Genome-wide association reveals novel genomic loci controlling rice grain yield and its component traits under water-deficit stress during the reproductive stage

Niteen N Kadam; Paul C Struik; Maria C Rebolledo; Xinyou Yin; S V Krishna Jagadish

doi:10.1093/jxb/ery186

Genome-wide association reveals novel genomic loci controlling rice grain yield and its component traits under water-deficit stress during the reproductive stage

Kadam, Niteen N; Struik, Paul C; Rebolledo, Maria C; Yin, Xinyou; Jagadish, S V Krishna 2018-07-18 00:00:00 A diversity panel comprising of 296 indica rice genotypes was phenotyped under non-stress and water-deficit stress conditions during the reproductive stage in the 2013 and 2014 dry seasons (DSs) at IRRI, Philippines. We investigated the genotypic variability for grain yield, yield components, and related traits, and conducted genome-wide association stud- ies (GWAS) using high-density 45K single nucleotide polymorphisms. We detected 38 loci in 2013 and 64 loci in 2014 for non-stress conditions and 69 loci in 2013 and 55 loci in 2014 for water-deficit stress. Desynchronized flowering time con- founded grain yield and its components under water-deficit stress in the 2013 experiment. Statistically corrected grain yield and yield component values using days to flowering helped to detect 31 additional genetic loci for grain yield, its components, and the harvest index in 2013. There were few overlaps in the detected loci between years and treatments, and when compared with previous studies using the same panel, indicating the complexity of yield formation under stress. Nevertheless, our analyses provided important insights into the potential links between grain yield with seed set and assimilate partitioning. Our findings demonstrate the complex genetic architecture of yield formation and we pro- pose exploring the genetic basis of less complex component traits as an alternative route for further yield enhancement. Keywords: A priori candidate genes, multi-locus analysis, Oryza sativa, reproductive-stage water-deficit stress, single-locus analysis, synchronized phenology. Introduction Rice (Oryza sativa L.) is a staple food crop for more than half the productivity of rice (Kadam et al., 2014; Reynolds et al., the world’s population. Maintaining its high yield potential 2016), as rice is more sensitive to water deficit than other cere- with sustained productivity is imperative for future food se- als (Kadam et al., 2015). Nearly 20% of global rice production curity. However, global climate change, with frequent epi- is affected by water deficit (Bouman et al., 2005). Water deficit sodes of abiotic stress (water deficit and heat stress), reduces can occur at any time during the growing season, but stress © The Author(s) 2018. Published by Oxford University Press on behalf of the Society for Experimental Biology. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4018 | Kadam et al. adaptation. This panel was assembled at the International Rice Research occurring within the reproductive phase (i.e. from meiosis Institute (IRRI), Philippines for the Phenomics of Rice Adaptation to flowering) causes the greatest grain yield losses (Liu et al., and Yield potential (PRAY) project in the context of the Global Rice 2006). The physiological effects of water deficit within the re- Phenotyping Network (http://ricephenonetwork.irri.org). Recent stud- productive phase have been discussed in detail by Saini and ies have reported GWAS analyses using this population for grain quality Lalonde (1997), Saini et al. (1999), and Barnabás et al. (2008). traits (Qiu et al., 2015), salinity tolerance (Al-Tamimi et al., 2016), panicle architecture (Rebolledo et al., 2016), yield traits under varying planting Increasing tolerance to water deficit has been considered as a densities (Kikuchi et al., 2017), and root plasticity (Kadam et al., 2017). major breeding target, although knowledge on phenotypic traits linked with stress tolerance is limited. Recent evidence in rice has Strategy to cope with variation in flowering phenology demonstrated that progress can be made through direct selection of grain yield, as a criterion under reproductive-stage water def- The PRAY panel was screened in non-stress and reproductive-stage water-deficit conditions under field experiments conducted at the upland icit (Venuprasad et al., 2007; Kumar et al., 2014). Physiologically, farm of IRRI, Philippines (14°11′N, 121°15′E; elevation 21 m above sea grain yield is a very complex trait determined by different com- level) in the 2013 and 2014 DSs. Seeds were sown from December of ponent traits (Slafer, 2003). Hence, exploring ideotype breeding the preceding year to late January or early February of each year (Fig. 1). based on selection for component traits is proposed as a com- As expected, a strong genotypic variation in flowering phenology was plementary route for further yield improvement (Donald, 1968). observed that confounds the true water-deficit response (Fukai et al., 1999) and inevitably induces bias with interpretation of genetic mapping Revealing the genetic basis of grain yield and its compo- outcomes (Pinto et al., 2010; Kumar et al., 2014). We followed staggered nent traits is essential for providing breeders with the tools for sowing in seedbeds and transplanting in main plots to synchronize flower- efficient development of stress-resilient cultivars. The genetic ing and thus minimize phenological differences under stress imposition control of grain yield under reproductive-stage water deficit (Fig. 1). Briefly, in the 2013 DS experiment, we divided 296 genotypes has been investigated extensively using linkage analysis of bi- into six groups with a 10 d interval based on days to flowering data col- lected from a previous experiment conducted in the 2012 wet season parental crosses in rice. This approach has proven to be power- (WS), our only source of flowering dates for this population grown at ful in the detection of quantitative trait loci (QTLs) for grain IRRI. While the expected range of flowering was 29 March to 8 April yield and its components under stress (Lanceras et al., 2004; 2013 (Fig. 1A), we observed deviation in days to flowering in the 2013 Bernier et al., 2007; Vikram et al., 2011; Mishra et al., 2013; DS experiment, where the staggered sowing was based on the 2012 WS Dixit et al., 2014; Kumar et al., 2014). A few of these QTLs data. Therefore, in the 2014 DS experiment, we regrouped the 296 geno- types into eight groups with a 7 d interval using 2013 DS flowering data regulating grain yield, for instance qDTY , have been intro- 12.1 to improve synchrony within the whole population. The expected date of gressed into elite cultivars to improve stress tolerance (Mishra flowering was 28 March to 5 April 2014 for these genotypes (Fig. 1B). In et al., 2013), but most of them are only based on a small frac- each year, the sowing date chosen for the stress treatment was the same as tion of the rice genetic diversity. Identifying the allelic vari- for the non-stress treatment of the same genotype. ations exhibited in a large genetic diversity panel as a result of divergent selection pressure provides an obvious alterna- Crop management tive that can have a greater potential in grain yield improve- The soil type of the upland farm at IRRI is Maahas clay loam, isohyper- ment under water deficit. These natural allelic variations have thermic mixed Typic Tropudalf. The experiments were laid out in a group been identified in rice under non-stress conditions for grain block design with three replications for each genotype in both treatments yield and its component traits through genome-wide associ- (Supplementary Fig. S1 at JXB online). Seeds were first exposed to 50 °C for 3 d to break dormancy and then hand sown in a seedbed nursery. ation studies (GWAS) (Agrama et al., 2007; Borba et al., 2010; Twenty-one-day-old seedlings were transplanted (two seedlings per hill) Huang et al., 2010, 2012; Zhao et al., 2011; Begum et al., 2015; for each genotype in four rows per replication. In both years, row dis- Spindel et al., 2015; Rebolledo et al., 2016; Yano et al., 2016). tance was 0.2 m and row length was 2.4 m. The seeds of one genotype Yet, very few studies are available for reproductive-stage water- in 2013 and eight genotypes in 2014 germinated poorly and hence were deficit conditions (Ma et al., 2016; Pantalião et al., 2016; Swamy excluded. In addition, four genotypes completed flowering and maturity before stress imposition in 2013 and were excluded. This resulted in final et al., 2017). This is partly due to the difficulty in implementing sets of 291 genotypes in 2013 and 288 genotypes in 2014, with three water deficit to coincide with reproductive stage under field replications and two treatments totalling 1746 and 1728 plots in 2013 conditions for a large diversity panel, which usually consists −1 and 2014, respectively. A day before transplanting, 30 kg P ha (as single −1 −1 of genotypes having diverse phenology. Only the study of Ma superphosphate), 40 kg K ha (as KCl), and 5 kg Zn ha (as zinc sulfate et al. (2016) followed a staggered sowing to account for vari- heptahydrate) fertilizers were manually applied. Nitrogen fertilizer as urea −1 −1 was applied in three splits: 45 kg ha before transplanting, 30 kg ha at ation in flowering phenology under stress. −1 mid-tillering, and 45 kg ha at panicle initiation. The IRRI standard Our study aimed to (i) explore the natural variation in grain management practices were followed to control weeds, insects, and dis- yield and yield component traits under non-stress and repro- eases. In both years, all plots were maintained like irrigated lowlands with ductive-stage water-deficit conditions; (ii) link the variation of ~5 cm standing water until maturity except for the water-deficit plots these phenotypic traits with single nucleotide polymorphisms during the stress period (see below). (SNPs) through GWAS; and (iii) identify the most likely under- lying candidate genes in close proximity to the significant SNPs. Reproductive stage water-deficit stress imposition There was variation in synchronizing days to flowering among rice geno- types in 2013, resulting in deviation from our expected flowering win- Materials and methods dow (29 March to 8 April). In rice, the reproductive stage ranges between 19 and 25 d, starting at panicle initiation and ending with flowering Association mapping population (Moldenhauer et al., 2013). Therefore, before imposing stress, we manu- We used a rice panel consisting of a diverse set of 296 indica genotypes ally dissected the main tillers of the middle two plants of border rows consisting of improved and traditional genotypes with (sub)tropical from water-deficit plots for all the genotypes, primarily to check the Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4019 Sowing Transplanting (21 days after sowing) Flowering Maturity 18 Dec 2012 8 Jan 2013 G1: 101-110 DTF (n=9) 28 Dec 18 Jan G2: 91-100 DTF (n=31) 7Jan 2013 28 Jan G3: 81-90 DTF (n=80) 2013 Experiment 7 Feb 17 Jan G4: 71-80 DTF (n=119) 18 Feb 27 Jan G5: 61-70 DTF (n=42) 6Feb 27 Feb G6: 51-60 DTF (n=15) Transplanting (21 days after sowing) Flowering Sowing Maturity 10 Dec 2013 31 Dec 2013 G1: 108-120 DTF (n=5) 17 Dec 07 Jan 2014 G2: 101-107 DTF (n=5) 24 Dec 14 Jan G3: 93-100 DTF (n=24) 31 Dec 21 Jan G4: 85-92 DTF (n=65) 2014 Experiment 28 Jan 07 Jan 2014 G5: 78-84 DTF (n=89) 14 Jan 04 Feb G6: 70-77 DTF (n=68) 21 Jan 11 Feb G7: 63-69 DTF (n=33) 28 Jan 18 Feb G8: 55-62 DTF (n=7) C D 2013 (n=291) 140 2014 (n=288) Non-stress Non-stress water-deficit stress Water-deficit stress y =0.8891x + 12.268 y =0.669x + 33.183 WD WD *** r²=0.85*** r²=0.46 y =0.9092x + 9.56 NS y =0.6879x + 26.976 NS *** r²=0.91*** 60 r²=0.53 50 50 TrtGYPNSPP SS TGW SP HI TrtGYPNSPP SS TGWSPHI 40 40 NS * ns ns ** ns ns ** NS * ns ns ** ** ns ns 30 30 WD ns ns ns ns ns ns ns WD *** * ** ****** ****** 20 20 20 30 40 50 60 70 80 90 100 110 120130 140 20 30 40 50 60 70 80 90 100 110 120 130 140 Expected DTF (2012) Expected DTF (2013) Fig. 1. Schematic representation of the staggered sowing and transplanting approach to synchronize flowering time that was followed for screening of an indica rice diversity panel under reproductive-stage water-deficit stress in the dry seasons (DSs) of 2013 (A) and 2014 (B). Days to flowering (DTF) was 10 d between groups (G) in 2013 and 7 days in 2014 DS experiments. (C, D) The expected and observed DTF in non-stress (NS) and water-deficit stress (WD) in the 2013 (C) and 2014 (D) DS experiments. ANOVA results with the effect of DTF (as a covariate in mixed linear model) on grain yield and 2 2 its key component traits are shown. GY, grain yield; HI, harvest index; n, number of genotypes; PN, panicles per m ; SP, spikelets per m ; SPP, spikelets per panicle; SS, seed set; TGW, thousand grain weight; Trt, treatments. Significance levels: *P<0.05, **P<0.01, ***P<0.001. To synchronize the flowering time, we used the 2012 wet season DTF data in the 2013 DS experiment (C). Similarly, for the 2014 DS experiment, we used DTF data from the 2013 DS experiment (D). (This figure is available in colour at JXB online.) reproductive-stage development. Stress was imposed on 23 March 2013 To quantify the stress intensity, 26 tensiometers were installed ran- when the majority of genotypes reached the agronomic panicle initiation domly across the entire stress field at 30 cm depth in each season. A poly- stage, by draining water out from the field. The stress continued for 14 thene sheet was inserted at 2 m depth by digging a deep and narrow d until 5 April 2013. In the 2014 experiment, the synchronization was trench in between stress and non-stress fields to prevent water seepage more precise with expected dates of flowering occurring between 28 during the stress period from the adjacent non-stress field. In addition, March and 5 April, as predicted. The same dissection approach as in 2013 the stress and the non-stress plots were separated by a distance of 2.3 was followed and stress was imposed on 26 March 2014 and continued m (Supplementary Fig. S1). The intensity of stress was higher in 2014 for 14 d until 8 April. than in 2013 (Supplementary Fig. S2A). There was no rainfall during Observed DTF (2013) Observed DTF (2014) 28 March 29 March 8April 5 April Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4020 | Kadam et al. the peak stress period in both seasons, except for rainfall during the first multiple linear regression analysis of grain yield with its component and day of the stress period in 2013 (Supplementary Fig. S2B). Higher stress related traits. intensity in 2014 compared with 2013 could be due to higher maximum temperature and higher vapour-pressure deficit (Supplementary Fig. S3B, Heritability estimates D), leading to quicker loss of soil moisture in 2014. A weather station was Broad-sense heritability (H ), capturing the proportion of phenotypic placed between the non-stress and water-deficit plots (see Supplementary variance explained by genetic factors that is due to dominance, epistatic, Fig. S1). Detailed weather data are given in Supplementary Fig. S3. and additive effects, was calculated across years and treatments separately using the below equation: Observations 2 2 G At maturity, plants of 16 hills from the middle two rows, i.e. 0.64 m H= plot area (excluding the border rows) were harvested to assess yield (14% σ + moisture), its components, and related traits in both experiments, fol- lowing Shi et al. (2016). Days to flowering was assessed as the interval 2 2 where σ and σ are the genotypic and residual variances, respect- G E between the date of sowing and the date when panicles of 50% of plants ively, and r is the number of replications. The restricted maximum likeli- per plot were fully exerted. Days to maturity was assessed as the interval hood estimate was used to calculate the variance components in Genstat between the flowering date and date when panicles on most plants in V17.1. The narrow-sense heritability (h ), capturing the proportion of a plot turned yellow and ready for harvest. Plant height was measured total phenotypic variance explained by the additive genetic variance, was from the base of the root–shoot junction to the tip of the flag leaf, which estimated using the equation in Genomic Association and Prediction was manually straightened to be aligned with the culm. Non-grain dry Integrated Tool (GAPIT) function: weight was assessed as the sum of leaf, stem and rachis dry weight. The total aboveground dry weight was the sum of non-grain and grain dry weight. Harvest index was the ratio of grain dry weight to total above- 2 a h= 2 2 ground dry weight. σσ + a e 2 2 where σ is the additive genetic variance and σ is the residual variance. a e Statistical analysis of phenotypic data Analysis of variance Genetic analysis of marker-trait associations A combined linear mixed model based analysis of variance (ANOVA) was performed to test the effect of genotype (G), treatment (T), and year Two hundred and ninety-one genotypes in 2013 and 288 genotypes in (Y) with their interactions using the following model in Genstat V17.1: 2014 had complete phenotypic data. However, 20 genotypes were miss- ing from the 45 699 (46K) SNPs dataset resulting in 271 genotypes in YG =+ μ ++ TY + RT [] () YG +× () TY × + E 2013 and 268 in 2014, used for GWAS analysis. The detailed genotype- ijkli jk lj k ijkl ijk by-sequencing protocol of SNP genotyping, population structure, and linkage disequilibrium (LD) for this population is explained in Kadam where Y is the phenotypic trait value recorded in a plot, µ is the over- ijkl et al. (2017). The GWAS was performed on a set of 271 (2013) and 268 all mean, G is the effect of the ith genotype, T is the effect of the jth i j (2014) genotypes separately, with 267 genotypes being common across treatment, Y is the effect of the kth year, R [T (Y )] is the effect of the lth k l j k both years. Two GWAS methods were used to test the marker–trait asso- replication within the jth treatment of the kth year, (G×T×Y) is the ijk ciations: single-locus and multi-locus analysis. effect of three-way interaction between the ith genotype, the jth treat- Single-locus analysis is a one-dimensional scan, typically identifying ment and the kth year, and E is the error. Apart from the three-way ijkl associations between single markers and traits. We performed this ana- interaction, we also consider two-way interactions of main factors in all lysis using a compressed mixed-linear model (CMLM; Zhang et al., possible combinations. 2010) in GAPIT (Lipka et al., 2012). In the mixed model, we included population structure and family kinship (family relatedness), which were Linear mixed model to estimate best linear unbiased estimators calculated by the GAPIT function using SNPs with ≥0.05 minor allele We estimated the best linear unbiased estimators (BLUEs) of phenotypic frequency (MAF). traits for an individual genotype across years and treatments separately. The following linear mixed model was used in Genstat V17.1 to estimate Y= XQ ++ Ke + αβ μ the BLUEs separately in non-stress and stress conditions across years, using genotypes as a fixed effect and replications as a random effect, where Y represents the vector of phenotype, X represents the vector of SNPs, Q is the PCA matrix and K is the relative kinship matrix. Y= μ+G +R +E ij ij ij X and Q are the fixed effects, and K and e represent random effects. α β μ The Q and K matrices help to reduce the spurious false positive asso- where Y is the phenotypic trait value recorded in a plot, µ is the overall ij ciations. Correction for population structure (Q) substantially reduces mean, G is the effect of the ith genotype, R is the effect of the jth repli- i j the false positives but it sometimes eliminates true positive associations cation, and E is the error. ij due to overcorrection. Therefore, the optimal number of principal Days to flowering had a strong confounding effect on grain yield and components was estimated for each trait before incorporating them its components under stress, particularly in 2013 (Fig. 1C). Therefore, for CMLM tests, based on the forward model selection method using we performed the linear mixed model-based ANOVA using the above the Bayesian information criterion. This method helps to control both equation with days to flowering as covariate. When the effect of days to false-positive and -negative associations more effectively although it flowering was significant on phenotypic traits, corrected BLUEs of trait cannot eliminate both completely. We used a lower suggestive threshold values were estimated in stress treatments. −4 probability P-value 1.0 × 10 (−log P=4) and superior Bonferroni corrected threshold as an upper limit (2013: −log (0.05/45 437)=6; Principal component analysis, trait correlation and multiple 2014: −log (0.05/45 414)=6) to detect significant associations. regression analysis The single-locus analysis corrects the confounding effects of popula- A multivariate principal component analysis (PCA) was performed in tion structure and family kinship but does not consider the confounding XLSTAT across years and treatments. The chart.Correlation() func- effect of causal loci. The multi-locus GWAS is a method that corrects not tion within the R package ‘Performance Analytics’ was used to gen- only the confounding effects of population structure and family kinship erate the correlation scatter plot. The lm() function in R was used for but also the confounding and/or interaction effects of causal loci present Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4021 in the genome due to LD (Segura et al., 2012). We performed the multi- In 2014, we restructured the synchronization based on 2013 locus GWAS using a modified version of the multi-locus mixed linear DS data. This resulted in better synchronization with only small model (MLMM) in R (R script for mlmm.cof.r available at https:// 2 deviation observed from expected days to flowering (r =0.91 cynin.gmi.oeaw.ac.at/home/resources/mlmm). We ran the complete in non-stress and r =0.85 in stress conditions; Fig. 1D). Further, model as recommended with stepwise forward inclusion of the strongest to test the effect of days to flowering, we performed the analy- significant markers (lower P-value) and stepwise backward elimination of the last forward model (that is, least significant markers). Significant mark- sis with days to flowering as a covariate in the mixed model. ers were selected based on the criteria explained by Kadam et al. (2017). The moderate to strong significant effect of days to flowering Briefly, in the first step (like single-locus GWAS without any marker as on yield, its components, and harvest index were detected in a cofactor), we manually checked the P-value of SNPs before including 2013 stress, most likely due to desynchronized flowering time. them as a cofactor in the model. Then we continued adding markers to −4 Conversely, the improved flowering synchronization caused no the model as cofactors based on cut-off threshold P-value≤1.00 × 10 . Once no significant loci appeared below the threshold P-value, the significant effect in 2014 stress. The marginal (P <0.05) to mod- model was stopped. All the significant cofactors identified were consid- erate (P<0.01) effect of days to flowering on yield, seed set, ered as significant loci. and harvest index was detected in both years under non-stress (Fig. 1C, D). This could be due to the pleiotropic effect of flow- Selecting a priori candidate genes underlying the genetic loci ering genes on panicle development (Crowell et al., 2016), a key The detailed protocol to select a priori candidate genes near to significant determinant of rice grain yield. SNPs was followed as explained in Kadam et al. (2017). Genotype effects and genotype-by-environment interactions accounted for variations in Results phenotypic traits The flowering time was sensitive to seasonal climate A combined mixed model ANOVA across years was carried variations out to divide the variation in genotype, treatment and year The flowering time synchronization approach was followed to components and their interactions (Table 1). The variation in reduce the confounding effect of flowering time differences of grain yield, its components, and other related traits differed rice genotypes on grain yield and its components (those meas- significantly between genotypes (G; P<0.001), treatments (T; ured in this study) and related traits under stress (Fig. 1A, B). P<0.001) and years (Y; P<0.01 to P<0.001). Further, the yield, However, we witnessed deviation of our observed days to flow- its component, and related traits of each genotype responded 2 2 ering from expected days (r =0.53 in non-stress and r =0.46 differently to treatment (G×T; P<0.001) and year (G×Y; in stress conditions; Fig. 1C) in 2013. As rice flowering time is P<0.001). The detailed descriptive statistics of these traits are regulated by internal genetic cues and external stimuli such as given in Supplementary Table S1. The traits showed different photoperiod and temperature (Yin et al., 1997), such deviations distributions in non-stress and stress conditions for both years −2 were expected, since the synchronization in 2013 was based on (Fig. 2). Yield ranged from 106.3 to 727.0 g m in non-stress, −2 2012 WS pre-experiment data due to lack of DS data. Many and from 16.7 to 622.6 g m under stress in 2013, and from −2 genotypes exhibited photothermal sensitivity across wet and 102.8 to 839.7 g m in non-stress, and from 78.1 to 761.1 g −2 dry seasons. Therefore, some genotypes experienced stress dur- m under stress conditions in 2014. Across all observations, 2 2 ing the flowering period (31%), whereas others experienced H and h estimates ranged from 0.73 to 0.99 and from 0.27 stress either before (60%) or immediately after flowering (8%). to 0.94, respectively, in 2013; and from 0.62 to 0.99 and from Table 1. Analysis of variance (ANOVA) in 2013 and 2014 dry season experiments of three groups of traits: grain yield, yield components, and other related traits Trait Unit G T Y G×T G×Y T×Y G×T×Y −2 Grain yield g m *** *** *** *** *** *** ** Grain yield component traits 2 −2 Panicles per m m *** *** *** *** *** ns *** Spikelets per panicle — *** *** *** *** *** ns Seed set % *** *** *** *** *** *** *** Thousand grain weight g *** *** ** *** *** ns *** 2 3 −2 Spikelets per m (×10 ) m *** *** *** *** *** *** Other related traits Harvest index — *** *** *** *** *** *** *** −2 Total dry weight kg m *** *** *** *** *** ns ns −2 Non-grain dry weight kg m *** *** *** *** *** *** *** Plant height cm *** *** *** *** *** * ns Days to flowering — *** *** *** *** *** *** *** Days to maturity — — — — — — — — 2 2 Spikelets per m is not an independent yield component but is the product of panicles per m and spikelets per panicle. G, genotype; T, treatment; Y, year. Significance level: *P<0.05, **P<0.01, ***P<0.001; ns, non-significant. Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4022 | Kadam et al. Fig. 2. Box-plot showing phenotypic distribution of grain yield and its components and related traits in non-stress (NS) and water-deficit stress (WD) during 2013 (n=271) and 2014 (n=268). Two-sample t-test P-value shows the significant difference between grain yield (A), its components (B–F), and related traits (G–J) in NS and WD conditions. n, number of genotypes. Inside boxplot, the bold line represents the median, box edges represent upper and lower quantiles, and whiskers are 1.5× the quantile of the data. Outliers are shown as open circles. Values in parentheses represent the significant percentage change (increase (+) or decrease (−)) in WD over NS conditions. Days to maturity across treatments in 2013 and data for non-grain tissue dry weight across treatments and years are given in Supplementary Table 1. The values of phenotypic traits given in the box-plot under 2013 water deficit are the original, not corrected for days to flowering to account variation in flowering synchronization. (This figure is available in colour at JXB online.) 0.69 to 0.93, respectively, in 2014 (Supplementary Table S1). years (Supplementary Figs S4, S5). However, non-significant The greater reduction of yield, seed set, and harvest index (P>0.05) correlations of yield were found with thousand grain under stress in 2014 was due to higher stress intensity during weight and non-grain dry weight in non-stress, and with pani- 2014 (−64 kPa) compared with 2013 (−46 kPa), driven by cle number in 2013 stress. Yield was not significantly (P>0.05) higher vapour-pressure deficit (Supplementary Figs S2A, S3D). correlated with non-grain dry weight across treatments in However, a higher reduction of spikelets per panicle and spike- 2014. The correlation of yield with spikelets per panicle was lets per m despite lower stress intensity was observed during higher in stress (2013: r=0.73; 2014: r=0.46) than in non-stress 2013 than during 2014 (Fig. 2C, E). This could be due to vari- conditions (2013: r=0.40; 2014: r=0.36) in both years, and ation in flowering time synchronization with more genotypes the increase was stronger in 2013. Similarly, the correlation experiencing stress before flowering in 2013 than in 2014. between yield and seed set increased from 0.62 in non-stress These results clearly illustrate that stress affects the number of to 0.75 in stress conditions in 2014. The increased correlation spikelets per m when imposed before flowering, but spike- of yield with spikelets per panicle in 2013 and with seed set let fertility when imposed during flowering (Lanceras et al., in 2014 in stress conditions reflects the effect of variation in 2004), as shown in Fig. 2C, E. The days to flowering differed days to flowering synchronization. The correlation of yield significantly (P=0.002) between non-stress and stress in 2013, with days to flowering was higher under stress (r=0.29) than but not (P>0.05) in 2014 (Fig. 2I). under non-stress conditions (r=0.16) in 2013, but was almost The first two principal components cumulatively explained the same (r=0.30) for both treatments in 2014. >55% in 2013 and >61% in 2014 of the total phenotypic vari- We also tested the relative contribution of each component ation across treatments (Fig. 3). The genotypic variation in and related trait to yield through multiple linear regression. All the first PC was mostly explained by yield, harvest index and the components and related traits significantly contributed to spikelets per m in non-stress and yield, harvest index, spikelets yield except for plant height and days to flowering in non- per m and total dry weight under stress in 2013 and 2014. The stress in 2013 and days to flowering in stress conditions during genotypic variation in the second PC was explained by non- 2014 (Supplementary Table S2). grain dry weight, days to flowering, and total dry weight under non-stress, and plant height, non-grain dry weight, and days to Treatment and year specific genetic loci for flowering under stress in 2013 and 2014. In addition, the prin- phenotypic traits cipal component variations for the phenotypic traits differed in Grain yield and its components and related traits followed a response to treatment and year (Fig. 3). This further confirms normal distribution (Supplementary Figs S4, S5), indicating the strong G×T and G×Y interactions. the quantitative pattern suitable for genetic analysis. A sum- mary of GWAS results using single-locus and multi-locus Phenotypic trait correlations and contribution of analysis methods is given in Table 2. The detailed results are component traits to grain yield in Supplementary Tables S3–S6. In total, we identified 38 Grain yield was significantly (P<0.05) correlated with most significant loci in non-stress conditions, and 69 loci in stress of its components and related traits across treatments and conditions during 2013, and 64 significant loci in non-stress Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4023 Biplot (PC1 and PC2: 55.61 %) Biplot (PC1 and PC2: 59.59 %) 2013:NS 2013:WD PH NGDW NGDW DTF DTF TDW 4 PH TDW SPP 2 TGW SP DTM TGW GY SPP SS DTM -2 -2 GY PN HI SS SP -4 -4 PN HI -6 -6 -8 -8 -10-8-6-4-2 02468 10 -10-8-6-4-20 24 68 10 PC1 (29.09 %) PC1 (35.51 %) Biplot (PC1 and PC2: 61.26 %) Biplot (PC1 and PC2: 63.66 %) 2014:NS 2014:WD NGDW DTF TDW NGDW DTF PH PH TDW SPP SP 2 SPP GY TGW P… SP 0 TGW SS GY SS -2 -2 HI PN HI -4 -4 -6 -6 -8 -8 -10-8-6-4-20 24 68 10 -10-8-6-4-20 2468 10 PC1 (34.69 %) PC1 (37.54 %) Fig. 3. The principal component analysis of grain yield, its components, and related traits with first two principal components (PC1 and PC2) in non- stress (NS) (A, C) and water-deficit stress (WD) (B, D) during 2013 (A, B) and 2014 (C, D) DS. The traits marked inside the solid circle/ellipses contributed more to the variation explained by PC1 and those marked inside the dashed ellipses to PC2. DTF, days to flowering; DTM, days to maturity; GY, grain 2 2 yield; HI, harvest index; NGDW, non-grain dry weight; PH, plant height; PN, panicles per m ; SP, spikelets per m ; SPP, spikelets per panicle; SS, seed set; TDW, total dry weight; TGW, thousand grain weight. (This figure is available in colour at JXB online.) conditions, and 55 loci in stress conditions during 2014. Most when corrected trait values were subjected to GWAS analy- loci were specific across treatments within years and within sis. This suggests that the trait variations associated with these treatments across the years. Nevertheless, we also detected 14 loci were mostly explained by variation in days to flowering. common loci (nine in 2013 and five in 2014) across treatments Only five genetic loci (one on chromosome 4 for yield (Q9); and eight common loci within treatments (six in non-stress one on chromosome 12 for spikelets per m (141 599) and and two in stress conditions) across years for the same compo- three loci on chromosome 11 for harvest index (10 627 944, nents and related traits (Supplementary Table S7). 10 131 062, 10 329 677) were common to corrected and non- corrected trait values (Supplementary Tables S4 and S8). The common (corrected vs non-corrected) loci detected for yield Genetic analysis after correcting for days to flowering (Q9; Table 3; Fig. 4) and harvest index (Supplementary Fig. S6; under stress conditions in 2013 Supplementary Tables S4, S8) recorded lower P-values for cor- Flowering time synchronization was strongly confounding the rected than the non-corrected trait value through single locus grain yield and its component traits in 2013 stress conditions analysis. Despite correction, the novel locus Q10 on chromo- (Fig. 1C, D). We corrected for yield, yield components, and some 3 for corrected yield, seed set, and harvest index over- other related traits (only harvest index in this group) using lapped with days to flowering (Table 3). In summary, statistical days to flowering as a covariate in the mixed model. The sin- correction helped to explain the confounding effect of days to gle and multi-locus analysis of corrected trait values identi- flowering and to some extent helped to eliminate its effect on fied 31 additional loci using similar threshold P-values as yield under water deficit. Unless otherwise mentioned, all the mentioned earlier (Table 2; Supplementary Table S8). Most mapping results discussed in the following sections were for genetic loci detected for non-corrected traits disappeared the corrected trait loci under 2013 stress. PC2 (26.56 %) PC2 (26.52 %) PC2 (26.12 %) PC2 (24.08 %) Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4024 | Kadam et al. Table 2. Summary of genetic loci detected in 2013 and 2014 Seven grain-yield loci revealed a small to medium under non-stress (NS) and water-deficit stress (WD) conditions allelic effect in response to reproductive-stage water for three groups of traits: grain yield, yield components, and other deficit related traits We identified two loci (Q9 and Q10) for grain yield under Traits 2013 2014 stress in 2013. The minor allele of both these loci had a negative effect on yield. Five significant loci Q11–Q15 were detected NS WD WD NS WD for yield under stress in 2014 (Fig. 5). The minor allele of Q11, Grain yield 2 4 2 6 5 Q12, and Q15 had a positive effect on yield, while the minor Grain yield component traits 2 allele of two loci, Q13 and Q14, had a negative effect on yield. Panicles per m 6 12 7 9 3 Q9 and Q10 harboured 18 and Q11–Q15 harboured 16 a Spikelets per panicle 5 9 6 2 na priori candidate genes within the expected LD block region Seed set 3 7 7 8 11 Thousand grain weight 3 4 na 6 8 (Table 3). Seven a priori candidate genes, mostly near signifi- Spikelets per m 1 4 1 4 3 cant SNPs, are given in Supplementary Table S9. The Q9 locus Subtotal 18 36 21 29 25 was close (13 kb) to the phosphomannomutase gene regulat- Other related traits ing L-ascorbic acid biosynthesis and response to abiotic stress Harvest index 6 7 8 4 2 stimulus (Gene Ontology (GO):0009628). L-Ascorbic acid Total dry weight 3 2 — 4 11 acts as a redox buffer to detoxify reactive oxygen species (ROS) Non-grain dry weight 3 3 — 5 2 (Arrigoni and De Tullio, 2002). Q11 was close to squalene Plant height 2 6 — 6 4 monooxygenase or epoxidase (16 and 23 kb; two copies in LD Days to flowering 3 11 — 10 6 block) and response to abiotic stress stimulus (GO:0009628). Days to maturity 1 na — — — This gene is known to regulate ROS, stomatal responses and Subtotal 18 29 8 29 25 Total 38 69 31 64 55 water-deficit tolerance in Arabidopsis (Posé et al., 2009). na, no marker trait association analysis performed Only three loci for grain yield acted via change in seed Marker-trait associations detected for corrected trait values in water- deficit stress (see the text for the correction method). set percentage Although rice grain yield is co-determined by panicle num- Eight grain-yield loci revealed small to medium allelic ber, spikelets per panicle, seed set percentage, and grain weight, effect in non-stress conditions very few loci of these component traits co-located with loci for We identified two (Q1 and Q2) and six (Q3–Q8) loci for grain yield per se. The seed set percentage is one of the most important yield in 2013 and 2014, respectively (Table 3). There were no yield components as indicated by its strong correlation with yield common loci across years, most likely due to significant vari- (Supplementary Figs S4, S5). Three loci were regulating yield ations in temperature (minimum and maximum) and vapour- through changes in seed set percentage, i.e. two loci designated as pressure deficit (VPD; Supplementary Fig. S3). These loci had Q2 (2013) and Q7 (2014) in non-stress, and Q10 (2013) in stress a positive or negative effect (small to medium) on yield with conditions. The major allele (allele refers to the 0.95 frequency regard to its minor allele (allele refers to the 0.05 frequency in in the studied population) of these loci had a respective positive the studied population). In 2013, the minor allele of Q1 had a effect on yield, seed set, and harvest index (Fig. 6). In addition, the positive effect on yield. Conversely, the minor allele of Q2 had Q10 was also detected for days to flowering. No loci were com- a negative effect on yield. In 2014, the minor allele of Q3, Q5, mon for yield and seed set in 2014 stress conditions, but one of and Q6 had a positive effect, while the minor allele of Q4, Q7, the loci on chromosome 1 (29 223 354) was commonly detected and Q8 had a negative effect on yield (Table 3). for seed set and harvest index (Supplementary Fig. S7). Similarly, Eighteen and sixty-eight a priori (known or characterized) the major alleles had a respective positive effect on seed set, har- candidate genes were harboured within the expected LD vest index, and yield (irrespective of genetic significance) (Fig. 7). block by Q1 and Q2 in 2013, and Q3–Q8 in 2014, respect- Hence, these loci were regulating yield through the effect of seed ively. Interestingly eight a priori candidate genes were identi- set on harvest index. fied (Supplementary Table S9). Q1 was close to OsPTR2 (6 Four a priori candidate genes were predicted within the and 31 kb; two copies in LD block). The rice homologue of expected LD block of these loci. The Q2 was close (55 kb this gene, short panicle 1 (OsPTR2), regulates panicle and grain from peak SNP) to the plastocyanin gene that regulates flower size and nitrate transport (Li et al., 2009). The homologue of development (GO:0009908) and pollination (GO:0009856) in OsPTR2 was recently detected at the q-28 locus (OsPTR9) rice (Supplementary Table S9). The Arabidopsis orthologue of for spikelet number per panicle (a key determinant of grain this gene regulates seed set and pollen tube growth (Dong et al., yield) in the same rice association panel as that used in this 2005). Q7 was within the novel expressed protein, which pro- study (Rebolledo et al., 2016). Likewise, Q4 was close (34 kb vides an entry point for future study. Sugar transport or uptake from peak SNP) to serine–threonine kinase (OsSTE). The is essential for normal pollen development (Reinders, 2016), Arabidopsis orthologue of OsSTE (AtSTE or BLUS1) is the while the lack of starch synthesis arrests the pollen develop- major regulator of stomatal opening (Takemiya et al., 2013; ment in water deficit conditions thereby regulating seed set Supplementary Table S9). (Sheoran and Saini, 1996). Our Q10 locus was within the sugar Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4025 Table 3. GWAS results for final set of genetic loci detected for grain yield in non-stress and water-deficit stress conditions during 2013 and 2014. Detailed GWAS results for yield components and related traits across treatments and years are given in Supplementary Tables S3–S6 and S8 a b c d e f g Treatment Year Locus SNP pos Chr Allele MAF P P AE LD block Size (kb) Known CMLM MLMM −2 h (mean grain name (g m ) genes Start End −2 yield (g m )) −5 Non-stress 2013 (451.1) Q1 10 101 900 11 C:G 0.336 2.72 × 10 — 30.13 10 101 900 10 173 685 71 2 −8 Q2 30 523 925 2 G:A 0.070 — 5.78 × 10 −175.90 30 397 910 30 541 202 143 16 −5 2014 (521.9) Q3 13 199 901 12 C:T 0.468 6.84 × 10 — 30.04 12 917 853 13 298 195 380 8 k −5 −6 Q4 26 796 595 3 C:T 0.097 9.91 × 10 2.20 × 10 −46.18 26 756 997 26 978 105 221 9 l −5 Q5 29 142 398 2 C:A 0.179 — 4.19 × 10 13.98 29 122 557 29 261 158 138 9 l −6 Q6 19 367 031 10 T:G 0.466 — 1.75 × 10 74.08 19 280 939 19 474 522 193 21 l −5 Q7 5 105 627 12 A:C 0.078 — 3.03 × 10 −186.24 5 101 105 539 0949 289 12 l −5 Q8 42 643 337 1 A:G 0.347 — 6.58 × 10 −97.80 42 587 683 42 643 699 56 9 i −5 −6 Water deficit 2013 (317.3) Q9 34 815 277 4 C:T 0.074 1.17 × 10 1.77 × 10 −81.29 34 815 277 34 833 179 17 5 −6 −6 — — — — — 1.29 × 10 3.05 × 10 j k −5 −6 Q10 5 113 428 3 T:C 0.424 3.55 × 10 5.17 × 10 −40.61 5 021 158 5 167 439.00 146 13 k −5 −6 2014 (319.5) Q11 6 934 188 3 A:G 0.397 8.26 × 10 8.73 × 10 31.47 6 908 684 7 020 707 112 7 l −7 Q12 42 144 827 1 T:C 0.366 — 1.86 × 10 6.10 42 123 552 42 144 993 21 2 l −6 Q13 16 038 003 10 T:C 0.358 — 5.46 × 10 −49.86 16 024 382 16 110 372 85 6 l −5 Q14 23 005 301 11 G:A 0.276 — 1.12 × 10 −23.54 22 976 390 23 005 386 28 0 l −5 Q15 27 115 652 11 G:A 0.075 — 4.92 × 10 33.65 27 115 609 27 123 090 7 1 Single nucleotide polymorphism (SNP) position. Chromosome. Minor allele frequency (MAF). P-value of single-locus compressed mixed linear model (CMLM). P-value of multi-locus mixed model (MLMM). Allelic effect with respect to minor allele=(average traits value of genotypes carrying minor allele−average traits value of genotypes carrying major allele). Linkage disequilibrium block. Total number of known characterized genes in LD block. Genetic locus detected for non-corrected and corrected grain yield value. Genetic locus detected for corrected grain yield value and coinciding with days to flowering. Genetic locus detected through CMLM and MLMM methods. Genetic locus detected through MLMM method only. All the unmarked loci were detected through CMLM method. The italic P-value is for corrected grain yield value. The genetic locus marked in bold (Q2) overlaps with panicle weight (equivalent to grain yield) from Kikuchi et al. (2017) (Supplementary Table S10). transporter gene that plays an important role in sugar distri- and grain yield in reproductive-stage stress (Pinto et al., 2010). bution. The rice grain yield MQTL (meta-analysis QTL) Our genetic analysis of statistically corrected trait values was 2.1 detected in water-deficit conditions also contained the sugar effective in minimizing the effect of desynchronized flower- transporter gene (Swamy et al., 2011). Similarly, the locus on ing time, as it led to detection of several novel loci that were chromosome 1 for seed set and harvest index in 2014 stress was not detected for non-corrected trait values. Despite statistical near (34 kb from peak SNP) to the nitrate transporter gene adjustment for flowering time, our novel Q10 for grain yield that plays a role in rice yield increment (Fan et al., 2016). was co-localized with flowering time (different SNPs but fall- ing within the same gene and LD block). In addition, it was also co-localized with seed set and harvest index. Previous Discussion studies in rice have identified several grain yield QTLs using linkage mapping under reproductive water-deficit stress con- The main aim of this study was to link phenotypic variation ditions (Bernier et al., 2007; Venuprasad et al., 2009; Vikram with genetic markers, thereby gaining insights about promis- et al., 2011; Swamy et al., 2013; Mishra et al., 2013), of which ing candidate genes and the genetic architecture controlling some co-localized with plant height (qDTY ), days to flow- yield traits. To the best of our knowledge, this is the first study 6.2 ering (qDTY ), or both (qDTY ). Interestingly, the major conducted on the rice PRAY association mapping panel under 3.2 1.1 effect of qDTY was consistent even after statistical correc- reproductive-stage water-deficit stress. The key findings from 1.1 tion of grain yield using flowering time and plant height as our study are discussed below. covariates (Vikram et al., 2011), and the recent detailed char- acterization confirmed the tight linkage and not the pleiot- Statistical trait adjustment can reduce confounding ropy of this QTL with plant phenology (Vikram et al., 2015). effect of desynchronized flowering on genetic analysis Our novel Q10 provided higher confidence of a causa- under water deficit tive SNP placed directly within the sugar transporter gene. The desynchronized flowering time may result in the identi- However, this SNP was just 5 kb away from the COP9 signa- fication of QTLs, often colocating with QTLs for phenology losome complex subunit 4 gene within the same LD block Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4026 | Kadam et al. Fig. 4. (A) GWAS results (Manhattan and quantile–quantile plot) detected through single-locus compressed mixed linear model (CMLM) and multi-locus mixed model (MLMM) for non-corrected and corrected (using days to flowering as covariate) grain yield in 2013 water-deficit stress (WD) conditions. Significant SNPs in the Manhattan plot of MLMM are numbered according to the order in which they were included as a cofactor in the regression model. 2 2 (B) Identified LD block (17 kb) based on r value between SNPs on chromosome 4 and the colour intensity of the box on the LD plot corresponds with r (multiplied by 100) according to legend. Significant SNP marked by first rectangle was detected by CMLM and MLMM and the next three rectangles only by CMLM approach. (This figure is available in colour at JXB online.) (Supplementary Table S9). The COP9 signalosome complex with flowering time and stress tolerance to test linkage versus gene is known to regulate flower development in Arabidopsis pleiotropy. Nevertheless, the effect of our consistent Q9 for (Wang et al., 2003), although the role of this gene in rice flow- grain yield (detected using either corrected or non-corrected ering has not been reported. Therefore, a further characteriza- values) was independent of flowering time stress conditions. tion of Q10 would be interesting to decipher the relationship More precise flowering time synchronization in 2014, which Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4027 Fig. 5. (A) GWAS results (Manhattan and quantile–quantile plot) detected through single-locus compressed mixed linear model (CMLM) and multi- locus mixed model (MLMM) for grain yield in 2014 water-deficit stress (WD) conditions. Significant SNPs on the Manhattan plot of MLMM are numbered according to the order in which they included as a cofactor in regression model. (B) Identified LD block (112 kb) based on r value between SNPs on chromosome 3 and the colour intensity of the box on the LD plot corresponds with r (multiplied by 100) according to the legend. Significant SNP marked by a rectangle was detected by CMLM and MLMM. (This figure is available in colour at JXB online.) allowed identification of the genetic loci without having any key yield components in cereals that are genetically less com- co-localization with flowering time in stress conditions, added plex than yield per se (Yin et al., 2002). In rice, grain yield is the value to the findings. To the best of our knowledge, this is the product of the panicle number or productive tiller (determined first report demonstrating the effectiveness of better synchro- during the vegetative phase), spikelets per panicle (determined nization of flowering time phenology on a large GWAS panel during panicle initiation), seed set percentage (determined dur- under stress conditions at field level. ing gametogenesis and anthesis), and individual grain weight (determined during grain filling). The genetic selection for each of these traits during rice domestication has given rise to rich Genetic control of grain yield, its components, genetic diversity (Doebley et al., 2006; Sweeney and McCouch, and related traits was mostly independent and 2007). To date, molecular genetic studies have detected QTLs environment-specific underlying these genetic changes in rice yield components Grain yield is a complex trait determined by many interactive (http://www.gramene.org/). From these QTLs some of the physiological processes changing temporally during the grow- candidate genes were successfully identified, notably display- ing period. These processes often match the development of the ing improvement in grain yield (Ashikari et al., 2005; Fan et al., Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4028 | Kadam et al. B C Q2 P=5.59E-10 80 0.45 P=4.15E-12 Q2 Q2 P=1.42E-13 75 0.40 0.35 0.30 0.25 0.20 150 0.15 GG (n=252)AA (n=19) GG (n=252)AA (n=19) GG (n=252)AA (n=19) D E F 600 0.50 80 P=1.03E-06 Q7 Q7 Q7 P=3.90E-09 P=4.04E-16 550 75 0.45 0.40 400 0.35 0.30 0.25 40 0.20 AA (n=247)CC (n=21) AA (n=247)CC (n=21) AA (n=247)CC (n=21) G H P=0.009 360 Q10 P=0.14 Q10 P=0.032 75 Q10 0.35 0.34 70 0.33 0.32 300 65 0.31 0.30 60 0.29 0.28 240 55 0.27 TT (n=156)CC (n=115) TT (n=156)CC (n=115) TT (n=156)CC (n=115) Fig. 6. Allelic effect of Q2 (A–C; 2013), Q7 (D–F; 2014) in non-stress and Q10 (G–I; 2013) in water-deficit stress conditions on grain yield, seed set, and harvest index. Allelic effect of Q7 on harvest index was significant regardless of GWAS significance. Two-sample t-test P-value shows significant allelic effect difference with reference to major and minor allele. The Q10 locus also coincided with days to flowering. A B C 400 60 0.35 SNP Pos: 29223354 SNP Pos: 29223354 SNP Pos: 29223354 0.33 P=3.50E-10 P=9.90E-11 P=1.03E-06 350 55 0.31 0.29 300 50 0.27 250 0.25 0.23 200 40 0.21 0.19 150 35 0.17 30 0.15 GG (n=224) CC (n=44) GG (n=224)CC (n=44) GG (n=224)CC (n=44) Fig. 7. Allelic effect of chromosome 1 locus (29 223 354) on grain yield (A), seed set (B), and harvest index (C) in 2014 water-deficit stress conditions. Allelic effect on grain yield was significant regardless of GWAS significance. Two-sample t-test P-value shows significant allelic effect difference with reference to major and minor allele. -2 -2 -2 Grain yield (g m ) Grain yield (g m ) Grain yield (g m ) -2 Grain yield (g m ) Seed set (%) Seed set (%) Seed set (%) Seed set (%) Harvest index Harvest index Harvest index Harvest index Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4029 2006; Song et al., 2007; Shomura et al., 2008; Huang et al., 2009; when comparing with other studies using different mapping Miura et al., 2010). For instance, the SPIKE gene/allele regu- panels under reproductive-stage water deficit (Ma et al., 2016; lating the spikelet numbers indicated 13–36% yield increment Pantalião et al., 2016; Swamy et al., 2017). The major reasons in rice (Fujita et al., 2013). In the present study, genetic dis- for this were different rice genotypes or population size and section of these yield components enabled us to detect more inherent environmental and field variation for stress treatment loci than yield per se that were directly or indirectly contribut- (QTL×environment interaction). Another possible reason ing to rice grain yield. The co-localization of grain yield loci could be use of indica subspecies genotypes in this study while with yield components was limited in this study compared with previous studies either used japonica subspecies (Pantalião other studies in rice (Lanceras et al., 2004). This could be due et al., 2016) or small population size (75 genotypes) with sim- to compensation among the yield components. In addition, ple sequence repeat markers (Swamy et al., 2017). In addition, these results emphasize the need for genetic analysis of yield it can be difficult to identify genomic regions or genes deter- components to identify additional genetic determinants having mining the trait difference across subspecies or genotypes. indirect effect on grain yield, providing alternative routes to enhance yield under water deficit. Seed set regulates the assimilate partitioning and Except for one locus on chromosome 12 for spikelets per m grain yield in 2014, the majority of the loci for grain yield and its com- Better optimization of assimilate partitioning to reproductive ponent traits were specific to non-stress or stress conditions in organs with minimal competition among reproductive organs both years. These results are in agreement with previous studies is essential to achieve stable and higher grain yield. So far, the in rice (Lanceras et al., 2004; Vikram et al., 2011; Kumar et al., physiological and genetic basis of the above processes have been 2014) and other crop species (Yin et al., 2002; Millet et al., 2016). poorly understood in rice and other cereal crops. Our study Hence, the greater dependence on environments appeared to showed that the co-localization of grain yield loci with its com- be a common characteristic of QTLs, although this does not ponents was rare. However, four genetic loci, namely Q2 and negate their importance in marker-assisted selection. Despite Q7 in non-stress, and Q10 and 29 223 354 (SNP position) in strong variation in weather, we also detected four consistent stress conditions, were regulating the grain yield and harvest loci: one each for panicles per m and spikelets per panicle on index through changes in the seed set (Figs 6, 7). This indicates chromosomes 10 (19 903 199) and 4 (23 423 399), respectively, that the seed set is a critical determinant of assimilate partition- and two loci on chromosomes 2 (30 699 332) and 5 (5 366 489) ing (harvest index), thereby regulating the final expression of for thousand-grain weight across years in non-stress condi- grain yield. A recent GWAS analysis confirmed these interac- tions (Supplementary Table S7). These consistent regions with tions in wheat (Guo et al., 2017). Hence, these identified loci favourable alleles could be used for improving yield potential. could be pyramided into an ‘ideotype’ at genomic level through marker-assisted selection to enhance rice grain yield in non- Few overlaps of genetic loci with previously identified stress and stress conditions. In addition, such loci could also be markers using same diversity panel of interest in identifying the physiological and molecular basis The PRAY population has been previously used in GWAS of assimilate partitioning to reproductive organs. for a range of phenotypic traits (Qiu et al., 2015; Al-Tamimi et al., 2016; Rebolledo et al., 2016; Kikuchi et al., 2017; Kadam Promising a priori candidate genes for grain yield and et al., 2017). When comparing our results with these previ- water-deficit stress resilience ous studies, we could not find any overlap between significant We detected a priori candidate genes near peak SNP(s) within markers, except a SNP marker detected for plant height (posi- the LD block for grain yield loci (Supplementar y Table S9). A tion: 38 286 772) on chromosome 1, which was detected in our priori candidate genes of grain yield loci can indicate possible previous study (Kadam et al., 2017). The most likely reasons for roles of underlying physiological (SET kinase, sugar and nitrate this lack of co-localization are difference in type and timing of transporter genes) and reproductive developmental (plasto- stress or growing environments (QTL×environment interac- cyanin gene) processes in regulating the grain yield. Likewise, tion), population size, and molecular marker data used by pre- the abiotic stress tolerance candidate genes were detected vious studies or novel GWAS analysis methods (multi-locus) near to grain yield loci in water-deficit conditions, of which that are used in this study. Therefore, to make a more logical genes regulating the detoxification of ROS (phosphoman- comparison for the same traits, we reanalysed the number of nomutase and squalene epoxidase genes) seem to be critical spikelets per panicle from Rebolledo et al. (2016) and yield in rice stress tolerance (Selote and Chopra, 2004; Pyngrope and yield components from Kikuchi et al. (2017), using the et al., 2013). These candidate genes need to be considered to same SNP datasets and analysis methods that are used in this detect the most likely causal genes. However, detailed large- study. This comparative analysis identified one locus on chro- scale molecular validations need to be conducted using the mosome 2 (30 518 548) for panicle weight (equivalent to grain available approaches of RNAi, knockout mutants transgenic yield) from Kikuchi et al. (2017) that overlapped with grain overexpression, and gene editing. Similarly, the loci for com- yield locus (Q2: 30 523 925; different SNP but falls within ponents and related traits that were not co-localized with the same LD block; Table 3; Supplementary Table S10) from yield per se could also be interesting candidates for further 2013 non-stress conditions. In addition, there was also no over- identification of novel genes. lap of a significant marker for grain yield and its components Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4030 | Kadam et al. water-deficit stress conditions using compressed mixed linear- Concluding remarks model and multi-locus mixed model methods. This study provides novel genetic loci for rice grain yield, its Table S5. The details of genetic loci detected for grain yield components, and related traits under non-stress and stress con- components and related traits in 2014 non-stress conditions ditions in field phenotyping experiments. We detected several using compressed mixed linear-model and multi-locus mixed favourable alleles regulating these traits that, upon validation, model methods. can be effectively used in improving yield. Additional genetic Table S6. The details of genetic loci detected for grain yield loci with less overlap of yield component traits to yield per se components and related traits in 2014 water-deficit stress con- clearly indicate the independent genetic regulation of these ditions using compressed mixed linear-model and multi-locus traits. Thus, many loci for component traits had an indirect mixed model methods. effect on yield, which cannot be detected while mapping yield Table S7. Common genetic loci detected across treatments directly. This indicates the complexity of yield as a trait despite (non-stress versus water-deficit stress) in 2013 or 2014 (A). moderate to high heritability, which is often used as a selec- Similarly, common genetic loci detected across years (2013 tion criterion to improve yield potential and stress tolerance. versus 2014) in NS or WD conditions (B). Hence, future studies should also explore the genetic basis of Table S8. The details of genetic loci detected for corrected individual component traits that are genetically less complex— grain yield components and related traits (only on harvest an approach expected to give additional useful information to index excluding the other traits in this group) in 2013 water- further enhance yield. deficit stress conditions using compressed mixed linear-model and multi-locus mixed model methods. Table S9. The list of a priori candidate genes within the link- Supplementary data age disequilibrium block of GWAS significant peak SNP/loci for grain yield in non-stress and water-deficit stress conditions. Supplementary data are available at JXB online. Table S10. The details of genetic loci detected from previ- Fig. S1. Field set-up of 296 genotypes screened under non- ously published data on grain yield and yield components from stress and reproductive-stage water-deficit stress in 2013 and Kikuchi et al. (2017), and number of spikelets per panicle (a 2014 experiments. key yield component) from Rebolledo et al. (2016), using the Fig. S2. Soil moisture tension measured using tensiometers same rice PRAY panel. in water-deficit stress field during 2013 and 2014, and rainfall pattern measured during stress period in 2013 and 2014. Fig. S3. Climate parameters observed during the growing Acknowledgements period. This work was supported by an anonymous private donor who pro- Fig. S4. Pearson correlation coefficient between grain yield vided the financial support, via Wageningen University Fund, to the and its components and related traits in 2013 non-stress and first author’s PhD fellowship. We also thank The Federal Ministry for water-deficit stress conditions. Economic Cooperation and Development, Germany, and the USAID- Fig. S5. Pearson correlation coefficient between grain yield Bill & Melinda Gates Foundation for their financial support. We also thank the GRiSP (Global Rice Science Partnerships; now renamed and its components and related traits in 2014 non-stress and to RICE CRP consortium) for establishing the PRAY Global water-deficit stress. Phenotyping Network. Dr C. G. van der Linden and Dr P. S. Bindraban Fig. S6. GWAS results (Manhattan and quantile–quantile are acknowledged for their valuable advice. plot) detected through single-locus compressed mixed linear model and multi-locus mixed model for non-corrected and corrected harvest index (using days to flowering as a covariate) Author contributions in 2013 water-deficit stress conditions. XY, PCS, and SVKJ conceived the project and its components; NNK Fig. S7. GWAS results (Manhattan and quantile–quantile and SVKJ implemented the experiment; NNK performed the pheno- plot) detected through single-locus compressed mixed linear typing; NKK performed the GWAS including both the conventional and multi-locus approach; NNK drafted the figures, tables, and manu- model and multi-locus mixed model for seed-set and harvest script; MCR provided data obtained from the same panel for compara- index in 2014 water-deficit stress conditions. tive GWAS analysis; XY, SVKJ, and PCS supervised the data processing Table S1. Summary statistics of grain yield and its compo- and the preparation of the drafts; NNK, MCR, SVKJ, XY, and PCS nents and related traits in 2013 and 2014 non-stress and water- interpreted the data and wrote the final paper. The authors declare no deficit stress conditions. competing financial interests. Table S2. Multiple linear regression of grain yield with its components and related traits in non-stress and water-deficit References stress conditions during 2013 and 2014. Table S3. The details of genetic loci detected for grain yield Agrama HA, Eizenga GC, Yan W. 2007. Association mapping of yield and its components in rice cultivars. Molecular Breeding 19, 341–356. components and related traits in 2013 non-stress conditions Al-Tamimi N, Brien C, Oakey H, Berger B, Saade S, Ho YS, Schmöckel using compressed mixed linear-model and multi-locus mixed SM, Tester M, Negrão S. 2016. Salinity tolerance loci revealed in rice using model methods. high-throughput non-invasive phenotyping. Nature Communications 7, 13342. Table S4. The details of genetic loci detected for uncor- Arrigoni O, De Tullio MC. 2002. Ascorbic acid: much more than just an rected grain yield, its components, and related traits in 2013 antioxidant. Biochimica et Biophysica Acta 1569, 1–9. Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4031 Ashikari M, Sakakibara H, Lin S, Yamamoto T, Takashi T, Nishimura Genome-wide association mapping for phenotypic plasticity in rice. Plant, A, Angeles ER, Qian Q, Kitano H, Matsuoka M. 2005. Cytokinin oxidase Cell & Environment 40, 1565–1575. regulates rice grain production. Science 309, 741–745. Kumar A, Dixit S, Ram T, Yadaw RB, Mishra KK, Mandal NP. 2014. Barnabás B, Jäger K, Fehér A. 2008. The effect of drought and heat Breeding high-yielding drought-tolerant rice: genetic variations and stress on reproductive processes in cereals. Plant, Cell & Environment 31, conventional and molecular approaches. Journal of Experimental Botany 11–38. 65, 6265–6278. Begum H, Spindel JE, Lalusin A, Borromeo T, Gregorio G, Hernandez Lanceras JC, Pantuwan G, Jongdee B, Toojinda T. 2004. Quantitative J, Virk P, Collard B, McCouch SR. 2015. Genome-wide association trait loci associated with drought tolerance at reproductive stage in rice. mapping for yield and other agronomic traits in an elite breeding population Plant Physiology 135, 384–399. of tropical rice (Oryza sativa). PLoS ONE 10, e0119873. Li S, Qian Q, Fu Z, et al. 2009. Short panicle1 encodes a putative PTR Bernier J, Kumar A, Ramaiah V, Spaner D, Atlin G. 2007. A large-effect family transporter and determines rice panicle size. The Plant Journal 58, QTL for grain yield under reproductive-stage drought stress in upland rice. 592–605. Crop Science 47, 507–516. Lipka AE, Tian F, Wang Q, Peiffer J, Li M, Bradbury PJ, Gore MA, Borba TCdO, Brondani RPV, Breseghello F, Coelho ASG, Mendonça Buckler ES, Zhang Z. 2012. GAPIT: genome association and prediction JA, Rangel PHN, Brondani C. 2010. Association mapping for yield and integrated tool. Bioinformatics 28, 2397–2399. grain quality traits in rice (Oryza sativa L.). Genetics and Molecular Biology Liu JX, Liao DQ, Oane R, Estenor L, Yang XE, Li ZC, Bennett J. 2006. 33, 515–524. Genetic variation in the sensitivity of anther dehiscence to drought stress in Bouman BAM, Peng S, Castañeda AR, Visperas RM. 2005. Yield rice. Field Crops Research 97, 87–100. and water use of irrigated tropical aerobic rice systems. Agricultural Water Ma X, Feng F, Wei H, Mei H, Xu K, Chen S, Li T, Liang X, Liu H, Luo Management 74, 87–105. L. 2016. Genome-wide association study for plant height and grain yield in Crowell S, Korniliev P, Falcão A, Ismail A, Gregorio G, Mezey rice under contrasting moisture regimes. Frontiers in Plant Science 7, 1801. J, McCouch S. 2016. Genome-wide association and high-resolution Millet EJ, Welcker C, Kruijer W, et al. 2016. Genome-wide analysis of phenotyping link Oryza sativa panicle traits to numerous trait-specific QTL yield in Europe: allelic effects vary with drought and heat scenarios. Plant clusters. Nature Communications 7, 10527. Physiology 172, 749–764. Dixit S, Singh A, Sta Cruz MT, Maturan PT, Amante M, Kumar A. 2014. Mishra KK, Vikram P, Yadaw RB, Swamy BP, Dixit S, Cruz MT, Maturan Multiple major QTL lead to stable yield performance of rice cultivars across P, Marker S, Kumar A. 2013. qDTY12.1: a locus with a consistent effect varying drought intensities. BMC Genetics 15, 16. on grain yield under drought in rice. BMC Genetics 14, 12. Doebley JF, Gaut BS, Smith BD. 2006. The molecular genetics of crop Miura K, Ikeda M, Matsubara A, Song XJ, Ito M, Asano K, Matsuoka domestication. Cell 127, 1309–1321. M, Kitano H, Ashikari M. 2010. OsSPL14 promotes panicle branching Donald CM. 1968. The breeding of crop ideotypes. Euphytica 17, 385–403. and higher grain productivity in rice. Nature Genetics 42, 545–549. Dong J, Kim ST, Lord EM. 2005. Plantacyanin plays a role in reproduction Moldenhauer K, Wilson EC, Counce P Jr, Hardke J. 2013. Rice growth in Arabidopsis. Plant Physiology 138, 778–789. and development. In: Hardke JT, ed. Arkansas rice production handbook. Little Rock: University of Arkansas Division of Agriculture Cooperative Fan C, Xing Y, Mao H, Lu T, Han B, Xu C, Li X, Zhang Q. 2006. GS3, Extension Service, 9–20. a major QTL for grain length and weight and minor QTL for grain width and thickness in rice, encodes a putative transmembrane protein. Theoretical Pantalião GF, Narciso M, Guimarães C, Castro A, Colombari JM, and Applied Genetics 112, 1164–1171. Breseghello F, Rodrigues L, Vianello RP, Borba TO, Brondani C. 2016. Genome wide association study (GWAS) for grain yield in rice cultivated Fan X, Tang Z, Tan Y, Zhang Y, Luo B, Yang M, Lian X, Shen Q, Miller under water deficit. Genetica 144, 651–664. AJ, Xu G. 2016. Overexpression of a pH-sensitive nitrate transporter in rice increases crop yields. Proceedings of the National Academy of Sciences, Pinto RS, Reynolds MP, Mathews KL, McIntyre CL, Olivares- USA 113, 7118–7123. Villegas JJ, Chapman SC. 2010. Heat and drought adaptive QTL in a wheat population designed to minimize confounding agronomic effects. Fujita D, Trijatmiko KR, Tagle AG, et al. 2013. NAL1 allele from a rice Theoretical and Applied Genetics 121, 1001–1021. landrace greatly increases yield in modern indica cultivars. Proceedings of the National Academy of Sciences, USA 110, 20431–20436. Posé D, Castanedo I, Borsani O, Nieto B, Rosado A, Taconnat L, Ferrer A, Dolan L, Valpuesta V, Botella MA. 2009. Identification of the Arabidopsis Fukai S, Pantuwan G, Jongdee B, Cooper M. 1999. Screening for dry2/sqe1-5 mutant reveals a central role for sterols in drought tolerance and drought resistance in rainfed lowland rice. Field Crops Research 64, 61–74. regulation of reactive oxygen species. The Plant Journal 59, 63–76. Guo Z, Chen D, Alqudah AM, Röder MS, Ganal MW, Schnurbusch T. 2017. Genome-wide association analyses of 54 traits identified multiple Pyngrope S, Bhoomika K, Dubey RS. 2013. Reactive oxygen species, loci for the determination of floret fertility in wheat. New Phytologist 214, ascorbate-glutathione pool, and enzymes of their metabolism in drought- 257–270. sensitive and tolerant indica rice (Oryza sativa L.) seedlings subjected to progressing levels of water deficit. Protoplasma 250, 585–600. Huang X, Qian Q, Liu Z, Sun H, He S, Luo D, Xia G, Chu C, Li J, Fu X. 2009. Natural variation at the DEP1 locus enhances grain yield in rice. Qiu X, Pang Y, Yuan Z, Xing D, Xu J, Dingkuhn M, Li Z, Ye G. 2015. Nature Genetics 41, 494–497. Genome-wide association study of grain appearance and milling quality in a worldwide collection of indica rice germplasm. PLoS ONE 10, e0145577. Huang X, Wei X, Sang T, et al. 2010. Genome-wide association studies of 14 agronomic traits in rice landraces. Nature Genetics 42, 961–967. Rebolledo MC, Peña AL, Duitama J, Cruz DF, Dingkuhn M, Grenier C, Tohme J. 2016. Combining image analysis, genome wide association Huang X, Zhao Y, Wei X, et al. 2012. Genome-wide association study studies and different field trials to reveal stable genetic regions related to of flowering time and grain yield traits in a worldwide collection of rice panicle architecture and the number of spikelets per panicle in rice. Frontiers germplasm. Nature Genetics 44, 32–39. in Plant Science 7, 1384. Kadam N, Tamilselvan A, Lawas LMF, et al. 2017. Genetic control of Reinders A. 2016. Fuel for the road—sugar transport and pollen tube plasticity in root morphology and anatomy of rice in response to water- growth. Journal of Experimental Botany 67, 2121–2123. deficit. Plant Physiology 174, 2302–2315. Reynolds MP, Quilligan E, Aggarwal PK, et al. 2016. An integrated Kadam NN, Xiao G, Melgar RJ, Bahuguna RN, Quinones C, approach to maintaining cereal productivity under climate change. Global Tamilselvan A, Prasad PVV, Jagadish KSV, Donald S. 2014. Agronomic Food Security 8, 9–18. and physiological responses to high temperature, drought, and elevated CO interactions in cereals. Advances in Agronomy 27, 111–156. Saini HS, Lalonde S. 1997. Injuries to reproductive development under water stress, and their consequences for crop productivity. Journal of Crop Kadam NN, Yin X, Bindraban PS, Struik PC, Jagadish KS. 2015. Does Production 1, 223–248. morphological and anatomical plasticity during the vegetative stage make wheat more tolerant of water deficit stress than rice? Plant Physiology 167, Saini HS, Westgate ME, Donald LS. 1999. Reproductive development in 1389–1401. grain crops during drought. Advances in Agronomy 68, 59–96. Kikuchi S, Bheemanahalli R, Jagadish KSV, Kumagai E, Masuya Segura V, Vilhjálmsson BJ, Platt A, Korte A, Seren Ü, Long Q, Y, Kuroda E, Raghavan C, Dingkuhn M, Abe A, Shimono H. 2017. Nordborg M. 2012. An efficient multi-locus mixed-model approach Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4032 | Kadam et al. for genome-wide association studies in structured populations. Nature Takemiya A, Sugiyama N, Fujimoto H, Tsutsumi T, Yamauchi S, Genetics 44, 825–830. Hiyama A, Tada Y, Christie JM, Shimazaki K. 2013. Phosphorylation of BLUS1 kinase by phototropins is a primary step in stomatal opening. Nature Selote DS, Chopra RK. 2004. Drought induced spikelet sterility is Communications 4, 2094. associated with an inefficient antioxidant defense in rice panicles. Physiologia Plantarum 121, 462–471. Venuprasad R, Dalid CO, Del Valle M, Zhao D, Espiritu M, Sta Cruz MT, Amante M, Kumar A, Atlin GN. 2009. Identification and characterization Sheoran IS, Saini HS. 1996. Drought-induced male sterility in rice: changes of large-effect quantitative trait loci for grain yield under lowland drought in carbohydrate levels and enzyme activities associated with the inhibition stress in rice using bulk-segregant analysis. Theoretical and Applied of starch accumulation in pollen. Sexual Plant Reproduction 9, 161–169. Genetics 120, 177–190. Shi W, Yin X, Struik PC, Xie F, Schmidt RC, Jagadish KSV. 2016. Venuprasad R, Lafitte HR, Atlin GN. 2007. Response to direct selection Grain yield and quality responses of tropical hybrid rice to high night-time for grain yield under drought stress in rice. Crop Science 47, 285–293. temperature. Field Crops Research 190, 18–25. Vikram P, Swamy BP, Dixit S, Ahmed HU, Teresa Sta Cruz M, Singh Shomura A, Izawa T, Ebana K, Ebitani T, Kanegae H, Konishi S, Yano AK, Kumar A. 2011. qDTY₁.₁, a major QTL for rice grain yield under M. 2008. Deletion in a gene associated with grain size increased yields reproductive-stage drought stress with a consistent effect in multiple elite during rice domestication. Nature Genetics 40, 1023–1028. genetic backgrounds. BMC Genetics 12, 89. Slafer GA. 2003. Genetic basis of yield as viewed from a crop physiologist’s Vikram P, Swamy BP, Dixit S, Singh R, Singh BP, Miro B, Kohli A, perspective. Annals of Applied Biology 142, 117–128. Henry A, Singh NK, Kumar A. 2015. Drought susceptibility of modern rice varieties: an effect of linkage of drought tolerance with undesirable traits. Song XJ, Huang W, Shi M, Zhu MZ, Lin HX. 2007. A QTL for rice grain Scientific Reports 5, 14799. width and weight encodes a previously unknown RING-type E3 ubiquitin ligase. Nature Genetics 39, 623–630. Wang X, Feng S, Nakayama N, Crosby WL, Irish V, Deng XW, Wei N. 2003. The COP9 signalosome interacts with SCF UFO and participates in Spindel J, Begum H, Akdemir D, Virk P, Collard B, Redoña E, Atlin Arabidopsis flower development. The Plant Cell 15, 1071–1082. G, Jannink JL, McCouch SR. 2015. Genomic selection and association mapping in rice (Oryza sativa): effect of trait genetic architecture, training Yano K, Yamamoto E, Aya K, et al. 2016. Genome-wide association study population composition, marker number and statistical model on accuracy of using whole-genome sequencing rapidly identifies new genes influencing rice genomic selection in elite, tropical rice breeding lines. PLoS Genetics 11, agronomic traits in rice. Nature Genetics 48, 927–934. e1004982. Yin X, Chasalow SD, Stam P, Kropff MJ, Dourleijn CJ, Bos I, Bindraban Swamy BP, Ahmed HU, Henry A, et al. 2013. Genetic, physiological, PS. 2002. Use of component analysis in QTL mapping of complex crop and gene expression analyses reveal that multiple QTL enhance yield of rice traits: a case study on yield in barley. Plant Breeding 121, 314–319. mega-variety IR64 under drought. PLoS ONE 8, e62795. Yin X, Kropff MJ, Horie T, Nakagawa H, Centeno HGS, Zhu D, Swamy BPM, Shamsudin NAA, Rahman SNA, Mauleon R, Ratnam Goudriaan J. 1997. A model for photothermal responses of flowering in W, Sta Cruz MT, Kumar A. 2017. Association mapping of yield and yield- rice I. Model description and parameterization. Field Crops Research 51, related traits under reproductive stage drought stress in rice (Oryza sativa 189–200. L.). Rice 10, 21. Zhang Z, Ersoz E, Lai CQ, et al. 2010. Mixed linear model approach Swamy BP, Vikram P, Dixit S, Ahmed HU, Kumar A. 2011. Meta- adapted for genome-wide association studies. Nature Genetics 42, analysis of grain yield QTL identified during agricultural drought in grasses 355–360. showed consensus. BMC Genomics 12, 319. Zhao K, Tung CW, Eizenga GC, et al. 2011. Genome-wide association Sweeney M, McCouch S. 2007. The complex history of the domestication mapping reveals a rich genetic architecture of complex traits in Oryza sativa. of rice. Annals of Botany 100, 951–957. Nature Communications 2, 467. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Experimental Botany Oxford University Press http://www.deepdyve.com/lp/oxford-university-press/genome-wide-association-reveals-novel-genomic-loci-controlling-rice-VcMmDXOt4b

Loading next page...

References (71)

A. Reinders (2016)
Fuel for the road – sugar transport and pollen tube growth
Journal of Experimental Botany, 67
Shengben Li, Q. Qian, Zhiming Fu, D. Zeng, Xiangbing Meng, J. Kyozuka, M. Maekawa, Xu-dong Zhu, Jian Zhang, Jiayang Li, Yonghong Wang (2009)
Short panicle1 encodes a putative PTR family transporter and determines rice panicle size.
The Plant journal : for cell and molecular biology, 58 4
Xiping Wang, S. Feng, N. Nakayama, W. Crosby, V. Irish, X. Deng, N. Wei (2003)
The COP9 Signalosome Interacts with SCFUFO and Participates in Arabidopsis Flower Development Article, publication date, and citation information can be found at www.plantcell.org/cgi/doi/10.1105/tpc.009936.
The Plant Cell Online, 15
Emilie Millet, C. Welcker, W. Kruijer, S. Negro, A. Coupel-Ledru, S. Nicolas, J. Laborde, C. Bauland, S. Praud, N. Ranc, T. Presterl, R. Tuberosa, Z. Bedo, X. Draye, B. Usadel, A. Charcosset, F. Eeuwijk, F. Tardieu (2016)
Genome-Wide Analysis of Yield in Europe: Allelic Effects Vary with Drought and Heat Scenarios.
Plant physiology, 172 2
Tereza Borba, R. Brondani, F. Breseghello, A. Coelho, J. Mendonça, P. Rangel, C. Brondani (2010)
Association mapping for yield and grain quality traits in rice (Oryza sativa L.)
Genetics and Molecular Biology, 33
N. Kadam, Xinyou Yin, P. Bindraban, P. Struik, K. Jagadish (2015)
Does Morphological and Anatomical Plasticity during the Vegetative Stage Make Wheat More Tolerant of Water Deficit Stress Than Rice?1[OPEN]
Plant Physiology, 167
Xiaorong Fan, Zhong Tang, Yawen Tan, Yong Zhang, Bingbing Luo, Meng Yang, Xingming Lian, Q. Shen, A. Miller, Guohua Xu (2016)
Overexpression of a pH-sensitive nitrate transporter in rice increases crop yields
Proceedings of the National Academy of Sciences, 113
J. Spindel, H. Begum, Deniz Akdemir, P. Virk, B. Collard, E. Redoña, G. Atlin, J. Jannink, S. McCouch (2015)
Genomic Selection and Association Mapping in Rice (Oryza sativa): Effect of Trait Genetic Architecture, Training Population Composition, Marker Number and Statistical Model on Accuracy of Rice Genomic Selection in Elite, Tropical Rice Breeding Lines
PLoS Genetics, 11
Gabriel Pantalião, M. Narciso, C. Guimarães, A. Castro, José Colombari, F. Breseghello, L. Rodrigues, R. Vianello, T. Borba, C. Brondani (2016)
Genome wide association study (GWAS) for grain yield in rice cultivated under water deficit
Genetica, 144
Zifeng Guo, Dijun Chen, Ahmad Alqudah, M. Röder, M. Ganal, T. Schnurbusch (2017)
Genome-wide association analyses of 54 traits identified multiple loci for the determination of floret fertility in wheat.
The New phytologist, 214 1
G. Slafer (2003)
Genetic basis of yield as viewed from a crop physiologist's perspective
Annals of Applied Biology, 142
H. Saini, M. Westgate (1999)
Reproductive Development in Grain Crops during Drought
Advances in Agronomy, 68
Keyan Zhao, Chih-Wei Tung, G. Eizenga, Mark Wright, M. Ali, A. Price, G. Norton, S. Islam, Andy Reynolds, J. Mezey, A. McClung, Carlos Bustamante, Carlos Bustamante, S. McCouch (2011)
Genome-wide association mapping reveals a rich genetic architecture of complex traits in Oryza sativa
Nature Communications, 2
O. Arrigoni, M. Tullio (2002)
Ascorbic acid: much more than just an antioxidant.
Biochimica et biophysica acta, 1569 1-3
Megan Sweeney, S. McCouch (2007)
The Complex History of the Domestication of Rice
Annals of Botany, 100
N. Al-Tamimi, C. Brien, H. Oakey, B. Berger, Stephanie Saade, Y. Ho, Sandra Schmöckel, M. Tester, S. Negrão (2016)
Salinity tolerance loci revealed in rice using high-throughput non-invasive phenotyping
Nature Communications, 7
H. Begum, J. Spindel, Antonio Lalusin, T. Borromeo, G. Gregorio, Jose Hernandez, P. Virk, B. Collard, S. McCouch (2015)
Genome-Wide Association Mapping for Yield and Other Agronomic Traits in an Elite Breeding Population of Tropical Rice (Oryza sativa)
PLoS ONE, 10
S. Kikuchi, R. Bheemanahalli, K. Jagadish, E. Kumagai, Yusuke Masuya, E. Kuroda, C. Raghavan, M. Dingkuhn, A. Abe, H. Shimono (2017)
Genome-wide association mapping for phenotypic plasticity in rice.
Plant, cell & environment, 40 8
Shalabh Dixit, Anshuman Singh, M. Cruz, Paul Maturan, M. Amante, Arvind Kumar (2014)
Multiple major QTL lead to stable yield performance of rice cultivars across varying drought intensities
BMC Genetics, 15
Arvind Kumar, Shalabh Dixit, T. Ram, R. Yadaw, K. Mishra, N. Mandal (2014)
Breeding high-yielding drought-tolerant rice: genetic variations and conventional and molecular approaches
Journal of Experimental Botany, 65
Xianzhong Huang, Q. Qian, Zhengbin Liu, Hongying Sun, Shuyuan He, D. Luo, G. Xia, Chengcai Chu, Jiayang Li, Xiangdong Fu (2009)
Natural variation at the DEP1 locus enhances grain yield in rice
Nature Genetics, 41
Samuel Crowell, Pavel Korniliev, A. Falcão, A. Ismail, G. Gregorio, J. Mezey, S. McCouch (2016)
Genome-wide association and high-resolution phenotyping link Oryza sativa panicle traits to numerous trait-specific QTL clusters
Nature Communications, 7
Xian-Jun Song, Wei Huang, Min Shi, Mei-Zhen Zhu, Hong-Xuan Lin (2007)
A QTL for rice grain width and weight encodes a previously unknown RING-type E3 ubiquitin ligase
Nature Genetics, 39
V. Segura, B. Vilhjálmsson, Alexander Platt, A. Korte, Ümit Seren, Q. Long, M. Nordborg (2012)
An efficient multi-locus mixed model approach for genome-wide association studies in structured populations
Nature genetics, 44
Mallikarjuna P., H. Ahmed, A. Henry, R. Mauleon, Shalabh Dixit, P. Vikram, Ram Tilatto, S. Verulkar, P. Perraju, N. Mandal, M. Variar, R. S, R. Chandrababu, O. Singh, J. Dwivedi, S. Das, K. Mishra, R. Yadaw, T. Aditya, B. Karmakar, Kouji Satoh, A. Moumeni, S. Kikuchi, H. Leung, Arvind Kumar (2013)
Genetic, Physiological, and Gene Expression Analyses Reveal That Multiple QTL Enhance Yield of Rice Mega-Variety IR64 under Drought
PLoS ONE, 8
S. Fukai, G. Pantuwan, B. Jongdee, M. Cooper (1999)
Screening for drought resistance in rainfed lowland rice
Field Crops Research, 64
Huang (2010)
Genome-wide association studies of 14 agronomic traits in rice landraces
Nature Genetics, 42
P. Vikram, P. Vikram, B. Swamy, Shalabh Dixit, Renu Singh, B. Singh, B. Miro, A. Kohli, A. Henry, N. Singh, Arvind Kumar (2015)
Drought susceptibility of modern rice varieties: an effect of linkage of drought tolerance with undesirable traits
Scientific Reports, 5
B. Barnabás, K. Jäger, A. Fehér (2007)
The effect of drought and heat stress on reproductive processes in cereals.
Plant, cell & environment, 31 1
J. Bernier, Arvind Kumar, V. Ramaiah, D. Spaner, G. Atlin (2007)
A large-effect QTL for grain yield under reproductive-stage drought stress in upland rice.
Crop Science, 47
Lipka (2012)
GAPIT: genome association and prediction integrated tool
Bioinformatics, 28
David Posé, Itziar Castanedo, O. Borsani, Benjamín Nieto, A. Rosado, L. Taconnat, A. Ferrer, L. Dolan, V. Valpuesta, M. Botella (2009)
Identification of the Arabidopsis dry2/sqe1-5 mutant reveals a central role for sterols in drought tolerance and regulation of reactive oxygen species.
The Plant journal : for cell and molecular biology, 59 1
Jian-Xiang Liu, D. Liao, R. Oane, L. Estenor, X. Yang, Zichao Li, J. Bennett (2006)
Genetic variation in the sensitivity of anther dehiscence to drought stress in rice
Field Crops Research, 97
M. Ashikari, H. Sakakibara, Shaoyang Lin, Toshio Yamamoto, T. Takashi, A. Nishimura, E. Angeles, Q. Qian, H. Kitano, M. Matsuoka (2005)
Cytokinin Oxidase Regulates Rice Grain Production
Science, 309
Kenji Yano, E. Yamamoto, Koichiro Aya, H. Takeuchi, Pei-ching Lo, Li Hu, M. Yamasaki, S. Yoshida, H. Kitano, K. Hirano, M. Matsuoka (2016)
Genome-wide association study using whole-genome sequencing rapidly identifies new genes influencing agronomic traits in rice
Nature Genetics, 48
K. Mishra, P. Vikram, R. Yadaw, B. Swamy, Shalabh Dixit, M. Cruz, Paul Maturan, S. Marker, Arvind Kumar (2013)
qDTY12.1: a locus with a consistent effect on grain yield under drought in rice
BMC Genetics, 14
H. Agrama, G. Eizenga, W. Yan (2007)
Association mapping of yield and its components in rice cultivars
Molecular Breeding, 19
Juan Dong, S. Kim, E. Lord (2005)
Plantacyanin Plays a Role in Reproduction in Arabidopsis1
Plant Physiology, 138
R. Suzuky, Pinto bullet, M. Reynolds, bullet Ky, L. Mathews, bullet Lynne, Mcintyre bullet, J. bullet, S. Chapman (2010)
Heat and drought adaptive QTL in a wheat population designed to minimize confounding agronomic effects
TAG. Theoretical and Applied Genetics. Theoretische Und Angewandte Genetik, 121
D. Fujita, K. Trijatmiko, A. Tagle, Maria Sapasap, Y. Koide, K. Sasaki, Nikolaos Tsakirpaloglou, R. Gannaban, Takeshi Nishimura, S. Yanagihara, Y. Fukuta, T. Koshiba, I. Slamet-Loedin, T. Ishimaru, N. Kobayashi (2013)
NAL1 allele from a rice landrace greatly increases yield in modern indica cultivars
Proceedings of the National Academy of Sciences, 110
C. Donald (1968)
The breeding of crop ideotypes
Euphytica, 17
C. Fan, Y. Xing, Hailiang Mao, Tingting Lu, B. Han, Cai-guo Xu, Xianghua Li, Qifa Zhang (2006)
GS3, a major QTL for grain length and weight and minor QTL for grain width and thickness in rice, encodes a putative transmembrane protein
Theoretical and Applied Genetics, 112
Kotaro Miura, M. Ikeda, Atsushi Matsubara, Xian-Jun Song, Midori Ito, K. Asano, M. Matsuoka, H. Kitano, M. Ashikari (2010)
OsSPL14 promotes panicle branching and higher grain productivity in rice
Nature Genetics, 42
A. Shomura, T. Izawa, K. Ebana, T. Ebitani, H. Kanegae, Saeko Konishi, M. Yano (2008)
Deletion in a gene associated with grain size increased yields during rice domestication
Nature Genetics, 40
Xinyou Yin, M. Kropff, T. Horie, H. Nakagawa, H. Centeno, De-feng Zhu, J. Goudriaan (1997)
A model for photothermal responses of flowering in rice I. Model description and parameterization
Field Crops Research, 51
(2012)
Genetics and population analysis Advance Access publication July 13, 2012
M. Rebolledo, A. Peña, J. Duitama, D. Cruz, M. Dingkuhn, C. Grenier, J. Tohme (2016)
Combining Image Analysis, Genome Wide Association Studies and Different Field Trials to Reveal Stable Genetic Regions Related to Panicle Architecture and the Number of Spikelets per Panicle in Rice
Frontiers in Plant Science, 7
S. Pyngrope, K. Bhoomika, R. Dubey (2012)
Reactive oxygen species, ascorbate–glutathione pool, and enzymes of their metabolism in drought-sensitive and tolerant indica rice (Oryza sativa L.) seedlings subjected to progressing levels of water deficit
Protoplasma, 250
R. Venuprasad, C. Dalid, M. Valle, Dule Zhao, M. Espiritu, M. Cruz, M. Amante, Arvind Kumar, G. Atlin (2009)
Identification and characterization of large-effect quantitative trait loci for grain yield under lowland drought stress in rice using bulk-segregant analysis
Theoretical and Applied Genetics, 120
P. Vikram, B. Swamy, Shalabh Dixit, H. Ahmed, Ma Cruz, Alok Singh, Arvind Kumar (2011)
qDTY1.1, a major QTL for rice grain yield under reproductive-stage drought stress with a consistent effect in multiple elite genetic backgrounds
BMC Genetics, 12
N. Kadam, Guilin Xiao, R. Melgar, R. Bahuguna, Cherryl Quiñones, Anandhan Tamilselvan, P. Prasad, K. Jagadish (2014)
Agronomic and Physiological Responses to High Temperature, Drought, and Elevated CO2 Interactions in Cereals
Advances in Agronomy, 127
N. Kadam, N. Kadam, Anandhan Tamilselvan, Anandhan Tamilselvan, L. Lawas, Cherryl Quiñones, R. Bahuguna, M. Thomson, M. Thomson, M. Dingkuhn, R. Muthurajan, P. Struik, Xinyou Yin, S. Jagadish (2017)
Genetic Control of Plasticity in Root Morphology and Anatomy of Rice in Response to Water Deficit1[OPEN]
Plant Physiology, 174
W. Shi, Xinyou Yin, P. Struik, Fangming Xie, Ralf Schmidt, K. Jagadish (2016)
Grain yield and quality responses of tropical hybrid rice to high night-time temperature
Field Crops Research, 190
(2013)
Arkansas rice production handbook
A. Takemiya, Naoyuki Sugiyama, H. Fujimoto, Toshifumi Tsutsumi, S. Yamauchi, Asami Hiyama, Y. Tada, J. Christie, K. Shimazaki (2013)
Phosphorylation of BLUS1 kinase by phototropins is a primary step in stomatal opening
Nature Communications, 4
Devarshi Selote, R. Khanna-Chopra (2004)
Drought‐induced spikelet sterility is associated with an inefficient antioxidant defence in rice panicles
Physiologia Plantarum, 121
J. Lanceras, G. Pantuwan, B. Jongdee, T. Toojinda (2004)
Quantitative Trait Loci Associated with Drought Tolerance at Reproductive Stage in Rice1
Plant Physiology, 135
B. Swamy, P. Vikram, Shalabh Dixit, H. Ahmed, Arvind Kumar (2011)
Meta-analysis of grain yield QTL identified during agricultural drought in grasses showed consensus
BMC Genomics, 12
R. Venuprasad, H. Lafitte, G. Atlin (2007)
Response to Direct Selection for Grain Yield under Drought Stress in Rice
Crop Science, 47
M. Reynolds, E. Quilligan, P. Aggarwal, K. Bansal, A. Cavalieri, S. Chapman, S. Chapotin, S. Datta, E. Duveiller, K. Gill, K. Jagadish, A. Joshi, A. Koehler, P. Kosina, S. Krishnan, R. Lafitte, R. Mahala, R. Muthurajan, A. Paterson, B. Prasanna, S. Rakshit, M. Rosegrant, I. Sharma, R. Singh, S. Sivasankar, V. Vadez, R. Valluru, P. Prasad, O. Yadav (2016)
An integrated approach to maintaining cereal productivity under climate change
Global Food Security, 8
Xuehui Huang, Yan Zhao, Xinghua Wei, Canyang Li, Ahong Wang, Qiang Zhao, Wenjun Li, Yunli Guo, Liuwei Deng, Chuanrang Zhu, Danlin Fan, Yiqi Lu, Qijun Weng, Kunyan Liu, Taoying Zhou, Y. Jing, Lizhen Si, Guojun Dong, Tao Huang, Tingting Lu, Qi Feng, Q. Qian, Jiayang Li, B. Han (2011)
Genome-wide association study of flowering time and grain yield traits in a worldwide collection of rice germplasm
Nature Genetics, 44
I. Sheoran, H. Saini (2005)
Drought-induced male sterility in rice: Changes in carbohydrate levels and enzyme activities associated with the inhibition of starch accumulation in pollen
Sexual Plant Reproduction, 9
R. Dunand, J. Saichuk
Rice Growth and Development
J. Doebley, B. Gaut, Bruce Smith (2006)
The Molecular Genetics of Crop Domestication
Cell, 127
Zhiwu Zhang, E. Ersoz, Chao-Qiang Lai, R. Todhunter, H. Tiwari, M. Gore, Peter Bradbury, Jianming Yu, D. Arnett, J. Ordovás, E. Buckler (2010)
Mixed linear model approach adapted for genome-wide association studies
Nature Genetics, 42
Xinyou Yin, S. Chasalow, P. Stam, M. Kropff, C. Dourleijn, I. Bos, P. Bindraban (2002)
Use of component analysis in QTL mapping of complex crop traits : A case study on yield in barley
Plant Breeding, 121
B. Swamy, N. Shamsudin, S. Rahman, R. Mauleon, W. Ratnam, M. Cruz, Arvind Kumar (2017)
Association Mapping of Yield and Yield-related Traits Under Reproductive Stage Drought Stress in Rice (Oryza sativa L.)
Rice, 10
H. Saini, S. Lalonde (1997)
INJURIES TO REPRODUCTIVE DEVELOPMENT UNDER WATER STRESS, AND THEIR CONSEQUENCES FOR CROP PRODUCTIVITY
Journal of crop production, 1
Xianjin Qiu, Yun-long Pang, Zhi-hua Yuan, Danying Xing, Jianlong Xu, M. Dingkuhn, Zhikang Li, G. Ye (2015)
Genome-Wide Association Study of Grain Appearance and Milling Quality in a Worldwide Collection of Indica Rice Germplasm
PLoS ONE, 10
B. Bouman, S. Peng, A. Castañeda, R. Visperas (2005)
Yield and water use of irrigated tropical aerobic rice systems
Agricultural Water Management, 74
Xiaosong Ma, Fangjun Feng, Haibin Wei, H. Mei, Kai Xu, Shoujun Chen, Tianfei Li, Xiaohua Liang, Hongyan Liu, L. Luo (2016)
Genome-Wide Association Study for Plant Height and Grain Yield in Rice under Contrasting Moisture Regimes
Frontiers in Plant Science, 7

Publisher: Oxford University Press
Copyright: Copyright © 2022 Society for Experimental Biology
ISSN: 0022-0957
eISSN: 1460-2431
DOI: 10.1093/jxb/ery186
Publisher site: See Article on Publisher Site

Abstract

A diversity panel comprising of 296 indica rice genotypes was phenotyped under non-stress and water-deficit stress conditions during the reproductive stage in the 2013 and 2014 dry seasons (DSs) at IRRI, Philippines. We investigated the genotypic variability for grain yield, yield components, and related traits, and conducted genome-wide association stud- ies (GWAS) using high-density 45K single nucleotide polymorphisms. We detected 38 loci in 2013 and 64 loci in 2014 for non-stress conditions and 69 loci in 2013 and 55 loci in 2014 for water-deficit stress. Desynchronized flowering time con- founded grain yield and its components under water-deficit stress in the 2013 experiment. Statistically corrected grain yield and yield component values using days to flowering helped to detect 31 additional genetic loci for grain yield, its components, and the harvest index in 2013. There were few overlaps in the detected loci between years and treatments, and when compared with previous studies using the same panel, indicating the complexity of yield formation under stress. Nevertheless, our analyses provided important insights into the potential links between grain yield with seed set and assimilate partitioning. Our findings demonstrate the complex genetic architecture of yield formation and we pro- pose exploring the genetic basis of less complex component traits as an alternative route for further yield enhancement. Keywords: A priori candidate genes, multi-locus analysis, Oryza sativa, reproductive-stage water-deficit stress, single-locus analysis, synchronized phenology. Introduction Rice (Oryza sativa L.) is a staple food crop for more than half the productivity of rice (Kadam et al., 2014; Reynolds et al., the world’s population. Maintaining its high yield potential 2016), as rice is more sensitive to water deficit than other cere- with sustained productivity is imperative for future food se- als (Kadam et al., 2015). Nearly 20% of global rice production curity. However, global climate change, with frequent epi- is affected by water deficit (Bouman et al., 2005). Water deficit sodes of abiotic stress (water deficit and heat stress), reduces can occur at any time during the growing season, but stress © The Author(s) 2018. Published by Oxford University Press on behalf of the Society for Experimental Biology. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4018 | Kadam et al. adaptation. This panel was assembled at the International Rice Research occurring within the reproductive phase (i.e. from meiosis Institute (IRRI), Philippines for the Phenomics of Rice Adaptation to flowering) causes the greatest grain yield losses (Liu et al., and Yield potential (PRAY) project in the context of the Global Rice 2006). The physiological effects of water deficit within the re- Phenotyping Network (http://ricephenonetwork.irri.org). Recent stud- productive phase have been discussed in detail by Saini and ies have reported GWAS analyses using this population for grain quality Lalonde (1997), Saini et al. (1999), and Barnabás et al. (2008). traits (Qiu et al., 2015), salinity tolerance (Al-Tamimi et al., 2016), panicle architecture (Rebolledo et al., 2016), yield traits under varying planting Increasing tolerance to water deficit has been considered as a densities (Kikuchi et al., 2017), and root plasticity (Kadam et al., 2017). major breeding target, although knowledge on phenotypic traits linked with stress tolerance is limited. Recent evidence in rice has Strategy to cope with variation in flowering phenology demonstrated that progress can be made through direct selection of grain yield, as a criterion under reproductive-stage water def- The PRAY panel was screened in non-stress and reproductive-stage water-deficit conditions under field experiments conducted at the upland icit (Venuprasad et al., 2007; Kumar et al., 2014). Physiologically, farm of IRRI, Philippines (14°11′N, 121°15′E; elevation 21 m above sea grain yield is a very complex trait determined by different com- level) in the 2013 and 2014 DSs. Seeds were sown from December of ponent traits (Slafer, 2003). Hence, exploring ideotype breeding the preceding year to late January or early February of each year (Fig. 1). based on selection for component traits is proposed as a com- As expected, a strong genotypic variation in flowering phenology was plementary route for further yield improvement (Donald, 1968). observed that confounds the true water-deficit response (Fukai et al., 1999) and inevitably induces bias with interpretation of genetic mapping Revealing the genetic basis of grain yield and its compo- outcomes (Pinto et al., 2010; Kumar et al., 2014). We followed staggered nent traits is essential for providing breeders with the tools for sowing in seedbeds and transplanting in main plots to synchronize flower- efficient development of stress-resilient cultivars. The genetic ing and thus minimize phenological differences under stress imposition control of grain yield under reproductive-stage water deficit (Fig. 1). Briefly, in the 2013 DS experiment, we divided 296 genotypes has been investigated extensively using linkage analysis of bi- into six groups with a 10 d interval based on days to flowering data col- lected from a previous experiment conducted in the 2012 wet season parental crosses in rice. This approach has proven to be power- (WS), our only source of flowering dates for this population grown at ful in the detection of quantitative trait loci (QTLs) for grain IRRI. While the expected range of flowering was 29 March to 8 April yield and its components under stress (Lanceras et al., 2004; 2013 (Fig. 1A), we observed deviation in days to flowering in the 2013 Bernier et al., 2007; Vikram et al., 2011; Mishra et al., 2013; DS experiment, where the staggered sowing was based on the 2012 WS Dixit et al., 2014; Kumar et al., 2014). A few of these QTLs data. Therefore, in the 2014 DS experiment, we regrouped the 296 geno- types into eight groups with a 7 d interval using 2013 DS flowering data regulating grain yield, for instance qDTY , have been intro- 12.1 to improve synchrony within the whole population. The expected date of gressed into elite cultivars to improve stress tolerance (Mishra flowering was 28 March to 5 April 2014 for these genotypes (Fig. 1B). In et al., 2013), but most of them are only based on a small frac- each year, the sowing date chosen for the stress treatment was the same as tion of the rice genetic diversity. Identifying the allelic vari- for the non-stress treatment of the same genotype. ations exhibited in a large genetic diversity panel as a result of divergent selection pressure provides an obvious alterna- Crop management tive that can have a greater potential in grain yield improve- The soil type of the upland farm at IRRI is Maahas clay loam, isohyper- ment under water deficit. These natural allelic variations have thermic mixed Typic Tropudalf. The experiments were laid out in a group been identified in rice under non-stress conditions for grain block design with three replications for each genotype in both treatments yield and its component traits through genome-wide associ- (Supplementary Fig. S1 at JXB online). Seeds were first exposed to 50 °C for 3 d to break dormancy and then hand sown in a seedbed nursery. ation studies (GWAS) (Agrama et al., 2007; Borba et al., 2010; Twenty-one-day-old seedlings were transplanted (two seedlings per hill) Huang et al., 2010, 2012; Zhao et al., 2011; Begum et al., 2015; for each genotype in four rows per replication. In both years, row dis- Spindel et al., 2015; Rebolledo et al., 2016; Yano et al., 2016). tance was 0.2 m and row length was 2.4 m. The seeds of one genotype Yet, very few studies are available for reproductive-stage water- in 2013 and eight genotypes in 2014 germinated poorly and hence were deficit conditions (Ma et al., 2016; Pantalião et al., 2016; Swamy excluded. In addition, four genotypes completed flowering and maturity before stress imposition in 2013 and were excluded. This resulted in final et al., 2017). This is partly due to the difficulty in implementing sets of 291 genotypes in 2013 and 288 genotypes in 2014, with three water deficit to coincide with reproductive stage under field replications and two treatments totalling 1746 and 1728 plots in 2013 conditions for a large diversity panel, which usually consists −1 and 2014, respectively. A day before transplanting, 30 kg P ha (as single −1 −1 of genotypes having diverse phenology. Only the study of Ma superphosphate), 40 kg K ha (as KCl), and 5 kg Zn ha (as zinc sulfate et al. (2016) followed a staggered sowing to account for vari- heptahydrate) fertilizers were manually applied. Nitrogen fertilizer as urea −1 −1 was applied in three splits: 45 kg ha before transplanting, 30 kg ha at ation in flowering phenology under stress. −1 mid-tillering, and 45 kg ha at panicle initiation. The IRRI standard Our study aimed to (i) explore the natural variation in grain management practices were followed to control weeds, insects, and dis- yield and yield component traits under non-stress and repro- eases. In both years, all plots were maintained like irrigated lowlands with ductive-stage water-deficit conditions; (ii) link the variation of ~5 cm standing water until maturity except for the water-deficit plots these phenotypic traits with single nucleotide polymorphisms during the stress period (see below). (SNPs) through GWAS; and (iii) identify the most likely under- lying candidate genes in close proximity to the significant SNPs. Reproductive stage water-deficit stress imposition There was variation in synchronizing days to flowering among rice geno- types in 2013, resulting in deviation from our expected flowering win- Materials and methods dow (29 March to 8 April). In rice, the reproductive stage ranges between 19 and 25 d, starting at panicle initiation and ending with flowering Association mapping population (Moldenhauer et al., 2013). Therefore, before imposing stress, we manu- We used a rice panel consisting of a diverse set of 296 indica genotypes ally dissected the main tillers of the middle two plants of border rows consisting of improved and traditional genotypes with (sub)tropical from water-deficit plots for all the genotypes, primarily to check the Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4019 Sowing Transplanting (21 days after sowing) Flowering Maturity 18 Dec 2012 8 Jan 2013 G1: 101-110 DTF (n=9) 28 Dec 18 Jan G2: 91-100 DTF (n=31) 7Jan 2013 28 Jan G3: 81-90 DTF (n=80) 2013 Experiment 7 Feb 17 Jan G4: 71-80 DTF (n=119) 18 Feb 27 Jan G5: 61-70 DTF (n=42) 6Feb 27 Feb G6: 51-60 DTF (n=15) Transplanting (21 days after sowing) Flowering Sowing Maturity 10 Dec 2013 31 Dec 2013 G1: 108-120 DTF (n=5) 17 Dec 07 Jan 2014 G2: 101-107 DTF (n=5) 24 Dec 14 Jan G3: 93-100 DTF (n=24) 31 Dec 21 Jan G4: 85-92 DTF (n=65) 2014 Experiment 28 Jan 07 Jan 2014 G5: 78-84 DTF (n=89) 14 Jan 04 Feb G6: 70-77 DTF (n=68) 21 Jan 11 Feb G7: 63-69 DTF (n=33) 28 Jan 18 Feb G8: 55-62 DTF (n=7) C D 2013 (n=291) 140 2014 (n=288) Non-stress Non-stress water-deficit stress Water-deficit stress y =0.8891x + 12.268 y =0.669x + 33.183 WD WD *** r²=0.85*** r²=0.46 y =0.9092x + 9.56 NS y =0.6879x + 26.976 NS *** r²=0.91*** 60 r²=0.53 50 50 TrtGYPNSPP SS TGW SP HI TrtGYPNSPP SS TGWSPHI 40 40 NS * ns ns ** ns ns ** NS * ns ns ** ** ns ns 30 30 WD ns ns ns ns ns ns ns WD *** * ** ****** ****** 20 20 20 30 40 50 60 70 80 90 100 110 120130 140 20 30 40 50 60 70 80 90 100 110 120 130 140 Expected DTF (2012) Expected DTF (2013) Fig. 1. Schematic representation of the staggered sowing and transplanting approach to synchronize flowering time that was followed for screening of an indica rice diversity panel under reproductive-stage water-deficit stress in the dry seasons (DSs) of 2013 (A) and 2014 (B). Days to flowering (DTF) was 10 d between groups (G) in 2013 and 7 days in 2014 DS experiments. (C, D) The expected and observed DTF in non-stress (NS) and water-deficit stress (WD) in the 2013 (C) and 2014 (D) DS experiments. ANOVA results with the effect of DTF (as a covariate in mixed linear model) on grain yield and 2 2 its key component traits are shown. GY, grain yield; HI, harvest index; n, number of genotypes; PN, panicles per m ; SP, spikelets per m ; SPP, spikelets per panicle; SS, seed set; TGW, thousand grain weight; Trt, treatments. Significance levels: *P<0.05, **P<0.01, ***P<0.001. To synchronize the flowering time, we used the 2012 wet season DTF data in the 2013 DS experiment (C). Similarly, for the 2014 DS experiment, we used DTF data from the 2013 DS experiment (D). (This figure is available in colour at JXB online.) reproductive-stage development. Stress was imposed on 23 March 2013 To quantify the stress intensity, 26 tensiometers were installed ran- when the majority of genotypes reached the agronomic panicle initiation domly across the entire stress field at 30 cm depth in each season. A poly- stage, by draining water out from the field. The stress continued for 14 thene sheet was inserted at 2 m depth by digging a deep and narrow d until 5 April 2013. In the 2014 experiment, the synchronization was trench in between stress and non-stress fields to prevent water seepage more precise with expected dates of flowering occurring between 28 during the stress period from the adjacent non-stress field. In addition, March and 5 April, as predicted. The same dissection approach as in 2013 the stress and the non-stress plots were separated by a distance of 2.3 was followed and stress was imposed on 26 March 2014 and continued m (Supplementary Fig. S1). The intensity of stress was higher in 2014 for 14 d until 8 April. than in 2013 (Supplementary Fig. S2A). There was no rainfall during Observed DTF (2013) Observed DTF (2014) 28 March 29 March 8April 5 April Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4020 | Kadam et al. the peak stress period in both seasons, except for rainfall during the first multiple linear regression analysis of grain yield with its component and day of the stress period in 2013 (Supplementary Fig. S2B). Higher stress related traits. intensity in 2014 compared with 2013 could be due to higher maximum temperature and higher vapour-pressure deficit (Supplementary Fig. S3B, Heritability estimates D), leading to quicker loss of soil moisture in 2014. A weather station was Broad-sense heritability (H ), capturing the proportion of phenotypic placed between the non-stress and water-deficit plots (see Supplementary variance explained by genetic factors that is due to dominance, epistatic, Fig. S1). Detailed weather data are given in Supplementary Fig. S3. and additive effects, was calculated across years and treatments separately using the below equation: Observations 2 2 G At maturity, plants of 16 hills from the middle two rows, i.e. 0.64 m H= plot area (excluding the border rows) were harvested to assess yield (14% σ + moisture), its components, and related traits in both experiments, fol- lowing Shi et al. (2016). Days to flowering was assessed as the interval 2 2 where σ and σ are the genotypic and residual variances, respect- G E between the date of sowing and the date when panicles of 50% of plants ively, and r is the number of replications. The restricted maximum likeli- per plot were fully exerted. Days to maturity was assessed as the interval hood estimate was used to calculate the variance components in Genstat between the flowering date and date when panicles on most plants in V17.1. The narrow-sense heritability (h ), capturing the proportion of a plot turned yellow and ready for harvest. Plant height was measured total phenotypic variance explained by the additive genetic variance, was from the base of the root–shoot junction to the tip of the flag leaf, which estimated using the equation in Genomic Association and Prediction was manually straightened to be aligned with the culm. Non-grain dry Integrated Tool (GAPIT) function: weight was assessed as the sum of leaf, stem and rachis dry weight. The total aboveground dry weight was the sum of non-grain and grain dry weight. Harvest index was the ratio of grain dry weight to total above- 2 a h= 2 2 ground dry weight. σσ + a e 2 2 where σ is the additive genetic variance and σ is the residual variance. a e Statistical analysis of phenotypic data Analysis of variance Genetic analysis of marker-trait associations A combined linear mixed model based analysis of variance (ANOVA) was performed to test the effect of genotype (G), treatment (T), and year Two hundred and ninety-one genotypes in 2013 and 288 genotypes in (Y) with their interactions using the following model in Genstat V17.1: 2014 had complete phenotypic data. However, 20 genotypes were miss- ing from the 45 699 (46K) SNPs dataset resulting in 271 genotypes in YG =+ μ ++ TY + RT [] () YG +× () TY × + E 2013 and 268 in 2014, used for GWAS analysis. The detailed genotype- ijkli jk lj k ijkl ijk by-sequencing protocol of SNP genotyping, population structure, and linkage disequilibrium (LD) for this population is explained in Kadam where Y is the phenotypic trait value recorded in a plot, µ is the over- ijkl et al. (2017). The GWAS was performed on a set of 271 (2013) and 268 all mean, G is the effect of the ith genotype, T is the effect of the jth i j (2014) genotypes separately, with 267 genotypes being common across treatment, Y is the effect of the kth year, R [T (Y )] is the effect of the lth k l j k both years. Two GWAS methods were used to test the marker–trait asso- replication within the jth treatment of the kth year, (G×T×Y) is the ijk ciations: single-locus and multi-locus analysis. effect of three-way interaction between the ith genotype, the jth treat- Single-locus analysis is a one-dimensional scan, typically identifying ment and the kth year, and E is the error. Apart from the three-way ijkl associations between single markers and traits. We performed this ana- interaction, we also consider two-way interactions of main factors in all lysis using a compressed mixed-linear model (CMLM; Zhang et al., possible combinations. 2010) in GAPIT (Lipka et al., 2012). In the mixed model, we included population structure and family kinship (family relatedness), which were Linear mixed model to estimate best linear unbiased estimators calculated by the GAPIT function using SNPs with ≥0.05 minor allele We estimated the best linear unbiased estimators (BLUEs) of phenotypic frequency (MAF). traits for an individual genotype across years and treatments separately. The following linear mixed model was used in Genstat V17.1 to estimate Y= XQ ++ Ke + αβ μ the BLUEs separately in non-stress and stress conditions across years, using genotypes as a fixed effect and replications as a random effect, where Y represents the vector of phenotype, X represents the vector of SNPs, Q is the PCA matrix and K is the relative kinship matrix. Y= μ+G +R +E ij ij ij X and Q are the fixed effects, and K and e represent random effects. α β μ The Q and K matrices help to reduce the spurious false positive asso- where Y is the phenotypic trait value recorded in a plot, µ is the overall ij ciations. Correction for population structure (Q) substantially reduces mean, G is the effect of the ith genotype, R is the effect of the jth repli- i j the false positives but it sometimes eliminates true positive associations cation, and E is the error. ij due to overcorrection. Therefore, the optimal number of principal Days to flowering had a strong confounding effect on grain yield and components was estimated for each trait before incorporating them its components under stress, particularly in 2013 (Fig. 1C). Therefore, for CMLM tests, based on the forward model selection method using we performed the linear mixed model-based ANOVA using the above the Bayesian information criterion. This method helps to control both equation with days to flowering as covariate. When the effect of days to false-positive and -negative associations more effectively although it flowering was significant on phenotypic traits, corrected BLUEs of trait cannot eliminate both completely. We used a lower suggestive threshold values were estimated in stress treatments. −4 probability P-value 1.0 × 10 (−log P=4) and superior Bonferroni corrected threshold as an upper limit (2013: −log (0.05/45 437)=6; Principal component analysis, trait correlation and multiple 2014: −log (0.05/45 414)=6) to detect significant associations. regression analysis The single-locus analysis corrects the confounding effects of popula- A multivariate principal component analysis (PCA) was performed in tion structure and family kinship but does not consider the confounding XLSTAT across years and treatments. The chart.Correlation() func- effect of causal loci. The multi-locus GWAS is a method that corrects not tion within the R package ‘Performance Analytics’ was used to gen- only the confounding effects of population structure and family kinship erate the correlation scatter plot. The lm() function in R was used for but also the confounding and/or interaction effects of causal loci present Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4021 in the genome due to LD (Segura et al., 2012). We performed the multi- In 2014, we restructured the synchronization based on 2013 locus GWAS using a modified version of the multi-locus mixed linear DS data. This resulted in better synchronization with only small model (MLMM) in R (R script for mlmm.cof.r available at https:// 2 deviation observed from expected days to flowering (r =0.91 cynin.gmi.oeaw.ac.at/home/resources/mlmm). We ran the complete in non-stress and r =0.85 in stress conditions; Fig. 1D). Further, model as recommended with stepwise forward inclusion of the strongest to test the effect of days to flowering, we performed the analy- significant markers (lower P-value) and stepwise backward elimination of the last forward model (that is, least significant markers). Significant mark- sis with days to flowering as a covariate in the mixed model. ers were selected based on the criteria explained by Kadam et al. (2017). The moderate to strong significant effect of days to flowering Briefly, in the first step (like single-locus GWAS without any marker as on yield, its components, and harvest index were detected in a cofactor), we manually checked the P-value of SNPs before including 2013 stress, most likely due to desynchronized flowering time. them as a cofactor in the model. Then we continued adding markers to −4 Conversely, the improved flowering synchronization caused no the model as cofactors based on cut-off threshold P-value≤1.00 × 10 . Once no significant loci appeared below the threshold P-value, the significant effect in 2014 stress. The marginal (P <0.05) to mod- model was stopped. All the significant cofactors identified were consid- erate (P<0.01) effect of days to flowering on yield, seed set, ered as significant loci. and harvest index was detected in both years under non-stress (Fig. 1C, D). This could be due to the pleiotropic effect of flow- Selecting a priori candidate genes underlying the genetic loci ering genes on panicle development (Crowell et al., 2016), a key The detailed protocol to select a priori candidate genes near to significant determinant of rice grain yield. SNPs was followed as explained in Kadam et al. (2017). Genotype effects and genotype-by-environment interactions accounted for variations in Results phenotypic traits The flowering time was sensitive to seasonal climate A combined mixed model ANOVA across years was carried variations out to divide the variation in genotype, treatment and year The flowering time synchronization approach was followed to components and their interactions (Table 1). The variation in reduce the confounding effect of flowering time differences of grain yield, its components, and other related traits differed rice genotypes on grain yield and its components (those meas- significantly between genotypes (G; P<0.001), treatments (T; ured in this study) and related traits under stress (Fig. 1A, B). P<0.001) and years (Y; P<0.01 to P<0.001). Further, the yield, However, we witnessed deviation of our observed days to flow- its component, and related traits of each genotype responded 2 2 ering from expected days (r =0.53 in non-stress and r =0.46 differently to treatment (G×T; P<0.001) and year (G×Y; in stress conditions; Fig. 1C) in 2013. As rice flowering time is P<0.001). The detailed descriptive statistics of these traits are regulated by internal genetic cues and external stimuli such as given in Supplementary Table S1. The traits showed different photoperiod and temperature (Yin et al., 1997), such deviations distributions in non-stress and stress conditions for both years −2 were expected, since the synchronization in 2013 was based on (Fig. 2). Yield ranged from 106.3 to 727.0 g m in non-stress, −2 2012 WS pre-experiment data due to lack of DS data. Many and from 16.7 to 622.6 g m under stress in 2013, and from −2 genotypes exhibited photothermal sensitivity across wet and 102.8 to 839.7 g m in non-stress, and from 78.1 to 761.1 g −2 dry seasons. Therefore, some genotypes experienced stress dur- m under stress conditions in 2014. Across all observations, 2 2 ing the flowering period (31%), whereas others experienced H and h estimates ranged from 0.73 to 0.99 and from 0.27 stress either before (60%) or immediately after flowering (8%). to 0.94, respectively, in 2013; and from 0.62 to 0.99 and from Table 1. Analysis of variance (ANOVA) in 2013 and 2014 dry season experiments of three groups of traits: grain yield, yield components, and other related traits Trait Unit G T Y G×T G×Y T×Y G×T×Y −2 Grain yield g m *** *** *** *** *** *** ** Grain yield component traits 2 −2 Panicles per m m *** *** *** *** *** ns *** Spikelets per panicle — *** *** *** *** *** ns Seed set % *** *** *** *** *** *** *** Thousand grain weight g *** *** ** *** *** ns *** 2 3 −2 Spikelets per m (×10 ) m *** *** *** *** *** *** Other related traits Harvest index — *** *** *** *** *** *** *** −2 Total dry weight kg m *** *** *** *** *** ns ns −2 Non-grain dry weight kg m *** *** *** *** *** *** *** Plant height cm *** *** *** *** *** * ns Days to flowering — *** *** *** *** *** *** *** Days to maturity — — — — — — — — 2 2 Spikelets per m is not an independent yield component but is the product of panicles per m and spikelets per panicle. G, genotype; T, treatment; Y, year. Significance level: *P<0.05, **P<0.01, ***P<0.001; ns, non-significant. Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4022 | Kadam et al. Fig. 2. Box-plot showing phenotypic distribution of grain yield and its components and related traits in non-stress (NS) and water-deficit stress (WD) during 2013 (n=271) and 2014 (n=268). Two-sample t-test P-value shows the significant difference between grain yield (A), its components (B–F), and related traits (G–J) in NS and WD conditions. n, number of genotypes. Inside boxplot, the bold line represents the median, box edges represent upper and lower quantiles, and whiskers are 1.5× the quantile of the data. Outliers are shown as open circles. Values in parentheses represent the significant percentage change (increase (+) or decrease (−)) in WD over NS conditions. Days to maturity across treatments in 2013 and data for non-grain tissue dry weight across treatments and years are given in Supplementary Table 1. The values of phenotypic traits given in the box-plot under 2013 water deficit are the original, not corrected for days to flowering to account variation in flowering synchronization. (This figure is available in colour at JXB online.) 0.69 to 0.93, respectively, in 2014 (Supplementary Table S1). years (Supplementary Figs S4, S5). However, non-significant The greater reduction of yield, seed set, and harvest index (P>0.05) correlations of yield were found with thousand grain under stress in 2014 was due to higher stress intensity during weight and non-grain dry weight in non-stress, and with pani- 2014 (−64 kPa) compared with 2013 (−46 kPa), driven by cle number in 2013 stress. Yield was not significantly (P>0.05) higher vapour-pressure deficit (Supplementary Figs S2A, S3D). correlated with non-grain dry weight across treatments in However, a higher reduction of spikelets per panicle and spike- 2014. The correlation of yield with spikelets per panicle was lets per m despite lower stress intensity was observed during higher in stress (2013: r=0.73; 2014: r=0.46) than in non-stress 2013 than during 2014 (Fig. 2C, E). This could be due to vari- conditions (2013: r=0.40; 2014: r=0.36) in both years, and ation in flowering time synchronization with more genotypes the increase was stronger in 2013. Similarly, the correlation experiencing stress before flowering in 2013 than in 2014. between yield and seed set increased from 0.62 in non-stress These results clearly illustrate that stress affects the number of to 0.75 in stress conditions in 2014. The increased correlation spikelets per m when imposed before flowering, but spike- of yield with spikelets per panicle in 2013 and with seed set let fertility when imposed during flowering (Lanceras et al., in 2014 in stress conditions reflects the effect of variation in 2004), as shown in Fig. 2C, E. The days to flowering differed days to flowering synchronization. The correlation of yield significantly (P=0.002) between non-stress and stress in 2013, with days to flowering was higher under stress (r=0.29) than but not (P>0.05) in 2014 (Fig. 2I). under non-stress conditions (r=0.16) in 2013, but was almost The first two principal components cumulatively explained the same (r=0.30) for both treatments in 2014. >55% in 2013 and >61% in 2014 of the total phenotypic vari- We also tested the relative contribution of each component ation across treatments (Fig. 3). The genotypic variation in and related trait to yield through multiple linear regression. All the first PC was mostly explained by yield, harvest index and the components and related traits significantly contributed to spikelets per m in non-stress and yield, harvest index, spikelets yield except for plant height and days to flowering in non- per m and total dry weight under stress in 2013 and 2014. The stress in 2013 and days to flowering in stress conditions during genotypic variation in the second PC was explained by non- 2014 (Supplementary Table S2). grain dry weight, days to flowering, and total dry weight under non-stress, and plant height, non-grain dry weight, and days to Treatment and year specific genetic loci for flowering under stress in 2013 and 2014. In addition, the prin- phenotypic traits cipal component variations for the phenotypic traits differed in Grain yield and its components and related traits followed a response to treatment and year (Fig. 3). This further confirms normal distribution (Supplementary Figs S4, S5), indicating the strong G×T and G×Y interactions. the quantitative pattern suitable for genetic analysis. A sum- mary of GWAS results using single-locus and multi-locus Phenotypic trait correlations and contribution of analysis methods is given in Table 2. The detailed results are component traits to grain yield in Supplementary Tables S3–S6. In total, we identified 38 Grain yield was significantly (P<0.05) correlated with most significant loci in non-stress conditions, and 69 loci in stress of its components and related traits across treatments and conditions during 2013, and 64 significant loci in non-stress Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4023 Biplot (PC1 and PC2: 55.61 %) Biplot (PC1 and PC2: 59.59 %) 2013:NS 2013:WD PH NGDW NGDW DTF DTF TDW 4 PH TDW SPP 2 TGW SP DTM TGW GY SPP SS DTM -2 -2 GY PN HI SS SP -4 -4 PN HI -6 -6 -8 -8 -10-8-6-4-2 02468 10 -10-8-6-4-20 24 68 10 PC1 (29.09 %) PC1 (35.51 %) Biplot (PC1 and PC2: 61.26 %) Biplot (PC1 and PC2: 63.66 %) 2014:NS 2014:WD NGDW DTF TDW NGDW DTF PH PH TDW SPP SP 2 SPP GY TGW P… SP 0 TGW SS GY SS -2 -2 HI PN HI -4 -4 -6 -6 -8 -8 -10-8-6-4-20 24 68 10 -10-8-6-4-20 2468 10 PC1 (34.69 %) PC1 (37.54 %) Fig. 3. The principal component analysis of grain yield, its components, and related traits with first two principal components (PC1 and PC2) in non- stress (NS) (A, C) and water-deficit stress (WD) (B, D) during 2013 (A, B) and 2014 (C, D) DS. The traits marked inside the solid circle/ellipses contributed more to the variation explained by PC1 and those marked inside the dashed ellipses to PC2. DTF, days to flowering; DTM, days to maturity; GY, grain 2 2 yield; HI, harvest index; NGDW, non-grain dry weight; PH, plant height; PN, panicles per m ; SP, spikelets per m ; SPP, spikelets per panicle; SS, seed set; TDW, total dry weight; TGW, thousand grain weight. (This figure is available in colour at JXB online.) conditions, and 55 loci in stress conditions during 2014. Most when corrected trait values were subjected to GWAS analy- loci were specific across treatments within years and within sis. This suggests that the trait variations associated with these treatments across the years. Nevertheless, we also detected 14 loci were mostly explained by variation in days to flowering. common loci (nine in 2013 and five in 2014) across treatments Only five genetic loci (one on chromosome 4 for yield (Q9); and eight common loci within treatments (six in non-stress one on chromosome 12 for spikelets per m (141 599) and and two in stress conditions) across years for the same compo- three loci on chromosome 11 for harvest index (10 627 944, nents and related traits (Supplementary Table S7). 10 131 062, 10 329 677) were common to corrected and non- corrected trait values (Supplementary Tables S4 and S8). The common (corrected vs non-corrected) loci detected for yield Genetic analysis after correcting for days to flowering (Q9; Table 3; Fig. 4) and harvest index (Supplementary Fig. S6; under stress conditions in 2013 Supplementary Tables S4, S8) recorded lower P-values for cor- Flowering time synchronization was strongly confounding the rected than the non-corrected trait value through single locus grain yield and its component traits in 2013 stress conditions analysis. Despite correction, the novel locus Q10 on chromo- (Fig. 1C, D). We corrected for yield, yield components, and some 3 for corrected yield, seed set, and harvest index over- other related traits (only harvest index in this group) using lapped with days to flowering (Table 3). In summary, statistical days to flowering as a covariate in the mixed model. The sin- correction helped to explain the confounding effect of days to gle and multi-locus analysis of corrected trait values identi- flowering and to some extent helped to eliminate its effect on fied 31 additional loci using similar threshold P-values as yield under water deficit. Unless otherwise mentioned, all the mentioned earlier (Table 2; Supplementary Table S8). Most mapping results discussed in the following sections were for genetic loci detected for non-corrected traits disappeared the corrected trait loci under 2013 stress. PC2 (26.56 %) PC2 (26.52 %) PC2 (26.12 %) PC2 (24.08 %) Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4024 | Kadam et al. Table 2. Summary of genetic loci detected in 2013 and 2014 Seven grain-yield loci revealed a small to medium under non-stress (NS) and water-deficit stress (WD) conditions allelic effect in response to reproductive-stage water for three groups of traits: grain yield, yield components, and other deficit related traits We identified two loci (Q9 and Q10) for grain yield under Traits 2013 2014 stress in 2013. The minor allele of both these loci had a negative effect on yield. Five significant loci Q11–Q15 were detected NS WD WD NS WD for yield under stress in 2014 (Fig. 5). The minor allele of Q11, Grain yield 2 4 2 6 5 Q12, and Q15 had a positive effect on yield, while the minor Grain yield component traits 2 allele of two loci, Q13 and Q14, had a negative effect on yield. Panicles per m 6 12 7 9 3 Q9 and Q10 harboured 18 and Q11–Q15 harboured 16 a Spikelets per panicle 5 9 6 2 na priori candidate genes within the expected LD block region Seed set 3 7 7 8 11 Thousand grain weight 3 4 na 6 8 (Table 3). Seven a priori candidate genes, mostly near signifi- Spikelets per m 1 4 1 4 3 cant SNPs, are given in Supplementary Table S9. The Q9 locus Subtotal 18 36 21 29 25 was close (13 kb) to the phosphomannomutase gene regulat- Other related traits ing L-ascorbic acid biosynthesis and response to abiotic stress Harvest index 6 7 8 4 2 stimulus (Gene Ontology (GO):0009628). L-Ascorbic acid Total dry weight 3 2 — 4 11 acts as a redox buffer to detoxify reactive oxygen species (ROS) Non-grain dry weight 3 3 — 5 2 (Arrigoni and De Tullio, 2002). Q11 was close to squalene Plant height 2 6 — 6 4 monooxygenase or epoxidase (16 and 23 kb; two copies in LD Days to flowering 3 11 — 10 6 block) and response to abiotic stress stimulus (GO:0009628). Days to maturity 1 na — — — This gene is known to regulate ROS, stomatal responses and Subtotal 18 29 8 29 25 Total 38 69 31 64 55 water-deficit tolerance in Arabidopsis (Posé et al., 2009). na, no marker trait association analysis performed Only three loci for grain yield acted via change in seed Marker-trait associations detected for corrected trait values in water- deficit stress (see the text for the correction method). set percentage Although rice grain yield is co-determined by panicle num- Eight grain-yield loci revealed small to medium allelic ber, spikelets per panicle, seed set percentage, and grain weight, effect in non-stress conditions very few loci of these component traits co-located with loci for We identified two (Q1 and Q2) and six (Q3–Q8) loci for grain yield per se. The seed set percentage is one of the most important yield in 2013 and 2014, respectively (Table 3). There were no yield components as indicated by its strong correlation with yield common loci across years, most likely due to significant vari- (Supplementary Figs S4, S5). Three loci were regulating yield ations in temperature (minimum and maximum) and vapour- through changes in seed set percentage, i.e. two loci designated as pressure deficit (VPD; Supplementary Fig. S3). These loci had Q2 (2013) and Q7 (2014) in non-stress, and Q10 (2013) in stress a positive or negative effect (small to medium) on yield with conditions. The major allele (allele refers to the 0.95 frequency regard to its minor allele (allele refers to the 0.05 frequency in in the studied population) of these loci had a respective positive the studied population). In 2013, the minor allele of Q1 had a effect on yield, seed set, and harvest index (Fig. 6). In addition, the positive effect on yield. Conversely, the minor allele of Q2 had Q10 was also detected for days to flowering. No loci were com- a negative effect on yield. In 2014, the minor allele of Q3, Q5, mon for yield and seed set in 2014 stress conditions, but one of and Q6 had a positive effect, while the minor allele of Q4, Q7, the loci on chromosome 1 (29 223 354) was commonly detected and Q8 had a negative effect on yield (Table 3). for seed set and harvest index (Supplementary Fig. S7). Similarly, Eighteen and sixty-eight a priori (known or characterized) the major alleles had a respective positive effect on seed set, har- candidate genes were harboured within the expected LD vest index, and yield (irrespective of genetic significance) (Fig. 7). block by Q1 and Q2 in 2013, and Q3–Q8 in 2014, respect- Hence, these loci were regulating yield through the effect of seed ively. Interestingly eight a priori candidate genes were identi- set on harvest index. fied (Supplementary Table S9). Q1 was close to OsPTR2 (6 Four a priori candidate genes were predicted within the and 31 kb; two copies in LD block). The rice homologue of expected LD block of these loci. The Q2 was close (55 kb this gene, short panicle 1 (OsPTR2), regulates panicle and grain from peak SNP) to the plastocyanin gene that regulates flower size and nitrate transport (Li et al., 2009). The homologue of development (GO:0009908) and pollination (GO:0009856) in OsPTR2 was recently detected at the q-28 locus (OsPTR9) rice (Supplementary Table S9). The Arabidopsis orthologue of for spikelet number per panicle (a key determinant of grain this gene regulates seed set and pollen tube growth (Dong et al., yield) in the same rice association panel as that used in this 2005). Q7 was within the novel expressed protein, which pro- study (Rebolledo et al., 2016). Likewise, Q4 was close (34 kb vides an entry point for future study. Sugar transport or uptake from peak SNP) to serine–threonine kinase (OsSTE). The is essential for normal pollen development (Reinders, 2016), Arabidopsis orthologue of OsSTE (AtSTE or BLUS1) is the while the lack of starch synthesis arrests the pollen develop- major regulator of stomatal opening (Takemiya et al., 2013; ment in water deficit conditions thereby regulating seed set Supplementary Table S9). (Sheoran and Saini, 1996). Our Q10 locus was within the sugar Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4025 Table 3. GWAS results for final set of genetic loci detected for grain yield in non-stress and water-deficit stress conditions during 2013 and 2014. Detailed GWAS results for yield components and related traits across treatments and years are given in Supplementary Tables S3–S6 and S8 a b c d e f g Treatment Year Locus SNP pos Chr Allele MAF P P AE LD block Size (kb) Known CMLM MLMM −2 h (mean grain name (g m ) genes Start End −2 yield (g m )) −5 Non-stress 2013 (451.1) Q1 10 101 900 11 C:G 0.336 2.72 × 10 — 30.13 10 101 900 10 173 685 71 2 −8 Q2 30 523 925 2 G:A 0.070 — 5.78 × 10 −175.90 30 397 910 30 541 202 143 16 −5 2014 (521.9) Q3 13 199 901 12 C:T 0.468 6.84 × 10 — 30.04 12 917 853 13 298 195 380 8 k −5 −6 Q4 26 796 595 3 C:T 0.097 9.91 × 10 2.20 × 10 −46.18 26 756 997 26 978 105 221 9 l −5 Q5 29 142 398 2 C:A 0.179 — 4.19 × 10 13.98 29 122 557 29 261 158 138 9 l −6 Q6 19 367 031 10 T:G 0.466 — 1.75 × 10 74.08 19 280 939 19 474 522 193 21 l −5 Q7 5 105 627 12 A:C 0.078 — 3.03 × 10 −186.24 5 101 105 539 0949 289 12 l −5 Q8 42 643 337 1 A:G 0.347 — 6.58 × 10 −97.80 42 587 683 42 643 699 56 9 i −5 −6 Water deficit 2013 (317.3) Q9 34 815 277 4 C:T 0.074 1.17 × 10 1.77 × 10 −81.29 34 815 277 34 833 179 17 5 −6 −6 — — — — — 1.29 × 10 3.05 × 10 j k −5 −6 Q10 5 113 428 3 T:C 0.424 3.55 × 10 5.17 × 10 −40.61 5 021 158 5 167 439.00 146 13 k −5 −6 2014 (319.5) Q11 6 934 188 3 A:G 0.397 8.26 × 10 8.73 × 10 31.47 6 908 684 7 020 707 112 7 l −7 Q12 42 144 827 1 T:C 0.366 — 1.86 × 10 6.10 42 123 552 42 144 993 21 2 l −6 Q13 16 038 003 10 T:C 0.358 — 5.46 × 10 −49.86 16 024 382 16 110 372 85 6 l −5 Q14 23 005 301 11 G:A 0.276 — 1.12 × 10 −23.54 22 976 390 23 005 386 28 0 l −5 Q15 27 115 652 11 G:A 0.075 — 4.92 × 10 33.65 27 115 609 27 123 090 7 1 Single nucleotide polymorphism (SNP) position. Chromosome. Minor allele frequency (MAF). P-value of single-locus compressed mixed linear model (CMLM). P-value of multi-locus mixed model (MLMM). Allelic effect with respect to minor allele=(average traits value of genotypes carrying minor allele−average traits value of genotypes carrying major allele). Linkage disequilibrium block. Total number of known characterized genes in LD block. Genetic locus detected for non-corrected and corrected grain yield value. Genetic locus detected for corrected grain yield value and coinciding with days to flowering. Genetic locus detected through CMLM and MLMM methods. Genetic locus detected through MLMM method only. All the unmarked loci were detected through CMLM method. The italic P-value is for corrected grain yield value. The genetic locus marked in bold (Q2) overlaps with panicle weight (equivalent to grain yield) from Kikuchi et al. (2017) (Supplementary Table S10). transporter gene that plays an important role in sugar distri- and grain yield in reproductive-stage stress (Pinto et al., 2010). bution. The rice grain yield MQTL (meta-analysis QTL) Our genetic analysis of statistically corrected trait values was 2.1 detected in water-deficit conditions also contained the sugar effective in minimizing the effect of desynchronized flower- transporter gene (Swamy et al., 2011). Similarly, the locus on ing time, as it led to detection of several novel loci that were chromosome 1 for seed set and harvest index in 2014 stress was not detected for non-corrected trait values. Despite statistical near (34 kb from peak SNP) to the nitrate transporter gene adjustment for flowering time, our novel Q10 for grain yield that plays a role in rice yield increment (Fan et al., 2016). was co-localized with flowering time (different SNPs but fall- ing within the same gene and LD block). In addition, it was also co-localized with seed set and harvest index. Previous Discussion studies in rice have identified several grain yield QTLs using linkage mapping under reproductive water-deficit stress con- The main aim of this study was to link phenotypic variation ditions (Bernier et al., 2007; Venuprasad et al., 2009; Vikram with genetic markers, thereby gaining insights about promis- et al., 2011; Swamy et al., 2013; Mishra et al., 2013), of which ing candidate genes and the genetic architecture controlling some co-localized with plant height (qDTY ), days to flow- yield traits. To the best of our knowledge, this is the first study 6.2 ering (qDTY ), or both (qDTY ). Interestingly, the major conducted on the rice PRAY association mapping panel under 3.2 1.1 effect of qDTY was consistent even after statistical correc- reproductive-stage water-deficit stress. The key findings from 1.1 tion of grain yield using flowering time and plant height as our study are discussed below. covariates (Vikram et al., 2011), and the recent detailed char- acterization confirmed the tight linkage and not the pleiot- Statistical trait adjustment can reduce confounding ropy of this QTL with plant phenology (Vikram et al., 2015). effect of desynchronized flowering on genetic analysis Our novel Q10 provided higher confidence of a causa- under water deficit tive SNP placed directly within the sugar transporter gene. The desynchronized flowering time may result in the identi- However, this SNP was just 5 kb away from the COP9 signa- fication of QTLs, often colocating with QTLs for phenology losome complex subunit 4 gene within the same LD block Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4026 | Kadam et al. Fig. 4. (A) GWAS results (Manhattan and quantile–quantile plot) detected through single-locus compressed mixed linear model (CMLM) and multi-locus mixed model (MLMM) for non-corrected and corrected (using days to flowering as covariate) grain yield in 2013 water-deficit stress (WD) conditions. Significant SNPs in the Manhattan plot of MLMM are numbered according to the order in which they were included as a cofactor in the regression model. 2 2 (B) Identified LD block (17 kb) based on r value between SNPs on chromosome 4 and the colour intensity of the box on the LD plot corresponds with r (multiplied by 100) according to legend. Significant SNP marked by first rectangle was detected by CMLM and MLMM and the next three rectangles only by CMLM approach. (This figure is available in colour at JXB online.) (Supplementary Table S9). The COP9 signalosome complex with flowering time and stress tolerance to test linkage versus gene is known to regulate flower development in Arabidopsis pleiotropy. Nevertheless, the effect of our consistent Q9 for (Wang et al., 2003), although the role of this gene in rice flow- grain yield (detected using either corrected or non-corrected ering has not been reported. Therefore, a further characteriza- values) was independent of flowering time stress conditions. tion of Q10 would be interesting to decipher the relationship More precise flowering time synchronization in 2014, which Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4027 Fig. 5. (A) GWAS results (Manhattan and quantile–quantile plot) detected through single-locus compressed mixed linear model (CMLM) and multi- locus mixed model (MLMM) for grain yield in 2014 water-deficit stress (WD) conditions. Significant SNPs on the Manhattan plot of MLMM are numbered according to the order in which they included as a cofactor in regression model. (B) Identified LD block (112 kb) based on r value between SNPs on chromosome 3 and the colour intensity of the box on the LD plot corresponds with r (multiplied by 100) according to the legend. Significant SNP marked by a rectangle was detected by CMLM and MLMM. (This figure is available in colour at JXB online.) allowed identification of the genetic loci without having any key yield components in cereals that are genetically less com- co-localization with flowering time in stress conditions, added plex than yield per se (Yin et al., 2002). In rice, grain yield is the value to the findings. To the best of our knowledge, this is the product of the panicle number or productive tiller (determined first report demonstrating the effectiveness of better synchro- during the vegetative phase), spikelets per panicle (determined nization of flowering time phenology on a large GWAS panel during panicle initiation), seed set percentage (determined dur- under stress conditions at field level. ing gametogenesis and anthesis), and individual grain weight (determined during grain filling). The genetic selection for each of these traits during rice domestication has given rise to rich Genetic control of grain yield, its components, genetic diversity (Doebley et al., 2006; Sweeney and McCouch, and related traits was mostly independent and 2007). To date, molecular genetic studies have detected QTLs environment-specific underlying these genetic changes in rice yield components Grain yield is a complex trait determined by many interactive (http://www.gramene.org/). From these QTLs some of the physiological processes changing temporally during the grow- candidate genes were successfully identified, notably display- ing period. These processes often match the development of the ing improvement in grain yield (Ashikari et al., 2005; Fan et al., Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4028 | Kadam et al. B C Q2 P=5.59E-10 80 0.45 P=4.15E-12 Q2 Q2 P=1.42E-13 75 0.40 0.35 0.30 0.25 0.20 150 0.15 GG (n=252)AA (n=19) GG (n=252)AA (n=19) GG (n=252)AA (n=19) D E F 600 0.50 80 P=1.03E-06 Q7 Q7 Q7 P=3.90E-09 P=4.04E-16 550 75 0.45 0.40 400 0.35 0.30 0.25 40 0.20 AA (n=247)CC (n=21) AA (n=247)CC (n=21) AA (n=247)CC (n=21) G H P=0.009 360 Q10 P=0.14 Q10 P=0.032 75 Q10 0.35 0.34 70 0.33 0.32 300 65 0.31 0.30 60 0.29 0.28 240 55 0.27 TT (n=156)CC (n=115) TT (n=156)CC (n=115) TT (n=156)CC (n=115) Fig. 6. Allelic effect of Q2 (A–C; 2013), Q7 (D–F; 2014) in non-stress and Q10 (G–I; 2013) in water-deficit stress conditions on grain yield, seed set, and harvest index. Allelic effect of Q7 on harvest index was significant regardless of GWAS significance. Two-sample t-test P-value shows significant allelic effect difference with reference to major and minor allele. The Q10 locus also coincided with days to flowering. A B C 400 60 0.35 SNP Pos: 29223354 SNP Pos: 29223354 SNP Pos: 29223354 0.33 P=3.50E-10 P=9.90E-11 P=1.03E-06 350 55 0.31 0.29 300 50 0.27 250 0.25 0.23 200 40 0.21 0.19 150 35 0.17 30 0.15 GG (n=224) CC (n=44) GG (n=224)CC (n=44) GG (n=224)CC (n=44) Fig. 7. Allelic effect of chromosome 1 locus (29 223 354) on grain yield (A), seed set (B), and harvest index (C) in 2014 water-deficit stress conditions. Allelic effect on grain yield was significant regardless of GWAS significance. Two-sample t-test P-value shows significant allelic effect difference with reference to major and minor allele. -2 -2 -2 Grain yield (g m ) Grain yield (g m ) Grain yield (g m ) -2 Grain yield (g m ) Seed set (%) Seed set (%) Seed set (%) Seed set (%) Harvest index Harvest index Harvest index Harvest index Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4029 2006; Song et al., 2007; Shomura et al., 2008; Huang et al., 2009; when comparing with other studies using different mapping Miura et al., 2010). For instance, the SPIKE gene/allele regu- panels under reproductive-stage water deficit (Ma et al., 2016; lating the spikelet numbers indicated 13–36% yield increment Pantalião et al., 2016; Swamy et al., 2017). The major reasons in rice (Fujita et al., 2013). In the present study, genetic dis- for this were different rice genotypes or population size and section of these yield components enabled us to detect more inherent environmental and field variation for stress treatment loci than yield per se that were directly or indirectly contribut- (QTL×environment interaction). Another possible reason ing to rice grain yield. The co-localization of grain yield loci could be use of indica subspecies genotypes in this study while with yield components was limited in this study compared with previous studies either used japonica subspecies (Pantalião other studies in rice (Lanceras et al., 2004). This could be due et al., 2016) or small population size (75 genotypes) with sim- to compensation among the yield components. In addition, ple sequence repeat markers (Swamy et al., 2017). In addition, these results emphasize the need for genetic analysis of yield it can be difficult to identify genomic regions or genes deter- components to identify additional genetic determinants having mining the trait difference across subspecies or genotypes. indirect effect on grain yield, providing alternative routes to enhance yield under water deficit. Seed set regulates the assimilate partitioning and Except for one locus on chromosome 12 for spikelets per m grain yield in 2014, the majority of the loci for grain yield and its com- Better optimization of assimilate partitioning to reproductive ponent traits were specific to non-stress or stress conditions in organs with minimal competition among reproductive organs both years. These results are in agreement with previous studies is essential to achieve stable and higher grain yield. So far, the in rice (Lanceras et al., 2004; Vikram et al., 2011; Kumar et al., physiological and genetic basis of the above processes have been 2014) and other crop species (Yin et al., 2002; Millet et al., 2016). poorly understood in rice and other cereal crops. Our study Hence, the greater dependence on environments appeared to showed that the co-localization of grain yield loci with its com- be a common characteristic of QTLs, although this does not ponents was rare. However, four genetic loci, namely Q2 and negate their importance in marker-assisted selection. Despite Q7 in non-stress, and Q10 and 29 223 354 (SNP position) in strong variation in weather, we also detected four consistent stress conditions, were regulating the grain yield and harvest loci: one each for panicles per m and spikelets per panicle on index through changes in the seed set (Figs 6, 7). This indicates chromosomes 10 (19 903 199) and 4 (23 423 399), respectively, that the seed set is a critical determinant of assimilate partition- and two loci on chromosomes 2 (30 699 332) and 5 (5 366 489) ing (harvest index), thereby regulating the final expression of for thousand-grain weight across years in non-stress condi- grain yield. A recent GWAS analysis confirmed these interac- tions (Supplementary Table S7). These consistent regions with tions in wheat (Guo et al., 2017). Hence, these identified loci favourable alleles could be used for improving yield potential. could be pyramided into an ‘ideotype’ at genomic level through marker-assisted selection to enhance rice grain yield in non- Few overlaps of genetic loci with previously identified stress and stress conditions. In addition, such loci could also be markers using same diversity panel of interest in identifying the physiological and molecular basis The PRAY population has been previously used in GWAS of assimilate partitioning to reproductive organs. for a range of phenotypic traits (Qiu et al., 2015; Al-Tamimi et al., 2016; Rebolledo et al., 2016; Kikuchi et al., 2017; Kadam Promising a priori candidate genes for grain yield and et al., 2017). When comparing our results with these previ- water-deficit stress resilience ous studies, we could not find any overlap between significant We detected a priori candidate genes near peak SNP(s) within markers, except a SNP marker detected for plant height (posi- the LD block for grain yield loci (Supplementar y Table S9). A tion: 38 286 772) on chromosome 1, which was detected in our priori candidate genes of grain yield loci can indicate possible previous study (Kadam et al., 2017). The most likely reasons for roles of underlying physiological (SET kinase, sugar and nitrate this lack of co-localization are difference in type and timing of transporter genes) and reproductive developmental (plasto- stress or growing environments (QTL×environment interac- cyanin gene) processes in regulating the grain yield. Likewise, tion), population size, and molecular marker data used by pre- the abiotic stress tolerance candidate genes were detected vious studies or novel GWAS analysis methods (multi-locus) near to grain yield loci in water-deficit conditions, of which that are used in this study. Therefore, to make a more logical genes regulating the detoxification of ROS (phosphoman- comparison for the same traits, we reanalysed the number of nomutase and squalene epoxidase genes) seem to be critical spikelets per panicle from Rebolledo et al. (2016) and yield in rice stress tolerance (Selote and Chopra, 2004; Pyngrope and yield components from Kikuchi et al. (2017), using the et al., 2013). These candidate genes need to be considered to same SNP datasets and analysis methods that are used in this detect the most likely causal genes. However, detailed large- study. This comparative analysis identified one locus on chro- scale molecular validations need to be conducted using the mosome 2 (30 518 548) for panicle weight (equivalent to grain available approaches of RNAi, knockout mutants transgenic yield) from Kikuchi et al. (2017) that overlapped with grain overexpression, and gene editing. Similarly, the loci for com- yield locus (Q2: 30 523 925; different SNP but falls within ponents and related traits that were not co-localized with the same LD block; Table 3; Supplementary Table S10) from yield per se could also be interesting candidates for further 2013 non-stress conditions. In addition, there was also no over- identification of novel genes. lap of a significant marker for grain yield and its components Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4030 | Kadam et al. water-deficit stress conditions using compressed mixed linear- Concluding remarks model and multi-locus mixed model methods. This study provides novel genetic loci for rice grain yield, its Table S5. The details of genetic loci detected for grain yield components, and related traits under non-stress and stress con- components and related traits in 2014 non-stress conditions ditions in field phenotyping experiments. We detected several using compressed mixed linear-model and multi-locus mixed favourable alleles regulating these traits that, upon validation, model methods. can be effectively used in improving yield. Additional genetic Table S6. The details of genetic loci detected for grain yield loci with less overlap of yield component traits to yield per se components and related traits in 2014 water-deficit stress con- clearly indicate the independent genetic regulation of these ditions using compressed mixed linear-model and multi-locus traits. Thus, many loci for component traits had an indirect mixed model methods. effect on yield, which cannot be detected while mapping yield Table S7. Common genetic loci detected across treatments directly. This indicates the complexity of yield as a trait despite (non-stress versus water-deficit stress) in 2013 or 2014 (A). moderate to high heritability, which is often used as a selec- Similarly, common genetic loci detected across years (2013 tion criterion to improve yield potential and stress tolerance. versus 2014) in NS or WD conditions (B). Hence, future studies should also explore the genetic basis of Table S8. The details of genetic loci detected for corrected individual component traits that are genetically less complex— grain yield components and related traits (only on harvest an approach expected to give additional useful information to index excluding the other traits in this group) in 2013 water- further enhance yield. deficit stress conditions using compressed mixed linear-model and multi-locus mixed model methods. Table S9. The list of a priori candidate genes within the link- Supplementary data age disequilibrium block of GWAS significant peak SNP/loci for grain yield in non-stress and water-deficit stress conditions. Supplementary data are available at JXB online. Table S10. The details of genetic loci detected from previ- Fig. S1. Field set-up of 296 genotypes screened under non- ously published data on grain yield and yield components from stress and reproductive-stage water-deficit stress in 2013 and Kikuchi et al. (2017), and number of spikelets per panicle (a 2014 experiments. key yield component) from Rebolledo et al. (2016), using the Fig. S2. Soil moisture tension measured using tensiometers same rice PRAY panel. in water-deficit stress field during 2013 and 2014, and rainfall pattern measured during stress period in 2013 and 2014. Fig. S3. Climate parameters observed during the growing Acknowledgements period. This work was supported by an anonymous private donor who pro- Fig. S4. Pearson correlation coefficient between grain yield vided the financial support, via Wageningen University Fund, to the and its components and related traits in 2013 non-stress and first author’s PhD fellowship. We also thank The Federal Ministry for water-deficit stress conditions. Economic Cooperation and Development, Germany, and the USAID- Fig. S5. Pearson correlation coefficient between grain yield Bill & Melinda Gates Foundation for their financial support. We also thank the GRiSP (Global Rice Science Partnerships; now renamed and its components and related traits in 2014 non-stress and to RICE CRP consortium) for establishing the PRAY Global water-deficit stress. Phenotyping Network. Dr C. G. van der Linden and Dr P. S. Bindraban Fig. S6. GWAS results (Manhattan and quantile–quantile are acknowledged for their valuable advice. plot) detected through single-locus compressed mixed linear model and multi-locus mixed model for non-corrected and corrected harvest index (using days to flowering as a covariate) Author contributions in 2013 water-deficit stress conditions. XY, PCS, and SVKJ conceived the project and its components; NNK Fig. S7. GWAS results (Manhattan and quantile–quantile and SVKJ implemented the experiment; NNK performed the pheno- plot) detected through single-locus compressed mixed linear typing; NKK performed the GWAS including both the conventional and multi-locus approach; NNK drafted the figures, tables, and manu- model and multi-locus mixed model for seed-set and harvest script; MCR provided data obtained from the same panel for compara- index in 2014 water-deficit stress conditions. tive GWAS analysis; XY, SVKJ, and PCS supervised the data processing Table S1. Summary statistics of grain yield and its compo- and the preparation of the drafts; NNK, MCR, SVKJ, XY, and PCS nents and related traits in 2013 and 2014 non-stress and water- interpreted the data and wrote the final paper. The authors declare no deficit stress conditions. competing financial interests. Table S2. Multiple linear regression of grain yield with its components and related traits in non-stress and water-deficit References stress conditions during 2013 and 2014. Table S3. The details of genetic loci detected for grain yield Agrama HA, Eizenga GC, Yan W. 2007. Association mapping of yield and its components in rice cultivars. Molecular Breeding 19, 341–356. components and related traits in 2013 non-stress conditions Al-Tamimi N, Brien C, Oakey H, Berger B, Saade S, Ho YS, Schmöckel using compressed mixed linear-model and multi-locus mixed SM, Tester M, Negrão S. 2016. Salinity tolerance loci revealed in rice using model methods. high-throughput non-invasive phenotyping. Nature Communications 7, 13342. Table S4. The details of genetic loci detected for uncor- Arrigoni O, De Tullio MC. 2002. Ascorbic acid: much more than just an rected grain yield, its components, and related traits in 2013 antioxidant. Biochimica et Biophysica Acta 1569, 1–9. Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 Genetic regulation of rice grain yield and its components | 4031 Ashikari M, Sakakibara H, Lin S, Yamamoto T, Takashi T, Nishimura Genome-wide association mapping for phenotypic plasticity in rice. Plant, A, Angeles ER, Qian Q, Kitano H, Matsuoka M. 2005. Cytokinin oxidase Cell & Environment 40, 1565–1575. regulates rice grain production. Science 309, 741–745. Kumar A, Dixit S, Ram T, Yadaw RB, Mishra KK, Mandal NP. 2014. Barnabás B, Jäger K, Fehér A. 2008. The effect of drought and heat Breeding high-yielding drought-tolerant rice: genetic variations and stress on reproductive processes in cereals. Plant, Cell & Environment 31, conventional and molecular approaches. Journal of Experimental Botany 11–38. 65, 6265–6278. Begum H, Spindel JE, Lalusin A, Borromeo T, Gregorio G, Hernandez Lanceras JC, Pantuwan G, Jongdee B, Toojinda T. 2004. Quantitative J, Virk P, Collard B, McCouch SR. 2015. Genome-wide association trait loci associated with drought tolerance at reproductive stage in rice. mapping for yield and other agronomic traits in an elite breeding population Plant Physiology 135, 384–399. of tropical rice (Oryza sativa). PLoS ONE 10, e0119873. Li S, Qian Q, Fu Z, et al. 2009. Short panicle1 encodes a putative PTR Bernier J, Kumar A, Ramaiah V, Spaner D, Atlin G. 2007. A large-effect family transporter and determines rice panicle size. The Plant Journal 58, QTL for grain yield under reproductive-stage drought stress in upland rice. 592–605. Crop Science 47, 507–516. Lipka AE, Tian F, Wang Q, Peiffer J, Li M, Bradbury PJ, Gore MA, Borba TCdO, Brondani RPV, Breseghello F, Coelho ASG, Mendonça Buckler ES, Zhang Z. 2012. GAPIT: genome association and prediction JA, Rangel PHN, Brondani C. 2010. Association mapping for yield and integrated tool. Bioinformatics 28, 2397–2399. grain quality traits in rice (Oryza sativa L.). Genetics and Molecular Biology Liu JX, Liao DQ, Oane R, Estenor L, Yang XE, Li ZC, Bennett J. 2006. 33, 515–524. Genetic variation in the sensitivity of anther dehiscence to drought stress in Bouman BAM, Peng S, Castañeda AR, Visperas RM. 2005. Yield rice. Field Crops Research 97, 87–100. and water use of irrigated tropical aerobic rice systems. Agricultural Water Ma X, Feng F, Wei H, Mei H, Xu K, Chen S, Li T, Liang X, Liu H, Luo Management 74, 87–105. L. 2016. Genome-wide association study for plant height and grain yield in Crowell S, Korniliev P, Falcão A, Ismail A, Gregorio G, Mezey rice under contrasting moisture regimes. Frontiers in Plant Science 7, 1801. J, McCouch S. 2016. Genome-wide association and high-resolution Millet EJ, Welcker C, Kruijer W, et al. 2016. Genome-wide analysis of phenotyping link Oryza sativa panicle traits to numerous trait-specific QTL yield in Europe: allelic effects vary with drought and heat scenarios. Plant clusters. Nature Communications 7, 10527. Physiology 172, 749–764. Dixit S, Singh A, Sta Cruz MT, Maturan PT, Amante M, Kumar A. 2014. Mishra KK, Vikram P, Yadaw RB, Swamy BP, Dixit S, Cruz MT, Maturan Multiple major QTL lead to stable yield performance of rice cultivars across P, Marker S, Kumar A. 2013. qDTY12.1: a locus with a consistent effect varying drought intensities. BMC Genetics 15, 16. on grain yield under drought in rice. BMC Genetics 14, 12. Doebley JF, Gaut BS, Smith BD. 2006. The molecular genetics of crop Miura K, Ikeda M, Matsubara A, Song XJ, Ito M, Asano K, Matsuoka domestication. Cell 127, 1309–1321. M, Kitano H, Ashikari M. 2010. OsSPL14 promotes panicle branching Donald CM. 1968. The breeding of crop ideotypes. Euphytica 17, 385–403. and higher grain productivity in rice. Nature Genetics 42, 545–549. Dong J, Kim ST, Lord EM. 2005. Plantacyanin plays a role in reproduction Moldenhauer K, Wilson EC, Counce P Jr, Hardke J. 2013. Rice growth in Arabidopsis. Plant Physiology 138, 778–789. and development. In: Hardke JT, ed. Arkansas rice production handbook. Little Rock: University of Arkansas Division of Agriculture Cooperative Fan C, Xing Y, Mao H, Lu T, Han B, Xu C, Li X, Zhang Q. 2006. GS3, Extension Service, 9–20. a major QTL for grain length and weight and minor QTL for grain width and thickness in rice, encodes a putative transmembrane protein. Theoretical Pantalião GF, Narciso M, Guimarães C, Castro A, Colombari JM, and Applied Genetics 112, 1164–1171. Breseghello F, Rodrigues L, Vianello RP, Borba TO, Brondani C. 2016. Genome wide association study (GWAS) for grain yield in rice cultivated Fan X, Tang Z, Tan Y, Zhang Y, Luo B, Yang M, Lian X, Shen Q, Miller under water deficit. Genetica 144, 651–664. AJ, Xu G. 2016. Overexpression of a pH-sensitive nitrate transporter in rice increases crop yields. Proceedings of the National Academy of Sciences, Pinto RS, Reynolds MP, Mathews KL, McIntyre CL, Olivares- USA 113, 7118–7123. Villegas JJ, Chapman SC. 2010. Heat and drought adaptive QTL in a wheat population designed to minimize confounding agronomic effects. Fujita D, Trijatmiko KR, Tagle AG, et al. 2013. NAL1 allele from a rice Theoretical and Applied Genetics 121, 1001–1021. landrace greatly increases yield in modern indica cultivars. Proceedings of the National Academy of Sciences, USA 110, 20431–20436. Posé D, Castanedo I, Borsani O, Nieto B, Rosado A, Taconnat L, Ferrer A, Dolan L, Valpuesta V, Botella MA. 2009. Identification of the Arabidopsis Fukai S, Pantuwan G, Jongdee B, Cooper M. 1999. Screening for dry2/sqe1-5 mutant reveals a central role for sterols in drought tolerance and drought resistance in rainfed lowland rice. Field Crops Research 64, 61–74. regulation of reactive oxygen species. The Plant Journal 59, 63–76. Guo Z, Chen D, Alqudah AM, Röder MS, Ganal MW, Schnurbusch T. 2017. Genome-wide association analyses of 54 traits identified multiple Pyngrope S, Bhoomika K, Dubey RS. 2013. Reactive oxygen species, loci for the determination of floret fertility in wheat. New Phytologist 214, ascorbate-glutathione pool, and enzymes of their metabolism in drought- 257–270. sensitive and tolerant indica rice (Oryza sativa L.) seedlings subjected to progressing levels of water deficit. Protoplasma 250, 585–600. Huang X, Qian Q, Liu Z, Sun H, He S, Luo D, Xia G, Chu C, Li J, Fu X. 2009. Natural variation at the DEP1 locus enhances grain yield in rice. Qiu X, Pang Y, Yuan Z, Xing D, Xu J, Dingkuhn M, Li Z, Ye G. 2015. Nature Genetics 41, 494–497. Genome-wide association study of grain appearance and milling quality in a worldwide collection of indica rice germplasm. PLoS ONE 10, e0145577. Huang X, Wei X, Sang T, et al. 2010. Genome-wide association studies of 14 agronomic traits in rice landraces. Nature Genetics 42, 961–967. Rebolledo MC, Peña AL, Duitama J, Cruz DF, Dingkuhn M, Grenier C, Tohme J. 2016. Combining image analysis, genome wide association Huang X, Zhao Y, Wei X, et al. 2012. Genome-wide association study studies and different field trials to reveal stable genetic regions related to of flowering time and grain yield traits in a worldwide collection of rice panicle architecture and the number of spikelets per panicle in rice. Frontiers germplasm. Nature Genetics 44, 32–39. in Plant Science 7, 1384. Kadam N, Tamilselvan A, Lawas LMF, et al. 2017. Genetic control of Reinders A. 2016. Fuel for the road—sugar transport and pollen tube plasticity in root morphology and anatomy of rice in response to water- growth. Journal of Experimental Botany 67, 2121–2123. deficit. Plant Physiology 174, 2302–2315. Reynolds MP, Quilligan E, Aggarwal PK, et al. 2016. An integrated Kadam NN, Xiao G, Melgar RJ, Bahuguna RN, Quinones C, approach to maintaining cereal productivity under climate change. Global Tamilselvan A, Prasad PVV, Jagadish KSV, Donald S. 2014. Agronomic Food Security 8, 9–18. and physiological responses to high temperature, drought, and elevated CO interactions in cereals. Advances in Agronomy 27, 111–156. Saini HS, Lalonde S. 1997. Injuries to reproductive development under water stress, and their consequences for crop productivity. Journal of Crop Kadam NN, Yin X, Bindraban PS, Struik PC, Jagadish KS. 2015. Does Production 1, 223–248. morphological and anatomical plasticity during the vegetative stage make wheat more tolerant of water deficit stress than rice? Plant Physiology 167, Saini HS, Westgate ME, Donald LS. 1999. Reproductive development in 1389–1401. grain crops during drought. Advances in Agronomy 68, 59–96. Kikuchi S, Bheemanahalli R, Jagadish KSV, Kumagai E, Masuya Segura V, Vilhjálmsson BJ, Platt A, Korte A, Seren Ü, Long Q, Y, Kuroda E, Raghavan C, Dingkuhn M, Abe A, Shimono H. 2017. Nordborg M. 2012. An efficient multi-locus mixed-model approach Downloaded from https://academic.oup.com/jxb/article/69/16/4017/4996191 by DeepDyve user on 13 July 2022 4032 | Kadam et al. for genome-wide association studies in structured populations. Nature Takemiya A, Sugiyama N, Fujimoto H, Tsutsumi T, Yamauchi S, Genetics 44, 825–830. Hiyama A, Tada Y, Christie JM, Shimazaki K. 2013. Phosphorylation of BLUS1 kinase by phototropins is a primary step in stomatal opening. Nature Selote DS, Chopra RK. 2004. Drought induced spikelet sterility is Communications 4, 2094. associated with an inefficient antioxidant defense in rice panicles. Physiologia Plantarum 121, 462–471. Venuprasad R, Dalid CO, Del Valle M, Zhao D, Espiritu M, Sta Cruz MT, Amante M, Kumar A, Atlin GN. 2009. Identification and characterization Sheoran IS, Saini HS. 1996. Drought-induced male sterility in rice: changes of large-effect quantitative trait loci for grain yield under lowland drought in carbohydrate levels and enzyme activities associated with the inhibition stress in rice using bulk-segregant analysis. Theoretical and Applied of starch accumulation in pollen. Sexual Plant Reproduction 9, 161–169. Genetics 120, 177–190. Shi W, Yin X, Struik PC, Xie F, Schmidt RC, Jagadish KSV. 2016. Venuprasad R, Lafitte HR, Atlin GN. 2007. Response to direct selection Grain yield and quality responses of tropical hybrid rice to high night-time for grain yield under drought stress in rice. Crop Science 47, 285–293. temperature. Field Crops Research 190, 18–25. Vikram P, Swamy BP, Dixit S, Ahmed HU, Teresa Sta Cruz M, Singh Shomura A, Izawa T, Ebana K, Ebitani T, Kanegae H, Konishi S, Yano AK, Kumar A. 2011. qDTY₁.₁, a major QTL for rice grain yield under M. 2008. Deletion in a gene associated with grain size increased yields reproductive-stage drought stress with a consistent effect in multiple elite during rice domestication. Nature Genetics 40, 1023–1028. genetic backgrounds. BMC Genetics 12, 89. Slafer GA. 2003. Genetic basis of yield as viewed from a crop physiologist’s Vikram P, Swamy BP, Dixit S, Singh R, Singh BP, Miro B, Kohli A, perspective. Annals of Applied Biology 142, 117–128. Henry A, Singh NK, Kumar A. 2015. Drought susceptibility of modern rice varieties: an effect of linkage of drought tolerance with undesirable traits. Song XJ, Huang W, Shi M, Zhu MZ, Lin HX. 2007. A QTL for rice grain Scientific Reports 5, 14799. width and weight encodes a previously unknown RING-type E3 ubiquitin ligase. Nature Genetics 39, 623–630. Wang X, Feng S, Nakayama N, Crosby WL, Irish V, Deng XW, Wei N. 2003. The COP9 signalosome interacts with SCF UFO and participates in Spindel J, Begum H, Akdemir D, Virk P, Collard B, Redoña E, Atlin Arabidopsis flower development. The Plant Cell 15, 1071–1082. G, Jannink JL, McCouch SR. 2015. Genomic selection and association mapping in rice (Oryza sativa): effect of trait genetic architecture, training Yano K, Yamamoto E, Aya K, et al. 2016. Genome-wide association study population composition, marker number and statistical model on accuracy of using whole-genome sequencing rapidly identifies new genes influencing rice genomic selection in elite, tropical rice breeding lines. PLoS Genetics 11, agronomic traits in rice. Nature Genetics 48, 927–934. e1004982. Yin X, Chasalow SD, Stam P, Kropff MJ, Dourleijn CJ, Bos I, Bindraban Swamy BP, Ahmed HU, Henry A, et al. 2013. Genetic, physiological, PS. 2002. Use of component analysis in QTL mapping of complex crop and gene expression analyses reveal that multiple QTL enhance yield of rice traits: a case study on yield in barley. Plant Breeding 121, 314–319. mega-variety IR64 under drought. PLoS ONE 8, e62795. Yin X, Kropff MJ, Horie T, Nakagawa H, Centeno HGS, Zhu D, Swamy BPM, Shamsudin NAA, Rahman SNA, Mauleon R, Ratnam Goudriaan J. 1997. A model for photothermal responses of flowering in W, Sta Cruz MT, Kumar A. 2017. Association mapping of yield and yield- rice I. Model description and parameterization. Field Crops Research 51, related traits under reproductive stage drought stress in rice (Oryza sativa 189–200. L.). Rice 10, 21. Zhang Z, Ersoz E, Lai CQ, et al. 2010. Mixed linear model approach Swamy BP, Vikram P, Dixit S, Ahmed HU, Kumar A. 2011. Meta- adapted for genome-wide association studies. Nature Genetics 42, analysis of grain yield QTL identified during agricultural drought in grasses 355–360. showed consensus. BMC Genomics 12, 319. Zhao K, Tung CW, Eizenga GC, et al. 2011. Genome-wide association Sweeney M, McCouch S. 2007. The complex history of the domestication mapping reveals a rich genetic architecture of complex traits in Oryza sativa. of rice. Annals of Botany 100, 951–957. Nature Communications 2, 467.

Journal

Journal of Experimental Botany – Oxford University Press

Published: Jul 18, 2018

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Genome-wide association reveals novel genomic loci controlling rice grain yield and its component traits under water-deficit stress during the reproductive stage

Genome-wide association reveals novel genomic loci controlling rice grain yield and its component traits under water-deficit stress during the reproductive stage

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Genome-wide association reveals novel genomic loci controlling rice grain yield and its component traits under water-deficit stress during the reproductive stage

Genome-wide association reveals novel genomic loci controlling rice grain yield and its component traits under water-deficit stress during the reproductive stage

References (71)

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies