Background

In eukaryotic, approximately 90% of the whole genomes are transcribed into RNA [1]. Among these transcripts, only ~ 2% of them can be translated into proteins, and majority of them are defined as non-coding RNAs (ncRNAs) [2, 3]. The ncRNAs are functional RNA molecules that do not encode proteins and possess key regulatory functions [4]. According to their functions, ncRNAs can be divided into housekeeping ncRNAs and regulatory ncRNAs [5]. Long non-coding RNAs (lncRNAs) are an important group of regulatory ncRNAs that are longer than 200 nucleotides [6]. According to their genomic positions, lncRNAs can be classified into long intervening noncoding RNA (lincRNA), antisense lncRNA (lncNAT), intron lncRNA, and sense lncRNA [7, 8]. Compared to protein-coding genes (PCgenes), most lncRNAs exhibit lower conservation across species, lower expression levels and strong tissue-specific expression [9,10,11,12,13,14]. In plants, more and more studies have shown that lncRNA plays a critical role in many biological processes, including development processes, reproduction processes and stress responses [15,16,17,18].

With the rapid development of high-throughput RNA sequencing, thousands of lncRNAs have been identified and characterized in several plants [10, 13, 14, 19,20,21,22,23,24,25]. Although only a few lncRNAs have known functions in current study, the functions and regulatory mechanisms of lncRNAs are diverse and complex [26,27,28]. For example, a NAT-lncRNA MAS can be induced by cold and activate of sense gene MADS AFFECTING FLOWERING4 (MAF4) for suppression of precocious flowering [29]. GARR2 can influence the plant height ideotype by involving in the modulation of the GA response in maize [30]. In addition, some lncRNAs also play pivotal roles in biotic and abiotic stress responses in plants. Enhanced expression of ALEX1 can activate the expression of jasmonic acid signaling pathway related genes in rice, and significantly improve rice resistance to Xanthomonas oryzae [31]. All in all, lncRNAs might play important biological roles during plant growth and development.

In recent years, the lncRNAs from seeds had been identified in many plants, including maize [32,33,34], Brassica napus [35], tree peony [36], castor bean [22], pigeonpea [37], Ginkgo biloba [38], and rice [39]. These lncRNAs might play a complex regulatory role in seed development. In Brassica napus and tree peony developing seeds, lncRNAs probably have effect on lipid metabolism [35, 36]. In maize and castor bean, lncRNA might play a part in regulating endosperm development by genomic imprinting [22, 33]. In plants, endosperm is a triploid tissue with a 2:1 maternal:paternal genome ratio [40]. Genomic imprinting, mainly occurring in endosperm, refers to allele-specific expression of genes depending on parental origin [41, 42]. So far, imprinted long noncoding RNAs were identified in endosperm of several plants [22, 33, 43]. Recently, a maternally expressed lncRNA MISSEN were reported as a regulator to modulate rice endosperm development [44]. In flowering plants, seed development is an intricate and ordered process that is regulated by both genetic and epigenetic factors [45]. DNA methylation, a heritable epigenetic mark, can affect gene transcription and influence development [46,47,48,49]. Understanding the regulation of DNA methylation requires consideration of the distribution of methylation across the gene and lncRNA. Hence, acquisition of lncRNAs and its DNA methylation pattern in sunflower endosperm will lay a solid foundation for further exploration its influence on seed development.

Sunflower (Helianthus annus L.) is the fourth most important oil crop in the world [50]. And the endosperm was easily separated from the embryo and other maternal tissues, which avoid surrounding tissue contamination. In this study, we analyzed RNA sequencing (RNA-seq) and DNA methylation data, and comprehensively characterized the genomic expression, DNA methylation and inheritance patterns of lncRNAs in endosperm tissues of sunflower. Together, our findings will be helpful for further research on the potential functions, parental effects and epigenetic regulation of lncRNAs in flowering plants.

Results

RNA sequencing and identification of lncRNAs in sunflower endosperm

In order to explore the characteristics of lncRNA expression in sunflower endosperm, the RNA-seq data of 12 days after pollination (DAP) endosperm tissue from reciprocal hybrid pairs of our previously published was performed to identify lncRNA [51]. About 45 million clean reads were acquired from each of the four libraries [SY1(138A × 398B), YS1(398A × 138B), SY2(723A × 6B), YS2(6A × 723B)] for further analysis (Additional file 1: Table S1). After reassembling and mapping, between 88.9 and 91.06% of the reads were successfully aligned to the sunflower genome (Additional file 1: Table S1). Then, the mapped clean reads were assembled as a transcript using StringTie, and we identified 153,342 transcripts (Fig. 1a). Subsequently, the transcripts were filtered based on their type of transcripts and sequence length (less than 200 nucleotides), and 55,231 transcripts were retained (Fig. 1a). Next, the protein-coding potential of remaining transcripts were predicted jointly by four analyses: CPC2 analysis (Coding Potential Calculator), CNCI analysis (Coding-Non-Coding Index), PLEK analysis (predictor of long non-coding RNAs and messenger RNAs based on an improved k-mer scheme) and Pfam protein domain analysis. After the four computational approaches prediction, 17,882 transcripts were obtained (Fig. 1a). Finally, we obtained 11,840 transcripts as putative lncRNAs by expression level [fragments per kilobase of transcript per million mapped reads (FPKM) ≥ 0.5] in sunflower endosperm (Fig. 1a). Thereinto, 11,840 lncRNAs were identified in all tissues (Additional file 2: Table S2), including 9534 and 10,640 lncRNAs in SY1/YS1 and SY2/YS2, respectively (Fig. 1b).

Fig. 1
figure 1

Identification and characterization of long non-coding RNAs (lncRNAs) in sunflower endosperm at 12 DAP. A Schematic pipeline for the identification of lncRNAs in sunflower endosperm; B Expressed lncRNAs in two crosses. Venn diagrams showing the number of common and specific lncRNAs in the four libraries; C Distribution of the lincRNA (red), sense lncRNA (purple), antisense lncRNA (blue) and intron lncRNA (green) on each chromosome; D Classification of total identified lncRNAs including lincRNA, antisense-lncRNA, intronic-lncRNA and sense-lncRNA; E Length density distributions of long non-coding RNAs (lncRNAs) and protein-coding genes (PCgenes); F Distribution of exon numbers in lncRNAs and PCgenes

Full size image

In order to explore the potential functions of lncRNAs, we defined co-expressed protein-coding genes (PCgenes) that located within 100 kb from all candidate lncRNAs. The functional annotation of these PCgenes were carried out by assignment of GO terms. There were 13 biological processes including “hormone-mediated signaling pathway”, “response to abscisic acid”, “response to lipid” and so on, and 11 molecular functions including “hormone binding”, “carboxylic acid binding” and so on (Additional file 3: Table S3).

We examined the overlap of the lncRNA transcripts in the four sunflower F1 hybrid endosperm. As shown, about 88.0% (7347) of the lncRNAs with a genome hit showed evidence of expression in SY1 and YS1 endosperm, and about 87.2% (7100) of the lncRNAs with a genome hit showed evidence of expression in SY2 and YS2 endosperm (Fig. 1b). But only half of (4849) lncRNAs were found in both of two crosses (Fig. 1b), which indicated that lncRNAs tend to be specific expression in intraspecies.

The genomic characteristics of lncRNAs in sunflower endosperm

Using the circus program, these lncRNAs were mapped to the 17 chromosomes of the sunflower genome, and we found that these lncRNAs were evenly distributed in all chromosomes with no obvious location preference (Fig. 1c). Based on their locations in the genome, the 11,840 lncRNAs in sunflower endosperm were divided into four types: 8988 (76%) lincRNAs, 348 (3%) lncNATs, 268 (2%) intronic-lncRNAs, and 2236 (19%) sense-lncRNAs, respectively (Fig. 1d). The lncRNA identified in both of two crosses were tend to be located in genic region compared with all lncRNAs (Additional file 4: Fig. S1). To more clearly characterize the lncRNA in sunflower endosperm, the identified lncRNAs were performed through comparing with that of PCgenes. The sequence length of lncRNA transcripts (average length of 647 nt) was shorter than the PCgenes (average length of 1474 nt) (Fig. 1e). The number of exons of the lncRNAs was significantly lower than that of the PCgenes (Fig. 1f). Approximately 86% of lncRNAs with 1–3 exons were significantly higher proportion than PCgenes (49%). As the number of exons increased, the proportion of lncRNAs decreased.

Association of the expression of lncRNAs and protein-coding genes

The overall expression levels of lncRNAs were significantly lower than those of PCgenes in endosperm of two sunflower crosses (Fig. 2a, Additional file 5: Fig. S2). LncRNA have been found to show tissue-specific expression in plants [10, 13, 14, 22]. To explore the expression patterns of lncRNAs in sunflower, we downloaded and analyzed publicly available RNA-seq data sets of other sunflower tissues, including pistil, stamen, ligule, mature leaf, root, and seed. We found that most of the lncRNAs exhibited strong tissue-specific expression patterns in endosperm, and a small number of lncRNAs showed constitutive expression (Fig. 2b).

Fig. 2
figure 2

Expression of long non-coding RNAs (lncRNAs) and protein-coding genes (PCgenes). A Expression levels of lncRNAs and PCgenes in YS1 endosperm as illustrated by the boxplot; B The expression profile of lncRNAs among tissues; C Summary of various types and numbers of lncRNA–PCgene pairs in sunflower endosperm; D The density distribution of the Pearson correlation coefficient for lincRNA–PCgene and lncNAT–PCgene pairs; E A heat map showing the enrichment of GO terms in the biological process (BP) category and molecular function (MF). The colors of the heat map represents the P-value for each GO term value

Full size image

LncRNA affect gene expression in a cis (neighboring genes) or trans (distant genes) manner. To analyze the potential functions of these lncRNAs, we predicted the cis- and trans-target genes within 100 kb upstream and downstream of the lncRNAs. Pearson correlation coefficient (rp) was used to estimate the expression correlation of lncRNA-PCgene pairs. PCgenes with low expression levels (FPKM< 0.5) were removed. Accordingly, 1792 lincRNA-PCgene and 78 lncNAT-PCgene pairs were identified (Fig. 2c). We observed a high percentage of positive correlations (rp ≥ 0.8, P-value < 0.01, t-test) in lincRNA-PCgene and lncNAT-PCgene pairs (Fig. 2d). The lncNAT-PCgenes pairs exhibited a stronger correlation than the lincRNA-PCgene pairs (Fig. 2d). A gene ontology analysis of those PCgenes showing strong correlation with the lncRNAs revealed that most lncRNAs were involved in methionine adenosyltransferase activity, auxin binding, carboxylic acid binding and so on (Fig. 2e).

DNA methylation of lncRNAs

Since lncRNAs are important regulatory roles in many biological processes, their expression must be tightly regulated. The regulation by DNA methylation of the expression of PCgenes and lncRNAs has not been well characterized in sunflower. The overall methylation levels within the 2-kb flanking region and body region of both expressed PCgenes and lncRNAs (FPKM ≥0.5) was examined. In YS1 endosperm, the PCgenes and lncRNAs displayed a relatively lower methylation levels near the transcription start and stop sites in the CG context (Fig. 3a). The methylation levels of lncRNAs were significantly higher than PCgenes in transcription start sites. In the CHG context, the overall DNA methylation levels within the 2-kb flanking region and body region was substantially higher for lncRNAs (Fig. 3b). In the CHH context, for both lncRNAs and PCgenes, the level of DNA methylation was decreased near the transcription start sites (Fig. 3c). The overall DNA methylation levels of PCgenes in the upstream was higher than lncRNAs, whereas lncRNAs in downstream and gene body regions had a higher level of DNA methylation (Fig. 3c). Similarity, the overall methylation profiles of PCgenes and lncRNAs in SY1 endosperm was similar to those in YS1 endosperm (Additional file 6: Fig. S3).

Fig. 3
figure 3

DNA methylation profiles of long non-coding RNAs (lncRNAs) and protein-coding genes (PCgenes) in sunflower endosperm. A-C Average DNA methylation levels of lncRNAs (blue lines) and PCgenes (red lines) in YS1 endosperm; D-F Association between DNA methylation and lncRNA expression in CG, CHG and CHH sequence contexts throughout the gene body and its 2-kb up- and downstream regions in YS1 endosperm. G, H Two examples of DNA methylation and gene expression at the PCgene (g) and lncRNA (f) were displayed, respectively. The expression level of transcribed regions is shown in green; The DNA methylation level of transcribed regions is shown in red

Full size image

To evaluate the relationship between DNA methylation level and expression levels of PCgenes and lncRNAs, we divided the PCgenes and lncRNAs into three groups according to their expression levels. The highly expressed lncRNAs displayed a relatively lower CG, CHG and CHH methylation levels at both their flanking and body regions (Fig. 3d-f). In contrast, the low expression level of lncRNAs had a higher methylation level for all three sequence contexts. The level of DNA methylation at the transcription start sites was negatively correlated with gene expression levels in lncRNAs. For example, areas near the TSS were about 40% methylation levels for the most highly expressed genes, but were nearly 70% methylation for the genes with lowest expression level. In the PCgenes, the results showed that mRNA transcript levels in endosperm were positively correlated to gene-body methylation levels, but were negatively significantly correlated to promoter methylation levels (Additional file 7: Fig. S4). In Fig. 3g and h, the integrated profiles of DNA methylation and gene expression at the HanXRQr2_Chr01g0000901 (PCgene) and HanXRQr2_lncRNA11165 (lncRNA) were displayed, respectively.

Identification and characters of imprinted lncRNAs

Some lncRNAs exhibit allelic expression which is regulated by the parent-of-origin effects in endosperm of flowering plants. To systematically identify imprinted noncoding RNAs in sunflower endosperm. A total of 36 imprinted lncRNAs in sunflower endosperm were got (Additional file 8: Table S4). Among them, 32 are maternally expressed lncRNAs (MNC), whereas four are paternally expressed lncRNAs (PNC). Most of imprinted lncRNAs were located in intergenic region, including 30 intergenic lncRNAs, one intronic lncRNA, five sense lncRNA (Additional file 9: Fig. S5). These imprinted long noncoding transcripts have an average length of 1049 bp, ranging from 308 bp to 2711 bp (Additional file 8: Table S4), as estimated from regions covered by the sequencing reads.

We assessed allelic imprinting variation in the two crosses (SY1/YS1 and SY2/YS2) as visualized in the Venn diagram (Fig. 4a). Although three (one MNC and two PNCs) imprinted lncRNAs were found to overlap in the two crosses, most of the imprinted lncRNAs identified in one cross tended to be imprinted in other reciprocal crosses (Fig. 4b). Imprinted lncRNAs found in only one set of reciprocal crosses usually lacked informative SNPs or had insufficient reads to identify if they were imprinted in other crosses (Fig. 4b). For example, among 24 imprinted lncRNAs (including 21 MNCs and three PNCs) identified in SY1/YS1 endosperm, four were MNCs/PNCs, one were non-imprinted gene and 19 (79.1%) had no polymorphisms or were not expressed in SY2/YS2 endosperm. Some of the examples of imprinted lncRNAs exhibited imprinting of alleles from some genotypes but not others. Figure 4c and d displays the expression profiles of two MNCs. As showed, all SNPs located at two MNCs exhibited significantly maternal bias.

Fig. 4
figure 4

Identification of imprinted lncRNAs in sunflower endosperm at 12 DAP. A Venn diagram analysis of imprinted lncRNAs. The number of imprinted lncRNAs identified in two crosses are shown in the blue (SY1/YS1) and red (SY2/YS2) circles, respectively.; B Comparison of imprinted lncRNAs in two crosses of sunflower. Non-imprinted: lncRNAs not showing significant deviation from 2:1 ratio of maternal allele to paternal allele in each reciprocal hybrid. Non-analyzed: lncRNAs without sufficient read counts. Low-stringency imprinted lncRNA: lncRNAs showing significant deviation from 2:1 ratio of maternal allele to paternal allele in each reciprocal hybrid. High-stringency imprinted lncRNAs: lncRNAs in which favorable alleles were at least five times more than those of non-favorable alleles in both directions of a reciprocal cross; C, D Two examples of imprinted lncRNAs. The expression level of transcribed regions is shown in green for SY1 and YS1; The percentages of allelic reads of two imprinted lncRNAs for specific SNP sites are shown, with red lines for the paternal allele and blue lines for the maternal allele; Black rectangle, exon; black line, intron. E-F DNA methylation level distribution in imprinted lncRNAs (E) and all lncRNAs (F) around the transcription start site (TSS) region, including CG, CHG methylation

Full size image

Genomic imprinting is generally regulated by epigenetic modifications [52, 53]. The availability of DNA methylome data allowed us to investigate the relationship between DNA methylation and expression of the imprinted noncoding RNAs. In the CG and CHG context, the overall DNA methylation levels of the imprinted noncoding RNAs in the upstream 1 kb and gene body 5′ regions were slightly lower in endosperm than those in embryo (Fig. 4e, f).

Identification of lncRNAs exhibiting allele-specific expression in cultivated sunflower lines for edible fruit and oil

LncRNAs exhibiting allele-specific expression (ASEG) may lead to phenotypic variation depending on the function of the genes. To better understand how parental alleles contribute to the development of endosperm, a genome-wide identification of lncRNAs exhibiting allele-specific expression were performed by comparing the read ratios of the parental alleles in RNA-sequencing data of hybrid endosperm. Consequently, the expression of 81 and 62 lncRNAs showed allelic bias toward cultivated lines for edible fruit (SA1 and SA2) and cultivated lines for oil (YA1 and YA2), respectively (Additional file 10: Table S5). Interestingly, lncRNAs showing allelic bias toward cultivated lines for edible fruit and cultivated lines for oil seem have different function in sunflower development. The functional annotation of these PCgenes located within 100 kb from lncRNAs showing allelic bias were carried out by assignment of GO terms. For lncRNAs showing allelic bias toward cultivated lines for edible fruit, there were three enriched GO term including “cysteine-type endopeptidase activity”, “polysaccharide binding” and “pattern binding” (Additional file 11: Table S6). For lncRNAs showing allelic bias toward cultivated lines for oil, there were 15 enriched GO term including “ATP binding”, “carbohydrate derivative binding” and “pattern binding” and so on (Additional file 12: Table S7).

Discussion

In recent years, growing evidence suggested that lncRNAs play an essential role in plant development and responses to stresses [54, 55]. So far, lncRNAs have been characterized in many plant species, such as Arabidopsis [13, 19], rice [21], maize [10] and wheat [21, 56]. Here, we undertook a genome-wide identification and characterization of lncRNAs and analyzed its methylation pattern in sunflower endosperm. In this study, 11,840 lncRNAs were identified by analyzing RNA-seq data of endosperm from two reciprocal crosses. The number of lncRNAs in sunflower endosperm is nearly twice more than that in caster bean [22]. The main reason may be the difference of genome size and complexity [57, 58]. Of course, the computational approaches prediction applied were different. Despite a large number of lncRNAs have been identified from many species, methods developed to date are not sufficiently accurate or comprehensive, which may cause incorrect and conflicting results [59]. In our study, we also found that the sequence lengths of lncRNAs are shorter, exon numbers are fewer, expression levels are lower, and have relatively specific tissue-specific expression when compared to PCgenes. These results are consistent with previous reports describing the common features in other plants [10, 21, 22, 60]. Also, we found that half of lncRNA tend to be expressed specifically in sunflower genotype. This implies that lncRNAs might share a common evolutionary pattern and have rapid turnover of lncRNA sequence.

LncRNAs can act in cis (neighboring genes) or in trans (distant genes) to regulate the expression of genes via transcriptional level, epigenetic modification level or post-transcriptional level [8, 61]. In previous work, about 20,000 lncRNA were identified in sunflower meiocytes [25]. And these lncRNAs potentially play roles in meiosis and may participate in the processes of chromatin modification [25]. In our study, a large number of lncRNAs were distant from PCgenes. Whether distant lncRNAs exert their function in trans, or as enhancers or insulators, needs to be further determined. A strong positive correlation was only present in a small number of lncRNA-PCgenes, suggesting that transcription of these genes may be coordinately regulated by adjacent lncRNA. It is tempting to speculate that coordinated transcription of lincRNAs with nearby PCgenes may be due to common regulatory sequences in their promoter regions, and/or that these lncRNAs themselves can positively regulate the transcription of nearby genes in cis. Seed oil content and quality is one of major breeding traits for sunflower [57]. We found that some genes homologous to Arabidopsis genes were metabolic pathways involved in oil synthesis and seed development (Additional file 13: Table S8). For example, we identified the lncNAT (HanXRQr2_lncRNA08192) located downstream of the gene HanXRQr2_Chr04g0171821, which was homologous to AT2G26640 (KCS11) in Arabidopsis, encoding KCS11, a putative member of the 3-ketoacyl-CoA synthase family involved in the biosynthesis of VLCFA (very long chain fatty acids). In eukaryotes, S-adenosylmethionine enzymes play roles in rRNA modifications [62, 63], tRNA modifications [64, 65], and lipid metabolism [66, 67]. The lncNAT (HanXRQr2_lncRNA09471) was expressed in downstream of the gene HanXRQr2_Chr06g0272911, which was homologous to AT4G13330 in Arabidopsis, encoding a putative S-adenosyl-L-methionine-dependent methyltransferases superfamily protein and may be related to fruit development. The target gene of lincRNA HanXRQr2_lncRNA00100 product is a putative FatA acyl-ACP thioesterase whose homologous gene in Arabidopsis is AT3G25110. Previous study showed that FatA is the dominant thioesterase during the period of oil accumulation in sunflower seeds [68]. Sunflower FatA acyl-ACP thioesterase is important not only for oil deposition in the seed but also, for the final oil composition [68]. These results suggested that lincRNA HanXRQr2_lncRNA00100 may regulate the expression of FatA, which could functions in the fatty acid biosynthesis pathway. Another lincRNA HanXRQr2_lncRNA02864 targets casein kinase I (CKI) gene, which encode a putative Ser/Thr kinase protein [69]. In rice, the activity of the lipase is controlled by the activity of riceCKI [70]. These may be involved in fatty acid biosynthesis pathway regulation. The protein-sequence homology of the target gene of lincRNA HanXRQr2_lncRNA09231 to Arabidopsis suggested that it encodes a putative Flavin-containing monooxygenase family protein (YUC10). In Arabidopsis, the YUC genes are mainly expressed in meristems, young primordia, vascular tissues, and reproductive organs, and it is essential for the formation of floral organs [71]. In maize, ZmYuc1 can affect endosperm development by regulating IAA biosynthesis [72]. These results suggested that lincRNA HanXRQr2_lncRNA09231 may be involved in seed development. The target gene of lincRNA HanXRQr2_lncRNA04387, which was homologous to AT4G00850 (GIF3) in Arabidopsis, encoding a putative GRF1-interacting factor 3. The GRF-INTERACTING FACTOR (GIF) family of Arabidopsis is an essential component required for the cell specification maintenance during reproductive organ development and, ultimately, for the reproductive competence [73]. This may imply that lincRNA HanXRQr2_lncRNA04387 is related to seed development. Transcription factors play important roles in plant development including floral organogenesis [74, 75], leaf initiation [76], lateral shoot initiation [77], gametogenesis [78] and seed development [79]. Those PCgenes showing strong correlation with the lncRNAs included 30 transcription factors (Additional file 14: Fig. S6). Hence, lncRNA potentially function in play roles in seed development. Along with the study of the coordinated transcription of lncRNA-PCgene pairs, additional mechanistic insights into the function of lncRNAs should be explored in future.

DNA methylation in plant have been focused on its regulation for gene expression [80]. In this study, we compared overall methylation levels between PCgenes and lncRNAs. We found that lncRNAs exhibited a much higher levels of DNA methylation than PCgenens, which might explain the low expression levels of lncRNAs. The similar expression pattern was also observed in castor bean [22]. Meanwhile, DNA methylation levels at transcription start sites were negatively correlated with lncRNAs expression levels, which was also the same with PCgenes. These finding indicate that DNA methylation may be related to regulation of lncRNAs expression in sunflower endosperm.

Genomic imprinting may be an important dosage control mechanism to regulate gene expression in a parent-of-origin-dependent manner [81]. Studies on the endosperm of rice, maize and castor bean identified a small number of imprinted lncRNAs [22, 33, 43]. Recently, a maternally expressed lncRNA MISSEN were reported as a regulator to modulate rice endosperm development [44]. Hence, identification and studies on the potential roles of imprinted lncRNAs in the triploid endosperm were meaningful for understanding the development of seed. In this study, we identified 36 imprinted lncRNAs by generating reciprocal crosses of different sunflower lines (Additional file 8: Table S4). Very similarly in rice and maize, the number of MNCs is significantly more than the number of PNCs [33, 43], suggesting that MNCs might have play more important roles in sunflower endosperm. In our study, we discovered most of imprinted lncRNAs showed parent-of-origin-dependent expression in certain genotypes but not in others. Major reason was due to lack of SNP. Hence, the density of SNPs was key limit for comparing the imprinting status of the lncRNAs in different reciprocal hybrids. Although the limited lncRNAs can be allelically analyzed in both of two crosses, imprinted lncRNAs show evidence of allelic variation for imprinting. However, how frequently imprinting variation of lncRNA is deserved to be research in future. The epigenetic profiles were also investigated for 36 imprinted non-coding RNAs. Result indicated the maternal demethylation at MNCs and the similar mechanism for epigenetic regulation of imprinted genes and non-coding RNAs.

In our study, 143 lncRNAs exhibiting allele-specific expression in cultivated sunflower lines for edible fruit and oil (Additional file 10: Table S5). Based on the result of GO analysis, we found lncRNAs showing allelic bias toward cultivated lines for edible fruit and cultivated lines for oil seem have different function in sunflower development. Serine carboxypeptidase (SCP) is a class of enzymes catalyzing proteolysis for functional protein maturation [82]. In rice, serine carboxypeptidase 46 has been reported to regulate grain filling [82]. lncRNAs showing allelic bias toward cultivated lines for oil are enriched in pathways related to serine-type peptidase activity. These results suggest that these lncRNAs may play a key role in grain filling in cultivated sunflower lines for oil. In peanut, differentially expressed genes in seed of different oil content varieties was analyzed for significant enrichment of GO terms [83]. Higher expression of generation of energy and metabolites was observed in peanut cv. Hanoch (high oil genotype) than 53 (low oil genotype) during seed development [83]. In grain filling in cultivated sunflower lines for oil, processes involving the generation of precursor metabolites and energy (e.g. ATP binding, adenyl nucleotide binding, carbohydrate derivative binding, oxidoreductase activity, pyrophosphatase activity) was significant enrichment (Additional file 12: Table S7). The result is similar to that reported in peanut. This might explain the differences in cultivated sunflower lines for edible fruit and oil.

Conclusions

We comprehensively identified and analyzed11,840 lncRNAs in sunflower endosperm. Base on genome-wide analyses we found that the lncRNAs were relatively short, had fewer exons and a very tightly controlled tissue-specific expression compared to PCgenes. And a small fraction of lncRNAs exhibited coordinated expression with nearby PCgenes. Moreover, Genomic DNA methylation analyses revealed that the expression level of lncRNAs was tightly linked to DNA methylation. We further characterized expressed imprinted lncRNA during hybridization. Importantly, these results provide valuable information pointing to potential roles for lncRNAs in the development of sunflower endosperm. Our findings also shed light on the inheritance patterns of lncRNA expression and the epigenetic regulation of lncRNA itself in plants.

Materials and methods

Data sources

The datasets in this study were obtained from NCBI (https://www.ncbi.nlm.nih.gov) BioProject PRJNA740059 [51]. The RNA-seq datasets, YS2 endosperm (SRR14885491), SY2 endosperm (SRR14885492), YS1 endosperm (SRR14885493), SY1 endosperm (SRR14885498), were used for filtering potential lncRNAs. YS1(SRR14885497). The DNA methylation datasets, YA1(398A) embryo (SRR14885495), SY1 ensosperm (SRR14885496), YS1 ensosperm (SRR14885497), were used to analyze the average methylation levels for lncRNAs.

Identification of lncRNAs and expression analysis

All raw reads containing adapter and low-quality reads were remove to obtain clean reads via Trim Galore (https://github.com/FelixKrueger/TrimGalore). The clean reads were used to align to reference genome of sunflower (https:// www.ncbi.nlm.nih.gov/assembly/GCF_ 00212 7325.2/), using HISAT2 [84]. After mapping to the reference genome of sunflower, the final transcriptome was assembled and quantified using StringTie [85].

After assembling and obtaining the transcripts, the process of lncRNA identification was based on their characteristics. The class-code of transcripts with ‘j’, ‘i’, ‘x’, ‘u’, ‘o’ and ‘e’ were chosen with Gffcompare for further analysis [85]. Then, we screened out the transcripts with length longer than 200 bp. Because lncRNA does not code protein, except the indictor of length and type, the transcript also should be evaluated whether it possessed the capability of coding protein. Based on the CPC2 (Coding Potential Calculator 2, identified label was ‘nocoding’) [86], CNCI (Coding-Non-Coding Index, identified label was ‘nocoding’) [87], PLEK (the Predictor of Long noncoding RNAs and mEssenger RNAs based on an improved K-mer scheme, identified label was ‘nocoding’) [88] and Pfam (E-value < 0.001) [89] analysis, the transcripts that could potentially code for a protein were removed. According to the FPKM values, transcripts that were less than 0.5 were discarded. The identified lncRNAs were further classified into four types of lncRNA by the genomic locations relative to PCgenes.

Target gene prediction and functional annotation

To explore the function of lncRNAs in sunflower endosperm, we predicted the target genes of lncRNAs. In this study, PCgenes in 100 kb up- and downstream from the lncRNA, were selected by bedtools [24, 90]. To further function analysis, we identified a set of transcript pairs between the lincRNAs and the PCgenes transcribed within a 100 kb upstream or downstream of lincRNAs [91], and between the lncNATs and the corresponding PCgenes [22]. And the correlation in expression was evaluated using Pearson’s correlation coefficient (|rp| > 0.8 and p < 0.01) [22]. Pearson’s correlation coefficient and two-tailed Student’s t-test were calculated.

GO annotation was performed by InterProScan. The GO term enrichment analysis was conducted for genes included in each cluster using website (https://www.genescloud.cn/chart/GOenrich). All PCgenes and lncRNA-associated PCgenes were divided into two groups. GO categories among molecular function and biological process that show significant (p < 0.01) enrichment were displayed.

Analysis of DNA methylation of lncRNA

DNA methylation data from the endosperm (SY1 and YS1) and embryo (398A) at 12 DAP were used to analyze the average methylation levels for lncRNAs, and the methylation ratios of CG, CHG and CHH sequence contexts were calculated as described in our previous study [51]. The methylation profiles in the 2-kb flanking regions and the lncRNA bodies were plotted based on the average methylation level for each 100-bp interval.

Identification of imprinted lncRNA in sunflower 12 DAP endosperm

The SNP calling were performed as previously described [51]. according to the information of SNPs, we can divide the short sequences aligned at the SNP site from maternal or paternal allele. A series of Perl programs were used to calculate read counts from maternal or paternal allele at each SNPs. For a lncRNA, the number of reads that mapped to each allele was summed across all SNPs. Only transcripts that had at least 10 reads that could be assigned to a particular allele in each direction of the reciprocal cross could be analyzed. lncRNAs sites with significant bias (greater than or less than 2:1) in both hybrid endosperm tissues were considered as potentially imprinted lncRNAs. To obtain a subset of high-confidence imprinted lncRNAs, the favorable alleles were at least five times more than those of non-favorable alleles in both directions of a reciprocal cross, similar to the standard used in our previous study [51].