Chinese kale seeds were germinated in plastic pots, and the seedlings were cultured in the field at 22–25°C at South China Agricultural University (Guangzhou, China). Bolting stems were harvested when plants were fully grown, with inflorescences as high as the apical leaves used for cDNA sequence cloning. For spatial expression pattern analysis, cotyledon (7 days), leaf (15 days), stem (30 days), flower (anthesis), and silique (15 days post-anthesis) were excised from 10 plants. For temporal expression pattern analysis, primary leaf (15 days), young leaf (30 days), mature leaf (60 days), and inflorescence leaf (anthesis) were excised from 10 plants.

Arabidopsis thaliana seeds (ecotype Col-0) were sterilized and plated on agar-solidified half-strength Murashige and Skoog (MS) medium containing 2.15 g/l MS salts and 0.5% sucrose, pH 5.8. Wild-type seeds were cold-treated at 4°C for 3 days in the dark and grown in a culture chamber at 21°C under 16/8 h light/dark cycle. Seedlings were transferred to soil and grown at 25°C under 16/8 h light/dark regime and at 40% humidity.

Plant materials were harvested in the early morning, weighed when fresh and washed with distilled water. All samples were immediately transported to the laboratory within 10 min, frozen in liquid nitrogen and lyophilised in an ultralow -80°C freezer for RNA isolation and/or for aliphatic glucosinolate content analysis.

Total RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA, United States) followed by treatment with DNase (Promega, MA, United States) to remove possible DNA contamination. RNA quality and content were determined by gel electrophoresis and biophotometer (Eppendorf, Germany). First-strand cDNA was synthesized from 1 μg total RNA with an oligo-dT primer according to the instructions of the Reverse Transcriptase M-MLV Kit (Takara, Japan).

The full-length coding sequence of BoaMYB28 was isolated from Chinese kale. The primers were designed based on a sequence identified in the BRAD Brassica database (Cheng et al., 2011). PCR amplification was performed using the primers listed in Supplementary Table S1. The PCR program was as follows: 95°C for 5 min; 35 cycles of 95°C for 1 min, 58°C for 1 min, and 72°C for 1 min; with a final extension at 72°C for 10 min. Five independent PCRs were used for isolating the gene. PCR products were ligated into the pMD18-T vector (Takara, Japan) and transformed into Escherichia coli DH5α. The BoaMYB28 sequence was submitted to GenBank and was given the accession number KP723785.

The expression patterns of the target genes were assayed by quantitative real-time PCR (qRT-PCR) using the primers listed in Supplementary Table S1. The reaction was performed using the MyiQ Real-Time PCR Detection System (Bio-Rad) platform using the SYBR Green Master Mix (Roche, Swiss) in a total 20 μl volume. The relative gene expression was calculated by the 2^-ΔΔCt method.

The amplification protocol was as follows: 95°C for 2 min; 35 cycles at 95°C for 15 s, 58°C for 30 s, and 72°C for 30 s; with a 5 min final extension at 72°C. β-actin was selected as the endogenous control gene for relative quantification. PCR reactions were performed from three independent biological replicates with three technical replicates for each. Data was analyzed using Bio-Rad CFX Manager software.

For generation of the BoaMYB28 over-expression and RNAi construct, the coding sequence of MYB28 was cloned into the plasmid vector, pFGC5941 (GenBank number AY310901) (Kerschen et al., 2004). The pFGC5941 vector contained the herbicide resistance gene (bar), which was used as selectable marker gene for generating transgenic lines (Supplementary Figures S10, S11). The different constructs were transformed into Agrobacterium strain GV3101 by the freeze-thaw method and further into A. thaliana by the floral dip method (Clough and Bent, 1998). Transgenic plants were selected on half-strength MS medium (0.5% sucrose, 0.8% agar, pH 5.8) (Gigolashvili et al., 2008).

Chinese kale seeds were surface sterilized with 70% ethanol for 5 min followed by 10% sodium hypochlorite for 10 min and then rinsed three times in sterile distilled water. The sterilized seeds were germinated for 1 week in half-strength MS medium with 30 g/l sucrose and 8 g/l agar at pH 5.8 (Murashige and Skoog, 1962). The hypocotyl explants were cultured for 3 days in MS medium with 1 mg/l 6-benzylaminopurine (BAP) and 1 mg/l naphthaleneacetic acid (NAA). The precultured hypocotyls were soaked in the overnight-cultured Agrobacterium tumefaciens strain EHA105 suspension for 5 min. The explants were cultured for 3 days in MS medium with the addition of 1 mg/l BAP and 1 mg/l NAA.

After washing any remaining bacteria away with sterile water, the explants were cultured for 1 week in MS medium with 500 mg/l cefotaxime sodium (Cef), 1 mg/l BAP, and 1 mg/l NAA. The explants were transferred to selective shoot regeneration medium containing MS basal medium with 5 mg/l phosphinothericin (ppt), 500 mg/l Cef, 1 mg/l BAP, and 1 mg/l NAA. After culturing for 2 weeks, the shoot buds were sub-cultured on MS medium with 10 mg/l ppt, 500 mg/l Cef, 1 mg/l BAP, and 1 mg/l NAA. The elongated green shoots were transferred to MS medium with 250 mg/l Cef and 2 mg/l NAA for rooting. After culturing for 4–5 weeks, plantlets with well-developed roots were hardened off and transplanted into the greenhouse for further analysis and recovery of seeds (Qian et al., 2015).

Southern blot was used to confirm that MYB28 was integrated into A. thaliana and Chinese kale genomes (Zang et al., 2009). Aliquots of DNA (25 μg) were digested overnight with EcoRI, fractionated by 0.8% (w/v) agarose gel electrophoresis and transferred to a Hybond-N+ nylon membrane (Amersham Pharmacia Biotech, United Kingdom). The probe used for Southern analysis was bar, a selectable marker gene. Probe labeling, pre-hybridisation, hybridisation and detection were performed according to the protocol of DIG High Prime DNA Labeling and Detection Starter Kit II (Roche Diagnostics GmbH, Germany).

Extraction and determination of aliphatic glucosinolates was performed in 5-week-old rosette leaves of transgenic and wild-type A. thaliana. Glucosinolates were extracted and analyzed as previously described (Gigolashvili et al., 2008). Freeze-dried samples were extracted twice with 70% methanol with the addition of glucotropaeolin (Applichem) as an internal standard. The methanol extracts were applied to a DEAE Sephadex A-25 column (acetic acid activated). The columns were sealed and desulphated overnight using purified Helix pomatia sulfatase (type H1, Sigma, Deisenhofen, Germany). HPLC analysis of desulphoglucosinolates was carried out as previously described (Gigolashvili et al., 2008). Detection of desulphoglucosinolates was performed at 229 nm by applying the response factors as described previously (Gigolashvili et al., 2007).

Glucosinolates of Chinese kale were extracted and analyzed as previously described (Sun et al., 2012a). The freeze-dried samples (500 mg) were mixed with 2 ml of 75% methanol at 80°C for 15 min and centrifuged at 10,000 × g for 5 min. The supernatant was further washed with 2 ml of 75% methanol and centrifuged at 10,000 × g for 5 min. The supernatant sample was applied onto DEAE-Sephadex A-25 column (acetic acid activated) and rinsed three times with 2 ml 0.02 M sodium acetate. The column was washed three times with 20 mM acetic acid and twice with water. The supernatant was mixed with 200 μl arylsulfatase solution. After incubation for 16 h at 35°C, the desulphoglucosinolates were eluted with 4 ml of Milli-Q water and filtered through a 0.45-μm membrane filter. HPLC analysis of desulphoglucosinolates was carried out as previously described (Qian et al., 2015). Glucosinolates were extracted using ortho-nitrophenyl-β-D-galactopyranoside (Sigma) as an internal standard.

The amino acid sequence was predicted using DNAman. Multiple sequence alignments of amino acid sequences were performed using ClustalW. Phylogenetic trees were generated using MEGA based on the NJ (neighbor-joining) method. The confidence value for the phylogeny analysis was calculated using the bootstrap method. Molecular weight and isoelectric point of the protein were calculated with ProtParam (Brown et al., 2003). The conserved domains were predicted by CDD (Brown et al., 2003). The secondary structures and tertiary structures were predicted by SOPMA and SWISS MODEL (Brown et al., 2003). The transmembrane domains were predicted by TMHMM 2.0. The topology was determined using TMpred software. Signal peptide prediction was performed using SignalP 4.1. The hydrophilic and hydrophobic regions of the protein were predicted using ProtScale. Putative N-glycosylation sites were predicted using NetNGlyc 1.0 (Yu et al., 2004). Putative phosphorylation sites were predicted using NetPhos 2.0. YinOYang 1.2 Server was used to predict the putative O-glycosylation sites.

The data were analyzed with the SPSS software package version 11.5 (SPSS Inc. Chicago, IL, United States). Data were analyzed for statistical significance by one-way ANOVA model. The means were determined by the Least Significant Differences (LSD) test at p = 0.05.

We got the reference sequence from BRAD Brassica database. BRAD Brassica database is a new database which focuses on the genetics and genomics of the Brassica plants¹. We designed the cloning primers by Primer 6.0 and determined the full-length of the target gene by PCR amplification. Two MYB28 sequences (MYB28-1 and MYB28-2) were cloned from different varieties of Chinese kale. The cDNAs of the two MYB28 sequences were 98.57% identical with a total 18 different nucleotides over 1257 bp (Supplementary Figure S1). The deduced amino acid sequences were 100% identical (Supplementary Figure S2). We submitted only one sequence (MYB28-1) to GenBank (BoaMYB28). Analysis of the nucleotide sequence indicated that the BoaMYB28 cDNA was 1257 bp with an ORF of 1020 bp (Supplementary Figure S3). The full-length cDNA contained a 138-bp 5′-untranslated region and 99-bp 3′-untranslated region. The deduced MYB28 protein sequence was 339 amino acids in length with a putative molecular mass of 38 kDa and a pI of 6.87. Homology searches against the Conserved Domain Database (CDD) revealed that BoaMYB28 contained Myb-like DNA-binding domains (positions 14–61 and 67–112) and belonged to the SANT (SWI3, ADA2, N-CoR, and TFIIIB) superfamily. Secondary structure prediction by SDPMA showed that BoaMYB28 contained 37.46% alpha helices, 45.72% random coils, 7.37% extended strand, and 9.44% beta turn (Supplementary Figure S4). A similar result was obtained after constructing the BoaMYB28 three-dimensional structure using SWISS-MODEL (Supplementary Figure S5).

Alignment of the deduced amino acid sequence of BoaMYB28 revealed 73–98% identity with MYB28 proteins from 17 selected plant species. BoaMYB28 showed the higher level of sequence identity with MYB28 proteins from other members of the genus Brassica and lower levels of sequence identity with MYB28 proteins form other genera (Table 1).

In order to evaluate the molecular evolutionary relationships, the NJ method was used to construct a phylogenetic tree of MYB28 protein sequences (Figure 1). The distances between different genera were relatively large while those within the same genus were relatively small. Multiple sequence alignment showed that BoaMYB28 was phylogenetically closer to MYB28 proteins from other Brassica vegetables than other genera. The phylogenetic relationships were in accordance with the classification and evolutionary status via morphological and biochemical characteristics.

Bioinformatic tools were used to conduct structural analysis, such as identification of putative structural domains in BoaMYB28 proteins. The TMHMM results suggested that BoaMYB28 had no obvious transmembrane domain, which implied that BoaMYB28 was neither a membranous acceptor nor was located in the membrane. TMpred analysis predicted a lack of a transmembrane helix in BoaMYB28, and SignalP analysis also revealed a lack of a signal peptide. We examined the hydrophobic/hydrophilic nature of BoaMYB28 using ProtScale software, with a predicted maximum value of 1.5 and a minimum value of -3.578. Hydrophilic amino acid residues predominated in the peptide chain, indicating that BoaMYB28 might be a hydrophilic protein. NetNGlyc analysis identified three single N-glycosylation sites at positions 134, 177, and 276. Prediction of phosphorylation sites revealed 28 potential serine, 9 potential threonine, and 2 potential tyrosine targets were present in BoaMYB28. YinOYang analysis revealed that 12 potential O-glycosylation sites exist in BoaMYB28.

qRT-qPCR was used to analyze the BoaMYB28 mRNA expression levels in five tissues (cotyledon, leaf, stem, flower, and silique) and four developmental stages of leaf (primary leaf, young leaf, mature leaf, and inflorescence leaf). BoaMYB28 was expressed in all of the collected tissues under normal growth conditions (Figure 2). However, BoaMYB28 expression levels differed among the tissues. BoaMYB28 was abundantly expressed in tissues that synthesize glucosinolates, such as leaves and stems, whereas cotyledons, flowers, and siliques showed lower transcript accumulation. BoaMYB28 had the highest transcript abundance in stems and the lowest transcript accumulation in siliques. The expression level of BoaMYB28 in stems was 9.52 times higher than that in siliques.

The temporal expression of BoaMYB28 was further determined at different developmental stages of Chinese kale leaves, tissues where glucosinolate synthesis predominantly occurs. During the developing stages of the leaves, BoaMYB28 transcript levels were higher in the mature and inflorescence leaves than those in the primary and young leaves (Figure 3). BoaMYB28 had the highest transcript abundance in the mature leaves and the lowest transcript abundance in the primary leaves. The expression level of BoaMYB28 in the mature leaves was 4.2 times higher than that in the primary leaves.

In order to illustrate that aliphatic glucosinolates contents can be controlled by BoaMYB28, several transgenic over-expression and RNAi plants were created using Agrobacterium tumefaciens-mediated transformation. About 30 primary transformants were used for BoaMYB28 over-expression constructs of A. thaliana. About 60 primary transformants were used for both BoaMYB28 over-expression and RNAi constructs of Chinese kale. Three representative over-expressing lines (OE.Ara1, OE.Ara2, and OE.Ara3) of A. thaliana were chosen for further analysis. Three representative over-expressing lines (OE.Boa1, OE.Boa2, and OE.Boa3) and RNAi lines (RNAi.Boa1, RNAi.Boa2, and RNAi.Boa3) of Chinese kale were chosen for further analysis. T2 transformants were used for phenotypic and molecular analysis. Southern blot and qRT-PCR analysis confirmed that BoaMYB28 had been integrated into A. thaliana and Chinese kale genomes (Supplementary Figures S6, S7 and Figures 4, 5). BoaMYB28 transgenic lines of A. thaliana (OE.Ara1) and Chinese kale (OE.Boa1 and RNAi.Boa1) showed no visible differences in plant morphology (Supplementary Figures S8, S9).

Over-expression lines of A. thaliana exhibited increased levels of both short- and long-chain aliphatic glucosinolates compared to those of the wild-type (Table 2). Levels of 4-methylsulfinylbutyl-GS (4MSOB), the major aliphatic glucosinolate in A. thaliana, were found to be increased by 1.5- to 1.8-fold in over-expression lines. Levels of other short-chain aliphatic glucosinolates, such as 3-methylsulfinylpropyl-GS (3MSOP) and 5-methylsulfinylpentyl-GS (5MSOP), were increased by 1.7- to 2.3-fold. Accumulation of the long-chain aliphatic glucosinolate, 8-methylsulfinyloctyl-GS (8MSOO), also increased by 1.4- to 1.9-fold compared with that of the wild-type plants. The levels of indolic glucosinolates such as indolyl-3-methyl-GS (I3M), 1-methoxyindolyl-3-methyl-GS (1MOI3M), 4-methoxyindolyl-3-methyl-GS (4MOI3M), and 4-hydroxyindolyl-3-methyl-GS (4HOI3M), were found to be unaltered in the transgenic lines.

Levels of the main aliphatic glucosinolates, such as 2-hydroxy-3-butenyl-GS (2HO3BE), 2-propenyl-GS (2PE), 3-butenyl-GS (3BE), and 4MSOB were analyzed in the bolting stems of transgenic Chinese kale. The over-expression lines contained approximately 2- to 3-fold higher levels of 4MSOB compared to that of the wild-type (Table 3). Levels of 3BE, the alkenyl product of 4MSOB, were increased by 1.6- to 2.3-fold in over-expression lines. The level of 2HO3BE, the hydroxyl product of 3BE, was also increased by 1.9- to 2.6-fold. Accumulation of 2PE increased by 1.4- to 2.0-fold compared with that of the wild-type. The levels of 2HO3BE, 2PE, 3BE, and 4MSOB significantly decreased in the RNAi lines (Table 3). Levels of indolic glucosinolates such as I3M, 1MOI3M, 4MOI3M, and 4HOI3M were not affected in the transgenic lines.

BoaMYB28 transgenic lines of A. thaliana (OE.Ara1) and Chinese kale (OE.Boa1 and RNAi.Boa1) were used for the qRT-PCR (Figures 4, 5). Increased transcript levels of genes involved in aliphatic glucosinolate biosynthesis corresponded well with the increased transcript level of BoaMYB28 in the over-expression lines. In over-expression lines of A. thaliana, the expression levels of MYB28, side-chain elongation genes and core structure formation genes were 1.63-fold, 1.75- to 2.23-fold, and 1.39- to 4.02-fold higher than those in the wild-type. The expression levels of MYB28, side-chain elongation genes and core structure formation genes in over-expression lines of Chinese kale were 1.45-fold, 1.62- to 2.34-fold, and 1.21- to 3.18-fold higher than those in the wild-type. In contrast, a significant decrease in the expression levels of the target genes was observed in BoaMYB28 RNAi lines. The results indicated that BoaMYB28 was a positive regulator of aliphatic glucosinolate biosynthetic pathway genes in Chinese kale.

BoaMYB28 up-regulated the aliphatic glucosinolate biosynthesis genes. MAM1 and MAM3 are involved in methionine chain elongation, the first phase of glucosinolate biosynthesis (Textor et al., 2007; Redovniković et al., 2012). MAM1 catalyses the production of short-chained aliphatic glucosinolates, whereas MAM3 is involved in the biosynthesis of long-chained aliphatic glucosinolates (Textor et al., 2007). The transcript levels of MAM1 and MAM3 in over-expression lines of A. thaliana were 1.82-fold and 2.05-fold higher than those in the wild-type, while the expression levels of MAM1 and MAM3 in over-expression lines of Chinese kale were 1.57-fold and 1.93-fold higher than those in the wild-type. However, the expression level of MAM1 was higher than that of MAM3 in over-expression lines of A. thaliana and Chinese kale. Although the expression level of MAM1 was higher than that of MAM3, the change interval of MAM3 was larger than MAM1. Consequently, MAM3 was more tightly regulated by BoaMYB28 compared with MAM1.

CYP79F1 and CYP79F2 are involved in biosynthesis of core structure of aliphatic glucosinolates (Chen et al., 2003). CYP79F1 is responsible for the generation of short-chained aliphatic glucosinolates, while CYP79F2 is capable of mediating the biosynthesis of long-chained aliphatic glucosinolates (Tantikanjana et al., 2004; Miao et al., 2013). In this study, CYP79F1 and CYP79F2 were differentially regulated by BoaMYB28. The transcript levels of CYP79F1 and CYP79F2 in over-expression lines of A. thaliana were 1.74-fold and 1.53-fold higher than those in the wild-type, while the expression levels of CYP79F1 and CYP79F2 in over-expression lines of Chinese kale were 1.98-fold and 1.62-fold higher than those in the wild-type. The expression level of CYP79F1 was higher than that of CYP79F2 in over-expression lines of A. thaliana and Chinese kale. The results indicated that expression of CYP79F1 was more tightly controlled by BoaMYB28 compared with CYP79F2.

Other core structure formation genes (CYB83A1, ST5B, and ST5C) also showed the same and dramatic change in BoaMYB28 transgenic lines of A. thaliana and Chinese kale, which was similar with the variation of BoaMYB28. The results indicated that CYB83A1, ST5B, and ST5C were positively regulated by BoaMYB28.