Genetic and transcriptome analyses of the effect of genotype-by-environment interactions on Brassica napus seed oil content

Abstract

The molecular basis underlying crop traits is complex, with gene-by-environment interactions (GEIs) affecting phenotypes. However, quantitative trait nucleotide (QTN)-by-environment interactions (QEIs) and GEIs for seed oil content (SOC) in oil crops are rare. Here, we detected 11 environmentally specific and 10 stable additive QTNs and 11 QEIs for SOC in rapeseed (Brassica napus) using genome-wide association studies. Weighted gene co-expression network analysis identified 8 Environmental-Developmental Gene co-expression Modules for which the eigengenes correlated with SOC and the environment explained a large proportion of the variance in gene expression. By incorporating information from the multi-omics dataset, 17 candidate genes and 11 candidate GEIs for SOC were predicted. We mined 1 GEI candidate, LIGHT-DEPENDENT SHORT HYPOCOTYLS5 (LSH5), around the environmentally specific QTN qspOC.A02.1 and QEI qeOC.A02.1 detected by climatic indices as covariates. BnaA02.LSH5 was highly expressed in early seed development, and its expression varied significantly across planting sites, with a trend opposite to light-related climatic indices. The BnaA02.lsh5 and BnaC02.lsh5 double mutants had lower SOC, hypocotyl length, photosynthesis, and carbon- and energy-related metabolites compared with wild type. Moreover, BnaA02.LSH5 transcriptionally directly repressed BnaA02.pMDH2 in fatty acid β-oxidation and photosynthetic electron transport. We propose that BnaLSH5 affects seed oil accumulation in response to light intensity. This study provides a basis for creating high-oil germplasm that is adapted to specific environments.

A systematic investigation of genotype-by-environment interactions for seed oil content in rapeseed revealed that BnaA02.LSH5 affects seed oil accumulation in response to light intensity.

Introduction

Agriculture and food security are being threatened by an increasing number of extreme weather events, including high temperature, drought, flood, and combined hot–wet events (Liu et al. 2023). However, plants have evolved multiple intrinsic mechanisms to maintain normal development in response to various climatic changes (Des Marais et al. 2013; Liu et al. 2021). Therefore, it is crucial to reduce climate-related yield risks by breeding cultivars with stable yields and quality.

Phenotypic plasticity refers to the phenotypic variation exhibited by the same genotype in different environments. The study of phenotypic plasticity helps breeders to make trade-offs between average phenotypes and plasticity, and helps to select phenotypes that are best suited to specific regions based on local climatic conditions (Allard and Bradshaw 1964; Wang et al. 2015; Kusmec et al. 2017). Differences in plasticity responses between genotypes have been described as genotype-by-environment interactions (G × E) (Pigliucci et al. 2006). Most of the agronomic traits of crops are controlled nonadditively by both genetic and environmental factors and their interactions (G × E).

Seed oil content (SOC) is a complex trait influenced by a combination of genetic and environmental factors, as seeds grow under varying climatic and abiotic stress. Light can influence seed oil accumulation by altering the carbon supply through photosynthesis (Yang et al. 2022). As the green seeds of most members of Brassicaceae family are photosynthetically active, light passing through the silique wall is a key driver for oil biosynthesis (Eastmond et al. 1996; Ruuska et al. 2004; Schwender et al. 2004; Hua et al. 2012). Drought generally limits seed filling and promotes embryo development, thereby reducing SOC (Singer et al. 2016). During the flowering stage, adequate rainfall extends the flowering period and favors the formation and accumulation of oil in the seeds (Si and Walton 2004). Temperature affects SOC and fatty acid (FA) composition in plants. In some cultivars, SOC tends to decrease significantly with increasing temperature during seed filling (Yaniv et al. 1995; Werteker et al. 2010), whereas SOC tends to be higher at cooler growing temperatures (Schulte et al. 2013). Although it has been recognized that various environmental cues affect seed oil biosynthesis, little is known about how plant sense, transmit, and respond to environmental signals in the regulatory network for oil biosynthesis in oilseeds.

Rapeseed (Brassica napus, AACC, 2n = 38) is an economically important oil crop grown in a wide range of regions. Stabilizing rapeseed oil production in the face of global climate change is a challenge for rapeseed industry (Hickey et al. 2019). Currently, several genes and TFs that regulate SOC have been cloned and validated in rapeseed, e.g. FATTY ACID DEHYDROGENASE2 and FATTY ACID ELONGASE1 (Shi et al. 2017), DIACYLGLYCEROL ACYLTRANSFERASE1 (DGAT1) (Taylor et al. 2009), BILE ACID: SODIUM SYMPORTER FAMILY PROTEIN2 (Tang et al. 2022a) and PHOSPHOENOLPYRUVATE/PHOSPHATE TRANSLOCATOR1 (Tang et al. 2022b), WRINKLED1 (Wu et al. 2014), LEAFY COTYLEDON1 (Elahi et al. 2016), FUSCA3 (Elahi et al. 2015), and SCARECROW-LIKE 31 (Tan et al. 2022a). With the quantitative genetics and omics approaches, an increasing number of quantitative trait loci (QTLs) and quantitative trait nucleotides (QTNs) related to SOC in rapeseed have been identified by QTL mapping (Jiang et al. 2014; Fu et al. 2016; Sun et al. 2016) and genome-wide association studies (GWASs) (Liu et al. 2016; Tang et al. 2021; Wang et al. 2021). Recently, Zhao et al. (2022) identified the QTL-by-environment interactions (QEIs) for SOC in multiple environments by QTL mapping. To date, some gene-by-environment interactions (GEIs) have been reported to affect SOC in response to the environment. BnWRI1 is reported to coordinate FA synthesis and photosynthesis pathways during seed development (Wu et al. 2014). DGAT1 is reported to be up-regulated under osmotic stress in Arabidopsis (Arabidopsis thaliana) (Lu and Hills 2002), while co-expression of CsDGAT1C and CsMYB96A shows enhanced drought stress resistance and SOC in Camelina (Camelina sativa) (Kim et al. 2019). Nevertheless, these QEIs and GEIs for SOC are still far from sufficiently investigated in rapeseed.

To explore the G × E underlying SOC, we integrate climatic, genetic, and transcriptomic analysis in rapeseed. Environmentally stable and specific QTNs, QEIs with climatic indices as covariates are detected. Environmental-Developmental Gene co-expression Modules (EDGMs) are identified by weighted gene co-expression network analyses (WGCNAs) of the transcriptomes of 9 inbred lines with high SOC plasticity at 2 seed developmental stages in 3 planting sites. Based on the EDGMs and a gene prioritization framework, the candidate genes and GEIs for SOC were predicted. Moreover, LIGHT-DEPENDENT SHORT HYPOCOTYLS5 (LSH5) is found to be the candidate gene for the environmentally specific QTN qspOC.A02.1 and QEI qeOC.A02.1 detected by light-related indices as covariates. We further investigated the functional and molecular mechanism of BnaLSH5 on seed oil accumulation and light responses using rapeseed CRISPR mutants and verified the transcriptional repression activity of BnaA02.LSH5 on the expression level of peroxisomal MALATE DEHYDROGENASE2 (pMDH2). This study will further elucidate the mechanisms of oil biosynthesis in response to environmental cues and help breed superior rapeseed cultivars with stable oil yield.

Results

Genetic analysis of environmentally stable and specific QTNs for SOC in 8 field trials

SOC variation was measured in a diversity group of 505 rapeseed inbred lines in 3 planting sites including 8 field trials (Wuhan [WH], Hefei [HF], and Chengdu [CD]; Fig. 1A). The mean population SOC in the 8 field trials range from 39.44% to 46.62%, with WH being higher than HF and CD (Supplementary Table S1; Fig. 1, B and C). R² values between different field trials range from 0.07 to 0.29, while R² between the different planting sites by using the Best Linear Unbiased Prediction (BLUP) range from 0.20 to 0.40 (Supplementary Table S1; Fig. 1, B and C). According to the principal component analysis (PCA) of SOC and the correlation of each pair of trials, the trials within WH are more closely correlated than those within HF and CD (Supplementary Fig. S1, A and B). The heritability for SOC ranges from 0.82 to 0.97. The coefficients of variation, skewness, and kurtosis of the trait in 8 trials illustrate that SOC is a typical quantitative trait (Supplementary Table S1).

Figure 1.

Genome-wide association studies for seed oil content at 3 planting sites of a diversity panel of 505 rapeseed inbred lines. A) Geographical sites of 8 field trials. WH, Wuhan. HF, Hefei. CD, Chengdu. B, C) The BLUP values of SOC at 3 planting sites (n = 505 inbred lines). The center line, box limits, and whiskers represent the median, upper and lower quartiles, and 1.5 times the interquartile range, respectively. P-value indicates the significant differences by two-tailed unpaired Student's t test. D, E) The Manhattan plot and the known genes for SOC-related environmentally specific and stable QTNs by using single-environment module in IIIVmrMLM software. The environmentally specific QTNs were identified in 2 environments of 1 site, while environmentally stable were identified in at least 3 out of 8 environments and in 2 out of 3 sites. Significant QTNs are above the black dash line at the critical P-value of 4.27e-09 (=0.05/m, where m is the number of markers), while suggested QTNs are above the gray dash line with the LOD score ≥3.0 but the P-value ≥0.05/m, where m is the number of markers.

3VmrMLM is a newly established GWAS method that considers all possible effects and controls for all possible polygenic backgrounds and population structure (k = 4). In this study, its single-environment and multi-environment modules were used to identify QTNs (Li et al. 2022b). In the single-environment module, we associated 11,700,689 SNPs with the phenotypes for SOC in WH2016, WH2017, WH2018, WH2019, HF2017, HF2018, CD2017, and CD2020. As a result, 46, 37, 50, 38, 36, 46, 47, and 41 significant QTNs, and 10, 11, 9, 13, 9, 10, 8, and 9 suggested QTNs were identified to be related to SOC in the above 8 trials, respectively (Supplementary Table S2; Supplementary Fig. S2). Of these QTNs, 11 were identified in at least 3 of the 8 trials and in 2 of the 3 sites, which are considered to be environmentally stable QTNs, while 9 were identified in 2 trials of 1 site, which are considered to be environmentally specific QTNs (Fig. 1, D and E; Supplementary Table S3; Supplementary Fig. S3), including 10 QTNs identified in previous studies (Supplementary Table S3). We further mined 8 known genes between 100 kb upstream and downstream regions of these QTNs according to the LD decay (48.4 kb) described in Tang et al. (2021), where BnaLEC1 (Tan et al. 2011), NUCLEOTIDE TRIPHOSPHATE TRANSPORTER1 (NTT1) (Hong et al. 2022), and PROBABLE METHYLTRANSFERASE6 (PMT6) (Tang et al. 2021) were validated in rapeseed, and LIPOAMIDE DEHYDROGENASE2 (LPD2) (Drea et al. 2001), VIVIPAROUS1/ABI3-LIKE1 (VAL1) (Suzuki et al. 2007), MFP2 (Arent et al. 2010), and MDH (Kong et al. 2018) were validated in Arabidopsis. In addition, DIOXYGENASE FOR AUXIN OXIDATION1 (DAO1) was predicted to be associated with SOC in Tang et al. (2021).

The multi-environment module was used to analyze the Datasets I, II, and III, respectively (see Methods). In total, 30, 40, and 54 significant QTNs and 16, 15, and 11 suggested QTNs were identified to be related to SOC in the above 3 datasets, respectively (Supplementary Table S4; Supplementary Fig. S4). All the environmentally stable QTNs are commonly identified among these QTNs, indicating the increased power in QTN detection (Supplementary Table S3; Fig. 1E). The LOD scores of all the QTNs ranged from 3.38 to 65.98, and the corresponding average R², representing the percentage of the total phenotypic variance explained by each QTN/QEI, ranged from 0.06% to 3.41%.

Dynamic expression profiles of SOC-related co-expression modules in 3 planting sites

We first captured the SOC variation of 505 genotypes across the 8 trials through the Finlay–Wilkinson regression model, in which the slope parameter represents the plasticity of each genotype to the different environments. As a result, 9 semi-winter ecotype inbred lines with high SOC plasticity and relatively consistent first flowering time were selected (Fig. 2A; Supplementary Table S5). To study the expression patterns of SOC-related genes in different field trials, we planted the 9 inbred lines in WH2022, HF2022, and CD2022, and collected 54 transcriptomes from these 3 sites at 2 developmental stages. Total FA content in developing seed was measured and used to calculate SOC. SOC is significantly different between the 3 sites at 20 and 40 days after flowering (DAF) of developing seed (Fig. 2B). Samples from the same sites are consistently clustered together based on PCA of the transcriptomes at 20 DAF and 40 DAF, respectively (Supplementary Fig. S5A).

Figure 2.

Dynamic expression profiles of SOC-related EDGMs in 3 planting sites. A) Nine inbred lines with high phenotype plasticity (n = 8 planting sites) were planted at WH2023, HF2023, and CD2023. B) Significant difference in SOC at 20 and 40 DAF of developing seed among 9 inbred lines at the 3 planting sites with 3 biological replicates (n = 27). DW, dry weight. C) WGCNA of SOC in developing seed at 20 and 40 DAF. The blue and turquoise module eigengenes were found to be related to SOC at 20 DAF, and the blue, turquoise, green, and yellow module eigengenes were found to be related to SOC at 40 DAF. 8 EDGMs were constructed based on the SOC correlated modules at 20 and 40 DAF. D) Four EDGMs were found to be correlated with SOC in turquoise module at 20 DAF. Hub genes were identified based on the scores calculated by network topological analysis using cytoHubba. E) GO enrichment analysis of turquoise20-blue40 EDGM showed that this EDGM significant enriched for fatty acid synthesis. F) Variance component analysis of genes in 4 EDGMs of turquoise module at 20 DAF. The closer the point is to a vertex, the higher the variance represented by that vertex is explained in that gene. Through any point in the triangle, the intercepted line segments of parallel lines to each of the 3 sides are equal to the variance proportions of the genotypes, environments, and G × E, respectively.

WGCNA was then performed by using 55,316 genes at 20 and 40 DAF, respectively (Supplementary Fig. S5B). We found that the turquoise and blue module eigengenes in WGCNA at 20 DAF are significantly correlated with the SOC in the consistent trend across 9 inbred lines, while the turquoise, blue, yellow, and green module eigengenes in WGCNA at 40 DAF are significantly correlated with the SOC (Supplementary Fig. S6). To obtain the dynamic gene co-expression patterns at 20 and 40 DAF, we defined the Environmental-Developmental Gene co-expression Module (EDGM) as the sub-module at the intersection of 2 modules of different developmental stages (20 and 40 DAF). The eigengenes of the 8 EDGMs were found to be associated with SOC at both stages. For example, the turquoise20-blue40 EDGM contains 1,655 genes co-expressed in the turquoise module at 20 DAF and in the blue module at 40 DAF (Fig. 2C; Supplementary Fig. S7). Interestingly, we found that the turquoise20-blue40 EDGM is significantly enriched for FA biosynthetic process (GO:0006633) (Fig. 2, D and E), whereas turquoise20-turquoise40 is significantly enriched for stress-related processes, e.g. response to water deprivation (GO:0009414) (Supplementary Table S6).

Through the integration of network topological analysis, the hub genes in each EDGM were identified by maximal clique centrality (MCC) scores (Supplementary Table S7; Chin et al. 2014). In addition, considering each gene expression value as phenotypic value, the phenotypic variance of 3 sites was partitioned into genotype, environment, and G × E by the variance component analysis (VCA). The environment variance was found to be indicative of hub genes, explaining a large proportion of the variance in gene expression (Fig. 2F, Supplementary Table S7). We found that the relationship between the environment variance component and the MCC scores in 8 EDGMs demonstrates a linear correlation (Supplementary Fig. S8). And we found that the mean of the environment variances of the hub genes in the 6 EDGMs is greater than the mean value of the random 10 genes within the 99% confidence interval based on 1,000 permutations (Supplementary Fig. S9). These results suggest that the more crucial genes in the network tend to account for a substantial proportion of the environment variance.

Detection of SOC-related QEIs in response to climatic indices after flowering

To uncover the environmental variation of SOC across different trials, the multi-environment module of 3VmrMLM was first used to analyze the Datasets I, II, and III, respectively. As a result, 16, 20, and 37 significant QEIs and 12, 11, and 10 suggested QEIs were identified as related to SOC in the 3 datasets above (Supplementary Fig. S4; Supplementary Table S4). Of these QEIs, 11 were commonly identified in at least 2 of the 3 datasets (Table 1; Fig. 3A; Supplementary Fig. S10).

Table 1.

Candidate gene-by-environment interactions for SOC in rapeseed

Genome-wide association studies					Candidate genes
QEI name	Chr	Method and markera	LOD	R 2	Gene ID	Symbol	Description	EDGM	Response to environment stress	Reference
qeOC.A02.1	A02	E1-BnvaA0205977571, E3-BnvaA0205686129, E2-BnvaA0205686129	3.73 to 8.10	0.16 to 0.58	BnaA02G0110000ZS	BnaA02.LSH5	Protein Light-dependent short hypocotyls 5	tyrquoise20-blue40	Light	(Zhao et al. 2004)
qeOC.A04.1	A04	E3-BnvaA0417128210, E2-BnvaA0417128210	24.87 to 27.31	0.77 to 1.12	BnaA04G0161300ZS	BnaA04.STYK	Serine/threonine/tyrosine protein kinase	turquoise20-yellow40	Drought	(Sun et al. 2022)
qeOC.A06.1	A06	E1-BnvaA0637095156, E3-BnvaA0636870929, E2-BnvaA0636797920	20.83 to 28.53	0.78 to 1.62	BnaA06G0266300ZS	BnaA06.SWEET11	Bidirectional sugar transporter SWEET11		Salt	(Wang et al. 2020, 2023a)
qeOC.A09.1	A09	E3-BnvaA0901910956, E3-BnvaA0901981940	8.33 to 20.41	0.2261 to 0.56	BnaA09G0026500ZS	BnaA09.MPH2	Thylakoid lumenal 16.5 kDa protein	turquoise20-blue40	Light	(Liu and Last 2017)
qeOC.A09.2	A09	E3-BnvaA0958024853, E2-BnvaA0957894842	6.25 to 6.64	0.17 to 0.29	BnaA09G0572900ZS	BnaA09.VPE	Vacuolar-processing enzyme alpha-isozyme	tyrquoise20-yellow40	Drought	(Kinoshita et al. 1999)
qeOC.A09.3	A09	E1-BnvaA0961460454, E2-BnvaA0961214867	6.12 to 7.67	0.27 to 0.61	BnaA09G0636200ZS	BnaA09.DAO1	2-oxoglutarate (2OG) and Fe (II)-dependent oxygenase superfamily protein		Drought and salt	(Tang et al. 2021; Chauhan et al. 2023)
qeOC.C02.1	C02	E3-BnvaC0204347682, E2-BnvaC0204321368	16.54 to 23.57	0.46 to 1.06	BnaC02G0069400ZS	BnaC02.ATPEP2	Transducin/WD40 repeat-like superfamily protein		Drought	(Xu et al. 2019)
qeOC.C05.1	C05	E3-BnvaC0540919786, E2-BnvaC0540919786	14.61 to 28.11	0.65 to 0.80	BnaC05G0364500ZS	BnaC05.SDG36	Histone-lysine N-methyltransferase	blue20-green40
qeOC.C06.1	C06	E3-BnvaC0636504698, E2-BnvaC0636521436	8.49 to 12.04	0.33 to 0.37	BnaC06G0257400ZS	BnaC06.RAP2	Ethylene-responsive transcription factor RAP2	turquoise20-blue40	Drought and light	(Yang et al. 2020; Iwase et al. 2021)
qeOC.C09.1	C09	E3-BnvaC0920997369, E2-BnvaC0920997369	13.78 to 15.02	0.42 to 0.61	BnaC09G0229700ZS	BnaC09.E1-OGDH	The E1 subunit of the 2-oxoglutarate dehydrogenase	blue20-yellow40	Light	(Condori-Apfata et al. 2019; Eprinstev et al. 2020)
qeOC.C09.2	C09	E1-BnvaC0949804875, E3-BnvaC0949631389, E3-BnvaC0949743995, E2-BnvaC0949661132	7.02 to 77.08	0.31 to 3.96	BnaC09G0390600ZS	BnaC09.SFH14	Phosphatidylinositol/phosphatidylcholine transfer protein SFH14	turquoise20-turquoise40

^aThree datasets for joint environments GWAS. E1, Dataset I, 3 planting sites in 2017 (WH, HF, and CD). E2, Dataset II, the Best Linear Unbiased Prediction (BLUP) values from the 3 planting sites. E3, Dataset III, the 8 field trials.

Figure 3.

Detection of QEIs and GEIs for SOC by multi-environment joint genome-wide association studies. A) The Manhattan plot for SOC-related QEIs and GEIs identified in 3 datasets by using multi-environments module in IIIVmrMLM software. Known GEIs are colored in red. B) Principal components analysis of climatic indices after flowering in 8 field trials. PC1 and PC2 accounted for 39.4% of the overall variance. C) Variance explained by first 40 climatic indices loadings on PC1 and PC2. D) Trends in 3 climate factors with a sliding window of 7 d in 8 field trials and schematic illustration of seed maturation. Rectangles represent 3 d of first 40 climatic indices loadings on PC1 and PC2. E) The Manhattan plot for SOC-related QEIs and GEIs identified with the 3 climatic factors as covariate in Dataset III by using multi-environment module in IIIVmrMLM software. Significant QTNs are above the black dash line at the critical P-value of 4.27e-09 (=0.05/m, where m is the number of markers), while suggested QTNs are above the gray dash line with the LOD score ≥3.0 but the P-value ≥0.05/m, where m is the number of markers.

In order to provide further characterization of the effect of the climatic factors on SOC, 22 climatic factors were obtained after flowering in 8 trials to explore the environmental determinants for SOC during seed development (Supplementary Fig. S11; Supplementary Table S8). Significant differences were found in ultraviolet-B (UVB) and clear-sky photosynthetically active radiation (CPAR) between CD and WH and between CD and HF (Supplementary Fig. S12). In addition, we performed PCA using 22 climate factors ranging from 0 to 70 DAF in 8 trials. As a result, PC1 and PC2 explained 39.6% of the overall variance and divided the 8 trials into 3 clusters, which is consistent with those from PCA of SOC (Supplementary Fig. S1B). Based on the largest variance (either positive or negative) explained by the loadings of PC1 and PC2, the diurnal temperature range at 23 DAF (MMR₂₃) and photothermal time at 14 DAF (dPTT₁₄) affects the clustering of HF2018 with HF2017 (Fig. 3, B and C; Supplementary Table S9). Subsequently, to detect SOC-related QEIs in response to climate factors, we selected the top 10 features as covariates for joint environments GWAS using the multi-environment module of 3VmrMLM (Supplementary Tables S9 and S10). The analysis reveals 7 QEIs to be associated with SOC by considering precipitation at 19 DAF (PR₁₉), diurnal temperature range at 29 DAF (DTR₂₉), and UVB at 15 DAF (UVB₁₅) as covariates, respectively. Notably, 5 of these QEIs are commonly identified in the aforementioned 14 QEIs and environmentally stable QTNs (Fig. 3, D and E; Supplementary Table S10), demonstrating the consistency of the findings.

Prediction of BnaA02.LSH5 as a candidate GEI in qspOC.A02.1 and QEI qeOC.A02.1

To mine candidate genes for QTNs and GEIs, POCKET was utilized for prioritizing the candidate genes by incorporating information of knowledge-based gene sets, effects of variants, GWAS, and transcriptome-wide association studies (TWASs); Tang et al. 2021) to determine candidate oil metabolism genes and used the EDGMs to further filter the candidate environmental interaction genes (Supplementary Table S11). In addition, we conducted a validation of tissue expression patterns for these GEIs, which resulted in identification of 4 genes (BnaA02.LSH5, BnaA04.STYK, BnaA09.VPE, and BnaA09.DAO1) that exhibited high seed expression and can be considered reliable GEIs (Supplementary Fig. S13). Consequently, 3 and 5 candidate genes were identified around the environmentally specific and stable QTNs, respectively (Supplementary Table S4), and 11 GEIs were identified around the QEIs (Table 1).

qspOC.A02.1 is specifically detected in CD planting sites and is commonly identified in qeOC.A02.1, which is associated with SOC by PR₁₉, UVB₁₅, and DTR₂₉. The initial investigation was directed towards the qspOC.A02.1 region on chromosome A02, and it was determined that BnaA02.LSH5 was among the top candidate genes that scored near these QTNs and in the turquoise20-blue40 EDGM (Supplementary Fig. S14). To dissect the natural variation of BnaA02.LSH5, 2 haplotypes of BnaA02.LSH5 were constructed from variants in the gene region and the 1-kb upstream region, including 6 variants in the upstream region and 1 synonymous variant (Fig. 4A). In the diversity panel of 505 rapeseed inbred lines, the sub-population with hap. A has higher nucleotide diversity (π) and SOC compared to the hap.B sub-population, in which the inbred lines are mainly spring ecotypes (Fig. 4, A and B). The SOC of CD2017 and the BLUPs of CD have the largest significant difference between hap.A and hap.B. (Figure 4B). The π around BnaA02.LSH5 is lower in the sub-population of the spring ecotype than in the semi-winter ecotype (Fig. 4C; Supplementary Fig. S15), suggesting selective pressure in the spring ecotype. Furthermore, this region has been identified to be located in the selective sweeps between spring and semi-winter ecotype in Han et al. (2022a), and in the selective sweeps between spring and winter ecotype, as well as in the selective sweeps of different breeding eras in China in Hu et al. (2022).

Figure 4.

Natural variation of BnaA02.LSH5 in populations and its expression in multiple tissues and environments. A) Two major haplotypes for BnaA02.LSH5 and nucleotide diversity (π) of each haplotype in the diversity population (n_hap.A = 276, n_hap.B = 73). S, Synonymous single nucleotide variant, U, upstream gene single nucleotide variant. B) Haplotype analysis of BnaA02.LSH5 for SOC at 8 trials and the BLUP values at 3 planting sites (n_hap.A = 276, n_hap.B = 73). C) Nucleotide diversity around BnaA02.LSH5 in spring and semi-winter ecotype rapeseeds in a diversity panel of 505 inbred lines (n_spring =562, n_semi-winter = 585). D) Significance at population expression level between haplotypes of BnaA02.LSH5 at 20 and 40 DAF at WH2016 (n = 48 to 136). E) Expression patterns of BnaA02.LSH5 homologous in multiple tissues and stages. Color scale bar, TPM. F) Significant difference at expression level for BnaA02.LSH5 at 20 and 40 DAF among 3 planting sites (n = 27). Normalized scores of 4 light-related climatic factors on the day of sampling are shown at the top of the figure. G) Correlation between BnaA02.LSH5 expression and SOC in 9 inbred lines with high plasticity in WH, HF, and CD at 20 and 40 DAF. DW, dry weight. Values are mean ± SD (two-tailed unpaired Student's t-test, *, **, and *** indicated significant probability levels at 0.05, 0.01, and 0.001, respectively). The center line, box limits, and whiskers represent the median, upper and lower quartiles, and 1.5 times the interquartile range, respectively.

On the other hand, an examination of the expression profiles of BnaA02.LSH5 was conducted at the population, tissue, and environment levels. First, it was observed that the sub-population expression level of hap.A was found to be significantly higher than hap.B (Figure 4D). Secondly, there are 6 LSH5 homologs in rapeseed, whereas only the BnaA02.LSH5 and BnaC02.LSH5 are highly expressed in the rosette, stem, silique, and seed, in particular, they were most expressed at the early developmental stage of the seed (14 to 30 DAF) (Fig. 4E). Based on the above 54 transcriptomes from 3 planting sites, BnaA02.LSH5 showed a significantly higher expression level in CD than in WH and HF at 20 DAF, but significantly lower expression levels in 40 DAF (Fig. 4F). The 3 light climate factors in CD are lower than in WH and HF on the day of sampling at 20 DAF and the 3 light climate factors in WH are lower than in HF and CD on the day of sampling at 40 DAF, suggesting the opposite trend between BnaA02.LSH5 expression and light climatic indices (Fig. 4F; Supplementary Fig. S16). Interestingly, correlation analysis of SOC and gene expression shows that BnaA02.LSH5 has a positive effect on the SOC at 20 DAF (r = 0.34, P = 1.99 × 10⁻³) and a negative effect on the SOC at 40 DAF (r = −0.72, P = 2.97 × 10⁻¹⁴; Fig. 4G). In addition, we respectively conducted Finlay–Wilkinson regression model in SOC and BnaA02.LSH5 across WH2022, HF2022, and CD2022 at 20 and 40 DAF (Supplementary Fig. S17A). These findings revealed a significant correlation between the plasticity of SOC and the plasticity of BnaA02.LSH5 expression at 40 DAF, but not at 20 DAF (Supplementary Fig. S17B). These results suggest that BnaA02.LSH5 may be the causal gene for plasticity in SOC around qspOC.A02.1 and QEI qeOC.A02.1 and this region may be subject to selection during rapeseed breeding.

Genome editing of BnaLSH5 in rapeseed affects SOC accumulation

LSH5 encodes a member of the ALOG (Arabidopsis LIGHT-SENSITIVE HYPOCOTYL 1, Oryza G1) protein family, which is conserved in plants (MacAlister et al. 2012; Vo Phan et al. 2023). In this research, we cloned BnaA02.LSH5 and BnaC02.LSH5 and determined their sequences. BnaA02.LSH5 fused with GFP at its C-terminus is transiently expressed in the epidermal cells of Nicotiana benthamiana. The LSH5–GFP is found to encapsulate the auto-fluorescence of the nucleus, suggesting that BnaA02.LSH5 is localized in the nucleus (Fig. 5A). To understand the function of BnaLSH5, we designed 2 sgRNAs targeting BnaA02.LSH5 and BnaC02.LSH5 at the same time and cloned them into a clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR-associated nuclease 9 (Cas9) vector to knock out BnaLSH5 s in the A and C sub-genomes (Supplementary Fig. S18A). As a result, 3 of the resulting CRISPR/Cas9-edited lines, designated CR34, CR36, and CR41, were identified to be homozygous double mutant by sequencing (Supplementary Fig. S18B). Moreover, BnaA02.LSH5 was overexpressed in rapeseed driven by the 35S promoter. Reverse transcription quantitative PCR (RT-qPCR) revealed that 5 overexpression lines (OE) exhibited higher expression levels of BnaA02.LSH5 compared to the wildtype (WT) (Supplementary Fig. S19), and OE1 and OE2 were selected for further investigation.

Figure 5.

Functional studies on BnaLSH5 by generating knockout mutants. A) Subcellular localization of BnaA02.LSH5 protein fused with GFP (35S:LSH5–GFP) in tobacco epidermal cells showing nuclear localization under confocal. Bars = 25 μm. B) Quantitative comparisons of seed oil content in WT, OE1, OE2, CR34, CR36, and CR41 mutant plants by near-infrared reflectance spectroscope in 2 field trials. Values are presented as means ± SD (n = 6 to 10), and each of 3 technological replicates per biological replicate contained from mature open-pollinated seeds of individual plant grown in different pots arranged randomly within 1 of 3 blocks. Asterisks indicate significant differences in seed oil content in the OE1, OE2, CR34, CR36, and CR41 mutant plants compared with those in the WT plants (two-tailed unpaired Student's t-test, *, **, and *** indicated significant probability levels at 0.05, 0.01, and 0.001, respectively). C) Volcano plots of DEGs between mutants of LSH5 and WT at 20 and 40 DAF. Grey lines are at |log₂(FoldChange)| = 1 or adjusted P-value = 0.05. D) GO enrichment of DEGs between mutants of LSH5 and WT at 20 and 40 DAF, respectively.

We first measured the SOC of WT and transgenic lines by near-infrared spectrometer (NIRS) in 2-year field replicates (Fig. 5B). In WH2022, SOC exhibited a significantly increase from an average of 39.79% in WT to 41.21% and 41.97% in OE1 and OE2, respectively, and a significantly decrease to 38.84%, 38.02%, and 35.04% in CR34, CR36, and CR41. In WH2023, SOC increased significantly from an average of 42.49% in WT to 43.36% and 44.01% in OE1 and OE2, and significantly decreased to 37.89%, 41.95%, and 41.58% in CR34, CR36, and CR41. Subsequently, an evaluation was conducted of the impact of BnaLSH5 on agronomic traits, including plant height, inflorescence length, number of effective branches, pod number, and 1,000 seed weight. We found that there were no obvious differences between WT and mutant lines for most of the agronomic traits (Supplementary Table S12). In particular, the mutants displayed slower growth and late flowering (Supplementary Fig. S20), which corresponded to the haplotype analysis of BnaA02.LSH5 in flowering time (Supplementary Fig. S21).

Besides, we further obtained the transcriptomes of WT and CR34 lines in developing seeds at 20 and 40 DAF (Supplementary Fig. S22). Differential expression gene (DEG) analyses were performed between mutant lines and WT at 20 and 40 DAF (Fig. 5C). Gene ontology (GO) enrichment analysis of differential expression genes (DEGs) involved in seed maturation, stress-related pathways, and lipid metabolism. Especially, DEGs between CR34 and WT are involved in response to high light intensity (GO:0009644) and response to abscisic acid (GO:0009737) at both 20 and 40 DAF, and involved in FA metabolic process (GO:0006631), ATPase activity (GO:0016887), and FA beta-oxidation (GO:0006635) at 20 DAF (Fig. 5D; Supplementary Table S13). To explore the metabolism changes of BnaA02.LSH5 in regulation of oil accumulation, we extracted primary metabolites from 20 DAF seeds and determined by liquid chromatography with the tandem mass spectrometry (LC–MS/MS) method (Supplementary Figs. S23 and S24). A brief model of carbon influx from the carbon supply into FA synthesis in lsh5 mutant seeds is shown in Supplementary Fig. S24B. The results demonstrate that significant differences occur between WT and LSH5 mutant lines with respect to carbon fixation, pentose phosphate pathway, glycolysis, and FA synthesis (Supplementary Fig. S23). The contents of the metabolites of glycerate, ribose 5-phosphate (R5P), D-xylulose 5-phosphate (XU5P), D-ribulose 5-phosphate (RU5P) were found to be significantly increased in mutants than that in WT. The contents of D-ribulose 1,5-bisphosphate (RUBP), glucose-6-phosphate (G6P), fructose-6-phosphate (F6P), 3-phosphoglycerate (3-PGA), D-glucose 1-phosphate (G1P), D-erythrose 4-phosphate (E4P), phosphoenolpyruvate (PEP), acetyl-CoA, and pyruvate related to glycolysis were found to be significantly decreased in mutants than that in WT. In chloroplasts, light energy is used for the synthesis of ATP and NADPH, which are necessary for FA synthesis. Interestingly, the contents of ATP are significantly decreased in mutants compared with ADP, and NADPH are significantly decreased in mutants compared with NADP (Supplementary Fig. S24). Thus, the above results suggest that BnaLSH5 may affect oil accumulation through ATP and NADPH to carbon and energy supply during seed development.

BnaLSH5 directly suppresses the expression of BnaA02.pMDH2

To further dissect the regulatory mechanisms of BnaLSH5 in oil accumulation, we performed a cleavage under targets and tagmentation (CUT&Tag) assay. We isolated protoplasts extracted from rapeseed leaf tissues to identify the potential targets of BnaA02.LSH5 and BnaC02.LSH5 at the genome-wide level. We obtained 730 and 1,059 target genes around the confidential peaks from 2 biological replicates of BnaA02.LSH5 and BnaC02.LSH5, respectively (Supplementary Table S14), with which 665 genes were commonly founded (Fig. 6A). GO analysis of the 665 genes revealed that the most significantly enriched term is transcription factor activity (GO:0003700), and other significantly enriched terms are genes involved in response to light stimulus (GO:0009416), photosynthetic electron transport in photosystem II (GO:0009772), electron carrier activity (GO:0009055), consistent with the GO enrichment of DEGs for lsh5 mutants. (Figure 6B; Supplementary Table S15). Intriguingly, we found BnaA02.pMDH2 (BnaA02G0032600ZS) was potential targets of BnaLSH5 revealed by CUT&Tag (Fig. 6C). Transient dual-luciferase assay shows that LUC expression levels driven by the promoter of BnaA02.pMDH2, is significantly down-regulated when the BnaA02.LSH5 effector and the reporter construct were co-transformed in Arabidopsis protoplasts (Fig. 6, D and E). By performing motif enrichment in all the peaks using meme-chip, we identified 94 significantly enriched motifs, including the “AATAGTR” and “YTYTYTYTYTYTYTY” in 3 promoter regions of BnaA02.pMDH2. (Figure 6, F and G) We further demonstrated using CUT&Tag-qPCR and electrophoretic mobility shift assay (EMSA) assay that BnaLSH5 can biochemically directly binds to the 3 promoter regions of BnaA02.pMDH2 (Fig. 6, H and I). Therefore, we concluded that BnaLSH5 serves as the negative regulator of these salt-induced TF genes. Therefore, these data support that BnaLSH5 targets the BnaA02.pMDH2 promoter to suppress BnaA02.pMDH2 expression in rapeseed.

Figure 6.

BnaLSH5 directly suppresses the expression of BnapMDH2. A) Venn diagram showing the numbers and overlap of target genes targeted by BnaA02.LSH5 and BnaC02.LSH5 in 2 biological replicates (rep-1 and rep-2). B) GO enrichment terms of 665 genes targeted by BnaLSH5. C) BnaLSH5 binding profiles and expression levels around BnaA02.pMDH2. Peaks are normalized by counts per million (CPM). Blue and red peaks are the binding profiles of BnaLSH5 on the promoter region of BnaA02.pMDH2. Purple and turquoise peaks are the expression levels of BnaA02.pMDH2. Orange rectangles indicate the 3 promoter regions for CUT&Tag-qPCR in H). D) Schematic diagram of vectors used in the transient dual-luciferase assay. The effector construct includes the BnaA02.LSH5 coding sequence driven by the 35S promoter. The reporter construct contains the REN gene driven by the 35S promoter and the LUC reporter gene driven by the BnaA02.pMDH2 promoter. E) LUC assays in rapeseed protoplasts demonstrate that BnaA02.LSH5 suppress the transcription of BnaA02.pMDH2. F, G) The “AATAGTR” and “YTYTYTYTYTYTYTY” motif are significantly enriched around the peaks targeted by BnaA02.LSH5 and BnaC02.LSH5 in 3 regions of the BnaA02.pMDH2 promoter. H) CUT&Tag-qPCR show in vivo binding of the 3 promoter regions of BnaA02.pMDH2 by BnaA02.LSH5. I) EMSAs showing that GST-BnaA02.LSH5 fusion protein directly binds to motifs within the promoters of BnaA02.pMDH2. In E) and H), values are mean ± SD (n = 3 independent experiments, two-tailed unpaired Student's t-test, *, **, and *** indicated significant probability levels at 0.05, 0.01, and 0.001, respectively).

Low light intensity induces the expression of BnaLSH5 which may confer SOC plasticity

Although transgenic lines of BnaLSH5 affect seed oil accumulation, it remains questionable how BnaLSH5 confers SOC plasticity beyond the direct targeting of BnaA02.pMDH2. Previous studies have reported the dominant mutant of AtLSH1 shows hypersensitive light responses in seedling development (Zhao et al. 2004). In the field trials, we found that the hypocotyls length of LSH5 mutants was significantly shorter than that of WT, suggesting a light-sensitive response for LSH5 in rapeseed (Fig. 7A). Hypocotyls and root lengths of the mutants were further examined under normal light, shade, and dark conditions in the greenhouse during seeding. The difference in hypocotyl length between the mutants and WT was significant under normal light, while the difference was not significant at shade and dark levels (Fig. 7, B and C). The reduction in root length is not significant under normal light (Supplementary Fig. S25).To explore the pattern of LSH5 affecting SOC in response to light, we explored the contribution to SOC of various combinations of CPAR and growth windows by CERIS (Li et al. 2021; Supplementary Fig. S26). We identified the environmental index, the window CPAR_173–186, which is strongly correlated with the environmental mean of SOC. GWAS for the slope of the reaction norm between the SOC of each genotype and the environmental index CPAR_173–186 commonly identified the qeOCA02.1. Thus, we propose the hypothesis that BnaLSH5 affects seed oil accumulations in response to light intensity.

Figure 7.

BnaLSH5 affects seed oil accumulations in respond of light. A–C) The length of hypocotyl and roots of mutants for LSH5. Significant reduction in hypocotyl length of mutants for LSH5 than WT during seedling in field trials A) and in the greenhouse B) (images were digitally extracted for comparison). Values are presented as means ± SD (n = 10 to 15). C) Significant hypocotyl length in the CR34, CR36, and CR41 mutant plants compared with those in the wild type plants. D, E) Artificial climatic chamber plantation of mutants for LSH5 in high- and low-light environment. High light level: 300 μmol m⁻² s⁻¹, low light level: 100 μmol m⁻² s⁻¹. D) Relative BnaA02.LSH5 transcript levels at high- and low-light level in rapeseed leaves of WT. Actin was used as an internal standard and values are means ± SD (n = 3 biological repeats). E) Seed oil content of mutants for LSH5 and WT at high- and low-light environment. Significant differences were observed in the regression slopes of PAR and SOC for CR34 and CR36 in comparison with WT (n = 6 to 8). F) Significant differences in net photosynthetic rate (P_n) of CR34 and CR36 mutant compared to WT in rosette leaves during seeding (n = 34 to 48). The center line, box limits, and whiskers represent the median, upper and lower quartiles, and 1.5 times the interquartile range, respectively. G) Tentative model explaining how transcription regulation of LSH5 on PMDH2 confers SOC plasticity in response to light intensity. In C-F, asterisks indicate significant levels of two-tailed unpaired Student's t-test, *, **, and *** indicated significant probability levels at 0.05, 0.01, and 0.001, respectively.

Rapeseed produce green seeds with photosynthetic activity, which has pleiotropic effects to seed metabolism (Hua et al. 2012). To validate the above hypothesis, mutants for LSH5 and WT were cultivated in an artificial climatic chamber at high and low light levels after flowering. RT-qPCR analysis showed that the expression of BnaA02.LSH5 is significantly higher in low light level environments (Fig. 7D). The difference in SOC between mutants and WT is significant under high light level, while the difference is not significant under low light level (Supplementary Fig. S27). The relative changes from low light level to high light level are decreased from 5.30% for WT to 1.89% and 4.12% for CR34 and CR41, respectively (Fig. 7E). Simultaneously, by construct the regress of PAR and SOC, the slope is significant decreased 0.029 and 0.012 from WT to CR34 and CR41, respectively, suggesting G × E of LSH5 on SOC (Fig. 7E). It is noted that several DEGs between mutants and WT are enriched in photosynthesis, i.e. photosystem II (GO:0009523), photosystem I (GO:0009522), and photosynthesis (GO:0015979). Because BnaA02.LSH5 and BnaC02.LSH5 are highly expressed in rosette, we therefore investigated 4 photosynthesis-related parameters between WT and mutants (CR34 and CR36) in rosette leaves (Fig. 7F; Supplementary Fig. S28). The net photosynthetic rate is significantly decreased in mutants than in WT (Fig. 7F; Supplementary Fig. S29). In addition, we predicted the upstream regulators for LSH5. We found that the PHYA is around the expression QTL (eQTL) region of BnaA02.LSH5 and BnaC02.LSH5 (Supplementary Fig. S26; Supplementary Tables S16 and S17). We further propose the model that BnaLSH5 affects seed oil accumulations in response to light intensity (Fig. 7G).

Discussion

The genetic basis and genes for SOC have been intensively studied by traditional linkage analysis and association studies in oil crops (Fu et al. 2016; Liu et al. 2016, 2020; Sun et al. 2016; Tang et al. 2021; Han et al. 2022b). In contrast, there are only a few reports on QEIs and GEIs of SOC in oil crops (Zhu et al. 2012; Zhao et al. 2022; Zuo et al. 2022), owing to the following reasons: the limitation of methodologies in QEI detection of GWAS (Zuo et al. 2022), the complex and dynamic effects of climatic indices on SOC during the seed development of oil crops (Guo et al. 2020), and the difficulties in prediction and functional validation of GEIs (Gupta et al. 2019). To address the above issues, we used the SOC phenotypes of 505 rapeseed inbred lines in 8 environments to detect environmentally stable and specific QTNs. The window of differential climatic factors for the 3 plantation sites was mined by PCA analysis. QEIs were detected by joint environments GWAS and by integrating the climate factors as covariates. By sampling 9 inbred lines with high SOC plasticity at 3 planting sites and 2 developing stages, we obtained 54 transcriptomes. By performing WGCNA with 54 transcriptomes, we identified SOC-related EDGMs and their hub genes. Known and additional candidate genes and GEIs were mined around these QTNs and QEIs by utilizing the EDGMs and the gene prioritization framework of Tang et al. (2021). These genes will provide useful resources for molecular marker-assisted breeding to create new high-oil germplasms adapted to the ecological environment of different main rapeseed production areas.

For QTNs identified using multi-locus method 3VmrMLM in this study, a total of 341 significant QTNs and 79 suggested QTNs are associated with SOC by single-environment GWAS in 8 trials (Supplementary Table S2), and a total of 124 significant QTNs and 42 suggested QTNs are found by joint environments GWAS of 3 datasets (Supplementary Table S3). Among these QTNs, 11 environmentally stable QTNs are commonly identified by at least 3 approaches, and 9 environmentally specific QTNs were commonly identified by at least 2 approaches at a specific planting site. Interestedly, all the environment stable QTNs were found around the QTNs detected by joint environment GWAS (Supplementary Table S4), and the environment-specific QTN qspOC.A02.1 were found to around QEI qeOC.A02.1. We compared environmentally stable QTNs and specific QTNs in this study with previously reported QTLs. As a result, 10 out of 20 QTNs are located in the genomic regions of known QTLs underlying SOC in rapeseed traits (Supplementary Table S4). For example, stable QTN qstOC.A05.2 is found simultaneously in the genomic region of 2 previously published QTLs Bn-A05-p21491115 (Liu et al. 2016) and qOC.A05.3 (Tang et al. 2021). In addition, the known genes are found around these QTNs, i.e. LEC1 (Tan et al. 2011), PMT6 (Tang et al. 2021), and NTT1 (Hong et al. 2022) were validated in rapeseed, LPD2 (Drea et al. 2001), VAL1 (Suzuki et al. 2007), MFP2 (Arent et al. 2010), GPDH (Zhao et al. 2021), and MDH (Kong et al. 2018) were validated in other plants.

Agriculture and food security are under long-term threat from increasing extreme weather events (Liu et al. 2023). Given the diverse environments in which rapeseed is cultivated, climate factors have complex and dynamic impacts on SOC (Si and Walton 2004; Zhang et al. 2013). We collected 22 climatic indices from 8 field trials at 3 planting sites for 71 DAF. The 8 trials were distinguishable based on climatic indices by PCA analysis, and the results are consistent with PCA analysis of SOC phenotypes, but with lower explained variance. This indicates that the climatic factors are very variable and have a complex effect on SOC. On the other hand, we detected G × E for SOC in rapeseed at the transcriptional and genetic analysis level. Firstly, we identified 8 EDGMs related to SOC by using WGCNA at 20 and 40 DAF during seed development. Some hub genes in the EDGM are known oil biosynthesis genes, e.g. WRI1 (Wu et al. 2014), FAB2 (Knutzon et al. 1992), and PDHC (De Meirleir et al. 1992) in turquoise20-blue40 EDGM, ABI3 (Tian et al. 2020) and OLEO1 (Zhang et al. 2019) in turquoise20-turquoise40 EDGM, suggesting diverse modules during oil accumulations. Besides, we detected 11 QEIs and 7 QEIs to be associated with SOC by considering PR₁₉, DTR₂₉, and UVB₁₅ as covariates. Using EDGM and POCKET (Tang et al. 2021), candidate genes were identified around these QEIs. These GEIs have potential for breeding applications to create high-oil germplasm resources with adaptations to specific environments. For example, BnaA04.STYK, which is specifically highly expressed in seeds (Supplementary Fig. S13), was found to be a hub gene in turquoise20-yellow40 EDGM and has been reported to regulate seed oil body biogenesis and to maintain growth capacity during postgerminative growth arrest under stress in Arabidopsis (Ramachandiran et al. 2018).

Light regulates plant growth and development and photosynthesis has pleiotropic effects on plant phenotypes (Jiao et al. 2007). In this study, we identified a GEI, BnaA02.LSH5 around the environmentally specific QTN qspOC.A02.1 and QEI qeOC.A02.1 detected with climatic indices as covariate, which affects seed oil accumulations in response to light intensity. LSH5 encodes a member of ALOG family of proteins containing a single conserved domain of unknown function, which have been reported to regulate inflorescence architecture and specify organ boundaries with spatial expression pattern in rice (Yoshida et al. 2009), tomato (MacAlister et al. 2012), and Arabidopsis (Zhao et al. 2004; Cho and Zambryski 2011; Takeda et al. 2011). However, the function of the ALOG family in seed-related traits is unknown. We validated the phenotype, transcriptome and metabolites of LSH5 CRISPR/Cas9 mutants and overexpression lines in field trials. We found the SOC and net photosynthetic rate are significantly decreased in BnaLSH5 mutants (Fig. 7F), and the ATP, NADPH, carbon and energy supply are limited (Supplementary Fig. S24). Then, through DEG analysis between mutants and WT, we inferred the potential transcription mechanisms that LSH5 affects the ATPase activity, photosynthesis, as well as the FA β-oxidation (Supplementary Fig. S24, A and B). Previous studies have shown that photosynthesis and Rubisco activity in the silique wall is highly associated with SOC (Hua et al. 2012). The light reactions of photosynthesis in the embryo provide both ATP and reductant, and the Rubisco bypass allows more efficient lipid synthesis (Ruuska et al. 2004; Schwender et al. 2004; Goffman et al. 2005).

To further dissect the mechanism by which LSH5 confers SOC plasticity, we first investigated downstream transcription regulation of BnaA02.LSH5 on SOC. BnaA02.LSH5 is highly expressed during early seed development and is localized in the nucleus as a transcription factor (Figs. 4, E and 5A). To date, the regulatory mechanisms of the ALOG family are not clear (Naramoto et al. 2020). In the previous study, ALOG was identified as an evolutionarily conserved plant-specific family that originated in algal evolution from a class of retrotransposons XerC/D-like recombinases (Naramoto et al. 2020). By using CUT&Tag, dual- luciferase, CUT&Tag-qPCR, and EMSA, we identified and validated the direct repression of BnaLSH5 on BnaA02.pMDH2. In Chlamydomonas, pMDH2 helps photoautotrophs cope with high light exposure by transmitting the redox state of the peroxisome to the chloroplast. The pmdh2 mutant accumulates triacylglycerol and starch, as well as maintains a higher photosynthetic activity (Kong et al. 2018). In this study, we found that BnaLSH5 mutants has significant reduction in SOC, ATP, and NADPH, an increase in BnaA02.pMDH2 expression, and a significant enrichment of DEGs in the peroxisomal FA catabolism. Thus, these results reveal the potential of BnaLSH5 in breeding for high SOC rapeseed.

Subsequently, we discovered the pattern of BnaLSH5 response to light. Although the BnaA02.LSH5 has a low expression level in leaves, we can still find the diurnal oscillations and the increased expression level at night (Xue et al. 2023). We found that the effects conferred by BnaLSH5 on SOC are environmentally and developmentally specific. BnaA02.LSH5 shows environmental expression variability, with its expression positively correlated with SOC at 20 DAF and negatively correlated with SOC at 40 DAF (Fig. 4, F and G). We also found a significant positive correlation between plasticity in SOC and plasticity in BnaA02.LSH5 expression at 40 DAF (Supplementary Fig. S17). Since the expression of BnaA02.LSH5 showed an opposite trend between light-related climatic factors on the day of sampling (Fig. 4F; Supplementary Fig. S16), we validated the light sensitivity of BnaLSH5 by measuring hypocotyls length in the field and by measuring SOC changes under high and low light levels using artificial climate chamber cultivation (Fig. 7E).

Finally, we rationalized the mechanism by which BnaLSH5 confers SOC plasticity. The variations in promoter region of BnaA02.LSH5 affect the SOC at different planting sites (Fig. 4B). We found that the SOC of the hap.B sub-population is significantly lower than hap.A at CD, whereas this was not the case at WH, where daylength, UVB and CPAR during seed development are higher than CD (Supplementary Fig. S12). Interestingly, most rapeseeds carrying hap.B are spring ecotype and have lower oil content, LSH5 expression, and π, which may attribute to the photoperiod response of spring ecotypes rapeseed. These results also suggest that the introgression of spring ecotype to semi-winter ecotype rapeseed is directly related to adaptation and plays a key role in canola improvement (Wang et al. 2023b). On the other hand, we found that PHYA were around the upstream eQTL of LSH5 (Supplementary Tables S16 and S17), which is consistent with AtLSH1 acting as a phytochrome-dependent endogenous factor in seeding mediating photomorphogenesis (Zhao et al. 2004). However, more experiments are needed to verify the precise upstream regulatory module of BnaA02.LSH5 during light-mediated developmental processes.

Breeders aim to identify markers or genes for SOC underlying G × E to stabilize the oil yield of varieties in different environments to cope with climate change and to breed varieties with a wide range of adaptations (Guo et al. 2020; Liu et al. 2021), whereas minimal efforts have been made in rapeseed (Tan et al. 2024). In this study, we systematically summarized that BnaLSH5 confers the SOC plasticity in 3 aspects, i.e. genotype, environment, and development (Fig. 7G). BnaLSH5 may emerge as a promising key gene for increasing SOC and adapting to low light environments in rapeseed breeding, holding the potential for future research and improvement efforts.

Materials and methods

Germplasm and data collection

A B. napus diversity panel of 505 rapeseed inbred lines collected from Tang et al. (2021) was used in this study. This panel was cultivated in 3 different field locations in 5 different years, resulting in 8 distinct environment-years field trials, i.e. WH for 4 years (WH2016, WH2017, WH2018, and WH2019), HF for 2 years (HF2017 and HF2018), CD for 2 years (CD2017 and CD2020). Additionally, 9 inbred lines for RNA sequencing (RNA-seq) were planted in Wuhan, Hefei, and Chengdu (WH2022, HF2022, and CD2022), at experimental fields of Huazhong Agricultural University, Anhui Academy of Agricultural Sciences, and Sichuan Academy of Agricultural Sciences, respectively. WT and CRISPR lines were planted in WH2022 and WH2023. The design of field trials of above materials was same as Tang et al. (2021).

The genotype data were obtained from Tang et al. (2021). We obtained 8,503,071 SNPs by filtering SNPs with a MAF (minor allele frequency) <0.01 and Geno (genotyping error) >0.1 using the PLINK software (Purcell et al. 2007). The rapeseed genome (ZS11 v0) from BnPIR (Song et al. 2020; http://cbi.hzau.edu.cn/bnapus/index.php) was used as the reference genome. The transcriptomes used for seed-specific expressed analysis and eQTL detection were obtained from Yu et al. (2023) and Tang et al. (2021), respectively. Environmental data were retrieved from the websites of the National Oceanic and Atmospheric Administration (NOAA: https://www.noaa.gov/weather) and the Astronomical Applications Department of the US Naval Observatory (https://www.usno.navy.mil/USNO/astronomical-applications).

Measurements of SOC and total FA content

Six SOC datasets (WH2016, WH2017, WH2018, HF2017, HF2018 and CD2017) in 505 lines were obtained from Tang et al. (2021), while 2 datasets (WH2019 and CD2020) were collected in this study. We measured the SOC of 505 inbred lines planted in WH2019 and CD2020. Mature open-pollinated seeds from 6 replicates of each accession were harvested and desiccated for SOC analysis with a Foss NIRSystems 5,000 near-infrared reflectance spectroscope (2000 to 3,000 seeds per scan) (Gan et al. 2003).

As described in Lu et al. (2016), the total FA content of developing seeds of 9 high plasticity lines in WH2022, HF2022, and CD2022 was measured by gas chromatography-flame ionization detector (GC-FID). Seeds were collected from 3 replicates at 20 and 40 DAF. Briefly, FA was methylated and extracted with methanol containing 5% (v/v) H₂SO₄ and 0.01% butylated hydroxyl toluene. The GC-FID parameters are described as following: oven temperature was maintained at 170 ℃ for 1 min and then increased to 210 ℃ with an increase rate of 3 ℃/min. FAs were identified according to their retention times, and the peak area was used to quantify FA composition by comparing with the internal standard heptadecanoic acid. Finally, FAs were identified according to their retention times and peak areas were analyzed. FAs were calculated as nmol% (Tang et al. 2021).

Detecting QTNs, QEIs, and eQTLs by GWASs

We associated 8,503,071 SNPs with SOC of 505 inbred lines by using the 3VmrMLM method implemented by the IIIVmrMLM software (Li et al. 2022a, 2022b). The single-environment module was used to identify QTNs for SOC in each of 8 trials, while the multi-environment module was used to identify QTNs and QEIs for SOC in the Datasets I, II, and III. Dataset I is 3 planting sites in 2017 (WH, HF, and CD), Dataset II is the BLUP values that obtained from different years in the same site, and Dataset III is the 8 trials. Population structure and kinship matrix were considered in all the analysis, where population structure matrix (k = 4) was derived from Tang et al. (2021), and kinship matrix was calculated by IIIVmrMLM. According to the first 10 features variance explained by loadings on PC1 and PC2, their climatic indices of 8 trials were used as covariates in Dataset III by using the multi-environment module. Using single-environment module, regression parameters of the trait phenotypes on CPAR_173–186 were used to indirectly identify QEIs for phenotypic plasticity of SOC, and the expressional levels of BnaLSH5 s were used to identify eQTLs. The “svpal” parameter which is the threshold parameter for the first step of 3VmrMLM in single-marker screening was set as 0.01. Significant and suggested QTNs are identified at the critical P-value of 4.27e-09 (=0.05/m, where m is the number of markers) and QTNs with the LOD score ≥3.0 but the P-value ≥0.05/m, respectively.

RNA extraction, RNA-seq, and RT-qPCR

We analyzed RNA-seq and total FA content of developing seeds of 9 inbred lines planted in WH2022, HF2022 and CD2022. The development seed at 20 and 40 DAF were collected for RNA-seq with 3 biological replicates. The detailed sampling processes were described in Tang et al. (2021). Total RNA was extracted using the TIANGEN RNAprep Pure Plant Kit. Sequencing libraries were generated using the NEBNext Ultra RNA Library Prep Kit for Illumina (NEB, USA, Catalog #: E7530L) and were sequenced on an Illumina Hiseq 4000 platform. We used MultiQC (Ewels et al. 2016) to perform quality control and Salmon (Patro et al. 2017) to quantify the RNA-seq reads of annotated genes in the reference ZS11 v0. The expression levels of genes were quantitatively estimated using the Transcripts Per Kilobase of exon model per Million mapped reads (TPM) method.

RT-qPCR was performed with SYBR green fluorescent labeling of double-stranded DNA by using a CFX Connect instrument (Bio-Rad). The expression level was normalized to that of ACTIN7 (BnaA06G0089200WE). PCRs were set up in 96-well Hard-Shell PCR plates (Bio-Rad, USA) with 0.4-μM primers using SYBR Premix TransStart Green qPCR SuperMix (TransGen, China) in 15 μL and run on a Bio-Rad CFX Connect. All the data were collected from 3 biological replicates and 3 technical replicates. PCR conditions were as follows: 1 cycle of 95 °C for 1 min; 45 cycles of DNA melting at 95 °C for 15 s, 60 °C for 15 s, and 72 °C for 20 s. Primers were designed using NCBI primer design software and are listed in Supplementary Table S18.

WGCNA and DEG analysis

Genes were excluded if the TPM value of expression level in all the samples was less than 1 and the missing value was exceeding 20% of the total number of samples. Two WGCNA were respectively performed with above 81 and 81 seed transcriptomes using the R package at 20 DAF and 40 DAF, with considering each biological replicate as a distinct data set (Langfelder and Horvath 2008). The networkType and TOMType parameters are set to “signed” for 20 and 40 DAF networks. Soft-thresholding power was set to be 9 to satisfy scale-free topology of the networks (Supplementary Fig. S5B). The modules were divided using the following parameters: deepSplit-4, minModuleSize-1000, and mergeCutHeight-0.25. The eigengene for each module was respectively correlated with SOC at 20 and 40 DAF. We define the EDGM as the sub-module at the intersection of 2 modules of different developmental stages (20 and 40 DAF), and the eigengene of these modules are associated with SOC at both stages, respectively. The EDGMs was constructed with the threshold of 0.3 in the TOM matrix by cytoscape (Shannon et al. 2003). Base on MCC values calculated by cytoHubba the top 10 of genes were taken to determine the hub genes of each EDGM (Chin et al. 2014). The FA content The R package DESeq2 was used for DEG analysis (Love et al. 2014). Threshold for DEG is set as the absolute value of log₂FoldChange greater than 1 and adjusted P-value less than 0.05 (two-tailed Student's t-test; Tan et al. 2022a, 2022b).

Candidate genes and GEIs identification

Genes between 100 kb upstream and downstream regions of each QTN or QEI were extracted according to LD interval described in Tang et al. (2021) to mine candidate genes. Known candidate genes were defined as the oil biosynthesis genes via transgenic experiments validation in previous research or the rapeseed homologs of Arabidopsis lipid metabolism genes from ARALIP (http://aralip.plantbiology.msu.edu/). Then, candidate genes for QTNs and GEIs were firstly scored using POCKET with multiple omics data described in Tang et al. (2021). In brief, the POCKET scores were calculated by using the following formula with 4 effects:

$$\mathrm{POCKE}{\mathrm{T}}_{\mathrm{score}}=\mathrm{G}{\mathrm{F}}_{\mathrm{g}}\times (\mathrm{E}{\mathrm{V}}_{\mathrm{e}}+\mathrm{E}{\mathrm{E}}_{\mathrm{t}}+2\mathrm{E}{\mathrm{H}}_{\mathrm{hap}})$$

GF_g is the gene function score representing the correlation of gene and trait. We used known key regulators of oil biosynthesis as positive set and 2 times lager other genes as negative set to train the model by SVM. After 10-fold cross-validation in each training, the mean value of probability was calculated as the final score. EV_e is the maximum value of variant effect among the variants located in the coding region. EE_i is the gene expression effect using the P-value of TWAS and cis-eQTL EH_hap is the P-value calculated by haplotype-based association studies. The code for POCKET is available at https://github.com/zhaouu/POCKET. Secondly, the candidate genes for environment-specific QTN and GEI were further screened based on the modular eigengene profiles of the EDGMs from the top 5 scoring genes of the POCKET.

Subcellular localization

The coding region of BnaA02.LSH5 was cloned and ligated into a pMDC83 vector contained a 2 × 35S promoter and a C terminal GFP tag. The primers used are listed in Supplementary Table S18. After transformation of the recombinant BnaA02.LSH5::GFP vector into Agrobacterium strain GV3101, it was infiltrated into N. benthamiana epidermal cells with P19 for transient expression (Yang et al. 2021). Two days after transfection, BnaA02.LSH5–GFP proteins were observed in epidermal cells at using confocal laser scanning microscopy (FV1200, Olympus, Japan).

Genomic editing and plant transformation of LSH5 in rapeseed

The Westar of rapeseed was used for genetic manipulations and physiological experiments. To generate BnaLSH5 mutants, 2 sgRNAs simultaneously targeting at the exon of BnaA02.LSH5 and BnaC02.LSH5 were designed by CRISPR-P2.0 (http://crispr.hzau.edu.cn/CRISPR2/) (Lei et al. 2014) and putative off-target sites were manually eliminated. The primers for the amplification of LSH5 were listed in Supplementary Table S18. Oligonucleotides for the target site annealed to form double-stranded DNA were inserted into pKSE401 vector carrying a Kanamycin selection marker (Zhang et al. 2020). The vectors were transformed into the Agrobacterium (Agrobacterium tumefaciens) strain GV3101. Transgenic plants were produced with the resulting plasmids via Agrobacterium-mediated transformation of rapeseed as previously described in (Dai et al. 2020).

T₀ plants were obtained by kanamycin screening (25 mg/L), and the Cas9 protein was identified by primer pairs Cas9-F/R. Then the positive plants with Cas9 were selected to amplify BnaA02.LSH5 and BnaC02.LSH5, respectively, and the amplified fragments were sequenced and analyzed to identify edited T₀ mutants. To obtain homozygous mutants, the T₀ mutants were self-crossed for T₁ and T₂ generations and the homozygosity of mutation was confirmed by sequencing. Primers used in the identification were listed in Supplementary Table S18.

Measurement of photosynthetic efficiency

Seeds of WT and mutants for LSH5 were planted in the 2022 to 2023 field trial in Wuhan. The photosynthetic efficiency was measured using a portable photosynthesis system (LI-6800; Li-Cor, Lincoln, NE). The data were saved after the measured parameters remained stable and under the following conditions: with the paraments of 600 μmol s⁻¹ flow rate, 60% relative humidity, 400 μmol mol⁻¹ CO₂, 1,000 rpm fan speed, 1,200 μmol m⁻² s⁻¹ light intensity, 25 °C.

Artificial climate chamber cultivation

Seeds of WT and mutants for LSH5 were planted in the growth chamber PGW40 (Conviron, Surrey, Bc, Canada) and its growth phenotype was observed indoors under a condition of 16 h light/8 h dark, 22/20 °C, and a relative humidity of 60%, The high light level was set as 300 µmol m⁻² s⁻¹. The low light level was set as 100 µmol m⁻² s⁻¹.

Metabolite extraction and analysis

Metabolites involved in primary carbon metabolism were extracted and measured by LC–MS/MS based on previous study with modifications (Tang et al. 2022a). Briefly, whole seeds of 20 DAF seeds were collected and freeze-dried (Millrock Technology, Kingston, NY). Five milligram of samples was weighed, and a mix of 1.8 mL of methanol:chloroform (7:3, v/v) was added. The samples were incubated for 2 h at −20 °C and gently mixed several times during the incubation. 1.6 mL of ddH₂O was added and samples were vortexed and centrifuged at 100 × g for 5 min. The upper was transferred into a new tube and dried using nitrogen. The extracts were re-dissolved with 200 μL ddH₂O and then filtered with 0.45 μM cellulose acetate centrifuge tube filters. The re-dissolved extracts were used for metabolites analysis by LC–MS/MS (QTRAP 6500 plus). Mass signal intensity normalized by dry weight was calculated for comparing primary metabolites.

Protoplast isolation, CUT&Tag assay, and sequencing analysis

We conducted transient transformations of rapeseed leaf protoplasts and subsequently performed CUT&Tag experiments utilizing the CUT&Tag kit (Vazyme, TD904) with 2 biological replicates. The method for extracting rapeseed leaf protoplasts was adapted from the previously reported method for Arabidopsis leaf protoplast extraction (Yoo et al. 2007). The full-length coding sequence of BnaA02.LSH5 was cloned into the pM999-GFP vector, which is driven by the 35S promoter of CaMV, through the XbaI restriction site. Subsequently, the recombinant plasmid was introduced into rapeseed protoplasts via PEG-mediated transformation. Following a 16-h culture period, the protoplasts that successfully expressed GFP fluorescence were lysed, and the sequencing library was constructed using the Hyperactive Universal CUT&Tag Assay Kit (Vazyme, TD904). The successfully constructed libraries were high-throughput sequenced on the Illumina NovaSeq platform by Novogene (Beijing, China). By using BWA (v0.7.17) and the MEM algorithm, clean reads were mapped into the ZS11 v0 reference genome (Li and Durbin 2010). High confidence peaks calling was performed via MACS2 (v2.2.7.1) with the parameter “CallPeak -g 1e + 9 -s 150 -B –nomodel –shift 100 –extsize 200 -q 0.01 -f BAMPE” (Zhang et al. 2008). Peaks were visualized via pyGenomeTracks (Ramírez et al. 2018). FIMO and meme-chip in the MEME Suite were used for the motif analysis (Bailey et al. 2015). The immunoprecipitated DNA was used for the qPCR assay to quantify the enrichment of the targeted fragment, which was performed with SYBR Premix TransStart Green qPCR SuperMix (TransGen, China) on a Bio-Rad CFX Connect. DNA Spike in was used as a negative control. Primers for the targeted fragments were designed based on the motif analysis or CUT&Tag-seq analysis, as listed in Supplementary Table S18.

Dual-luciferase assay

The coding regions of BnaA02.LSH5 were amplified and inserted into the pM999-YFP vector to generate the effector construct. The promoters of BnaA02.pMDH2 was amplified and inserted into the pGreenII-0800-LUC vector as the reporter construct. The method for extracting Arabidopsis protoplasts was based on (Yoo et al. 2007). Dual-luciferase assays were performed using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA). After extracting the Arabidopsis protoplasts and dark culture for 16 h, the Arabidopsis protoplast liquor was dissolved into 200 μL of passive lysate. 50 μL of each sample was added to the 96-well plate. The SPARKRMULTIMODE MICROPLATE (TECAN, Swiss) was used for luminescence measurements. 50 μL of the firefly luciferase reagent (LARII) was added to the test sample, with a 10-s equilibration time and measurement of luminescence with a 10-s integration time, followed by the addition of 50 μL of the REN reagent and firefly quenching (Stop and Glow TM buffer), 10-s equilibration time, and measurement of luminescence with a 10-s integration time. All data were saved as the ratio of firefly to Renilla luciferase activity (Fluc/Rluc). Each group has at least 3 biological replicates in each assay.

EMSA

The full-length coding sequence of BnaA02.LSH5 was amplified from the cDNA of ZS11 and cloned into the pGEX-6p-1 vector. The GST-BnaA02.LSH5 fusion protein was expressed in Escherichia coli Rosetta for 16 h at 16 °C in the presence of 0.4 mm isopropyl β-D-thiogalactoside (IPTG). GST-BnaA02.LSH5 protein was purified using GST.Bind Resin (70541, Merck) according to the protocol provided by manufacturer. The promoter fragments of BnaA02.pMDH2 containing motifs were synthesized and labeled with FAM at the 5′-end. The forward and reverse probes were heated at 98 °C for 10 min and then cooled to room temperature to generate double-strand probes. The FAM-labeled and unlabeled probes were incubated with GST-BnaA02.LSH5 protein in the EMSA/Gel-shift Binding Buffer (GS005, Beyotime) for 30 min at 23 °C, respectively. The reaction mixture was loaded on a 6% (w/v) native polyacrylamide gel. The electrophoresis was performed at 80 V for 1 h in 0.5×TBE buffer (45 mm Tris-base, 45 mm boric acid, and 0.5 mm EDTA, pH 8.3) at 4 °C in the dark. FAM-labeled DNA was detected using Amersham Typhoon (Cytiva). All primers used for EMSA are listed in Supplementary Table S18.

Statistical analysis

Phenotypic plasticity is indicated by the slope of the Finlay–Wilkinson regression model, which was implemented by the FW package (Lian and de los Campos 2016) by regressing 505 phenotypes with environmental index in 8 trials (WH2016, WH2017, WH2018, WH2019, HF2017, HF2018, CD2017, and CD2020), which is calculated using the CERIS algorithm (https://github.com/jmyu/CERIS_JGRA; Li et al. 2021). In brief, the algorithm calculates the average SOC of the 505 genotypes in each environment (environmental mean of SOC in 8 field trials). The pipeline then calculates the average value of CPAR for the window from DAF_i to DAF_j, where DAF_i is the ith day after planting, DAF_j is the jth day after planting, and 5 < j-i < 35. The correlation between the vector of environmental mean and the vector if each parameter–window value was calculated separately. The parameter–window combination with the strongest correlation was then selected as the environmental index and used for further analyses. Subsequently, the Finlay–Wilkinson regression model was adopted based on the SOC of each genotype and the environmental index (CPAR_173–186). The slope of the reaction norm parameter representing the phenotype plasticity of SOC in response to the CPAR_173–186 was as a phenotype for GWAS. The relative difference of SOC at 2 different light level is calculated as $(\mathrm{SO}{\mathrm{C}}_{\mathrm{high}}-\mathrm{SO}{\mathrm{C}}_{\mathrm{low}})/\mathrm{SO}{\mathrm{C}}_{\mathrm{mean}}\times 100\text{\%}$. Based on Nekola and White (1999), the function “diffslope” in R package “simba” is used to calculate the statistical inference of the difference in slope between 2 regression lines.

Based on the kinship matrix, the SNP-based heritability was calculated for each environment with SOC as the phenotype by using GCTA (Yang et al. 2011), with default parameters: “–bfile –make-grm –make-grm-alg 0 –out; –grm –pheno –reml –out”. BLUP values for SOC were estimated using the R package lme4 (https://github.com/lme4). PCA was performed by R package FactoMineR with function “PCA” (https://cran.r-project.org/web/packages/FactoMineR). The VCA was performed as described in Guo et al. (2020) by using R package VCA with function “anovaVCA” (https://cran.r-project.org/web/packages/VCA). The t-test was adopted in the hypothesis testing for haplotype analysis, *, **, and *** indicated significant probability levels at 0.05, 0.01, and 0.001, respectively. Statistical analyses in this study are provided in Supplementary Table S19.

Accession numbers

Sequence data from this article can be found in the Brassica napus multi-omics information resource (BnIR) (https://yanglab.hzau.edu.cn/) and National Center for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov/)under the following accession numbers: BnaA02.LSH5 (BnaA02G0110000ZS), BnaC02.LSH5 (BnaC02G0134800ZS), BnaA02.pMDH2 (BnaA02G0032600ZS), BnaA04.STYK (BnaA04G0161300ZS), BnaA09.VPE (BnaA09G0572900ZS), and BnaA09.DAO1 (BnaA09G0636200ZS).

Author contributions

L.G., Y.M.Z., and T.G. conceived this study. L.G. supervised all the experiments. Y.M.Z. supervised the statistical methods. T.G. supervised the bioinformatics analysis. X.H., X.W., Q.T., Y.L., K.H., S.F., W.L., H.L., S.T., Y.P. performed most of the experiments. X.H., K.H., S.F., W.L., H.L., S.T. performed most of the field trials. X.H., Y.Z., Z.L., Y.X., and H.Z. performed the bioinformatics analysis. X.H. and X.W. wrote the manuscript. X.H., X.Y., T.G., Y.M.Z., and L.G. revised the manuscript. All authors approved the manuscript.

Supplementary data

The following materials are available in the online version of this article.

Supplementary Figure S1. Frequency distributions and correlations for seed oil contents (SOC) of 505 rapeseed inbred lines in 8 field trials.

Supplementary Figure S2. The Manhattan plots of QTNs for seed oil content in 8 trials using the single-environment module in IIIVmrMLM software.

Supplementary Figure S3. QTNs for seed oil content identified by at least 2 approaches in 8 trials using single-environment module of the IIIVmrMLM software.

Supplementary Figure S4. The Manhattan plots of QTNs and QEIs for seed oil content in 3 datasets by using the multi-environment module of IIIVmrMLM software.

Supplementary Figure S5. Principal component analyses and soft power determination of weighted gene co-expression network analyses (WGCNA) of 54 transcriptomes at 2 seed development stages in 3 planting sites.

Supplementary Figure S6. Identification of modules associated with the seed oil content at 20 and 40 DAF.

Supplementary Figure S7. Four Environmental-Developmental Gene co-expression Modules (EDGMs) with eigengenes negatively correlated with SOC in the blue module at 20 DAF by WGCNA.

Supplementary Figure S8. Correlations between maximal clique centrality (MCC) scores and 3 variance components (genotype, environment, and genotype-environment interactions) for genes in 8 EDGMs.

Supplementary Figure S9. The distribution of the average variance of 10 genes randomly chosen 10,000 times in each EDGM.

Supplementary Figure S10. QTNs and QEIs for seed oil content identified in 3 datasets by using multi-environment module of IIIVmrMLM software.

Supplementary Figure S11. 22 climate factors from 0 to 70 DAF in 8 field trials (Supports Fig. 2).

Supplementary Figure S12. Significant differences in light-related climate factors from 0 to 70 DAF in 3 planting sites.

Supplementary Figure S13. Expression patterns of candidate gene for QTL-environment interactions in multiple tissues.

Supplementary Figure S14. Candidate gene identification of qspOC.A02.1 and qeOC.A02.1 by using POCKET (prioritizing the candidate genes by incorporating information of knowledge-based gene sets, effects of variants, GWAS, and transcriptome-wide association studies).

Supplementary Figure S15. Nucleotide diversity (π) among spring ecotype and semi-winter ecotype rapeseeds in A02 chromosome.

Supplementary Figure S16. Trends in 4 light climate factors in WH2023, HF2023, and CD2023 after flowering.

Supplementary Figure S17. The correlation between plasticity in SOC and plasticity in expression of BnaA02.LSH5 at 20 and 40 DAF.

Supplementary Figure S18. Acquisition of mutant lines based on CRISPR/Cas9 technology.

Supplementary Figure S19. Overexpression lines of BnaA02.LSH5 was constructed in rapeseed driven by 35S promoter.

Supplementary Figure S20. Late flowering phenotypes of LSH5 mutants between CR34 and CR41 compared to wild type and overexpression lines.

Supplementary Figure S21. Haplotype analysis of flowering time of BnaA02.LSH5 (n_hap.A = 276, n_hap.B = 73).

Supplementary Figure S22. Principal component analysis (PCA) of transcriptomes of developing seed of mutant plants of LSH5.

Supplementary Figure S23. Mutants of BnaLSH5 impact primary carbon metabolism in developing seeds at 20 DAF.

Supplementary Figure S24. Mutants of BnaLSH5 impact on ATP/ADP, NADH/NAD, and NADPH/NADP in developing seeds at 20 DAF.

Supplementary Figure S25. Significant reduction in root length in the CR34, CR36, and CR41 mutant plants compared with those in the wild type plants.

Supplementary Figure S26. The pattern of photosynthesis active radiation (CPAR) affecting SOC during rapeseed growth.

Supplementary Figure S27. Quantitative comparisons of seed oil content in wild type (WT), CR34, and CR41 mutant plants by near-infrared reflectance spectroscope in artificial climate chamber.

Supplementary Figure S28. Quantitative comparisons of photosynthesis-related parameters in wild type (WT), CR34 and CR36 mutant plants.

Supplementary Figure S29. The Manhattan plots of additive expression QTNs and their candidate genes for BnaLSH5 s at 20 and 40 DAF by using single-environment module of IIIVmrMLM software.

Supplementary Table S1. Phenotypic characteristics of SOC in 8 field trials of 505 rapeseed accessions.

Supplementary Table S2. QTNs for rapeseed SOC in 8 trials by using single-environment module of IIIVmrMLM software.

Supplementary Table S3. Environmentally stable and specific QTNs for rapeseed SOC identified by 8 trials.

Supplementary Table S4. QTNs and QEIs for rapeseed SOC in 3 datasets by using multi-environment module of IIIVmrMLM software.

Supplementary Table S5. Nine accessions for RNA sequencing in development seed at 20 and 40 DAF.

Supplementary Table S6. Significant enriched GO terms in each EDGM.

Supplementary Table S7. The network topological analysis and variance components analysis of genes in each EDGM.

Supplementary Table S8. Characterization of 25 climatic indices from 0 to 70 DAF in 8 field trials.

Supplementary Table S9. Variance explained by climatic indices loadings on PC1 and PC2.

Supplementary Table S10. SOC-related QEIs identified with the 3 climatic factors as covariate in Dataset III by using multi-environment module in IIIVmrMLM software.

Supplementary Table S11. POCKET results of each QTN and QEI for SOC.

Supplementary Table S12. Measurement of agronomic traits for LSH5 mutants after harvest.

Supplementary Table S13. GO enrichment analysis of differential expressed genes between mutants and WT at 20 and 40 DAF.

Supplementary Table S14. Peak annotation of CUT&Tag assay from 2 biological replicates.

Supplementary Table S15. GO enrichment analysis of 665 target genes commonly identified of BnaA02.LSH5 and BnaC02.LSH5.

Supplementary Table S16. Expression QTNs for BnaA02.LSH5 in development seed by using single-environment module of IIIVmrMLM software.

Supplementary Table S17. Expression QTNs for BnaC02.LSH5 in development seed by using single-environment module of IIIVmrMLM software.

Supplementary Table S18. Primers for gene cloning and PCR confirmation.

Supplementary Table S19. Statistical analyses in this study.

Funding

This work was supported by grants from Joint Funds of the National Natural Science Foundation of China (U2102217), National Natural Science Foundation of China (32401862), Hubei Provincial Natural Science Foundation (2024AFB117) and Fundamental Research Funds for the Central Universities (2662022ZKYJ003). We thank the high-performance computing platform at the National Key laboratory of Crop Genetic Improvement in Huazhong Agriculture University.

Data availability

All data generated or analyzed during this study are included in this published article, its supplementary information files and publicly available repositories. The raw sequencing data of RNA-Seq and CUT&Tag libraries analyzed in this study have been deposited in the National Genomics Data Center under the accession number PRJCA035155.

Dive Curated Terms

The following phenotypic, genotypic, and functional terms are of significance to the work described in this paper:

• lsh5 Gramene: AT5G58500

• lsh5 Araport: AT5G58500

• MDH Gramene: AT3G47520

• MDH Araport: AT3G47520

• MFP2 Gramene: AT3G06860

• MFP2 Araport: AT3G06860

• ABI3 Gramene: AT3G24650

• ABI3 Araport: AT3G24650

• LEC1 Gramene: AT1G21970

• LEC1 Araport: AT1G21970

• OLEO1 Gramene: AT4G25140

• OLEO1 Araport: AT4G25140

• PHYA Gramene: AT1G09570

• PHYA Araport: AT1G09570

• WRI1 Gramene: AT3G54320

• WRI1 Araport: AT3G54320

• VAL1 Gramene: AT2G30470

• VAL1 Araport: AT2G30470

• FUSCA3 Gramene: AT3G26790

• FUSCA3 Araport: AT3G26790