Integrating Multi-omics Analysis to Decipher the Genetic Basis of Heterosis in Xieyou9308 Authors: Feng Lin, Tainle Zhu, Muhammad Kamran, Xiong Zheng, Mingzhe Suo, Haiming Xu, Liyong Cao, Yongbo Hong Published: 2025 ABSTRACT The intersubspecific rice hybrid Xieyou9308 exhibits striking hybrid vigor and serves as a valuable resource for dissecting the genetic basis of yield heterosis in indica-japonica crosses. However, the molecular mechanisms underlying this hybrid vigor remain poorly understood. Leveraging a set of recombinant inbred lines (RILs) and an immortalized F2 (IMF2) population derived from Xieyou9308, 1,300,425 genomic variants were identified, which were consolidated into 3818 BIN markers. Genomic regions inherited from Zhonghui9308 were positively associated with traits including plant height, heading date, panicle length, and grain number per panicle. In contrast, regions derived from XieqingzaoB, particularly those located on chromosome 3, positively associated with grain length and 1000-grain weight. In the RIL population, 43 additive QTLs and 3 pairs of epistatic QTLs were identified for 10 agronomic traits, with the majority mapped to chromosomes 3 and 10, which harbor candidate genes including Ehd1, GS3, and OsTB1. In the IMF2 population, 118 significant QTLs exhibiting partial dominance and 11 epistatic interactions were identified, underscoring the contributions of both partial dominance and epistasis to trait expression. Transcriptomic analysis further corroborated these findings, showing that alleles predominantly from Zhonghui9308 contributed to a prolonged growth duration, while those from XieqingzaoB shortened the growth duration. Collectively, these findings indicate that the hybrid vigor exhibited in Xieyou9308 is attributable to the synergistic effects of superior parental alleles and their epistatic interactions. These insights offer a valuable foundation for molecular breeding strategies aimed at improving yield in indica-japonica hybrid rice.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12284-025-00875-4. Background Heterosis, or hybrid vigor, refers to the phenomenon in which hybrid offspring outperform their homozygous parents in growth, yield, and other desirable traits (Shull 1908, 1914). Heterotic crop development, especially in rice and maize, is a major genetic achievement in agriculture, with hybrid seeds now accounting for over half of global production and driving substantial yield gains (Cheng et al. 2007; Zhou et al. 2012). Hybridization between Xian/indica (XI) and Geng/japonica (GJ) rice, combined with the strategic adoption of plant ideotypes, has played a pivotal role in enhancing the yield of modern GJ rice in China over the past five decades (Chen et al. 2007). Despite over a century of research and debate, the biological mechanisms underlying indica-japonica intersubspecific hybridization and their role in heterosis remain poorly understood, which continues to constrain their full utilization in rice breeding. The genomic basis of heterosis has been widely investigated across various crop species, leading to the development of several classical explanatory models. Evidence from multiple crops supports three non-mutually exclusive hypotheses regarding the genetic loci contributing to heterosis: the dominance model, the single-locus overdominance model, and the epistasis model (Hollick and Chandler 1998; Krieger et al. 2010; Li et al. 2015; Minvielle 1987). High-throughput sequencing technologies have facilitated comprehensive investigations into the molecular basis of heterosis at the whole-genome level. In one study, 1495 hybrid rice varieties and their parental lines were genotyped to identify key genomic factors associated with heterosis (Huang et al. 2015). The findings revealed that overall genome-wide heterozygosity had minimal contribution to heterosis, indicating that a high level of heterozygosity across the genome does not directly correlate with superior hybrid performance. In another genome-wide association study (GWAS) using high-throughput sequencing and phenotypic data from 10,074 F₂ lines derived from 17 representative hybrid rice crosses, the accumulation of numerous rare superior alleles with positive dominance was found to play a major role in heterosis (Huang et al. 2016). The advent of RNA sequencing (RNA-Seq) has further advanced heterosis research by enabling high-resolution transcriptional profiling (Chen et al. 2018a, b; Shen et al. 2015; Yang et al. 2017). Wei et al. conducted a comparative gene expression analysis across seven tissues, including leaves and panicles, in the elite hybrid Liang-You-Pei-Jiu and its parental lines (Wei et al. 2009). They found that the F₁ hybrid exhibited a substantial number of differentially expressed genes, with expression levels significantly higher than those in either parent. Allele-specific expression (ASE) refers to the preferential transcription of one parental allele in a hybrid, while the other remains silent or is expressed at a significantly lower level. This phenomenon can contribute to phenotypic variation and enhanced traits such as improved growth, yield, and stress tolerance. Several studies in species such as maize (Guo et al. 2003, 2006), Arabidopsis (Ng et al. 2014; Zhang and Borevitz 2009), and rice (Chodavarapu et al. 2012; He et al. 2010; Lin et al. 2020) suggest that ASE is a key regulatory mechanism contributing to heterosis. The advancement of RNA sequencing technologies has enabled unbiased and highly reproducible profiling of whole transcriptomes, allowing the detection of single-nucleotide polymorphisms (SNPs) that distinguish parental alleles and facilitate the identification of ASE in heterozygotes (Guo et al. 2008). Lin et al. conducted a genome-wide analysis of ASE by comparing the read ratios of parental alleles in RNA-sequencing data from the elite rice hybrid Shanyou63 and its parents, Zhenshan 97 (ZS97) and Minghui 63 (MH63), using three tissues collected under four distinct environmental conditions (Lin et al. 2020). The analysis identified 3270 genes exhibiting diverse ASE patterns in response to developmental and environmental cues, providing a comprehensive view of the ASE landscape within the hybrid genome. This extensive dataset offers valuable insights into the potential role of ASE in the genetic basis of heterosis, particularly in the context of classical dominance and overdominance models. In the present study, we investigated the late-stage, high-vigor super-hybrid rice variety Xieyou9308 (XY9308), which is characterized by its high grain yield and was designated as a super rice by the Chinese Ministry of Agriculture in 2005. XY9308 was developed from a cross between the restorer line Zhonghui9308 (ZH9308, with 25% japonica genetic composition) and the maternal line Xieqingzao B (XQZB, indica). To explore the genetic basis of the super hybrid rice XY9308, we sequenced 139 recombinant inbred line (RIL) populations derived from XY9308 and constructed an immortalized F2 (IMF2) population comprising 570 hybrid combinations through a pairwise crossing. From these 570 hybrid combinations, 122 samples were randomly selected for RNA-seq analysis. Genome constitution comparisons and GWAS analysis based on binned genotypes were conducted to identify alleles positively associated with increased plant height, heading date, panicle length, grain number per panicle, grain length, 1,000-grain weight, and seed-setting rate. These analyses were followed by ASE profiling to further verify the potential contribution of these factors to the exceptionally high yield of XY9308. Main Text Phenotypic Performance of Ten Agronomic Traits We conducted a comprehensive comparative analysis of the two parental lines, XQZB and ZH9308, across ten key agronomic traits under two distinct environmental conditions. The results revealed significant differences between the two parental lines in seven of the ten traits across both environments, as shown in Fig. 1. These findings underscore the distinct phenotypic profiles of the two parents, with ZH9308 exhibiting superior performance in overall plant architecture and reproductive capacity. ZH9308 exhibited superior performance in traits related to plant architecture and reproductive development, including increased plant height, delayed heading date, longer panicle length, and a higher number of grains per panicle. Conversely, XQZB showed clear advantages in grain-related traits critical for yield and quality, such as greater grain length, higher 1,000-grain weight, and an elevated seed-setting rate. These contrasting trait profiles underscore the complementary genetic potential of the two parental lines. Phenotypic evaluation of the hybrid line XY9308 was restricted to the Fuyang region due to limited seed availability. Statistical analysis (Fig. S1) demonstrated that XY9308 significantly outperformed both parents in plant height, growth duration, panicle length, and overall yield, exhibiting strong heterosis. Comparative images of plant and panicle morphology (Fig. S2) further corroborated the hybrid’s superior agronomic advantages. The pronounced heterosis observed in XY9308 is likely attributed to the integration of favorable alleles from both ZH9308 and XQZB. Variance components were estimated for the ten agronomic traits in both the RIL and IMF2 populations. Broad-sense heritability in the RILs ranged from 14.03% for yield to 89.17% for grain length, while in the IMF2 population, estimates ranged from 3.55% for heading date to 65.27% for 1,000-grain weight. All traits were significantly influenced by environmental factors, as reflected by genotype-by-environment interaction heritability values, which ranged from 8.60% for grain length to 89.68% for heading date. Notably, plant height, heading date, and seed-setting rate exhibited heightened sensitivity to environmental variation (Table 1). To investigate the genetic relationships among traits, we estimated genetic correlations based on phenotypic values predicted using the best linear unbiased prediction (BLUP) method. In both environments, traits such as grain length, grain width, 1,000-grain weight, grain number per panicle, plant height, heading date, seed-setting rate and panicle length showed mostly significant positive correlations in both the RIL and IMF2 populations (Fig. 2B, D). In the RIL population, panicle number showed significant negative correlations with plant height, heading date, panicle length, and grain number per panicle in both environments (FY: −0.38, −0.18, −0.40, −0.55; LS: −0.41, −0.29, −0.33, −0.61). However, in the IMF2 population, these correlations were weakly negative (FY: −0.04, −0.16, −0.07, −0.11; LS: −0.10, −0.07, −0.11, −0.33). Furthermore, grain number per panicle exhibited weak but significant negative correlations with grain length in both populations across the two environments (FY: −0.16, −0.18; LS: −0.24, −0.12), but no significant correlation with grain width. All agronomic traits exhibited positive correlations with yield in both the RIL and IMF2 populations. Among these, panicle number exhibited the strongest positive correlation with yield in the IMF2 population, with coefficients of 0.71 in FY and 0.59 in LS. In LS, both RIL and IMF2 populations showed positive correlations between heading date and seed-setting rate (0.54 and 0.18, respectively), whereas in FY, these correlations were negative (−0.19 and − 0.53, respectively). The correlations between heading date and grain number per panicle, panicle length, and grain width demonstrated notable differences across the two environments and populations. For example, in FY, the correlation coefficients between heading date and grain number per panicle in the RIL and IMF2 populations were 0.19 and 0, respectively, while in LS, they were 0.44 and 0.47, respectively. This highlights that heading date is significantly influenced by environmental conditions, consistent with the findings from the variance component analysis. In general, the correlation patterns among the ten agronomic traits in both the RIL and IMF2 populations were largely consistent across the two environments. The genetic correlation patterns (Fig. 2E, F) were also consistent with the phenotypic correlation patterns, indicating that these traits are governed by stable underlying genetic relationships. This coherence underscores the potential for coordinated genetic regulation and warrants further investigation. Fig. 1Phenotypic comparison between XQZB and ZH9308 for ten traits in two environments (FY: Fuyang in red, LS: Lingshui in green). ns: p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 Phenotypic comparison between XQZB and ZH9308 for ten traits in two environments (FY: Fuyang in red, LS: Lingshui in green). ns: p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 Table 1Analysis of variance for ten agronomic traits in two environmentsPopTraitsVariance Components (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma}$$\end{document}2)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{h}}_{{g}}^{2}$$\end{document}(%)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{h}}_{{g}{e}}^{2}$$\end{document}(%) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\sigma}}_{g}^{2}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\sigma}}_{e}^{2}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\sigma}}_{ge}^{2}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\sigma}}_{{\upepsilon}}^{2}$$\end{document} RILPH91.4023.1051.1116.1057.6232.22HD32.4796.5342.845.2240.3353.19PN2.207.791.745.2324.0018.96PL4.035.631.680.6763.1726.27GL0.290.030.030.0189.178.60GW1.69E-021.79E-022.38E-031.35E-0381.9211.54TGW7.260.751.150.4482.0912.98GNP732.6929.28309.75299.6054.6023.08SSR3.00E-031.28E-036.04E-032.33E-0326.4053.14YD10.9462.7719.7547.3114.0325.32IMF2PH67.3955.3253.2529.2644.9635.52HD4.5840.16115.788.753.5589.68PN1.310.053.4810.858.4022.23PL2.495.251.381.0151.0728.18GL0.120.020.080.0352.3336.31GW8.61E-031.92E-021.01E-023.10E-0339.5246.26TGW3.861.061.150.9165.2719.38GNP508.5917.71398.08644.6932.7825.66SSR2.69E-031.19E-035.14E-035.45E-0320.2238.73YD18.160.0080.34194.706.1927.40 Table 1Analysis of variance for ten agronomic traits in two environmentsPopTraitsVariance Components (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma}$$\end{document}2)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{h}}_{{g}}^{2}$$\end{document}(%)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{h}}_{{g}{e}}^{2}$$\end{document}(%) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\sigma}}_{g}^{2}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\sigma}}_{e}^{2}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\sigma}}_{ge}^{2}$$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\sigma}}_{{\upepsilon}}^{2}$$\end{document} RILPH91.4023.1051.1116.1057.6232.22HD32.4796.5342.845.2240.3353.19PN2.207.791.745.2324.0018.96PL4.035.631.680.6763.1726.27GL0.290.030.030.0189.178.60GW1.69E-021.79E-022.38E-031.35E-0381.9211.54TGW7.260.751.150.4482.0912.98GNP732.6929.28309.75299.6054.6023.08SSR3.00E-031.28E-036.04E-032.33E-0326.4053.14YD10.9462.7719.7547.3114.0325.32IMF2PH67.3955.3253.2529.2644.9635.52HD4.5840.16115.788.753.5589.68PN1.310.053.4810.858.4022.23PL2.495.251.381.0151.0728.18GL0.120.020.080.0352.3336.31GW8.61E-031.92E-021.01E-023.10E-0339.5246.26TGW3.861.061.150.9165.2719.38GNP508.5917.71398.08644.6932.7825.66SSR2.69E-031.19E-035.14E-035.45E-0320.2238.73YD18.160.0080.34194.706.1927.40 Analysis of variance for ten agronomic traits in two environments Traits abbreviations: PH, HD, PL, PN, GL, GW, GNP, TGW, SSR and YD denote plant height, heading date, panicle length, panicle number, grain length, grain width, grain number per panicle, 1,000-grain weight, seed-setting rate and yield, respectively. Variance Components (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\sigma\:$$\end{document}2): \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{g}^{2}$$\end{document} is for genotypic variance, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{e}^{2}$$\end{document} for environmental variance, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{ge}^{2}$$\end{document} for genotype-by-environment interaction variance, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{{\upepsilon\:}}^{2}$$\end{document} for error variance, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{g\:}^{2}$$\end{document} for heritability: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{g}^{2}$$\end{document}/(\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{g}^{2}$$\end{document}+\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{ge}^{2}$$\end{document}+\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{{\upepsilon\:}}^{2}$$\end{document}), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{ge}^{2}$$\end{document} for genotype-by-environment interaction heritability: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{ge}^{2}$$\end{document}/(\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{g}^{2}$$\end{document}+\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{ge}^{2}$$\end{document}+\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{{\upepsilon\:}}^{2}$$\end{document}). Fig. 2Phenotypic and genetic correlations between traits in the RIL and IMF2 population. A, B Heat map showing the phenotypic correlation coefficients between ten traits in FY (A) and LS (B) for RIL population. D, E Heat map showing the phenotypic correlation coefficients between ten traits in FY (D) and LS (E) for IMF2 population. C, F Heat map showed the genetic correlation coefficients between ten traits in the RIL (C) and IMF2 (F) population. *, **, *** denote significance level at 0.05, 0.01 and 0.001, respectively Phenotypic and genetic correlations between traits in the RIL and IMF2 population. A, B Heat map showing the phenotypic correlation coefficients between ten traits in FY (A) and LS (B) for RIL population. D, E Heat map showing the phenotypic correlation coefficients between ten traits in FY (D) and LS (E) for IMF2 population. C, F Heat map showed the genetic correlation coefficients between ten traits in the RIL (C) and IMF2 (F) population. *, **, *** denote significance level at 0.05, 0.01 and 0.001, respectively Genomic Architecture and Bin Map Construction of RIL Populations SNP calling was performed using whole-genome resequencing data from the two parental lines (at 100× coverage) and 139 RILs (at 10× coverage), with the ZS97 genome serving as the reference. A total of 3,357,331 SNPs and indels were initially identified. Variants were then filtered based on the following criteria: homozygosity in both parents, polymorphism between the parental lines, genotype missing rate < 10%, heterozygosity rate < 20%, and minor allele frequency (MAF) > 0.2. After filtering, a total of 1,300,425 high-quality variants were retained for downstream analysis, including 1,104,349 SNPs and 196,076 indels. Variant annotation was performed using SnpEff. Among the genic SNPs, 24.54% were distributed within exons which contained 142,630 moderate-to-high effect, and 116,828 low-effect variants, 56.58% within introns, 10.81% within 5’-untranslated regions (UTRs) and 8.71% within 3’-UTRs. Of the intergenic SNPs, 43.27% are located in putative regulatory regions (PPRs) within 2 kb of a transcriptional start site. The snpbinner software was used to construct the bin maps (Fig. S3), and a total of 3,818 bin markers were obtained. The genotype and bin marker information of IMF2 are inferred from the mating information of the RIL population. Correlation Between japonica-related Lineage Proportions and Agronomic Traits Previous studies have demonstrated that the proportion of japonica-derived genomic segments is significantly associated with multiple agronomic traits in RIL populations (Li et al. 2018; Yu et al. 2020; Zhang et al. 2020). The RIL population in our study was developed from a cross between the indica variety XQZB and the japonica restorer line ZH9308. To assess the impact of japonica genomic contributions, we quantified the proportion of japonica-related segments inherited from ZH9308 across each chromosome in all RILs and analyzed their correlations with agronomic traits under two environmental conditions. Our results revealed that grain length and 1,000-grain weight were significantly negatively correlated with japonica-related lineage proportions on chromosomes 3 and 6 in both environments (Fig. 3A, B). Specifically, correlation coefficients for grain length were − 0.67, −0.64, −0.29, and − 0.22, while those for 1000-grain weight were − 0.52, −0.47, −0.40, and − 0.28. In contrast, grain number per panicle exhibited significant positive correlations with japonica lineage proportions on chromosome 3 in both environments (r = 0.34 and 0.38), and on chromosome 8 (r = 0.29 in FY and 0.24 in LS). Panicle number generally exhibited negative correlations with the proportion of japonica-related genomic segments. In the FY environment, the correlation coefficients were − 0.19 for chromosome 7 and − 0.26 for chromosome 8. In the LS environment, negative correlations were observed on chromosomes 3, 5, 7, 8, and 12, with coefficients of −0.28, −0.27, −0.28, −0.28, and − 0.19, respectively. Similar negative correlation patterns were also evident for grain length and 1,000-grain weight. In contrast, traits such as plant height, heading date, panicle length, and grain number per panicle were predominantly positively correlated with japonica-related lineage across multiple chromosomes. For instance, in the FY environment, heading date showed positive correlations with japonica-derived segments on chromosomes 1, 7, and 12, with coefficients of 0.21, 0.30, and 0.27, respectively. In the LS environment, heading date exhibited a strong positive correlation with the japonica segment on chromosome 10, with a coefficient of 0.38. These results suggest that ZH9308 contributes positively to traits such as plant height, heading date, panicle length, and grain number per panicle, whereas XQZB provides alleles favorable for grain length and 1,000-grain weight. This pattern is consistent with the distinct phenotypic strengths of the two parental lines and supports the hypothesis that complementary allele contributions underline the observed heterosis in XY9308 . Similarly, the correlation between the japonica-related lineage proportion in the IMF2 population and agronomic traits was analyzed under both environmental conditions (Fig. 3C, D). Consistent with the findings in the RIL population, grain length and 1,000-grain weight exhibited significant negative correlations with the japonica-related lineage on chromosome 3 in both environments. Specifically, the correlation coefficients for grain length were − 0.50 and − 0.39, while those for 1,000-grain weight were − 0.52 and − 0.47. In addition, grain length and 1,000-grain weight showed strong negative correlations with the japonica-related lineage on chromosome 6 under the FY environment, with coefficients of −0.23 and − 0.25, respectively. Moreover, heading date demonstrated strong positive correlations with the japonica-related lineage on chromosomes 7 and 8 in the FY environment (r = 0.36 and 0.16, respectively), but negative correlations with chromosomes 2 and 9 (r = − 0.20 and − 0.17). In the LS environment, heading date was positively correlated with the japonica-related lineage on chromosome 10, with a coefficient of 0.35. Furthermore, plant height showed strong positive correlations with japonica-related segments on chromosomes 7 and 8 in the FY environment (r = 0.31 and 0.32), and weaker but still positive correlations on chromosomes 1, 5, 8, and 10 in LS (r = 0.14, 0.14, 0.19, and 0.18). In contrast, plant height exhibited consistent negative correlations with chromosome 2 across both environments (r = − 0.28 and − 0.32). In addition, panicle length was positively associated with japonica-related lineage on chromosomes 5 and 6 under the FY environment (r = 0.24 and 0.27) and on chromosomes 1 and 5 in the LS environment (r = 0.34 and 0.25). Conversely, negative correlations were observed with chromosomes 2 and 9 in both environments (FY: − 0.19 and − 0.23; LS: − 0.20 and − 0.21). Similarly, grain number per panicle exhibited strong positive correlations with the japonica-related lineage on chromosomes 3 and 8 across both environments (FY: 0.32 and 0.24; LS: 0.18 and 0.19), and negative correlations with chromosome 2 (r = − 0.20 in FY and − 0.15 in LS). Furthermore, this study also analyzed the correlation between the indica-japonica lineage rate (heterozygosity) on each chromosome of the IMF2 population and agronomic traits in two environments (Fig. 3E, F). The results showed that most agronomic traits had no significant or weak correlations with the indica-japonica lineage rate on each chromosome. Grain length showed a strong positive correlation with the indica-japonica lineage rate on chromosome 3 in both the FY and LS environments, with correlation coefficients of 0.28 and 0.25, respectively, but the correlation was much lower than that with the japonica-related lineage rate, indicating that the genes controlling grain length on chromosome 3 mainly have additive effects. Grain width showed a strong positive correlation with the indica-japonica lineage rate on chromosome 7 in the FY environment, with a correlation coefficient of 0.20, but no significant correlation with the japonica-related lineage rate on chromosome 7, suggesting that the genes controlling grain width on chromosome 7 may have a dominant effect. Fig. 3Correlation between ZH9308 and the heterozygous ratio of each chromosome with ten traits across two environments. A, B Heat maps showing the correlation between ZH9308 ratio of each chromosome and ten traits in the RIL population at the FY (A) and LS (B) loci. C, D Heat maps showing the correlation between ZH9308 ratio of each chromosome and ten traits in the IMF2 population at the FY (C) and LS (D) loci. E, F Heat maps showing the correlation of the heterozygous ratio with ten traits in the IMF2 population at the FY (E) and LS (F) loci Correlation between ZH9308 and the heterozygous ratio of each chromosome with ten traits across two environments. A, B Heat maps showing the correlation between ZH9308 ratio of each chromosome and ten traits in the RIL population at the FY (A) and LS (B) loci. C, D Heat maps showing the correlation between ZH9308 ratio of each chromosome and ten traits in the IMF2 population at the FY (C) and LS (D) loci. E, F Heat maps showing the correlation of the heterozygous ratio with ten traits in the IMF2 population at the FY (E) and LS (F) loci Identification of QTLs Across RIL Population Based on Bin Map Using the high-quality bin map, QTL association mapping for ten agronomic traits in the RIL population was conducted using QTXNetwork software, which employed a mixed linear model to detect both additive and epistatic effects, as well as their interactions with the environment. In total, 43 additive QTLs, as reported in Table 2, and three pairs of epistatic QTLs, described in Table 3, were identified. Chromosome 3 contained the greatest number of QTLs, with 12 loci detected, and followed by chromosome 6 with seven QTLs and chromosome 9 with five. The remaining chromosomes generally harbored between two and three QTLs each. Of note, the positive or negative effects of loci depend on the coding of genetic variants. In this study we coded the additive coefficient of paternal allele homozygotes (QQ, ZH9308) as “1” and the maternal allele homozygotes (qq, XQZB) as “−1”. The reported effect is for the genotype of QQ and it should be the opposite value for genotype of qq. A total of four QTLs exhibiting additive effects were identified for plant height. Among them, qPH6, qPH10, and qPH11, located on chromosomes 6, 10, and 11, respectively, had additive effects of 3.126, 3.446, and − 3.015, accounting for 13.12%, 14.72%, and 11.27% of the phenotypic variance. Notably, qPH10 was environmentally sensitive, with its additive-by-environment interaction contributing an additional 4.15% to variation. Similarly, four QTLs with additive effects were detected for heading date. Of these, qHD1 and qHD6 showed additive effects of 2.506 and − 2.210, explaining 12.87% and 10.01% of the variance, respectively. Two other QTLs, qHD7 and qHD10, exhibited significant additive–additive interaction effects (−1.341), contributing 7.37% to phenotypic variation. Both qHD7 and qHD10 also displayed substantial additive-by-environment interactions, explaining 4.32% and 17.26% of the total variance, respectively. For grain length, seven QTLs exhibiting additive effects were identified. Among these, qGL3-2 and qGL9-1 demonstrated significant additive-additive interaction effects (0.1044), jointly accounting for 11.11% of the phenotypic variation. Notably, qGL3-2 was identified as a major-effect locus, with an additive effect of −0.294 and a substantial contribution of 43.93% to grain length variation. Similarly, four QTLs with additive effects were detected for grain width. Of these, three loci located on chromosomes 3, 4, and 6 contributed 17.30%, 15.13%, and 15.75% to phenotypic variance, respectively, and were considered major contributors. Furthermore, four QTLs were identified with additive effects for grain number per panicle. Among them, qGNP3-2 and qGNP8 showed strong additive–additive interaction effects, contributing a combined 17.56% to phenotypic variation. In addition, seven QTLs influencing panicle length were also identified. A major locus on chromosome 6 exhibited an additive effect of 0.783 and accounted for 21.96% of the variation. Additionally, qPL3 was detected solely through its additive-by-environment interaction effects, with interaction values of 0.307 and − 0.304, highlighting the influence of environmental factors on panicle length expression. For panicle number, four QTLs exhibiting additive effects were identified. Two were located on chromosome 3, one on chromosome 8, and one on chromosome 11, contributing 10.09%, 16.12%, 6.28%, and 13.95% to phenotypic variance, respectively. Six loci were associated with 1,000-grain weight through additive effects. Among them, qTGW3-2 exhibited negative effect (−1.073), explaining 38.02% of the phenotypic variation. Finally, two loci influencing seed-setting rate and one affecting yield were identified, with contribution rates of 10.82%, 17.43%, and 16.01%, respectively. Apart from plant height, heading date, and panicle length, no significant additive-by-environment interaction effects were observed for the remaining seven agronomic traits. This supports earlier findings, reinforcing that the genetic regulation of plant height and heading date is highly sensitive to environmental conditions. In total, seven pleiotropic QTLs were identified. On chromosome 3, two regions showed overlapping effects on 1000-grain weight and grain length: the region from 17,247,867 to 17,696,465 (qTGW3-1 and qGL3-2) accounted for 38.02% and 43.93% of phenotypic variation, respectively, while the region from 29,841,590 to 30,158,060 (qTGW3-2 and qGL3-4) explained 4.37% and 1.36%, respectively. All four loci showed negative additive effects (−1.073, −0.294, −0.364, and − 0.052), supporting the strong positive phenotypic (0.74 and 0.73) and genetic (0.73) correlations between 1,000-grain weight and grain length. Another pleiotropic QTL on chromosome 3 (30,746,025–30,816,291), encompassing qGNP3-2 and qPN3-1, explained 11.24% of the variation in grain number per panicle and 10.09% in panicle number. These loci exhibited opposite additive effects (8.733 and − 0.375), providing genetic evidence for the strong negative genetic correlation (−0.67) between these two traits. A multi-trait QTL on chromosome 11 (769,292–882,097), containing qPH11, qGNP11, and qPN11, explained 11.27%, 6.62%, and 13.95% of the variation in plant height, grain number per panicle, and panicle number, respectively. The corresponding additive effects (−3.015, −6.699, and 0.441) suggest that this locus enhances tillering and panicle production while limiting vertical growth and grain number per panicle. Additionally, a QTL on chromosome 10 (19,087,711 − 19,951,550), comprising qPH10 and qHD10, exhibited significant additive and additive-by-environment interaction effects on both plant height and heading date, with consistent effect directions across environments. Based on the estimated effects of the above QTLs, the additive effects of the alleles from the original parents XQZB and ZH9308 can be inferred. In general, alleles from ZH9308 tend to increase plant height, prolong the growth period, promote panicle length, increase the number of grains per panicle, and improve seed-setting rate. In contrast, alleles from XQZB tend to promote tillering, increase panicle number, improve grain length and 1,000-grain weight, and enhance single-plant yield (Fig. S4). Although not all alleles follow this trend, the overall direction of these effects is consistent with the phenotypic characteristics of the two parents: ZH9308 is more favorable for improving growth- and development-related traits, whereas XQZB contributes more to grain quality and yield-related traits. Table 2Effects and heritability of additive QTLs and additive-by-environment interaction detected for ten traits in RIL populationTraitQTLBin markeraae1ae2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\boldsymbol{h}}_{\boldsymbol{a}}^{2}$$\end{document}(%)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\boldsymbol{h}}_{\boldsymbol{a}\boldsymbol{e}}^{2}$$\end{document}(%)Candidate genesPlant height qPH1 Chr01:25902539–25,982,1552.501––7.76–– qPH6 Chr06:8024603–8,263,7653.126––12.12– LOC_Os06g16370 HD1 qPH10 Chr10:19723260–19,951,5503.446−1.8191.83414.724.15 LOC_Os10g32600 Ehd1 qPH11 Chr11:769292–816,186−3.015––11.27––Heading date qHD1 Chr01:23126519–23,299,9092.506––12.87– LOC_Os01g38530 OsEF3 qHD6 Chr06:3256517–3,837,756−2.210––10.01– LOC_Os06g16370 HD1 qHD7 Chr07:9261251–9,778,3222.1491.455−1.4499.474.32– qHD10 Chr10:19087711–19,723,2591.966−2.9432.8617.9217.26 LOC_Os10g32600 Ehd1 Grain length qGL2 Chr02:10529555–17,162,0720.124––7.78– LOC_Os02g25230 SG2 qGL3-1 Chr03:15212685–15,317,172−0.059––1.75–– qGL3-2 Chr03:17588326–17,696,465−0.294––43.93– Os03g0407400 GS3 qGL3-3 Chr03:27824988–27,833,438−0.064––2.08– LOC_Os03g44500 GL3.1 qGL3-4 Chr03:29841590–30,158,060−0.052––1.36–– qGL9-1 Chr09:16948935–18,105,017−0.131––8.8– LOC_Os09g26999 DEP1 qGL9-2 Chr09:22078668–22,122,744−0.082––3.45––Grain width qGW2 Chr02:20164037–20,403,824−0.020––5.28–– qGW3 Chr03:24654255–24,878,8770.036––17.3–– qGW4 Chr04:36216464–36,400,1450.034––15.13–– qGW6 Chr06:22428988–22,436,848−0.034––15.75––Grain number per panicle qGNP3-1 Chr03:14660028–14,780,0729.953––14.6–– qGNP3-2 Chr03:30746025–30,816,2918.733––11.24– LOC_Os03g49880 OsTB1 qGNP8 Chr08:27506145–29,565,3676.073––5.44– LOC_Os08g07740 HD5 qGNP11 Chr11:769292–816,186−6.699––6.62––Panicle length qPL3 Chr03:32105736–32,137,473–0.307−0.304–3.34– qPL4 Chr04:20996354–21,077,8460.350––4.39–– qPL5-1 Chr05:17505091–17,684,0930.333––3.97–– qPL5-2 Chr05:21113936–21,159,5420.341––4.17–– qPL6 Chr06:9759517–9,766,0610.783––21.96–– qPL9-1 Chr09:13696279–13,768,418−0.409––5.99–– qPL9-2 Chr09:21002719–21,021,576−0.358––4.6––Panicle number qPN3-1 Chr03:30746025–30,816,291−0.375––10.09– LOC_Os03g49880 OsTB1 qPN3-2 Chr03:32426246–32,602,075−0.474––16.12–– qPN8 Chr08:9043846–10,125,775−0.296––6.28–– qPN11 Chr11:816187–882,0970.441––13.95––Seed setting rate qSSR9 Chr09:20994341–21,002,7180.018––10.82–– qSSR10 Chr10:18327331–18,424,8670.023––17.43––1,000-grain weight qTGW1 Chr01:6736248–6,892,059−0.442––6.46– LOC_Os01g12390 OsSRLK qTGW3-1 Chr03:17247867–17,522,516−1.073––38.02– Os03g0407400 GS3 qTGW3-2 Chr03:29841590–30,158,060−0.364––4.37–– qTGW6-1 Chr06:1407720–1,577,646−0.471––7.33– LOC_Os06g04200 Wx qTGW6-2 Chr06:22581318–22,974,842−0.263––2.28––- qTGW11 Chr11:2453121–2,482,746−0.399––5.25– LOC_Os11g40590 OsGRDP1 Yield qYD6 Chr06:1918239–2,029,016−1.324−–16.01– LOC_Os06g04540 DP1 QTLs are named in the form of “q”, “Trait”, “chromosome”,“-Rank” and are italicized, for example, qGL3-1denotes a QTL of grain length which is the first QTL on the chromosome 3; a, additive effect; ae, additive-by-environment interaction effects, of which ae1 denotes the interactions between additive and environment FY, ae2 denotes the interactions between additive and environment LS; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{a}^{2}$$\end{document}: the proportion of phenotypic variance explained by the additive QTL. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{ae}^{2}$$\end{document}, the proportion of phenotypic variance explained by the additive-by-environment interaction. The annotation of the candidate genes is derived from the National Rice Data Center (https://www.ricedata.cn/). a = additive effect for paternal allele homozygotes (QQ, ZH9308), -a for maternal allele homozygotes (qq, XQZB) Table 2Effects and heritability of additive QTLs and additive-by-environment interaction detected for ten traits in RIL populationTraitQTLBin markeraae1ae2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\boldsymbol{h}}_{\boldsymbol{a}}^{2}$$\end{document}(%)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\boldsymbol{h}}_{\boldsymbol{a}\boldsymbol{e}}^{2}$$\end{document}(%)Candidate genesPlant height qPH1 Chr01:25902539–25,982,1552.501––7.76–– qPH6 Chr06:8024603–8,263,7653.126––12.12– LOC_Os06g16370 HD1 qPH10 Chr10:19723260–19,951,5503.446−1.8191.83414.724.15 LOC_Os10g32600 Ehd1 qPH11 Chr11:769292–816,186−3.015––11.27––Heading date qHD1 Chr01:23126519–23,299,9092.506––12.87– LOC_Os01g38530 OsEF3 qHD6 Chr06:3256517–3,837,756−2.210––10.01– LOC_Os06g16370 HD1 qHD7 Chr07:9261251–9,778,3222.1491.455−1.4499.474.32– qHD10 Chr10:19087711–19,723,2591.966−2.9432.8617.9217.26 LOC_Os10g32600 Ehd1 Grain length qGL2 Chr02:10529555–17,162,0720.124––7.78– LOC_Os02g25230 SG2 qGL3-1 Chr03:15212685–15,317,172−0.059––1.75–– qGL3-2 Chr03:17588326–17,696,465−0.294––43.93– Os03g0407400 GS3 qGL3-3 Chr03:27824988–27,833,438−0.064––2.08– LOC_Os03g44500 GL3.1 qGL3-4 Chr03:29841590–30,158,060−0.052––1.36–– qGL9-1 Chr09:16948935–18,105,017−0.131––8.8– LOC_Os09g26999 DEP1 qGL9-2 Chr09:22078668–22,122,744−0.082––3.45––Grain width qGW2 Chr02:20164037–20,403,824−0.020––5.28–– qGW3 Chr03:24654255–24,878,8770.036––17.3–– qGW4 Chr04:36216464–36,400,1450.034––15.13–– qGW6 Chr06:22428988–22,436,848−0.034––15.75––Grain number per panicle qGNP3-1 Chr03:14660028–14,780,0729.953––14.6–– qGNP3-2 Chr03:30746025–30,816,2918.733––11.24– LOC_Os03g49880 OsTB1 qGNP8 Chr08:27506145–29,565,3676.073––5.44– LOC_Os08g07740 HD5 qGNP11 Chr11:769292–816,186−6.699––6.62––Panicle length qPL3 Chr03:32105736–32,137,473–0.307−0.304–3.34– qPL4 Chr04:20996354–21,077,8460.350––4.39–– qPL5-1 Chr05:17505091–17,684,0930.333––3.97–– qPL5-2 Chr05:21113936–21,159,5420.341––4.17–– qPL6 Chr06:9759517–9,766,0610.783––21.96–– qPL9-1 Chr09:13696279–13,768,418−0.409––5.99–– qPL9-2 Chr09:21002719–21,021,576−0.358––4.6––Panicle number qPN3-1 Chr03:30746025–30,816,291−0.375––10.09– LOC_Os03g49880 OsTB1 qPN3-2 Chr03:32426246–32,602,075−0.474––16.12–– qPN8 Chr08:9043846–10,125,775−0.296––6.28–– qPN11 Chr11:816187–882,0970.441––13.95––Seed setting rate qSSR9 Chr09:20994341–21,002,7180.018––10.82–– qSSR10 Chr10:18327331–18,424,8670.023––17.43––1,000-grain weight qTGW1 Chr01:6736248–6,892,059−0.442––6.46– LOC_Os01g12390 OsSRLK qTGW3-1 Chr03:17247867–17,522,516−1.073––38.02– Os03g0407400 GS3 qTGW3-2 Chr03:29841590–30,158,060−0.364––4.37–– qTGW6-1 Chr06:1407720–1,577,646−0.471––7.33– LOC_Os06g04200 Wx qTGW6-2 Chr06:22581318–22,974,842−0.263––2.28––- qTGW11 Chr11:2453121–2,482,746−0.399––5.25– LOC_Os11g40590 OsGRDP1 Yield qYD6 Chr06:1918239–2,029,016−1.324−–16.01– LOC_Os06g04540 DP1 QTLs are named in the form of “q”, “Trait”, “chromosome”,“-Rank” and are italicized, for example, qGL3-1denotes a QTL of grain length which is the first QTL on the chromosome 3; a, additive effect; ae, additive-by-environment interaction effects, of which ae1 denotes the interactions between additive and environment FY, ae2 denotes the interactions between additive and environment LS; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{a}^{2}$$\end{document}: the proportion of phenotypic variance explained by the additive QTL. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{ae}^{2}$$\end{document}, the proportion of phenotypic variance explained by the additive-by-environment interaction. The annotation of the candidate genes is derived from the National Rice Data Center (https://www.ricedata.cn/). a = additive effect for paternal allele homozygotes (QQ, ZH9308), -a for maternal allele homozygotes (qq, XQZB) Effects and heritability of additive QTLs and additive-by-environment interaction detected for ten traits in RIL population LOC_Os06g16370 HD1 LOC_Os10g32600 Ehd1 LOC_Os01g38530 OsEF3 LOC_Os06g16370 HD1 LOC_Os10g32600 Ehd1 LOC_Os02g25230 SG2 Os03g0407400 GS3 LOC_Os03g44500 GL3.1 LOC_Os09g26999 DEP1 LOC_Os03g49880 OsTB1 LOC_Os08g07740 HD5 LOC_Os03g49880 OsTB1 LOC_Os01g12390 OsSRLK Os03g0407400 GS3 LOC_Os06g04200 Wx LOC_Os11g40590 OsGRDP1 LOC_Os06g04540 DP1 QTLs are named in the form of “q”, “Trait”, “chromosome”,“-Rank” and are italicized, for example, qGL3-1denotes a QTL of grain length which is the first QTL on the chromosome 3; a, additive effect; ae, additive-by-environment interaction effects, of which ae1 denotes the interactions between additive and environment FY, ae2 denotes the interactions between additive and environment LS; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{a}^{2}$$\end{document}: the proportion of phenotypic variance explained by the additive QTL. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{ae}^{2}$$\end{document}, the proportion of phenotypic variance explained by the additive-by-environment interaction. The annotation of the candidate genes is derived from the National Rice Data Center (https://www.ricedata.cn/). a = additive effect for paternal allele homozygotes (QQ, ZH9308), -a for maternal allele homozygotes (qq, XQZB) Table 3Heritability of additive-additive and additive-additive by environment interaction effects of QTLs detected for agronomic traits in RIL populationTraitQTLiBin marker iQTLjBin marker jaaaae1aae2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\boldsymbol{h}}_{\boldsymbol{a}\boldsymbol{a}}^{2}$$\end{document}(%)HD qHD1 Chr01:23126519–23299909 qHD10 Chr10:19087711–19723259−1.341−−7.37GL qGL3-2 Chr03:17588326–17696465 qGL9-1 Chr09:16948935–181050170.104−−11.11GNP qGNP3-2 Chr03:30746025–30816291 qGNP8 Chr08:27506145–295653677.718−−17.56aa, additive-additive effect; ae, additive-additive by environment interaction effects, of which aae1 denotes the interactions between additive-additive and environment FY, aae2 denotes the interactions between additive-additive and environment LS; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{aa}^{2}$$\end{document}: the proportion of phenotypic variance explained by the additive-additive QTL. Table 3Heritability of additive-additive and additive-additive by environment interaction effects of QTLs detected for agronomic traits in RIL populationTraitQTLiBin marker iQTLjBin marker jaaaae1aae2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\boldsymbol{h}}_{\boldsymbol{a}\boldsymbol{a}}^{2}$$\end{document}(%)HD qHD1 Chr01:23126519–23299909 qHD10 Chr10:19087711–19723259−1.341−−7.37GL qGL3-2 Chr03:17588326–17696465 qGL9-1 Chr09:16948935–181050170.104−−11.11GNP qGNP3-2 Chr03:30746025–30816291 qGNP8 Chr08:27506145–295653677.718−−17.56aa, additive-additive effect; ae, additive-additive by environment interaction effects, of which aae1 denotes the interactions between additive-additive and environment FY, aae2 denotes the interactions between additive-additive and environment LS; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{aa}^{2}$$\end{document}: the proportion of phenotypic variance explained by the additive-additive QTL. Heritability of additive-additive and additive-additive by environment interaction effects of QTLs detected for agronomic traits in RIL population aa, additive-additive effect; ae, additive-additive by environment interaction effects, of which aae1 denotes the interactions between additive-additive and environment FY, aae2 denotes the interactions between additive-additive and environment LS; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{aa}^{2}$$\end{document}: the proportion of phenotypic variance explained by the additive-additive QTL. Identification of QTLs Across IMF2 Population Using the bin marker and mating information derived from the RIL population, the bin genotypes of the IMF2 population were inferred. QTL association mapping for 10 agronomic traits in the IMF2 population was performed using QTXNetwork software, which estimated the additive, dominant, epistatic effects of the QTLs and their interaction effects with the environments (Tables S1, S2 and S3). A total of 118 significant QTLs were identified for the 10 agronomic traits, including 11 pairs of epistatic QTLs. A total of 22 QTLs were identified on chromosome 3, representing the highest number among all chromosomes, followed by 18 QTLs on chromosome 2. Chromosomes 1 and 6 each harbored 14 QTLs, while 10 QTLs were detected on both chromosomes 7 and 10. In contrast, chromosome 12 exhibited the lowest number of QTLs, with only two loci identified. A total of 11 significant QTLs were identified for heading date, including six pairs that exhibited significant epistatic effects. The additive and dominance effects of these QTLs accounted for 0.25% to 2.30% and 0.13% to 1.93% of the phenotypic variation, respectively. The environmental interaction effects of these genetic components ranged from 0.46% to 2.55% for additive effects and 0.14% to 2.11% for dominance effects. Among them qHD7 and qHD8-2 exhibited significant additive and dominance-by-environment interaction effects, suggesting that the expression of these two QTLs in phenotype is greatly influenced by the environment. Moreover, qHD3 and qHD6 showed negative partial dominance, with dominance/|additive| ratios of −0.56 and − 0.61, suggesting small dominance effects for both loci. qHD7 and qHD8-2 exhibited overdominance, with dominance/|additive| ratios of 1.52 and − 3.78, respectively, indicating that dominance effects exceeded additive effects. Notably qHD7 exhibited positive overdominance and qHD8-2 exhibited negative overdominance. Three QTLs associated with heading date, qHD2-1, qHD10-1, and qHD10-2, exhibited significant dominance effects but no significant additive effects, indicating that their expression is primarily governed by dominance. The epistatic effects and their environmental interactions explained 40.35% and 23.20% of the phenotypic variation for the heading date. Notably, qHD3 and qHD8-3 showed significant additive-additive and dominance-dominance interactions. Moreover, QTLs qHD1, qHD7, and qHD10-2 were also identified in the RIL population, showing consistent patterns of environmental interaction across both populations. A total of 16 significant QTLs were identified for plant height, exhibiting notable additive effects. Among these, qPH2-2, qPH2-3, qPH3, qPH4, qPH8-2, qPH11-1, and qPH11-2 exhibited only additive contributions. A total of eight QTLs exhibited dominance effects, with six showing partial dominance and two showing overdominance. Among them, qPH2-1 and qPH10-2 had dominance/|additive| ratios of 1.90 and 3.87, respectively. qPH10-3 was a major-effect QTL with an additive effect of 5.912 and contributing 15.28% to the phenotypic variance. Its gene expression was influenced by environmental conditions, as reflected by environmental interaction effects of −2.749 and 2.659. In addition, significant epistatic interactions were observed between qPH10-2 and qPH10-3, including additive-by-dominance, dominance-by-additive, and dominance-by-dominance effects. These interactions contributed 9.70%, 11.58%, and 5.42% to the phenotypic variation, with effect sizes of 4.710, − 5.147, and − 4.978, respectively. Both qPH6 and qPH10-3 were also detected in the RIL population, reinforcing their relevance across genetic backgrounds. For grain length, 16 significant QTLs and two pairs of epistatic QTLs were detected. Among these, 15 QTLs exhibited significant additive effects, with qGL1-3, qGL3-3, qGL4, qGL9, and qGL12 showing only additive effects without any detectable dominance component. Additionally, eight QTLs exhibited dominance effects, including three with partial dominance and four with overdominance. QTLs qGL1-2 and qGL6-1 showed positive overdominance, with dominance/|additive| ratios of 1.26 and 1.42, respectively, while qGL2-1 and qGL2-2 showed negative overdominance with ratios of −1.07 and − 1.22. qGL3-1 and qGL3-2 explained a relatively high proportion of phenotypic variation for grain length, with additive effect values of −0.199 and − 0.156, contributing 8.43% and 5.15%, respectively. Furthermore, qGL3-1 and qGL3-2 exhibited significant additive-by-additive, additive-by-dominance, and dominance-by-additive epistatic effects, with effect values of 0.174, 0.237, and − 0.109, contributing 12.83%, 11.94%, and 2.54% to the phenotypic variation, respectively. The phenotypic expression of the QTLs qGL3-2 and qGL6-1 was influenced by environmental factors, with their additive-by-environment interactions contributing 6.61% and 6.44% to the phenotypic variation, respectively. qGL2-2, qGL3-1, qGL3-3, and qGL9 were also detected in the RIL population. A total of 14 significant QTLs were identified for grain width. Among these, eight QTLs exhibited significant dominance effects, including two with partial dominance and five with overdominance. Moreover, qGW3-5 and qGW7-1 exhibited positive overdominance, with dominance/|additive| ratios of 1.05 and 1.23, respectively. In contrast, qGW3-1, qGW3-3, and qGW6-1 showed negative overdominance, with ratios of − 1.91, − 1.58, and − 1.27, respectively. QTL qGW6-2 exhibited the highest additive effect among the identified loci, with an effect value of −0.035, accounting for 13.18% of the phenotypic variation in grain width. This QTL was also confirmed in the RIL population. Moreover, no significant additive-by-environment interaction effects were detected for qGW6-2, suggesting that its gene expression is environmentally stable and genetically robust. Total nine QTLs were detected for grain number per panicle, with only qGNP10-2 exhibiting significant additive-by-environment interaction effects. This suggests that the gene controlling grain number per panicle is stably expressed in both FY and LS environments. All nine QTLs exhibited significant additive effects, with their contributions to phenotypic variance ranging from 0.78% to 20.17%. Four QTLs exhibited dominance effects, including three with partial dominance and one with negative overdominance (d/|a| = − 1.96). Five QTLs contributed more than 5% to the phenotypic variation, with qGNP3-2 contributing the most at 20.17% and an additive effect value of 12.442, which was also confirmed in the RIL population. The next highest contributors were qGNP10-1, with a contribution of 10.48% and an additive effect value of 8.971. A total of 22 QTLs associated with panicle length were identified, with no detection of epistatic interactions. Of these, nine QTLs exhibited dominance effects, all with dominance-to-additive ratios below 1, indicating partial dominance. Therefore, genetic improvement of panicle length should prioritize the selection of superior homozygous genotypes. QTL qPL6-1 contributed the most to phenotypic variation in panicle length (12.33%), with an additive effect of 0.668. This QTL also exhibited significant additive-by-environment and dominance-by-environment interactions, with additive-by-environment effect values of − 0.157 and 0.157, and dominance-by-environment effect values of 0.1779 and − 0.1751. These findings suggest that the phenotypic expression associated with qPL6-1 is influenced by environmental conditions. Notably, QTLs qPL4-1, qPL5-2, qPL6-1, qPL9-1, qPL9-2 were consistently identified in the RIL population. Six significant QTLs and one epistatic QTL pair were identified for panicle number. QTL qPN3 contributed the most to phenotypic variation in panicle number (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{a}^{2}$$\end{document}= 17.28%) with an additive effect of −0.739. This QTL was also detected in the RIL population. Four QTLs exhibited dominance effects, among which QTL qPN10 showed negative overdominance (d/|a| = −2.75), suggesting that selection should favor superior homozygous genotypes. The remaining three QTLs demonstrated partial dominance. Significant additive-by-dominance and dominance-by-dominance epistatic interactions were observed between QTLs qPN5 and qPN10, contributing 3.07% and 9.30%, respectively, to the phenotypic variance in panicle number. A total of 19 QTLs and one epistatic QTL pair were identified for 1,000-grain weight. 11 QTLs exhibited significant additive and dominance effects, with six showing partial dominance and three exhibiting positive overdominance. QTL qTGW1-3 contributed the most to phenotypic variation in 1,000-grain weight, with an additive effect of − 0.698 (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{a}^{2}$$\end{document}= 10.84%), a dominance effect of 0.7564 (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{d}^{2}$$\end{document}= 6.27%), and additive-by-environment interaction effects of − 0.144 and 0.139 (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{ae}^{2}$$\end{document}= 0.44%). These results indicate that qTGW1-3 enhances grain weight in the LS environment, with a net additive contribution of 0.837. Additionally, QTLs qTGW1-1 and qTGW3-3 each accounted for over 5% of the phenotypic variation in this trait. Three QTLs associated with seed setting rate were identified, each contributing more than 5% to the phenotypic variation. QTL qSSR1 exhibited both additive (−0.020) and dominance (0.008) effects, explaining 15.05% and 1.27% of the phenotypic variation, respectively. It also showed significant additive-by-environment interaction effects, increasing seed setting rate in the FY environment (0.009) and decreasing it in LS (−0.009). QTL qSSR2 exhibited significant additive effects, accounting for 10.30% of the phenotypic variation. Moreover, QTL qSSR3 exhibited significant additive effects and dominance-by-environment interaction effects. For yield, two QTLs were identified, both exhibiting significant additive effects. QTLs qYD1 and qYD6 had additive effects of 1.665 and − 1.808, accounting for 8.93% and 10.53% of the phenotypic variation, respectively. Epistatic effects are widely recognized as a key component of heterosis. Based on estimated additive effects in the IMF2 population, which were confirmed in the RIL population, ZH9308 alleles at most QTLs showed positive additive effects for traits such as plant height, heading date, grain number per panicle, and panicle length. In contrast, XQZB alleles enhanced grain length, 1000-grain weight, and seed setting rate. These allele–trait associations were consistent with the phenotypic advantages observed in the respective parents, suggesting that targeted mining and utilization of superior alleles has the potential to further improve yield performance (Fig. 4). We also analyzed the genetic effects of heterozygous loci in the IMF2 population and found that relatively few alleles exhibited overdominance effects (Fig. 5). Fig. 4Additive effects of the QTL detected by ten agronomic traits in IMF2 population. The horizontal axis displays the detected QTLs, while the vertical axis represents the additive effects of the parental alleles of the QTLs, XQZB in red, ZH9308 in green. Traits abbreviations: PH, HD, PL, PN, GL, GW, GNP, TGW, SSR and YD denote plant height, heading date, panicle length, panicle number, grain length, grain width, grain number per panicle, 1,000-grain weight, seed-setting rate and yield, respectively Additive effects of the QTL detected by ten agronomic traits in IMF2 population. The horizontal axis displays the detected QTLs, while the vertical axis represents the additive effects of the parental alleles of the QTLs, XQZB in red, ZH9308 in green. Traits abbreviations: PH, HD, PL, PN, GL, GW, GNP, TGW, SSR and YD denote plant height, heading date, panicle length, panicle number, grain length, grain width, grain number per panicle, 1,000-grain weight, seed-setting rate and yield, respectively Fig. 5Dominance/|Additive| effects of the QTL detected by ten agronomic traits in IMF2 population Dominance/|Additive| effects of the QTL detected by ten agronomic traits in IMF2 population Candidate Genes in QTL Regions Based on the physical intervals of the QTLs and gene annotations from the ZS97 reference genome, genes located within each QTL interval were first identified. Candidate gene screening was then refined by integrating protein function annotations, known trait-associated QTLs, and the predicted impact of sequence variants on protein function (categorized as HIGH, MODERATE, LOW, or MODIFIER). By prioritizing variants with HIGH and MODERATE impacts, the number of candidate genes was substantially reduced. The results of candidate gene screening for QTLs in the RIL population are presented in Table 2. The candidate gene identified in QTL qGNP3-2 and qPN3-1 was LOC_Os03g49880 (OsTB1), which has a codon deletion (AGCG-> A) and a splice donor mutation (G-> T). Studies had found that OsTB1 influences rice tillering mainly by regulating the growth and development of axillary buds (Takeda et al. 2003). When OsTB1-overexpressing transgenic lines were constructed, axillary bud growth was significantly inhibited, leading to a significant reduction in tiller number. In contrast, OsTB1 loss-of-function mutants showed a noticeable increase in tillering. The candidate gene identified in QTL qPH10 and qHD10 was LOC_Os10g32600 (Ehd1), which has three nonsynonymous coding mutations with moderate effects on its protein structure. Studies had shown that Ehd1 encodes a B-type photoperiod regulatory protein composed of 341 amino acids (Doi et al. 2004). The N-terminal of this protein contains a signal receiving domain, and the central region has a GARP (G-box binding transcription factor Pintail-like) domain that can bind to DNA. Sequence analysis of the Ehd1 in different rice varieties revealed a key amino acid substitution in the GARP domain of T65 rice: glycine (Gly) was replaced by arginine (Arg), which is a major cause of delayed flowering in T65 rice. This mutation directly affects the function of Ehd1 and regulates the genetic traits related to flowering time. Relationship Between Gene Expression and Agronomic Traits Principal component analysis (PCA) was first conducted on the gene expression profiles of 122 IMF2 accessions. The results revealed that accessions sharing the same maternal parent exhibited a clear clustering trend in expression space, suggesting that maternal genetic background may significantly influence offspring gene expression (Fig. S5A). In contrast, accessions with the same paternal parent did not exhibit a similar clustering pattern (Fig. S5B). Subsequently, a generalized linear model was utilized to establish an association between gene expression levels (FPKM values) and the heading date trait (Fig. 6A). This yielded 854 genes exhibiting a significant association with heading date. To delineate the biological functions underlying these trait-associated genes, Gene Ontology (GO) enrichment analysis was subsequently conducted (Fig. 6B). The findings indicated a prominent enrichment of heading date-associated genes in pathways encompassing energy metabolism, electron transport, redox reactions, photosynthesis and light energy capture, ion transport and homeostasis, redox regulation, and overarching metabolic processes. Fig. 6Association analysis of gene expression levels with heading date in 122 IMF2 accessions. A Manhattan plot showing the associations between heading date and gene expression levels. B GO enrichment analysis of genes significantly associated with heading date Association analysis of gene expression levels with heading date in 122 IMF2 accessions. A Manhattan plot showing the associations between heading date and gene expression levels. B GO enrichment analysis of genes significantly associated with heading date Gene Expression Pattern Analysis Based on the bin marker analysis of the IMF2 population, genes were divided into two categories: homozygous, which include gene fragments entirely derived from the parent XQZB (type a) or entirely derived from the parent ZH9308 (type b); and heterozygous (h), where one chromosomal fragment is from XQZB and the other is from ZH9308. Then, this study calculated the mid-parent dominance value of the expression levels for each gene. It was found that most of the genes had a mid-parent dominance value around 0, indicating that the expression of most genes follows an additive model. Only about 2% of the genes had a mid-parent dominance value greater than 20% (Fig. 7A). Meanwhile, the mid-parent dominance values of the genes significantly associated with agronomic traits were generally also below 20% (Fig. 7B). Fig. 7The distribution of mid-parent heterosis in gene expression levels. A The distribution of parental advantage values in gene expression levels. B The distribution of parental advantage values in the expression of significant genes associated with agronomic traits The distribution of mid-parent heterosis in gene expression levels. A The distribution of parental advantage values in gene expression levels. B The distribution of parental advantage values in the expression of significant genes associated with agronomic traits Parental Allele-Specific Expression Analysis Using RNA-seq data from 122 IMF2 lines, a comprehensive variant site analysis was performed, resulting in the identification of 223,793 candidate variants (Fig. S6). To ensure the reliability of these variant calls, they were cross validated against resequencing data from the RIL population. Following stringent filtering, a final set of 111,931 high-confidence and consistent variant sites was retained. Next, allele-specific expression (ASE) weights were calculated based on the sequencing depth of the original parental alleles at each variant site. These weights were then integrated with gene expression data to estimate the transcript levels of the original parental alleles across the 122 IMF2 lines. Regression analysis was performed on allele-specific expression levels to determine their relationship with the heading date trait. This analysis revealed that multiple alleles exhibiting preferential expression in relation to heading date were predominantly located on chromosome 10 (Fig. 8A). Further examination of the regression coefficients (beta values) indicated a significant negative effect on heading date from XQZB alleles, while the majority of ZH9308 alleles showed a significant positive effect (Fig. 8B). Fig. 8Association analysis between allele-specific expression levels and heading date trait. A Manhattan plot showing the associations between the expression levels of XQZB and ZH9308 alleles and heading date. B Regression beta values for XQZB and ZH9308 alleles that are significantly associated with heading date Association analysis between allele-specific expression levels and heading date trait. A Manhattan plot showing the associations between the expression levels of XQZB and ZH9308 alleles and heading date. B Regression beta values for XQZB and ZH9308 alleles that are significantly associated with heading date Discussion This study aimed to explore the genetic basis underlying the super hybrid rice variety Xieyou9308. To this end, a RIL population derived from Xieyou9308 was sequenced, and an IMF2 population was generated through a pairwise crossing scheme among the RILs. Because the genotype of each IMF2 line can be inferred from its RIL parents, this design improves mapping resolution and detection power, enabling high-quality QTL identification using binmaps (Yao et al. 2019). Genomic constitution analysis revealed that the overall effect of heterozygous genotypes was weakly correlated with phenotypic traits, consistent with previous findings (Huang et al. 2015). This suggests the potential to develop higher-yielding inbred lines without relying on hybridization between inbred. In other words, the new hybrids would benefit more from homozygous favorable alleles rather than heterozygous alleles, leading to better yield performance. These findings indicate that selecting and pyramiding homozygous favorable alleles may be a more effective strategy for developing high-yielding inbred rice lines than relying on heterozygosity derived from hybridization. To understand the genetic variation of complex traits, we employed a full-QTL model that integrates additive effects, dominance effects, epistasis, and gene-environment interactions into a single mapping framework. Such a comprehensive model has not been considered in most previous QTL studies of agronomic traits. In this framework, QTL main effects are treated as fixed, assuming they are consistent across environments, while QTL-environment interaction effects are treated as random. The random interaction effects capture deviations from the average QTL effect across environments, based on the assumption that their sum equals zero. As a result, both positive and negative QTL-environment interaction effects may be observed. For ten traits, a total of 8 epistatic QTL pairs were detected in IMF2 population, especially, three epistasis QTL pairs with additive-additive, dominance-dominance interaction effects, contributed more than 10% to the phenotypic variance. These findings highlight the significant role of epistatic interactions in contributing to heterosis, a phenomenon that has been well-documented in numerous studies (Ma et al. 2007; Wei et al. 2024; Yu et al. 1997). Furthermore, we identified several environment-specific QTL effects across both environments, which could be leveraged in the genetic improvement of rice to develop elite varieties with either environment-specific traits or strong environmental adaptability in future breeding programs. Based on the estimated allele additive effects, the ZH9308 alleles showed positive additive effects on most QTLs associated with traits such as plant height, heading date, grain number per panicle, and panicle length. In contrast, the XQZB alleles enhanced traits like grain length, 1000-grain weight, and seed setting rate. Our results suggest that a few loci exhibited positive partial dominance, while even fewer displayed overdominance effects that favored heterozygosity. These findings suggest that identifying and utilizing superior alleles could significantly contribute to further yield improvement. Therefore, hybrids between the indica and japonica subspecies can effectively combine these beneficial alleles, optimizing their advantages for enhanced yield potential. Based on the results of QTL association mapping, candidate genes controlling traits can be narrowed down to the regions of the identified QTLs. However, when using a preliminary mapping population, such as the RILs in our study, distinguishing candidate genes for a specific trait from the large number of annotated genes in the QTL regions remains a challenge. Therefore, further studies, such as fine mapping using secondary mapping populations, are necessary to more precisely define the target QTL. Secondary mapping populations, including single-segment substitution lines (SSSLs) and near-isogenic lines (NILs), are genetically identical to the recurrent parent except for the target QTL, which helps eliminate genetic background interference and improves the sensitivity and accuracy of candidate gene prediction. Using RNA-seq data to evaluate gene expression is an effective approach for functionally validating candidate genes. However, due to the spatiotemporal specificity of gene expression, the leaf RNA-seq data from IMF2 population, while helpful, may not be sufficient for comprehensive screening of candidate genes for grain characteristics and panicle traits. Research on the genetic mechanisms of heterosis has mostly focused on the overall gene expression levels in hybrids and their parents. However, the differential regulation of parental alleles in hybrids has not been thoroughly studied at the population level. We performed a population-level analysis of parent allele-specific allele expression based on RNA-seq data from the IMF2 population, showed that allele expression from XQZB shortened the heading date, while that from ZH9308 increased it. which was consistent with the QTL association mapping results. Materials and Methods Plant Materials and Phenotypic Investigation The recombinant inbred line (RIL) mapping population used in this study consisted of 139 lines, derived from the super hybrid rice cross between the maintainer line XieqingzaoB (XQZB) and the restorer line Zhonghui9308 (ZH9308) through the single seed descent method. The IMF2 mapping population comprising 570 hybrid combinations was generated from the RILs using a pairwise crossing scheme. Phenotypic evaluations were conducted at two experimental farms of the China National Rice Research Institute (CNRRI): one located in Fuyang City, Zhejiang Province (120.2° E, 30.3° N) in 2019, and the other in Lingshui City, Hainan Province (110.0° E, 18.5° N) in 2020. For each accession, 18 seedlings were germinated and transplanted in three rows, with row and column spacing set to 20 cm and 26.7 cm, respectively. Plants located in the middle of each plot were selected for phenotypic assessment including morphological characteristics (plant height), yield components and relevant factors (grain length, grain width, panicle number, panicle length, grain number per panicle, 1000-grain weight, seed setting rate, and yield per plant), and physiological traits (e.g., heading date). Among these, heading date and morphological characteristics were measured in the field. The heading date was recorded as the number of days from sowing to the emergence of the first inflorescences above the flag leaf sheath of five plants per accession. The other field traits were investigated after heading, with the tallest panicle selected to measure plant height. Grain-related traits were observed in the laboratory after harvest. The full/total grain number per plant was counted manually, and grain number per panicle was estimated by dividing the total grain number by the panicle number. Seed setting rate was calculated as the ratio of full grain number to total grain number per plant. The 1,000-grain weight was determined by weighing 400 fully filled grains and then converting this weight to the 1,000-grain weight. Yield per plant was recorded as the total weight of fully filled grains per plant. Sequence Variation Calling DNA was extracted from fresh leaves of 139 RIL lines and the original parents using a routine protocol. Resequencing paired-end reads were generated using Illumina HiSeq 4000. Raw data were first filtered by NGSQC Toolkit (v2.3.348). Paired-end reads were aligned to the rice ZS97 reference genome using the MEM algorithm of BWA (Li and Durbin 2009). SAMtools (version 1.7) (Danecek et al. 2021) was used to sort BAM files resulting from BWA output. The sorted alignment files were then input into the Genome Analysis Toolkit (GATK, version 4.1) (McKenna et al. 2010) to call the SNPs. Variations were called using ‘HaplotypeCaller’, ‘GenomicsDBImport’ and ‘GenotypeGVCFs’ functions in GATK. High-quality variations were contained by the ‘VariantFiltration’ function in GATK with parameters of ‘–cluster-size 3 –cluster-window-size 10 QD < 10.00 FS > 15.000 AC < 3 DP > 200||DP < 5’ for SNPs and ‘QD < 10.00 FS > 30.000 DP > 200||DP < 5’ for InDels. SNPs assigned as ‘PASS’ by GATK and SNPs and Indels with a missing rate less than 0.2 and a minor allele frequency greater than 0.2 were retained for further analysis. All variants were annotated by SnpEff (Cingolani et al. 2012). Using the Python program SNPbinner (Gonda et al. 2019), bin maps were constructed for the RIL population, based on the SNPs and Indels genotyped by the GATK pipeline to further decrease the impact of sequencing errors and computational consumption. Statistical Analysis of Phenotypes Based on the following mixed linear model, variance component estimation and correlation analysis of phenotypic traits are performed:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{Y}_{hkm}=\mu\:+{G}_{k}+{L}_{h}+{GL}_{hk}+{\epsilon\:}_{hkm},$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{Y}_{hkm}$$\end{document} represents the phenotypic value of the k-th accession in the h-th environment and the m-th replicate; µ is the population mean; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{G}_{k}$$\end{document} represents the genetic effect of the k-th accession, random effect, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${G}_{k}\sim N(0,{\sigma}_{G}^{2})$$\end{document}; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{L}_{h}$$\end{document} represents the effect of the h-th environment, random effect, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${L}_{h}\sim N(0,{\sigma}_{L}^{2})$$\end{document}; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{GL}_{hk}$$\end{document} represents the interaction effect between the k-th genotype and the h-th environment, random effect, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${GL}_{hk} \sim N (0,{\sigma}_{GL}^{2})$$\end{document}; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\epsilon\:}_{hkm}$$\end{document} is the residual effect of the individual, random effect, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon}_{hkm}\:\sim N(0,{\sigma}_{\varepsilon}^{2})$$\end{document}. Based on the above mixed linear model, the R package lmer (Bates 2016) is used to estimate the variances of the random effects \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\widehat{\sigma}}_{G}^{2},\:{\widehat{\sigma}}_{L}^{2},\:{\widehat{\sigma}}_{GL}^{2},\:{\widehat{\sigma}}_{\varepsilon}^{2}$$\end{document}. The best linear unbiased prediction (BLUP) method is used to predict the random effects \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\widehat{G}}_{k},\,{\widehat{L}}_{h},\,{\widehat{GL}}_{hk}$$\end{document}. The broad-sense heritability was estimated with the formula \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{g}^{2}={\widehat{\sigma\:}}_{G}^{2}/({\widehat{\sigma\:}}_{G}^{2}+{\widehat{\sigma\:}}_{GL}^{2}+{\widehat{\sigma\:}}_{\epsilon\:}^{2})$$\end{document} and the interaction genetic rate was estimated with the formula \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{h}_{ge}^{2}={\widehat{\sigma\:}}_{GL}^{2}/({\widehat{\sigma\:}}_{G}^{2}+{\widehat{\sigma\:}}_{GL}^{2}+{\widehat{\sigma\:}}_{\epsilon\:}^{2})$$\end{document}. The rcorr module of Hmisc R package (https://cran.r-project.org/web/packages/Hmisc/index.html) was employed to calculate Pearson correlation coefficients between 10 agronomic traits: (1) phenotypic correlation coefficients with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{Y}_{hkm}$$\end{document} for each environment, respectively; (2) genetic correlation coefficients with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\widehat{Y}}_{k}=\widehat{\mu\:}+{\widehat{G}}_{k}$$\end{document}, where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\widehat{Y}}_{k}$$\end{document} is the corrected genotypic value of the k-th individual by population mean, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\widehat{\mu\:}$$\end{document} is the estimated population mean, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\widehat{G}}_{k}$$\end{document} is the genotypic value of the k-th individual predicted by BLUP. Genetic Model and Statistical Methods for QTL Association Mapping For the RIL population, the full model for QTL association mapping based on the mixed linear model includes the additive effects of each QTL, the additive-by-additive epistasis effects of paired QTLs, as well as the interaction effects of each effect with the environment. Suppose a trait is controlled by s QTLs, with t pairs of QTLs exhibiting epistasis effects. Then, the phenotypic value\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\:{y}_{hkm}$$\end{document} of the k-th genotype in the h-th environment and the m-th replicate can be represented by the following mixed linear model:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} y_{{hkm}} = \mu + \sum\limits_{{i = 1}}^{s} {a_{i} } x_{{ik}} \hfill \\ + \mathop \sum \limits_{{\begin{array}{*{20}l} {i,j \in \left\{ {1,2 \ldots ,s} \right\}} \hfill \\ {l \ne n} \hfill \\ \end{array} }}^{t} aa_{{ij}} x_{{ik}} x_{{jk}} + e_{h} \hfill \\ + \sum\limits_{{i = 1}}^{s} a e_{{hi}} u_{{hik}} \hfill \\ + \mathop \sum \limits_{{\begin{array}{*{20}l} {i,j \in \left\{ {1,2 \ldots ,s} \right\}} \hfill \\ {l \ne n} \hfill \\ \end{array} }}^{t} aae_{{hij}} u_{{hijk}} + \varepsilon _{{hkm}} \# \hfill \\ \end{gathered} $$\end{document} where, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\mu\:$$\end{document} is the population mean; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{a}_{i}$$\end{document} is the additive effect of the i-th QTL with coefficient \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{x}_{ik}$$\end{document}, fixed effect; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{aa}_{ij}$$\end{document} is the additive-by-additive epistatic effect of the i-th QTL and the j-th QTL with coefficient \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{x}_{ik}{.x}_{jk}$$\end{document}, fixed effect; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{e}_{h}$$\end{document} is the effect of the h-th environment, random effect, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{e}_{h}\sim\left(0,{\sigma\:}_{E}^{2}\right)$$\end{document}; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{ae}_{ih}$$\end{document} is the additive-by-environment interaction effect of the i-th QTL and the h-th environment with coefficient \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{x}_{hik}$$\end{document}, random effect, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{ae}_{ih}\sim\left(0,\:\:{\sigma\:}_{{A}_{i}E}^{2}\right)$$\end{document}; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{aae}_{hij}$$\end{document} the interaction effect of the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{aa}_{ij}$$\end{document}and the h-th environment with coefficient \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{u}_{hijk}$$\end{document}=\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{x}_{hik}.{x}_{hjk}$$\end{document}, random effect, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{aae}_{ijh}\sim(0,{\sigma\:}_{{AA}_{ij}E}^{2})$$\end{document}; and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\epsilon\:}_{hkm}$$\end{document} is the residual effect of the individual, random, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\epsilon\:}_{hkm}\sim(0,\:{\sigma\:}_{\epsilon\:}^{2})$$\end{document}. For the IMF2 population, the full model for QTL association mapping based on the mixed linear model should include the additive, dominance, and epistasis effects (additive-by-additive, additive-by-dominant, dominant-by-additive, and dominant-by-dominant) of each QTL, as well as the interaction effects of each of these effects with the environment. Considering a trait that is controlled by s QTLs, with t pairs of QTLs exhibiting epistasis effects, the phenotypic value \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{y}_{hkm}$$\end{document} of the k-th genotype in the h-th environment and the m-th replicate can be represented by the following mixed linear model:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \begin{gathered} y_{{hkm}} = \mu + \mathop \sum \limits_{{i = 1}}^{s} a_{i} x_{{A_{{ik}} }} + \mathop \sum \limits_{{i = 1}}^{s} d_{i} x_{{D_{{ik}} }} \hfill \\ + \mathop \sum \limits_{{\begin{array}{*{20}l} {i,j \in \left\{ {1,2 \ldots ,s} \right\}} \\ {l \ne n} \\ \end{array} }}^{t} aa_{{ij}} x_{{AA_{{ijk}} }} \hfill \\ + \mathop \sum \limits_{{\begin{array}{*{20}c} {i,j \in \left\{ {1,2 \ldots ,s} \right\}} \\ {l \ne n} \\ \end{array} }}^{t} ad_{{ij}} x_{{AD_{{ijk}} }} \hfill \\ + \mathop \sum \limits_{{\begin{array}{*{20}l} {i,j \in \left\{ {1,2 \ldots ,s} \right\}} \\ {l \ne n} \\ \end{array} }}^{t} da_{{ij}} x_{{DA_{{ijk}} }} \hfill \\ + \mathop \sum \limits_{{\begin{array}{*{20}l} {i,j \in \left\{ {1,2 \ldots ,s} \right\}} \\ {l \ne n} \\ \end{array} }}^{t} dd_{{ij}} x_{{DD_{{ijk}} }} \hfill \\ + e_{h} + \mathop \sum \limits_{{i = 1}}^{s} ae_{{hi}} u_{{A_{{ik}} E_{h} }} \hfill \\ + \mathop \sum \limits_{{i = 1}}^{s} de_{{hi}} u_{{D_{{ik}} E_{h} }} + \mathop \sum \limits_{{i = 1}}^{s} aae_{{hij}} u_{{AA_{{ijk}} E_{h} }} \hfill \\ + \mathop \sum \limits_{{i = 1}}^{s} ade_{{hij}} u_{{AD_{{ijk}} E_{h} }} + \mathop \sum \limits_{{i = 1}}^{s} dae_{{hij}} u_{{DA_{{ijk}} E_{h} }} \hfill \\ + \mathop \sum \limits_{{i = 1}}^{s} dde_{{hij}} u_{{dd_{{ijk}} E_{h} }} + \varepsilon _{{hkm}} \hfill \\ \end{gathered} $$\end{document} where, µ is the population mean; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{a}_{i}\:$$\end{document}and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{d}_{i}$$\end{document} represent the additive and dominance effects of the i-th QTL, with coefficients \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{x}_{{A}_{ik}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{x}_{{D}_{ik}}$$\end{document}, fixed effects; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{aa}_{ij}$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{ad}_{ij}$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{da}_{ij}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{dd}_{ij}$$\end{document} represent the additive-by-additive, additive-by-dominant, dominant-by-additive, and dominant-by-dominant epistasis effects of the i-th and j-th QTLs, with coefficients \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{x}_{{AA}_{ijk}}={x}_{{A}_{ik}}.{x}_{{A}_{jk}}$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{x}_{{AD}_{ijk}}={x}_{{A}_{ik}}.{x}_{{D}_{jk}}$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{x}_{{DA}_{ijk}}={x}_{{D}_{ik}}.{x}_{{A}_{jk}}$$\end{document}and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{x}_{{DD}_{ijk}}={x}_{{D}_{ik}}.{x}_{{D}_{jk}}$$\end{document}, fixed effects; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{e}_{h}$$\end{document} is the environmental effect of the h-th environment, random effect, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{e}_{h}\sim\left(0,{\sigma\:}_{E}^{2}\right)$$\end{document}; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{ae}_{hi}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{de}_{hi}$$\end{document}​ represent the interaction effects between the additive and dominance of the i-th QTL and the h-th environment, with coefficients \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{u}_{{A}_{ik}{E}_{h}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{u}_{{D}_{ik}{E}_{h}}$$\end{document}, random effects, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{ae}_{ih}\sim\left(0,\:\:{\sigma\:}_{{A}_{i}E}^{2}\right)$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{de}_{ih}\sim\left(0,\:\:{\sigma\:}_{{D}_{i}E}^{2}\right)$$\end{document}; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{aae}_{hij}$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{ade}_{hij}$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{dae}_{hij}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{dde}_{hij}$$\end{document} represent the interaction effects between the additive-by-additive, additive-by-dominant, dominant-by-additive, and dominant-by-dominant epistasis of the i-th and j-th QTLs with the h-th environment, random effects, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{aae}_{ijh}\sim(0,{\sigma\:}_{{AA}_{ij}E}^{2})$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{ade}_{ijh}\sim(0,{\sigma\:}_{{AD}_{ij}E}^{2})$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{dae}_{ijh}\sim(0,{\sigma\:}_{{DA}_{ij}E}^{2})$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{dde}_{ijh}\sim(0,{\sigma\:}_{{DD}_{ij}E}^{2})$$\end{document}; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\epsilon\:}_{hkm}$$\end{document} represents the residual effect of the individual, random effect, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\epsilon\:}_{hkm}\sim(0,\:{\sigma\:}_{\epsilon\:}^{2})$$\end{document}. The coefficients for the QTLs depend on the genotype of the QTL. If the genotype of the QTL is QQ, Qq, or qq, the coefficients for the additive effects are 1, 0, and − 1, respectively, and the coefficients for the dominance effects are 0, 1, and 0, respectively. For the above model, QTL association analysis for the RIL and IMF2 populations under two environments and ten agronomic traits was performed using the mixed linear model-based software QTXNetwork (Zhang et al. 2015). Firstly, QTXNetwork uses one-dimensional and two-dimensional genome scanning strategies to detect all bin markers with significant effects and pairs of epistatic bin markers. Then, a full model is constructed to estimate all parameters in the above model. The minimum quadratic unbiased estimation (MINQUE) method (Rao 1971) is used to estimate the variance components of the random effects in the model. Based on the estimated variance components, the adjusted unbiased prediction (AUP) method is used to predict the random effect parameters. Meanwhile, the least squares estimation method is used to estimate the fixed effect parameters in the model. The obtained estimates are then used as prior values in the MCMC process. Using Gibbs sampling, the parameters are estimated, and significance tests are conducted to make the corresponding statistical inferences (Smith and Roberts 1993). RNA-seq Data Analysis A total of 122 hybrid combinations were selected for RNA sequencing from the IMF2 mapping population. Total RNA was extracted from flag leaf tissues collected at the booting stage from plants grown in Lingshui using Trizol reagent (Invitrogen, CA, USA) and purified using a RNeasy Plant Mini Kit (Qiagen, CA, USA) according to the manufacturer’s instructions. RNA degradation and contamination were checked by 1% agarose gel electrophoresis. The quality and integrity of RNA were assessed using an Agilent Bioanalyzer 2100 system (Agilent, CA, USA); RNA Integrity Number (RIN) values were greater than 8.5 for all samples. After total RNA extraction, mRNA was enriched by Oligo (dT) beads. Sequencing libraries were prepared using an Illumina TruSeq RNA Library Preparation Kit (Illumina, CA, USA) as per the manufacturer’s protocol and sequenced on an Illumina HiSeq 2500 platform, and 150 bp paired-end reads were generated. After removing low-quality reads with > 5 bases aligned to the adapter sequence or > 50% bases having Phred quality < 20 or contained ≥ 10% unidentified nucleotides by fastp (Chen et al. 2018a, b), high-quality reads were mapped to the ZS97 genome using HISAT2 (Kim et al. 2015). HTseq (Anders et al. 2015) was used to obtain the mapping reads count for each gene, and StringTie (Pertea et al. 2016) was used to obtain the FPKM value for each gene. Given that the transcriptome data were derived from a single environment, we applied a generalized linear model (GLM) using R software to assess the relationship between gene expression and agronomic traits. In this model, the phenotypic value of each trait was treated as the dependent variable (y), while the normalized gene expression levels were used as independent variables (x). The model was implemented using the glm() function in R, and significance was evaluated based on the model coefficients and p-values. Allele Specific Expression Analysis Variant detection was performed on the RNA-seq data of 122 IMF2 lines. The specific process included: constructing a reference genome index, followed by data quality control and cleaning to remove low-quality sequences and adapter contamination, with filtering parameters consistent with those used in resequencing data analysis. STAR (Dobin et al. 2013) was then used for alignment in a two-pass mode. The first alignment was performed based on the initial index, and the new splice site information discovered in the first alignment was added to the index for the second alignment. Next, variant detection, filtering, and annotation were performed to identify high-quality variants consistent with resequencing data from the RIL population. Finally, the allele-specific sequencing depth for each variant site was extracted from the VCF file, and the preference expression weight of the two alleles was calculated. This was combined with gene expression levels to obtain the transcript expression levels of the original parents in the 122 IMF2 strains, followed by regression analysis with agronomic traits. GO Enrichment Analysis InterProScan software was used to obtain Gene Ontology (GO) annotation information for protein sequences (Jones et al. 2014). R package ClusterProfiler was used to performing GO enrichment analysis (Yu et al. 2012). Supplementary Information Supplementary Material 1 Supplementary Material 1