Genetic Basis Underlying Correlations Among Growth Duration and Yield Traits Revealed by GWAS in Rice (Oryza sativa L.)

Avoidance of disadvantageous genetic correlations among growth duration and yield traits is critical in developing crop varieties that efficiently use light and energy resources and produce high yields. To understand the genetic basis underlying the correlations among heading date and three major yield traits in rice, we investigated the four traits in a diverse and representative core collection of 266 cultivated rice accessions in both long-day and short-day environments, and conducted the genome-wide association study using 4.6 million single nucleotide polymorphisms (SNPs). There were clear positive correlation between heading date and grain number per panicle, and negative correlation between grain number per panicle and panicle number, as well as different degrees of correlations among other traits in different subspecies and environments. We detected 47 pleiotropic genes in 15 pleiotropic quantitative trait loci (pQTLs), 18 pleiotropic genes containing 37 pleiotropic SNPs in 8 pQTLs, 27 pQTLs with r² of linkage disequilibrium higher than 0.2, and 39 pairs of interactive genes from 8 metabolic pathways that may contribute to the above phenotypic correlations, but these genetic bases were different for correlations among different traits. Distributions of haplotypes revealed that selection for pleiotropic genes or interactive genes controlling different traits focused on genotypes with weak effect or on those balancing two traits that maximized production but sometimes their utilization strategies depend on the traits and environment. Detection of pQTLs and interactive genes and associated molecular markers will provide an ability to overcome disadvantageous correlations and to utilize the advantageous correlations among traits through marker-assisted selection in breeding.

Introduction

Rice (Oryza sativa L.) is the staple food of a large proportion of the world population. Development of elite varieties that use light and soil resources efficiently to produce high yield is an important way to eliminate food shortages resulting from population growth and reductions in the area of arable land (Takeda and Matsuoka, 2008). However, in practice, it is difficult to develop varieties that produce high yields under conditions of shortened growth duration (Okada et al., 2017). Rice breeders also often face the problem of negative correlations among different yield traits (Sandhu et al., 2013; Ranawake and Amarasinghe, 2014), such as grain number per panicle (GNP), panicle number per plant (PN), kilo-grain weight (KGW), and rate of seed setting (Sakamoto and Matsuoka, 2008; Zhao et al., 2016). Many published works suggest that there is a disadvantageous positive correlation between heading date (HD) and GNP, but negative correlations among yield traits (Das and Sarma, 2015; Kharb et al., 2016). Most high-yield varieties have long growth duration in regions with adequate light and energy resources because longer growth duration provides more metabolic activity and nutrition for grain filling (Okada et al., 2017). However, longer growth duration becomes a key limiting factor for widespread production planting, especially in high latitude and high altitude regions. Modern breeding has therefore targeted development of high-yielding varieties with moderate growth duration, such as Shanyou 63 (Zhu et al., 2016) with optimal growth duration of 138 days, 120–130 GNP, medium PN and KGW of 29 g, but also maintains higher yield than those with short growth duration (http://www.ricedata.cn/variety/index.htm). Thus it is feasible to balance the relationships among traits by breaking constraints and to uncover the genetic basis underlying correlations among HD and yield.

Mapping of quantitative trait locus (QTL) has revealed many pleiotropic QTLs (pQTLs) that contributing to the correlations among HD and yield traits at the genomic level. Approximately 734, 209, 239, and 223 widely dispersed QTLs for HD, GNP, PN, and KGW were mapped using different segregating populations respectively (http://www.gramene.org/), which densely distributing across the 12 chromosomes (Ni et al., 2009). By searching their shared markers at those QTLs, we detected more than 27-42 pQTLs per combination of two traits (Supplementary Figure 1 and Supplementary Table 6), namely the common QTLs between two traits. However, only few of them were reported to be pleiotropy by mapping in the same population and in a single environment. For example, one pQTL that contributed to the phenotypic variations of both HD and GNP was detected using a recombination inbred line derived from the cross Zhenshan 97B × Miyang 46 (Zhuang et al., 2002). However, most of the other pQTLs detected are not yet to be confirmed by mapping in the same population and in a single environment.

With increasing numbers of QTLs being cloned (Supplementary Tables 11–14), several pleiotropic genes (designated pGenes, the genes with pleiotropy for different traits of interest, that is, commonly associated with the traits of interest in GWAS results) related to the four traits in pQTLs have become evident. Among them, Ghd7, DTH8, and RFL have important roles in regulation of HD and GNP (Rao et al., 2008; Xue et al., 2008; Wei et al., 2010). AID1 and RFL are pGenes that control HD and PN. AID1 mutant plants have fewer tillers and show delayed flowering by up to 15 days compared to the wild type (Zhu et al., 2004). The expression level of RFL affects HD and vegetative axillary meristems (Rao et al., 2008). HGW has a fundamental role in the ubiquitination pathway in control of HD and KGW (Li et al., 2012). LAX1, OsSPL14 and RFL affect GNP and PN by having opposite roles in production of tillers and panicle branches (Komatsu et al., 2003; Rao et al., 2008; Jiao et al., 2010; Miura et al., 2010). GSD1 is an important regulator of GNP and KGW, with a role in regulating photoassimilate translocation through the symplastic pathway to impact grain setting (Gui et al., 2014). The literature about related to pQTLs and pGenes suggests that pleiotropic effects among agronomic traits are very common, but study on its discovery and our understanding of the genetic basis is still quite limited.

Facilitated with high-throughput sequencing technology, genome-wide association study (GWAS) using high quality single nucleotide polymorphisms (SNPs) has been proved to be a good approach to explain the genetic basis of complex quantitative traits (Huang et al., 2010, 2011). It may also provide an efficient tool to discover the genetic basis of correlations among agronomic traits since GWAS has demonstrated the advantages in improving the efficiency of detecting natural variations, and in saving time for QTL mapping and gene cloning, and especially in detecting QTLs controlling multiple agronomic traits in different environments (Myles et al., 2009). In this study, we estimated the phenotypic correlations among HD and yield traits for different rice subspecies in two typical environments using a representative mini core collection (MCC) of the global cultivated rice (Zhang et al., 2011), and carried out GWAS in indica, japonica and the full populations using high quality SNPs from the whole genome. Our aim was to identify QTLs simultaneously controlling growth duration and yield traits, to analyze the genetic bases contributing to correlations among them, and to investigate possible utilization pattern for the QTLs with different genetic bases in breeding practice so as to gain from the positive correlation or avoid the negative correlation. The study sheds light on understanding of the genetic basis underlying the correlations among four traits and provides QTLs and markers to breeders for further improving the important agronomic traits that are subject to both positive and negative correlations.

Materials and Methods

Populations for GWAS

The 266 rice varieties of the MCC for GWAS included 87 non-Chinese cultivated varieties that represented a global collection of more than 9,000 foreign rice accessions, and 179 varieties selected from more than 50,000 Chinese varieties conserved in the Chinese National Gene Bank, and representing more than 70% of the genetic diversity in the respective germplasm sets (Zhang et al., 2011). The MCC thus has a wide geographic and morphological representation (Supplementary Table 1 and Supplementary Figure 2A). It comprises 151 landraces, 114 improved varieties from 24 provinces in China and 35 other countries; it covers all maturity types, including 70 early-season, 100 intermediate-season, and 87 late season varieties.

The phylogenetic relationships of the 266 varieties were determined using genetic distance calculated from 1.8 million high quality SNPs, and the resulting neighbor-joining tree (Supplementary Figure 2B) as well as the principal-component analysis (PCA) (Supplementary Figure 3) showed that they comprised two divergent subspecies groups, Oryza sativa ssp. indica with 157 varieties, and ssp. japonica with 109 varieties.

Phenotyping for MCC

Phenotypes of MCC were evaluated in two rice-growing environments (Supplementary Figure 4); one was grown from November 2012 to April 2013 at Sanya (SY) in Hainan province (18°20′N), representing a short-day environment; the other was grown from May to October 2013 at Changsha (CS) in Hunan province (28°11′N), representing a long-day environment. All varieties were planted in a randomized complete block design with two replications. Strong and uniform seedlings were transplanted according to a standard design of three rows × 10 plants / row, with a relatively low density of 150,000 plants per hectare so as to avoid interaction between density and genotype. Field management, including irrigation, fertilization, and disease and pest control, followed local plot trial management.

The middle five plants in the middle row for each variety were sampled for phenotyping that included HD and the three major yield factors, PN, GNP and KGW. HD of each accession was recorded when the first inflorescence of more than 50% of plants had emerged above the flag leaf sheath; GNP was the mean of grains number on the main panicles of five plants; PN was the mean number of productive panicles from the five plants; and KGW was estimated from the weights of 100 fully filled grains.

Genome Sequencing and SNP Identification

As a subset of the 3,000 Rice Genomes Project (3KRGP), the whole genome sequence of the entire MCC, was carried out using high-throughput sequencing on the HiSeq2000 platform of BGI-Shenzhen (Li, 2014; Alexandrov et al., 2015; Sun et al., 2017). Clean reads showed an average sequencing depth of 14 ×, covering 94.0% of the Nipponbare reference genome (Kawahara et al., 2013). Using the same method as 3KRGP, we called 10.7 million non-redundant SNPs in the MCC. By quality control, we obtained 4.6 million high quality SNPs with minor allele frequency ≤ 5% and missing rate ≤ 50% across the 266 accessions, 12.35 SNPs per kb in the rice genome on average approximately.

GWAS and QTL Detection

GWAS for indica, japonica and full populations on traits HD, GNP, PN and KGW were separately implemented using the compressed mixed linear model (CMLM) that took into account population structure estimated by principal component (PC) and kinship so as to reduce false positives (Kang et al., 2008; Price et al., 2010; Zhang Z. et al., 2010; Yang et al., 2014). The principal component (PC) and kinship were conducted using the 4.6 million SNPs by an R implementation (www.R-project.org) available as part of the GAPIT software package (version 3.0.2) (Lipka et al., 2012).

Given that most of multi-test adjustment methods especially the Bonferroni correction are too stringent because not all markers are independent due to linkage disequilibrium (LD) among markers, missing QTL with small effect usually makes GWAS not as straightforward as previously thought (Liu et al., 2016). Instead of multi-test adjustment method, in this study we used the permutation test, which was proven to be an effective strategy to define the genome-wide significant threshold of a complex agronomic trait controlled by multiple alleles with minor effect, and had better control of population structure (Huang et al., 2015; Yano et al., 2016). We divided the phenotype (Y) of each accession into the original genotypic effect (G) and the fixed effect of population structure (Ps). Ps was estimated by the average effect of each PC on each individual through regression analysis of each PC on Y; the remainder was G after excluding Ps from Y. G was randomly reshuffled as Gr, and the new phenotype of each accession was Ps + Gr. We executed permutation test using CMLM with the same parameters of GWAS, a total of 1,000 sets of Ps + Gr were performed for the four traits with the same number of PC. To improve the computational efficiency, the SNPs whose P-values were greater than 2 in the GWAS using original phenotypes were included in the permutation test. Any SNP whose -log₁₀(P) was higher than -log₁₀(P_perm), where P_perm was the lowest P-value in 1,000 permutation tests, was considered to be real association between the SNP and the original phenotype in the GWAS.

LD length in cultivated rice is approximately 120 and 170 kb for indica and japonica, respectively (Mather et al., 2007; Huang et al., 2010). We therefore defined one QTL by the bin that included at least three significant SNPs with a physical distance <70 kb between two adjacent significant SNPs.

Hitch-Hiking Effects Among SNPs Within pQTL on Phenotypic Correlation

To investigate the role of LD on phenotypic correlation (such as HD vs. GNP) through hitch-hiking effects of QTL controlling one trait (such as HD) with other traits (such as GNP), we firstly calculated the average r² (LD) of all pairs between significant SNPs in HD QTL and significant SNPs in GNP QTL using HAPLOVIEW, the detail is demonstrated in the schematic diagram (Supplementary Figure 11A). We calculated the Pearson correlation coefficient (r_HD2GNP) between average r² of LD and differences between phenotypic variance contributions (PVC) to GNP for HD QTL and GNP QTL (i.e., PVC to GNP for significant SNPs within GNP QTL in GNP GWAS minus PVC to GNP for significant SNPs within HD QTL in GNP GWAS). This r_HD2GNP thus accounted for the hitch-hiking effect of HD QTL on GNP. Similarly, we also calculated the hitch-hiking effect of GNP QTL on HD QTL, r_GNP2HD, and other combinations.

Effects of SNP or Haplotypes and Their Distribution Among Different Variety Types

We investigated the significance of effects on related traits of alleles of each pleiotropic SNP (pSNP; that is, a SNP simultaneously associated with two different traits), haplotypes of pQTLs, and interactive genes (iGenes, the genes that simultaneously control different traits of interest by interaction among those genes) by multiple comparison analysis of GLM in SPSS (version 19.0). Alleles with significantly higher effect on some traits such as HD were denoted as increased effect alleles for that trait, and those with significantly lower effect were denoted as decreased effect alleles. For pSNPs or haplotypes, for example, pleiotropy between HD and GNP, we classified their effects into four types: the same direction (SD) with both increased effect on HD and GNP, or SD with both decreased effect on HD and GNP, the opposite direction (OD) with increased effect on HD but decreased effect on GNP, or OD with decreased effect on HD but increased effect on GNP. We then calculated the distribution of alleles with different kinds of effects among four types of varieties, including indica landraces, improved indica varieties, japonica landraces, and improved japonica varieties.

Results

Phenotypic Variation and Correlation

Phenotypes of the four traits for the whole MCC and two subspecies at SY and CS showed continuous Gaussian distribution (Supplementary Figure 5) and wide variation. The coefficient of variation at SY was higher than that at CS; however, the phenotypic means for the MCC indicated longer HD, higher GNP and PN, and lower KGW at CS than at SY. For subspecies (Supplementary Table 2), japonica had longer HD, more GNP and PN, and lower KGW than indica in both environments. Pearson's correlation analysis (Figure 1 and Supplementary Table 3) indicated that in general the correlations among traits were stronger at SY than that at CS. This may be attributed to the smaller range of variation in HD due to high temperature at CS, but some correlations also appeared as subgroup-specific or environment-specific. There was a significant positive correlation between HD and GNP and a significant negative correlation between GNP and PN in all populations and environments. Longer HD significantly decreased PN and KGW in both subspecies at SY, but its impact on these two traits depended on subspecies at CS, where HD was negatively correlated with KGW, but not PN in indica and the reverse in japonica. The interplay between GNP and KGW depended on subspecies, being significantly negative only in japonica. The interplay between PN and KGW was very weak, being only significantly negative in indica at SY. These findings implicate the genetic foundation underlying phenotypic correlation, which was also impacted by environment to some extent.

FIGURE 1

Figure 1. Phenotype distribution of grain number per panicle (GNP), panicle number (PN) and kilo-grain weight (KGW) along with increase of heading date (HD) in different subspecies at Sanya (A) and Changsha (B), and correlations among the four traits at Sanya (C) and Changsha (D). *, significant correlation at the 0.05 level (two tails); **, significant correlation at the 0.01 level (two tails).

Detection of pQTL by GWAS

GWAS was performed to dissect the underlying genetic basis of correlations among the four agronomic traits for the full population (266 lines) and each subspecies, the differences of which contributed to 94% of population structure and differentiation using CMLM (Huang et al., 2010; Lipka et al., 2012; Figure 2A and Supplementary Figures 6–9). With the permutation test threshold to evaluate significant SNP, we identified 77, 65, 71 and 81 QTLs associated with HD, GNP, PN and KGW, respectively (Supplementary Tables 4, 5), and the reported QTL clusters had high coverage percentages to GWAS QTL (Figure 2A and Supplementary Table 6). Among them, only a few QTLs were repeatedly detected in both environments, which had 2, 6, 3, 7 QTLs for HD, GNP, PN, KGW in the full population, 4, 1, 0, 1 QTLs in indica, and 0, 3, 1, 1 QTLs in japonica, respectively. The low proportion of common QTLs between SY and CS was partly ascribed to the different light and temperature conditions. The number of QTLs in indica was higher than in japonica, and was consistent with lower variation in japonica than in indica (Garris et al., 2005). Within these QTLs, 6 cloned genes were significantly associated including OsMADS50, RFT1, Hd3a regulating HD, TUT1 relating to PN, and OsPPKL2, FLO7 involved in KGW, and a further 8, 2, 4, and 8 cloned genes located in the scope of QTLs.

FIGURE 2

Figure 2. Display of pleiotropic quantitative trait loci (QTLs) identified in the genome-wide association study (GWAS). (A) Circos diagram illustrates QTLs contributing to different traits from GWAS and linkage analysis (LA). Here, the inner four layers from the inside represent QTLs for HD, GNP, PN, and KGW, respectively; the outer four layers represent the frequency of each bin for HD, GNP, PN and KGW covered by LA-QTL respectively; the outside red fonts represent pleiotropic QTLs and the black fonts represent cloned genes. (B) Venn diagram of pleiotropic QTLs from GWAS among the four traits in same population and environment.

There were 46 pQTLs, each of which was shared in GWAS by two different traits in the same population and environment (Figure 2B and Supplementary Figure 10). Among them, 16, 9, 9, 8, 11, 11 pQTLs were detected between HD and GNP, HD, and PN, HD and KGW, GNP and PN, GNP and KGW, and PN and KGW, respectively. There was high correspondence between pQTL number and the degree of phenotypic correlation among traits. For example, the number of pQTLs between HD and GNP was the highest (Figure 3), corresponding to the highest correlation between HD and GNP. These pQTLs may shed light on the genetic basis underling the correlations among traits.

FIGURE 3

Figure 3. Percentages of pleiotropic QTLs for different combinations relative to the total QTL number of each corresponding trait revealed by GWAS. Here, “A”, all pleiotropic QTLs in the same population and environment; “B”, pleiotropic QTLs with r² of Llinkage disequilibrium (LD) > 0.2; “C”, pleiotropic QTLs containing pleiotropic genes; “D”, pleiotropic QTLs containing pleiotropic genes with pleiotropic SNPs. Two bars in each pair of traits (such as HD GNP) in “D” and elsewhere are percentages of pleiotropic QTLs relative to the QTLs of the first trait (such as HD) and the second trait (such as GNP), respectively.

Genetic Analysis of pQTL

There were at least two genetic bases to explain the pleiotropic phenomenon: pGene (Rao et al., 2008; Miura et al., 2010; Li et al., 2012; Lu et al., 2012) or linkage of genes controlling different traits (Bai et al., 2011; Luo et al., 2013). To distinguish these two fundamental genetic bases, we annotated all significant SNPs, and found 47 pGenes in 15 pQTLs, 18 pGenes containing 37 pSNPs in 8 pQTLs, while other 38 pQTLs did not contain any pSNP. We hardly distinguished the real pGenes from the linkage between genes when two pGenes were found in one QTL, unless functions of all pGenes were validated. Here we focused on two sets of genes or SNPs, one set was the cloned pGenes and pSNPs within pQTLs and the other was pQTLs with unknown pGenes or without any pGene or pSNP, in order to investigate the contribution of pGenes or pSNPs and the contribution of linkage between different genes or SNPs within pQTLs to the phenotypic correlations among traits and their utilization in breeding.

Among the 37 pSNPs (Supplementary Tables 7, 8), the most pSNPs and pGenes were between HD and GNP (26 pSNPs and 12 pGenes containing pSNPs) in 5 pQTLs, followed by PN and KGW (8 and 4), and GNP and PN (2 and 2) in 1 and 2 pQTLs. The pQTLs between HD and GNP accounted for 6.49 and 7.69% of the QTLs for HD and GNP), respectively, pQTLs between PN and KGW accounted for 1.41 and 1.25%, and pQTLs between GNP and PN accounted for 3.08 and 2.82%, respectively. Among the 47 pGenes (Supplementary Tables 7, 8), the most pQTLs and pGenes were between HD and GNP (8 pQTLs containing 22 pGenes), followed by GNP and PN (4 and 15), PN and KGW (2 and 7), HD and PN (2 and 5), GNP and KGW (2 and 2), and HD and KGW (1 and 1). The pQTLs between HD and GNP accounted for 10.39 and 12.31% of the QTLs for HD and GNP, respectively, pQTLs between GNP and PN accounted for 6.15 and 5.63%, pQTLs between PN and KGW accounted for 2.82 and 2.47%, pQTLs between HD and PN accounted for 2.60 and 2.82%, pQTLs between GNP and KGW accounted for 3.08 and 2.47%, and pQTLs between HD and KGW accounted for 1.30 and 1.23%. The number of pQTLs containing pSNPs and pGenes represented high correspondence with the strength of phenotypic correlation (Figure 3). Among these pGenes, Ghd7 and HYR were cloned. HYR (LOC_Os03g02650) was reported to control the chlorophyll levels and chloroplast number, and to further affect plant biomass including GNP and PN (Ambavaram et al., 2014). In our study, HYR was pleiotropy between HD and GNP (Figure 4A). Haplotype analysis (Figures 4B,C) indicated that this gene clearly differentiated between the two subspecies, that Hap1 and Hap 3 mainly presented in indica and Hap2 was mainly in japonica. The varieties carrying Hap1 showed significantly longer HD and more GNP than Hap3, possibly explaining the positive correlation between HD and GNP. Indica varieties grow in the low latitudes where their growth is not limited by the light and energy resources (Londo et al., 2006), thus the Hap1 of HYR was mainly selected in indica in order to sufficiently make use of the long season to increase yield, especially GNP. In contrast, japonica varieties grow in the higher latitudes where their growth duration is usually limited (Londo et al., 2006), consequently Hap2 of HYR was mainly selected in japonica to maintain a balance between HD and GNP. The effect of pSNPs (Figures 4D,E) showed that the percentage of pSNPs with SD was 100% for HD vs GNP, completely explaining the strong positive phenotypic correlation between HD and GNP. But for PN vs KGW, 62.50% pSNPs appeared OD effect, being consistent with the moderate strength of negative phenotypic correlation between PN and KGW. The distribution of pSNPs with SD or OD in landraces and improved varieties (Figure 4F) indicated that genotypes with earlier HD and fewer GNP were higher in proportion in japonica than in indica. The subspecific differentiation attributed to the adaptions of two subspecies to different light and energy resources (Khush, 1997; Garris et al., 2005), that is, japonica has decreased its growth duration in order to ripen normally at the cost of decreasing GNP to some extent. For pSNPs between PN and KGW, breeders tended to select the genotypes with more PN but lower KGW in indica, but with the reverse situation in japonica.

FIGURE 4

Figure 4. Analysis of pleiotropic genes and SNPs. (A) Local Manhattan plots from 450,000 to 1,450,000 on Chromosome 3 for GWAS of HD and GNP in indica (ind.) at Sanya (SY). Here, colored dots represent SNPs in pleiotropic gene HYR (LOC_Os03g02650). (B) Gene structure of HYR and its haplotype distribution between the two subspecies. (C) Differences for HD and GNP at SY among different haplotypes in pleiotropic gene HYR, different letters in box plots label significant differences (P < 0.05) using Duncan's test. (D) Correlations of effects calculated using pleiotropic SNPs among different traits. (E) Haplotype percentages of pleiotropic SNPs with SD or OD. (F) Distribution of pleiotropic SNPs with SD or OD in landrace (LAN) and improved varieties (IMP) for HD and GNP, PN and KGW in indica and japonica. I, increased effect; D, decreased effect; SD, same direction effect; OD, opposite direction effect.

To estimate the effect of linkage between genes or QTLs on phenotypic correlation between pairs of traits, we calculated the average r² (LD) between QTLs of two traits belonging to the same pQTL. Among 46 pQTLs, the LD of approximately 5% of combinations exceeded 0.6 (Supplementary Figure 11B). The number of pQTLs indicated the consistent trend of distribution among six combinations at each of four r² (LD) levels (> 0.2, > 0.4, > 0.6 and > 0.8) (Supplementary Figure 11C). Among 27 pQTLs with LD stronger than 0.2, there were 9, 3, 4, 6, 4, and 5 pQTLs in combinations between HD and GNP, HD and PN, HD and KGW, GNP and PN, GNP and KGW, PN and KGW, respectively. pQTLs between HD and GNP with LD stronger than 0.2 accounted for 11.69 and 13.85% of the QTLs for HD and GNP, respectively; those between HD and PN accounted for 3.90 and 4.23%, those between HD and KGW accounted for 9.23 and 8.45%, those between GNP and PN accounted for 6.15 and 4.94%, those between GNP and KGW accounted for 5.19 and 4.94%, and those between PN and KGW accounted for 7.04 and 6.17%, respectively. These percentages represented roughly the strength of phenotypic correlation between the corresponding traits. The correlation analysis (Figure 5A) between PVC differences and r² (LD) of pQTLs showed that the interplay influence decreased along with the LD decay in one pQTL, indicating the contribution of LD to the phenotypic correlation. The effect of these pQTLs on different traits (Figures 5B,C) indicated that 96.15% of haplotypes showed SD effects for HD vs GNP, only 30.77 and 75% of haplotypes showed OD effects for GNP vs PN and PN vs KGW, implying their weaker contributions to phenotypic correlation than pSNPs. The distribution of haplotypes with SD or OD in landraces and improved varieties (Figure 5D) revealed the possibility of breaking the phenotypic correlations by recombination, which may result to produce haplotypes with earlier HD but increased GNP, and haplotypes with more PN and higher KGW simultaneously. The above results indicated that linkage between genes for different traits contributes partially to the phenotypic correlation, but their constraints on improving different traits simultaneously can be broken by recombination.

FIGURE 5

Figure 5. Analysis of pleiotropic QTLs. (A) Correlations between LD of two adjacent QTLs and differences of phenotypic variance contributions (PVC). Red, yellow, blue and green represent the traits of calculated PVC respectively, and each combination has two corresponding results of correlation analysis. (B) Correlations of effect which were calculated using pleiotropic QTLs (r² of LD > 0.2) among different traits. (C) Haplotype percentage of pleiotropic QTLs (r² of LD > 0.2) with SD or OD. (D) Distribution of pleiotropic QTLs (r² of LD > 0.2) with SD or OD in LAN and IMP for HD and GNP, GNP and PN, PN and KGW in indica and japonica. I, increased effect; D, decreased effect; SD, same direction effect; OD, opposite direction effect; LAN, landrace varieties; IMP, improved varieties.

Shared Pathways Contribute to Phenotypic Correlation

In addition to pQTL, it was reported that interaction between D53 and IPA1 in the strigolactone signal pathway can simultaneously control plant height and tiller number in rice (Song et al., 2017). We investigated the contribution of genetic interaction among genes involved in the same pathway to the phenotypic correlation. By pathway searching in the KEGG database (http://www.kegg.jp/) for all cloned genes and significantly associated genes related to the four traits, we discovered 15 pathways in which two or more genes were involved in regulation of multiple traits and in which at least one of them was cloned. In the 15 pathways (Supplementary Table 9) there were 75 new candidate genes (14 for HD, 28 for GNP, 15 for PN, and 18 for KGW) and 27 cloned genes. These pathways included arginine and proline metabolism, biosynthesis of amino acids, carbon fixation in photosynthetic organisms, carbon metabolism, carotenoid biosynthesis, circadian rhythm-plant, diterpenoid biosynthesis, glycolysis-gluconeogenesis, nitrogen metabolism, plant hormone signal transduction, plant-pathogen interaction, porphyrin and chlorophyll metabolism, starch and sucrose metabolism, ubiquitin mediated proteolysis and zeatin biosynthesis. Through interaction analysis among genes within the same pathway (Supplementary Table 10), we detected 39 pairs of iGenes in 8 pathways, including 11 iGenes in 5 pathways for HD and GNP, 3 in 2 for HD and PN, 4 in 3 for HD and KGW, 7 in 5 for GNP and PN, 6 in 4 for GNP and KGW, and 8 in 4 for PN and KGW. The higher number of iGenes and pathways represented a trend with stronger correlations among traits (Figure 6A), and apparently indicating their contribution to phenotypic correlations among the four traits. For example, in the carotenoid biosynthetic pathway (Figure 6E) we found that OsCCD7 (LOC_Os04g46470) and CYP97A4 (LOC_Os02g57290) interactively control PN and GNP by determining the production of carotenoid, i.e., strigolactone or abscisic acid. It has been shown that ABA is a regulator of strigolactone biosynthesis (López-Ráez et al., 2010; Cheng et al., 2017). OsCCD7 (LOC_Os04g46470) was reported to negatively control tiller number, PN and GNP (Kulkarni et al., 2014), CYP97A4 was expressed in the root and stalk junction from which tillers emerge (Lv et al., 2012). Data in the RiceXpro database (http://ricexpro.dna.affrc.go.jp/) also showed high expression level of CYP97A4 in inflorescences of Nipponbare. The above results suggest that OsCCD7 (LOC_Os04g46470) and CYP97A4 (LOC_Os02g57290) might jointly control GNP and PN by interaction in the carotenoid biosynthetic pathway. The effect of haplotypes for each pair of iGenes on the traits (Figures 6B,C) indicated that 100.00% of haplotypes were SD for HD vs. GNP, and 100, 96.55, and 100% of haplotypes were OD for HD vs KGW, GNP vs. PN, and PN vs. KGW. These results implied that the contribution of iGenes to phenotypic correlation was higher than pQTLs with r² LD > 0.2. The distribution of haplotypes with SD or OD in landraces and improved varieties (Figure 6D) indicated that breeders tended to select genotypes with longer growth duration with a hope to increase GNP of indica varieties grown at low latitudes, but select genotypes with shorter growth duration so as to guarantee timely maturity of japonica varieties grown at high latitudes (Londo et al., 2006). Moreover, breeders preferred the haplotypes with more GNP and fewer PN when the effects of their haplotypes were opposite direction, because increased planting density can compensate for disadvantages of fewer PN. Consequently, only a small proportion of varieties containing the haplotypes in iGenes that increase both GNP and PN.

FIGURE 6

Figure 6. Analysis of shared pathways containing interactive genes. (A) Numbers of pathways, pathways containing interactive genes and pairs of interactive genes detected by GWAS. (B) Correlations of effect which were calculated using interactive genes among different traits. (C) Haplotype percentages of interactive genes with SD or OD. (D) Distribution of interactive genes from pathways with SD or OD in LAN and IMP for HD and GNP, HD and KGW, GNP and PN, GNP and KGW in indica and japonica. (E) Pleiotropy of genes regulating GNP (yellow) and PN (blue) by genetic interaction in the partial sketchy pathway of carotenoid biosynthesis based on the KEGG database (http://www.kegg.jp/), including in known genes and candidate genes from GWAS. Black solid arrows, direct synthetic steps; black dotted arrows, indirect synthetic steps; brown dotted lines, significant interactions among participating genes; I, increased effect; D, decreased effect; SD, same direction effect; OD, opposite direction effect; LAN, landrace varieties; IMP, improved varieties.

Discussion

A common phenomenon in crop production is phenotypic correlations among agronomic traits. Many of these phenotypic correlations are constraints to crop production and breeding progress. For example, the positive correlation between HD and yield traits usually limits planting region or reduces the production efficiency in regions with restricted light and heat resources, and the negative correlations among yield traits have been the critical constraints when yield reach certain levels. Using the diverse rice MCC and whole genome sequence, we investigated the possible genetic basis underlying correlations among HD and yield traits based on GWAS and interaction analysis. Our results provided not only more information about pSNP, pGene, or pQTL, but also theoretical guidance to develop the varieties with short growth duration and high yield.

Genetic Bases of Correlations Among HD and Yield Traits

Previous study supported at least two genetic bases underlying correlations among traits in rice, i.e., pGenes (Supplementary Figures 12–15) and linkage between two genes controlling different traits (Bai et al., 2011; Luo et al., 2013). Our results not only confirmed the two genetic bases, but also suggested a third one, i.e., the interaction between genes within the same biological pathway. Although it was not definitively reported that interactive genes within the same biological pathway contribute to correlations between traits, Song et al. (2017) reported that interaction between D53 and IPA1 in the strigolactone signal pathway can simultaneously control the plant height and tiller number. We identified 39 pairs of iGenes in 8 pathways contributing to phenotypic correlation among growth duration and yield traits. Among those pathways five contain at least five candidate genes involved in controlling correlations among traits by gene-gene interaction. These five pathways involved two classes of metabolism, one related to energy metabolism, including carbon metabolism, glycolysis / gluconeogenesis, and starch and sucrose metabolism; the other one related to plant hormones, including plant hormone - signal transduction and carotenoid biosynthesis. Energy metabolism is fundamental for development of almost all agronomic traits, and thus contributes to phenotypic correlations among traits. For example, the Moc2 mutant has significantly reduced tiller numbers and a dwarf phenotype; Moc2 encodes cytosolic fructose-1,6-bisphosphatase 1 (FBP1), which is a key enzyme in the sucrose biosynthesis pathway (Koumoto et al., 2013). Several genes involving in plant hormone biosynthesis and signal transduction are reported to control different traits. For example, D10 encodes a carotenoid cleavage dioxygenase and functions in strigolactone biosynthesis, and can regulate the tiller number and plant height simultaneously (Zhang S. et al., 2010). Our results indicated that OsCCD7 (LOC_Os04g46470) and CYP97A4 (LOC_Os02g57290) might together control GNP and PN by interaction in carotenoid biosynthetic pathway. The contributions of these three genetic bases to correlations among traits varied. Percentages of SD and OD indicated that in general the contributions of pSNPs or pGenes were the strongest, followed by iGenes within pathways, and linkage between genes was the lowest. The number of pSNPs, pQTLs and iGenes among different traits (Supplementary Table 8) also indicated different genetic bases for correlations among traits. Correlations between HD and GNP, GNP, and PN can attribute to all three genetic bases, and correlation between PN and KGW is mainly ascribed to pSNPs or pGenes. Of course, the resolution of GWAS cannot distinguish pGenes from pQTL that contain more than two tightly linked genes and cannot validate iGenes, these need to be confirmed using additional mapping populations and molecular approaches.

Implementation of Genetic Correlations in Molecular Breeding by Marker Assisted Selections (MAS)

Disadvantageous correlations among agronomic traits are important constraints to maximizing the potential of each trait. Therefore, breaking or weakening correlations is one of the problems faced by breeders. Our results reveal the diverse genetic bases of phenotypic correlation in rice and different strategies for improving the related traits. It is a difficult task to develop varieties with short growth duration and more GNP due to the high proportion of pGenes, pQTLs and interactions in the same pathways, in which the genotype with short growth duration is accompanied by few GNP in majority of the cases. To overcome the contradictory effect of pGenes or iGenes, farmers and breeders tend to utilize genotypes with weak sensitivity to day-length so as to develop the varieties that adapt the long-day environment with the cost of moderately decreased GNP during the domestication of japonica (Xue et al., 2008). In pQTLs, genotypes with short growth duration and higher GNP were developed by breaking the linkage between genes controlling growth duration and genes controlling GNP (Figure 5D), except for the selection of genotypes with weak effect. This practice suggests a potential to overcome disadvantageous correlations between growth duration and GNP, that is, we can monitor the genotypic relationship between genes with tight linkage by molecular markers and thereby identify the rare advantageous genotypic combinations during breeding. It is undeniable that the positive correlation between growth duration and GNP is advantageous in environments where the light and temperature resources are adequate for the growth and harvest of indica varieties in low latitudes and altitudes (Khush, 1997; Londo et al., 2006). Our results indicated that the phenotypic correlations and their genetic bases (including pGenes, pQTLs and iGenes) vary in different environments and sometimesare environment-specific, and their utilization strategies during domestication and breeding as indicated above depends on the geographic location, especially the latitude and altitude. Our reported QTLs, pGenes, pQTLs and iGenes were clearly labeled by the locations where they detected (Supplementary Tables 4, 6, 7, 9, 10), this will facilitate their utilization in breeding and further study such as the molecular mechanism of GxE interaction.

In summary, phenotypic correlations among growth duration and yield traits can be advantageous or disadvantageous depending on variety types and environments. Selection for pGenes or iGenes needs to focus on the genotypes with weak effects when two traits show disadvantageous negative correlations or on genotypes with strong effect when two traits have advantageous positive correlations. Discovery of pQTLs attributed to different genes with strong LD or iGenes will provide breeders the opportunities to overcome the disadvantageous phenotypic correlations among traits by breaking the disadvantageous linkages or genotype combination for two genes, and also provide the opportunities to improve the breeding efficiency by simultaneously selecting two trait pairs with advantageous phenotypic correlation through MAS.

Author Contributions

FL and HZ designed the experiments. FL wrote the manuscript and performed data analysis. JX, XZ, XW, XM, ZhiZ, LZ, and SZ helped in data analysis. YZ, ZhaZ, and MR helped to prepare the manuscript. HZ, ZL, and JL conceived and supervised the project. HZ and ZL conceptualized and guided the study.

Funding

This research was supported by grants from Ministry of Science and Technology of China (2015BAD02B01).

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

We thank Professor Robert McIntosh (Plant Breeding Institute, University of Sydney) and Dr. Ming Li (Beijing Biohuge Biological Science and Technology Co Ltd) for critical reading and suggested revisions to the manuscript.

Supplementary Material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2018.00650/full#supplementary-material

Abbreviations

3KRGP, 3000 Rice Genomes Project; CMLM, Compressed mixed linear model; CS, Changsha; GNP, Grain number per panicle; GWAS, Genome-wide association study; HD, Heading date; iGene, Interactive gene; KGW, Kilo-grain weight; LD, Linkage disequilibrium; MAS, Marker-assisted selection; MCC, Mini core collection; OD, Opposite direction effect; pGene, Pleiotropic gene; PN, Panicle number per plant; Pqtl, Pleiotropic QTL; pSNP, Pleiotropic SNP; PVC, Phenotypic variance contribution; QTL, Quantitative trait locus; SD, Same direction effect; SY, Sanya; SNP, Single nucleotide polymorphism.

References

Alexandrov, N., Tai, S., Wang, W., Mansueto, L., Palis, K., Fuentes, R. R., et al. (2015). SNP-seek database of SNPs derived from 3000 rice genomes. Nucleic Acids Res. 43, 1023–1027. doi: 10.1093/nar/gku1039

PubMed Abstract | CrossRef Full Text | Google Scholar

Ambavaram, M. M., Basu, S., Krishnan, A., Ramegowda, V., Batlang, U., Rahman, L., et al. (2014). Coordinated regulation of photosynthesis in rice increases yield and tolerance to environmental stress. Nat. Commun. 5, 1–14. doi: 10.1038/ncomms6302

PubMed Abstract | CrossRef Full Text

Bai, X. F., Luo, L. J., Yan, W. H., Kovi, M. R., and Xing, Y. Z. (2011). Quantitative trait loci for rice yield-related traits using recombinant inbred lines derived from two diverse cultivars. J. Genet. 90, 209–215. doi: 10.1007/s12041-011-0057-y.

PubMed Abstract | CrossRef Full Text | Google Scholar

Cheng, X., Floková, K., Bouwmeester, H., and Ruyter-Spira, C. (2017). The role of endogenous strigolactones and their interaction with ABA during the infection process of the parasitic weed phelipanche ramosa in tomato plants. Front. Plant Sci. 8:392. doi: 10.3389/fpls.2017.00392

PubMed Abstract | CrossRef Full Text | Google Scholar

Das, S., and Sarma, D. (2015). Correlation and path analysis of morpho physiological and yield traits in boro rice (Oryza sativa L.). Agric. Sci. Digest. 35, 256–262. doi: 10.18805/asd.v35i4.6855

CrossRef Full Text | Google Scholar

Garris, A. J., Tai, T. H., Coburn, J., Kresovich, S., and McCouch, S. (2005). Genetic structure and diversity in Oryza sativa L. Genetics 169, 1631–1638. doi: 10.1534/genetics.104.035642

PubMed Abstract | CrossRef Full Text | Google Scholar

Gui, J., Liu, C., Shen, J., and Li, L. (2014). Grain setting defect1, encoding a remorin protein, affects the grain setting in rice through regulating plasmodesmatal conductance. Plant Physiol. 166, 1463–1478. doi: 10.1104/pp.114.246769

PubMed Abstract | CrossRef Full Text | Google Scholar

Huang, X., Wei, X., Sang, T., Zhao, Q., Feng, Q., Zhao, Y., et al. (2010). Genome-wide association studies of 14 agronomic traits in rice landraces. Nat. Genet. 42, 961–967. doi: 10.1038/ng.695

PubMed Abstract | CrossRef Full Text | Google Scholar

Huang, X., Yang, S., Gong, J., Zhao, Y., Feng, Q., Gong, H., et al. (2015). Genomic analysis of hybrid rice varieties reveals numerous superior alleles that contribute to heterosis. Nat. Commun. 6, 1–9. doi: 10.1038/ncomms7258

PubMed Abstract | CrossRef Full Text | Google Scholar

Huang, X., Zhao, Y., Wei, X., Li, C., Wang, A., Zhao, Q., et al. (2011). Genome-wide association study of flowering time and grain yield traits in a worldwide collection of rice germplasm. Nat. Genet. 44, 32–39. doi: 10.1038/ng.1018

PubMed Abstract | CrossRef Full Text | Google Scholar

Jiao, Y., Wang, Y., Xue, D., Wang, J., Yan, M., Liu, G., et al. (2010). Regulation of OsSPL14 by OsmiR156 defines ideal plant architecture in rice. Nat. Genet. 42, 541–544. doi: 10.1038/ng.591

PubMed Abstract | CrossRef Full Text | Google Scholar

Kang, H. M., Zaitlen, N. A., Wade, C. M., Kirby, A., Heckerman, D., Daly, M. J., et al. (2008). Efficient control of population structure in model organism association mapping. Genetics 178, 1709–1723. doi: 10.1534/genetics.107.080101

PubMed Abstract | CrossRef Full Text | Google Scholar

Kawahara, Y., Bastide, M., Hamilton, J. P., Kanamori, H., McCombie, W. R., Ouyang, S., et al. (2013). Improvement of the Oryza sativa Nipponbare reference genome using next generation sequence and optical map data. Rice 6:4. doi: 10.1186/1939-8433-6-4

PubMed Abstract | CrossRef Full Text | Google Scholar

Kharb, A., Jain, S., and Jain, R. K. (2016). Phenotypic analysis, correlation studies and linkage mapping of QTL for traits promoting cultivation under dry direct seeded aerobic conditions for the development of water-efficient high yielding rice lines. Cereal Res. Commun. 44, 658–668. doi: 10.1556/0806.44.2016.037

CrossRef Full Text | Google Scholar

Khush, G. S. (1997). Origin, dispersal, cultivation and variation of rice. Plant Mol. Biol. 35, 25–34. doi: 10.1023/A:1005810616885

PubMed Abstract | CrossRef Full Text | Google Scholar

Komatsu, K., Maekawa, M., Ujiie, S., Satake, Y., Furutani, I., Okamoto, H., et al. (2003). LAX and SPA: major regulators of shoot branching in rice. Proc. Natl. Acad. Sci. U.S.A. 100, 11765–11770. doi: 10.1073/pnas.1932414100

PubMed Abstract | CrossRef Full Text | Google Scholar

Koumoto, T., Shimada, H., Kusano, H., She, K. C., Iwamoto, M., and Takano, M. (2013). Rice monoculm mutation moc2, which inhibits outgrowth of the second tillers, is ascribed to lack of a fructose-1,6-bisphosphatase. Plant Biotechnol. 30, 47–56. doi: 10.5511/plantbiotechnology.12.1210a

CrossRef Full Text | Google Scholar

Kulkarni, K. P., Vishwakarma, C., Sahoo, S. P., Lima, J. M., Nath, M., Dokku, P., et al. (2014). A substitution mutation in OsCCD7 cosegregates with dwarf and increased tillering phenotype in rice. J. Genet. 93, 389–401. doi: 10.1007/s12041-014-0389-5

PubMed Abstract | CrossRef Full Text | Google Scholar

Li, J., Chu, H., Zhang, Y., Mou, T., Wu, C., Zhang, Q., et al. (2012). The rice HGW gene encodes a ubiquitin-associated (UBA) domain protein that regulates heading date and grain weight. PLoS ONE. 7:e34231. doi: 10.1371/journal.pone.0034231

PubMed Abstract | CrossRef Full Text | Google Scholar

Li, J. (2014). The 3,000 rice genomes project. Giga Sci. 3, 3–7. doi: 10.1186/2047-217X-3-8

PubMed Abstract | CrossRef Full Text | Google Scholar

Lipka, A. E., Tian, F., Wang, Q., Peiffer, J., Li, M., Bradbury, P. J., et al. (2012). GAPIT: genome association and prediction integrated tool. Bioinformatics 28, 2397–2399. doi: 10.1093/bioinformatics/bts444

PubMed Abstract | CrossRef Full Text | Google Scholar

Liu, X., Huang, M., Fan, B., Buckler, E. S., and Zhang, Z. (2016). Correction: iterative usage of fixed and random effect models for powerful and efficient genome-wide association studies. PLoS Genet. 12:e1005957. doi: 10.1371/journal.pgen.1005957

CrossRef Full Text | Google Scholar

Londo, J. P., Chiang, Y. C., Hung, K. H., Chiang, T. Y., Schaal, B. A. (2006). Phylogeography of Asian wild rice, Oryza rufipogon, reveals multiple independent domestications of cultivated rice, Oryza sativa. Proc. Natl. Acad. Sci. U.S.A. 103, 9578–9583. doi: 10.1073/pnas.0603152103

PubMed Abstract | CrossRef Full Text | Google Scholar

López-Ráez, J. A., Kohlen, W., Charnikhova, T., Mulder, P., Undas, A. K., Sergeant, M. J., et al. (2010). Does abscisic acid affect strigolactone biosynthesis? New Phytol. 187, 343–354. doi: 10.1111/j.1469-8137.2010.03291.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Lu, L., Yan, W., Xue, W., Shao, D., Xing, Y. (2012). Evolution and association analysis of Ghd7 in rice. PLoS ONE 7:e34021. doi: 10.1371/journal.pone.0034021

PubMed Abstract | CrossRef Full Text | Google Scholar

Luo, X., Ji, S., Yuan, P., Lee, H., Kim, D., Balkunde, S., et al. (2013). QTL mapping reveals a tight linkage between QTLs for grain weight and panicle spikelet number in rice. Rice 6:33. doi: 10.1186/1939-8433-6-33

PubMed Abstract | CrossRef Full Text | Google Scholar

Lv, M. Z., Chao, D. Y., Shan, J. X., Zhu, M. Z., Shi, M., Gao, J. P., et al. (2012). Rice carotenoid beta-ring hydroxylase CYP97A4 is involved in lutein biosynthesis. Plant Cell Physiol. 53, 987–1002. doi: 10.1093/pcp/pcs041

PubMed Abstract | CrossRef Full Text | Google Scholar

Mather, K. A., Caicedo, A. L., Polato, N. R., Olsen, K. M., McCouch, S., and Purugganan, M. D. (2007). The extent of linkage disequilibrium in rice (Oryza sativa L.). Genetics 177, 2223–2232. doi: 10.1534/genetics.107.079616

PubMed Abstract | CrossRef Full Text | Google Scholar

Miura, K., Ikeda, M., Matsubara, A., Song, X. J., Ito, M., Asano, K., et al. (2010). OsSPL14 promotes panicle branching and higher grain productivity in rice. Nat. Genet. 42, 545–549. doi: 10.1038/ng.592

PubMed Abstract | CrossRef Full Text | Google Scholar

Myles, S., Peiffer, J., Brown, P. J., Ersoz, E. S., Zhang, Z., Costich, D. E., et al. (2009). Association mapping: critical considerations shift from genotyping to experimental design. Plant Cell 21, 2194–2202. doi: 10.1105/tpc.109.068437

PubMed Abstract | CrossRef Full Text | Google Scholar

Ni, J., Pujar, A., Youens-Clark, K., Yap, I., Jaiswal, P., Tecle, I., et al. (2009). Gramene QTL database: development, content and applications. Database 2009, 1–13. doi: 10.1093/database/bap005

PubMed Abstract | CrossRef Full Text | Google Scholar

Okada, R., Nemoto, Y., Endo-Higashi, N., Izawa, T. (2017). Synthetic control of flowering in rice independent of the cultivation environment. Nat. Plants 3, 1–7. doi: 10.1038/nplants.2017.39

PubMed Abstract | CrossRef Full Text | Google Scholar

Price, A. L., Zaitlen, N. A., Reich, D., and Patterson, N. (2010). New approaches to population stratification in genome-wide association studies. Nat. Rev. Genet. 11, 459–463. doi: 10.1038/nrg2813.

PubMed Abstract | CrossRef Full Text | Google Scholar

Ranawake, A. L., and Amarasinghe, U. G. S. (2014). Relationship of yield and yield related traits of some traditional rice cultivars in sri lanka as described by correlation analysis. J. Sci. Res. Rep. 3, 2395–2403. doi: 10.9734/JSRR/2014/12050

CrossRef Full Text | Google Scholar

Rao, N. N., Prasad, K., Kumar, P. R., Vijayraghavan, U. (2008). Distinct regulatory role for RFL, the rice LFY homolog, in determining flowering time and plant architecture. Proc. Natl. Acad. Sci. USA. 105, 3646–3651. doi: 10.1073/pnas.0709059105

PubMed Abstract | CrossRef Full Text | Google Scholar

Sakamoto, T., and Matsuoka, M. (2008). Identifying and exploiting grain yield genes in rice. Curr. Opin. Plant Biol. 11, 209–214. doi: 10.1016/j.pbi.2008.01.009

PubMed Abstract | CrossRef Full Text | Google Scholar

Sandhu, N., Jain, S., Kumar, A., Mehla, B. S., Jain, R. (2013). Genetic variation, linkage mapping of QTL and correlation studies for yield, root, and agronomic traits for aerobic adaptation. BMC Genet. 14:104. doi: 10.1186/1471-2156-14-104

PubMed Abstract | CrossRef Full Text | Google Scholar

Song, X., Lu, Z., Yu, H., Shao, G., Xiong, J., Meng, X., et al. (2017). IPA1 functions as a downstream transcription factor repressed by D53 in strigolactone signaling in rice. Cell Res. 27, 1–14. doi: 10.1038/cr.2017.102

PubMed Abstract | CrossRef Full Text | Google Scholar

Sun, C., Hu, Z., Zheng, T., Lu, K., Zhao, Y., Wang, W., et al. (2017). RPAN: rice pan-genome browser for approximately 3000 rice genomes. Nucleic Acids Res. 45, 597–605. doi: 10.1093/nar/gkw958

PubMed Abstract | CrossRef Full Text

Takeda, S., and Matsuoka, M. (2008). Genetic approaches to crop improvement: responding to environmental and population changes. Nat. Rev. Genet. 9, 444–457. doi: 10.1038/nrg2342

PubMed Abstract | CrossRef Full Text | Google Scholar

Wei, X., Xu, J., Guo, H., Jiang, L., Chen, S., Yu, C., et al. (2010). DTH8 suppresses flowering in rice, influencing plant height and yield potential simultaneously. Plant Physiol. 153, 1747–1758. doi: 10.1104/pp.110.156943

PubMed Abstract | CrossRef Full Text | Google Scholar

Xue, W., Xing, Y., Weng, X., Zhao, Y., Tang, W., Wang, L., et al. (2008). Natural variation in Ghd7 is an important regulator of heading date and yield potential in rice. Nat. Genet. 40, 761–767. doi: 10.1038/ng.143

PubMed Abstract | CrossRef Full Text | Google Scholar

Yang, J., Zaitlen, N. A., Goddard, M. E., Visscher, P. M., Price, A. L. (2014). Advantages and pitfalls in the application of mixed-model association methods. Nat. Genet. 46, 100–106. doi: 10.1038/ng.2876

PubMed Abstract | CrossRef Full Text | Google Scholar

Yano, K., Yamamoto, E., Aya, K., Takeuchi, H., Lo, P. C., Hu, L., et al. (2016). Genome-wide association study using whole-genome sequencing rapidly identifies new genes influencing agronomic traits in rice. Nat. Genet. 48, 927–934. doi: 10.1038/ng.3596

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhang, H., Zhang, D., Wang, M., Sun, J., Qi, Y., Li, J., et al. (2011). A core collection and mini core collection of Oryza sativa L. in China. Theor. Appl. Genet. 122, 49–61. doi: 10.1007/s00122-010-1421-7

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhang, S., Li, G., Fang, J., Chen, W., Jiang, H., Zou, J., et al. (2010). The interactions among DWARF10, auxin and cytokinin underlie lateral bud outgrowth in rice. J. Integr. Plant Biol. 52, 626–638. doi: 10.1111/j.1744-7909.2010.00960.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhang, Z., Ersoz, E., Lai, C. Q., Todhunter, R. J., Tiwari, H. K., Gore, M. A., et al. (2010). Mixed linear model approach adapted for genome-wide association studies. Nat. Genet. 42, 355–360. doi: 10.1038/ng.546

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhao, F., Zhu, H., Zeng, R., Zhang, G., Xu, S. (2016). Detection of additive and additive*environment interaction effects of QTLs for yield-component traits of rice using single-segment substitution lines (SSSLs). Plant Breed. 135, 452–458. doi: 10.1111/pbr.12385

CrossRef Full Text | Google Scholar

Zhu, D., Zhou, G., Xu, C., Zhang, Q. (2016). Genetic components of heterosis for seedling traits in an elite rice hybrid analyzed using an immortalized F2 population. J. Genet. Genomics. 43, 87–97. doi: 10.1016/j.jgg.2016.01.002

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhu, Q. H., Ramm, K., Shivakkumar, R., Dennis, E. S., Upadhyaya, N. M. (2004). The anther indehiscence1 gene encoding a single MYB domain protein is involved in anther development in rice. Plant Physiol. 135, 1514–1525. doi: 10.1104/pp.104.041459

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhuang, J. Y., Fan, Y. Y., Rao, Z. M., Wu, J. L., Xia, Y. W., Zheng, K. L. (2002). Analysis on additive effects and additive-by-additive epistatic effects of QTLs for yield traits in a recombinant inbred line population of rice. Theor. Appl. Genet. 105, 1137–1145. doi: 10.1007/s00122-002-0974-5

PubMed Abstract | CrossRef Full Text | Google Scholar