Integration of metabolome and transcriptome analyses reveals the mechanism of anthocyanin accumulation in purple radish leaves

Abstract

Anthocyanins are natural pigments and play significant roles in multiple growth, development, and stress response processes in plants. The vegetables with high anthocyanin content have better colours, higher antioxidant activity than green vegetables and are potent antioxidants with health benefits. However, the mechanism of anthocyanin accumulation in purple and green leaves of Raphanus sativus (radish) is poorly understood and needs further investigation. In the present study, the pigment content in a green leaf cultivar “RA9” and a purple-leaf cultivar “MU17” was characterized and revealed that the MU17 had significantly increased accumulation of anthocyanins and reduced content of chlorophyll and carotenoid compared with that in RA9. Meanwhile, these two cultivars were subjected to a combination of metabolomic and transcriptome studies. A total of 52 massively content-changed metabolites and 3463 differentially expressed genes were discovered in MU17 compared with RA9. In addition, the content of significantly increased flavonoids (such as pelargonidin and cyanidin) was identified in MU17 compared to RA9 using an integrated analysis of metabolic and transcriptome data. Moreover, the quantitative real-time polymerase chain reaction results also confirmed the differences in the expression of genes related to pathways of flavonoids and anthocyanin metabolism in MU17 leaves. The present findings provide valuable information for anthocyanin metabolism and further genetic manipulation of anthocyanin biosynthesis in radish leaves.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12298-022-01245-w.

Introduction

Anthocyanins are the most important plant natural flavonoid pigments, conferring purple, red, and blue colors to plant tissues, such as leaves, stems, roots, flowers, and fruits (Passeri et al. 2016). It has been well documented that flavonoids play essential roles in a variety of plant biology, such as UV protection, pigmentation, and plant defence (Harborne and Williams 2000). More importantly, anthocyanins have attracted considerable interest due to their antioxidant properties and health benefits, which reduce cardiovascular diseases and cancer risks (Butelli et al. 2008; Toufektsian et al. 2008). Therefore, cultivating colored vegetable varieties with high levels of anthocyanins has higher commercial value than the colorless cultivars.

In the flavonoid metabolic pathway and the anthocyanin biosynthetic pathway, anthocyanins are one of the major products which has been well studied and illuminated (Zhang et al. 2019). Briefly, the phenylalanine ammonia-lyase (PAL) which has ammonia-lyase activity catalyzed synthesis of cinnamic acid represents the initial reaction step (Vogt 2010). Then, several enzymes are evolved in sequential reactions in anthocyanin biosynthesis, such as the chalcone synthase (CHS), chalcone isomerase (CHI), dihydroflavonol 4-reductase (DFR) and anthocyanidin synthase (ANS) (Shirley 1996; Grotewold 2006; Guo et al. 2014). As a result of glycosylation, unstable anthocyanidins catalyzed by ANS could be converted into stable anthocyanins immediately. Then, these stable anthocyanins will be transported into vacuoles from the cytosol (Koes et al. 2005). In addition to pH, metal cations, co-pigmentation, and modified backbones, anthocyanins exhibit a variety of colors due to their glycosylated form (Hichri et al. 2011).

The anthocyanins biosynthesis in plants is heavily regulated by various transcription factors at multiple levels, such as R2R3 MYB, bHLH, WRKY, bZIP and MADS-box family members (Gonzalez et al. 2008; Jaakola et al. 2010). At the molecular level, a transcriptional activator MBW complex (MYB-bHLH-WD40) consisting of R2R3MYB, bHLH, and WD40 regulates the activation of genes associated with late anthocyanin biosynthesis in plants (Zhao et al. 2013). Among the regulatory proteins, the R2R3 MYB genes play an important regulatory role in influencing the transcriptional level of the anthocyanin biosynthetic genes at the early biosynthesis stage (Zhao et al. 2013; Niu et al. 2010). In general, the bHLH transcriptional factors regulate the structural genes responsible for anthocyanin biosynthesis (Zhang et al. 2019; Tang et al. 2020). However, WD40 proteins can’t directly activate the transcript of anthocyanin biosynthesis genes, although it has been shown that the WD40 proteins regulate the expression of anthocyanin biosynthesis genes by interacting with bHLH and MYB and forming the MYB-bHLH-WD complex (An et al. 2012). Meanwhile, it has been well identified and characterized that several catalysts, such as phenylalanine ammonia-lyase, flavonoid 3′-hydroxylase, dihydroflavonol 4-reductase, UDP-flavonoid glucosyl transferase are involved in the biosynthesis of anthocyanins (Schaart et al. 2013; Liu et al. 2016).

Identifying the metabolic pathway and the regulation of anthocyanin biosynthesis is crucial for enabling us to develop anthocyanin-enriched vegetables. Recently, a number of studies have revealed how anthocyanins contributes to color variation in plant tissues during the different growth stages (Zhang et al. 2019; He et al. 2020; Li et al. 2020). Meanwhile, several transcription factors and structural genes have been reported to be involved in the anthocyanidin biosynthetic pathway together controlled the anthocyanin accumulation in leaves of sweet potato and purple bok choy (Zhang et al. 2014; Li et al. 2019). Radish belongs to the Brassicaceae family and it is widely cultivated worldwide (Wang et al. 2015). Recently, a pan-genome of Raphanus has been sequenced and de novo assembled covering most of domesticated, wild and weedy radish varieties (Zhang et al. 2021). Currently, a study on anthocyanin components and the potential molecular mechanism associated with radish taproots have been reported (Muleke et al. 2017). The fresh radish leaves could be used to make the sauerkraut with a rich flavor and taste. Homoplastically, they could be processed into high nutritional fodder for livestock. Consequently, the purple radish leaves with anthocyanin may have high antioxidant activity, high benefit and will have a potential applied value. However, it is still unclear how anthocyanin is biosynthesized in the different color variants of radish leaves even the pan-genome of Raphanus and accumulation of anthocyanin in radish taproots have been reported (Zhang et al. 2021; Muleke et al. 2017).

In present study, the fluctuations in the gene expression level and alterations in metabolite levels in radish leaves were investigated by transcriptome and metabolomic analysis, respectively. The metabolic pathways and key differentially expressed genes (DEGs) on the anthocyanin biosynthesis pathway were identified. Thus, the study results provide a broad perspective for further functional characterization of the underlying mechanisms of anthocyanin biosynthesis in radish leaves.

Materials and methods

Plant materials

Two radish cultivars: “MU17” with purple leaves and “RA9” with green leaves were selected for this study. The radish leaves of different developmental stages were harvested from the experimental farm of Nanchong Academy of Agricultural Sciences (Sichuan Nanchong, China). Leaves at three stages, i.e., 15 days after seeding, 30 days after seeding, and 45 days after seeding, were collected. At each sampling point, the leaves from about 10 plants were collected and sampled. As soon as the leaves were collected, they were immediately frozen using liquid nitrogen, and transferred to a freezer at − 80 ℃ for store until using.

Determination of pigment content

Anthocyanins were extracted from radish samples collected as described above. Briefly, with liquid nitrogen, frozen samples were ground into powder. A separate extraction with 4 ml of methanol–HCl extraction solution (80% methanol/37% HCl/pure water = 160:27:13, v/v/v) was performed on 1 g powder, and extraction was continued overnight at 4 °C. After centrifuged at 12,000 g and 4 °C for 10 min, the supernatants were analyzed for absorbance using a VIS spectrophotometer at 530 nm, 620 nm, and 650 nm (A₅₃₀, A₆₂₀, A₆₅₀) (Lambda 850 + UV/Vis, Perkin Elmer). The total anthocyanin content was defined from the equation:

$$\text{Total}\text{anthocyanin}\text{content}\left(\text{nmol}{g}^{-1}\right)=\frac{\left(\text{A}530-\text{A}620\right)-0.1\times (\text{A}650-\text{A}620)}{4.62}\times \frac{V}{m}$$

in which the V is the volume of the extraction solution used for anthocyanin in terms of milliliters, and the m is the weight of the sample used in grams (Gou et al. 2011). The chlorophyll and carotenoids were extracted from frozen leaf samples. In brief, samples of known weight were ground into powder in low temperature condition and then placed into 80% acetone. Chlorophyll was extracted overnight in the darkroom and the absorbance measurement was performed at wavelengths of 470 nm, 663 nm and 645 nm (A470, A663, and A645) using a spectrophotometer (Lambda 850 + UV/Vis, Perkin Elmer). The total chlorophyll content was defined from the equation: 

$$\text{Total}\text{chlorophyll}\left(\mathrm{\mu }\text{g}{\text{ml}}^{-1}\right)=20.2\times \text{A}645+8.02\times \text{A}663$$

in which the A645 and A663 represent the absorbance of samples at 645 and 663 nm wavelength (Arnon 1949). The total carotenoid content was defined from the equation:

$$\text{Total}\text{carotenoid}\text{content}\left(\mathrm{\mu }g{\text{g}}^{-1}\right)=\frac{1000\times \text{A}470-1.82\times \text{ChlA}-85.02\times \text{ChlB}}{198\times m},$$

in which m represents the weight of the sample used in gram (Vis 2001), and the ChlA (chlorophyll A) and ChlB (chlorophyll B) content were calculated using the following equation:

$$\text{ChlA}\text{content}\left(\mathrm{\mu }\text{g}{\text{ml}}^{-1}\right)=12.7\text{A}663-2.69\text{A}645.$$

$$\text{ChlB}\text{content}\left(\mathrm{\mu }\text{g}{\text{ml}}^{-1}\right)=22.9\text{A}645-4.68\text{A}663.$$

Metabolite extraction

The freeze-dried leaf samples (~ 30-day-old) were ground into powder. For each sample, a 1000 µl methanol/water mixture (3:1, v:v) was used to extract 10 mg powder. After vortexed for 30 s, samples were ultra-sonicated for 15 min in an ice bath, followed by overnight shaking at 4 °C. Afterwards, supernatants were obtained by centrifuging extracted samples at 12,000 rpm and 4 °C for 15 min. As a final step, the supernatants were transferred into 2 ml glass vials and stored in a freezer of − 80 °C until the UHPLC-MS/MS was performed. An equal aliquot of supernatants from each sample was used to prepare the quality control sample (QC).

UHPLC-MS analysis

The UHPLC-MS metabolite profiling was detected using an ACQUITY UPLC HSS T3 column (100 × 2.1 mm, 1.8 μm) (Waters) and the QTOF MS system (AB Sciex). Besides, 0.1% formic acid was used as mobile phase A, and the acetonitrile was used for mobile phase B. Elution profile was used as follows: 98:2 vA/vB at 0 min, 50:50 vA/vB at 10 min, 5:95 vA/vB at 11 min, 5:95 vA/vB at 13 min, 98:2 vA/vB at 13.1 min, and 98:2 vA/vB at 15 min. The column temperature was 40 °C and the flow rate held at 0.4 ml min^− 1. An auto-sampler temperature of 4 °C and injection volume (2 µl) were used. The effluent was connected to QTOF mass spectrometer using electrospray positive ion and negative ion mode: ion spray voltage: + 5500/− 4500 V, curtain gas: 35 psi, temperature: 600 °C, declustering potential: ± 100 V, and the ion source gas I and gas II were set at 60 psi.

Metabolome data preprocessing and annotation

The MS raw data were processed and the metabolite identification was carried out using an inhouse MS2 database based on the preprocessing results. And the MRM data were calculated with Skyline software. The metabolites with Variable Importance in the Projection (VIP) > 1 and Student’s t-test P value < 0.05 were analyzed in the comparison between RA9 and MU17 samples.

RNA-seq, annotation and data analysis

For the RNA-seq analysis, the fresh leaves from at least 10 plants of each variety were collected and pooled. RNA-seq was performed according to the previously described method (Chai et al. 2017). In brief, the total RNA of radish leaves for RNA-seq was extracted using RNAiso (Takara, Dalian, China). RNA integrity was assessed in a Bioanalyzer 2100 system (Agilent Technologies, USA) and 1% agarose gel electrophoresis. Separating mRNA from total RNA was carried out with magnetic beads containing oligo (dT). The first cDNA strand was synthesized using 6 nt random primers and linking the sequencing adapter to both ends. The library quality was checked using Agilent Bioanalyzer 2100 system and paired-end reads sequencing was performed on an Illumina HiSeq 2500 platform. The HISAT2 package was used to map clean reads for the  radish reference genome (https://www.ncbi.nlm.nih.gov/genome/?term=radish). Then, FPKM (Fragments Per Kilo bases per Million fragments) was normalized to the read counts and DEGs were identified by |log₂ (fold change) | ≥ 1 and student t-test P ≤ 0.05. All DEGs were enriched by Gene Ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis (Gotz et al. 2008).

qRT-PCR analysis

Frozen radish leaves were ground into a fine powder in liquid nitrogen. Subsequently, the total RNA of each samples was independently extracted using RNAiso Plus (Takara, China) and reverse transcription of the the first cDNA strand was performed with M-MLV (Promega, China) including 2 µg of RNA in 25 µl reaction and diluted 3-times with nuclease-free water. The qRT-PCR was performed using a CFX96™ Real-Time system (Bio-Rad, USA), while the reaction was performed using GoTaq qPCR Master Mix (Promega, China) with 1 µl 20 ng µl^− 1 cDNA in 20 µl reaction following the product recommend protocol. For gene expression analysis, the RNA polymerase-II (RPII) was used as an internal reference and 2^−∆∆Ct method was used for calculation of the expression level. All qRT-PCR tests were performed in three replicates. All primer information is listed in Table S7.

Statistical analysis

For statistical significance analysis, an one-way analysis of variance (ANOVA) was conducted using IBM SPSS statistical software 20.0 (SPSS Inc, USA). Statistical data are presented as the mean ± standard error (SE).

Results

Phenotype and physiological characterization of the RA9 and MU17 radish plants

In this study, two radish varieties RA9 and MU17 were cultivated and investigated. The seedlings of radish RA9 had green leaves with purple petioles while the MU17 seedlings had purple leaves and petioles (Fig. 1). The MU17 had reduced plant height and width compared with the RA9 in 15 days, 30 days, and 45 days old seedlings (Table S1). Since the significant variation in leaf color in these plants was observed, the pigments of their leaves were further investigated. As shown in Fig. 2, the MU17 leaves had higher anthocyanin content compared to the RA9 leaves. However, the MU17 leaves had reduced the total carotenoids and chlorophyll content compared with the RA9 leaves. These results indicated that the significantly increased anthocyanin content should be responsible for the purple leaves in MU17 radish.

Fig. 1.

Phenotypes of leaves from the RA9 and MU17. Bars = 1 cm

Fig. 2.

The anthocyanin, total carotenoids and chlorophyll content in the RA9 and MU17. Data are the mean ± SE of three independent experiments. Asterisks indicate significant differences from the RA9 versus MU17 (P < 0.05)

Analysis of metabolic profiling

In order to analyse the metabolites from different varieties, the leaf samples of MU17 and RA9 were subjected to UHPLC-ESI-MS/MS. In this study, a total of 113 metabolites were found in MU17 and RA9 leaves (Table S2). Principal components analysis indicated that the RA9 and MU17 were clearly separated within the PCA score plots (Fig. 3A). The volcano plot analysis of metabolite content revealed significant differences between the MU17 and RA9 varieties (Fig. 3B). Among them, 27 (23.9%) different metabolites were markedly higher and 25 (22.1%) were lower in MU17 compared to those in RA9. The metabolite content of these varieties significantly changed and showed different patterns in the five investigated RA9 samples and five MU17 samples (Fig. 3C).

Fig. 3.

Metabolomic analysis of the leaves in the RA9 and MU17. A Score scatter plot of PCA model for group RA9 versus MU17. B Volcano plot for group RA9 versus MU17. C Heatmap of hierarchical clustering analysis for group RA9 versus MU17. D Pathway analysis for group RA9 versus M17. E Network analysis for group RA9 versus MU17

Metabolomic analysis of anthocyanin-related metabolite content

A significant metabolic difference was detected between RA9 and MU17. KEGG analysis of these differential metabolites showed that they were enriched for 5 pathways, including phenylalanine metabolism, flavonoid biosynthesis, cysteine and methionine metabolism, phenylpropanoid biosynthesis, and glycerophospholipid metabolism (Fig. 3D). The content of phenylethylamine and sinapic acid which were involved in phenylalanine metabolism and phenylpropanoid biosynthesis pathway was significantly higher in purple line MU17 than that in RA9. Meanwhile, compounds of the flavonoid biosynthesis pathway had a significant difference in these tissues. The MU17 showed a higher accumulation of anthocyanins (pelargonidin and cyanidin) and isoflavonoid (genistein) compared to the RA9 (Table 1). However, it had a significantly lower content of flavonol kaempferol derivatives (Table 1). The changes for 6-Caffeoylsucrose, chalconaringenin, flavone, flavonol, and anthocyanin indicated a kinetic process for anthocyanin accumulation in MU17 leaves. Based on the metabolite difference between MU17 and RA9, a metabolic network indicating the compounds involved in pathways and the potential enzyme-catalyzed reactions were constructed (Fig. 3E, Table S3). The network showed that the five KEGG-enriched metabolic pathways were all linked together by the enzyme factors. These results suggested that the anthocyanin accumulation process was related to the expression of genes closely related to flavonoid metabolism.

Table 1.

Anthocyanins detected with significantly changed content between two radish varieties (MU17 and RA9)

Compound	m/z	Log2(Fold change)	VIP(Importance in the Projection)
Phenylethylamine	122.10	0.711	1.26
6-Caffeoylsucrose	505.15	− 1.012	1.33
m-Coumaric acid	165.05	− 1.522	1.39
Chalconaringenin	271.06	− 1.334	1.42
Camellikaempferoside B	887.25	− 3.115	1.40
Dihydrowogonin 7-O-glucoside	449.13	− 0.701	1.31
Caffeic acid O-glucoside	341.08	2.568	1.42
Quercetin-3-O-beta-glucopyranosyl-7-O-alpha-rhamnopyranoside	609.15	1.258	1.19
Kaempferol 3-(6’’-caffeoylglucosyl)-(1->4)-rhamnoside	757.19	− 2.776	1.43
Kaempferol 3-O-arabinoside	419.09	− 2.209	1.45
Kaempferol 3-rhamnosyl-(1->4)-xyloside	565.15	− 3.241	1.45
Kaempferol-3-Galactoside-6’’-Rhamnoside-3’’’-Rhamnoside	741.22	− 3.987	1.01
Kaempferol-3-O-alpha-L-arabinoside	417.08	− 1.399	1.41
Kaempferol-3-O-arabinopyranoside	441.07	− 1.226	1.40
Kaempferol-3-Rhamnoside-4’’-Rhamnoside-7-Rhamnoside	725.23	− 1.307	1.43
Kaempferol-7-neohesperidoside	595.16	3.845	1.44
Pelargonidin	271.06	4.593	1.41
Cyanidin	287.05	1.068	1.42
Genistein	271.06	4.218	1.41

Transcriptome analysis of the purple (MU17) and preen (RA9) radish cultivars

To better understanding the molecular basis for anthocyanin biosynthesis and metabolism, the third leaves from 30 days-old plants were used for comparative transcriptome analysis. A total of 29,754 to 36,555 genes were detected in the RA9 and MU17 leaves with FPKM > 0 (Fig. 4A, Table S4). As a results of the principal component analysis (PCA), the purple and green radish cultivars for the same line showed a distinct separation (Fig. 4B). In addition, three biological replicates were applied for RNA-Seq and all biological replicates showed correlation coefficients over 0.9 (Fig. 4B, Figure S1), demonstrating that the RNA-seq data for the three biological replicates were reproducible and consistent (Fig. 4C). Moreover, significance level of P-value < 0.05 was applied to the transcript levels of genes identified by RNA-sequencing. As compared the RA9 and MU17, 1456 DEGs were down-regulated in MU17 leaves and 2007 DEGs had up-regulated expression in MU17 leaves (Fig. 5A, B). Comparatively, 346 DEGs were clustered according to their expression patterns into nine clusters, with similar expression trends found within each cluster (Figure S2).

Fig. 4.

Transcriptome analysis of the leaves in the RA9 and MU17. A Numbers of the expressed genes in differential leaves. B Correlation analysis of the differential leaves. C Principal component analysis (PCA) of miRNA profiles

Fig. 5.

RNA-seq analysis of the leaves in the RA9 and MU17. A Heat map analysis of DEGs of the RA9 and MU17. B The volcano diagram indicating the common DEGs in the RA9 and MU17. C Gene Ontology (GO) classification analysis of the main up- and down-regulated DEGs. D KEGG pathway enrichment analysis of the main up- and down-regulated DEGs in MU17. ‘Rich factor’ indicates the ratio of the number of DEGs associated with one KEGG pathway to the total number of all DEGs.

GO and KEGG enrichment analysis

GO analysis was used to identify the functions of the DEGs (Fig. 5C, D, Table S5). A total of three main categories of DEGs were identified, including molecular function, cellular component, and biological process. For cellular component, many of DEGs were enriched in the ribosomal subunit. For molecular function, many of the DEGs were connected to oxidoreductase and ammonia-lyse activity. For biological process, many DEGs were enriched in the oxidation-reduction process (Fig. 5C). Meanwhile, KEGG analysis exhibited that most of the DEGs were enriched in pathways related to ribosome and phenylpropanoid biosynthesis (Fig. 5D, Table S5). In addition, the pathways closely related to flavonoid-anthocyanin biosynthesis, such as flavonoid biosynthesis, phenylpropanoid biosynthesis, flavone and flavonol biosynthesis, phenylalanine metabolism, were enriched in the purple MU17 radish, which is consistent with the high accumulation of anthocyanins in MU17.

Mapping of DEGs and metabolites involved in flavonoid-anthocyanin biosynthesis

According to the transcriptome analysis of the purple (MU17) and green (RA9) radish cultivars, 1456 DEGs were significantly down-regulated and 2007 DEGs were up-regulated in MU17 leaves (Fig. 5B). Since the MU17 had a higher accumulation of anthocyanins in its leaves (Figs. 1, 2A), while the biosynthetic pathway for flavonoid-anthocyanins in plants has been well-characterized (Jaakola 2013), the DEGs involved in the flavonoid-anthocyanin metabolic pathway were analyzed based on the gene expression pattern in the MU17 leaves. As shown in Table 2, more than 40 DEGs that participated in flavonoid-anthocyanin biosynthesis were enriched in MU17 leaves compared with the green RA9 lines. The genes coding flavonoid-anthocyanin biosynthetic enzymes, such as Chalcone synthase, Dihydroflavonol-4-reductase, chalcone-flavonone isomerase and Leucoanthocyanidin dioxygenase, had a significantly high expression level in MU17 compared with that in RA9 (Table 2). Consistent with the expression of flavonoid-anthocyanin biosynthesis genes, the metabolites were enriched in phenylalanine metabolism, phenylpropanoid biosynthesis, and flavonoid biosynthesis pathways which all had a close relation with the biosynthesis of flavonoid and anthocyanin (Fig. 6A). Meanwhile, the comparison of the changed metabolites and corresponding genes revealed that 160 compounds and genes had significantly changed content or expression level (P < 0.05) (Table S6). In these selected compounds and genes, 11 genes with significantly changed expression in MU17 and their related metabolites showed a consistent or opposite trend (Fig. 6B). For instance, the 4-coumarate-CoA ligase 3 (LOC108817856), flavone 3′-O-methyltransferase 1 (LOC108821435), and indole glucosinolate O-methyltransferase 1 (LOC108825157) had the same trend with the accumulation of sinapic acid in MU17. However, the reduced content of chalconaringenin was contrary to the higher-level expression of chalcone-flavonone isomerase 3 (LOC108814129) and chalcone synthase 3 (LOC108814337). These suggested the complicated regulatory network for the homeostasis of biosynthesis of anthocyanin and the expression of genes involved in this process.

Table 2.

The DEGs related to anthocyanin biosynthesis in the MU17 and RA9

Pathway	Gene	Log2FC	Q value	Annotation
Anthocyanin biosynthesis	LOC108821920	6.95	3.44E−162	Anthocyanidin 3-O-glucoside 2′-O-xylosyltransferase
Flavone and flavonol biosynthesis	LOC108814778	1.02	2.91E−4	Flavonoid 3′-monooxygenase
Phenylalanine metabolism	LOC108848885	1.09	1.81E−8	Trans-cinnamate 4-monooxygenase-like
	LOC108849060	1.09	2.58E−3	Caffeoyl-coa O-methyltransferase 1
	LOC108819811	6.63	8.91E−19	Probable caffeoyl-coa O-methyltransferase
	LOC108847879	3.42	1.66E−81	Trans-cinnamate 4-monooxygenase-like
	LOC108855076	2.01	1.32E−25	Phenylalanine ammonia-lyase 1-like
	LOC108862526	1.62	1.44E−20	Phenylalanine ammonia-lyase 2
	LOC108838754	5.28	1.09E−16	Phenylalanine ammonia-lyase 4-like
	LOC108817856	1.27	3.53E−14	4-coumarate-coa ligase 3
	LOC108857697	4.97	5.92E−28	Phenylalanine ammonia-lyase 4
	LOC108854852	1.13	1.81E−11	Phenylalanine ammonia-lyase 2-like
	LOC108862454	3.48	5.10E−27	4-coumarate-coa ligase 4
	LOC108829906	2.14	8.87E−35	Phenylalanine ammonia-lyase 1-like
Flavonoid biosynthesis	LOC108814337	1.33	7.64E−16	Chalcone synthase 3-like
	LOC108805892	1.73	5.79E−26	Naringenin,2-oxoglutarate 3-dioxygenase
	LOC108826061	6.62	1.09-48	Dihydroflavonol-4-reductase
	LOC108814129	1.23	7.12E−7	Probable chalcone-flavonone isomerase 3
	LOC108848981	6.73	1.33E−227	Leucoanthocyanidin dioxygenase-like
	LOC108843686	6.84	1.22E−75	Leucoanthocyanidin dioxygenase
	LOC108814290	− 1.48	2.87E−13	Flavonol synthase/flavanone 3-hydroxylase
	LOC108848401	− 1.89	4.50E−5	Cytochrome P450 98A3-like
Phenylpropanoid biosynthesis	LOC108857825	7.42	5.82E−32	Peroxidase 3
	LOC108860949	1.01	4.5E−2	Peroxidase 29
	LOC108829918	1.38	2.96E−08	UDP-glycosyltransferase 72E1-like, partial
	LOC108854832	1.09	1.4E−2	Flavin-dependent oxidoreductase FOX2-like
	LOC108823133	4.58	4.9E−2	Peroxidase P7
	LOC108847438	8.27	2E−30	Beta-glucosidase 8-like
	LOC108817684	1.07	8.3E−3	Uncharacterized protein
	LOC108829983	6.15	4.82E−05	Beta-glucosidase 30-like
	LOC108847745	3.37	1.19E−10	Flavone 3′-O-methyltransferase 1
	LOC108844285	2.57	3.31E−2	Peroxidase 22-like
	LOC108808805	5.38	1.1E−3	Beta-glucosidase 28-like
	LOC108817685	− 3.29	1.67E−11	Uncharacterized protein LOC108817685
	LOC108851454	− 1.01	4.57E−3	Indole glucosinolate O-methyltransferase 4-like
	LOC108850268	− 1.36	8.3E−3	Peroxidase 3-like
	LOC108849034	− 5.54	2.93E−09	Flavone 3′-O-methyltransferase 1-like isoform X1
	LOC108850047	− 1.80	6.97E−07	Peroxidase 2-like
	LOC108807220	− 2.47	9.64E−11	Peroxidase 19
	LOC108828710	− 1.08	2.8E−05	Peroxidase 32
	LOC108834421	− 1.01	2.2E−2	Peroxidase 34-like
	LOC108820816	− 1.15	1.25E−4	Peroxidase 66-like

Fig. 6.

Combined analysis of the transcriptomic and metabolomic data. A Enrichment of KEGG terms for the selected DEGs (anthocyanin-associated genes). B Scheme of changes in the anthocyanin-associated genes and metabolites

Expression of genes associated with anthocyanin accumulation

To understand the reason for significant differences in anthocyanin accumulation between the green and purple radish cultivars, the expression of anthocyanin accumulation-related genes were further verified by qRT-PCR. Notably, purple cultivars exhibited increased expression of anthocyanin accumulation-related structural genes (ANS, DFR, CYP73A, CHS, F3H, UGT79B1, PAL4, PXP7, OMT1, IRL-P3 like, and 4CL4) (Fig. 7 A). In addition, the transcript levels of anthocyanin biosynthetic related transcription factors (WRKY44-like, MYB308-like, RAV2, TT8, CPC, bHLH28, MYB114, MYB108-like, MYB305, MYB6 and bHLH93-like) were also up-regulated in the purple cultivar (Fig. 7B). Moreover, the transcriptome data were verified by qRT-PCR using genes related to anthocyanin accumulation. As shown in Fig. 7 C, RNA-seq and qRT-PCR results were consistent.

Fig. 7.

Expression levels of the anthocyanin synthesis-related genes. A Relative expression of anthocyanin synthesis-related structural genes in RA9 and MU17 leaves. B Relative expression of anthocyanin synthesis-related transcription factors in RA9 and MU17 leaves. Data are the mean ± SE of three independent experiments. C Expression levels of twelve genes as determined by qRT-PCR are closely correlated with those according to RNA-seq. The fold change calculated from the qRT-PCR analysis was compared with the fold change from RNA-seq analysis. The R-value represents the correlation coefficient

Discussion

Anthocyanins are important natural pigment, which belongs to a big group of flavonoid compounds (Petroni and Tonelli 2011). At the same time, anthocyanins have been known for a variety of biological functions, such as reducing the harm of UV radiation and improving stress resistance in plant growth and development (Steyn et al. 2002; Nakabayashi et al. 2014). Moreover, anthocyanin containing vegetables are considered highly beneficial due to the health-promoting effects of their constituents (Mink et al. 2007; He and Giusti 2010; Crowe et al. 2011). In the present study, “MU17” was a variety of conventional cultivars and had a purple phenotype due to the abundance of anthocyanins in leaves, while the “RA9” had green leaves with an extremely low content of anthocyanins (Figs. 1, 2A). These results suggested that the anthocyanin accumulation in the MU17 leaves was induced and contributed to the phenotype of purple leaves. In addition, the total carotenoid and chlorophyll contents were significantly reduced in MU17 leaves (Fig. 2B, C), suggesting that the anthocyanin accumulation may affect the total carotenoid and chlorophyll content in purple leaves.

Radish, an important vegetable crop, is widely cultivated worldwide (Wang et al. 2015). To our knowledge, numerous studies on radish taproots have been done and they provide a wide range of nutrients, including fibers, sugars, vitamins, and phytochemicals (Mitsui et al. 2015; Kim et al. 2016). Nevertheless, no study on comprehensive identification of anthocyanin accumulation has yet been conducted in radish leaves. Here, MU17 (a variety of conventional cultivars rich in anthocyanin in the leaves) and RA9 (a radish variety with green leaves) were first subjected to metabolomic analysis using UHPLC-ESI-MS/MS. Volcano plot of the metabolite content revealed significant differences between the two varieties. It has been shown that 113 metabolites were identified in RA9 and MU17 leaves (Table S2). Among these metabolites, 27 (23.9%) were sharply up-regulated and 25 (22.1%) were down-regulated in MU17 compared to RA9 (Fig. 3B). In addition, the content of these significantly changed metabolites showed different patterns in RA9 and MU17 leaves (Fig. 3C). These results indicated that the two varieties (RA9 and MU17) have great metabolic differences. Furthermore, KEGG analysis has shown that the differential metabolites of RA9 and MU17 are enriched in flavonoid biosynthesis, phenylalanine metabolism, cysteine and methionine metabolism, phenylpropanoid biosynthesis and glycerophospholipid metabolic pathways (Fig. 3D). Notably, Phenylethylamine and sinapic acid are known to involve in the metabolic pathway of phenylalanine and phenylpropanoid biosynthesis (Zheng et al. 2019; Yang et al. 2020), which was remarkably increased in MU17 leaves (Table 1). Moreover, the compounds that participated in the flavonoid biosynthesis pathway also had obvious differences in the MU17 leaves compared to RA9. Correspondingly, a higher accumulation of anthocyanin (pelargonidin and cyanidin) and isoflavonoid (genistein) was also observed in MU17 leaves, while a lower content of flavonol kaempferol derivatives was founded in MU17 leaves (Table 1). It has been shown that the changes in 6-Caffeoylsucrose, chalconaringenin, flavone, flavonol, and anthocyanin imply the accumulation of anthocyanin (Nabavi et al. 2020). In our study, the metabolomic data showed that several anthocyanin-related metabolites (precursors, anthocyanin accumulation process and metabolite-related enzymes) were up-regulated in MU17 leaves (Fig. 3E). These results suggested that the potential changes in anthocyanin metabolism resulted in the purple phenotype in MU17 leaves.

A transcriptome analysis in the RA9 and MU17 was performed to determine the molecular mechanism of anthocyanin accumulation. The RNA-seq data showed that 1456 DEGs were down-regulated and 2007 DEGs were up-regulated in MU17 leaves, respectively (Fig. 5A). KEGG analysis showed that most of the DEGs from purple leaves of MU17 were enriched in pathways related to the ribosome, phenylpropanoid biosynthesis and flavonoid-anthocyanin biosynthesis (Fig. 5C, D, Table S5). Among them, more than 40 DEGs that participated in flavonoid-anthocyanin biosynthesis were enriched in MU17 leaves (Table 2). Simultaneously, the anthocyanin and flavonoid biosynthesis genes (Bai et al. 2019; Zhang et al. 2019), such as Chalcone synthase (CHS), Dihydroflavonol-4-reductase (DFR), Chalcone-flavonone isomerase (CHI) and Leucoanthocyanidin dioxygenase(LDOX), had significantly high expression in MU17 compared to RA9 (Table 2), which correlates with a higher accumulation of anthocyanins (pelargonidin and cyanidin) in MU17 radish (Fig. 2A, Table 1). Meanwhile, 160 compounds content and gene expression levels had significantly changed (P < 0.05) (Table S6). Among them, the expression of 11 significant DEGs and their metabolites showed a consistent or opposite trend (Fig. 6B). For instance, the 4-coumarate-CoA ligase (4CL, LOC108817856), flavone 3’-O-methyltransferase 1 (OMT1, LOC108821435), and indole glucosinolate O-methyltransferase 1 (IGMT, LOC108825157) had the same trend with the accumulation of sinapic acid in MU17 leaves, in which the 4CL and OMT1 promoted the synthesis of p-Coumaroyl-CoA and methylated anthocyanins in MU17 leaves, respectively. However, the sinapic acid had high accumulation in MU17 leaves, which is a commonly existing compound in radish leaves and taproots. Meanwhile, the reduced content of chalconaringenin was contrary to the higher-level expression of CHS3 (LOC108814337) and CHI3 (LOC108814129). The high expression of CHI would facilitate the biosynthesis of naringenin from chalconaringenin. These data suggested that the network for homeostasis between the anthocyanin biosynthesis pathway and the expression of genes involved in this process was complicated. Moreover, we detected the expressions of biosynthetic and structural genes during the anthocyanin metabolism using the qRT-PCR. Conspicuously, the expressions of the transcripts of anthocyanin accumulation-related structural genes (ANS, DFR, CYP73A, CHS, F3H, UGT79B1, PAL4, PXP7, OMT1, IRL-P3 like, and 4CL4) were increased in the MU17 compared to RA9. In addition, the expressions of anthocyanin biosynthetic related transcription factors (WRKY44-like, MYB308-like, RAV2, TT8, CPC, bHLH28, MYB114, MYB108-like, MYB305, MYB6 and bHLH93-like) were significantly up-regulated in MU17 cultivar (Fig. 7A, B). Meanwhile, the up-regulation of anthocyanin biosynthetic genes in MU17 was accompanied by a high level of anthocyanin concentration (Fig. 2A, Table 1). Correspondingly, similar results were displayed in RNA-seq results (Fig. 7C). As a consequence of the high expression of anthocyanin metabolism-related transcription factors and structural genes in the MU17 leaves, anthocyanin accumulation was predicted to occur.

In summary, we combined metabolomics, transcriptomic and qRT–PCR analyses to uncover the mechanisms underlying the differential accumulation of anthocyanin in RA9 and MU17 radish. The significantly increased anthocyanin content causes purple leaves in MU17 radish. Transcriptome analysis revealed more than 40 DEGs which participated in anthocyanin accumulation enriched in MU17 leaves. Moreover, the expression of anthocyanin accumulation-related genes were significantly up-regulated contributing to the high anthocyanin content in MU17 leaves, which was consistent with qRT–PCR analysis. These results provide us with a better understanding of anthocyanin mechanism in MU17 leaves and a series of candidate genes would be generated that could be applied in breeding of anthocyanins-rich cultivars.

Conclusion

In this study, the purple MU17 radish was identified to have a high content of anthocyanins in the leaves. Meanwhile, a comparative metabolite and transcription profiling of purple MU17 and green RA9 showed metabolic and transcriptional changes regarding phenylpropanoid and flavonoid metabolism, which is critical for the accumulation of anthocyanins in MU17 leaves. In addition, 23 homologous flavonoid genes were selected for their expression pattern using RNA-seq and qRT–PCR analysis, the structural genes (such as the ANS, DFR, CHS, F3H, PAL4, OMT1 and 4CL4) and key transcription factors (such as the TT8 and WRKYs) all showed significantly high expression in purple MU17. Therefore, this study provides a new light into the understanding of anthocyanin mechanism in radish and potential candidate genes for breeding improvements for radish.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, QP and LZ; Funding acquisition, QP and LZ; Investigation, QP, ZH and LZ; Project administration, CX and PY; Software, SS; Writing—original draft, QP; Writing—review and editing, LZ.

Data availability

The RNA-seq raw data for the current study were deposited in the National Genomics Data Center (https://ngdc.cncb.ac.cn/gsa/browse/CRA008485) under accession ID CRA008485.

Declarations

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Consent to participate

All authors have agreed to participate in the manuscript.

Consent for publication

All authors have agreed to publish the manuscript.