Metabolite analysis reveals flavonoids accumulation during flower development in Rhododendron pulchrum sweet (Ericaceae)

Plant material

The flowers of Rhododendron pulchrum sweet used in this study were harvested from Huanggang Botanical Garden in Hubei Province, China. The flowers of different developmental stages were respectively collected on Marth 20^th, April 6^th, and April 22^th, 2018 (Wang et al., 2018). The samples (at least six flowers of each sample) were immediately frozen in liquid nitrogen and stored at −80 °C for subsequent analysis. There were two biological replicates per sample.

Metabolite extraction

Samples of R. pulchrum flowers at three stages of development were selected for metabolomic analysis using ultra-performance liquid chromatography electrospray ionization tandem mass spectrometry (UPLC-ESI-MS/MS) system. The freeze-dried samples were ground to powder. Metabolites were extracted from 100 mg powder using 1.0 mL of 70% aqueous methanol solution overnight at 4 °C. The extracted solution was then filtered through a 0.22 μM microporous membrane and injected into LC-MS vials.

Metabolic profiling of flavonoids

Exactly 5 μL of the sample were injected and analyzed using a Shim-pack UFLC SHIMADZU CBM30A system (Kyoto, Japan). Chromatographic separation conditions were as follows: column, Waters ACQUITY UPLC HSS T3 C18 column; the temperature of column oven, 40 °C; solvent system, A (0.04% formic): B (acetonitrile with 0.04% formic); flow rate, 0.4 mL/min. The eluting gradient program consisted of 0–11 min, 95–5% A; 11–12 min, 5% A; 12–12.1 min, 5–95% A; 12.1–15 min, 95% A. The QC sample was prepared by mixing aliquots of all the sample extracts and was injected after every set of 10 samples.

The effluent was connected to a triple quadrupole-linear ion trap mass spectrometer, the AB4500 Q TRAP system, and controlled by Analyst 1.6.3 software. The effluent was connected to a triple quadrupole-linear ion trap mass spectrometer, the AB4500 Q TRAP system, and controlled by Analyst 1.6.3 software. Metabolites qualitative and quantitative analysis followed the methods of Chen et al. (2013). Based on the self-built database (Metware Biotechnology Co. Ltd., Wuhan, China), high-through quantification of metabolites was carried out by multiple reaction monitoring (MRM) for widely targeted metabolomics analysis. Based on the fragmentation pattern, retention time (RT), and mass-to-charge-ratio (m/z) values, metabolites in Metware’s database were annotated by comparing with that of commercial standards or purified compounds, or searching the public databases, including MassBank, KNApSAcK, HMDB, MoTo DB, and METLIN. The QQQ scans were acquired as MRM experiments with the collision gas (nitrogen) set to 5 psi. Declustering potential (DP) and collision energy (CE) for each precursor-product ion (Q1–Q3) were done with further DP and CE optimization.

Data analysis

To investigate the flower development-controlled accumulation of flavonoids, hierarchical clustering analysis (HCA), principal component analysis (PCA), and partial least squares-discriminate analysis (PLS-DA) were conducted. PCA and PLS-DA were performed using the SIMCA-P version 14.0 software. For HCA and PCA, the metabolite data were log2-transformed and followed by a min-max normalization. The heatmap was generated by the “heatmap.2” function in the “gplot” R-package. Differentially accumulated metabolites (DAMs) were identified based on the thresholds of log2 (fold change) ≥1 and variable importance for the projection (VIP) ≥1 in PLS-DA. One-way ANOVA was used in this study (p ≤ 0.5).

Transcriptome data were derived from previously published studies (Wang et al., 2018). We mapped the differentially expressed genes and DAMs simultaneously to the KEGG pathway database. Canonical correlation analysis (CCA) was carried out using the Vegan R package.

Flavonoid accumulation of R. pulchrum

Liquid chromatography tandem mass spectrometry (LC-MS) based metabolome analysis of R. pulchrum flowers at three stages (bud stage, partially open flower stage, and fully open flower stage), was performed to investigate the changes in flavonoid accumulation (Fig. 1). A total of 199 flavonoids, including 57 flavone, 42 flavonol, 32 flavone C-glycosides, 21 flavanone, and 12 isoflavone, were detected in our study (Table S1). To further investigate flavonoid accumulation in the different development stages, the flavonoid profile was statistically analyzed by hierarchical clustering analysis (HCA). Samples were divided into three groups according to the development stage, indicating that the accumulation of flavonoids displayed a clear development stage variation in terms of their abundance in different stages. In the fully open flower stage, almost half flavonoids reached the highest levels, followed by the bud and partially open flower stage (Fig. 2A).

We also analyzed the metabolite accumulation patterns by principal component analysis (PCA). The PCA results also showed a clear grouping of the metabolites into three distinct groups (Fig. 2B). The three main PCs (PC1, PC2, and PC3) accounted for 86% of the total system variability. The variables that also contributed to the PCs are listed in Table S2. The first component (PC1, 49.24%) separated the fully open flower stage from the bud and partially open flower stages, whereas the second component (PC2, 22.21%) separated the bud stage from the partially open flower and fully open flower stage, reflecting a major difference in metabolite levels among these three stages.

Developmentally-controlled flavonoids in R. pulchrum

To determine the flavonoid accumulation patterns of R. pulchrum in flower development, a comparative metabolite analysis was performed. To identify flavonoids that mainly contribute to the separation of the different flower stages, partial least squares discriminant analysis (PLS-DA) was conducted, and the variable importance for the projection (VIP) values were used to identify the differentially altered metabolites (DAMs). Based on fold change (FC) ≥2 or ≤0.5 and VIP ≥ 1, we identified a total of 25 DAMs between the bud and partially open flower stage. 63 DAMs between and fully open flower stage, and 60 DAMs between bud and fully open flower stage. Compared with the bud stage, there were 14 DAMs upregulated and 11 downregulated in the partially open flower stage. From the partially open flower stage to the fully open flower stage, 49 DAMs were upregulated and 11 DAMs were downregulated (Fig. 3A). The numbers of metabolites upregulated were comparable to or higher than the number of downregulated metabolites. Most anthocyanins, flavones, flavone C-glycosides, and flavonols were significantly higher at the fully open flower stage than the bud stage including delphinidin, pelargonin, chrysoeriol O-glucuronic acid, and kaempferol, indicating developmentally-controlled flavonoids accumulation in R. pulchrum.

Furthermore, the developmentally-controlled accumulation pattern of DAMs was compared among the three stages, and the results of clustering analyses showed that the metabolites were grouped into three clusters (Fig. 3B). A deeper analysis of the metabolites in cluster 2 showed that anthocyanins, flavone, flavone C-glycoside, and isoflavone accumulated at higher levels at stage than at other two stages.

To visualize the differences during flower development, the 78 DAMs were presented in Venn diagrams (Fig. 4A, Table S3). Several DAMs were shared between two comparisons: 12, 15, and 54 were shared between the bud/early and early/full, bud/early and early/full, and bud/early and bud/full comparisons, respectively. A total of 11 DAMs, including six flavones, three flavonols, one isoflavone, and one flavone C-glycoside, were in common during flower development. Among the 11 DAMs, the accumulation of niene metabolites significantly increased in R. pulchrum from bud to partially open flower to fully open flower stage and displayed the highest accumulation levels in the fully open flower stage, including 3,7-di-O-methyl quercetin, acacetin O-glucuronic acid, chrysoeriol O-hexosyl-O-pentoside, biochanin A, chrysin, chrysoeriol 6-C-pentosyl-O-rutinoside, velutin, amentoflavone and kumatakenin (Fig. 4B). Chrysoeriol O-glucuronic acid and isorhamnetin were downregulated in the partially open flower stage and then upregulated in the fully open flower stage.

Differences in development stages may result in different accumulations of flavonoids. Nine metabolites specifically exhibited a significant difference accumulation between the bud and partially open flower stage, indicating that these metabolites may play roles in azalea from the bud to partially open flower stage (Fig. 4C). Syringetin, rosinidin O-hexoside, prunetin, glycitin, kaempferol 3-O-robinobioside (Biorobin), and nobiletin were down-regulated in the partially open flower stage, while C-hexosyl-apigenin O-hexosyl-O-hexoside, chrysoeriol 6-C-hexoside, and kaempferol 3-O-rutinoside (Nicotiflorin) were upregulated in the partially open flower stage.

During the flowering of R. pulchrum, five metabolites including tricin 5-O-feruloylhexoside, isosakuranetin-7-neohesperidoside (Poncirin), tricin O-saccharic acid, naringenin chalcone, and cyanidin, were characterized by a change exclusively in comparison between partially open flower and fully open flower stage, which may play roles in azalea from partially open flower stage to fully open flower (Fig. 4D).

Differentially altered flavonol during flower development

Multiple functional roles of flavonoids in plant development have been reported. Flavonol is a major class of UV-absorbing compounds in plants, which are present in the upper epidermis of plant organs consistent with their role in UV protection. Copigmentation is an association between flavonols and anthocyanin pigments. The 78 DAMs identified in R. pulchrum during flowering included 17 flavonols, 23 flavones, 12 flavone C-glycoside, eight isoflavones, nine flavanones, and seven anthocyanins. From the bud to the partially open flower stage, eight flavonols were differentially accumulated. The content of 3,7-di-O-methyquercetin and chrysoeriol 6-C-pentosyl-O-rutinoside was 425 and 152-fold higher in the partially open flower stage, respectively (Fig. 5A).

During R. pulchrum flower development, significant differences were observed between the fully open flower stage and the other stage. The top ten elevated metabolites between the partially open flower stage and fully open flower stage included three flavonols: isorhamnetin, quercetin-3,4′-O-di-beta-glucopyranoside, and kumatakenin, which was 30, 21, 15-fold higher in the fully open flower stage. In addition, the top five reduced metabolites include three flavonols: quercetin 5-O-malonylhexosyl-hexoside, syringetin 3-O-hexoside, and myricetin (Fig. 5B).

Identification of candidate enzymes involved in isoflavone biosynthesis during R. pulchrum flower development

To further explore the molecular mechanisms of flavonoid biosynthesis during flower development, integrative analyses of metabolomic and transcriptomic were conducted. The impacts of flower development on gene expression have been studied by comparative transcriptome before (Wang et al., 2018). The transcriptomic data were used to explore changes in flavonoids biosynthesis. We examined the expression of genes and the accumulation of metabolites during flower development. Based on the KEGG pathway assignment, the isoflavone and flavonol biosynthesis pathway in R. pulchrum was constructed (Fig. 5). Significantly differential unigenes encoding putative enzymes involved in flavonoid biosynthesis have been shown. As illustrated in Fig. 6, transcription analysis showed that shikimate O-hydroxycinnamoyl transferase (HCT) and caffeoyl-CoA O-methyltransferase unigenes were down-regulated during flower development. While the expression level of naringenin 3-dioxygenase and flavonoid 3′-monooxygenase (CYP75B1) increased significantly during flower development. In addition, naringenin chalcone, biochanin A, kaempferol, and myricetin contents were highly accumulated in the fully open flower stage, whereas the contents of naringin, and 2-hydroxy genistein showed the highest level in the bud stage.

In particular, gene-metabolites networks analysis showed that the expression of HIDH (2-hydroxyisoflavanone dehydratase, K13258, TRINITY_DN75801_c0_g1) was 18-fold higher at the fully open flower stage than bud stage, a result consistent with the high content of biochanin A, while the production of hydroxylation modification of genistein, 2′-hydroxygenistein was significantly downregulated in the fully open flower stage. The expression of HIDH (TRINITY_DN66446_c1_g3) was five fold change at the partially open flower stage than bud stage, which is likely to participate in the higher accumulation of prunetin in the partially open flower stage (Fig. 6 and Fig. S1). These findings suggest that HIDH encodes function dehydratase, which is a critical determinant of isoflavone productivity during R. pulchrum flower development.