Genes, Putative Long-Lived mRNAs and Pathways Underlying Genotypic Differences in Rice Seed Storability and Seed Dormancy

Abstract

Weak seed dormancy (SD) in rice tends to induce pre-harvest sprouting and impair seed quality and yield pre-harvest, whereas poor seed storability (SS) reduces these traits during post-harvest storage. Although multiple genes associated with these two traits have been cloned, the molecular genetic regulatory relationship between them remains unclear. To dissect the SD-SS correlation, this study compared SD and SS characteristic of 9311 (Xian/Indica) and Nipponbare (NIP, Geng/Japonica) via transcriptomic and metabolomic analyses. Results showed that NIP had strong SD but poor SS, while 9311 exhibited the opposite. Differentially accumulated metabolite (DAM) analysis showed 42 DAMs specific to dormant seeds, 141 to stored seeds, and 93 common to both. Transcriptomic analysis identified 1,334 (13.0%) differentially expressed genes (DEGs) and 11 metabolic pathways (28.9%) commonly associated with SD and SS, including key ones like hormone signaling and secondary metabolism. The biological functions of two core DEGs were further validated using CRISPR/Cas9 technology, among which OsGA2ox8 regulates SD and OsLEA5 (Late embryogenesis abundant protein) affects SS. Validation of DEGs in the gibberellin (GA) pathway demonstrated that knockout of OsGA2ox8 (gibberellin 2-oxidase) significantly reduced SD, whereas its overexpression markedly enhanced SD, confirming the core regulatory role of OsGA2ox8 in SD. Haplotype analysis in natural populations showed that Haplotype 1 of OsGA2ox8 was dominant in Xian subspecies, while Haplotype 2 prevailed in Geng subspecies. Additionally, the analysis of long-lived mRNAs (LLRs) identified 2,938 putative LLRs, of which 309 were associated with both SD and SS. Functional validation of a late embryogenesis abundant protein (OsLEA5) showed that knockout of this gene in NIP significantly decreased SS. This study preliminarily elucidated the differentiation mechanisms of SD and SS, and provided potential targets for breeding rice varieties with enhanced pre-harvest sprouting and superior SS.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12284-026-00931-7.

Background

Seeds are pivotal throughout the life cycle of plants, starting at sowing in the field and ending at harvest (Yuan et al. 2019; Liu et al. 2023). Seed vigor encompasses the capacity of seeds to transition from a silent state to a germinated state, as well as their ability to establish seedlings under either favourable or challenging water and soil environments (Mahender et al. 2015). Seed vigor ability can be reflected through various indicators, including the capacity to delay and prevent seed germination due to dormancy (Finkelstein et al. 2008; Mizuno et al. 2018), the tolerance to abiotic stresses during the germination stage (Shim et al. 2020), and seed germination ability after a period of storage (Lin et al. 2015; Zheng et al. 2025). Seed dormancy (SD), which refers to the inability of freshly-harvested seeds to germinate under favorite conditions (Chang et al. 2018), is a critical determinant of seed quality and resistance to pre-harvest sprouting (PHS) (Nagel et al. 2019). However, extremely strong SD leads to low or failed germination, higher seed usage of sowing, and poor crop establishment. Conversely, weak SD can result in PHS in high temperature and humid environments during seed maturation (Zhang et al. 2020; Chen et al. 2025). After SD is broken, seed germination is considered as the trait marking a rapid transition to the developmental phase in the life cycle (Bewley et al. 1997; Li et al. 2011; Gao et al. 2024). Rapid and uniform germination and seedling emergence are highly desirable for optimal seedling growth and early crop establishment. Seed storability (SS), is the capacity or duration for seeds to maintain a certain level of vitality during storage (Rajjou et al. 2008). In agricultural practices, seeds are often stored for a long period of time before they are sown or processed (Zhou et al. 2024). Over time, seeds can undergo deterioration during storage, particularly those with poor SS. This can result in poor and failed germination, and the production of non-viable seedlings. Furthermore, the nutritional quality of seeds may be compromised, which can affect their edible and economic value (Yoshida et al. 2010; Tang et al. 2016; Pang et al. 2018).

The environmental factors that induce seed germination and accelerate seed deterioration are strikingly similar, implying that SD and SS may be affected by some common environmental and genetic factors (Yu et al. 2024). For example, ABI3 is a key gene that initiates seed germination, and the increased permeability of cell membranes in abi3 mutant seeds can mitigate SD and reduce SS (Clerkx et al. 2003). Studies have shown that mutations of genes dog1 and ats can significantly diminish SD and SS, indicating a positive correlation between these two traits (Debeaujon et al. 2000; Bentsink et al. 2006; Sugliani et al. 2009). However, a negative correlation between SS and SD was reported in Arabidopsis through natural variation analysis (Nguyen et al. 2012). In rice, three genes (Rc, OsVP1, and OsC1) involved in the biosynthesis of anthocyanins were reported to be able to enhance the perception of the abscisic acid (ABA) signaling, thus increase seed sensitivity to ABA, and ultimately inhibit PHS (Wang et al. 2020). Seed storage experiments using near-isogenic lines, SD7-1D (Rc) and SD7-1d (rc), confirmed that rice Rc can enhance drought tolerance through elevated pressure of oxygen aging (Manjunath et al. 2023). Thus, the relationship between SD and SS in rice appears to be complex and remains to be better understood at the molecular level.

The accumulation of carbohydrates, lipids, proteins, and mRNA in seeds is essential for maintaining seed germination potential (Jang et al. 2009; Zhang et al. 2010; Sano et al. 2015). Normally, the lifespan of mRNA is only a few minutes, but some mRNA accumulated in seeds can be preserved for several months or even years. These mRNAs that can be stored long-term without degradation are called long-lived mRNAs (LLRs) There are many types of LLRs, mainly including some mRNA involved in protein biosynthesis, energy metabolism, and stress responses, such as small heat shock proteins and LEA (late embryogenesis abundant) family proteins (Liang et al. 2023). Among these, LEAs are originally identified as being highly accumulated in seeds during the late embryogenesis stage, which serve as membrane protectants, protein stabilizers, and antioxidants in dry seeds (Sano et al. 2013). It was reported that LLRs related to SD is gradually degraded during the seed maturation process (Galau et al. 1986). SS during long-term storage is reportedly closely linked to the random degradation of some mRNA fragments (Kimura et al. 2010; Sano et al. 2012). The residual mRNA within seeds is translated into proteins that facilitate rapid germination during the initial phase of seed imbibition (Sano et al. 2019). To date, transcriptomic technology has identified numerous LLRs that individually influence SD and SS in plants. However, no report in rice has yet employed transcriptomic analysis to concurrently examine the relationship between SD and SS.

In the present study, we compared the germination rates of freshly-harvested seeds and seeds stored for different lengths of periods of two rice varieties which showed contrast differences in SD and SS. By analyzing their phenotypic, transcriptomic, and metabolomic changes in SD and SS, we were able to provide insights into the genotypic differences in genetic and molecular mechanisms underlying SD and SS in rice, and their implications for developing rice cultivars with enhanced resistance to PHS, improved capacity for direct-seeding of rice and improved SS of rice.

Materials and Methods

Plant Materials

Two rice varieties, Nipponbare (NIP), a Geng (japonica) variety, and 9311, an elite Xian (indica) hybrid restorer line, were utilized to assess SD and SS. The two varieties were cultivated at the Nanfan experimental farm in Sanya, China, during the winter season (December to May). Seeds of the knockout (KO) and overexpression (OE) of OsGA2ox8 in the NIP (wild-type, WT) genetic background (Xiong et al. 2022) were kindly provided by Wensheng Wang (State Key Laboratory of Crop Gene Resources and Breeding/Institute of Crop Sciences, Chinese Academy of Agricultural Sciences, Beijing 100081, China). Seeds of the OsLEA5 KO lines generated via CRISPR/Cas9 gene-editing technology, with NIP serving as the WT genetic background. Each line was sown in a four-row plot, with eight plants per row, and a spacing of 20.0 × 25.0 cm between rows and plants.

Experimental Design

To ensure uniform flowering and maturity of harvested seeds, NIP and 9311 were sown in two stages based on their heading dates. The flowering time of each plant was recorded. At the maturity stage, at least five plants from the middle of each row were harvested approximately 35–40 days after flowering when virtually all seeds were fully filled and reached complete maturity. The collected seeds were dried under the natural sunlight for three days and then allowed to equilibrate for 5–7 days in the room temperature to achieve a consistent moisture content of approximately 12–13% (referred to as freshly harvested seeds, FH). Then, the equilibrated seeds of NIP and 9311 were stored under the conditions of constant temperature of 24℃–26℃ at a relative humidity of 30–40% (Yuan et al. 2019). Subsequently, five types of seeds were obtained for both NIP and 9311, including FH seeds to assess SD, after-ripened seeds (AR, freshly harvested seeds after dormancy was broken by placed under 48 °C for 72 h in the oven) as the control, the AR seeds that had been stored for 9 months, 12 months, and 18 months to evaluate SS. After each sampling, the seeds were placed in a − 20 °C refrigerator to maintain the state of seeds until the seeds were used for subsequent germination experiments and sequencing analyses.

Phenotyping SD and SS

In the germination test to evaluate SD, 30 seeds from each sample were placed in a Petri dish with two sheets of filter paper, and then 15 ml of distilled water was added to the dish, which was then placed in the growth chamber at 25 °C under a 14-hour light and 10-hour dark cycle with three independent replicates for each type of seed treatments (Yuan et al. 2021). The number of germinated seeds and established seedlings in each Petri dish were monitored daily for seven consecutive days. A seed was considered germinated when its radicle exceeded 2 millimeters as previously described (Yuan et al. 2021). The seed germination and early seedling-related parameters were used to assess SD and SS of NIP and 9311. These parameters included: the percentage of germinated seeds and seedlings at 2 to 7 days after imbibition; the germination index (GI) and seedling establishment index (SI) were calculated as GI = ∑(Gt/t) and SI = ∑(St/t), respectively, where Gt and St represented the numbers of germinated seeds and seedlings on day t.

RNA Extraction and Transcriptome Analysis

Total RNA was extracted from embryos from the seeds, which were separated from dehulled seeds and mixed directly, with each collection considered as one replicate for a total of three biological replicates. The tissues were promptly transferred into liquid nitrogen for storage at − 80 °C to prevent RNA degradation. Total RNA was extracted using the HP Plant RNA kit (Omega, Atlanta, GA) according to the manufacturer’s instructions. RNA-seq was conducted on the Illumina RNA sequencing (Paired-end, PE) (Li et al. 2024; Wang et al. 2020). The expression level of each gene was calculated based on the comparison results. Differentially expressed genes (DEGs) were identified using two criteria: |log₂FoldChange| ≥ 1 and P value < 0.05. Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were performed with a threshold of P < 0.05.

Endogenous Hormone Quantification in Rice Embryos

Endogenous phytohormones, including ABA, GA1, GA3, GA4, GA7, IAA, Jasmonic Acid (JA), and Salicylic Acid (SA) were quantified by LC-MS/MS (n = 3 biological replicates). The endogenous phytohormone detection service was provided by Wuhan ProNets Testing Technology Co., Ltd. Embryos dissected from dehulled seeds and pooled. 0.2 g sample + 2 mL acetonitrile + 30 µL IS stock, extracted overnight at 4 °C. 7000 rpm × 5 min, collect supernatant. Pellet re-extracted with 2 mL acetonitrile (sonication 30 min); combine supernatants. Add 200 mg C18, vortex 30 s, same centrifugation, collect supernatant. Evaporate to dryness, reconstitute in 150 µL methanol, filter 0.22 μm organic membrane. 500 µg/mL individual standards in methanol, diluted to 5 µg/mL mixed hormone and 0.1 µg/mL mixed IS working solutions. Serially diluted to 0.1–200 ng/mL (8 points, 20 ng/mL IS each). Column: Waters XSelect^® HSS T3 (2.1 × 150 mm, 2.5 μm); 30 °C, 5 µL injection, 0.35 mL/min flow. ESI-MRM; curtain gas 35 psi, spray voltage ± 4500 V, nebulizer/auxiliary gas 60 psi, 500 °C.

Metabolome Detection and Analysis

The metabolome detection and analysis were carried out through the following steps: (1) sample collection and preparation: embryos from the respective seeds were immediately frozen in liquid nitrogen. The samples were slowly thawed at 4 °C and then an appropriate amount of each sample was added to a pre-chilled methanol/acetonitrile/H₂O (2:2:1, v/v/v). The mixture vortexed and then subjected to low-temperature sonication for 30 min. After allowing the mixture to stand at − 20 °C for 10 min, it was centrifuged at 14,000 g for 20 min at 4 °C. The supernatant was then vacuum-dried and 100 µL acetonitrile/H₂O solution (1:1, v/v) was added to the dried residue to reconstitute it. The reconstituted solution was vortexed and then centrifuged again at 14,000 g for 15 min at 4 °C. Finally, the supernatant was collected for injection into the mass spectrometer for analysis. (2) Chromatography-Mass spectrometry: the samples were analyzed using the ultra-high-performance liquid chromatography (UHPLC) system (1290 Infinity LC) with a C-18 column for separation. To mitigate the impact of instrument signal fluctuations, samples were analyzed in a random sequence continuously. QC (Quality Control) of the samples were interspersed within the sample queue to monitor and assess the system’s stability and the reliability of the experimental data. The AB Triple TOF 6600 mass spectrometer was utilized for the acquisition of both primary and secondary spectra from the samples. (3) Data processing: the data were converted to MzXML files using ProteoWizard. Subsequently, software MSDAIL was used for peak alignment, retention time correction, and peak area extraction. (4) Data statistical analyses: principal component analysis (PCA) and orthogonal partial least-squares discriminant analysis (OPLS-DA) were utilized to conduct multidimensional statistical analysis. Seven-fold cross-validation and response permutation testing were used to evaluate the robustness of the model. To ensure its effectiveness, a permutation test was employed to validate the model. By employing OPLS-DA with a variable importance in the projection (VIP) threshold > 1 and P < 0.05 as the criteria for identifying DAM. The identified DAMs were then annotated using the KEGG compound database and mapped to the KEGG pathway database (http://www.genome.jp/kegg/pathway.html).

The Identification of Putative Long-Lived mRNAs (LLRs)

The putative LLRs for SS were identified using the following stringent criteria according to previous studies with minor modifications (Sano et al. 2015; Liang et al. 2023): (1) mRNAs commonly detected in 9311 and NIP; (2) FPKM-fragments per kilobase per million (FPKM) values cutoff value ≥ 1 for gene expression in the AR seeds groups from both 9311 and NIP; (3) log₂FC(NIP-12–18 m) ≤ − 1 and log₂FC (9311–12 m or 18 m) < 0; and (4) log₂FC(9311–12 m or 18 m)log₂FC(NIP-12–18 m) > 0. For the identification of putative LLRs affecting SD: (1) mRNAs commonly detected in 9311 and NIP; (2) FPKM cutoff values ≥ 1 for gene expression in the FH seeds groups from both 9311 and NIP; (3) log₂FC (9311-AR or 9 m) < 0 and log₂FC(NIP-AR or 9 m) ≤ − 1; and (4) log₂FC(9311-AR or 9 m)log₂FC(NIP-AR or 9 m) > 0.

Gene Expression Evaluated by qRT-PCR

To ensure amplification efficiency and detection sensitivity, ten genes meeting the criteria of |log₂FoldChange| ≥ 1.0, P < 0.05, and FPKM ≥ 10 in at least one sample were selected from the RNA-seq-identified DEGs for quantitative reverse transcription polymerase chain reaction (qRT-PCR) validation. The “actin” gene was utilized as an internal control. The primers used for the tested ten DEGs in this study are listed in Table S1. The data presented for gene expression analysis was calculated using the average values from three independent biological replicates for each sample. The relative expression of each DEG was evaluated using the 2^−ΔΔCt method (Livak et al. 2001).

Results

Phenotype Differences Between NIP and 9311 for SD and SS

Figure 1 shows the differences between NIP and 9311 for all measured SD and SS traits. Significant differences were observed between NIP and 9311 in the levels of SD and SS. The germination rate at day 7 after imbibition (G7d) and Seedling establishment rate at day 7 after imbibition (S7d) of FH seeds of NIP were 33.0% and 25.6%, much lower than that (97.0% and 94.4%) of the 9311 FH seeds. Additionally, NIP exhibited significantly lower GI and SI than 9311 (Fig. 1A-E). After dormancy was broken, the germination rates of the NIP and 9311 seeds reached over 90.0% (Fig. 1A, B), implying that the NIP FH seeds exhibited a much stronger SD than those of 9311. During the period of storage, the NIP and 9311 seeds maintained a high germination rate (G7d > 90.0%) even after 9 months of the natural storage (Fig. 1F-H), indicating that SS of both NIP and 9311 was not significantly decreased. However, with the increase of storage time, the 9311 seeds were able to maintain relatively high germination rates after 12 and 18 months of natural storage, whereas the germination and seedling establishment rates of the NIP seeds decreased significantly after 12 months and 18 months of storage (Fig. 1I-N), indicating that 9311 exhibited much higher SS than NIP. In summary, NIP showed relatively strong SD but much lower SS, while 9311 exhibited weaker SD and higher SS.

Fig. 1.

Differences in seed dormancy (SD), germination and seed storability (SS) between 9311 and Nipponbare (NIP). A, B The germination rate at day 7 after imbibition (G7d) and seedling establishment rate at day 7 after imbibition (S7d) behavior of the freshly-harvested (FH) NIP and 9311 seeds, the after-ripened seeds (AR), and the seeds after 9-month (9 m), 12-month (12 m), and 18-month (18 m) storage under the natural conditions. C, F, I, L Germination curves of dormant and stored seeds of NIP and 9311 monitored for seven consecutive days. Germination (%) represents germination rate. The vertical axis represents germination rate and seedling establishment rate, and the horizontal axis represents the number of days after imbibition. D, G, J, M Germination index (GI) of dormant and stored seeds in NIP and 9311. E, H, K, N Seedling establishment index (SI) of dormant and stored seeds in NIP and 9311. GI and SI were calculated as GI = ∑(Gt/t) and SI = ∑(St/t), respectively, where Gt and St represented the numbers of germinated seeds and seedlings on day t. The error bar indicates the mean ± standard deviation (n = 3). Double asterisks represent significant differences at P value < 0.01

Transcriptomic Differences Between NIP and 9311 Related to SD and SS

To elucidate the molecular mechanisms underlying SD and SS in rice, we conducted a comprehensive transcriptome analysis of the NIP and 9311 seed samples in ten groups, which included the FH and AR seeds of NIP and 9311, as well as their seeds stored for 9, 12, and 18 months. To discern the differential expression among the seed groups and the consistency across the three biological replicates, we conducted quality assessment of the RNA-seqing (RNA-seq) data using the principal component analysis (PCA) method (Fig. 2A). The PCA of the transcriptome data indicated that the first two PCs classified the transcriptomic data into four clusters (Fig. 2A). PC1 explained 97.0% of the total gene expression data and separated the seed samples of NIP from those of 9311. PC2 explained 2.0% of the total gene expression data and separated FH seeds and AR seeds and/or stored seeds of both NIP and 9311. Consistent results could be seen in the heatmap of the Fragments Per Kilobase of transcript per Million mapped fragments (FPKM) values (Fig. 2B), which depicted ten samples into two primary clusters (NIP and 9311). Furthermore, seeds stored for 9 m, 12 m, and 18 m clustered together, indicating that a large number of genes were stably expressed during storage. In contrast, FH seeds and AR seeds formed distinct clusters, showing clear differences in gene expression profiles between dormant and after-ripened seeds (Fig. 2B). To further verify the reliability of the transcriptome results, we validated the accuracy of the RNA-seq data by detecting ten selected DEGs using the qRT-PCR method (Fig. 2C). The results showed that the expression patterns of the randomly selected DEGs detected by the two methods were highly consistent (r² = 0.98).

Fig. 2.

Transcriptome analyses of the dormant seeds, AR seeds, stored and non-stored seeds of NIP (Geng) and 9311 (Xian). A The principal component analysis (PCA) plot of the RNA sequencing results. B The hierarchical cluster analysis of all expressed genes in all seed samples. C Validation of the accuracy of transcriptome data using quantitative reverse transcription polymerase chain reaction (PCR). Expression was transformed by log₂ fold change in expression. D The number of up-regulated and down-regulated differentially expressed genes (DEGs) in each group. E The different and common DEGs between NIP-FH vs. NIP-AR and 9311-FH vs. 9311-AR. F The different and common DEG between NIP-12 m vs. NIP-AR and NIP-18 m vs. NIP-AR. G The different and common DEG between 9311–12 m vs. 9311-AR and 9311–18 m vs. 9311-AR. H The different and common DEGs between SD and SS. I, J The different and common pathways between SD and SS

DEGs and Pathways for SD

Based on the threshold of two-fold difference in gene expression level between the FH and AR (CK) seeds, a total of 2420 DEGs were detected in NIP-FH, with 738 being significantly up-regulated and 1,682 significantly down-regulated. In 9311-FH, 1,692 DEGs were detected, including 738 significantly up-regulated and 954 significantly down-regulated. A total of 380 DEGs were detected in both NIP-FH and 9311-FH (Fig. 2D, E). Among these, 149 up-regulated genes and 203 down-regulated genes were commonly identified in both NIP-FH and 9311-FH, respectively (Fig. S1A, B). For all SD-related DEGs detected, the top 20 pathways where DEGs were significantly enriched were shown (Fig. S2A). For instance, DEGs were significantly enriched in the metabolic pathways and plant hormone signal transduction pathways (Fig. S2A). However, when respectively analyzing up-regulated and down-regulated DEGs in NIP and 9311, KEGG analysis revealed that the up-regulated genes for SD of NIP were significantly (P < 0.05) involved in two pathways of protein processing in endoplasmic reticulum (ko04141) and starch and sucrose metabolism (ko00500), while those in 9311 for SD were significantly involved in six pathways with only protein processing in endoplasmic reticulum overlapped between NIP and 9311 (Table S2). In contrast, those down-regulated genes for SD in NIP were significantly enriched in 15 pathways, while those down-regulated genes in 9311 were significantly involved in eight pathways with six pathways overlapped between NIP and 9311. These commonly down-regulated pathways included phenylpropanoid biosynthesis (ko00940), biosynthesis of secondary metabolites (ko01110), metabolic pathways (ko01100), glutathione metabolism (ko00480), sulfur metabolism (ko00920) and cysteine and methionine metabolism (ko00270), indicating their important roles in SD. Notably, pathway starch and sucrose metabolism (ko00500) was up-regulated in NIP but down-regulated in 9311 (Table S2).

DEGs and Pathways for SS

DEGs for SS were identified between the AR (CK) and 12-month/18-month stored seeds. In NIP-12 m and NIP-18 m, a total of 3,776 and 4,594 DEGs were detected respectively, with 437 overlapping between them (Fig. 2F). In 9311–12 m and 9311–18 m, 759 and 1,357 DEGs were detected respectively, with 343 overlapping between them (Fig. 2G). Additionally, 2,071 and 716 genes were up-regulated for SS in NIP and 9311 with 199 ones overlapped between NIP and 9311 (Fig. S1C). The number of down-regulated genes for SS was 3,621 in NIP and 1,076 in 9311 with 327 overlapped ones (Fig. S1D). KEGG analysis indicated that the up-regulated genes for SS in NIP were enriched in eight pathways, while those in 9311 were enriched in ten pathways with four pathways overlapping, including biosynthesis of various plant secondary metabolites (ko00999), amino sugar and nucleotide sugar metabolism (ko00520), and glutathione metabolism (ko00480), and protein processing in endoplasmic reticulum (ko04141) (Table S2). The down-regulated genes for SS in NIP were enriched in 22 pathways and those in 9311 were enriched in 12 pathways with seven common ones, i.e. phenylpropanoid biosynthesis (ko00940), biosynthesis of secondary metabolites (ko01110), metabolic pathways (ko01100), starch and sucrose metabolism (ko00500), arachidonic acid metabolism (ko00590), taurine and hypotaurine metabolism (ko00430) and betalain biosynthesis (ko00965) (Table S2).

Regulatory Genes Underlying SD and SS

When comparing the DEGs between SD and SS groups, 1,334 (13.0%) DEGs were found to affect SD and SS simultaneously (Fig. 2H). Interestingly, compared to the 9311 groups, more DEGs were detected in NIP. Additionally, the number of DEGs in NIP increased gradually with storage time in the naturally aged seeds, but not so in 9311 (Fig. 2D). These results suggested that there were significant changes in gene expression in NIP, while gene expression in 9311 was relatively stable. Among the DEGs, six cloned genes, OsABI3, OsABI5, OsGA2ox9, OsEm1, OsLEA3, and Sdr4 associated with SD or seed germination were identified. The expression of five ABA-related genes was significantly up-regulated in the 9311 FH seeds, and the opposite was true for one GA-related gene (OsGA2ox9). In addition, Sdr4 was significantly up-regulated in the FH seeds of both 9311 and NIP (Fig. S3A). The identified DEGs included eight cloned genes for SS (OsGH3-2, OsHSP18.0-CI, OsGLYI3, OsLOX2, OsLOX3, OsFAH2, Os4BGlu14, and OsCSD2). Among these, OsLOX2 and OsLOX3 encoding lipoxygenase, and OsFAH2 encoding fatty acid 2-hydroxylase, were significantly down-regulated in the 12- and 18-month stored NIP seeds, but significantly up-regulated in the AR or 12- and 18-month stored seeds of 9311, and the opposite was true for OsGLYI3 encoding glyoxalase. In contrast, Os4BGlu14 (encoding monolignol β-glucosidase) and OsHSP18.0 encoding a heat shock protein were significantly up-regulated in the AR and/or 12- or 18-month stored seeds of NIP, but significantly down-regulated in the corresponding seeds of 9311. OsGH3-2 encoding indole-3-acetic acid-amido synthetase and OsCSD2 encoding copper/zinc superoxide dismutase were the only two showing similar expression patterns in NIP and 9311. The former was significantly up-regulated in the 18 m stored seeds of both NIP and 9311, while latter were significantly up-regulated in the AR seeds of NIP and 9311 (Fig. S3B).

To further investigate whether SD and SS share the same DEGs, we first performed correlation analysis using the FPKM values of all expressed genes across all groups (Fig. S4). The correlation analysis revealed that the expression of genes was significantly positively correlated within each group in both NIP and 9311 (r > 0.90). Furthermore, there was also a significant correlation between NIP and 9311 groups, with the correlation between genes in dormant seeds and aged seeds reaching 0.63 and 0.79 (P < 0.01) in the 25 common DEGs between SD and SS (Table S4). These findings suggest that some genes may participate in the same pathways that regulate SD and SS.

Regulatory Pathways Underlying SD and SS

KEGG analysis was firstly conducted on all detected DEGs related to SD and SS, respectively (Fig. 2I; Fig. S2). The DEGs for SD were significantly (P < 0.05) involved in 16 pathways, and the DEGs for SS were significantly involved in 23 pathways. Among these pathways, five were specifically associated with SD, 12 were specifically associated with SS, and 11 (39.3%) pathways were common to both (Fig. 2I, J). For the 11 common pathways, 618 DEGs for both SD and SS were significantly enriched (Table S4), including 25 common DEGs (20.8%) in plant hormone signal transduction (ko04075), eight common DEGs (27.6%) in fatty acid elongation (ko00062), 19 common DEGs (22.4%) in amino sugar and nucleotide sugar metabolism (ko00520), 180 common DEGs (26.0%) in biosynthesis of secondary metabolites (ko01110), eight common DEGs (34.8%) in cutin, suberine and wax biosynthesis (ko00073), ten common DEGs (30.3%) in diterpenoid biosynthesis (ko00904), 17 common DEGs (45.9%) in DNA replication (ko03030), 22 common DEGs (28.9%) in glutathione metabolism, 253 common DEGs (24.4%) in metabolic pathways (ko00480), 53 common DEGs (33.5%) in phenylpropanoid biosynthesis (ko00940), and 23 common DEGs (23.7%) in and starch and sucrose metabolism (ko00500) pathways (Table S4).

The DEGs for SD and SS in Plant Hormone Signal Transduction Pathway

The functional genes involved in plant hormone signal transduction pathway have been extensively documented as being closely linked to seed germination. Within the common pathways, plant hormone signal transduction pathway was found to harbor 49 and 71 DEGs for SD and SS, respectively. To understand this, we found 25 common DEGs were involved in plant hormone signal transduction pathway, and these genes could be classified into six categories of plant hormones based on their annotated functions (Fig. 3A-C; Table S4). Based on their expression patterns, the common regulatory DEGs could be grouped into two clusters (Fig. S5). The first cluster consisted of 11 DEGs whose expression was significantly down-regulated in either dormant seeds and/or stored seeds of NIP, but up-regulated in either dormant seeds and/or stored seeds of 9311. This group contained seven genes (OsIAA7, OsIAA14, OsSAUR8, OsSAUR22, OsSAUR27, OsbZIP64, and OsbZIP70) involved in the regulation of IAA, GA (OsPIL16), and BR (OsBZR1), plus OsCYCD3 (a D-type cyclin). The second cluster contained 14 DEGs, including OsPYL9, OsPP2C8, OsPP2C51, OsbZIP46 and OsbZIP62 involved in ABA, OsEIL5, OsETR4 and OsERF_087 involved in ET, OsJAZ10, OsJAZ11 and OsJAZ13 involved in JA and OsSAUR21 in IAA plus Os05g00439100 and Os06g0696566 (Fig. S5). These genes, except for OsSAUR18, were significantly up-regulated in either dormant and/or stored seeds of NIP, but down-regulated in one or more seed samples of 9311. Similarly, 19 additional DEGs involved in IAA for both SD and SS showed contrasting expression patterns in dormant and stored seeds of NIP and 9311, and so for 24 genes specifically for SD related to ABA (5), JA (4), IAA (6), ET (1), GA (1) and SA (1), exhibited contrasting expression patterns in NIP and 9311 (Fig. 3C; Fig. S5).

Fig. 3.

The DEGs for SD and SS involved in plant hormone signaling pathways. A The different and common DEGs between SD and SS involved in plant hormone signaling pathways. B The proportions of the 25 DEGs commonly detected in both SD and SS involved in plant hormone pathways. The DEGs without clearly identified pathways were excluded. C The positions of the corresponding DEGs in the abscisic acid (ABA), gibberellic acid (GA), and indole-3-acetic acid (IAA) pathways. The bold indicates that the DEGs belong to LLRs

In addition to the 25 common genes, there were 46 DEGs specifically associated with SS and 24 DEGs specifically associated with SD were presented in the plant signaling pathway (Fig. 3A). Among the 46 DEGs for SS, four were found to be related to the ABA pathway. Among them, two bZIP transcription factors (OsbZIP46 and OsbZIP62) were detected. OsbZIP62 was significantly up-regulated in the seeds after 18-month storage in NIP, while OsbZIP46 was down-regulated in both the seeds after 12- and 18-month storage in NIP (Fig. 3C). Furthermore, two genes related to ET and one gene related to GA pathways were identified, and all three genes were observed to be down-regulated in NIP seeds after 12 and/or 18-month of storage. Two genes related to JA pathway were identified, with OsJAZ8 and OsJAZ9 displaying distinct expression patterns in the NIP seeds after storage. A total of 19 DEGs related to IAA, the majority of IAA genes exhibited significant changes in expression in NIP seeds after storage, but not so in 9311 seeds (Fig. S6A). Among the 24 DEGs that specifically affect SD, there were five genes related to ABA, one gene each related to ET, GA, and SA, four genes associated with JA, six genes associated with IAA. The majority of genes related to IAA, JA, and ET pathways were significantly down-regulated in freshly harvested NIP seeds, while genes related to the ABA pathway were significantly up-regulated (Fig. 3C; Fig. S6B).

Differentially Expressed Transcription Factors (TFs) for SD and SS

The DEGs also included 98 TFs specifically identified for SS (30.2%), 162 TFs specifically for SD (50.0%), and 64 TFs for both (19.8%). These TFs belong to 41 gene families (Fig. S7; Table S5). For example, OsBZR1, encoding a BR signaling factor, was identified as a common DEG for SD and SS. OsBZR1 has been found to significantly influence PHS, but its role in controlling SS remains unclear. We found this gene was differentially expressed in dormant and aged seed groups. Compared to the AR control in NIP, OsBZR1 was significantly down-regulated in NIP-FH, and in the NIP-12 m and NIP-18 m stored seeds. However, OsBZR1’s expression remained unchanged all the 9311 samples, though much lower than its expression in the NIP samples (Fig. S7B, C).

Putative Long-Lived RNAs (LLRs) for SD and SS

The stability of long-lived RNAs in seeds is known to be required for seed germination and thus are closely related to both SD and SS. Here, we identified 1271 potential candidates of LLRs for SD, and 1976 putative LLRs for SS, with 309 common LLRs (Fig. 4A). Based on the expression changes of all detected genes in all assayed seed samples, these putative LLRs could be divided into nine clusters, each representing a different expression pattern (Fig. 4B-D), including 32.2% and 25.0% of the LLRs affecting SD in clusters 3 and 4, and 23.7% and 52.4% of the LLRs affecting SS in clusters 5 and 7. KEGG analysis indicated that these putative LLRs for SD were significantly (P < 0.05) enriched in five pathways, including plant hormone signal transduction, terpenoid backbone biosynthesis, glycerolipid metabolism, sesquiterpenoid and triterpenoid biosynthesis, and protein processing in endoplasmic reticulum pathway (Fig. 4E; Table S6). For SS, 515 out of the 1,976 putative LLRs were significantly involved in 11 pathways, including plant hormone signal transduction, terpenoid backbone biosynthesis, biosynthesis of secondary metabolites, metabolic pathways, phenylpropanoid biosynthesis, amino sugar and nucleotide sugar metabolism, butanoate metabolism, MAPK signaling pathway-plant, riboflavin metabolism, pentose phosphate, and sphingolipid metabolism pathway (Fig. 4F; Table S6). Only two common pathways were involved in SD and SS, including terpenoid backbone biosynthesis and plant hormone signal transduction. For terpenoid backbone biosynthesis pathway, 11 and 7 putative LLRs were identified for SS and SD, respectively, and three putative LLRs were commonly identified. In the plant hormone signal transduction pathway, 31 and 20 putative LLRs were identified for SS and SD, respectively. Among these putative LLRs, five (OsPYL9, OsSIPP2C1, OsSAPK3, OsbZIP46, and OsABI5) were involved in the ABA pathway; three (Os-ETR3, OsEIN2L, and OsEIL5) were involved in the ET pathway; and 13 LLRs (OsIAA6, OsIAA7, OsIAA10, OsIAA11, OsIAA12, OsIAA13, OsIAA14, OsSAUR8, OsSAUR21, OsSAUR22, OsSAUR25, OsSAUR27, and OsSAUR31) were involved in the IAA pathway (Table S6).

Fig. 4.

Putative long-lived RNAs (LLR) analysis and gene expression patterns across all samples. A The number of putative LLRs identified for SD and SS. B, C Proportions of putative LLRs in the classifications for SS (B) and SD (C). D Classification of the genes based on expression patterns in ten groups. E Kyoto encyclopedia of genes and genomes (KEGG) analysis of putative LLRs identified in SS and SD. Red texts indicates two common pathways (plant hormone signal transduction and terpenoid backbone biosynthesis) identified in SD and SS

Long-Lived Late Embryogenesis Abundant (LEA) Proteins Involved in SD and SS

The genes encoding LEA proteins have been reported to belong to LLRs and affect SS in both rice and wheat. We preliminarily filtered out 41 OsLEAs from RNA-seq. After removing the genes with low or no expression, non-LLRs, and non-DEGs, 15 putative LLR-OsLEAs were finally identified (Fig. 5). According to their protein sequences, these OsLEAs were categorized into two groups, with OsLEA15, OsEm1, OsLEA30, OsLEA33, OsLEA31, OsLEA12, OsLEA16, OsLEA3-2, OsLEA14, and OsLEA3-1 in group I and OsLEA1a, OsLEA5, OsLEA3, OsLEA1, and OsLEA4 in group II (Fig. 5A). These OsLEAs exhibited distinct expression patterns between NIP and 9311, as well as in dormant and stored seeds. Compared to their expression levels in corresponding controls, all 15 OsLEAs, except for OsLEA5, were significantly up-regulated in 9311-FH seeds (average log₂FC = 3.2), whereas their expression remained unchanged in the NIP-FH seeds (average log₂FC = 0.49). Additionally, their expression decreased significantly in the NIP-12 m (average log₂FC = − 0.77) and NIP-18 m (average log₂FC = − 1.69) stored seeds as compared to the control (Fig. 5B). Detailed examination of the expression profiles of these 15 OsLEAs in 11 tissues of NIP revealed that the majority of these genes showed no or low expression in 20-day leaves, post- and pre-emergence inflorescence, anthers, pistils, shoots, and four-leaf seedlings. However, all 15 OsLEAs, except for OsLEA5, expressed strongly and specifically in embryos-25 DAP (Fig. 5C). Fig. S8A-C show the expression profiles of these OsLEAs under different abiotic stress and hormone treatment conditions. When treated with ABA, 14 of these OsLEAs showed significantly and gradually increased expression in roots over time, indicating they were associated with the ABA pathway. Under the Cd ion stress condition, the 14 OsLEAs showed significantly increased expression, except for OsLEA30 which exhibited decreased expression. Under the drought treatment, 13 of the OsLEAs exhibited significantly increased expression in roots, while the expression of OsLEA12 and OsLEA31 were decreased and no change in response to drought. Under the cold stress, 11 of the OsLEAs showed significantly increased expression in roots. In addition, expression of ten OsLEAs was significantly up-regulated under the salt treatment. Eight of the OsLEAs showed significantly increased expression in response to all five external treatments (Fig. S8A-C), while expression of OsLEA14 in roots were strongly up-regulated in response to ABA, drought, cold, salt, and Cd treatments (Fig. S8D).

Fig. 5.

Analysis of the rice gene family of 15 long-lived mRNA-OsLEAs (LLR-LEAs). A Neighbor-joining (NJ) tree plot of the 15 detected LLR-LEAs based on their protein sequence. B Expression pattern analysis of the 15 LLR-LEAs in dormant and aged seeds. C Expression pattern analysis of the LLR-LEAs in 11 tissues from NIP. S1, Leaves-20 days; S2, Post-emergence inflorescence; S3, Pre-emergence inflorescence; S4, Anther; S5, Pistil; S6, Seed-5 days after pollination (DAP); S7, Embryo- 25 DAP; S8, Endosperm- 25 DAP; S9, Seed- 10 DAP; S10, Shoots; S11, Seedling four-leaf stage. Red font indicates that these genes specifically or highly expressed in the corresponding tissue. (D) Schematic diagram of gene editing loci. E, F Germination (E) and seedling establishment rate (F) over seven consecutive days in wild-type NIP and Oslea5 mutant seeds after 9 months of storage. G, H Germination (G) and seedling establishment rate (H) over seven consecutive days in wild-type NIP and Oslea5 mutant after 12 months of storage. Germination (%) and seedling (%) represent germination rate and seedling establishment rate, respectively. The vertical axis represents germination rate and seedling establishment rate, and the horizontal axis represents the number of days after imbibition. The error bar indicates the mean ± standard deviation (n = 3). Double asterisks represent significant differences at P value < 0.01

OsLEA5 Regulates Seed Storability

A stably expressed DEG (OsLEA5) from the LEA gene family was selected for gene knockout to characterize its biological function (Fig. 5D-H). OsLEA5 is stably and widely expressed in various rice tissues, but exhibits genotype-specific changes during seed storage. Expression analysis showed that compared with unstored seeds, OsLEA5 expression was significantly down-regulated in seeds of the NIP cultivar stored for 12-months, while no significant change was observed in seeds of the 9311 cultivar stored for the same period. Two homozygous mutants (designated KO1 and KO2) were obtained via gene editing. Both KO1 and KO2 had an “A” nucleotide inserted at the first editing site. At the second editing site, KO1 had a “T” nucleotide inserted whereas KO2 had an “A” nucleotide inserted, resulting in premature translation termination. Germination curve analysis revealed that after 9-months of storage, the seed germination rate of Oslea5 knockout lines was significantly lower than that of the WT-NIP. Compared with seeds stored for 9-months, the germination rates of both WT and KO lines further decreased after 12-months of storage, but the germination rate of KO lines remained consistently lower than that of WT. Based on the full-length OsLEA5 (2 kb promoter and CDS regions) SNP data from 242 rice accessions and their 18 m storability phenotypes, we identified 3 SNPs (1 coding, 2 promoter) defining 3 major haplotypes (frequency > 5%). These haplotypes showed distinct subspecies distribution, Hap1 was predominantly in Geng rice accessions, while Hap2 and Hap3 were almost exclusive to Xian accessions. The mean germination rate of Hap1 (83.7%) was significantly lower than Hap2 (97.1%) and Hap3 (98.4%), with 13.4% and 14.7% increases respectively, confirming OsLEA5 as a key regulator of rice seed storability (Table S7). Taken together, phenotypic validation of the knockout lines confirms that OsLEA5 is a key regulator of seed storability in rice.

Metabolomic Differences for SD and SS

To elucidate the metabolic profiles in rice seeds, the same seed samples subjected to RNA-seq analysis were used for untargeted metabolomics analysis utilizing the UHPLC-Q-TOF MS technology. PCA and heatmap representations of the metabolite abundance in the NIP and 9311 seed samples revealed similar clustering patterns to those obtained from the RNA-seq data, with PC1 explaining 52.9% of the total variation in metabolite abundance and separating the seed samples from NIP and 9311, and PC2 explaining 22.9% of the total metabolite variation and separating the FH seeds from the stored seeds (Fig. 6A, B). The identified metabolites included diverse chemical categories, comprising lipids and lipid-like molecules (25.2%), phenylpropanoids and polyketides (17.3%), benzenoids (10.2%), organoheterocyclic compounds (9.9%), organic acids and derivatives (8.2%), organic oxygen compounds (8.0%), nucleosides and analogues (2.4%), organic nitrogen compounds (2.1%), alkaloids and derivatives (1.7%), lignans, neolignans and related compounds (0.9%), and others (14.0%) (Fig. 6C).

Fig. 6.

Metabolomic analysis of AR, dormant, stored, and non-stored seeds. A the PCA plot of all detected metabolites in different seed samples. B The hierarchical clustering of all metabolites in different seed groups. The red color shows relatively high metabolite abundance, while blue indicates relatively low metabolite abundance. C Classification of the 1,280 metabolites in all assayed seed samples. D The differential accumulation metabolites (DAMs) identified in different seed groups. E-G The different and common DAMs between different seed groups. H The different and common DAMs between SD and SS. I-K The different and common pathways between different groups. L The different and common pathways between SD and SS

To pinpoint the factors underlying the differences between the control (AR) and treatment (SD and SS) seed groups, we performed the orthogonal partial least squares discriminant analysis (OPLS-DA) on the metabolomic data to assess the relative contributions of various metabolite groups in different seed samples. Compared with the control group (AR), 62 DAMs were detected in NIP-FH and 89 DAMs were detected in 9311-FH with 16 DAMs shared between the FH seeds of NIP and 9311. After 12 months of natural storage, 82 DAMs were detected in NIP, and 61 DAMs were identified in 9311. After 18-months natural storage, 74 and 76 DAMs were detected in NIP and 9311, respectively (Fig. 6D). Among these, 38 and 26 DAMs were commonly detected in both NIP and 9311 after 12- and 18-months of natural storage, respectively. After comparing NIP and 9311, 101 DAMs for SD were detected between NIP-FH and 9311-FH. Following storage for 12 months, 127 DAMs for SS were identified between NIP and 9311. After 18 months of storage, 104 DAMs for SS were detected between NIP and 9311 (Fig. 6E-G). Analysis of metabolites that influence SD and SS revealed that 42 DAMs were specifically detected in dormant seeds, 141 were specifically detected in stored seeds, and 93 were detected in both SD and SS (Fig. 6H).

KEGG analysis revealed that the DAMs for SD were significantly (P < 0.05) involved in 25 pathways, while the DAMs for SS were significantly involved in 18 pathways for 9311 (Fig. 6I-L). Among the pathways, 12 for NIP group and 24 pathways for 9311 group were detected for SD, and 11 common pathways were found to be shared between NIP and 9311 (Fig. 6I). For SS, nine and six pathways were identified in NIP-12 m vs. CK and NIP-18 m vs. CK, respectively (Fig. 6J). In 9311 groups, 11 and seven pathways were identified in 9311–12 m vs. CK and 9311–18 m vs. CK, respectively, and four were common (Fig. 6K). When comparing the all groups for SD and SS, the results indicated that nine pathways were specifically identified for SD, two were specifically identified for SS, and 16 pathways (59.3%) were common (aminoacyl-tRNA biosynthesis, phenylpropanoid biosynthesis, pyruvate metabolism, biosynthesis of amino acids, histidine metabolism, glyoxylate and dicarboxylate metabolism, ABC transporters, phenylalanine, tyrosine and tryptophan biosynthesis, 2-Oxocarboxylic acid metabolism, glucosinolate biosynthesis, biosynthesis of various secondary metabolites-part 2, carbon metabolism, citrate cycle (TCA cycle), C5-Branched dibasic acid metabolism, cyanoamino acid metabolism, and phenylalanine metabolism) (Fig. 6L; Table S8).

Endogenous Phytohormone Content Analysis

LC-MS/MS was used to determine the endogenous hormone contents in embryos of rice cultivars 9311 and NIP under three conditions: fresh dormancy, 18-month storage, and dormancy breaking (Table S9). The results showed that the ABA content of NIP in freshly dormant seeds was 13.51 ng/g, 2.48-fold higher than that of 9311 (5.45 ng/g). After dormancy breaking, the ABA contents of both cultivars decreased significantly with no obvious genotypic difference. After 18-month room-temperature storage, the ABA levels further declined in both varieties, with NIP maintaining a significantly lower ABA concentration than 9311. Analysis of four active gibberellins (GA1, GA3, GA4, GA7) revealed that the contents of all GA components were markedly higher in 9311 than in NIP under fresh dormancy. Upon dormancy release, the levels of GA1, GA4 and GA7 increased obviously in 9311; by contrast, only GA1 slightly increased in NIP, while other GA components remained at low levels. After 18-month storage, 9311 still maintained a relatively high endogenous GA level, whereas most GA fractions in NIP remained consistently low. For endogenous IAA, the concentration in NIP (169.80 ng/g) was significantly higher than that in 9311 (86.29 ng/g) at the fresh dormant stage, and decreased in both cultivars after dormancy breaking. Following 18-month storage, the IAA content of 9311 dropped to 22.35 ng/g and stayed at a low level, while IAA accumulated abnormally up to 303.49 ng/g in NIP, which was substantially higher than that in 9311. The endogenous JA content was slightly higher in NIP than in 9311 in freshly dormant seeds. JA levels decreased distinctly in both cultivars after dormancy breaking, and further declined to a range of 1.95–11.61 ng/g after 18-month storage. In terms of SA content, 9311 exhibited a significantly higher level (622.06 ng/g) than NIP (319.29 ng/g) under fresh dormancy, and this intervarietal difference persisted after dormancy breaking. After 18-month storage, SA content continued to rise and remained at a high level in 9311, while SA in NIP decreased further to a relatively low level (Table S9). Collectively, dormancy status and room-temperature storage markedly modulate the accumulation of endogenous hormones in rice seeds, and 9311 and NIP exhibit distinct genotypic differences in phytohormone response patterns.

Combined Transcriptome and Metabolome Analyses

Data from the metabolomic and transcriptomic analyses were combined to explore the differences in metabolome contents and gene expression patterns for SD and SS. To identify the KEGG pathways commonly enriched in the transcriptome and metabolome, all the identified DEGs and DAMs were mapped to 151 pathways. Among the pathways, linoleic acid metabolism pathway was enriched in all groups and represented distinctive patterns among the groups. By examining the genes involved in this pathway, nine OsLOX genes with differential expression were found. Interestingly, the LOX gene family has been widely reported to be associated with SS in many plants. The nine OsLOXs could be categorized into two major groups based on their protein sequences (Fig. S9A). The first group included OsLOX1, OsHI-LOX, OsLOX6, OsLOX1-r9, and OsLOX2;L-2, and the second group consisted of OsLOX8, OsLOX7, OsLOX5, and Os03g0708000. Analysis of the expression profiles of nine OsLOXs in 11 tissues from NIP revealed that the majority of genes exhibit a constitutive expression pattern. Among them, five genes were highly expressed in roots (Fig. S9B). According to their expression patterns from NRA-seq results (Fig. S9C), two genes (OsLOX5 and OsLOX7), were specifically down-regulated in the dormant NIP-FH and 9311-FH seeds. In addition, five OsLOXs were specifically detected in stored seeds. Among these, three genes (OsLOX1, OsLOX1-r9, and Os03g0708000) were significantly down-regulated in the NIP and/or 9311 aging seeds. Two genes (OsHI-LOX and OsLOX6) were significantly down-regulated in NIP-18 m and/or NIP-12 m, but showed no changes in the 9311 aging seeds. In addition, two genes (OsLOX8 and OsLOX2;L-2) were identified in both dormant and aging seeds (Fig. S9C). For example, compared to their respective controls, OsLOX8 was significantly up-regulated in 9311-FH, NIP-12 m, and NIP-18 m, but showed no changes in NIP-FH, 9311–12 m, and 9311–18 m. We further analyzed the correlation between the expression levels of all nine OsLOXs and their SS phenotypes. The results showed that six genes (OsLOX6, OsLOX1-r9, OsLOX2;L-2, OsLOX7, OsLOX8, and Os03g0708000) had a significant correlation between their FPKM values and the SS phenotypic values. Among them, the expression levels of OsLOX6 and OsLOX8 were significantly negatively correlated with the seed germination phenotype. Additionally, the expression levels of the other three OsLOXs were significantly positively correlated with the seed germination phenotype (Fig. S9D), indicating that these six genes may be associated with seed germination at the transcriptional level. For example, OsLOX2;L-2 was significantly down-regulated in five groups. In NIP, OsLOX2;L-2 expression gradually increased with maturation, reaching a peak during breaking dormancy, and decreased sharply with increasing storage time. Similarly, in the 9311 group, OsLOX2;L-2 expression gradually increased with maturation, reached a peak at 9311–12 m, and significantly decreased at 9311–18 m (Fig. S9E, F). By comparing the expression trends of this gene in NIP and 9311 with their germination trends, we found that they highly coincided. Through the correlation analysis between OsLOX2;L-2’s FPKM and the phenotype, we found that OsLOX2;L-2 was significantly positively correlated with seed germination rates (G_3d and G_7d) and GI (P < 0.05) (Fig. S9G).

Transgenic Validation of the Dormancy Function of OsGA2ox8

Based on the above transcriptome analysis, we selected OsGA2ox8, a key GA metabolic DEG whose differential expression suggests its involvement in SD regulation, as the research target to investigate its biological functions (Fig. 7A, B). Based on the results of RNA-seq, the expression level of OsGA2ox8 in the NIP-FH seeds was significantly higher than that in the NIP-AR seeds. However, there was no significant difference in the expression level of OsGA2ox8 between the 9311-FH and 9311-AR seeds (Fig. 7B). Gene knockout (KO) and overexpression (OE) vectors were constructed for genetic transformation of NIP. Through genotyping identification, two independent homozygous mutant lines (KO1 and KO2) were obtained. Sequence analysis of the mutants revealed that the OsGA2ox8 gene in KO1 harbored a 2-bp deletion, while KO2 exhibited a 1-bp deletion in the coding sequence region (CDS), leading to a premature translation termination codon, thus forming loss-of-function mutations. Germination assay results showed that functional alterations of OsGA2ox8 significantly impacted NIP SD (Fig. 7A). Compared with the wild type (WT-NIP), the KO lines displayed significantly increased seed germination and seedling establishment rates: the G7d and S7d of fresh-harvested WT seeds were 35.6% and 19.9%, respectively, whereas those of KO1 (G7d, 74.1%; S7d, 61.7%) and KO2 lines (G7d, 48.2%; S7d, 33.7%) were significantly higher, indicating that gene function loss markedly weakened SD. Conversely, the OE lines showed a significantly delayed germination phenotype: G7d and S7d of OE1 (G7d, 22.2%; S7d, 14.4%) and OE2 (G7d, 22.4%; S7d, 13.4%) lines were both significantly lower than those of the WT-NIP, suggesting that OsGA2ox8 positively regulates SD at the transcriptional level (Fig. 7C-F). Taken together, the phenotypic validation results of the KO and OE transgenic lines confirmed that OsGA2ox8 is a key regulator of rice SD.

Fig. 7.

Functional validation of the key DEG (OsGA2ox8) in the GA pathway influencing seed dormancy. A Phenotypes of dormant seeds from gene knockout (KO), overexpression (OE) transgenic lines, and the wild type (WT) at 10 days after germination. Scale bar, 1 cm. B Comparison of OsGA2ox8 transcript levels in FH and AR seeds of NIP and 9311 based on RNA-seq results. C Germination curves and D germination indices of WT and OsGA2ox8 transgenic seeds. Germination (%) represents germination rate. G7d, germination rate at day 7 after imbibition. E Seedling emergence curves and F emergence indices of WT and OsGA2ox8 transgenic seeds. Seedling (%) represents seedling establishment rate. S7d, seedling establishment rate at day 7 after imbibition. The vertical axis represents germination rate and seedling establishment rate, and the horizontal axis represents the number of days after imbibition. (G) Phenotypic differences among haplotypes (Hap) of OsGA2ox8. H, I Proportions of haplotype 1 and haplotype 2 in the Xian H and Geng subspecies. Different lowercase letters in the boxplots indicate significant differences between the two haplotypes at P value < 0.01 based on Duncan’s multiple range test. I The error bar indicates the mean ± standard deviation (n = 3). Double asterisks represent significant differences at P value < 0.01

Based on the genetic variation information and phenotypic data of 277 rice accessions from the 3,000 Rice Genomes Project (obtained via download from the website http://101.201.107.228:8002/#/Rice3KGS/home/), we re-analyzed all high-quality SNPs in the 2 kb promoter region and full-length coding sequence (CDS) of the OsGA2ox8 gene (Fig. 7G, H). The results showed that a total of 31 SNPs were identified in this gene within the tested population, among which 29 SNPs were located in the promoter regulatory region, and 2 SNPs were located in the CDS region with non-synonymous mutations, leading to amino acid substitutions of alanine to threonine and valine to isoleucine, respectively. Effects of 29 SNPs in the promoter region on transcription factor binding sites were analyzed, and a total of 5 key SNPs were identified to cause alterations in the core conserved sequences of core regulatory elements and stress-responsive elements: SNP27913365 (G) affects the CAAT-box, SNP27913447 (C) and SNP27913636 (T) affect the TATA-box, SNP27913616 (G) affects the 3-AF3 binding site, and SNP29071135 (C) affects the STRE stress-responsive element. These variations may significantly influence the binding efficiency of transcription factors in the corresponding signaling pathways, thereby regulating the expression level of OsGA2ox8. After excluding rare haplotypes with a frequency of less than 5%, the gene could be classified into four major haplotypes based on the 31 SNPs. Haplotype distribution analysis across subspecies revealed significant differentiation in the haplotype composition of OsGA2ox8 between the Xian and Geng subspecies (Fig. 7G, H). In the Xian subspecies, Hap1 was the dominant haplotype (accounting for 59.3%), followed by Hap2 (21.2%) and Hap3 (19.5%), and Hap4 was not detected. In the Geng subspecies, Hap4 was the overwhelmingly dominant haplotype (69.9%), followed by Hap3 (25.7%) and Hap1 (only 4.4%), and Hap2 was absent. Phenotypic association analysis showed that seed dormancy was significantly lower in accessions carrying Hap4 than in those carrying the other three haplotypes (P value < 0.01), indicating that haplotype differentiation of the OsGA2ox8 gene is closely associated with the seed dormancy phenotype.

Discussion

SD and SS are two related traits contributing to rice seed quality and indirectly to eating and economic values of rice. In the present study, 9311 and NIP exhibited striking differences in SD and SS. 9311 displayed weak SD but strong SS, whereas NIP showed strong SD but poor SS. Through comprehensive transcriptomic and metabolomic analyses, we found that the differential expression of distinct regulatory gene sets in NIP and 9311 correlated with divergent expression profiles of large numbers of downstream genes in multiple pathways, which may contribute to their observed phenotypic differences in SD and SS. While our findings provide clear insights into the molecular differences underlying SD and SS between these two genotypes under the tested conditions, it is important to recognize that all seeds were produced during the winter growing season in Sanya. Given the well-documented sensitivity of SD and SS to maternal environmental cues, the observed phenotypic and molecular patterns may exhibit context-dependent variation across different ecological regions. Multi-year and multi-location trials will be conducted in subsequent work to further assess the stability and broader applicability of our conclusions. While 9311 and NIP represent the most widely used model genotypes for Xian and Geng rice functional genomics research, this study was limited to only two genetic backgrounds. Consequently, the observed regulatory mechanisms of SD and SS may reflect genotype-specific characteristics rather than universal subspecies-level patterns. To address this gap, subsequent studies should incorporate a broad range of rice accessions from different ecological regions to validate and extend our findings to the subspecies level.

The molecular mechanisms underlying the 9311 and NIP differences in SD and SS were reflected in the following three aspects. First, although large numbers of genes were differentially expressed in SD and SS of rice (Table S2), a substantial set of common genes and pathways were shared between SD and SS, indicating coordinated regulation. Downregulation of the five pathways resulted in enhanced SD/SS in both 9311 and NIP, while the contrasting regulation of the former four pathways and involved genes appeared to be the major contributor to the phenotypic divergences in SD and SS between NIP and 9311. The starch/sucrose metabolism pathway showed a slightly different regulation pattern and was down-regulated for SS in both the varieties, and for SD in 9311 only, but up-regulated for SD in NIP. Thus, genes implicated in the five pathways should be prioritized for future functional validation.

Second, a total of 2778 DEGs enriched in 16 pathways were specifically involved in rice SD. Consistent with previous reports (Bewley et al. 1997), the phenotypic difference in SD between NIP and 9311 could primarily be attributed to the contrasting regulation of the ABA and GA/IAA/JA signaling pathways, reflected by the balance of ABA, GA, and IAA signaling serving as a core node that distinguishes the establishment of SD from the maintenance of SS. Obviously, the strong SD of NIP was primarily achieved by the ABA-regulated expression of LLRs, especially the LEA family, contribute to SS by stabilizing membranes and protecting macromolecules during storage. The role of GA in regulating SD was also supported by our transgenic experiments, in which the OsGA2ox8 knockout lines showed reduced SD, and the OsGA2ox8 overexpression lines showed significantly enhanced SD. GA, together with ABA, ET and IAA, constitutes the core hormonal regulatory network for rice seed dormancy. The antagonistic interaction between GA and ABA forms the core regulatory backbone: ABA promotes dormancy maintenance, while GA promotes germination, and ET and IAA exert synergistic effects by regulating these two pathways. Our results showed that the GA metabolic pathway was significantly differentially expressed between the strong dormancy variety NIP and the weak dormancy variety 9311. OsGA2ox8 was stably expressed in 9311, but significantly up-regulated in dormant NIP seeds. This gene inhibits the ubiquitination and degradation of DELLA protein (OsSLRL1) by degrading active GAs, and the accumulated DELLA protein further suppresses the activity of germination-related transcription factors such as OsPIF14, thereby maintaining seed dormancy. These results indicate that OsGA2ox8 is a key functional node regulating endogenous GA homeostasis in seeds. The promoter is a core region for gene transcriptional regulation, and variations in cis-elements directly determine gene expression and function. In this study, five key SNPs were identified in the promoter region of OsGA2ox8, which alter the conserved sequences of CAAT-box, TATA-box, 3-AF3 binding site and STRE elements, affect the binding efficiency of transcription factors, and thereby regulate gene transcription. Although these elements are not GA-specific responsive elements, they can indirectly affect OsGA2ox8 expression and alter the GA metabolic homeostasis in rice through basal transcription, light-GA crosstalk, and stress-GA crosstalk. Combined with the association results between haplotype differentiation and seed dormancy, we speculate that these functional SNPs affect seed GA/ABA homeostasis by regulating OsGA2ox8 expression. This study provides key clues and targets for dissecting the cis-regulatory mechanism of OsGA2ox8 and improving seed dormancy traits in rice. However, this study has certain limitations: the expression data of OsGA2ox8 in the investigated natural population have not been obtained, so the above functional predictions need to be further verified by experiments.

Third, genes and pathways related to rice SS were highly abundant. Substantial differences in SS between the two genotypes were mainly associated with distinct regulation patterns of secondary metabolite biosynthesis, nitrogen metabolism, and fatty acid-related pathways. Consistently, the metabolomic analysis further revealed corresponding differences in phenylpropanoid biosynthesis and fatty acid metabolism between 9311 and NIP. The former was reflected by the accumulation of lignin precursors in the 9311 seeds, which could enhance the seed coat barrier function, whereas the latter was shown by a rapid degradation of unsaturated fatty acids in the NIP seeds, leading to possible membrane system damage. This would imply that 9311 and NIP adopted divergent metabolic strategies to adapt to distinct ecological environments. Additionally, many genes involved in the ET and JA signaling pathways (e.g., OsEIL5, OsJAZ10) exhibited antagonistic expression patterns between 9311 and NIP, suggesting they were potentially participating in maintaining high SS by regulating downstream stress-responsive genes. For instance, sustained expression of the ET signaling genes in 9311 might contribute its enhanced tolerance to long storage. The stronger SS of 9311 was also partially attributable to ABA mediated upregulation of 15 putative LLR-OsLEAs, whose expression remained stable in NIP. These proteins play critical roles in seed dehydration and aging by protecting RNA stability and cell membrane integrity. Interestingly, the expression of these OsLEAs was negatively correlated with some members of the LOX gene family (such as OsLOX2 and OsLOX3) (Lei et al. 2024). LOX family genes may partially explain the poorer storability in NIP compared with 9311. Finally, our results suggest that targeted manipulation of the key regulatory alleles identified in this study may serve as potential strategies for simultaneously improving PHS resistance and SS in rice breeding.

Although the integrative analysis of transcriptome and metabolome suggests strong associations between specific pathways (e.g., hormone signaling and starch/sucrose metabolism) and the phenotypic differences in SD and SS, we note that, except for OsGA2ox8 and OsLEA5, the causal roles of most candidate genes and pathways remain to be confirmed by functional experiments. Most pathway-level conclusions in this study are based on correlative transcriptomic and metabolomic analyses. Their exact regulatory mechanisms and causal relationships with SD and SS await further genetic functional validation, which will be addressed in our future work targeting the key pathways and genes identified here.

Conclusion

The present study explored the molecular mechanisms underlying SD and SS in rice using 9311 and NIP with contrasting SD/SS phenotypes. Phenotypic analysis confirmed NIP had strong SD but poor SS, while 9311 showed the opposite. Transcriptomic and metabolomic integration revealed 1,334 common DEGs and 11 shared pathways (e.g., hormone signaling) co-regulating SD and SS, along with 309 overlapping LLRs. Functional validation demonstrated OsGA2ox8 is an important regulator of SD, and OsLEA5 controls SS. Our findings preliminarily revealed the differentiation mechanisms of SD and SS and supplement the understanding of their genetic correlation. Importantly, the key genes and pathways identified in this study may serve as potential targets for breeding rice varieties with PHS resistance and enhanced SS.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

SD

Seed dormancy

SS

Seed storability

G7d

Germination rate at day 7 after imbibition

S7d

Seedling establishment rate at day 7 after imbibition

DAM

Differentially accumulated metabolite

DEG

Differentially expressed gene

LLR

Long-lived mRNA

PHS

Pre-harvest sprouting

LEA

Late embryogenesis abundant

GA

Gibberellic acid

ABA

Abscisic acid

JA

Jasmonic acid

SA

Salicylic acid

WT

Wild type

DAP

Days after pollination

OPLS-DA

Orthogonal partial least squares-discriminant analysis

GI

Germination index

SI

Seedling establishment index

KO

Knockout

OE

Overexpression

FH

Freshly harvested

AR

After-ripened

qRT-PCR

Quantitative reverse transcription polymerase chain reaction

FPKM

Fragments per kilobase of transcript per million mapped fragments

KEGG

Kyoto Encyclopedia of Genes and Genomes

PCA

Principal component analysis

UHPLC

Ultra-high-performance liquid chromatography

QC

Quality control

TF

Transcription factor

SNP

Single nucleotide polymorphism

Hap

Haplotype

Author Contributions

ZC designed and conceived the research; He XY, Ye JW, Yu TT performed the investigation and wrote the original draft; Shi YS, Xu K, Huang YL performed the investigation and formal analysis; Zhang LT performed the investigation; Liu EB performed the investigation and formal analysis; Zhang L, Li ZK, Wang WS, Li M acquired the funding and performed the investigation; Wang WS, Li M performed the formal analysis and reviewed and edited the paper; Zhang CP conceptualized the study, acquired the funding, developed the methodology, and reviewed and edited the paper. All authors reviewed the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (32301783 and U21A20214), the College Students’ Innovative Entrepreneurial Training Plan Program (S202510364038), the Natural Science Foundation of Anhui Province (2308085QC91, 2408085MC058, and 2408085QC084), and the Natural Science Foundation of Universities of Anhui Province (2025AHGXZK30253 and 2025AHGXZK31561). This work was supported by the open Foundation of State Key Laboratory of Crop Gene Resources and Breeding (CGRB-2025-05).

Data Availability

The data sets supporting the results of this article are included within the article and its supporting files.

Declarations

Ethics Approval and Consent to Participate

Not applicable.

Consent for Publication

Not applicable.

Competing Interests

The authors declare no competing interests.