Introduction

Rhododendrons, which are commonly referred to as rosage, are decorative shrubs with beautiful flowers that are widespread around the world in countries characterized by mild climates (Chosson et al. 1998). While there is a remarkably broad range of rhododendron flower colors, such as red, pink, purple, and white, there is no yellow-flowered cultivar of evergreen rhododendron (Ureshino et al. 2016). A yellow-flowered evergreen rhododendron has been the dream and goal of many rhododendron breeders. Rhododendron molle is a deciduous Rhododendron species with unique yellow flowers. Variations in the morphological characteristics and pigments of wild evergreen rhododendron flowers have been studied in detail (Huyen et al. 2016; Mizuta et al. 2009; Nakatsuka et al. 2008). It has been reported that the flower color of Rhododendron japonicum with yellow-flower is attributed to the presence of a carotenoid (Miyajima et al. 2000). Carotenoids are 40-carbon isoprenoids with polyene chains, which are usually responsible for yellow-orange petal colors and play an important role in furnishing flowers and fruits with distinct colors that are designed to attract pollinators (Hirschberg 2001). During the past decade, nearly all of the genes involved in the carotenoid biosynthesis pathway have been identified and characterized (Pandurangaiah et al. 2016; Rezaei et al. 2016; Yang et al. 2016). The differential expression of these genes may lead to differences in petal color (Sagawa et al. 2016; Xin; Li et al. 2015). However, there are few genetic data concerning carotenoids of R. molle. Therefore, the study of transcrioptome sequencing will provide an important molecular basis for study of R. molle and understanding its flowering mechanisms.

In recent years, the number of EST data sets has rapidly increased (Han et al. 2017; Xia et al. 2017). However, rhododendrons have little data available currently in DNA or protein sequence databases. It hampers the molecular breeding research of rhododendron. Sequencing of large genome remains expensive, even using next generation sequencing technologies. On the contrary, de novo transcriptome sequencing has proved to be accurate and efficient with much higher throughput and relatively low cost, providing opportunities to discover massive expressed sequence data for model and non-model species, such as chrysanthemum (Wang et al. 2017), tea (Li et al. 2017), and lettuce (Zhang et al. 2016). Moreover, the transcriptome is a complete set of transcripts in a cell at a specific developmental stage, providing information on amino acid content of proteins, gene expression and gene regulation. Thus the transcriptome analysis is essential to interpret the functional elements of the genome and reveal the molecular constituents of cells and tissues (Liu et al. 2012; Wang et al. 2009). In addition, transcriptome sequencing can provide a simple and effective method to develop a large number of simple sequence repeats (SSRs) marker and improved the efficiency of molecular marker development in many non-model species (Chen et al. 2016; Guo et al. 2016; Wang et al. 2016). To our knowledge, this is the first report on the characterization of transcrioptome of R. mollle. Hence, the study of transcrioptome sequencing will provide an important molecular basis for study of R. molle and understanding its flowering mechanisms.

In the present study, we used the Illumina sequencing platform to conduct the transcriptome sequencing analysis of mixed RNA separately extracted from flowers at four developmental stages of R. molle flower. A total of 123,120 different transcripts and 66,026 unigenes were identified. Meanwhile, a large number of SSRs were determined. Particularly, the transcripts involved in carotenoid biosynthesis were sequenced, aiming to facilitate color-regulating mechanisms research in rhododendron.

Materials and methods

Plant materials and RNA extraction

Rhododendron molle was grown in the field in Nanjing (China). Samples of flowers at four developmental stages were collected in 2016 as follows: tight bud stage (S1), soft bud stage (S2), initial flowering stage (S3) and full flowering stage (S4). The samples were collected, frozen immediately in liquid nitrogen, and stored in − 80 °C for RNA extraction. Total RNA was extracted using the Quick RNA isolation kit (Bioteke Corporation, Beijing, China) following the manufacturer’s instructions. RNA quality and purity were checked using a Nanodrop 2000 spectrophotometer (Thermo, Wilmington, DE, USA) and Qubit RNA Assay Kit in Qubit 2.0 Fluorometer (Life Technologies, CA, USA). The RNA integrity was assessed using the RNA Nano 6000 Assay Kit of Agilent Bioanalyzer 2100 system (Agilent Technologies, CA, USA).

cDNA library construction and sequencing

A pooled sample containing equal amounts of purified total RNA from each of the four samples was sent to Biomarker Technology Co. Ltd (Beijing, China). Construction of the libraries and the RNA-Seq were performed using the Illumina HisSeq 2500 platform. Firstly, mRNAs were isolated using oligo (dT) magnetic beads and fragmented into small pieces using a fragmentation buffer. First strand cDNA was reverse transcribed by random hexamers, followed by second strand cDNA synthesis. The double-stranded cDNA was subjected to an end repair process and the addition of single nucleotide A. Then paired-end adapters were ligated to the ends of the cDNA fragments. Suitable fragments were selected as templates for polymerase chain reaction amplification. Finally, the library preparations were sequenced on an Illumina HiSeq 2500 platform.

De novo transcriptome assembly

Clean reads were obtained by removing reads with adapters, reads with unknown nucleotides larger than 5% and low quality reads from raw data. Transcriptome de novo was carried out by program Trinity (Grabherr et al. 2011). All transcripts were clustered and the longest transcript in each cluster was defined as the unigene for functional annotation. The initial assembly quality was evaluated using the total number of contigs, mean contig length, maximum contig length and N50 (Martin and Wang 2011). The Trinity utility was used to predict ORFs of at least 100 amino acids from the reconstructed contigs.

Gene annotation and pathway assignments

All assembled unigenes were searched against NCBI non-redundant protein sequences (Nr) to determine the putative mRNA functions using the BLAST algorithm with an E-value lower than 10−5 (Altschul et al. 1997). GO (Gene ontology) terms were extracted from the best hits obtained from the BLASTX against Nr using the Blast2GO program (Götz et al. 2008). The sequences with corresponding enzyme commission (EC) numbers obtained from blast2GO were mapped to the Kyoto Enzyclopedia of Genes and Genomes (KEGG) Web server (Kanehisa et al. 2008). The WEGO software (http://wego.genomics.org.cn/cgi-bin/wego/index.pl) was used to obtain GO functional classifications. In addition, Unigenes were also aligned to KOG/COG (Clusters of Orthologous Groups of proteins) database, Nt (NCBI non-redundant nucleotide database) and Swiss-Prot to predict and classify possible functions of unigenes (Tatusov et al. 2001).

Extraction and measurement of carotenoids

The samples of flowers at four development stages were collected from five trees. And each stage contained five flowers with three replicates. The carotenoids were extracted and measured according to the previous study (Yan et al. 2015). Frozen flower petals (0.8 g) were homogenized in 4 mL of acetone, and extracted for 24 h away from light at 4 °C. Then the extracts were centrifuged at 16,000×g for 10 min at 4 °C, and filtered through a 0.22 µm filter for analysis in the Agilent 1100 series HPLC system (Agilent, USA). Samples (20 µL of extract) were analyzed using a YMC-C30 chromatographic column coupled with a UV detector at 450 nm, using methyl tertiary butyl ether and methyl alcohol (V/V = 30:70) as solvents: the flow rate was 1.0 mL min−1 and temperature was 25 °C. The lutein, zeaxanthin, α-carotene and β-carotene were quantified by comparing the peak areas.

Quantitative real-time PCR analysis

Quantitative real-time PCR (qRT-PCR) analysis was used to verify the expression levels of genes identified in sequencing. The RNA samples used for qRT-PCR assays were the same as the sequencing samples. Housekeeping gene actin was used as the control gene (Xiao et al. 2016). All primers are listed in Table S1.The qRT-PCR reactions were performed using a SYBR Premix Ex Taq Kit (TaKaRa, Dalian, China) following manufacturer’s instructions and was analyzed in the ABI 7500 Real-Time PCR system (Applied Biosystems, CA, USA). In a white 96-well plate (Axygen, CA, USA), 10 µL SYBR Master Mix and 0.4 µL of 10 µM forward primer and reverse primer was used in combination with 2 µL cDNA template, resulting in a total volume of 20 µL. Cycling conditions involved 10 s at 95 °C followed by 40 cycles of 5 s 95 °C and 34 s 60 °C. A melting dissociation curve was created to identify amplicon characteristics. The final threshold cycle values were average values. Each reaction was performed in triplicate. Analysis of the relative gene expression data was conducted using the 2−ΔΔCt method.

Development of SSR markers

Simple sequence repeats (SSRs) were detected using the MISA software (http://pgrc.ipk-gatersleben.de/misa/). The parameters were adjusted for identification of perfect di-, tri-, tetra-, penta-, and hexa-nucleotide motifs with a minimum of 6, 5, 5, 5 and 5 repeats, respectively.

Results

Sequencing and de novo assembly

A total of 20,143,879 (5.07 G data) clean reads were obtained after removing duplicate reads, trimming adapters and low quality sequences. The Q30 percentage and GC-content were 88.47 and 47.69%, respectively. The sequencing raw data were deposited in the NCBI database and can be accessed in the Short Reads Archive (SRR6056607).

The trinity software generated 124,120 all-transcripts with length > 200 bp, mean length of 1093 bp and an N50 of 1862 bp (Table 1). 66,026 all-unigenes were achieved with an N50 of 1243 bp. There were 26,539 (40.19%) unigenes of 200–300 bp; 16,007 (24.24%) unigenes of 300–500 bp; 10,656 (16.14%) unigenes of 500–1000 bp; 8053 (12.20%) unigenes of 1000–2000 bp; and the remaining 4771 (7.23%) unigenes were > 2000 bp (Table 1). The length distribution of unigenes is shown in Fig. 1.

Table 1 Length distribution of assemble transcripts and unigenes
Full size table
Fig. 1
figure 1

Length distribution of de novo assembled unigenes

Full size image

Functional annotation and classification

All of the contigs generated by Illumina sequencing were aligned to public protein and gene databases. The overall functional annotation is shown in Table 2. The contigs were compared to the sequences in the NR database, using BLASTX (E-value < 1e−3). A total of 31,298 unigenes (47.40%) returned a significant BLASTX result. The contigs homologous to sequences in the NR database were annotated using Blast2GO. A total of 18,680 (28.29%) unigenes was assigned to 4658 GO term annotations, which could be classified into three categories: biological processes, molecular functions and cellular components (Fig. 2). Among biological processes, “oxidation–reduction process” represented the most abundant subcategory followed by “protein phosphorylation” and “regulation of transcription”. In the molecular function category, “ATP binding” and “Zinc ion binding” were the most represented GO terms. Regarding cellular components, the major GO terms were “plasma membrane” and “internal component of membrane”.

Table 2 Functional annotation of the R. molle transcriptome
Full size table
Fig. 2
figure 2

Gene ontology (GO) classification of assembled unigenes. The 66,026 matched unigenes were classified into three functional categories: biological process, cellular component and molecular function

Full size image

The KOG analysis allowed the functional classification of 17,739 unigenes (Fig. 3). The most frequently identified classes were “general function” (4455, 25.11%), followed by “posttranslational modification” (1976, 11.14%) and “signal transduction mechanisms” (1723, 9.71%).

Fig. 3
figure 3

Clusters of eukaryotic orthologous groups (KOG) classification

Full size image

KEGG pathway and predicted coding sequences (CDS)

A total of 7177 unigenes had significant matches in KEGG, mapped with 118 KEGG pathways (Table S2). Figure 4 summarizes the top 20 represented biological pathways. The most highly represented pathways were “Ribosome” (349 unigenes, 4.86%, ko03010) and “Oxidative phosphorylation” (268 unigenes, 3.73%, ko00190). Four pathways were found to be related to plant color: “Anthocyanin biosynthesis” (2 unigenes, ko00942), “Carotenoid biosynthesis” (32 unigenes, ko00906), “Flavone and flavonol biosynthesis” (7 unigenes, ko00944) and “Flavonoid biosynthesis” (38 unigenes, ko00941).

Fig. 4
figure 4

KEGG classification of unigenes

Full size image

Based on the public protein database, a total of 55,456 CDSs was obtained. Among these CDSs, 47,695 were between 100 and 1000 bp, 5934 were between 1000 and 2000 bp, and 1827 were over 2000 bp (Fig. 5).

Fig. 5
figure 5

Length distribution of predicted CDS

Full size image

Key enzymes related to the carotenoid pathway

There were 32 unigenes annotated as encoding enzymes involved in carotenoid synthesis based on the KEGG pathway assignments while only 2 unigenes encoding enzymes of the anthocyanin biosynthesis pathway were identified (Table S2). Phytoene synthase (PSY), a rate-limiting enzyme in the carotenoid biosynthesis pathway, catalyzes a head-to-head coupling of two geranylgeranyl diphosphate (GGPP) molecules to produce phytoene (Fig. 6). Then phytoene is catalyzed to lycopene by phytoene desaturase (PDS), ζ-carotene desaturase (ZDS) and carotenoid isomerase (CRTISO). All of these enzymes were identified: c18470.graph_c0 annotated as PSY, c25193.graph_c0 as PDS, c21632.graph_c0 as ZDS and c29096.graph_c0 as CRTISO. Then the carotenoid biosynthetic pathway was split into two routes. In one route, lycopene ε-cyclase (LCYE) and lycopene β-cyclase (LCYB) catalyzed the cyclization of lycopene to produce α-carotene, which was subsequently converted to lutein (Chen et al. 2015). Transcript c30693.graph_c0 and c25695.graph_c0 were annotated as LCYB and LCYE, respectively. Enzymes in the other route were essential for the biosynthesis of abscisic acid (ABA). The route yielded β-carotene with catalysis by LCYB. Then β-carotene was catalyzed by β-ring hydroxylase (CHYB, c18642.graph_c0), zeaxanthin epoxidase (ZEP, c12244.graph_c0), and violaxanthin de-epoxidase (VDE, c11312.graph_c0) produced zeaxanthin and violaxanthin. Finally, violaxanthin was converted to ABA by neoxanthin synthase (NSY) and 9-cis-epoxycarotenoid dioxygenase (NCED, c24238.graph_c0, c18704.graph_c0, c46847.graph_c0 and c62319.graph_c0).

Fig. 6
figure 6

Simplified pathways for the carotenoid biosynthesis. Numbers in the brackets indicated numbers of candidate unigenes annotated in our research

Full size image

To study whether high carotenoid percentage results from high expression levels of carotenoid biosynthesis related genes in R. molle, the expression of transcripts encoding key enzymes for carotenoids was analyzed. Transcript c21632.graph_c0_seq1, annotated as Zeta-carotene desaturase (ZDS), reached the highest expression level (267.79 FPKM). PSY (C18470.graph_c0), PDS (c25193.graph_c0), CHYB (c18642.graph_c0) and ZEP (c12244.graph_c0) were 104.92 FPKM, 71.38 FPKM, 35.56 FPKM and 19.4 FPKM, respectively. Meanwhile, the expression of LCYB (c30693.graph_c0) was 26.01 FPKM, 3.22-fold higher than LCYE (c25695.graph_c0). These finds impel us to hypothesize that most of lycopenes are converted into β-carotenes by LCYB, and β-carotenes are further modified by hydroxylation or epoxidation, providing a variety of structural features.

Changes of carotenoid content during flower development

The results showed that there were a large number of carotenoids accumulated in the petal during the flower early development. The concentration ranges of lutein, zeaxanthin, α-carotene and β-carotene were 4.24–48.15 µg g−1 FW, 1.72–18.71 µg g−1 FW, 1.05–2.63 µg g−1 FW, 46.22–95.25 µg g−1 FW, respectively. The content of β-carotene was much higher than the other carotenoids during the flower development (Fig. 7). The α-carotene concentration remained low throughout flower development, and declined to the lowest at stage S3. The content of β-carotene tended to decline at stage S2 and rise to the highest at stage S4. On the contrary, the content of lutein and zeaxanthin tended to rise at stage S2 and decline to the lowest at stage S4. It was indicated that β-carotene played an important role in the coloring of R. molle flower.

Fig. 7
figure 7

Changes of carotenoids content during flower development in R. molle. Four flowers developmental stages were examined in this study as follows: tight bud stage (S1), soft bud stage (S2), initial flowering stage (S3), full flowering stage (S4). Error bars show the standard error calculated from three biological replicates

Full size image

Validation of Illumina sequencing result by qRT-PCR

To confirm the accuracy and reproducibility of the Illumina expression profiles, qRT-PCR analysis was carried out to analyze the expression level of five genes related to carotenoid catabolism (Fig. 8). The expression levels of each gene were measured through qRT-PCR and compared with its abundance from sequencing data. The qRT-PCR analysis showed that PSY (C18470.graph_c0) and PDS (c25193.graph_c0) were highly expressed during the flower development, followed by CHYB (c18642.graph_c0), LCYB (c30693.graph_c0), and LCYE (c25695.graph_c0). It is generally agreed well with the sequencing result, thus confirmed the Illumina expression profile analysis.

Fig. 8
figure 8

Expression levels of five genes related to carotenoid catabolism throughout the flower development of R. molle. a Expression levels of PSY and PDS at different stages of flower development. b Expression levels of LCYB, LCYE and CHYB at different stages of flower development. Error bars show the standard error calculated from three biological replicates

Full size image

SSR discovery

Approximately 3–7% of expressed genes in the plant genome contain putative SSR motifs, which are mainly located in the untranslated regions of mRNA (Liu et al. 2012). SSRs are widely used for genetics, evolution and breeding research. To explore SSR profiles in the R. molle, the assembled sequences (≥ 1 kb) were mined for SSRs. In total, 8266 SSRs were obtained from 12,824 unigenes with 1707 unigene sequences contained more than one SSR. The di-nucleotide repeat motif was the most abundant, accounting for 56.15%, followed by tri-nucleotide repeat motif (15.56%), tetra-nucleotide (0.57%), penta-nucleotide (0.17%) and hexa-nucleotide repeat units (0.08%) (Table 3).

Table 3 Distribution of identified SSRs using the MISA software
Full size table

The distribution and frequency of classified SSRs were analyzed (Fig. 9). Di-nucleotide was the most common repeat motif with a frequency of 56.15%, followed by tri-nucleotide (15.56%), tetra-nucleotide (0.57%), penta-nucleotide (0.17%), and hexa-nucleotide (0.08%). Among the dinucleotide repeats, the TC/GA (27.44%) and AG/CT (24.28%) were the most frequency motifs, accounting for 92.11% of the total dinucleotide repeats, much higher than that of Dendrocalamus latiflorus Munro and Huperzia serrate Thunb (Liu et al. 2012; Luo et al. 2010). Among the trinucleotide repeats, GAA/CTT (1.44%) and AGA/TCT (1.16%) were the most frequent motifs. The SSRs mining from this database will be valuable resources for genetic manipulations in R. molle.

Fig. 9
figure 9

Frequency distribution of SSRs based on motif types

Full size image

Discussion

The high-throughput sequencing technology is an efficient and reliable tool for the transcriptome analysis in organisms without a reference genome. Prior to our work, there was no sequencing information on R. molle public database. The cDNA library was constructed using pooled RNA samples. In this study, a total of 5.07 Gb clean data was generated and assembled into 66,026, 31,542 of which were annotated in public databases, and 32 carotenoid structural genes was found. In addition, 8266 SSRs were identified from 12,824 unigenes. This study provides valuable information for future molecular research on R. molle.

In total, 7177 unigenes were assigned to different pathways in the KEGG classification system. Four pathways, flavone and flavonol biosynthesis (ko00944), flavonoid biosynthesis (ko00941), anthocyanin biosynthesis (ko00942), and carotenoid biosynthesis (ko00906), involved in the color formation of plant flower were identified. The carotenoid biosynthesis pathway is important for the synthesis of plant pigments. The color of yellow flowers is usually caused by the presence of carotenoids. The carotenoid metabolic pathway has been extensively studied (Yuan et al. 2015). In pale-yellow petals of marigold, expression levels of nearly all the carotenogentic genes are relatively lower than those in orange petals (Ohmiya 2013). In chrysanthemum petals, no significant difference in the expression levels of carotenoid biosynthesis genes was detected between the white and yellow petals. However, carotenoids are degraded into colorless compounded for CmCCD4a highly expressed in the white-flowered chrysanthemum, which results in the white petal color (Ohmiya et al. 2006). The yellow color fades in azalea petals are also contributed by high expression of CCD4 (Ureshino et al. 2016). Disruption of CCD4 gene can convert flower color from white to yellow in Brassica species (Zhang et al. 2015). White petals of chrysanthemum turned yellow after RNAi construct of CmCCD4a was introduced (Ohmiya et al. 2006). However, the mechanisms that control carotenoid catabolism in R. molle flower are far from conclusive. Therefore, the estimation of candidate enzymes in the biosynthesis of carotenoids will facilitate further research on the underlying mechanisms. In this study, 32 unigenes belonging to the carotenoid synthesis were discovered. According to the sequencing result, the most highly expressed genes related to carotenoid biosynthesis were ZDS, followed by PSY and PDS while no unigene related to CCD4 was detected. In addition, the qRT-PCR results showed that the expression levels of PSY and PDS were low during the early flowering stage, and increased as the flowers developed. The expression level of LCYB was lower than LCYE during the early flowering stage, but higher than LCYE during the later stages. This result was consistent with the changes of carotenoid content in the flower petals of R. molle during the flower development. The content of β-carotene was much higher than the other carotenoids during the flower development, which declined at the stage S2 and rose to the highest at stage S4. The α-carotene concentration remained low throughout flower development. The identification of enzymes and related genes involved in the carotenoid biosynthesis and metabolic pathway would facilitate both color-regulating mechanisms and molecular breeding research in rhododendron.

SSRs were developed as co-dominant and multi-allelic, powerful molecular markers for comparative genetic mapping and genotyping since they are co-dominant, multi-allelic, highly polymorphic, locus-specific, and transportable among species within genera (Bruni et al. 2012; Varshney et al. 2005a, b). Six SSRs have been successfully used to analyze the genetic structure of rhododendron hybrid population for better characterize the evolution of hybridization events (Bruni et al. 2012; Gismondi and Canini 2013). However, only a few SSRs were developed in rhododendron species due to deficient in genomic information. Therefore, it is important to develop more SSR markers. In this study, a total of 8266 SSRs from the transcriptome of R. molle flower were identified. The dinucleotide repeat motifs were the most frequent SSR motifs. The TC/GA/AG/CT motifs were the most frequency motifs, accounting for 92.11% of the total dinucleotide SSRs. Among the di-nucleotide repeat motifs, GA repeats were the most common, which is different from that of Arabidopsis in which AG repeats were the most frequent (Lawson and Zhang 2006). It may due to the introduction of additional repeats during chromosome replication (Eckert et al. 2002). Furthermore, (GA)n serve as regulatory elements that contains a series of overlapped GAG motifs involved in light regulation (Liu et al. 2012). The unigenes obtained from R. molle transcriptome have provided a valuable genetic resource for SSRs mining and applications in gene mapping, cultivar fingerprinting, and selection of desirable genotypes in rhododendron breeding programs.

We conducted Illumina sequencing to profile the transcriptome of R. molle. To the best of our knowledge, this study presents the first next-generation sequencing of R. molle. The resource information would lay an important foundation for future genetic or genomic studies on rhododendron species. Meanwhile, a large number of putative R. molle homologs that are probably involved in carotenoid biosynthesis pathway were identified. We have identified a large amount of SSRs that are ready for marker development.