Plant materials
A mapping population of 250 BC2F3-derived lines developed from the cross Moroberekan/3* Swarna was used in this study. Moroberekan, the tolerant donor, is an upland- adapted tropical japonica [29] landrace from New Guinea. It is a long-duration cultivar with sturdy plant type, deep roots, and is tolerant to drought and rice blast. However, this variety has poor yield potential because of its low tillering ability and lower number of grains per panicle. On the other hand, Swarna (MTU 7029), the drought-susceptible recipient parent, is a lowland-adapted high-yielding indica variety [30, 31] derived from the cross Vashishtha X Mahsuri. It is a long-duration semi-dwarf variety with high tillering ability and grain yield. This variety is grown on a large area in rainfed and irrigated ecologies across India, Nepal, and Bangladesh and is regarded as a mega-variety of rice.
Experimental conditions and field management
Four field experiments were conducted in upland and lowland conditions at the experiment station of the International Rice Research Institute (IRRI), Los Baños, Laguna, Philippines (14°11′N, 121° 15′E) in the dry season (DS) and wet season (WS) of 2013 (Additional file 13). Throughout the study, the term ‘upland’ is used for field experiments conducted under direct-seeded, non-flooded, aerobic conditions while the term ‘lowland’ refers to field experiments conducted under flooded, puddled, transplanted and anaerobic conditions. Experiment 1 (1A, 1B, and 1C) was conducted under reproductive-stage drought stress conditions with early-, medium-, and late-maturing lines, respectively, due to heterogeneity in maturity duration in the population and with the aim of applying drought stress at the reproductive stage. Experiments 2 and 3 were conducted under non-stress conditions in lowland and upland, respectively, and Experiment 4 was conducted under upland seedling-stage drought stress. Experiments 2–4 were conducted with the full set of 250 lines and had no groupings based on maturity. All experiments were conducted in an α lattice design with three replicates each for Experiments 1 and 2 and two replicates each for Experiments 3 and 4 (Additional file 13).
In the lowland experiments, lines were grown in a wet bed nursery for 21 days before being transplanted in fields that were kept well-watered up to a month after transplanting. A spacing of 20 and 25 cm was maintained between plants and rows, respectively, with two to three seedlings transplanted per hill. In the upland experiments, lines were direct seeded in non-puddled soil at a density of 2.0–2.5 g m−1 and a depth of approximately 3 cm with a row spacing of 25 cm. Fields were sprinkler-irrigated to initiate seed germination and were surface-irrigated starting at one week after seedling emergence to maintain lowland-like conditions. All control treatments were irrigated 2–3 times per week throughout the crop duration. The stress experiments were also irrigated 2–3 times per week during crop establishment and early vegetative growth, and the drought stress treatment was initiated by withholding irrigation starting from 45 to 75 days after sowing (DAS), depending on the maturity group (Additional file 13). In Experiment 4, no irrigation was provided up to 21 DAS, after which full emergence was observed in all plots. The field was then re-irrigated and maintained well-watered until crop maturity. Complete fertilizer (14-14-14) was applied 13 days after transplanting in Experiments 1 and 2 (both the stress and control treatments) at a rate of 45 kg NPK ha−1, and a second application as topdressing was made before panicle initiation using ammonium sulfate at a rate of 45 kg N ha−1. Experiment 3 received 45 kg NPK ha−1 at 7 DAS, followed by topdressings of 45 kg N ha−1 on 34 and 38 DAS. No fertilizer was applied to Experiment 4. Manual weeding was done regularly in all experiments.
Phenotypic data collection
Traits related to drought tolerance (27 traits), yield potential (9 traits), lodging resistance (8 traits), and adaptation to direct seeding (22 traits) were recorded (Additional file 14).
Trait group 1: drought tolerance
Traits related to the ability to maintain shoot and root growth, water uptake, flowering, and grain yield under drought were classified in Trait group 1: Drought tolerance. These traits were recorded under lowland reproductive stage drought-stress conditions (Experiments 1A–C).
Shoot growth and groundcover dynamics were monitored according to NDVI measured around mid-day using a Greenseeker Hand-held Sensor (NTech Industries, CA, USA). Canopy temperature was measured at three locations per plot using a hand-held data-logging infrared (IR) sensor (Apogee Instruments, Logan UT, USA) after stress initiation. The increase in canopy temperature throughout the stress period was calculated as the slope (X) of the following equation:
$$ CT=\left(ICT\ X\ d\right)+b $$
where, CT is the canopy temperature, ICT is the increase in canopy temperature, d is the days after stress initiation, and b is the y intercept.
Root samples were taken 2–3 days after re-watering following severe stress symptoms at the grain-filling stage using a 4 cm-diameter core sampler (fabricated at IRRI, Los Baños, Philippines) to a depth of 60 cm. Soil cores were divided into 15-cm segments, and roots were washed by repeatedly mixing the soil with water in a container, and pouring the root-water suspension over a 1-mm plastic sieve. Only roots identified as living rice roots were retained for measurement. All samples were dried and weighed. Root mass density was calculated as the root mass per 15-cm soil core segment divided by the volume of the soil core segment. Percentage of deep roots was calculated as the mass of roots in the core below 30 cm divided by the total mass of roots inside the core. Bleeding rate measurements were carried out as described by Morita and Abe [32] and measured at mid-stress when soil water tension fell below −30 kPa. Sap exuded from the root zone was quantified in one hill per plot in both control and drought treatments. Starting at 7:00 am, shoots were cut at ~15 cm from the soil surface, and cut stems connected to the undisturbed root system were wrapped in a 625 cm2 cotton towel, then covered with a polyethylene bag, sealed at the base with a rubber band, and left for 4 h to absorb xylem sap that flowed from the cut stems. The towel, bag, and rubber band used for each hill were weighed before use. After 4 h, the bags and towels were removed from the stems, sealed, and immediately weighed to quantify the bleeding rate from the intact root system. Leaves from each hill were collected and kept inside a cold box for leaf area measurement. Tiller number was counted and leaves were separated from the stem and were dried and weighed to determine the biomass and leaf:stem ratio at mid-stress. In order to account for variation in plant size within and among genotypes, all sap exudation values were normalized by the dry shoot biomass of the hill from which sap was collected to calculate the bleeding rate. Specific leaf area was calculated as the leaf area, measured using a roller-belt-type leaf area meter (Li-Cor, Model LI-3100C, Li-Cor, Lincoln, NE, USA) divided by the leaf dry weight. Days to flowering (DTF) was recorded when about 50 % of the plants in the plot had flowered. Plant height (PH) of three plants from each plot was measured at maturity from ground level to the tip of the tallest tiller and averaged to get the mean PH for analysis. At physiological maturity, three hills were sampled in each plot for the measurements of the yield components including number of tillers and panicle, panicle length (cm), spikelet fertility (%), 1000-grain weight (gm) and rachis-stem-leaf dry weight (gm). Grain yield was measured from a sampled area of 1.5 m2 and dried to 14 % moisture content. The weight of grains was then used to calculate the kg ha−1 yield for each plot for further analysis.
Trait group 2: yield potential
Grain yield (kg ha−1) and yield-related traits such as number of panicles and tillers at harvest, spikelet fertility percentage (by weight), panicle length (cm), biomass (kg ha−1), DTF and plant height (cm) were classified in Trait group 2: Yield potential. These traits were recorded under lowland well-watered conditions (Experiment 2). The harvested area for grain yield was 1.5 m2.
Trait group 3: lodging resistance
Traits related to lodging resistance were measured under lowland and upland non-stress conditions (Experiments 2 and 3, respectively). These included stem strength, stem thickness, and fresh and dry weight per plant. Stem strength was measured using the prostrate tester (Daiki Rika Kogyou Co., Tokyo). All data were recorded from three plants from each plot. At physiological maturity, plants were cut off at 40 cm height, with the prostrate tester set perpendicularly at the middle (20 cm), and the pushing resistance of the lower part of the plant was measured by pushing the plants to the point at which the stem broke and the scale displacement (mm) due to pushing resistance was recorded. The stem diameter was measured from the same three plants at a height of 40 cm using a screw gauge. The plants were then harvested from the base to measure fresh weight per plant and then oven dried for three days at 70 °C and weighed to estimate the dry weight per plant. Average values for all parameters were calculated and used for further analysis.
Trait group 4: adaptation to direct seeding
Visual observations of the time (DAS) to first emergence (when the first seedlings of a plot emerged) and full emergence (most of the seedlings in each plot had emerged) were recorded in the seed bed nursery of Experiment 1 and of the direct-seeded experiments (3 and 4). Shoots from five seedlings per plot were sampled at a two-week interval (18 DAS and 32 DAS) to determine the relative growth rate (RGR) in Experiment 2. Shoots were dried and weighed to determine the biomass for each sampling date and the RGR was calculated as:
$$ RGR=\frac{\left[ \ln (B2)- \ln (B1)\right]}{\left(D2-D1\right)} $$
Where, B2 is the shoot biomass on date 2, B1 is the shoot biomass on date 1, D2 is date 2 and D1 is date 1.
In addition, grain yield and traits related to yield potential and phenology under direct-seeded conditions such as DTF, panicle number at harvest, tiller number at harvest, spikelet fertility, 1000-grain weight, and panicle length were recorded in Experiment 3. The harvested area for grain yield in Experiment 3 and 4 was 1 and 0.25 m2, respectively.
Statistical analysis
Statistical analysis for the computation of means and standard error of difference (SED) were conducted using CROPSTAT version 7.2.3. A mixed model analysis of data from individual years was carried out using the model:
$$ {y}_{ijk}=\mu +{g}_i+{r}_j+{b}_k\left({r}_j\right)+{e}_{ijk} $$
where yijk is the measurement recorded in a plot, μ is the overall mean, g
i
is the effect of the ith genotype, r
j
is the effect of the jth replicate, b
k
(r
j
) is the effect of the kth block within the jth replicate, and e
ijk
is the error. Genotypic effects were considered fixed and the replicates and block effects were random.
Correlations between the traits were estimated using the ‘cor’ function in R 3.1.0 [33]. For better visualization, the distance matrix was calculated using the correlation values between the traits and were used to conduct a MDS analysis using STAR (version 2.0.1). To perform PCA, all traits were first standardized to a mean of zero and standard deviation of one, and missing values were filled with zero (population mean). PCA was calculated with the prcomp function in R [33].
Genotypic data
Fresh leaves for all lines were collected and freeze-dried. DNA was extracted from the freeze-dried leaf samples by a modified CTAB method in deep-well plates. The DNA was then quantified and purified and lines were genotyped using KASPar SNP assays. These SNPs were selected as subsets from the set of 1536 and 44 K SNP chips [34, 35], converted to SNP assays and made available through the integrated breeding platform (https://www.integratedbreeding.net/482/communities/genomics-crop-info/crop-information/gcp-kaspar-snpmarkers). A total of 2015 SNP markers were screened for polymorphism between the two parents. Out of these 2015 SNPs, 591 polymorphic SNP loci were identified. The genotypic data from a set of 193 polymorphic SNP markers was used to generate the genotypic profile of the population.
Genetic analysis
Composite interval mapping (CIM) was conducted using QTL Network 2.1 [36] based on a mapping methodology outlined by Yang et al. [37]. Putative regions within the QTLs were identified with this software based on a one-dimensional genome scan taking selected candidate intervals as cofactors. A mixed linear model framework was used to perform the mapping procedure with an F-statistic based on Henderson method III for hypothesis testing. A total of 1,000 permutation tests were used to minimize the genome-wise type I error and to calculate the critical F-value. Apart from the CIM analysis, MLSIM analysis using the procedure detailed in Anderson et al. [38] was also conducted to identify putative QTLs controlling multiple traits simultaneously. Briefly, the QTL allele frequency (Pqm) conditional on the flanking marker genotypes for each point in the genome of each line was calculated. These Pqm values were then used as predictors [39] for multivariate analysis of variance (MANOVA) across the genome. Statistical significance was determined with randomization tests using 1,000 permutations. The identification of multiple QTLs was conducted sequentially, each conditional on all previously identified QTLs until no further significant QTL was found. To further understand QTL effects, a one-dimensional ‘composite trait’ was also calculated for each QTL by identifying the linear trait combination best explained by QTL genotypes with discriminant function analysis. Circle plot showing the chromosome map and QTLs was developed using Circos [40].
Availability of supporting data
The data sets supporting the results of this article are included within the article and its additional files.