Genome-wide analysis of long noncoding RNAs in response to drought stress in the seeds of Brassica napus

Abstract

Background

Long noncoding RNAs (lncRNAs) have been shown to play important roles in plant abiotic stress response and adaptation. However, the identification and characterization of genome-wide drought-responsive lncRNAs in rapeseed (Brassica napus L.) have been limited. Therefore, this study was the first to identify the expression profile of lncRNAs in rapeseed seeds responding to prolonged drought stress and subsequent short-term rewatering.

Results

A total of 6 000 lncRNAs were identified, among which 181 were classified as differentially expressed lncRNAs (DELs) in response to either drought stress or subsequent rewatering. Comparative analysis revealed that only 14 DELs were shared between the 159 DELs identified during drought stress and the 27 DELs detected upon rewatering. GO enrichment analysis showed that the co-expressed DEGs, primarily involved in photosynthesis, central carbon metabolism, stomatal movement, and strigolactone metabolism, were significantly down-regulated under drought stress but markedly up-regulated in subsequent rewatering. Furthermore, drought‑responsive competing endogenous RNA (ceRNA) networks were constructed based on the identified DETs and DEGs. Two ceRNA modules, MSTRG.57345.1-bna-miR164a/b/c/d-HSP2 and MSTRG.57345.1-bna-miR395d/e/f-HSFA7a, which are based on the newly identified lncRNA MSTRG.57345.1, were detected for the first time in rapeseed under drought stress and subsequent rewatering.

Conclusions

The present study advances our understanding of the expression patterns and functional role of rapeseed lncRNAs in the response to drought stress and the subsequent rewatering. It provides novel insights into lncRNA-mRNA networks and lncRNA-miRNA-mRNA networks in the seeds of B. napus. These findings offer a valuable reference for further genetic research and molecular breeding programs aimed at rapeseed improvement.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-025-12496-8.

Introduction

Drought stands as one of the most significant abiotic constraints in agricultural production, adversely affecting physiological metabolism, morphogenesis, and reproductive development in plants, and leading to considerable losses in biomass and economic yield [1]. To mitigate drought stress, plants have evolved adaptive strategies to reduce water loss, such as optimizing water allocation to essential organs and preserving cellular hydration. These adjustments elicit physiological changes reflected in leaf relative water content, chlorophyll concentration, and photosynthetic efficiency [2]. Additionally, drought stress promotes the overproduction of reactive oxygen species, which disrupts redox homeostasis and induces oxidative stress [3]. This condition is marked by lipid peroxidation, severe cellular damage, reduced membrane stability, and enhanced protein denaturation [4]. In severe cases, prolonged drought may ultimately result in plant mortality. Therefore, enhancing plant drought resistance is of utmost importance. Improving crop resilience is not only crucial for increasing agricultural output but also imperative for promoting environmental sustainability. Achieving this objective requires a thorough understanding of drought tolerance mechanisms and the elucidation of intricate molecular networks that enable plants to survive and adapt to increasingly arid conditions.

In eukaryotic genomes, over 90% of the DNA is transcribed, yet only a small portion (2%–25%) encodes proteins [5, 6]. The remaining parts are classified into different classes of noncoding RNAs (ncRNAs), which were long regarded as transcriptional noise without any biological significance [7]. Over the past few years, numerous significant findings have emerged about the involvement of short RNAs (like miRNA) and long noncoding RNAs (lncRNAs) in epigenetic regulation [8]. LncRNAs are defined as transcripts longer than 200 nucleotides with little or no protein-coding capacity, yet they demonstrate significant potential in modulating gene expression [9]. They function through cis-regulatory actions on neighboring genes or trans-regulatory effects on targets distributed genome-wide. Additionally, lncRNAs can function as miRNA “sponges” that competitively bind to miRNAs, thereby alleviating miRNA-mediated repression of target mRNAs and indirectly regulating gene expression [10]. LncRNAs contribute critically to plant growth, development, and stress adaptation, highlighting their promising applications in crop breeding. Advances in next-generation sequencing have enabled the sequencing of hundreds of plant genomes. Based on these genomic datasets, advanced whole-transcriptome RNA sequencing endeavors have deciphered the crucial regulatory dynamics governed by both protein-coding RNAs and ncRNAs in plants [5]. For example, Meng et al. (2023) conducted a comprehensive transcriptomic analysis on Malus spectabilis leaves under low-nitrogen and control-nitrogen conditions, and found that lncRNAs eTM858-1 and eTM858-2 serve as endogenous target mimics (eTM) for miR858 and bind to it to prevent the cleavage of the target gene MsMYB62-like, thereby inhibiting anthocyanin biosynthesis under the low-nitrogen condition [11]. Accumulating evidence supports the model that lncRNAs act as central regulators of drought tolerance, where they interact with target genes to form sophisticated regulatory networks that fine-tune multi-gene expression programs, thereby shaping the plant’s integrated stress response strategy.

Rapeseed (Brassica napus L., AACC, 2n = 4x = 38) is the third-largest oilseed crop in the world, accounting for about 13% of global oil production (USDA). Rapeseed is highly sensitive to drought stress, which leads to significant yield losses and reduced seed oil content [12]. Breeding drought-tolerant varieties represents a viable strategy to maintain yield stability in drought-prone environments. Growing evidence indicates that lncRNAs participate in regulating plant responses to multiple abiotic stresses, including drought. For instance, drought-responsive lncRNAs have been shown to be associated with photosynthesis, chlorophyll synthesis, fatty acid synthesis and degradation in Arabidopsis [13]. A critical lncRNA (DROUGHT INDUCED lncRNA, DRIR) was identified to be a novel positive regulator of the plant response to drought and salt stress in Arabidopsis [14]. Additionally, drought-responsive lncRNAs have been demonstrated in various crops such as cotton [15], barley [16], switchgrass [17], peanut [18], and wheat [19]. For example, Li et al. (2023) identified a novel drought-responsive lncRNA 1 (DRL1) that functions as a competing endogenous RNA (ceRNA) by sequestering miR477b, which in turn relieves the miRNA-mediated repression of its target genes, GhNAC1 and GhSCL3, thereby positively regulating drought tolerance in cotton seedlings [20]. Xu et al. (2024) revealed that lncRNA35557 acts as an eTM of tae-miR6206, effectively preventing tae-miR6206 from cleaving the TaNAC018, thereby further enhancing drought tolerance in wheat [21]. In B. napus, Tan et al.. (2020) demonstrated that pathways related to plant hormone signaling, defense/stress responses, and photosynthesis were significantly activated under both drought and rewatering conditions in two contrasting genotypes: drought-tolerant Q2 and drought-sensitive Qinyou8 [22]. Further investigation by Tan et al.. (2024) revealed 184 differentially expressed lncRNAs responding to both drought and rehydration treatments, with plant hormone signal transduction playing a pivotal role in the resilient Q2 genotype seedlings [23]. Despite these advances, the functional mechanisms of lncRNAs in the widely cultivated rapeseed cultivar ZhongShuang 11 under drought stress remain largely unexplored.

In this study, we conducted a comprehensive genome-wide identification and characterization of lncRNAs in mature seeds of the rapeseed cultivar Zhongshuang 11 subjected to drought stress from the flowering stage through seed maturation. The potential functions of the differentially expressed lncRNAs (DELs) and their target differentially expressed genes (DEGs) were analyzed based on lncRNA-mRNA co-expression networks. And some key DETs and ceRNA modules were analyzed and identified based on lncRNA-miRNA-mRNA networks. These results would provide valuable knowledge for elucidating the functional mechanisms of lncRNAs in conferring drought tolerance in rapeseed.

Materials and methods

Plant material, growth conditions, and treatments

Seeds of the rapeseed cultivar Zhongshuang 11 were provided by the Oil Crops Research Institute (OCRI) of the Chinese Academy of Agricultural Sciences (CAAS) in Wuhan, China. The experiment was carried out from March 18 to August 3, 2025, on the exterior corridor of the second floor of Chongli Building at the Dongshan Campus of Shanxi University. As a result, the light intensity and temperature conditions for plant growth mirrored the ambient environmental fluctuations. Rapeseed seeds were initially sown in a sand/vermiculite mixture (1:1, v/v) for germination under greenhouse conditions. Four-week-old seedlings were then transplanted into plastic pots (24 × 27 × 31 cm) filled with the same substrate mixture, with two seedlings per pot. Accordingly, this study employed a pot-based cultivation system combined with controlled water withholding to simulate drought stress.

Until the flowering stage, all pots were watered regularly to maintain approximately 70% soil humidity during the vegetative growth phase. After two months, when plants had reached the whole flowering stage, they were randomly assigned to three treatment groups: control group (CK), drought stress group (DS), and rewatering group (RW). The initial soil moisture content of CK continued to receive adequate irrigation (maintaining 70–80% soil humidity), while watering was withheld entirely for the DS group until soil moisture decreased to approximately 20%. Soil moisture content was measured every two days to maintain the target levels. For the RW, plants were rewatered after undergoing 45 days of drought stress during the seed maturation stage. Seeds from CK, DS, and RW were collected, immediately frozen in liquid nitrogen, and stored at − 80 °C for subsequent analysis. Three biological replicates were included for each treatment condition.

RNA extraction, library construction, sequencing, and mapping to the reference genome

All experimental procedures involving total RNA extraction, library construction, Illumina sequencing, and read mapping were conducted by Shanghai Majorbio Bio-pharm Biotechnology Co., Ltd. (Shanghai, China). Specifically, total RNA was extracted from approximately 100 mg of seed samples (with three biological replicates for each group) using TRIzol^® Reagent according to the manufacturer’s protocol (Invitrogen, USA). Subsequently, ribosomal RNA was removed to construct longRNA-seq libraries, which were then sequenced on the NovaSeq X Plus platform (PE150) with NovaSeq Reagent Kits [24]. Raw reads were processed through quality control and adapter trimming using SeqPrep (https://github.com/jstjohn/SeqPrep) and Sickle (https://github.com/najoshi/sickle) under default parameters. The resulting clean reads were aligned to the B. napus reference genome ZS11_v10 in orientation mode using TopHat software (http://tophat.cbcb.umd.edu/, version 2.0.0).

LncRNA identification and classification

The adaptor-polluted reads, low-quality reads, splice error, etc. were filtered out from raw data to produce clean data controlled by Fastp with default parameters [25]. The clean reads were aligned to the reference genome using HiSat2 [26]. Transcript assembly for each sample was performed with StringTie based on a probabilistic model [27]. The assembled transcripts were compared against reference transcripts and lncRNA databases to identify known lncRNAs, followed by the removal of transcripts resembling other ncRNAs or mRNAs. To obtain candidate lncRNAs, transcripts meeting the following criteria were retained: length > 200 bp and exon number ≥ 2, in accordance with canonical lncRNA characteristics. Furthermore, the coding potential of candidate lncRNAs was systematically assessed using four independent approaches: Coding Potential Calculator (CPC) [28], Coding-Non-Coding Index (CNCI) [29], and Pfam protein domain analysis [30]. Finally, transcripts consistently classified as non-protein-coding by all three methods were designated as novel lncRNAs.

Differential expression analysis and functional enrichment

For interlibrary comparisons, the expression level of each transcript was calculated according to the transcripts per million reads (TPM) method by utilizing the RSEM package [31], and the genes with TPM ≥ 1 were identified as expressed. Differential expression analysis was performed using the DESeq2 [32], and genes or lncRNAs with a fold change ≥ 2 and an adjusted P-value < 0.05 were regarded as significantly differentially expressed genes (DEGs) or differentially expressed lncRNAs (DELs).

In addition, functional enrichment analysis, including GO and KEGG enrichment, was performed with Goatools (https://github.com/tanghaibao/Goatools) and KOBAS (http://kobas.cbi.pku.edu.cn/home.do). The DEGs significantly enriched in GO and KEGG pathways were conducted at Bonferroni-corrected P-value < 0.05 compared with the whole-transcriptome background.

Prediction of cis- and trans-target genes

The cis-target genes of the DELs were predicted by identifying its neighboring DEGs located within a 10 kb upstream or downstream of a lncRNA using the BEDtools v2.29.2 software [33]. And the trans-target genes were predicted by analyzing its co-expressed DEGs attained through correlation analysis or co-expression analysis of DELs and DEGs with Pearson correlation coefficient (PCC). The targeted DEGs were identified with |PCC value|≥0.8 and Bonferroni-corrected P-value < 0.05.

Construction of the lncRNA-miRNA-mRNA network

Given that the functional roles of lncRNAs remain incompletely characterized, a widely adopted approach for inferring their mechanistic actions relies on the analysis of co-expressed mRNAs. To identify potential lncRNA-mRNA pairs, the lncRNA-mRNA co-expression network was constructed according to Yu et al.. (2023) [34]. DEGs and DELs with a correlation at |PCC| ≥ 0.8 and a Bonferroni-corrected p < 0.05 were selected for co-expression network construction. And the interactions between miRNA-lncRNA and miRNA-mRNA were predicted through psRNATarget (https://www.zhaolab.org/psRNATarget/analysis), following the scoring schema V2 (2017 release). The network was then visualized using Cytoscape (v3.8.0; https://cytoscape.org/).

Quantitative real-time PCR (qRT-PCR) validation

To obtain the first-strand cDNA, the total RNA of each sample was reverse transcribed using the EasyScript^® One-Step gDNA Removal and cDNA Synthesis SuperMix (TransGen, China). The relative expression levels of lncRNAs were performed with PerfectStart^® Green qPCR SuperMix (TransGen, China) on CFX Connect Real-Time System (Bio-Rad, USA) in accordance with the provided protocol. The designed primers of selected lncRNAs are listed in Supplementary Table S1, and the BnTIF4 gene was used as the reference gene. Relative gene expression level was calculated using the 2^−ΔΔCt method. For each sample in the qRT-PCR reaction, three biological and three technical replicates were performed.

Results

Phenotypic characterization of rapeseed in response to drought stress

To comprehensively investigate the effects of drought tolerance on rapeseed, we conducted a commercial rapeseed variety Zhongshuang 11 under drought treatment throughout its reproductive growth stage (Fig. 1). The relative soil humidity of CK maintained approximately 70% throughout the plant development (Fig. 1B). Stop watering from the flowering stage, and the relative soil humidity in both the DS and RW groups decreased to 20–30% after about 40 days. Distinct morphological differences were observed between well-watered and drought-stressed plants (Fig. 1A). Compared to the fully expanded leaves of well-watered plants, the drought-stressed plants exhibited varying degrees of wilting, with particularly pronounced effects during the seed development stage (30 days after drought stress) and the seed maturation stage (45 days after drought stress). After 30 days of drought treatment, no significant difference in relative plant height was detected between well-watered and drought-stressed treatments; however, the drought-stressed plants exhibited leaf yellowing, early leaf abscission, and premature silique yellowing (Fig. 1A). By 45 days of continuous drought stress, these plants displayed complete leaf abscission, severe wilting, along with accelerated silique maturation accompanied by seed abortion (Fig. 1A and Supplementary Figure S1). To further explore rapeseed genomic responses to drought, plants subjected to 45 days of drought were rewatered for 24 h, simulating natural recovery conditions to evaluate regenerative capacity.

Fig. 1.

Phenotype of rapeseed under drought stress and well-watered conditions. A Physiological response of rapeseed under 30 and 45 days after drought stress treatment. Bar = 10 cm. B Relative humidity of the rapeseed in CK, DS and RW groups

Whole transcriptome sequencing analysis of different samples

To investigate the response of lncRNAs in rapeseed to different drought conditions, seeds were collected from the CK, DS, and RW groups. Next-generation sequencing was performed to generate over 578.33 million raw reads across the different treatment groups. After filtering out low-quality sequences, an average of approximately 64 million clean reads per sample was obtained (Table 1). In total, we obtained approximately 87.33 Gb of raw data and 87.18 Gb of high-quality clean data (Supplementary Table S2). Among these clean reads, more than 99% were successfully mapped to the ZhongShuang 11 reference genome, with unique mapping rates ranging from 78% to 83% (Table 1). Furthermore, the average quality scores (Q20 and Q30) of the clean reads reached 99.44% and 96.70%, respectively. These results collectively demonstrate the high reliability and quality of the sequencing data.

Table 1.

Statistical data of the RNA-Seq reads under different treatments

Sample	Raw Reads	Clean Reads	Q20(%)	Q30(%)	Total Mapped Reads	Unique Mapped Reads
Seed_CK1	60262306	60233802	99.48	96.89	60207680(99.96%)	47952918(79.61%)
Seed_CK2	65833792	65798504	99.46	96.76	65766492(99.95%)	53874972(81.88%)
Seed_CK3	57468264	57442206	99.48	96.89	57415663(99.95%)	47686626(83.02%)
Seed_DS1	68901648	68867344	99.29	95.65	68825177(99.94%)	57258380(83.14%)
Seed_DS2	55069162	55045134	99.49	96.95	55026907(99.97%)	45414110(82.5%)
Seed_DS3	65588546	65550974	99.46	96.78	65519798(99.95%)	54184831(82.66%)
Seed_RW1	68800594	68761994	99.46	96.8	68729881(99.95%)	56384584(82.0%)
Seed_RW2	68205584	68167966	99.47	96.83	68139823(99.96%)	53614423(78.65%)
Seed_RW3	68202498	68165982	99.45	96.76	68141187(99.96%)	56553020(82.96%)

Identification of LncRNAs and mRNAs in rapeseed

After removal of viral genomic sequences and subsequent assembly, annotation, and filtering of all transcripts from the RNA-Seq data, approximately 6 000 lncRNAs (Supplementary Table S3) and 44 974 mRNAs (Supplementary Table S4) was identified based on the ZhongShuang 11 reference genome. All of the lncRNAs were more evenly distributed across all 19 chromosomes in rapeseed (Fig. 2A). Based on their genomic locations, 2 895 (48%) were classified as sense_exon_overlap lncRNAs overlapping with exonic regions of protein-coding genes, while 89 (1%) were sense_intron_overlap lncRNAs residing within intronic regions. In addition, 2 015 (34%) were intergenic lncRNAs, 815 (14%) were antisense lncRNAs, and 186 (3%) were bidirectional lncRNAs located within 1 kb upstream of coding genes and transcribed in the opposite direction (Fig. 2B). Moreover, lncRNAs consistently exhibited lower expression levels compared to mRNAs (Fig. 2C). The average transcript length of lncRNAs was shorter than that of mRNAs (Fig. 2D). Specifically, 76% of lncRNAs ranged between 500 bp and 2 000 bp in length, whereas 70% of mRNAs fell within the 500 bp to 2 500 bp range. In terms of exon composition, the majority of lncRNAs contained two or three exons: 47% of lncRNAs had two exons and 17% had three. In contrast, mRNAs generally possessed more exons, with 22%, 14%, and 12% of mRNAs containing over ten, two, and three exons, respectively (Fig. 2E). These differences in exon numbers may indicate or correlate with distinct functions between lncRNAs and mRNAs.

Fig. 2.

Comparison of structural features between lncRNAs and mRNAs. A The expression level of lncRNAs (log10 TPM) along 19 rapeseed chromosomes. It comprises three concentric rings, and each corresponds to a different sample. From outer to inner, they are CK, DS and RW samples, respectively. B Classification of identified lncRNAs. C Expression level comparison between lncRNAs and mRNAs. D Length distribution of lncRNAs and mRNAs. E Exons distribution in lncRNAs and mRNAs. In panels C to E, green represents lncRNAs, and blue represents mRNAs

Differentially expressed LncRNAs and mRNAs responsive to drought stress and rewatering

In this study, a total of 6 465 DEGs and 181 DELs were identified among three compare groups (Fig. 3). Specifically, in the DS vs. CK comparison, 425 DEGs were up-regulated and 5 822 were down-regulated. The RW vs. CK comparison showed 193 up-regulated and 165 down-regulated DEGs, while the RW vs. DS comparison revealed 538 up-regulated and 42 down-regulated DEGs. Among all DEGs, only four (ZS11A02G036940, ZS11C01G008480, ZS11C04G043410, and ZS11C07G023200) were common to all three comparison groups. Regarding DELs, the DS vs. CK comparison contained 82 up-regulated and 77 down-regulated transcripts, whereas the RW vs. DS comparison showed 8 up-regulated and 19 down-regulated DELs. The RW vs. CK group had 18 up-regulated and 8 down-regulated DELs, with only one DEL (MSTRG.10213.2) shared across all three groups. Comparative analysis demonstrated that the DS vs. CK group contained significantly more DEGs/DELs than either the RW vs. CK or RW vs. DS comparisons, indicating that prolonged drought stress elicited extensive transcriptomic changes in rapeseed. Furthermore, the predominance of down-regulated over up-regulated DEGs suggests a prevalence of transcriptional suppression under drought conditions.

Fig. 3.

Identification and characterization of differentially expressed genes (DEGs) and lncRNAs (DELs) under drought stress and subsequent rewatering.

A Number of down-regulated and up-regulated DEGs. B Number of down-regulated and up-regulated DELs. C Venn diagrams of the DEGs among three different groups. D Venn diagrams of the DELs among three different groups

Different LncRNA expression patterns under drought conditions in rapeseed

To characterize the dynamics expression pattern of lncRNAs in response to drought stress, we performed cluster analysis of all identified lncRNAs. As shown in Fig. 4, all the lncRNAs could be categorized into six distinct expression profiles. Subcluster 2, comprising the largest number of lncRNAs (5 243), showed minimal expression variation under drought stress (Fig. 4A), consistent with the limited number of DELs identified in this study (Fig. 3B). Subcluster 3, the second largest group (268 lncRNAs), was rapidly down-regulated upon rewatering following drought stress. Subclusters 1 and 7 exhibited contrasting expression patterns: subcluster 1 (102 lncRNAs) was up-regulated under drought stress and down-regulated after rewatering, whereas subcluster 7 (60 lncRNAs) showed the opposite trend. Subcluster 4 (91 lncRNAs) maintained sustained up-regulation during both drought and rewatering phases, suggesting a potential positive regulatory function in the drought response. Additionally, subcluster 5 (52 lncRNAs) displayed more pronounced expression changes after rewatering compared to subcluster 3. Together, these dynamic expression profiles support the involvement of lncRNAs in mediating drought stress responses in rapeseed.

Fig. 4.

Transcriptomic profiling of rapeseed lncRNAs in response to drought stress and subsequent rewatering. A Expression patterns of LncRNAs under different treatments. B GO pathway analysis for the target DEGs of the lncRNAs in different subclusters

To further investigate the functional relevance of these lncRNAs, we predicted their target DEGs and performed GO enrichment analysis. A total of 11 533 lncRNA-mRNA pairs were identified in subcluster 1, involving 4 622 DEGs primarily associated with photosynthesis (Fig. 4B and Supplementary Table S5). Similarly, subclusters 2, 4, and 7 contained 832 856, 15 658, and 107 012 lncRNA-mRNA pairs, encompassing 5 732, 4 857, and 5 409 DEGs, respectively. Most target DEGs in these subclusters were also related to photosynthesis and chloroplast thylakoid membrane processes (Fig. 4B and Supplementary Table S5). It indicated that photosynthesis plays a significant role in the drought resistance of rapeseed. In subclusters 3 and 5, 4 681 and 543 lncRNA-mRNA pairs were identified, involving 1 118 and 164 DEGs, respectively. DEGs in subcluster 3 were significantly enriched in extracellular region, carbohydrate metabolic process, and oxidoreductase activity, while those in subcluster 5 were primarily associated with DNA integration and DNA polymerase activity (Fig. 4 and Supplementary Table S5).

Target gene prediction and functional annotation of DETs under drought stress

To elucidate the functions of the identified DELs, we identified putative target DEGs located within 100 kb of DEL loci and constructed interaction networks based on predicted lncRNA-mRNA relationships. In the DS vs. CK comparison, 159 DELs were found to regulate 122 cis-target DEGs and 300 069 trans-target DEGs. The RW vs. DS comparison showed that 27 DELs regulated 9 cis-target and 2 147 trans-target DEGs, while in RW vs. CK, 26 DELs regulated 4 cis-target and 484 trans-target DEGs (Supplementary Tables S6-S8). Notably, the regulatory relationships exhibited a many-to-many pattern, with individual lncRNAs often targeting multiple mRNAs and vice versa.

In DS vs. CK, 159 DELs formed 300 191 lncRNA-mRNA pairs with 5 483 DEGs (Supplementary Tables S6). Among them, 122 lncRNA-mRNA pairs are in cis-regulation, and others are in trans-regulation. Interestingly, 21 DELs can target their own corresponding mRNAs. Besides, two novel lncRNAs (MSTRG.55134.2 and MSTRG.60019.1) targeted the largest number of mRNAs (4 737), whereas MSTRG.63748.2 targeted the fewest (49). Similarly, two mRNAs (ZS11A09G013270 and ZS11C01G018230) were each targeted by 100 lncRNAs. In RW vs. CK, 26 DELs formed 2 714 pairs with 319 DEGs (Supplementary Table S7). LncRNA MSTRG.34713.2 targeted the most mRNAs (214), while MSTRG.53958.1 targeted only 43 mRNAs; the mRNA ZS11A03G018890 was regulated by the highest number of lncRNAs (18). In RW vs. DS, 27 DELs formed 2 156 pairs with 588 DEGs (Supplementary Table S8). LncRNA MSTRG.30356.3 targeted 489 mRNAs - the highest among this comparison - while MSTRG.32668.3 targeted as few as 4; the mRNA ZS11A01G039500 was associated with the most lncRNAs (19). Together, these indicate that a relatively extensive regulatory network was formed in DS vs. CK in response to drought stress for up to 45 days, whereas a comparatively compact network in RW vs. DS group was constituted during the 24‑hour short‑term rewatering.

Additionally, we performed GO and KEGG enrichment analyses on target DEGs for all the DETs. GO enrichment analysis revealed that DEGs associated with photosystems, chloroplast function, and stomatal movement were highly enriched under both drought stress and 24-hour rewatering conditions (Fig. 5A, Supplementary Table S9). A key distinction was observed in their regulation patterns: the genes involved in these pathways were predominantly down-regulated in long-term drought stress conditions but up-regulated in 24-hour rewatering conditions. This might suggest that long-term drought stress represses stomatal movement-related genes to limit water loss, while short-term rewatering activates their expression to restore stomatal conductance, thereby enhancing photosynthetic efficiency and energy supply. In addition, down-regulated DEGs in long-term drought stress condition and up-regulated DEGs in short-term rewatering condition were significantly enriched in central carbon metabolism and energy-related pathways, aligning with changes in photosynthetic output. Similarly, lipid and fatty acid metabolic pathways were significantly suppressed under prolonged drought stress conditions (Fig. 5A, Supplementary Table S9), implying that drought may compromise seed oil accumulation during development. Notably, up-regulated DEGs under drought showed significant enrichment in the glyoxylate cycle, a key pathway in oilseed plants that supports the conversion of lipids to carbohydrates during seed development. It might indicate that rapeseed mobilizes its storage lipids and converts them into energy substrates to sustain essential metabolic processes under drought stress.

Fig. 5.

GO and KEGG enrichment analysis for the co-expressed DEGs of the DELs. A GO pathway analysis of co-expressed DEGs of the DELs in response to drought stress and subsequent rewatering. B KEGG pathway analysis of co-expressed DEGs of the DELs under drought stress and subsequent rewatering

Likely, KEGG pathway analysis showed that DEGs involved in photosynthetic systems, carbon fixation, cyanoamino acid metabolism, and nitrogen metabolism were strongly enriched under both drought stress and subsequent rewatering (Fig. 5B, Supplementary Table S10). It might suggest that the photosynthetic system and associated carbon fixation processes constitute one of the primary and most rapidly responsive biochemical systems during extended drought and short-term rehydration. However, pathways related to carbohydrate metabolism, central carbon metabolism, and plant secondary metabolism-such as starch and sucrose metabolism, glyoxylate and dicarboxylate metabolism, amino sugar and nucleotide sugar metabolism, phenylpropanoid biosynthesis, and tryptophan metabolism-were significantly enriched during rewatering compared to the control (Fig. 5B, Supplementary Table S10). Taken together, the divergence in enriched pathways ultimately influenced the composition and accumulation of storage reserves in rapeseed seeds following drought and recovery.

LncRNA-miRNA-mRNA regulatory networks in the seed of rapeseed response to drought stress

LncRNAs can also function as ceRNA for miRNAs, thereby influencing the expression of mRNAs. Given that dehydration and rehydration represent opposing biological processes in plants, this study aims to focus on the genes that are either up-regulated during drought and down-regulated after rewatering, or conversely, down-regulated under dehydration and up-regulated upon rehydration. These candidate drought-responsive genes will serve as targets for investigating their ceRNA regulatory networks.

In this study, 2 DETs and 466 DEGs were down-regulated in DS vs. CK but up-regulated in RW vs. DS, and 10 DETs and 23 DEGs were up-regulated in DS vs. CK but down-regulated in RW vs. DS (Fig. 6A). Only one of the two DETs, MSTRG.57654.1, regulated 402 DEGs (all down-regulated in DS vs. CK and up-regulated in RW vs. DS), of which 149 DEGs were the predicted targets of 88 miRNAs (Supplementary Table S11). However, no miRNA was predicted to bind to MSTRG.57654.1 and MSTRG.19715.4. This may be attributed to (1) he limited number of up‑regulated DETs under the brief (24‑hour) rewatering condition, and (2) the possibility that these two DETs indirectly influence the miRNA‑target gene network by modulating other ceRNAs or transcription factors (TFs). For the drought-responsive genes that up-regulated in DS vs. CK but down-regulated in RW vs. DS, a regulatory network was constructed consisting of 10 DETs, 30 miRNAs, and 22 DEGs in the seeds of rapeseed under drought stress and rewatering (Fig. 6B, Supplementary Table S12). Among them, 10 lncRNAs could target all 22 mRNAs. Notably, 9 lncRNAs (MSTRG.10213.2, MSTRG.16004.1, MSTRG.65271.1, MSTRG.50916.1, MSTRG.16742.3, MSTRG.16742.2, MSTRG.57345.1, MSTRG.10800.1, MSTRG.65589.1) were predicted to bind at least 10 mRNAs. And 30 miRNAs were predicted to bind to 8 of 22 mRNAs. Interestingly, a novel identified lncRNA MSTRG.57345.1 participates simultaneously in two distinct ceRNA networks. That is, bna-miR164a, bna-miR164b, bna-miR164c, and bna-miR164d all potentially regulate ZS11A05G014640 (Heat shock protein 2-like, HSP2) by targeting MSTRG.57345.1. Similarly, bna-miR395d, bna-miR395e, and bna-miR395f collectively target MSTRG.57345.1 to modulate the expression of ZS11A03G043600 (Heat stress transcription factor A-7a, HSFA7a) (Fig. 6B, Supplementary Table S12). Coincidentally, both HSP2 and HSFA7a are heat responsive proteins. Here, MSTRG.57345.1 may function by competitively binding to specific miRNAs during drought stress, thereby relieving their repression of heat-responsive protein genes. It may suggest a potential link between drought and heat stress as physiologically synergistic processes in plants. This finding further points to a potential role for MSTRG.57345.1 in mediating rapeseed responses to drought and heat stress.

Fig. 6.

LncRNA-miRNA-mRNA regulatory network based on the differentially expressed drought responsive genes. A Venn diagram showing the number of unique and common DETs and DEGs between DS vs.CK group and RW vs. DS group. B LncRNA-miRNA-mRNA network based on DETs and DEGs that up-regulated in DS vs. CK but down-regulated in RW vs. DS. Hexagon indicates lncRNA, diamond indicates miRNA, and bllipse indicates mRNA

Identification of differentially expressed TFs under drought stress and rewatering

TFs including MYB, NAC, WRKY, and bZIP are well-established regulators of drought response in plants. We systematically analyzed the TF families represented among the DEL-targeted DEGs (Fig. 7). Under drought stress, 478 differentially expressed TFs were identified and classified into 36 families. Nine major families accounted for 71.76% of the total, including MYB and MYB-related (71 TFs), NAC (46), M-type (35), bHLH (34), HB-other (34), bZIP (32), WRKY (31), ERF (30), and MIKC (30) (Fig. 7A, Supplementary Table S13). Following rewatering after drought stress, 46 differentially expressed TFs were identified, distributed across 18 families. Seven of these families comprised 67.39% of the total (31 TFs), including HB-other (5), M-type (5), MIKC (5), MYB-related (5), bHLH (4), ERF (4), and NAC (3) (Fig. 7C, Supplementary Table S13). Notably, 39 differentially expressed TFs were co-expressed under both drought and rewatering conditions, suggesting their potential roles in a core regulatory program responsive to water stress dynamics.

Fig. 7.

Distribution of TF families in the seeds of rapeseed in response to drought stress and subsequent rewatering. A Distribution of TF families in DS vs. CK. B Distribution of TF families in RW vs. CK. C Distribution of TF families in RW vs. DS

Validation of target genes of drought‑responsive LncRNAs

Recently, a key transcription factor BnNAC038 involved in drought stress response was reported in Brassica napus, which could coordinate the molecular network between drought response and yield balance, and achieve the breeding goal of “drought tolerance without yield reduction” [35]. In our study, we found BnNAC29 (ZS11C06G037860) is significantly down-regulated under drought stress and significantly up-regulated under subsequently rewatering process. Here, we constructed a lncRNA-miRNA-mRNA network associated with BnNAC29 (Fig. 8A, Supplementary Table S14). Within this network, 22 lncRNAs were predicted to target BnNAC29, and bna-miR2111c and other 22 miRNAs were predicated to bind to BnNAC29 and 22 lncRNAs, respectively. However, no significantly expressed ceRNA network based on BnNAC29 was identified in rapeseed seeds. This may be because the related ceRNA interactions possess higher condition- or tissue-specificity and were not sufficiently activated during the seed developmental stage examined in this study. Besides, 15 of 22 lncRNAs could target a considerable network of 3 230 DEGs related to drought response in seeds, including TFs (such as bHLH25, bZIP11 and WRKY23), calcium signaling genes (CAM7, CML35 and CBP1), hormone signaling genes (such as ABAI5, ABAI7 and ERF104), sugar transporting genes (such as STP1 and STP11) and so on.

Fig. 8.

The regulatory network of BnNAC29 in response to drought stress and subsequent rewatering. A Calculated interaction network of BnNAC29. Circle in yellow indicates BnNAC29, hexagon in blue indicates lncRNA, diamond in red indicates miRNA, and ellipse in pink indicates mRNA. B The gene expression validation of BnNAC29 and its associated lncRNAs

To verify the quality of the high-throughput sequencing of lncRNAs, 12 differentially expressed DETs were randomly selected for qRT-PCR verification. The R² value (R² = 0.92 > 0.9) indicated a significant correlation between the expression levels of the DE-lncRNAs quantified by TPM and qRT-PCR results (Figure S6). We also performed qRT-PCR to examine the expression patterns of BnNAC29 and 2 lncRNAs (the other 13 lncRNAs were lowly expressed in RNA-seq). The expression patterns of two lncRNAs, MSTRG.7792.1 and MSTRG.57654.1, were strongly correlated with BnNAC29, particularly under drought stress (Fig. 8B). Further experiments are required to confirm their relationship with BnNAC29 and to elucidate their roles in the response to drought stress.

Discussion

Plants encounter numerous abiotic stressors, including drought, low temperature, and high salinity, which collectively constrain their growth and developmental processes. Drought tolerance is a complex physiological trait controlled by numerous genes, making it essential to study plant stress responses in order to advance agricultural productivity and practice. In recent years, lncRNAs have emerged as key molecular regulators in mediating crops adaptation to environmental challenges [36, 37]. With the development of modern molecular breeding tools, it has become increasingly feasible to enhance crop resilience to abiotic constraints. For example, advances in next-generation sequencing have enabled genome-wide identification of lncRNAs across diverse plant species [38]. Growing evidence indicates that lncRNAs participate in a range of fundamental biological processes, including the regulation of mRNA processing and transcriptional mechanisms such as splicing, editing, localization, translation, and turnover [39]. LncRNAs are also involved in transcriptional gene silencing [40], regulation of protein-coding gene expression [41], and the modulation of plant epigenetics and chromatin structure [42].

Although lncRNAs have been extensively studied in many plant species, research on lncRNAs in rapeseed remains comparatively underexplored. Previous studies in rapeseed have explored lncRNAs related to clubroot resistance [43], lipid metabolism [44], seed oil accumulation [24], and anther and pollen development [45]. More recently, Tan et al. (2020, 2024) conducted a detailed comparative analysis of lncRNAs in rapeseed seedlings under drought stress and subsequent rewatering [22, 23]. Nevertheless, as a major oilseed crop, it is essential to investigate how drought stress initiated during the flowering stage influences seed development in rapeseed. In the present study, high-throughput whole-transcriptome sequencing was performed to analyse lncRNAs in seeds of the rapeseed cultivar Zhongshuang 11 under drought stress and subsequent rewatering, compared with well-watered controls. We identified approximately 6 000 novel lncRNAs, among which 82 were up-regulated and 77 down-regulated under long-term-drought condition, while 8 were up-regulated and 19 down-regulated following rewatering (Fig. 3). By comparison, previous studies reported 369 down-regulated and 108 up-regulated DELs in Q2 seedlings, and 449 and 257 DELs in Qinyou8 [22]. The notable differences in DEL numbers between Zhongshuang 11 and the previously studied genotypes Q2 and Qinyou8 may be attributed to variations in plant materials, tissue types, timing and duration of drought stress, or differences in lncRNA filtering criteria.

Despite variations among plant species and stress conditions, lncRNAs generally share conserved characteristics, such as relatively short transcript length, low expression abundance, and a predominance of transcripts containing only one or two exons [46]. Consistent with earlier reports in tobacco and other species [47, 48], most lncRNAs identified in this study were shorter than 2 000 nt and consisted of only 2–3 exons (Fig. 2). Unlike previous studies that focused on drought-stressed seedlings, our work specifically addresses the effect of drought on seed development. Interestingly, the number of DELs detected in seeds was comparable to that in three-leaf-stage seedlings [23]. Moreover, the DS vs. CK comparison showed a higher number of DELs than the RW vs. CK and RW vs. DS comparisons, a trend consistent with the observed DEG patterns (Fig. 3). This likely reflects the broad transcriptomic reprogramming induced by prolonged drought, whereas the 24-hour rewatering triggered only a short-term rapid adjustment in drought-adapted plants, leading to the differential abundance of DELs and DEGs across comparisons. In addition, most DEGs were down-regulated after drought stress, whereas the majority were up-regulated following rewatering (Fig. 3). This pattern suggests that rapeseed may mitigate the effects of prolonged drought by suppressing physiological processes, while rapidly restoring partial metabolic pathways in response to short-term rewatering.

LncRNAs are known to execute their function through the cis and trans regulation of protein-coding genes. In this study, we performed GO and KEGG enrichment analyses on protein-coding genes targeted by the identified DELs. Consistent with findings in Arabidopsis and other plant species, DEGs associated with photosystem I and II, chlorophyll biosynthesis, chloroplast thylakoid formation, and stomatal movement were significantly enriched following prolonged drought stress in rapeseed [13, 49, 50]. This may be attributed to the cellular dehydration under drought conditions, which disrupts chloroplast and thylakoid membrane integrity, thereby impairing light harvesting and electron transport efficiency. Moreover, stomatal closure as a water-conservation strategy limits CO₂ uptake, thereby constraining carbon fixation during photosynthesis. Phytohormones, particularly strigolactones, appear to play distinctive roles in rapeseed’s response to drought and subsequent recovery after rewatering (Fig. 6). This functional specialization, differing from classical ABA-mediated signaling [51, 52], may provide novel targets for improving drought tolerance in rapeseed breeding. As expected, lipid metabolic pathways were markedly suppressed by drought stress in seeds-the primary lipid storage organs (Fig. 6). This may result from drought-induced reduction in carbon fixation capacity, limiting the substrate supply for lipid biosynthesis. Alternatively, drought may activate stress-responsive transcriptional cascades that collectively inhibit lipid biosynthesis genes, resulting in attenuated lipid metabolism. Following rewatering, most DEGs were up-regulated, primarily involving photosynthetic light and dark reactions that were impaired during drought, as well as central carbon metabolism pathways that responded rapidly to water restoration. This recovery profile differs from pathways such as valine, leucine, isoleucine degradation, and tyrosine/tryptophan metabolism previously reported in rapeseed seedlings [23], suggesting that drought response mechanisms may vary across developmental stages and tissue types.

LncRNAs are also known to interact with miRNAs via diverse routes-such as functioning as miRNA precursors, serving as target mimics, or acting as direct miRNA targets-especially during abiotic stress responses. The ceRNA network has emerged as a recognized regulatory model in a wide range of plant species. However, no identified ceRNAs have been reported in Brassica napus. In our present study, we found two ceRNA modules that were up-regulated under drought stress (DS vs. CK) but down-regulated upon rewatering (RW vs. DS): MSTRG.57345.1-bna-miR164a/b/c/d-HSP2 and MSTRG.57345.1-bna-miR395d/e/f-HSFA7a (Fig. 6). In these two ceRNA modules, a novel lncRNA MSTRG.57345.1 can simultaneously act as a “sponge” for two independent miRNA families (miR164 and miR395) and regulate heat-responsive protein genes (HSP2 and HSFA7a), respectively. This suggests that it may be a key node for integrating different upstream signals and coordinating various downstream stress responses. And the two targeted DEGs of MSTRG.57345.1, HSP and HSFA, are known to play essential roles in plant heat responses [53]. Indeed, drought and heat stress exhibit a synergistic interaction, where each stressor can exacerbate the effects of the other. Under high-temperature conditions, increased water loss and enhanced soil evaporation are more likely to intensify drought. Conversely, under drought stress, reduced transpiration rates lead to elevated leaf surface temperatures [54]. Thus, the heat-responsive proteins HSP2 and HSFA7a may be involved in the processes of dehydration and rehydration in rapeseed. Additionally, miR164 and miR395 play critical roles in plant responses to abiotic stress. Plant miR164 is a conserved microRNA that plays essential roles in plant growth, development, and stress responses by post-transcriptionally regulating its target genes [55]. For example, foxtail millet miR164b-SiNAC015 module, apple miR164g-MsNAC022, and wheat miR164-TaNAC14 could reduce reactive oxygen species to improve drought tolerance and plant development [56–58]. Besides, tomato miR164a-NAM3-HSFA4b module and petunia miR164-NAC are involved in plant heat stress response [59, 60]. This demonstrates the dual function of miR164 in conferring both drought tolerance and heat tolerance. Plant miR395 is conserved, and both the ends of the bidirectional promoters of miR395 harbors several abiotic stresses (nutrient, salt, drought) responsive cis-motifs in five Brassica species [61]. In A. thaliana, miR395a-HSFA1b is also involved in plant thermotolerance [62]. This may suggest a potential role for the miR395-HSFA module in rapeseed drought stress responses. However, few studies have linked the miR164-HSP and miR395-HSFA modules with lncRNAs in the context of plant drought or heat tolerance. In this study, a novel lncRNA, MSTRG.57345.1, was identified in rapeseed seeds by whole transcriptome sequencing, and the regulatory network of MSTRG.57345.1-bna-miR164a/b/c/d- HSP2 and MSTRG.57345.1-bna-miR395d/e/f-HSFA7a was constructed for the first time in rapeseed under drought stress and subsequent rewatering. This “one-to-many” regulatory module is rarely reported in plant stress responses, suggesting MSTRG.57345.1 may be a key regulatory node in rapeseed. Further investigation into the two aforementioned ceRNA networks focused on the specific roles of MSTRG.57345.1 in the rapeseed drought stress would be valuable.

Conclusion

In summary, a comprehensive analysis of lncRNAs involved in the drought stress response in rapeseed was conducted. A total of 6 000 lncRNAs were discovered, with 181 of them being differentially expressed in response to drought stress and the subsequent rewatering. Moreover, lncRNA-mRNA and ceRNA networks elucidated molecular mechanisms of drought stress adaptation in rapeseed and revealed promising candidate genes for improving drought tolerance in rapeseed. Furthermore, we found that the up-regulated DEGs targeted by DELs were primarily associated with pathways related to photosynthesis, central carbon metabolism, stomatal movement, and strigolactone metabolism under prolonged drought stress. In contrast, down-regulated DEGs targeted by DELs were also enriched in these same pathways during subsequent rewatering. Besides, a novel lncRNA, MSTRG.57345.1, was identified in rapeseed seeds, and the regulatory network of MSTRG.57345.1-bna-miR164a/b/c/d- HSP2 and MSTRG.57345.1-bna-miR395d/e/f-HSFA7a was constructed for the first time in rapeseed under drought stress and subsequent rewatering. The present study provides significant insights about the roles of lncRNAs in the drought stress and subsequently rewatering treatment of rapeseed. Nevertheless, further functional analysis is required to confirm these observations and clarify the precise mechanisms through which these lncRNAs function in drought stress tolerance.

Abbreviations

LncRNA

Long noncoding RNA

DEGs

Differentially expressed genes

DELs

Differentially expressed lncRNAs

CK

Control group

DS

Drought stress group

RW

Rewater stress group

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

TF

Transcription factors

HSP2

Heat shock protein 2-like

HSFA7a

Heat stress transcription factor A-7a

Authors’ contributions

KZ designed the experiments and wrote the manuscript; WBW and JRL contributed to the material planting and sample collection; WBW also analyzed the RNA-seq data; ZL revised the manuscript. KZ and WBW contributed equally to this paper. MTL and KZ designed, led and coordinated the overall study. All authors read and approved the final manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (NO. 32501951), the “Shanxi Province Basic Research Program - Free Exploration Category” (NO. 202403021222015), and “Supported by Scientific and Technological Innovation Programs of Higher Education Institutions in Shanxi” (NO. 2024L006).

Data availability

The raw sequence reads are available for download from the China National Center for Bioinformation (CNCB) Genome Sequence Archive (GSA) database (CRA032486).

Declarations

Ethics approval and consent to participate

The experiments did not involve endangered or protected species. The data collection of plants was carried out with permission of related institutions, and complied with national or international guidelines and legislation.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.