Transcriptome Analysis and Metabolic Profiling of Green and Red Mizuna (Brassica rapa L. var. japonica)

Mizuna (Brassica rapa L. var. japonica), a member of the family Brassicaceae, is rich in various health-beneficial phytochemicals, such as glucosinolates, phenolics, and anthocyanins. However, few studies have been conducted on genes associated with metabolic traits in mizuna. Thus, this study provides a better insight into the metabolic differences between green and red mizuna via the integration of transcriptome and metabolome analyses. A mizuna RNAseq analysis dataset showed 257 differentially expressed unigenes (DEGs) with a false discovery rate (FDR) of <0.05. These DEGs included the biosynthesis genes of secondary metabolites, such as anthocyanins, glucosinolates, and phenolics. Particularly, the expression of aliphatic glucosinolate biosynthetic genes was higher in the green cultivar. In contrast, the expression of most genes related to indolic glucosinolates, phenylpropanoids, and flavonoids was higher in the red cultivar. Furthermore, the metabolic analysis showed that 14 glucosinolates, 12 anthocyanins, five phenolics, and two organic acids were detected in both cultivars. The anthocyanin levels were higher in red than in green mizuna, while the glucosinolate levels were higher in green than in red mizuna. Consistent with the results of phytochemical analyses, the transcriptome data revealed that the expression levels of the phenylpropanoid and flavonoid biosynthesis genes were significantly higher in red mizuna, while those of the glucosinolate biosynthetic genes were significantly upregulated in green mizuna. A total of 43 metabolites, such as amino acids, carbohydrates, tricarboxylic acid (TCA) cycle intermediates, organic acids, and amines, was identified and quantified in both cultivars using gas chromatography coupled with time-of-flight mass spectrometry (GC-TOFMS). Among the identified metabolites, sucrose was positively correlated with anthocyanins, as previously reported.


Introduction
Plant species of the family Brassicaceae, previously called Cruciferae, have long been considered health-boosting vegetables and are commonly cultivated worldwide because they provide high amounts of dietary minerals, amino acids, carbohydrates, fatty acids, dietary fibers, vitamins (tocopherol, ascorbate, and folate), and bioactive molecules, including glucosinolates, phenylpropanoids, anthocyanins, and carotenoids. Furthermore, they have been commercially important as a main element of the daily human diet as well as a source of vegetable oil for the food and biodiesel industry [1][2][3][4][5].
provide information on metabolic differences between green and red mizuna through transcriptome and metabolome analyses.

Plant Material
Asia Seed Co., Ltd. (Seoul, Korea) provided plant samples of green and red cultivars of mizuna (B. rapa L. var. japonica) cultivated at an experimental field of the Asia Seed R&D Center (Icheon, Gyeonggi-do, Korea). Mizuna plants ( Figure 1) were harvested after four months, treated with liquid nitrogen (−196 °C), and then stored at −80 °C. Both samples were ground into fine powders for RNA extraction and HPLC analysis.

RNA Extraction and Illumina Sequencing
RNA extraction and Illumina sequencing were performed according to our previously reported method (Jeon et al., [26]). Total RNA was extracted using the Plant Total RNA Mini Kit. The extracted RNA was quantitated using a Nanodrop, and its integrity was measured through denaturing 1% agarose gel electrophoresis. Subsequently, mRNA was separated and purified using Sera-Mag Magnetic Oligo (dT) beads, and cDNA was synthesized for Illumina sequencing using the Illumina/Solexa HiSeq2000 platform. RNA-sequencing analysis was performed to comprehensively compare the characteristics of red and green mizuna transcriptomes. In total, 81,330,120 and 46,286,878 reads were obtained for green and red mizuna, making up 6.18 Gb (giga-bases) and 3.52 Gb, respectively, with an average length of 76 nucleotides per read. After filtering, 61,035,862 and 36,350,466 clean reads were obtained for the green and red mizuna samples, respectively, with Q30 values of 89.69% and 92.31%, respectively, (Table S1). The Bowtie2 software and the DESeq library were used to estimate the expression abundance.

Assembly and Functional Annotation
After quality trimming of the raw RNA-sequencing reads, 81,330,120 and 46,286,878 highquality clean reads of green and red cultivars of mizuna, respectively, were prepared for transcriptome assembly. The Trinity de novo assembly program (https://github.com/trinityrnaseq/trinityrnaseq/wiki) was used to combine the overlapping reads into contigs without gaps. Furthermore, the filtered reads were aligned back to the de novo assembled

Plant Material
Asia Seed Co., Ltd. (Seoul, Korea) provided plant samples of green and red cultivars of mizuna (B. rapa L. var. japonica) cultivated at an experimental field of the Asia Seed R&D Center (Icheon, Gyeonggi-do, Korea). Mizuna plants ( Figure 1) were harvested after four months, treated with liquid nitrogen (−196 • C), and then stored at −80 • C. Both samples were ground into fine powders for RNA extraction and HPLC analysis.

RNA Extraction and Illumina Sequencing
RNA extraction and Illumina sequencing were performed according to our previously reported method (Jeon et al. [26]). Total RNA was extracted using the Plant Total RNA Mini Kit. The extracted RNA was quantitated using a Nanodrop, and its integrity was measured through denaturing 1% agarose gel electrophoresis. Subsequently, mRNA was separated and purified using Sera-Mag Magnetic Oligo (dT) beads, and cDNA was synthesized for Illumina sequencing using the Illumina/Solexa HiSeq2000 platform. RNA-sequencing analysis was performed to comprehensively compare the characteristics of red and green mizuna transcriptomes. In total, 81,330,120 and 46,286,878 reads were obtained for green and red mizuna, making up 6.18 Gb (giga-bases) and 3.52 Gb, respectively, with an average length of 76 nucleotides per read. After filtering, 61,035,862 and 36,350,466 clean reads were obtained for the green and red mizuna samples, respectively, with Q30 values of 89.69% and 92.31%, respectively, (Table S1). The Bowtie2 software and the DESeq library were used to estimate the expression abundance.

Assembly and Functional Annotation
After quality trimming of the raw RNA-sequencing reads, 81,330,120 and 46,286,878 high-quality clean reads of green and red cultivars of mizuna, respectively, were prepared for transcriptome assembly. The Trinity de novo assembly program (https://github.com/trinityrnaseq/trinityrnaseq/wiki) was used to combine the overlapping reads into contigs without gaps. Furthermore, the filtered reads were aligned back to the de novo assembled transcriptome sequences to validate the assembly and then mapped into the B. rapa reference genome from the Brassica database (BRAD) (http://brassicadb.org/brad) using TopHat2 (http://tophat.cbcb.umd.edu) Martin [27] and bowtie2 (http://bowtie-bio.sourceforge.net) Foods 2020, 9, 1079 4 of 13 (Kim et al. [28]). After normalizing expression levels by using the DESeq package in R, the transcript expression levels were calculated using the Fragments Per Kilobase of transcript per Million mapped reads (FPKM), and a false discovery rate (FDR) was applied to determine the significance cutoff of 0.05 via DESeq. Gene ontology (GO) was performed for the differentially expressed unigenes (DEGs) with an FDR < 0.05 in both cultivars. The Kyoto Encyclopedia of Genes and Genomes (KEGG) and Plant Transcription Factor (PlantTFDB) databases were used for the further analysis of genome networks. The statistical significance of DEGs was determined using the threshold setting (log2-fold changes of ≤−1 or ≥1 and an adjusted p-value of ≤0.05).

HPLC Analysis of Desulfo-Glucosinolates
Desulfoglucosinolates were extracted and analyzed as per previously reported procedures by Park et al. [29]. Briefly, 100 mg dried green and red mizuna cultivars were extracted with 1.5 mL of 70 • C boiling aquatic methanol 70% (v/v) for 5 min and then centrifuged at 11,000× g at 4 • C for 15 min. The supernatants were transferred into fresh tubes. The remaining pellets were re-extracted twice more in the same manner. The final collected supernatants were loaded onto a mini-column filled with DEAE-Sephadex A-25, followed by desulfation using 75 µL of an aryl sulfatase solution for 12 h. The resulting desulfoglucosinolate extract was further eluted with 0.5 mL of HPLC-grade water. An Agilent Technologies 1200 Series HPLC System (Palo Alto, CA, USA) equipped with an Inertsil ® ODS-3 column (150 × 3.0 mm i.d., particle size 3 µm; GL Sciences, Tokyo, Japan) and an Inertsil ® ODS-2 guard column (10 × 2.0 mm i.d., particle size 5 µm) was used to separate glucosinolate components. The HPLC conditions were set as follows: detection wavelength, 227 nm; oven temperature, 40 • C; flow rate, 0.2 mL min −1 ; injection volume, 20 µL. The gradient program used was as follows: solvent A, HPLC-grade water; solvent B, HPLC-grade acetonitrile; 0-18 min, 7-24% B; 18-32 min, 24% B; 32.01 min, rapid drop to 7% B; and 32.01-40 min, 7% B (total 40 min). Quantification of desulfoglucosinolates was performed as previously described by Park et al. [29]. Each compound was quantified using sinigrin (internal standard) and the response factor of individual desulfoglucosinolates relative to the internal standard.

HPLC Analysis of the Phenolic Compounds
Phenolics were extracted as per the previously reported method by Park et al. [30]. Briefly, 1 mL of aqueous methanol (80% v/v) was added to a fresh tube containing fine powders (100 mg) of dried green and red mizuna cultivars and vortexed thoroughly for 20 s. The extract was sonicated at temperatures below 25 • C for 60 min and centrifuged at 11, 000× g for 15 min. Then, the supernatants were collected into fresh tubes. The remaining sludge was used to re-extract phenolics two more times in the same manner. The collected supernatants were dried using a vacuum concentrator and then dissolved in 0.5 mL of methanol. An NS-4000 HPLC system (Futecs, Daejeon, Korea) with a UV−Vis detector and a C 18 column (250 × 4.6 mm, 5 µm; RStech, Daejeon, Korea) was used to isolate gallic acid, chlorogenic acid, caffeic acid, catechin, epi-catechin, vanillin, and benzoic acid. The gradient program was used according to our previously reported method [31]. The mobile phase consisted of solvent A, ultrapure water containing 0.2% acetic acid, solvent B, and methanol at a flow rate of 1 mL/min. The HPLC running time was 98 min, the oven temperature was 30 • C, and the detection wavelength was 280 nm. Peak identification was performed by comparison with the retention times of the standard chemicals and spike test, and quantification was carried out according to each calibration curve of eight phenolic and organic compounds (Table S2). The standards used for analysis of the phenolic compounds are gallic acid (≥99%), vanillin (99%), catechin hydrate (≥98%), (−)-catechin (≥98%), benzoic acid (≥99.5%), caffeic acid (≥98%), chlorogenic acid (≥95%), 4-hydroxybenzoic acid (99%) and were purchased from Sigma-Aldrich Co., Ltd. (St. Louis, MO, USA).

HPLC and LC-ESI-MS/MS Analysis of Anthocyanins
Anthocyanins were extracted as per our previously reported method (Park et al. [31]). Anthocyanins were extracted from 100 mg fine powders (100 mg) of dried green and red mizuna cultivars, with 2 mL water/formic acid (95:5 v/v), followed by gentle sonication for 20 min. After centrifugation at 11,000× g for 15 min, the extract was syringe-filtered into an LC vial. Different anthocyanins were separated using an Agilent 1200 series HPLC linked to a 4000 Q-Trap LC-ESI-MS/MS system with a Synergi 4 µm POLAR-RP 80A column (250 × 4.6 mm i.d., particle size 4 µm) combined with a Security Guard Cartridges Kit AQ C18 column. The operating conditions and gradient program were used as in our previously reported method (Park et al. [31]). Different anthocyanin quantifications were performed using a standard calibration curve depicted from the commercial anthocyanin (Cyanidin-3-O-glucoside (C3G)). C3G was used as an external standard and each anthocyanin was expressed as milligram per C3G gram dry weight (mg/C3G g dry weight).

GC-TOFMS Analysis of Polar Metabolites
Hydrophilic were extracted as per our previously reported method (Park et al. [29]). Fine powders (50 mg) of green and red mizuna were extracted with 1 mL of a chloroform-water-methanol mixture (1:1:2.5 v/v/v) to which 60 µL of ribitol (0.2 g/L) was added as an internal standard. The extract was shaken at 37 • C and 1200× g for 30 min and then centrifuged at 12,000× g for 15 min. After the supernatants were evaporated in a SpeedVac vacuum concentrator for 3 h, the extract was derivatized by adding 80 µL of methoxy-amine hydrochloride/pyridine (20 g L −1 ) and shaken at 37 • C and 1200× g for 2 h. Thereafter, 80 µL of N-methyl-N-(trimethylsilyl)trifluoroacetamide was added and the resulting mixture was heated at 37 • C for 30 min. After centrifugation, the final extract was transferred to a 2 mL glass vial with a micro insert. The analysis equipment and operation conditions of GC-TOFMS were as described in our previous study (Park et al. [29]). The quantification of the polar metabolite was performed using selected ions, and the Chroma-TOF software was used to locate the peaks. The peak identification of the GC-TOF/MS data was performed by comparing their retention times and mass spectrum with standard compounds, in-house library, and MS library (Nist database). ChromaTOF software was used to identify the hydrophilic compounds in mizuna cultivars. The results were filtered with retention time, signal-to-noise ratio (>5:1), and mass spectral matching (based on a match >700) by using reference compounds and the use of an in-house library. As a result, a total of 43 metabolites were identified (i.e., Metabolomics Standards Initiative (MSI) level 1) Viant et al. [32]. The corresponding retention times and their fragment patterns were agreed with our previous data (Park et al. [29]).

Statistical Analysis of Metabolites
A Student's t-test was performed using the Statistical Analysis System (SAS, system 9.4, 2013; SAS Institute, Inc., Cary, NC, USA). A volcano plot and a hierarchical cluster analysis (HCA) were carried out, and Pearson correlations for 76 metabolites identified in these analyses were determined using MetaboAnalyst 4.0 (http://www.metaboanalyst.ca/) with auto-scaling.

Functional Annotation and Classification of the Green and Red Mizuna Transcriptomes
The cDNA libraries of the green and red mizuna were mapped to the B. rapa reference genome with coverages of 75.5% and 75.4%, respectively, (Table S3). Among the 257 DEGs with an FDR < 0.05, 94 were expressed at higher levels and 163 were expressed at relatively lower levels in the red cultivar than in the green cultivar ( Figure 2). A GO analysis of the DEGs was carried out using DAVID (http://david.abcc.ncifcrf.gov/tools.jsp). The DEGs were assigned to the GO terms of biological process (BP) and molecular function (MF) categories as shown in Table 1. In the BP category, the DEGs were assigned to cell death (10 DEGs), apoptosis (eight DEGs), toxin catabolic process (four DEGs), toxin metabolic process (four DEGs), sucrose metabolic process (three DEGs), defense response (19 DEGs), innate immune response (eight DEGs), response to temperature stimulus (nine DEGs), secondary metabolic process (10 DEGs), response to bacterium (seven DEGs), multidrug transport (four DEGs), oxidation reduction (19 DEGs), drug transport (four DEGs), response to drug (four DEGs), and response to reactive oxygen species (five DEGs). In the MF category, the DEGs were assigned to heme binding (11 DEGs), tetrapyrrole binding (11 DEGs), oxygen binding (eight DEGs), iron ion binding (15 DEGs), glutathione transferase activity (four DEGs), electron carrier activity (14 DEGs), and drug transporter activity (four DEGs). This result was consistent with the previous results of Park et al. [33].

Identification of Secondary Metabolite Biosynthetic Genes from the Green and Red Mizuna Transcriptomes
Transcriptome analysis of mizuna cultivars revealed the identification of genes involved in the biosynthesis of secondary metabolites, including glucosinolates, phenolics, and anthocyanins. Based

Quantification of Glucosinolates in Green and Red Mizuna
Fourteen glucosinolates were detected in green mizuna, whereas only nine glucosinolates were found in red mizuna ( Table 2). The level of total glucosinolates was significantly higher in green mizuna than in red mizuna. Specifically, glucoraphanin, glucoalyssin, and gluconapoleiferin were detected only in green mizuna, and progoitrin and 4-hydroxyglucobrassicin levels were significantly higher in green than in red mizuna. In contrast, gluconasturtiin was detected only in red mizuna and glucobrassicin, 4-methoxyglucobrassicin, and neoglucobrassicin were significantly higher in red than in green mizuna. In particular, the highest accumulation patterns of most aliphatic glucosinolates (glucoiberin, progoitrin, glucoraphanin, glucoalyssin, gluconapoleiferin, and gluconapin) were observed in the green cultivar, whereas the red cultivar showed higher production patterns of indolic glucosinolates, represented mainly by glucobrassicin, 4-methoxyglucobrassicin, and neoglucobrassicin. These results were consistent with the results of the DEG analysis, revealing that the expression of the regulatory gene MYB28 and structural genes (BCAT4, MAM1, CYP79F1, CYP83A1, SUR1, AOP3, ST5c, and ST5b) responsible for aliphatic glucosinolate biosynthesis was higher in the green than in red mizuna, but revealed that the expression of a regulatory gene MYB34 involved in indolic glucosinolate biosynthesis was significantly higher in red than in green mizuna. Furthermore, these findings were supported by previous studies reporting that BoaMYB28 overexpressing lines of Chinese kale showed increased transcript levels of the aliphatic glucosinolate biosynthesis genes and increased levels of aliphatic glucosinolates. In contrast, the transcript levels and aliphatic glucosinolate levels decreased in the BoaMYB28 RNAi transgenic lines of Chinese kale [34]. Similarly, Frerigmann and Gigolashvilli [35] reported that MYB34 is the main transcription factor regulating indolic glucosinolate biosynthesis.

Quantification of Phenolic and Organic Compounds in Green and Red Mizuna
Five phenolic and two organic compounds related to the biosynthesis of phenolic compounds were detected in both green and red mizuna cultivars using HPLC (Table 3), whereas a total of 12 anthocyanins were identified only in red mizuna (Table 4). For the individual compounds, the levels of caffeic acid, (−)-epicatechin, and vanillin were higher in green mizuna, whereas those of gallic acid and catechin were higher in red mizuna. Furthermore, cyanidins were the major anthocyanins ubiquitous in red mizuna, and the red color might be derived from the identified cyanidin derivatives in the red cultivar [36]. These HPLC results were in accordance with the results of the analysis of DEGs which revealed that the expression of structural genes (C4H, DFR, ANS, UF3GT, 5GT, and TT19) responsible for the phenylpropanoid and flavonoid biosynthesis was significantly higher in red mizuna. Similarly, Jeon et al. [26] reported that higher levels of cyanidin derivatives were present in red kale than in green kale, which is consistent with the higher expression of anthocyanin biosynthetic genes in red kale via metabolome and transcriptome analysis. Table 3. Phenolic and organic compounds in green and red mizuna (µm/g dry weight).

Class Name Green Mizuna Red Mizuna
Phenolic acid  Table 4. Anthocyanin content in green and red kale. (µmol/g dry weight).

Quantification of Phenolic and Organic Compounds in Green and Red Mizuna
A total of 43 hydrophilic (two phenolic acids, three photorespiration intermediates, four TCA cycle intermediates, five organic acids, 18 amino acids, and 11 sugars) were identified and quantified in both cultivars using GC-TOFMS (Table S6). A comparison of amino acid levels between the green and red mizuna cultivars indicated that the levels of valine, serine, leucine, isoleucine, proline, glycine, threonine, β-alanine, aspartic acid, methionine, pyroglutamic acid, asparagine, glutamine, phenylalanine, tryptophan, and glutamic acid were significantly higher in green mizuna, whereas 4-aminobutyric acid levels were only higher in red mizuna. In particular, the glutamine content was consistent with the transcript levels of glutamine synthetase 1;4 (Table S7). Among the identified sugars, xylose, fructose, glucose, mannose, and glycerol levels were significantly higher in green mizuna. However, the levels of sucrose, maltose, trehalose, raffinose, and inositol were higher in red mizuna ( Figure S1). Sucrose synthesis involves a two-step process catalyzed by two different enzymes, sucrose-6-phosphate synthase (SPS) and sucrose-6-phosphate phosphatase (SPP), in plants. In the first step of the sucrose biosynthetic pathway catalyzed by the SPS, sucrose-6-phosphate is synthesized from uracil-diphosphate glucose (UDP-glucose) and fructose-6-phosphate through the SPS activity. Next, SPP rapidly dephosphorylates sucrose-6-phosphate to sucrose and inorganic phosphate [37]. The produced sucrose can be further degraded to fructose and UDP-glucose by the activity of sucrose synthase [38]. In this study, the sucrose level was consistent with the analysis of DEGs indicating the higher expression of sucrose-phosphate synthase family protein (SPS4F) and sucrose-phosphatase 1 (SPP1) and lower expression of sucrose synthase 3 (SUS3) in red mizuna.

Conclusions
To our knowledge, this is the first study to provide a comprehensive transcriptome and metabolome analysis of primary and secondary metabolites in green and red mizuna. Based on the high-throughput transcriptome data, the primary metabolite biosynthesis genes, including three genes related to sucrose metabolism and one gene related to glutamine metabolism, as well as secondary metabolite biosynthesis genes, include 12 genes involved in glucosinolate biosynthesis and 10 genes responsible for phenylpropanoid and flavonoid biosynthesis in mizuna. Furthermore, 14 glucosinolates, 12 anthocyanins, five phenolics, two organic acids, and 43 hydrophilic were detected in red and green mizuna cultivars using the HPLC, ESI-LC/MS/MS, and GC-TOFMS analyses. Through a comparative transcriptome and metabolome analysis, this study showed that the green mizuna contained a higher content of aliphatic glucosinolates in accordance with the expression of genes involved in aliphatic glucosinolate biosynthesis. In contrast, the red cultivar had a higher content of indolic glucosinolates and anthocyanin, consistent with the expression of genes responsible for indolic glucosinolate and flavonoid biosynthesis. Furthermore, a strong positive correlation between sucrose and anthocyanins was observed to support the positive effect of sucrose on anthocyanin biosynthesis. Taken together, these findings may help to develop breeding strategies and also to improve the biosynthesis of glucosinolates and anthocyanins in mizuna.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2304-8158/9/8/1079/s1. Figure S1: Volcano plot of differentially accumulated metabolites between green and red mizuna. The y-axis is the negative log10 of p values and the x-axis is log2 of fold change. Red indicates that the metabolites identified in green mizuna were significantly higher than those of red mizuna, blue indicates that the metabolites identified in green mizuna were significantly lower than those of red mizuna, and grey indicates metabolites with no significant difference. Table S1: Summary of RNA sequence data, Table S2: Validation of HPLC analysis of phenolic compounds, Table S3: Summary of genome mapping, Table S4: Putative glucosinolate biosynthetic genes in the mizuna transcriptome., Table S5: Putative phenylpropanoid and anthocyanin biosynthetic genes in the mizuna transcriptome., Table S6: Identified metabolites in GC-TOFMS chromatograms from mizuna extract., Table S7: Putative sucrose and glutamine biosynthetic genes in the mizuna transcriptome., Table S8. The fold changes of the red relative to the green mizuna corresponding to Figure S1.