RIL maturity. The delayed maturity effect of the U09-C2 Minsoy allele boosted yield grand mean, yield beta (except in 1994), plant height, and lodging. This U09-C2 maturity QTL and maturity locus E1 were judged synonymous, given that E1 maps 4 cM below Satt277 (Cregan et al., 1999), which is where this QTL mapped (Fig. 6).
That allelic segregation at two maturity QTLs could have such a significant impact on RIL yield per se and RIL yield-to-water response deserves explanation. Specht et al. (1986) noted that Midwestern U.S. cultivars with a full-season maturity (for a given rainfed environment) usually yield better than earlier-maturing ones under drought, but generally yield less under irrigation. This performance differential arises because the reproductive development of the early-maturing genotypes occurs during the hotter part of the growing season and those high temperatures tend to exacerbate the impact of any water stress.
The third-strongest QTL with a major effect on yield was located near the bottom of LG U14-L (Fig. 6). Recombinant inbred lines in this population differed in stem growth habit and were qualitatively scored in 1994 as either determinant (dt1dt1) or indeterminant (Dt1Dt1). The U14-L QTL and the Dt1 locus were clearly coincident loci (Fig. 6) and accounted for 20 to 40% of the phenotypic variation in plant height. The Minsoy dt1 allele reduced plant height by about 9 to 10 cm each year, and 1995 lodging score by 0.25 units. Yield was substantially reduced by the dt1 allele in 1994, when plant densities were low due to limited amounts of RIL seed. Determinant genotypes in low plant densities display shortened internodes. When pods at the lower nodes are lower than the combine cutter bar, yield loss occurs. Internode shortening is avoided if determinant genotypes are grown in densities >325 000 plants ha-1, which was the case in 1995.
Plant height must be carefully managed in cornucopian environments to prevent lodging from impeding the realization of the high-yield potential in these environments. In that regard, breeders might be interested in the significant height QTL detected in the upper part of LG U14-L, since it allows plant height to be reduced without the imposition of a determinant stem habit. Recombinant inbred lines that were homozygous for the Minsoy allele at this QTL were about 6 cm shorter each year and had 0.4 better lodging scores in 1995 (lodging not scored in 1994). This QTL, which had been identified in prior studies (Lark et al., 1995; Mansur et al., 1996; Orf et al., 1999b), may be the S gene, which alters plant height by shortening the main stem internodes. However, the S locus has not yet been positioned on the soybean map (Cregan et al., 1999).
Three other QTLs for yield had LOD peaks exceeding the 3.4 significance criterion. A 1995 yield QTL was detected at the bottom of U06-F, in a two-marker region separated from the rest of the linkage group by a large gap (Fig. 6). The additive effect of the Noir 1 allele was 80 kg ha-1. A 1995 yield QTL was also detected near the Ps locus on U10-H. Recombinant inbred lines that were homozygous for the Noir 1 Ps-s allele (semi-sparse pubescence) yielded 168.2 kg ha-1 less than RILs homozygous for the Minsoy ps allele, probably because semi-sparse phenotypes tend to be spindly and lodging-prone. Finally, a 1994 yield QTL was detected on the U06-N, at the Rpg4 locus for resistance/susceptibility to bacteria blight (Pseudomonas glycinea Coerper). Recombinant inbred lines that were homozygous for the Noir 1 allele at this QTL yielded 181.2 kg ha-1 less in 1994 than the RILs with the Minsoy allele. Noir 1 is known to be susceptible to bacterial blight (Mansur et al., 1996), which was observed only late in the 1994 growing season. The RILs segregated for the Rpg1 locus in LG U13-F (Cregan et al., 1999), but Rpg1 had no apparent effect on yield in either year.
Leaf 13C is purported to be a prescriptive indicator of genotypic TE, so its genetic basis is of considerable interest (Martin et al., 1989). The strongest QTL detected for leaf 13C was on LG U14-L (Fig. 6), and it accounted for 25% of the phenotypic variation in leaf 13C. However, this QTL and the Dt1 locus were coincident. Leaf 13C was made more negative by the Minsoy dt1 allele, indicating a lower TE for the shorter, determinant RILs (Fig. 5). Hall et al. (1994) noted that 13C selection was difficult, since genotypic differences in 13C were often confounded with genotypic differences in stem determinancy and maturity. A LOD score peak for leaf 13C in the U09-C2 region of maturity locus E1 was not statistically significant, but a significant QTL for leaf 13C was detected near the top of LG U09-C2 and it accounted for 7% of the phenotypic variation in leaf 13C (Fig. 6). Leaf 13C was 0.408*% less negative in RILs homozygous for the Minsoy allele at this QTL, suggesting that these RILs had a greater TE. However, this leaf 13C QTL had no pleiotropic impact on genotypic WUE, given that a QTL for 1995 yield beta was not detected in the same genomic vicinity. A significant QTL for 100-seed weight (i.e., 0.3-g enhancement by the Minsoy allele), about 24 cM below the leaf 13C QTL, was detected in both years. A QTL for 1995 yield grand mean (i.e., 98.1 kg ha-1 enhancement by the Minsoy allele) was also detected about 35 cM below the leaf 13C QTL. Recombinant inbred lines in this study were not evaluated for flowering date, but Orf et al. (1999b) detected QTLs at the top of U09-C2 for seed weight and reproductive period length. LOD peaks for leaf 13C on U12-D2 and U13-F did not exceed the LOD 3.4 significance criterion. The LG U05-G QTL for 13C reported by Mansur et al. (1993) was not detected in the present study.
Of the significant QTLs detected for both yield beta and leaf 13C (Fig. 6), only those at Dt1 on U14-L congruently enhanced 1995 yield beta and made leaf 13C less negative (i.e., higher TE). The detection of only one such pleiotropic QTL was disappointing, but not surprising, given the low genotypic correlation between these two traits.
The confounding of leaf 13C with plant-stem termination has a plausible explanation. On the date RILs were sampled for leaf 13C, we collected juvenile leaflets just emerging from the stem apex. However, the apical meristems of nearly all determinate RILs had already shifted from a vegetative to reproductive state, immediately arresting the expansion of the most recently emerged leaflet. Thus, leaflet smallness on a determinant was not always an indicator of juvenility. Because a 13C value is a time-integrated estimate of leaf TE (Ehleringer et al., 1993), it follows that genotypic differences in leaf 13C are not comparable unless the sampled leaves are at the same point in their developmental ontogenies. Much of the genotypic variation in leaf 13C was thus more attributable to genotypic variance in developmental ontogeny than to genotypic differences in leaf TE per se. The solution to this problem—sampling leaves earlier in the season when all RILs were still vegetative—was not congruent with our objective of comparing leaf 13C measurements in the 0% and 100% ET extremes, since plants in the 0% ET treatment did not experience water stress until early July of 1995 (Fig. 1), when the shift to determinancy was well underway.
Using pot-grown plants, Mian et al. (1996a)(1998) detected six WUE QTLs (i.e., one QTL on LG G, two on LG H, three on LG J) in one population, and two additional QTLs (one on LG C1, one on LG L) in another (Fig. 6). Aside from the WUE QTL on LG L (near the Dt1 locus), none of the WUE QTLs reported by Mian et al. (1996a)( 1998) mapped to the vicinity of the QTLs for yield beta or leaf 13C detected in this study (Fig. 6). The WUEs reported by Mian et al. (1996a)(1998) were based on pot-grown plants, in which the 36-d accumulation of plant dry matter in a pot was divided by water added to the pot. However, these data were apparently not adjusted for transpiratory leaf area differences among the genotypes. It is thus likely that leaf area was the trait Mian et al. (1996a)(1998) actually mapped. Indeed, several of their WUE QTLs had map positions near the leaf-dimension QTLs mapped by others (Mansur and Orf, 1995; Mansur et al., 1993, 1996; Orf et al., 1999a, 1999b).
Every QTL we detected for seed protein and oil in our study mapped near protein and oil QTLs identified in prior studies evaluating this, or other, populations (Brummer et al., 1997; Diers et al., 1992; Lee et al., 1996; Mansur and Orf, 1995; Mansur et al., 1993, 1996; Orf et al., 1999a, 1999b). Increasing water stress tends to decrease protein and increase oil (Table 1), so beta QTLs for these traits would be useful in reducing unwanted changes in seed protein and oil induced by variation in seasonal water amounts. Unfortunately, we detected no significant beta QTLs for protein or oil that would be useful in that regard.
Quantitative trait loci were detected in this study for mean seed size, in one or both years (Fig. 6), but mapped to genomic positions where seed size QTLs have been reported by others (Mian et al., 1996b; Maughan et al., 1996; Mansur and Orf, 1995; Mansur et al., 1993, 1996; Orf et al., 1999a, 1999b). The translation of a summer rainfall deficit into a stored soil water deficit usually does not gain momentum until mid-summer, after flowering and early podding, so seed size is almost always the yield component reduced by drought (Table 1). By definition, a seed size beta indicates the steepness of the 100-seed weight response to water. Stabilization of seed size is of particular value in food-grade soybean production systems, where market acceptability may be limited to a rather narrow window of seed-size variation. A significant 100-seed weight beta was detected near the bottom of U26-B2, but its usefulness is suspect, given that the Noir 1 allele at this QTL reduced the sensitivity of seed size per unit change in seasonal water by just 0.1 gm per 10 cm of seasonal water.
|
SUMMARY |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARYREFERENCES |
|---|
13C estimate of its TE. Our data indicates that yield beta and leaf 13C can be added to the list of traits affected by the three major QTLs described in prior studies (Mansur and Orf, 1995; Mansur et al., 1993, 1996; Orf et al., 1999b). The map positions of these three QTLs are coincident to those of loci governing stem growth habit (Dt1 on LG U14-L) and maturity (E1 on U09-C2 and an E? locus on U11-M). Still, we believe our data warrant the following conclusions about the genomic nature of relative drought tolerance, which by definition is a lower yield beta (i.e., less sensitivity of yield to changes in seasonal water amount). First, the correlation of leaf 13C with yield beta may be too low (r = 0.26) to convince soybean breeders that selection for leaf 13C (i.e., greater leaf TE) will improve seasonal WUE, as measured by yield beta. The low correlation was consistent with the observation that the significant QTLs for leaf 13C and for yield beta had no in-common map positions (except for the QTLs that mapped to the Dt1 locus), and it is pleiotropism or linkage that leads to genotypic correlations. However, because of the wide range in maturity and determinancy in this population, we are currently evaluating these two traits in a different population of RILs that do not segregate for stem growth habit or maturity. Second, the exceptionally high correlation between genotypic yields in the 0% and 100% ET extremes conclusively demonstrates that genotypic yield rankings in a well-watered environment can be quite predictive of the genotypic yield rankings in a water-stressed environment (Fig. 4, bottom). Third, most of the G x W variance was attributable to its G x WL component, revealing a G x W interaction that was mainly attributable to genotypic heterogeneity in yield beta. Fourth, because yield beta is, by definition, a linear parameter, genotypes with a low yield beta, while less sensitive to water scarcity, are also less responsive to water abundance. Does this mean drought tolerance and yield responsiveness to water are mutually exclusive? Yes, if the selection is targeted solely at a lower yield beta (cf., Rosielle and Hamblin, 1981), but not necessarily if one selects genotypes in which a high yield grand mean is coupled with a high (instead of low) yield beta. This would be a feasible objective, given the positive covariance of these two traits at the phenotypic (Fig. 4, top right) and the genotypic (Fig. 6) levels. In fact, the selective packaging of these two traits into one genotype creates a form of drought tolerance that is generally under-appreciated, but critically needed in production areas where a year of abundant rainfall is no less predictable, or probable, than a year of drought. In absolute terms, the substantive yield advantage of this genotype in an abundant rainfall year will almost always offset any yield disadvantage it suffers in drought year. |
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARYREFERENCES |
|---|
Received for publication October 25, 1999.
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONSUMMARYREFERENCES |
|---|
