Genome-wide identification of long non-coding RNAs revealing their regulatory roles under seed development in ricebean (Vigna umbellata)

Abstract

Background

Long non-coding RNAs (lncRNAs) are crucial components of regulatory networks that govern plant growth, development, and various stress responses. Vigna umbellata (ricebean) is a nutritious and underutilized legume with significant potential to enhance sustainable food security. However, the lack of characterization of lncRNAs and gene expression patterns in ricebean limits understanding of its regulatory mechanisms.

Results

High-throughput RNA-seq was used to identify high-confidence lncRNAs in contrasting genotypes of ricebean at 5 and 10 days post-anthesis (DPA). A total of 3,369 potential lncRNAs were identified. Among these, 955 were differentially expressed lncRNAs (DE-lncRNAs), which either cis-regulate or trans-regulate the expression of adjacent or distant genes. Further, 151 lncRNAs were acting as putative precursors of 228 known miRNAs, targeting 475 protein-coding genes. The competing endogenous RNA network analysis revealed that 8 lncRNAs interacted with 19 miRNAs and competed for 446 mRNAs. We also identified endogenous target mimics (eTMs) and found that 78 lncRNAs interacted with 59 miRNAs, forming a total of 151 lncRNA–miRNA mimicry interactions. Functional enrichment analysis confirmed the involvement of these lncRNAs in seed development, having functions like DAR-1 protein, ubiquitin-conjugating enzymes, LEUNIG_HOMOLOG (LUH), Auxin-responsive protein IAA8, Expansin-like B1 (EXLB1), and Nuclear Transcription Factor Y. The target genes enrichment analysis showed that these genes were involved in epigenetic regulation and the control of gene expression. Notably, several targets belonged to key transcription factor families, including SQUAMOSA PROMOTER-BINDING-LIKE (SPL), ethylene-responsive (RAP2), and MYB. In addition, they were enriched in signalling pathways including jasmonate, brassinosteroid, Gibberellin (GA), abscisic acid (ABA), Auxin, and MAPK, which play a crucial role in seed development, germination, and stress response.

Conclusions

This study reports the first identification of long non-coding RNAs in ricebean. It provides new insights into their possible regulatory roles during seed development. The findings improve our understanding of the molecular processes involved in seed development in ricebean. They also lay the groundwork for future studies on function and comparison in legume crop improvement.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-026-08820-3.

Background

Grain legumes and pulses are vital for global dietary needs, especially in developing countries, as they supply protein and important nutrients. Legumes are a primary protein source for rural communities, underscoring their significance in sustainable diets and eating habits. Besides the usual legumes, less common varieties such as ricebean have begun to gain recognition for their significant nutritional benefits and high yield [1].

Ricebean (Vigna umbellata), commonly known as thumb bean, is an underutilized multipurpose legume belonging to the Fabaceae family with high nutritional value [2]. The genus Vigna includes several economically important species, such as Vigna angularis (adzuki bean), Vigna unguiculata (cowpea), and Vigna radiata (mungbean) [3]. Ricebean is mainly cultivated in Nepal and Northeast India and is also grown in Myanmar, Northern Thailand, Bhutan, Vietnam, Southern China, Laos, East Timor, and Indonesia, typically at elevations of 700–1300 m above sea level [4]. It serves as an important dietary protein source and contributes to nutritional security [5]. The ricebean genome size is approximately 440 Mb, and it is a warm-season annual diploid legume (2n) [2, 6]. Its seeds contain about 25.57% protein and are rich in fiber, carbohydrates, vitamins, minerals, and unsaturated fatty acids such as linolenic and linoleic acids [1, 7].

Although the central dogma states that DNA is transcribed to mRNA, genomes also transcribe abundant non-coding RNAs (ncRNAs) like microRNAs (miRNAs) (22–33 nt) and lncRNAs (> 200 nt) — that regulate gene expression without being translated [8]. Transcriptome analysis indicates that 90% of the genome undergoes transcription into RNA. However, only 2% of these transcripts encode proteins, while the remaining are sRNAs or lncRNAs. Small and long non-coding RNAs are expressed in both tissues and developmental stages, specifically [9–11]. The lncRNAs appear to control expression at the chromatin, transcriptional, and post-transcriptional levels, forming a complex regulatory network that includes pathways regulated by methylation [12].

Long non-coding RNAs are longer than 200 nucleotides in length and do not encode proteins. They are expressed in specific tissues during plant developmental stages or after biotic and abiotic stresses [11]. LncRNAs are crucial in numerous biological functions, such as cell growth, proliferation, intracellular actions, developmental activities, pathways for disease resistance, and regulating transcription. They do not possess open reading frames (ORFs) and might be associated with regulating target gene expression [13]. LncRNAs are generated by RNA polymerase II, much like genes that code for proteins. Nevertheless, in plants, these can also be transcribed by RNA polymerases IV and V [14] and also have cis and trans-acting sites. In many plant species, genome-wide identification of lncRNAs has been reported, such as Arabidopsis [15], peanut [13], cluster beans [16], chickpea [17], cucumber [18], populus [19], pigeonpea [11], soybean [20], sunflower [10], jatropha [21], etc. In recent years, the lncRNAs from seeds have been identified in many plants, including maize [22], Brassica napus [23], castor bean [24], pigeonpea [11], tree peony [25], Ginkgo biloba [26], Jatropha curcas [21], and rice [27]. These lncRNAs might play a complex regulatory role in seed development. For the identification of tissue/cell-specific lncRNAs, transcriptomic data is useful. For example, in peanuts, 50,873 lncRNAs were found from the 124-sample data involving 15 different tissues (flowers, seeds, pods, leaves, fruits, etc.) [28]. Additionally, in developing seeds of Jatropha curcas, 1,850 novel lncRNAs were discovered across three different stages, of which 196 were differentially expressed during the seed’s development [21]. In Brassica napus and tree peony developing seeds, lncRNAs probably affect lipid metabolism [23, 25].

LncRNAs influence gene expression through various pathways or mechanisms, which include modulating translation, altering chromatin structure, or functioning as molecular sponges for miRNAs. Conversely, miRNAs control gene expression by repressing transcription, degrading mRNAs, or restricting mRNA translation.

MicroRNAs play a crucial role in plant development. These are small non-coding RNAs having 20-24nt in length and are transcribed by the polymerase II into primary miRNAs (pri-miRNA) [29]. This distinct group of non-coding RNAs plays a significant role in various biological processes. These processes include signal transduction, growth, development, hormone regulation, homeostasis, innate immune responses, and the responses of plants to various abiotic and biotic stresses [30]. They primarily regulate gene expression by either degrading the target mRNA or preventing its translation. Many studies have been done that lncRNAs can act as precursors of miRNAs, like in Jatropha curcas [21], cucumbers [8], black pepper [31], pearl millet [32] etc., Extensive research is being conducted on ricebean due to its high nutritional value and ability to grow easily, leading us to the identification of lncRNAs in this legume.

Previous studies have demonstrated that lncRNAs are involved in developmental and metabolic processes by functioning as endogenous target mimics and can bind to specific miRNAs. These findings underscore the significant regulatory roles of lncRNAs in seed development [11, 22].

Despite recent progress in elucidating protein-coding genes and miRNAs in Vigna umbellata, the roles of lncRNAs in this species remain largely uncharacterized. Given the low conservation of lncRNAs among plant species and the ambiguity surrounding their functions in seed development, identifying novel lncRNAs and investigating their roles in ricebean is crucial.

This study identified lncRNAs and analyzed their differential expression in rice bean. High-throughput sequencing was performed at different developmental stages, and functional annotation of differentially expressed lncRNA targets was undertaken to infer their possible roles in seed development. Also, the ceRNA network between lncRNA-miRNA-mRNA has been studied.

Materials and methods

Transcriptome data and lncRNA identification pipeline

A total of 12 raw RNA-seq datasets of two contrasting genotypes of ricebean (Vigna umbellata), i.e., IC426787 (bold-seeded) and IC552985 (small-seeded), were downloaded from NCBI (https://www.ncbi.nlm.nih.gov/), linked to BioProject ID PRJNA765494 [6]. The data were collected at two crucial stages of seed development: 5 days post-anthesis (DPA) and 10 DPA. Table 1 shows the SRA IDs and the replicates of the bold and small-seeded data collected at 5 DPA and 10 DPA.

Table 1.

SRA IDs and the replicates of the bold and small-seeded data used in this study

Days Post Anthesis	Replicates	Bold-seeded genotype (IC426787)	Small-seeded genotype (IC552985)
5 DPA	Replicate 1 Replicate 2 Replicate 3	SRR16122599 SRR16122608 SRR16122607	SRR16122605 SRR16122604 SRR16122603
10 DPA	Replicate 1 Replicate 2 Replicate 3	SRR16122602 SRR16122601 SRR16122600	SRR16122610 SRR16122609 SRR16122606

The SRA files obtained from NCBI were first converted into FASTQ format using the SRA Toolkit. Further, the raw data were subjected to quality control using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/), followed by adapter trimming and the removal of low-quality bases using Trimmomatic v0.36 [33] with standard parameters. High-quality paired reads were selected for further analysis. The cleaned reads were aligned to the Vigna umbellata reference genome (414 Mbp in size, containing 31,276 genes) using BWA-MEM [34]. The SAM files obtained after the reads alignment were converted to BAM format using SAMtools. The transcriptome assembly of the individual BAM files was performed using the StringTie software to give out GTF files for each of the reads [35]. Then, a single GTF file from each sample was created using the stringtie --merge command.

Prediction of lncRNAs

All merged transcripts were subsequently processed using Plant-LncPipe [36], a pipeline designed exclusively for identifying lncRNAs. Initially, the transcripts shorter than 200 bp in length were discarded because of their low functional significance, small RNAs or other non-functional transcriptional products, and transcripts longer than 200 nucleotides were retained with no protein-coding potential as candidate lncRNAs. The transcripts having FPKM (fragments per kilobase of transcript per million mapped reads) more than 1 were retained for further downstream analysis. The coding potential of the transcripts was predicted using CPAT, CP > = 0.46 indicates coding sequence, and CP < 0.46 indicates non-coding sequence. To improve the identification of lncRNAs, we included filtering from the Clusters of Orthologous Groups of proteins (COG), Pfam, and Rfam databases in the default pipeline. BLASTX was used against the COG and Pfam databases to exclude protein-coding transcripts, while BLASTN was used against the Rfam database to eliminate non-coding RNAs such as tRNAs, snRNAs, and other small non-coding RNAs. The pipeline used for the identification of lncRNAs and their miRNA precursors in V. umbellata is shown in Fig. 1.

Fig. 1.

Schematic flowchart representing the pipeline used for the identification of lncRNAs and their miRNA precursors in ricebean

Additionally, the results obtained from CPAT were used to preserve transcripts that contain open reading frames (ORFs) shorter than 100 amino acids. The final set of high-confidence lncRNAs was identified using R to intersect the outcomes from all filtering processes. Based on strand information and genomic locations, we identified lncRNAs as intergenic, intronic, NATs, sense, and antisense.

Prediction of mRNAs targeted by lncRNAs

LncRNAs regulate specific genes through two main mechanisms, including cis and trans regulation. Cis-targets were identified within a 100 kb region of lncRNAs. Trans-targets refer to mRNAs that show sequence complementarity with lncRNAs, but are produced by genes situated far from the locations of the lncRNAs. For cis-acting targets, we used bedtools window, and for trans-targets, RIblast [37] (hybridization energy ≤ -30 and interaction energy ≤ -20) was used to find potential target sequences that matched the lncRNAs [8, 38].

LncRNA homologs in other species

Long non-coding RNAs are known to be very poorly conserved among species, so a homology search was performed for the identified DE-lncRNAs that were found in ricebean. This search aimed to examine their sequence conservation across various plant species. The search used the CANTATAdb 3.0 database [39], which is a large collection of lncRNA annotations from 108 plant species. To ensure reliable matches, a strict threshold e-value of 0.001 was applied during the sequence alignment process.

Prediction of lncRNAs-derived miRNA

LncRNAs usually show low sequence conservation across different species, while miRNAs are highly conserved in plants. miRNAs control gene expression by binding to target mRNAs, which suppresses the expression. These small RNAs originate from longer pri-miRNAs, which are processed into mature miRNAs. This means that the predicted ricebean lncRNAs might also serve as precursors of miRNAs (pre-miRNAs).

To explore this, we retrieved all known mature and precursor miRNA sequences from the sRNAanno database [40]. We used the identified lncRNAs as query sequences for miRNA prediction. The downloaded miRNA and pre-miRNA sequences were aligned with the lncRNAs using BLASTn with the following parameters: word size = 7, E-value < 0.001.

Their secondary structures were then compared with the corresponding miRNA precursors using RNAfold (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi) for validation. Furthermore, we characterized the functions of the mRNAs linked to these miRNAs using Omicsbox, which sorted them into relevant biological processes, molecular functions, and cellular components.

Identification of miRNAs targeting the lncRNAs and their target mRNAs

LncRNAs can also act as target mimics of miRNAs, imitating the corresponding mRNAs. To predict miRNA targets, we analyzed all identified miRNA sequences with the psRNATarget tool, with an expectation less than or equal to 2 and unpaired energy (UPE) less than 25. For target prediction, we used the Vigna umbellata lncRNAs sequences as the target query set against the predicted miRNAs. Similarly, we identified potential lncRNA–mRNA interactions using the same method.

Identification of lncRNAs as candidate Endogenous Target Mimics

Endogenous target mimics are particular lncRNAs that interact with miRNAs via incorrect base pairing. The criteria for recognizing eTMs include a single three-nucleotide bulge between the miRNA’s 9th and 12th positions, as well as perfect pairing at the 2nd to 8th nucleotides. Furthermore, there should be no more than three total mismatches and G/U pairs, excluding the central bulge [41]. The psRobot program was used to predict putative eTMs [42].

Gene ontology annotation of DE-lncRNAs, DE-lncRNAs targeted mRNAs, miRNAs targeted mRNAs

Gene Ontology (GO) analysis was performed to functionally characterize the identified transcripts in terms of biological processes (BP), molecular functions (MF), and cellular components (CC). Because lncRNAs do not encode proteins, their functions remain largely uncharacterized. In our study, DE-lncRNAs were indirectly annotated based on sequence similarity to existing protein entries, rather than through direct functional assignments. Functional annotation was performed using sequence similarity searches against the UniProtKB Viridiplantae database with the OmicsBox (Blast2GO) pipeline, which also facilitated GO and KEGG pathway analyses of miRNA and lncRNA-targeted mRNAs, enabling interpretation of their roles in cellular and developmental processes.

Results and discussion

Identification of long non-coding RNAs in ricebean

To elucidate the landscape and expression profile of lncRNAs in ricebean, 12 RNA-seq libraries (BioProject ID PRJNA765494) generated from total RNA of two contrasting genotypes, IC426787 (bold-seeded) and IC552985 (small-seeded), at two developmental stages: 5 DPA and 10 DPA, with three biological replicates for each stage, were used. We obtained 120.6 GB of clean data with a Q30 score of more than 95% after removing the adapters and bad-quality reads.

We used the Burrows-Wheeler Aligner (BWA) to map the clean reads to the Vigna umbellata reference genome after trimming. All samples had a mapping rate of more than 95% to the reference genome.

Transcripts were initially filtered based on sequence length and exon number, resulting in the retention of 46,662 transcripts. LncRNAs were identified using the Plant-LncPipe, integrating Plant LncFinder and the CPAT model to assess coding potential. Transcripts with open reading frames (ORFs) < 100 nt and CPAT coding probability < 0.46 were retained, yielding 7,332 putative lncRNA candidates. To exclude protein-coding transcripts, BLASTx searches were performed against UniProt, COG, and Pfam databases. Additionally, housekeeping non-coding RNAs, including rRNAs and tRNAs, and others, those in the Rfam database, were removed through BLASTn analysis. The filtered transcript sets were subsequently intersected using R, resulting in the identification of 3,369 high-confidence lncRNAs. The number of identified lncRNAs is comparable to that reported for other members of the Fabaceae family, including pigeonpea [11] and soybeans [20]. Whereas, some plants have been found to contain significantly more lncRNAs as compared to ricebean, such as tomatoes [43], pearl millet [32], populus [44], and Arabidopsis thaliana [15] while some plants have a lower number of lncRNAs, such as cucumber [8] and peach [45] (Supplementary Table 1). Genome-wide chromosomal mapping of the 3,369 identified ricebean lncRNAs (as shown in Fig. 2) revealed a non-uniform distribution, with chromosome 1 exhibiting the greatest density (688 lncRNAs), while chromosome 10 encoded the fewest (124). Though not evenly distributed, the identified ricebean lncRNAs were found to be distributed across all the chromosomes. Similar trends have been observed in various plants, including pigeonpea [11], Glycine max [20], pearl millet [32], wheat [46], rice [47], and maize [48].

Fig. 2.

Chromosomal distribution of identified 3369 lncRNAs in Ricebean

Characteristic features of lncRNAs

Functional classification of the identified lncRNAs was carried out using the FEELnc-classifier Perl module integrated within the Plant-LncPipe framework, which categorizes lncRNAs into six genomic-context classes. Among these, antisense-exonic lncRNAs were the most prevalent, comprising 1,949 transcripts (57%), reflecting their widespread presence in plant genomes. This dominance aligns with previous observations by Lin et al. (2020) in Glycine max (~ 91%) and Glycine soja (~ 90%) [49], indicating an evolutionarily conserved feature among legumes. The intergenic lncRNAs represented the second largest group, with 528 transcripts (15%), consistent with the proportions reported for G. max (~ 8%) and soybean (~ 9%) [49]. In contrast, upstream, bidirectional, and downstream lncRNAs appeared at lower frequencies, while intronic lncRNAs were relatively rare (Fig. 3). The pronounced enrichment of antisense-exonic lncRNAs in ricebean underscores their potential regulatory roles, possibly mediated through transcriptional interference, RNA–DNA hybrid formation, or stabilization/destabilization of complementary mRNAs.

Fig. 3.

Classification of different types of lncRNAs in Ricebean

In ricebean, the identified lncRNAs varied in size from 200 to 2,045 nucleotides, with an average length of around 403 bp. Similar observations have been reported in other plants, such as chickpea (495 bp) [50], cucumber (322 bp), and rice (800 bp) [51], suggesting that shorter lncRNA lengths are a common feature across plant species. Among the identified lncRNAs, one transcript contained 37 exons, while others had 36, 35, and 29 exons (three transcripts). Despite these few multi-exonic examples, the majority of lncRNAs were mono-exonic [11]. Furthermore, the identified lncRNAs showed a higher AU content than mRNAs, a pattern consistent with observations in various plant species such as Arabidopsis [52], pigeonpea [11], maize [48], rice [51], and populus [19].

Conservation of lncRNAs in other species

LncRNAs are generally poorly conserved across distant taxa [53, 54]. To investigate this in ricebean (Vigna umbellata), we performed BLASTn searches of the identified ricebean lncRNAs against the CANTATAdb 3.0 database [39], which contains 571,688 plant lncRNAs from 108 species. Using an E-value cutoff of 0.001, we found that 328 ricebean lncRNAs had significant similarity to 565 lncRNAs in 32 other species (Table 2). However, with a more stringent identity threshold (≥ 90%), only 247 ricebean lncRNAs (matching 369 lncRNAs) had homologs, and no conserved lncRNAs were detected in the remaining 76 species (Supplementary Table 2).

Table 2.

List of ricebean lncRNA homologs in other plant species from the CANTATA3.0 database

Number of ricebean lncRNAs with homologs in other species	Conservation ratio of ricebean lncRNAs (%)	Total lncRNAs in other species	Other Species
23/31	0.71237756	3189	Cajanus cajan
22/37	0.653012763	4275	Glycine max
23/35	0.682695162	1866	Phaseolus vulgaris
2/2	0.059364797	13,968	Pistacia vera
9/11	0.267141585	6876	Pisum sativum
4/5	0.118729593	5407	Prunus dulcis
3/3	0.089047195	5268	Quercus lobata
3/3	0.089047195	3089	Trifolium pratense
2/3	0.059364797	15,321	Triticum aestivum
179/269	5.313149302	2103	Vigna angularis
91/125	2.701098249	5605	Vigna radiata
6/11	0.17809439	3815	Vigna unguiculata

The strongest conservation was observed within legumes. In particular, 152 ricebean lncRNAs (~ 4.5% of the total) had homologs in adzuki bean (Vigna angularis), and 216 V. angularis lncRNAs matched ricebean loci; this corresponds to the largest overlap. The next highest was mung bean (Vigna radiata), with 91 ricebean lncRNAs (≈ 2.7%) matching 125 mung lncRNAs. These two species are known to be the closest relatives of ricebean [55], explaining the higher lncRNA homology. In contrast, more distantly related plants had very few lncRNA homologs. Notably, detectable conservation was essentially restricted to Fabaceae: for example, we observed similarity with soybean (Glycine max), common bean (Phaseolus vulgaris), pigeon pea (Cajanus cajan), cowpea (Vigna unguiculata), lupin (Lupinus angustifolius), and pea (Pisum sativum) (Fig. 4). This pattern mirrors previous findings that plant lncRNAs are largely lineage-specific. For instance, in soybean, nearly all lncRNAs match its wild progenitor (Glycine soja), but most are absent in more distant legumes [54]. In other words, detectable sequence conservation was observed for only a small fraction of ricebean lncRNAs, primarily limited to closely related species. This pattern aligns with previous findings in soybean, where lncRNAs showed higher sequence conservation with Glycine max, Phaseolus vulgaris, and Vigna angularis compared to more distantly related legumes like Medicago truncatula, Lotus japonicus, and Cicer arietinum [49].

Fig. 4.

Synteny plot representing the homologs of identified lncRNAs of ricebean and the other members of the Fabaceae family

Identification and functional annotation of differentially expressed lncRNAs in ricebean during seed development

Plant lncRNAs play vital roles in regulating various aspects of plant growth and development. To explore their functional significance in ricebean, lncRNA expression profiles were examined in two contrasting seed types, small and bold, at two developmental stages. Expression levels, quantified as FPKM values, showed comparable distributions across samples. Among the 3,369 identified lncRNAs, about 28% (955 lncRNAs) were differentially expressed (p ≤ 0.05, fold-change ≥ ± 2). To compare expression differences between developmental stages, the variation in lncRNA expression was represented as the log₂ fold change between the 5 and 10 DPA stages. In the 955 DE-lncRNAs, highly DE-lncRNAs were observed in S10_vs_S5 (576: 214 down-regulated, 362 up-regulated), followed by B10_vs_S10 (442: 260 up-regulated and 182 down-regulated), B10_vs_B5 (398: 188 up-regulated and 210 down-regulated), and B5_vs_S5 (384: 175 up-regulated, 209 down-regulated), with many DE-lncRNAs common between them (Fig. 5a and 5b). Interestingly, more DE-lncRNAs were observed during later stages (10 DPA) of seed development as compared to the early stages (5 DPA) of seed development. Detailed expression pattern of DE-lncRNAs between 5 and 10 DPA is provided in the Supplementary Table 3. Their expression dynamics demonstrated a fold-change distribution ranging from + 14 to − 13.

Fig. 5.

a Graph showing up-regulated and down-regulated DE-lncRNAs in ricebean; (b) Venn diagram showing Differentially expressed lncRNAs (Source: https://bioinfogp.cnb.csic.es/tools/venny/)

GO annotations of DE-lncRNAs were inferred indirectly based on sequence similarity to previously annotated protein-coding genes [56]. The annotations revealed associations with key regulators of plant development. Notably, they included auxin-responsive IAA8, which in Arabidopsis acts as an auxin signalling repressor that promotes seed germination by inhibiting ABI3 transcription [57], and CYP714A1, a gibberellin/brassinosteroid-related gene functioning redundantly in plant development through gibberellin deactivation in Arabidopsis [58]. Transcription factors such as MADS-box and NF-YA are both known to play significant roles in seed development [59, 60]. Additionally, UNE10 is associated with seed desiccation sensitivity in Quercus [61], while LEUNIG positively regulates seed growth [62]. Collectively, these genes are known to influence embryo and seed maturation, highlighting their involvement in hormone-mediated signalling during seed and pod development [63]. Cellulose synthase-like proteins and UDP-glucose 4-epimerase contribute to cell wall biosynthesis. In contrast, lipid metabolism-related genes such as OBL1 and ABC1K7 exert positive regulatory effects under various abiotic stress conditions via the ABA signalling pathway [64]. The cation/proton exchanger CAX1 in Arabidopsis is crucial for maintaining metal ion homeostasis, particularly calcium storage and distribution, during seed development [65]. Together, these findings suggest that DE-lncRNAs regulate multiple pathways vital for seed and pod development in legumes. The corresponding heatmap (Fig. 6) depicts the expression patterns of the annotated DE-lncRNAs. Additionally, we studied the comparative expression of DE-lncRNAs during early and later developmental stages of small and bold seeds of ricebean. For example, lncRNA MSTRG.21031.1 in small and bold seeds shows higher expression in early stages than in later developmental stages, and it is annotated for Vicianin hydrolase. This enzyme plays an indirect but important role in seed and pod development by providing protection to the developing seeds from predators. It is highly expressed in seeds, where it breaks down a compound called vicianin to release hydrogen cyanide (HCN), a toxic substance [66]. This reaction acts as a chemical defence mechanism—like a hidden trap—discouraging insects or animals from eating the seeds or pods. Another lncRNA, MSTRG.8317.1, has lower expression in early stages than in later developmental stages in both bold and small seeds. This lncRNA was mapped to EXLB1, a cell wall–loosening protein associated with root architecture modulation, photosynthetic efficiency, drought resilience, and seed germination promotion [67]. Another lncRNA, namely, MSTRG.7479.1, a Ubiquitin-conjugating enzyme E2, showed lower expression in early stages and higher expression in later stages. The ubiquitin-conjugating enzyme (E2) plays a crucial role in transferring ubiquitin from the ubiquitin-activating enzyme (E1) to the ubiquitin-ligase enzyme (E3). It is mainly involved in regulating various cellular functions, including responses to stress (like heat, drought, and pathogens), hormonal signalling (such as ABA), and development [68]. In small seeds, there is no expression of lncRNA MSTRG.34999.1, but in bold type seeds, it shows higher expression. This lncRNA was mapped to the protein TIC236, which maintains the normal morphology and function of proplastids and promotes the differentiation of proplastids to chloroplasts during early embryo development. The study of TIC236 in Arabidopsis shows that disruption of TIC236 can cause seed abortion [69]. LncRNA-MSTRG.1571.1, associated with Nuclear-transcription factor Y (NFYA1), shows strong upregulation at the late stage and higher expression in the small-seed line. This is consistent with the known role of NF-YA transcription factors in promoting seed maturation, storage reserve accumulation, and embryo developmental progression [60]. Lnc-MSTRG.42963.1 mapped to transcription factor E2FA, which was strongly upregulated at the later stage and shows the highest expression in the small-seed line. This TF coordinates cell proliferation with differentiation and the accumulation of seed reserves during seed and embryo development [70].

Fig. 6.

Heatmap showing expression patterns of the annotated DE-lncRNAs

Prediction of cis- and trans-acting mRNA Targets of lncRNAs

Long non-coding RNAs can modulate gene expression by sequestering gene-specific microRNAs or by recruiting regulatory elements to the 5’ untranslated regions (UTRs) of target genes. Their distinctive secondary structures offer binding sites for miRNAs, mRNAs, transcription factors, and other regulators, facilitating either the activation or suppression of gene expression [11]. LncRNAs regulate target genes through two primary modes: cis and trans [21]. In the cis mode, lncRNAs influence the expression of neighbouring genes located within a 10 kb window upstream or downstream of their genomic loci, and such genes are considered cis-targets. In contrast, trans-targets are mRNAs encoded by genes located outside the vicinity of the lncRNA locus, but which share sequence complementarity with the lncRNA [11, 21]. These interactions are typically predicted based on base-pairing potential and free energy of hybridization. A total of 328 cis-target genes were identified from the 298 DE-lncRNAs and 24,214 trans-targets from 683 DE-lncRNAs, the DE-lncRNAs averaging 35 targets each.

Comparative GO and KEGG pathway analysis of mRNA targets of lncRNAs

We performed GO and KEGG enrichment analysis of the identified cis and trans-targets to explore their roles in the growth and development of seeds in ricebean. The GO terms of the cis and trans targets of DE-lncRNAs were enriched across all GO categories: cellular component, biological process, and molecular function.

Several target genes were found that belong to well-known transcription factor families, including MYB, WRKY, bHLH, GATA, TF-Y, and TFIID. Previous studies have described the roles of these transcription factors in seed and pod development. For example, in rice, MYB transcription factors play vital regulatory roles in seed germination and stress responses during seedling establishment [71]. Members of the WRKY family are associated with seed germination [72], while ethylene-responsive transcription factors contribute to iron storage during seed maturation [73]. The bHLH transcription factors are involved in photosynthesis, light signaling, pigment biosynthesis, and seed pod development [74]. Similarly, GATA transcription factors act as positive regulators of seed germination in Arabidopsis [75], nuclear TF-Y [60], and TFIID participate in the early stages of gene expression by recognizing promoters of protein-coding genes [76]. Collectively, these transcription factors have been reported to play crucial roles in plant embryo and seed development [77].

Among the identified cis-acting targets, approximately 80–85% was associated with metabolic processes, whereas 10–15% was involved in gene regulation, including transcription factors and RNA-binding proteins. A smaller fraction, about 5–10%, was linked to developmental processes such as embryogenesis and cell differentiation. In contrast, the trans-acting targets exhibited a more balanced distribution, with around 40% related to gene expression regulation, 20–25% involved in metabolic pathways, and 10–15% associated with developmental functions. The remaining targets were categorized under general or uncharacterized functions. Cis-targets were mainly found to encode enzymes and metabolic proteins, whereas trans-targets included a wide range of transcriptional regulators. The cis-targets often consisted of DNA or RNA-binding proteins and kinases, while the trans-targets were largely represented by enzymatic proteins such as terpene synthases, dehydrogenases, and UDP-glycosyltransferases. This pattern aligns with findings in other plant species; for instance, Arabidopsis lncRNAs like COOLAIR and COLDAIR interact with transcriptional regulators [78], whereas many oilseed lncRNAs co-express with metabolic enzymes (e.g., fatty acid biosynthesis genes) [79] Fig. 7.

Fig. 7.

GO annotations and KEGG pathway of cis-targeted mRNAs: a Biological Process, (b) Cellular Component, (c) Molecular Function, (d) KEGG Pathway

The functional characterization of cis and trans-targets of differentially expressed lncRNAs revealed their putative involvement in multiple developmental and physiological processes. These include embryo dormancy, root and lateral root formation, organ morphogenesis, guard cell differentiation, endosperm development, and seed and pod morphogenesis [77]. A substantial proportion of these targets were enriched in transcriptional regulation and phytohormone signaling pathways. These included pathways involving cytokinin, ABA, and auxin, as well as gibberellin biosynthesis and auxin-responsive F-box proteins. Implying that lncRNAs may modulate regulatory networks governing cellular differentiation, embryo patterning, and pod elongation.

Moreover, several targets exhibited strong enrichment in primary and secondary metabolic pathways, including fatty acid and carbohydrate metabolism, as well as in the biosynthesis of terpenoids, flavonoids, and phenylpropanoids. These associations indicate potential roles in nutrient mobilization, seed filling, and pod defense through the accumulation of secondary metabolites. A considerable number of targets encoded catalytic proteins, including glycosyltransferases, GDSL lipases, glycerol-3-phosphate 2-O-acetyltransferases, terpene synthases, oxidoreductases, F-box-LRR repeat proteins, serine/threonine-protein kinases, and B-box zinc finger proteins. Suggesting active participation in cell wall remodeling, lipid metabolism, and stress adaptation during seed maturation [11, 80]. These findings are concordant with previous reports in oilseed species such as Jatropha curcas [21], Brassica napus [79], and castor bean [24], wherein lncRNAs have been demonstrated to regulate lipid biosynthesis and seed development.

In addition to developmental functions, stress-responsive processes were prominently represented, with targets linked to responses against heat, cold, drought, salinity, oxidative stress, wounding, and diverse pathogenic challenges—underscoring their broad involvement in plant defense mechanisms. Metabolic functions encompass pathways related to the biosynthesis of amino acids, carbohydrates, lipids, and secondary metabolites. They also include targets associated with the transport of ions, sugars, and metabolites, including ABC transporters and sodium/potassium/zinc transporters. Furthermore, several targets were implicated in cell wall remodeling and in the biosynthetic pathways of lignin and flavonoids, contributing to structural stability and environmental resilience. Collectively, these findings suggest that lncRNA–mRNA interactions coordinate developmental regulation, metabolic reprogramming, and stress adaptation, thereby enhancing plant growth, reproduction, and ecological fitness.

KEGG pathway analysis revealed that the majority of cis-target genes were enriched in oxidative phosphorylation, glycolysis/gluconeogenesis, and BR signaling pathways. BRs, a class of plant steroid hormones, are known to regulate organ size [81] and participate in the metabolism of fructose, mannose, purine, and thiamine. These findings suggest that differentially expressed lncRNAs modulate genes involved in energy metabolism, potentially influencing seed development and carbohydrate biosynthesis.

Similarly, KEGG analysis of trans-targeted protein-coding genes demonstrated enrichment in pathways associated with seed size regulation, seed germination, the tricarboxylic acid (TCA) cycle, ABA-mediated signaling, brassinosteroid and MAPK signaling, ubiquinol biosynthesis, gibberellin biosynthesis, the ubiquitin–proteasome system, gluconeogenesis, and glycolysis. Notably, gibberellin biosynthesis plays a central role in seed germination, stem elongation, leaf development, and floral induction [82], while the ubiquitin–proteasome pathway regulates seed development by modulating cell growth [81]. These results collectively highlight the multifaceted roles of lncRNAs in integrating metabolic and hormonal signaling networks governing seed and plant development Fig. 8.

Fig. 8.

GO annotations and KEGG pathway of trans-targeted mRNAs: a Biological Process, (b) Cellular Component, (c) Molecular Function, (d) KEGG Pathway

Identified ricebean lncRNAs acting as miRNA precursors

LncRNAs can function as precursors of microRNAs (miRNAs; 18–35 nucleotides). To identify ricebean lncRNAs potentially serving as miRNA precursors, a BLASTn search (e-value ≤ 0.001, word size 7) was conducted against the sRNAanno database, which contains miRNA sequences from 143 plant species. A total of 151 lncRNAs were identified as putative precursors of 228 known miRNAs, sharing more than 90% sequence identity and collectively targeting 475 protein-coding genes. Some lncRNAs were predicted to generate a single miRNA (e.g., miR169a, miR156h, miR396e, miR319a, miR319c, miRN29c), whereas others—such as MSTRG.11585.1, MSTRG.14783.1, MSTRG.16019.1, MSTRG.1195.1, MSTRG.1605.3, MSTRG.16119.1, MSTRG.1627.1, MSTRG.17644.1, MSTRG.18028.1, MSTRG.26683.1, MSTRG.27796.1, and MSTRG.9006.1—were predicted to serve as precursors for multiple miRNAs. Chromosome 1 was identified as the major locus harboring lncRNAs with miRNA precursor potential (Supplementary Table 4).

To further validate these findings, the identified mature miRNAs were aligned with the lncRNA sequences, revealing 28 lncRNAs that acted as confirmed pre-miRNA precursors with > 90% sequence identity to mature miRNAs. Secondary structures of these lncRNAs and their corresponding miRNA precursors were predicted using the RNAfold web server. For instance, the secondary structure of MSTRG.26683.1 displayed multiple stem-loop formations, one of which serves as the precursor for miR482b of Phaseolus vulgaris (Fig. 9).

Fig. 9.

Multiple stem-loop structure of lncRNA MSTRG.26683.1, which serves as the precursor for miR482b

Functional enrichment analysis and KEGG pathway of miRNA–mRNA target pairs was performed (Fig. 10) to elucidate the biological roles of lncRNA-derived miRNAs. The results revealed that several targets were associated with seed and pod development. For example, bHLH transcription factors were linked to the negative regulation of oil accumulation in Arabidopsis seeds, thereby influencing seed reserve formation [83]. WRKY transcription factors were implicated in seed development and in mediating physiological responses to biotic and abiotic stressors [84]. Additionally, Nuclear Transcription Factor Y and F-box proteins were connected to AUXIN RESPONSE FACTOR 8 (ARF8), which is known to uncouple fruit development from pollination and fertilization, leading to seedless (parthenocarpic) fruit formation [85]. ABSCISIC ACID-INSENSITIVE 4 (ABI4) was found to play a critical role in ABA signaling during seed development and germination [86]. Several targets mapped to the APETALA2 (AP2) domain, a negative regulator of seed size repressed by BR activation [87]. These findings indicate that ricebean lncRNAs can act as miRNA precursors that regulate epigenetic and transcriptional processes, contribute to seed and pod development, and enhance adaptation to environmental stresses.

Fig. 10.

GO annotation and KEGG pathway of miRNA-targeted mRNAs: a Biological Process (b) Cellular Component (c) Molecular Function (d) KEGG Pathway

Identification of Endogenous Target Mimics and analysis of lncRNA–miRNA–mRNA interaction networks

MicroRNAs are short, non-coding RNA molecules typically ranging from 18 to 23 nucleotides in length [30]. They primarily regulate gene expression by binding to the 3′ untranslated regions (3′ UTRs) of protein-coding mRNAs. When miRNAs exhibit perfect complementarity with their target mRNAs, they induce mRNA cleavage and gene silencing; in contrast, partial complementarity suppresses translation and decreases gene expression [88].

Long non-coding RNAs can modulate this regulation by functioning as miRNA sponges, sequestering miRNAs and preventing them from binding to their target mRNAs [89]. Some lncRNAs act as eTMs, competing with mRNAs for miRNA binding and thereby relieving miRNA-mediated repression [11]. To explore this regulatory complexity in ricebean, the identified lncRNAs were analyzed as potential miRNA targets for identified miRNAs using the psRNATarget server [90]. Predictions were performed with default parameters, setting a maximum unpairing energy (UPE) of 25 and an expectation value below 2 [31]. The analysis revealed potential interactions among miRNAs, mRNAs, and lncRNAs. Specifically, 31 miRNA–lncRNA interactions were identified, involving 20 miRNAs targeting 14 lncRNAs. In contrast, 1,287 miRNA–mRNA interactions were predicted, where 220 miRNAs targeted 475 mRNAs. Within the ceRNA network, 19 miRNAs were found to be shared between 8 lncRNAs and 446 mRNAs (Supplementary Table 5).

These targets were enriched in pathways associated with auxin transport, which controls seed and pod growth patterning [91], and in hormone signal integration (ABA/Ethylene/BR/GA), regulating seed dormancy and growth [92]. They were also involved in transcriptional regulation of seed and pod organ formation. Notably, TCP3, a key regulator of leaf and fruit shape, modulates silique elongation in Brassicaceae [93] through miRNA-mediated regulation linked to seed shattering [93]. Other targets included NF-YB3, influencing seed size and nutrient loading [94], as well as chromatin modifiers such as SWI3C and SDG41, which regulate gene expression timing during seed maturation [95, 96]. Additionally, these targets were implicated in stress response and redox homeostasis during seed development.

Furthermore, the prediction of eTMs was performed using psRobot [42], which identified 78 lncRNAs interacting with 59 miRNAs, forming a total of 151 eTMs (lncRNA–miRNA mimicry interactions). These findings reinforce the hypothesis that lncRNAs can indirectly regulate mRNA expression by functioning as eTMs, sequestering miRNAs and thereby modulating their activity (Supplementary Table 6). Some ETM examples are shown in Fig. 11a.

Fig. 11.

a eTM binding sites for lncRNAs and miRNAs. b CeRNA networks of lncRNA-miRNA-mRNA. Interaction networks of miRNA-lncRNA and miRNA-mRNA. Peach colour eclipses are mRNAs; yellow diamonds are miRNAs, and orange colour eclipses are of lncRNAs

Comparable regulatory mechanisms have been documented in earlier studies. For instance, a degradome-based analysis in maize by Fan et al. [97] identified 34 lncRNAs acting as eTMs with binding sites for 33 miRNAs. The expression of these lncRNAs disrupted miRNA–mRNA interactions, impacting biological processes such as hormone signaling, sugar transport, and fatty acid and starch biosynthesis [11].

Notably, several lncRNAs were found to act as eTMs for members of the same miRNA family across different species. For example, MSTRG.16696.1, MSTRG.20765.1, MSTRG.25555.1, MSTRG.40506.1, and MSTRG.4665.1 were targeted by miRN1 (a–e); MSTRG.14167.1, MSTRG.14783.1, MSTRG.21789.1, MSTRG.23919.1, MSTRG.24352.1, MSTRG.26247.1, MSTRG.42377.1, MSTRG.42437.1, and MSTRG.8680.1 were targeted by miR156 family members; and MSTRG.10971.1, MSTRG.135.1, MSTRG.21002.1, and MSTRG.32642.1 were targeted by miR65, among others. Similar patterns of eTM–miRNA interactions have also been reported in pigeonpea [11] and cluster bean [16], suggesting a conserved regulatory mechanism among legumes.

The lncRNA–miRNA–mRNA interaction network was constructed and visualized using Cytoscape software [98], illustrating the complex interplay between lncRNAs and their target genes (Fig. 11b).

Role of lncRNAs in seed and pod development

Our analysis revealed that several lncRNAs are potentially implicated in the molecular regulation of seed and pod development. For instance, lncRNA MSTRG.25920.2 was aligned to Sterol 3-beta-glucosyltransferase (UGT80A2), an enzyme responsible for the biosynthesis of sitosteryl and stigmasteryl glucosides, which are essential components of seed lipid metabolism and membrane stabilization [99]. MSTRG.8317.1 was mapped to EXLB1, a cell wall–loosening protein associated with root architecture modulation, drought resilience, photosynthetic efficiency, and the promotion of seed germination [67]. Similarly, MSTRG.15783.1 was aligned with CAP-D3, a subunit of the Condensin II complex, which plays a crucial role in chromatin compaction, sister chromatid cohesion, fertility, and the maintenance of genomic integrity during development [100]. Three additional lncRNAs—MSTRG.26958.1, MSTRG.26963.1, and MSTRG.6808.1—were predicted to correspond to SNAP33, a t-SNARE component known to be highly expressed in pods, seeds, and roots of Glycine soja. In Arabidopsis thaliana, overexpression of SNAP33 enhances germination efficiency, root elongation, photosynthetic capacity, biomass accumulation, and expression of stress-responsive marker genes [101]. LncRNA MSTRG.47152.1 was mapped to ABSCISIC ACID-INSENSITIVE 5 (ABI5), a bZIP transcription factor involved in abscisic acid signaling. Hyperactivation of ABI5/DPBF genes results in altered ABA responses that tightly regulate seed maturation and germination processes [102]. Likewise, MSTRG.34996.1 was associated with GTP-binding protein At2g22870 (EMB2001), a regulator of embryo development and seed dormancy establishment [103]. MSTRG.1571.1 corresponded to Nuclear Transcription Factor Y (NF-Y), which functions redundantly in male gametophyte development, embryogenesis, seed morphogenesis, and post-germinative growth [60].

Moreover, MSTRG.7815.1 showed similarity to DAR1 (DA1-related protein), known to modulate organ and seed size through ubiquitin-mediated control of cell proliferation [104]. MSTRG.15095.1 was mapped to Auxin-responsive protein IAA8, which accumulates during seed germination and promotes the process by repressing ABI3 transcription within the auxin–ABA signaling nexus [57] (Table 3).

Table 3.

Role of lncRNAs in seed development in ricebean

lncRNA Transcript	Associated Protein-Coding Gene	Developmental Role & Function
MSTRG.25920.2	Sterol 3-beta-glucosyltransferase (UGT80A2 )	Seed lipid metabolism and membrane stabilization by producing sitosteryl and stigmasteryl glucosides, essential for seed components.
MSTRG.8317.1	Expansin-like B1 (EXLB1)	Seed germination promotion root architecture modulation, and general growth (drought resilience, photosynthetic efficiency).
MSTRG.15783.1	CAP-D3 (Condensin II Complex subunit)	Genomic integrity and fertility through chromatin compaction and sister chromatid cohesion during cell division, crucial for development.
MSTRG.26958.1, MSTRG.26963.1, MSTRG.6808.1	SNAP33	Germination efficiency, root elongation, biomass accumulation, and general growth (highly expressed in pods and seeds).
MSTRG.47152.1	ABSCISIC ACID-INSENSITIVE 5 (ABI5)	Seed maturation and germination regulation by controlling ABA signaling responses.
MSTRG.34996.1	GTP-binding protein (EMB2001)	Embryo development and seed dormancy establishment (a critical regulator of developmental timing).
MSTRG.1571.1	Nuclear Transcription Factor Y (NF-Y)	Embryogenesis, seed morphogenesis, male gametophyte development, and post-germinative growth (fundamental developmental processes).
MSTRG.7815.1	DAR1 (DA1-related protein)	Organ and seed size modulation via ubiquitin-mediated control of cell proliferation.
MSTRG.15095.1	Auxin-responsive protein IAA8	Seed germination promotion by repressing the ABA signaling pathway (specifically ABI3) within the auxin-ABA signaling nexus.

Taken together, these findings suggest that a subset of ricebean lncRNAs may act as regulatory intermediates influencing seed and pod morphogenesis. This may occur through cis- or trans-regulatory mechanisms targeting genes involved in hormone signaling, cell proliferation, and developmental timing. The identified lncRNAs correspond to orthologs with experimentally validated functions in model species such as Arabidopsis, Brassica, and Glycine. This underscores the evolutionary conservation of lncRNA-mediated regulatory networks governing seed and pod development in legumes.

Hypothesis: pathway regulation in seed development

Seed development in plants is a complex and highly regulated process that involves a network of hormonal, transcriptional, metabolic, and environmental signaling pathways. Hormonal pathways, including GA, ABA, jasmonic acid, auxin, and cytokinin signaling, are essential to this developmental process. These pathways coordinate the various phases of embryo formation, reserve accumulation, and dormancy [105]. Our analysis identified several developmental pathways, including jasmonic acid signaling, which limits integument proliferation and influences seed size [106]. ABA and GA pathways are involved in plant development as well as the regulation of seed dormancy and germination. Their balance is regulated by specific transporters and transcription factors, including ABI5, ABI3, and DELLA proteins, which ensure that embryo growth and seed coat development proceed correctly and respond to environmental cues such as light and temperature [107]. Protein degradation pathways, particularly the ubiquitin-proteasome system with SCF and APC/C complexes, regulate the cell cycle. They facilitate appropriate cell division and differentiation during seed development [108]. Transcription factors and epigenetic modifications play a crucial role in regulating gene expression and chromatin structure. These processes directly influence nutrient transport, reserve biosynthesis (such as starch), and the progression from embryogenesis to maturation [109, 110]. Metabolic pathways involved in carbohydrate metabolism, such as starch biosynthesis and nitrogen utilization, are interconnected with hormonal signaling to enhance the accumulation and subsequent mobilization of reserves. The brassinosteroid signaling pathway, where BRs are essential plant steroid hormones that influence many aspects of seed development. They regulate seed size, shape, and cellular differentiation by modulating the transcription of key developmental pathways [87]. The activity of microRNAs provides essential post-transcriptional regulation, modulating the expression of genes related to cell division, hormone signaling, and stress responses. Additionally, long non-coding RNAs can disrupt miRNA function, thereby affecting pathway outcomes. These pathways function collectively, allowing the seed to adapt, grow, and germinate under various environmental conditions, thereby ensuring effective seed development and the successful establishment of the subsequent generation.

We identified several early-stage pathways associated with seed development in V. umbellata (Fig. 12), including auxin–IAA, brassinosteroid, and ubiquitin-dependent E3 ligase signalling. The auxin pathway is central to early embryogenesis, as auxin produced in the integuments directs embryo patterning in Arabidopsis [111], and miR393-mediated auxin regulation influences seed germination and seedling establishment in rice under submergence [112, 113]. In our study, miRN1 (a, b, c) and miRN65 targeted the auxin-related genes Vu_16900.1, Vu_16900.2, and Vu_20287.1, along with lncRNAs MSTRG.6908.1, MSTRG.36342.1, MSTRG.135.1, and MSTRG.211481.1, indicating that these lncRNAs may act as miRNA sponges to modulate auxin availability during early seed development. The BR pathway also contributes to early seed growth, with elevated BR levels in the seed coat and endosperm promoting cell expansion and with BR signalling later repressing maturation programs [114]; miR172 has been shown to suppress BR-signalling defects in bak1 mutants [115]. Consistently, we found that miR156 (e, k, y) targeted Vu_17005.1 and also bound lncRNA MSTRG.8332.1, suggesting BR regulation through a miRNA–lncRNA sponge interaction. Ubiquitin-dependent E3 ligase pathways likewise play a fundamental role in early seed size determination [116]; for instance, the DA1–DA2 module regulates seed and organ size in Arabidopsis [117], and the F-box protein TIR1, as part of the SCFᵀᴵᴿ¹ E3 ligase, mediates Aux/IAA degradation to maintain auxin signalling during seed formation [118]. In our dataset, miRN11 targeted an E3 ubiquitin-ligase gene (Vu_23367.1) and the lncRNA MSTRG.27869.1, representing a potential novel miRNA–E3 regulatory module contributing to early seed development in ricebean.

Fig. 12.

Hypothetical pathway depicting the role of lncRNA, miRNA, and mRNA in seed development regulation at early and later stages in ricebean. Eclipse are miRNAs, dashed rectangles are mRNAs, and blue sketched boxes are lncRNAs

During the later stages of seed development, several hormone- and stress-responsive pathways were identified in V. umbellata. The ABA signalling pathway plays a crucial role in establishing seed dormancy and conferring desiccation tolerance through LEA protein accumulation. In addition, miRNAs such as miR398, miR156, and miR159 are known regulators of ABA and drought-responsive genes in other species [119]. In our study, one miRNA targeted the ABA-related gene Vu_09155.1, along with lncRNAs MSTRG.36342.1 and MSTRG.6908.1, suggesting that these lncRNAs may act as sponges for an ABA-repressive miRNA. The jasmonic acid (JA) signalling pathway, which promotes lipid/protein accumulation and defence responses during seed filling, is regulated by miR319 through TCP transcription factors in Arabidopsis [93]. Consistently, we identified three JA-related targets of miRN1 (a, b, c), with lncRNAs MSTRG.6908.1 and MSTRG.36342.1 likely functioning as miRNA sponges. MAPK signalling, which integrates ABA/GA balance and ROS cues during dormancy release and germination, has been shown to involve MKK3 as a key dormancy regulator in cereals [120] and contributes to ROS-mediated signalling during seed imbibition [121]. In our dataset, miRN1 (a, b, c) and miRN65 targeted Vu_01494.1, Vu_15885.1, and Vu_22546.1, along with lncRNAs MSTRG.6908.1, MSTRG.36342.1, MSTRG.135.1, and MSTRG.211481.1, indicating a multilayered MAPK–miRNA–lncRNA regulatory network. Finally, the GA pathway, which promotes dormancy release and germination by degrading DELLA repressors such as RGL2 [122]. This pathway is regulated by miR159 through GAMYB/MYB33/MYB65 [123, 124] and by miR156 through SPL transcription factors [125]. In line with this, we found that miRN65 targeted Vu_19702.1, a GA-related gene, representing a potential miRNA–GA regulatory module during late seed development in ricebean.

Conclusion

A total of 3,369 lncRNAs were identified in ricebean, underscoring their potential roles in cis- and trans-regulation of key transcription factors and genes associated with cell division, hormone signaling, seed development, and maturation. Additionally, miRNAs that form eTMs were found to modulate mRNA expression, collectively influencing seed-related traits. Comprehensive annotation of seed-associated lncRNAs and the construction of lncRNA–miRNA–mRNA regulatory networks provide a focused set of candidate genes for application in molecular breeding, functional genomics, and genome-editing strategies to improve seed size, nutritional quality, and stress-resilient yield in ricebean and related legume crops.

Abbreviations

lncRNAs

long non-coding RNAs

miRNA

microRNA

DE-LncRNAs

lncRNAs-differentially expressed lncRNAs

DPA

days post anthesis

eTMs

Endogenous target mimics, CeRNA-Competing endogenous RNA

ABA

abscisic acid

GA

gibberellins

CPAT

Coding potential assessment tool

Pfam

Protein families database

Rfam

RNA families database

FPKM

Fragments per kilobase of transcript per million

KEGG

Kyoto encyclopedia of genes and genomes

GO

Gene ontology

COG

Clusters of Orthologous Groups of proteins

Authors’ contributions

SM: conceptualization, manuscript review & editing; PR: methodology & manuscript writing; EG: reviewed the manuscript. All authors have read and approved the final manuscript.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

RNA-Seq reads used in the present study can be accessed from the NCBI SRA database under the project accession number PRJNA765494. All data generated or analyzed during this study are included in this article and its additional files.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.