Published online 6 February 2007
Published in Crop Sci 47:285-293 (2007)
© 2007 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
CROP BREEDING & GENETICS
Response to Direct Selection for Grain Yield under Drought Stress in Rice
R. Venuprasada,b,
H. R. Lafittea and
G. N. Atlina,*
a International Rice Research Institute (IRRI), Los Baños, Philippines
b MAS Lab., Dept. of Genetics and Plant Breeding, Univ. of Agric. Sciences, Bangalore-560 065, India
* Corresponding author (G.Atlin{at}cgiar.org)
 |
ABSTRACT
|
|---|
Drought is a major cause of yield loss in rain-fed rice (Oryza sativa L.), grown on over 40 million ha in Asia. The objective of this study was to evaluate the effectiveness of direct selection for yield under drought stress in upland rice in populations derived from crosses between irrigated high-yielding cultivars and upland-adapted cultivars. Random F2:4 lines from five populations were screened for grain yield in fully irrigated lowland fields under nonstress conditions and in uplands under severe reproductive-stage drought stress. Stress caused mean yield reduction of 64% across populations. Broad-sense heritability for yield was not consistently lower in stress than in nonstress trials. Response to selection was evaluated in two crosses in subsequent seasons. Stress-selected lines had a yield advantage of 25 to 34% over random lines when evaluated at stress levels similar to those in which they were selected. Yield gains under very severe stress occurred only in a population derived from a highly tolerant parent. Direct selection usually gave greater response under stress than indirect selection under nonstress conditions. Direct selection under dry-season stress also gave response under naturally occurring wet-season stress. These results support the hypothesis that selection for yield under reproductive-stage drought stress is effective in rice, and that choice of donor is very important in breeding drought-tolerant rice.
Abbreviations: DAS, days after sowing • DS, dry season • IRRI, International Rice Research Institute • TPE, target population of environments • WAS, weeks after sowing • WS, wet season
|
INTRODUCTION
|
|---|
OVER 50% OF the world's rice (Oryza sativa L.) is rain-fed, but these nonirrigated lands produce only a quarter of total rice output (McLean et al., 2002). Drought is a major production constraint in rain-fed rice. In Asia, approximately 34 million ha of shallow rain-fed lowland rice and 8 million ha of upland rice, totaling approximately one-third of the total Asian rice area (Huke and Huke, 1997), are subject to occasional or frequent drought stress. Rain-fed lowland rice is cultivated in bunded fields that are flooded for part or all of the season, depending on rainfall, while upland rice is generally grown in aerobic soils. In the eastern Indian states of Jarkhand, Orissa, and Chhattisgarh alone, total rice production losses in severe droughts (about 1 yr in 5) average 40%, valued at $650 million (Pandey et al., 2005). The poorest rice farmers are most affected. In these areas, drought risk reduces productivity even in nondrought years, because farmers avoid purchasing inputs when they fear crop loss, becoming mired in a cycle of low productivity, poverty, and food insecurity (Pandey et al., 2000). Rice is particularly sensitive to drought stress during reproductive growth, when even moderate stress can result in drastic reduction in grain yield (Hsiao, 1982; O'Toole, 1982). The development of rice cultivars with improved drought tolerance is thus an important element in reducing risk, increasing productivity, and alleviating poverty in communities dependent on rain-fed production.
There is no consensus on the most effective approach to screening for reproductive-stage drought stress tolerance in rice. Screening under natural drought stress in the target population of environments (TPE) is difficult and often inefficient because of the sporadic and unpredictable occurrence of drought stress in the wet season. On the other hand, screening in managed-stress environments, including dry-season trials, rain-out shelters, and drained upper paddies, is more controllable, but the accuracy with which these screens would predict performance under natural drought stress is unknown. Selection response in the TPE under natural stress can be considered a correlated response to selection in the managed stress environment. Theory developed by Falconer (1989) and extended to the analysis of plant breeding programs by Pederson and Rathjen (1981) and Atlin and Frey (1989, 1990) permits breeding strategies to be evaluated on the basis of the predicted response in the target environment resulting from selection conducted in a selection environment. The efficiency of selection in a managed-stress screen is a function of the repeatability or broad-sense heritability with which the trait is measured in the selection environment and its genetic correlation with yield under stress in the TPE (Atlin and Frey, 1990). There are few estimates of these parameters in the literature published on breeding for drought-tolerant rice. Furthermore, grain yield under drought stress is a complex quantitative trait whose repeatability is thought to be low relative to yield in nonstress environments, reducing selection efficiency (Rosielle and Hamblin, 1981; Blum, 1988; Edmeades et al., 1989; Fukai and Cooper, 1995). Hence much effort has been focused on the genetic analysis of secondary traits such as root system architecture, leaf water potential, panicle water potential, osmotic adjustment and relative water content (Fukai et al., 1999; Price and Courtois, 1999; Jongdee et al., 2002; Pantuwan et al., 2002c; Toorchi et al., 2003). However, these traits rarely have a higher broad-sense heritability than yield under drought stress and are often not highly correlated with it (Atlin and Lafitte, 2002). So far, in no case have gains in yield by selecting for these traits been clearly demonstrated in rice.
In several recent studies of unselected populations of doubled-haploid lines, broad-sense heritability of grain yield under reproductive-stage drought stress was observed to be comparable to that of grain yield estimated in nonstress conditions (Blum et al., 1999; Lafitte and Courtois, 2000; Atlin and Lafitte, 2002; Babu et al., 2003; Lanceras et al., 2004). This indicates that direct selection for yield under drought stress is likely to be effective. Selection indices that give substantial weight to grain yield under drought stress have proven highly effective in improving drought tolerance in maize (Edmeades et al., 1999; Monneveux et al., 2006). However, there have been no conclusive reports on response to selection for yield under conditions of drought stress in rice. The objective of the present study was therefore to assess the effectiveness of direct selection for yield under severe imposed drought stress for improving rice grain yield under natural and artificially imposed drought stress. Direct selection for yield under drought stress in the dry season was compared with indirect selection for yield under nonstress conditions in segregating populations derived from crossing high-yield cultivars developed for irrigated lowland conditions with drought-tolerant lines adapted to upland conditions.
|
MATERIALS AND METHODS
|
|---|
The study was conducted on the experiment station of the International Rice Research Institute (IRRI), Los Baños, Laguna, Philippines, in the dry seasons (DS) of 2003 and 2004, and the wet season (WS) of 2004. IRRI is located at 14°13' N latitude, 121°15' E longitude at an elevation of about 21 m above mean sea level. The soil type is a Maahas clay loam, isohyperthermic mixed typic tropudalf.
Definition of Stress and Nonstress Environments
In this article, we refer to trials sown under upland (nonflooded, nonpuddled, direct-sown) conditions in the dry season, in which drought stress was imposed through reduced irrigation frequency, as stress environments. Rain-fed trials sown under upland management in the wet season in which stress occurred naturally due to periods of low rainfall are referred to as natural stress environments. Trials transplanted into fully irrigated lowland (puddled) fields with standing water are referred to as nonstress environments.
Selection in Stress and Nonstress Environments in 2003
Population Development and Screening in the 2003 Dry Season
Two semidwarf high-yielding indica lines, IR64 and IR72, developed in and for irrigated lowland conditions, wherein the crop is transplanted into a puddled, flooded field with standing water maintained throughout the season, were the lowland parents. The cultivars Azucena, Apo, and Vandana were the upland parents. Azucena is a deep-rooted tropical japonica variety (Price et al., 1997) with good vegetative drought tolerance (De Datta and Seshu, 1982). Apo (IR55423-01) is an improved indica upland variety with high yield potential (George et al., 2002) and moderate reproductive-stage drought tolerance (G. N. Atlin, unpublished data). Vandana is a drought-tolerant improved Indian upland variety developed from a tropical japonica/aus cross (Singh et al., 2000). Our previous unpublished studies indicate that the yield potential of Vandana and Azucena is low compared with Apo. Five populations of random F2:4 lines derived from the crosses IR64*2/Azucena, Apo/IR64, Vandana/IR64, Apo/IR72, Vandana/IR72 were used in the experiment. The IR64*2/Azucena population consisted of 400 BC1F2:4 lines developed by backcrossing 10 moderately drought-tolerant doubled-haploid lines derived from IR64/Azucena to IR64 (doubled-haploid population described by Guiderdoni et al., 1992). The numbers of F2:4 lines used in Apo/IR64, Apo/IR72, Vandana/IR64, and Vandana/IR72 populations were 481, 223, 283, and 263, respectively.
In DS 2003, the populations were evaluated under upland drought stress and lowland nonstress conditions. In the lowland trials, all the lines of IR64*2/Azucena population were evaluated, while in the other four populations, (Apo/IR64, Apo/IR72, Vandana/IR64 and Vandana/IR72), a maximum of 215 random lines was used. In upland trials, all the available lines in each population were included. Owing to the large populations used, and to exercise better control over within-experiment residual variation, each cross was divided into sets of approximately equal size (about 75 lines) for screening purposes.
All sets were laid out in an alpha-lattice design (Patterson and Williams, 1976). All the experiments had two replications with single row plots measuring 1.31 m2 in lowland and 0.5 m2 in upland sites.
Management of Stress Trials
In the stress trials, dry seed was direct sown at a rate of 8 g m–2 into level, unpuddled, unflooded upland fields. Rows were 2 m long and spaced 0.25 cm apart. Inorganic NPK fertilizer was applied at the rate of 100-40-40 kg ha–1. P and K were applied at sowing, and N was applied in three equal splits: at sowing, 30 d after sowing (DAS) and 60 DAS. Weeds were controlled by application of herbicide Nominee (bispyribac sodium [2,6-bis[(4,6-dimethoxy-2-pyrimidinyl)oxy]benzoic acid], 0.02 kg a.i. ha–1) 5 DAS and at later stages by hand weeding.
For the first 4 wk after sowing (WAS), trials were irrigated by overhead sprinklers once in 3 d for 4 h. Approximately 90 mm of water was applied at each irrigation. From 4 to 7 WAS, basin irrigation was provided once in 4 to 5 d to maintain the soil near field capacity. Basin size was 10 m x 12 m. Drought stress was imposed around flowering by reducing the frequency of irrigation to once every 10–12 d beginning at 7 WAS and continuing until harvest. Tensiometers were installed in the field, and soil water status was monitored. Irrigation was withheld until soil water tension reached about –70 kPa at 15 cm and –40 kPa at 30 cm soil depth. Trials were then basin-irrigated until soil was saturated in the root zone. The stress cycle was then repeated. This irrigation regime resulted in stress levels that caused leaf rolling and tip burning at the end of each drying cycle. The repeated cycles of stress ensured that all entries experienced stress during the sensitive stage of 15 d around flowering. The total amount of rainfall received during 2003 crop growth was 34 mm, while the estimated evapotranspiration was 780 mm.
Management of Nonstress Trials
In the nonstress trials, seeds were sown in the nursery, and 21-d-old seedlings were transplanted to the main field. Length of the rows was 5.25 m, with spacing of 0.25 m between rows and 10 cm between hills in a row. One seedling was transplanted per hill. After transplanting, approximately 5 cm of standing water was maintained in the field until drainage before harvest. Basal fertilizer was applied at the rate of 30-30-30 kg NPK ha–1, and trials were top-dressed twice with 30 kg N ha–1 at 2 and 6 wk after planting. Weeds were controlled by application of postemergence herbicide Nominee (bispyribac sodium, 0.02 kg a.i. ha–1).
In both stress and nonstress trials, the insecticide Regent (BASF Philippines, Inc., Canlubang, Laguna) (friponil [5-amino-[2,6-dichloro-4-[(1R,S)-(trifluoromethyl)sulfinyl]-1H-pyrazole-3-carbonitrile], 50 g a.i. ha–1) was applied at 24 DAS, followed by Cymbush (distributed by Syngenta Philippines, Inc., Makati City) (cypermethrin [cyano(3-phenoxyphenyl)methyl-3-(2,2-dichloroethenyl)-2,2-dimethylcyclopropanecarboxylate], 1 L ha–1) and Dimotrin (BASF Philippines, Inc.) (cartap hydrochloride [S,S'-[2-(dimethylamino)-1,3-propanediyl] dicarbamothioate], 0.25 kg a.i. ha–1) at 30 DAS to control stem-borer and sucking insects.
Data Collection and Statistical Analysis
In all the trials, grain yield, plant height and flowering date were recorded. Flowering date was recorded when the panicle was exserted in approximately 50% of the plants in a plot. Final plant height was measured as the distance from the ground to the panicle tip of three random plants from each plot. Grain yield from each plot was harvested, dried (50°C, 3 d), weighed, and adjusted to a moisture content of 14%. For measurement of harvest index, all the plants in one linear meter of row were sampled and dried to zero percent moisture. From this sample, harvest index was computed as the ratio of grain yield to total harvested biomass. Relative yield reduction (RYR) under stress was computed on a trial-mean basis as
 | [1] |
Statistical analysis was performed using SAS v8.2 (SAS Institute, Inc., 1999). For all analyses, all factors were considered random. For the estimation of line means for selection, data within sets were analyzed using the REML option of the SAS MIXED procedure. For broad-sense heritability (H) estimates, variance components were estimated using the REML option of the SAS VARCOMP procedure. Variance component estimates were averaged over sets within trials and used to calculate H for a selection unit consisting of the mean of a line over two replicates as
 | [2] |
where 
2G and 2e are the genotypic and residual variance components, and r is the number of replicates.
Phenotypic correlations across stress levels were computed based on line means. Genetic correlations (rG) across stress levels for grain yield for each set were calculated as
 | [3] |
where, rP is the phenotypic correlation between line means estimated in the stress and nonstress environments and HS and HNS are the broad-sense heritabilities estimated in the stress and nonstress environments, respectively (Cooper et al., 1996). Estimates were averaged across sets within trials.
Evaluation of Selection Response in 2004
In two populations, Apo/IR64 and Vandana/IR64, response to selection was evaluated under both nonstress and stress conditions during DS 2004 and under natural stress during WS 2004. From among the lines evaluated in both stress and nonstress trials during DS 2003 in each population, the 25 highest-yielding lines under stress conditions, the 25 highest-yielding lines under nonstress conditions and 25 random lines in each population were selected. The subsets from the Apo/IR64 and Vandana/IR64 populations were evaluated in separate experiments. Dry-season experiments were laid out in two replicates, as single-row plots, with plot area of 0.5 m2 in upland and 0.75 m2 in lowland, respectively. Wet-season experiments conducted in upland conditions consisted of two replications with two-row plots measuring 1.8 m2. Alpha-lattice designs were used for all experiments.
Management of Field Trials
Field management practices for stress trials in DS 2004, including water management, weed control, and fertilizer management, were similar to stress trials of DS 2003, except that NPK fertilizer was applied at the rate of 90-30-30 kg ha–1. During crop growth 120 mm of rainfall was received, while total evaporation was 680 mm.
Field management practices for nonstress trials in DS 2004 were similar to nonstress trials of DS 2003. Length of the rows was 3 m. Fertilizer was applied at the rate of 90-30-30 kg NPK ha–1. Weed control was done by the application of postemergence herbicide Sofit (Syngenta Philippines, Inc.,) (pretilachlor [2-chloro-2',6'-diethyl-N-(2-propoxyethyl) acetanilide] + safener, 0.3 kg a.i. ha–1) 4 d after transplanting.
In both stress and nonstress trials, the insecticide Furadan (FMC Corp., Philadelphia) (carbofuran [2,3-dihydro-2,2-dimethyl-7-benzofuranyl methylcarbamate], 1 kg a.i. ha–1) was applied at 5 d after transplanting.
In trials conducted under natural stress in WS 2004, dry seed was direct sown at a rate of 8 g m–2 into unpuddled, unflooded upland fields. Spacing between rows was 0.30 m and length of the rows was 3 m. Sprinkler irrigation was provided once in 3 d for 4 h duration for the first 7 wk. Subsequently, the crop was solely dependent on rainfall. The experiment received approximately 790 mm of rainfall, but severe drought stress occurred during the period 15–30 August, when only 37.5 mm of rain was received but cumulative evaporation was 69 mm. Average flowering date in this trial was 21 Aug., so most lines experienced drought stress during flowering. Rainfall in September was also much lower than average (70 mm compared to the long-term average of 244), which affected grain filling.
Inorganic NPK fertilizer was applied at the rate of 80-40-40 kg ha–1. N was applied in three equal splits: at sowing, 30 DAS, and 60 DAS. Weeds were controlled by preemergence application of the herbicide Ronstar (Bayer Crop Science, Inc., Canlubang, Laguna) (oxadiazon [3-(2,4-dichloro-5-(1-methylethoxy)phenyl)-1,3,4-oxadiazol-2(3H)-one], 0.5 kg a.i. ha–1), followed by interrow cultivation at 14 d after emergence and subsequently by hand weeding.
Data Collection and Analysis
Data on grain yield, plant height, and flowering were collected as in 2003. Data were analyzed using a mixed model in which selection protocols (stress, nonstress, and random) were considered fixed effects, while replicates, blocks within replicates and lines within selection protocols were all considered random. Least-squares estimates of mean grain yield of each selection protocol and variances for the random strata were obtained using the REML option of the SAS MIXED procedure. Using linear contrasts, the mean yields of the stress and nonstress sets were compared with the random unselected set. The error term for this contrast includes the variation among lines within selection protocols.
|
RESULTS AND DISCUSSION
|
|---|
Performance of Parents in 2003 Selection Trials
Mean grain yield of the parents, averaged over all the sets and populations, are presented in Table 1. On average, yield under upland drought stress was reduced by 64% relative to yield in the irrigated lowland treatment, but yield under stress varied greatly among the parents, as did the mean reduction due to stress. In the stress treatment, Apo produced the greatest yield. Vandana was somewhat lower yielding than Apo under stress, although in other studies it has outperformed Apo under severe stress at flowering (unpublished data). The lowland parents IR64 and IR72 had much lower yields under stress than Vandana and Apo. Azucena, an upland-adapted tropical O.japonica traditional variety with deep roots (Price et al., 1997), was highly susceptible to drought stress at flowering and yielded no more under stress than the lowland-adapted parents. Azucena appears to have vegetative-stage drought stress tolerance (De Datta and Seshu, 1982; Price et al., 1997), but our results show that it has very poor tolerance to reproductive-stage drought stress. In general, the percentage reduction in yield under stress was not closely related to absolute yield under stress conditions. For example, Apo and Azucena both experienced yield reductions of about 60%, but Apo's yield under stress was nearly three times that of Azucena. This result confirms the importance of choosing drought-tolerant parents on the basis of absolute yield under stress rather than on the basis of small yield reduction. Apo was also the highest-yielding line under lowland nonstress conditions, indicating that a high level of stress tolerance can be combined with high yield potential. Yield of Vandana under nonstress conditions could not be accurately measured because its early flowering and maturity resulted in severe rat and bird damage.
View this table:
[in this window]
[in a new window]
|
Table 1. Mean grain yield (g m–2) of parents in stress and nonstress environments: International Rice Research Institute, 2003 dry season.
|
|
Performance of Random Lines in 2003 Selection Trials
Mean grain yield of the populations and parents are presented in Table 2 (means of the parents are not equal to the overall means of checks presented in Table 1 because they include only the data from parents planted in the same sets as their progeny lines). Averaged over all populations, yields were 209 g m–2 in the nonstress treatment and 63 g m–2 under stress. Stress caused mean reductions in yield of over 80% in the IR64/Azucena and Apo/IR64 populations, and over 60% in Apo/IR72 and Vandana/IR72. The Vandana/IR64 population exhibited the least reduction in yield (24%).
View this table:
[in this window]
[in a new window]
|
Table 2. Mean yields (g m–2) of random F2:4 lines from five upland/lowland populations and yields expressed as a percentage of the yield of the parents in stress and nonstress environments: International Rice Research Institute, 2003 dry season.
|
|
Progeny lines equaling or exceeding the superior parent for grain yield were observed both in stress and nonstress screens (Fig. 1
and 2
) in 2003. Only in the Vandana/IR72 population was no line yielding less than Vandana under nonstress conditions observed. Under stress, the highest-yielding line in all populations yielded substantially more than the respective tolerant parent. In the Vandana/IR72 and IR64*2/Azucena populations, the best line out-yielded the tolerant parent by two- and threefold, respectively. Under nonstress conditions, the best line out-yielded the best parent by at least 50% in all populations. In the Apo/IR64 population, the best line yielded twice as much as IR64.
View larger version (27K):
[in this window]
[in a new window]
|
Fig. 1. Frequency distribution of grain yield (g m–2) of five rice populations in nonstress selection experiments: International Rice Research Institute, 2003 dry season.
|
|
View larger version (23K):
[in this window]
[in a new window]
|
Fig. 2. Frequency distribution of grain yield (g m–2) of five rice populations in stress selection experiments: International Rice Research Institute, 2003 dry season.
|
|
There was a strong influence of the upland parent on the mean yield of random lines from upland/lowland crosses evaluated under stress conditions. Under stress, populations of random F2:4 lines generated from the stress-tolerant upland parents Vandana and Apo were, on average, equivalent or superior in yield to the tolerant upland parent and out-yielded the susceptible lowland parent three- to sevenfold. Lines derived from the less tolerant upland parent Azucena did not, on average, out-yield the lowland parent under stress. These results strongly support the use of donors with superior yield under stress when the objective is to develop cultivars with improved yield under severe upland drought stress. However, under nonstress conditions, the mean yield of the populations derived from the most stress-tolerant donor, Vandana, which was also lower-yielding and shorter in duration than Apo, was about 40% less than that of the Apo-derived populations (Table 2). Thus, to combine high yield potential with improved tolerance, care needs to be taken in selecting donors that do not depress yield potential. Prebreeding may be needed to exploit drought-tolerance donors that transmit poor yield potential to progeny.
Broad-Sense Heritability of Yield under Stress and Nonstress Conditions
Over the five populations, H estimates for line mean yield from single-row plots in two-replicate trials under stress conditions ranged from 0.10 to 0.43; in the nonstress trials, the range was 0.25 to 0.67 (Table 3). Mean H across the five populations was slightly lower in stress than in nonstress trials (0.28 versus 0.42), but in two of the five populations (IR64*2/Azucena and Vandana/IR72), H was greater under stress than in the nonstress screen. There appears to be no consistent reduction in H under severe upland drought stress relative to lowland nonstress conditions. These results confirm other reports that the heritability of grain yield under reproductive-stage stress is comparable to that in nonstress trials (Blum et al., 1999; Lafitte and Courtois, 2000; Atlin and Lafitte, 2002; Babu et al., 2003; Atlin et al., 2004; Lanceras et al., 2004). These studies support the hypothesis that yield evaluation under reproductive drought stress in rice can be conducted with a precision roughly equivalent to that obtained for nonstress trials, and indicate that direct phenotypic selection for grain yield under stress will result in gains if screening trials are well managed (Atlin and Lafitte, 2002).
View this table:
[in this window]
[in a new window]
|
Table 3. Heritability (H) within and phenotypic (rP) and genotypic correlations (rG) across stress levels for grain yield of five populations in selection experiments: International Rice Research Institute, 2003 dry season.
|
|
Correlations across Stress Levels
Phenotypic and genotypic correlations for grain yield across stress levels are presented in Table 3. A moderately positive genetic correlation, averaging 0.48, was observed in all the crosses, although stress reduced yield by about 64% on average. This correlation indicates that partitioning and yield potential differences expressed in nonstress environments are responsible for about 23% of the yield variation under drought stress. The positive genetic correlation confirms that it is possible to select genotypes combining high nonstress yield with a high degree of stress tolerance, but the relatively low proportion of stress yield variation explained by nonstress performance indicates that screening under stress is important in identifying lines with improved drought tolerance.
The relationship between relative yield reduction (RYR) and rG for grain yield across stress levels was examined by regressing rG estimates from individual sets within populations on RYR (Fig. 3
). There was a negative association between RYR and the genetic correlation between yields in stress and nonstress environments. RYR explained about 22% of the variation among sets in genetic correlation across stress levels. Bänziger et al. (1997) observed that in maize, RYR under low nitrogen stress explained 27% of variation in the genetic correlation between stress and nonstress trials. The relationship between RYR and rG indicates that it is important that drought tolerance screening be conducted at stress levels that reduce yield substantially relative to a nonirrigated control, to ensure that stress trials assay stress tolerance rather than yield potential. In upland rice, these results indicate that the yield of stress trials must be reduced by 70–80% relative to nonstress trials to ensure that rG is reduced to a level where variation in drought tolerance, rather than in yield potential, explains more than 50% of the genetic variation in yield under stress.
View larger version (8K):
[in this window]
[in a new window]
|
Fig. 3. Relationship between genetic correlation across stress levels (rG) and relative grain yield reduction (RYR) in selection experiments: International Rice Research Institute, 2003 dry season.
|
|
Selection Response
Response in Stress and Nonstress Environments during DS 2004
Stress under upland conditions was severe in DS 2004 due to high evaporative demand during flowering, with the result that the susceptible parent, IR64, set nearly no seed (Table 4). Stress reduced yield by 93% and 70% in Apo/IR64 and Vandana/IR64 lines, respectively, relative to nonstress trials. The moderately tolerant upland-adapted parent, Apo, yielded an average of only 18 g m–2, compared to 111 g m–2 in DS 2003. The superior tolerance of Vandana was evident in the DS 2004 trials, where it yielded an average of 83 g m–2, as opposed to 86 g m–2 in 2003. The superior tolerance of Vandana was also expressed in its unselected progeny, which yielded an average of 56 g m–2, or about four times as much as unselected lines derived from Apo/IR64 (Table 4). The superior tolerance of Vandana as a donor was also reflected in the fact that there was a significant response to direct selection for yield under stress; stress-selected lines from this cross out-yielded both random lines and lines selected in nonstress environments by 25%. In Apo/IR64, on the other hand, variation for tolerance expressed in the less stressful 2003 screening experiment did not result in a positive response in DS 2004. These results indicate the importance of using highly tolerant donors if the objective is to make gains under severe drought stress; the tolerance level of the donor appeared to be more important than the environment in which the progeny lines were selected.
View this table:
[in this window]
[in a new window]
|
Table 4. Grain yield (g m–2) of stress-selected, non-stress-selected, and random lines and checks in Apo/IR64 and Vandana/IR64 populations in selection response trials: International Rice Research Institute, 2004 dry and wet season.
|
|
Although no effect of selection was observed in the Apo/IR64 population in the DS 2004 stress environment, transgressive segregation was again observed for yield under stress, as it was in DS 2003. In DS 2004, 23 of 75 lines tested equaled or exceeded the yield of the tolerant parent, Apo; the three highest-yielding lines outyielded it by at least two standard errors of the difference, with the highest-yielding line yielding over three times as much as Apo (data not shown). Similar levels of transgressive segregation were not observed in the Vandana/IR64 population, in which five lines equaled the mean yield under stress of the tolerant parent but did not significantly exceed it. These results confirm that lines with yields under stress equaling or exceeding the tolerant parent can be identified by direct selection for yield under managed stress.
In the nonstress environment, direct selection for yield under nonstress conditions was more effective than indirect selection under stress (Table 4). Non-stress-selected lines out-yielded random and stress-selected lines by 17% in Vandana/IR64, and by 9% in Apo/IR64 (although the advantage was not significant at p = .05 in the latter cross). In both the populations, the stress-selected lines yielded as much as random lines. Thus selection for yield under stress did not reduce yield potential on average.
Response under Natural Stress in WS 2004
In the WS 2004 evaluation experiment, natural stress was milder than the managed stress experienced in the DS 2004 evaluation environment. Relative to the DS 2004 nonstress trials, natural stress in WS 2004 reduced trial yield by 64% and 50% in Apo/IR64 and Vandana/IR64, respectively. Mean yield levels were similar to those in the 2003 selection environment. In the Apo/IR64 population, stress-selected lines had a yield advantage of 35% over either random or non-stress-selected lines. In Vandana/IR64, gains from selection were not statistically significant, although stress-selected lines out-yielded random- and non-stress-selected lines by 7% and 14%, respectively.
Although positive genetic correlations between yield in stress and nonstress environments were observed in all populations in DS 2003, selection based on yield under nonstress conditions was not effective in increasing yield in either the DS 2004 managed stress or WS 2004 natural stress evaluation environments. These results therefore confirm that selection for yield under upland drought stress in the dry-season results in greater gains under either managed dry-season drought stress or natural wet-season drought stress than does indirect selection in nonstress environments. However, in Vandana/IR64, the effect of the highly tolerant donor Vandana was more important than that of the selection environment. Although direct comparisons must be made with caution because the populations were screened in separate (albeit adjacent) trials, it is noteworthy that random Vandana/IR64 lines out-yielded random Apo/IR64 lines by 87%, 200%, and 30% in the DS 2003, DS 2004, and WS 2004 stress environments, respectively, with the magnitude of the advantage increasing with the severity of the stress.
Selection for higher grain yield under stress did not alter days to flowering or plant height (Table 5). Thus selection was for drought tolerance rather than drought escape. The yield advantage of stress-selected lines under stress seems to be due to better maintenance of harvest index under stress (Table 5).
View this table:
[in this window]
[in a new window]
|
Table 5. Days to flowering, height (cm), and harvest index of stress-selected, non-stress-selected, and random lines in Apo/IR64 and Vandana/IR64 populations in selection response trials: International Rice Research Institute, 2004 dry and wet season.
|
|
Line mean correlation of performance between screening environments (DS 2003) and evaluation environments (DS and WS 2004) is presented in Table 6. In the Vandana/IR64 population, performance in both the stress-evaluation environments was significantly and positively correlated with performance in the DS 2003 stress screening environment. In the Apo/IR64 population, the correlation was positive but nonsignificant. Yield under stress in the evaluation environments was not related with yield in nonstress trials in DS 2003. Correlation between yields across seasons in nonstress environments was positive and highly significant in both populations.
|
CONCLUSIONS
|
|---|
The results of this study confirm that upland rice yield under reproductive-stage drought stress is a moderately heritable trait, and that therefore direct selection for yield under stress is likely to be effective. These results were confirmed in empirical selection experiments, in which selection under severe managed drought stress in the dry season resulted in yield gains under both artificially imposed stress in the dry season and natural stress in the wet season, when stress levels (expressed in terms of trial mean yield) were similar in selection and evaluation environments. Stress and nonstress yields were positively correlated, indicating that it is possible to develop genotypes combining high yield potential with improved yield under drought stress, but the correlation was not high enough that selection under nonstress conditions resulted in significant gains in stress environments where mean yield was reduced by 60% or more. This supports the observation of Pantuwan et al. (2002a, 2002b) that potential "spillovers" from selection for yield potential under nonstress conditions are likely to be limited to environments where stress is relatively mild, and mean yields are greater than 50% of nonstress yields (about 1.5 t ha–1 in upland rice, which has a yield potential under favorable conditions of 4–5 t ha–1 (Atlin et al., 2006). The use of a highly tolerant donor appears to be critical to achieving gains in the most stressful environments, i.e., those in which the susceptible parent fails to set any seed. Use of such donors, and screening under stress that reduces yield of susceptible parents by at least 60%, appears to be an effective approach to improving drought tolerance of upland rice.
The selection experiments described in this report indicated that direct selection, whether for yield under stress or yield under nonstress conditions, was more effective than indirect selection in the contrasting environment, but that gains in one environment did not result in reduced yield in the other. These results indicate that an effective strategy for developing rice cultivars combining improved drought tolerance with acceptable yield potential under favorable conditions would be to screen breeding lines for yield under both stress and nonstress conditions, with selection based on an index combining information from both environments. This approach has been used successfully in maize to develop populations combining improved drought tolerance with high yield potential under favorable conditions (Bänziger and Cooper, 2001). It is likely to be successful in rice as well.
|
ACKNOWLEDGMENTS
|
|---|
We appreciate the research assistance provided by Mr. Modesto Amante, Mr. Roger Magbanua, and Mr. Rolly Torres, IRRI, in conducting the experiments.
|
NOTES
|
|---|
This research work was funded by BMZ, Germany, and The Rockefeller Foundation, USA.
Received for publication March 19, 2006.
|
REFERENCES
|
|---|
- Atlin, G.N., and K.J. Frey. 1989. Breeding crop varieties for low-input agriculture. Am. J. Alternative Agric. 4:53–57.
- Atlin, G.N., and K.J. Frey. 1990. Selecting oat lines for yield in low-productivity environments. Crop Sci. 30:556–561.[Abstract/Free Full Text]
- Atlin, G.N., and H.R. Lafitte. 2002. Marker-assisted breeding versus direct selection for drought tolerance in rice. p. 208. In N.P. Saxena and J.C. O'Toole (ed.) Field screening for drought tolerance in crop plants with emphasis on rice. Proc. Int. Workshop onField Screening for Drought Tolerance in Rice, Patancheru, India. 11–14 Dec 2000. ICRISAT, Patancheru, India, and The Rockefeller Foundation, New York.
- Atlin, G.N., H.R. Lafitte, D. Tao, M. Laza, M. Amante, and B. Courtois. 2006. Developing rice cultivars for high-fertility upland systems in the Asian tropics. Field Crops Res. 97:43–52.[CrossRef]
- Atlin, G.N., H.R. Lafitte, R. Venuprasad, R. Kumar, and B. Jongdee. 2004. Heritability of mean grain yield under reproductive-stage drought stress and correlations across stress levels in sets of selected and unselected rice lines in the Philippines, Thailand, and India: Implications for drought tolerance breeding. p. 85–87. In D. Poland et al. (ed.) Resilient crops for water limited environments. Proceedings of a workshop, Cuernavaca, Mexico. 24–28 May 2004. CIMMYT (International Maize and Wheat Improvement Center), Mexico D.F.
- Babu, R.C., B.D. Nguyen, V.P. Chamarerk, P. Shanmugasundaram, P. Chezhian, S.K. Jeyaprakash, A. Ganesh, S. Palchamy, S. Sadasivam, S. Sarkarung, L.J. Wade, and H.T. Nguyen. 2003. Genetic analysis of drought resistance in rice by molecular markers. Crop Sci. 43:1457–1469.[Abstract/Free Full Text]
- Bänziger, M., F.J. Betran, and H.R. Lafitte. 1997. Efficiency of high nitrogen selection environments for improving maize for low nitrogen target environments. Crop Sci. 37:1103–1109.[Abstract/Free Full Text]
- Bänziger, M., and M. Cooper. 2001. Breeding for low-input conditions and consequences for participatory plant breeding: Examples from tropical maize and wheat. Euphytica 122:503–519.[CrossRef]
- Blum, A. 1988. Plant breeding for stress environments. CRC Press, Boca Raton, FL.
- Blum, A., J. Mayer, G. Golan, and B. Sinmena. 1999. Drought tolerance of a doubled haploid line population of rice in the field. p. 310–330. In O. Ito et al. (ed.) Genetic improvement of rice for water-limited environments. IRRI, Los Baños, Philippines.
- Cooper, M., I.H. DeLacy, and K.E. Basford. 1996. Relationship among analytical methods used to analyse genotypic adaptation in multi-environment trials. In M. Cooper and G. L. Hammer (ed.) Plant adaptation and crop improvement. Univ. Press, Cambridge, UK.
- De Datta, S.K., and D.V. Seshu. 1982. Evaluating rices for drought tolerance using field screening and multilocation testing. p. 245–263. In Drought resistance in crops with emphasis on rice. IRRI, Los Baños, Philippines.
- Edmeades, G.O., J. Bolanos, S.C. Chapman, H.R. Lafitte, and M. Bänziger. 1999. Selection improves drought tolerance in tropical maize populations: I. Gains in biomass, grain yield and harvest index. Crop Sci. 39:1306–1315.[Abstract/Free Full Text]
- Edmeades, G.O., J. Bolanos, H.R. Lafitte, S. Rajaram, W. Pfeiffer, and R.A. Fischer. 1989. Traditional approaches to breeding for drought resistance in cereals. p. 353. In F. W. G. Baker (ed.) Drought resistance in cereals. CAB International, Wallingford, Oxon, UK.
- Falconer, D.S. 1989. Introduction to quantitative genetics. 3rd ed. Longman, London.
- Fukai, S., and M. Cooper. 1995. Development of drought-resistant cultivars using physiomorphological traits in rice. Field Crops Res. 40:67–86.[CrossRef]
- Fukai, S., G. Pantuwan, B. Jongdee, and M. Cooper. 1999. Screening for drought resistance in rainfed lowland rice. Field Crops Res. 64:61–74.[CrossRef]
- George, T., R. Magbanua, D.P. Garrity, B.S. Tubana, and J. Quinton. 2002. Rapid yield loss of rice cropped successively in aerobic soil. Agron. J. 94:981–989.[Abstract/Free Full Text]
- Guiderdoni, E., E. Gallinato, J. Luistro, and G. Vergara. 1992. Anther culture of tropical japonica/indica hybrids of rice (Oryza sativa L.). Euphytica 62:219–224.[CrossRef]
- Hsiao, T.C. 1982. The soil plant atmosphere continuum in relation to drought and crop production. p. 39–52. In Drought resistance in crops with emphasis on rice. IRRI, Los Baños, Philippines.
- Huke, R.E., and E.H. Huke. 1997. Rice area by type of culture: South, Southeast, and East Asia. IRRI, Los Baños, Philippines.
- Jongdee, B., S. Fukai, and M. Cooper. 2002. Leaf water potential and osmotic adjustment as physiological traits to improve drought tolerance in rice. Field Crop Res. 76:153–163.[CrossRef]
- Lafitte, H.R., and B. Courtois. 2000. Genetic variation in performance under reproductive-stage water deficit in a doubled haploid rice population in upland fields. p. 97–102. In J. M. Ribaut and D. Poland (ed.) Workshop on molecular approaches for the genetic improvement of cereals for stable production in water-limited environments, El Batan, Mexico. 21–25 June 1999. CIMMYT (International Maize and Wheat Improvement Center), Mexico, D.F.
- Lanceras, J., G. Pantuwan, B. Jongdee, and T. Toojinda. 2004. Quantitative trait loci associated with drought tolerance at reproductive stage in rice. Plant Physiol. 135:384–399.[Abstract/Free Full Text]
- McLean, J.L., D.C. Dawe, B. Hardy, and G.P. Hettel. 2002. Rice almanac. p. 253. IRRI, Los Baños, Philippines; WARDA, Bouaké, Côte d'Ivoire; CIAT, Cali, Colombia; and FAO, Rome.
- Monneveux, P., C. Sánchez, D. Beck, and G.O. Edmeades. 2006. Drought tolerance improvement in tropical maize source populations: Evidence of progress. Crop Sci. 46:180–191.[CrossRef]
- O'Toole, J.C. 1982. Adaptation of rice to drought prone environments. p. 195–213. In Drought resistance in crops with emphasis on rice. IRRI, Los Baños, Philippines.
- Pandey, S., D. Behura, R. Villano, and D. Naik. 2000. Economic costs of drought and farmers' coping mechanisms: A study of rainfed rice systems in eastern India. IRRI discussion paper series number 39. IRRI, Los Baños, Philippines.
- Pandey, S., H. Bhandari, R. Sharan, D. Naik, S.K. Taunk, and A.D.R.A.S. Sastri. 2005. Economic costs of drought and rainfed rice farmers' coping mechanisms in eastern India. Final project report. IRRI, Los Baños, Philippines.
- Pantuwan, G., S. Fukai, M. Cooper, S. Rajatasereekul, and J.C. O'Toole. 2002a. Yield response of rice (Oryza sativa L.) to drought under rainfed lowlands: 1. Grain yield and yield components. Field Crop Res. 73:153–168.[CrossRef]
- Pantuwan, G., S. Fukai, M. Cooper, S. Rajatasereekul, and J.C. O'Toole. 2002b. Yield response of rice (Oryza sativa L.) to drought under rainfed lowlands: 2. Selection for drought resistance genotypes. Field Crop Res. 73:169–180.[CrossRef]
- Pantuwan, G., S. Fukai, M. Cooper, S. Rajatasereekul, and J.C. O'Toole. 2002c. Yield response of rice (Oryza sativa L.) to drought under rainfed lowlands: 3. Plant factors contributing to drought resistance. Field Crop Res. 73:181–200.[CrossRef]
- Patterson, H.D., and E.R. Williams. 1976. A new class of resolvable block designs. Biometrika 63:83–92.[Abstract/Free Full Text]
- Pederson, D.G., and A.J. Rathjen. 1981. Choosing trial sites to maximize selection response for grain yield in spring wheat. Aust. J. Agric. Res. 32:411–424.[CrossRef]
- Price, A., and B. Courtois. 1999. Mapping QTLs associated with drought resistance in rice: Progress, problems, and prospects. Plant Growth Regul. 29:123–133.[CrossRef]
- Price, A.H., A.D. Tomos, and D.S. Virk. 1997. Genetic dissection of root growth in rice (Oryza sativa L.): I. A hydrophonic screen. Theor. Appl. Genet. 95:132–142.[CrossRef]
- Rosielle, A.A., and J. Hamblin. 1981. Theoretical aspects of selection for yield in stress and non-stress environments. Crop Sci. 21:943–946.[Abstract/Free Full Text]
- SAS Institute, Inc. 1999. SAS/stat user's guide, version 8.2. SAS Institute, Inc., Cary, NC.
- Singh, R.K., J.K. Roy, K. Prasad, S. Mallik, R.K. Sahu, N.K. Sarma, and J.L. Dwivedi. 2000. Identification of donor for rice breeding. p. 131–135. In Rainfed rice: A sourcebook of best practices and strategies in eastern India. IRRI, Los Baños, Philippines.
- Toorchi, M., H.E. Shashidhar, T.M. Gireesha, and S. Hittalmani. 2003. Performance of backcross involving trangressant doubled haploid lines in rice under contrasting moisture regimes: Yield components and marker heterozygosity. Crop Sci. 43:1448–1456.[Abstract/Free Full Text]
TOP
ABSTRACT
INTRODUCTION
