Rose breeders select roses according to particular criteria, which include cold and disease resistance, flower form, recurrent flowering, and to some degree, scent . In spite of the importance of phenotypes for roses, only a few studies have addressed flower features at the molecular level. Prior to this study, only a few thousand rose unigenes had been deposited in the EST database (dbEST) at NCBI (search query: "Rosa"[Organism]). With the aim of providing valuable resources for molecular studies of rose flowers, we constructed the Rose transcriptome database and analyzed miRNA sequences from three R. hybrids (Vital, Maroussia, and Sympathy) and Haedang.
These rose cultivars were selected because of a range of structures (from a single to multiple layers of petals) and colours (from white to dark red) in their flowers (Figure 1). The database includes 35,657, 31,434, 34,725, and 39,722 unigenes for ‘Vital’, ‘Maroussia’, ‘Sympathy’, and ‘Haedang’, respectively (Table 1), covering an average of 55.96%, 60.70%, 59.14%, and 61.57% of apple, strawberry, peach, and grape protein-coding genes, respectively. We annotated unigenes up to similar levels of related-species, especially in contigs (Table 2). However, overall numbers of identified domains were slightly lower than those identified in related species (Table 2). These domains may be expressed at lower levels in rose flowers or not expressed in flower tissues. In addition, the Rose database provides additional information for the unigenes, including gene description, domains, GO, PO, metabolic pathway, FunCat, and homologous gene families among roses, Rosaceae families (apple, strawberry, and peach) and grape (Figure 1). The sequence dataset and high-quality annotation enabled us to analyse and compare the transcripts which are specialized in rose flowers and related species.
With the help of large-scale sRNA sequencing, a large number of miRNAs have been identified, characterized, and reported [48–54]. However, these techniques are best applied to plant species whose genomes have been assembled, annotated and published (i.e., Arabidopsis, Rice and Grape [51, 52, 54]) because it is necessary to predict and construct hairpin precursors of potential miRNAs, using neighbouring genomic sequences of the mapped sRNAs to distinguish high quality miRNAs from other sRNAs [24, 28]. Therefore, in the absence of the genome sequences, the strategy to identify miRNAs is limited [55, 56].
In this study, we set out to identify highly qualified conserved and novel miRNAs from floriculture plants Rosa × hybrida (Vital, Maroussia, and Sympathy) and Haedang, for which genomes are not available. To predict hairpin precursors for potential miRNAs, we mapped sRNA tags to the strawberry genome, the most closely related species among genome sequenced plants, and identified mature miRNAs using modified miRDeep, a program that employs a probabilistic model of miRNA biogenesis [24, 29]. Mapping sRNA to cross-species is possible because secondary structures for pre-miRNAs are highly conserved between species . Identification of miRNAs by mapping of sRNA to related species has strengths in that it i) maximizes the screening of novel miRNAs in plants for which genome is not available, ii) allows identification of the conserved and novel miRNAs between related species, and iii) provides an incremental improvement in accuracy of miRNA identification by utilizing secondary structures. We might have missed some novel miRNAs that are specific to roses; however, these novel miRNAs are less likely to be important in functional studies because they are expressed at low levels and lack target-genes, whereas conserved miRNAs are usually highly expressed [23, 57].
With this approach, we identified 122, 125, 125, and 120 miRNAs for ‘Vital’, ‘Maroussia’, ‘Sympathy’, and ‘Haedang’, respectively (Table 3). Of these miRNAs, 21, 23, 21, and 20 were novel miRNAs for ‘Vital’, ‘Maroussia’, ‘Sympathy’, and ‘Haedang’, respectively (Table 3). These novel miRNAs correspond to 16–18% of the total identified miRNAs. We also identified 101, 102, 104, and 100 conserved miRNAs by aligning sRNA tags to known miRNAs for ‘Vital’, ‘Maroussia’, ‘Sympathy’, and ‘Haednag’, respectively (Table 3). This hybrid method maximized the identification of expressed miRNAs for rose flowers, therefore providing many miRNA candidates (Figure 1). In addition, this allows us to understand the conserved and unique regulatory processes occurring in rose flowers of different species and hybrids.
miRNA research is advancing from analysis of single tissue or species to comparison of miRNAs between species , varieties , tissues [48, 51–53, 59], or different development stages . Today, NGS technologies allow for the comparison of miRNA profiles with statistical analysis of the redundancy of miRNA tags and enables discussion of tissue- or organ-specific miRNAs [51, 52, 54, 58, 59]. We first compared miRNA profiles among three R. hybrida cultivars and Haednag to analyze the conservation and variation of miRNA tags in the four roses (Figure 4). We identified miRNA tags detected in single rose cultivar, leading to specific function to Rosa. We also identified 297 conserved miRNAs in all four roses (Figure 4A), including 274 known miRNAs and 23 novel miRNAs. Among 274 conserved miRNA, 70 miRNA tags were also identified in strawberry. In addition, 30 conserved miRNAs were also conserved in all Rosaceae (Figure 4B), suggesting that Rosaceae share many conserved miRNAs between ornamental and fruit-bearing plants. Furthermore, among the conserved miRNAs, seven miRNA families were experimentally confirmed to regulate their target genes by cleavage mechanism using 5' RACE assay (Figure 5), suggesting that the conserved miRNAs identified in this study are actually functional.
Most of the target genes validated in this study were transcription factors (Figure 5A-D, F), and their mutant phenotypes were characterized in many model plants (Table 4). Over-expression of miR156 (Figure 5A) and miR159 (Figure 5B) induced delayed flowering in Arabidopsis by negatively regulating SPL and MYB family transcription factors genes, respectively [60, 61]. The expression of miR167-resistant ARF6 (Figure 5A) leads to arrested ovule development and indehiscent anthers . miR172 (Figure 5C) is crucial for development of reproductive organs and for timely termination of floral stem cells by regulating AP2 RNA stability . The expression of miR172-resistant AP2 induces the formation of variable numbers of floral organs with numerous petals and lacking inner whorl organs [63, 64]. miR160 regulates development by altering expression of auxin-induced genes through ARF families [65, 66]. miR164 targets CUC1 and CUC2 transcripts in Arabidopsis and controls leaf margin development. Therefore, we assumed that the conserved miRNAs, which were experimentally validated in this study, may have important roles in floral organ identity or flower developments.
In addition, these eight miRNAs are evolutionary conserved and abundantly expressed miRNAs in roses. According to previous studies, miR156, miR159, and miR160 are evolutionary conserved in all land plants, and miR164, and miR172 are conserved in seed-bearing plants . Evolutionarily conserved miRNAs in plants tend to regulate ancestral transcription factors that specify basic meristem functions, organ polarity and separation, cell division, or hormonal control (reviewed by Garcia ). Based on experimental validation of conserved miRNAs and the current discussions [22, 23, 67], we might expect that novel and un-validated miRNAs identified in this study (Table 4) possibly play important roles such as flower development or hormonal control. Our analysis suggests that the Rose database is a useful tool to search for candidate target transcripts or miRNAs that play roles in flower development in rose and for those have a variety of other specific functions.
Flower colour in most angiosperms is one of the most important targets for plant breeders and many different-coloured cultivars have been bred using natural mutants or genetically-related species. Flower colours are determined by an accumulation of secondary metabolites such as flavonoids, carotenoids, and betalains [68, 69]. We examined miRNA profiles in which target genes were involved in colour-metabolite related biosynthesis pathways to gain insight into the regulation of the white flowered cultivar, Maroussia (Figure 1), which seems to be regulated by miRNAs (Table 5). We hypothesized that miRNA enrichment in Maroussia may negatively regulate the colour-related genes, leading to white colour of Maroussia. Although miRNA enrichments were also observed in other Rosa (Vital, Sympathy, Haedang), the most interesting thing was that five miRNA tags were enriched in Maroussia. Especially, miR356e was expressed up to five times more than other Rosa. Thus, the function of its target, CYP, which involves catalyzing the biosynthesis of flavonoids and cyanidin (red to magenta) and delphinidin (violet to blue) [44, 45], would be more negatively regulated in Maroussia, which may possibly lead to lack of colour . Unfortunately, we were not able to validate miRNA-directed cleavage of these targets due to high penalty score of the target genes (Table 5). However, we would rather expect that miRNA-directed regulation of target gene possibly lead to low read counts of target genes (i.e. low level of expression) in the transcriptome library, which makes target identification more challenging.
However, previous studies reported that two transcription factors, SPL and R2R3-MYB, both of which regulate expression of antocyanins-related genes. Moreover, over-expressed miR156 directly prevent the expression of anthocyanin biosynthetic genes (Additional file 5) by targeting SPL9, in Arabidopsis. In this study, we identified nine miR156 members from all Rosa (Additional file 2), and their target genes, SPL transcription factors, were experimentally validated by 5’ RACE assay (Figure 5). The miR159 were among the most frequent in our library (187,579; 271,208; 264,412 and 327,436 for ‘R.thunb.’, ‘Marcia’, ‘Sympathy’, and ‘Vital’, respectively) and its sequencing frequencies were 10 to 100 times more than other relatively abundant miRNA families, including miR156, miR157, and miR167 (Table 4). Along with this, it has been previously reported that differential expression of R2R3-MYB gene determine colour patterning in plants that are linked with anthocyanin production [70, 71]. Given the fact that the miR159 is very highly expressed in Rosa and miR159-directed cleavage of R2R3-MYB gene is confirmed using 5' RACE, our results raise an intriguing possibility that miRNAs in roses may be involved in pigment synthesis pathway. In addition, miR828, and miR858, which are also involved in R2R3-MYB regulation was predicted target MYB genes in Rosa (Additional file 2).
Based on the miRNA profiles (Table 5, Additional file 5) and the previous research, flower colour seems to be regulated by combinatorial mechanisms. Therefore, it is difficult to elucidate the molecular mechanism or miRNA-directed regulation of colour determinacy of Rosa flowers. Nevertheless, miRNA profiles give potential clues to examine the colour of Rosa flowers. Moreover, there are limited numbers of unigenes currently available. Therefore, we would expect to reveal more direct evidence of miRNA-mediated regulation of colour development in Rosa when genome resources become more available.
In addition to the miRNA profiles, we also compared the expression profiles of colour-related transcripts in Rosa and Rosaceae families (Table 6). It shows that Maroussia contains less number of unigenes (similar to strawberry or peach) than other Rosa cultivars. The higher number of colour-related gene in apple arose by current whole genome duplication . In addition, the average number of reads per gene in Maroussia was also smaller than other Rosa cultivars because all of them were singlets (Table 6). On the whole, based on expression profiles of miRNA and the smaller number of transcripts involved in colour-metabolite related biosynthesis in Maroussia, it is possible to expect that colour-related genes are systemically repressed in Maroussia.