SNP in DFR1 Coding Sequence Is Tightly Associated with Anthocyanin Accumulation in Cabbage ( B. oleracea var. capitata f. alba ) at Low Temperature

: Keeping green leaf color at the time of harvest is one of the important traits for breeding of Brassica oleracea var. capitata f. alba , and this trait is related to low anthocyanin contents. To understand the di ﬀ erential accumulation of anthocyanins in cabbage, we selected high anthocyanin accumulators (HAAs) and low anthocyanin accumulator (LAAs) of cabbages and examined the anthocyanin content and the expression of anthocyanin biosynthesis-related genes. Among many genes investigated, BoDFR1 was found to be closely related to anthocyanin accumulation, even under low temperature (LT) conditions. BoDFR1 sequence analyses between HAAs and LAAs revealed that there is a single nucleotide polymorphism (SNP) (1118T / A) in the coding sequence, which substitutes one amino acid from Leu261 to His261; we named BoDFR1 with His261 substitution as BoDFR1v . This amino acid substitution did not a ﬀ ect dihydroﬂavonol 4-reductase (DFR) activity and substrate speciﬁcity, but the polymorphism showed tight association to the BoDFR1 expression, i.e., high level expression of BoDFR1 and low level expression of BoDFR1v under LT conditions. The high levels of BoDFR1 expression were due to the high levels of BoMYB114 and BobHLH expressions combined with low level expression of BoMYBL2 , a repressor MYB . On the other hand, low levels of BoDFR1v expression seemed to be related to very low level expressions of BoMYB114 and BobHLH combined with a high level expression of BoMYBL2 . It seems that di ﬀ erent expression levels of these regulatory genes for MBW (MYB-bHLH-WD40) complex between HAAs and LAAs regulate BoDFR expression and anthocyanin accumulation. Using a single nucleotide polymorphism (SNP) between BoDFR1 and BoDFR1v , molecular markers for PCR and high resolution melt analyses were developed and validated to distinguish between HAAs and LAAs. Combined use of the BoDFR1 SNP marker with other stress markers, such as a cold tolerant marker, will greatly improve cabbage breeding.


Introduction
Anthocyanins are water-soluble plant pigments that range from blue to red in color. Anthocyanins impart color to leaves, flowers, and fruits, thus attracting pollinators and seed distributors [1][2][3][4], accumulators (HAAs) and low anthocyanin accumulators (LAAs), under LT conditions, and selected BoDFR1 in both cabbage types as the target gene that regulates anthocyanin accumulation. BoDFR1 expression levels were closely related to a single nucleotide polymorphism (SNP) identified in the gene (BoDFR1 in HAAs vs. BoDFR1v in LAAs) as well as to the level of anthocyanins. This SNP showed perfect association with the expression of upstream regulatory genes and was used to discriminate between HAAs and LAAs.

Plant Materials and Growth Conditions
Cabbage (Brassica oleracea var. capitata) varieties used in this study were divided into two groups: one used for sequence and expression analyses (Table 1) and the other used for marker validation (Table S1). Seed and leaf samples of cabbage varieties were kindly provided by Korean Seed Companies or the GSP Horticultural Center of Suncheon National University, Chonnam, Korea. Initially, eight inbred lines, including two purple cabbage lines, were obtained from the Biotechnology and Breeding Institute of Asia Seed Co., Icheon-si, Gyeonggi-do, Republic of Korea (Table 1). Then, additional lines were selected for detailed analysis, including Speed King, Golden Cross, CT-415, T-523, and T-530, from more than 40 inbred lines and cultivars grown in the field under LT conditions. Seeds of these additional lines were kindly provided by Joen Seed Co., Goesan-gun, Chungbuk, Korea, and Asia Seed Co. These lines were usually classified into two groups under LT conditions, HAAs and LAAs. Seeds were germinated, and plants were grown for 3 weeks in a growth chamber under controlled conditions (23 • C, 16 h light/8 h dark photoperiod and 140 µmol m −2−1 light intensity). Then, plants were grown in a growth chamber under LT (<10 • C) or normal conditions (control) for 1 month. To conduct experiments under natural conditions, plants were cultivated inside and outside the greenhouse of Chungnam National University, Daejeon, Korea. LT treatment (<10 • C) under natural conditions (outside) was conducted for 30 days, whereas plants grown under normal conditions (inside) were maintained in the greenhouse at set minimum (20 • C) and maximum (30 • C) temperatures. Seedlings of the inbred line BN106 were grown in the growth room for 3 weeks and exposed to LT (10 • C, 4 • C, or 0 • C) for 2 h.

Analysis of BoDFR1 Gene and Promoter Sequences
Genomic DNAs were extracted from the leaves of cabbage genotypes listed in Table 1, except BN106, using the DNeasy Plant Mini Kit (QIAGEN, Hilden, Germany). The promoter and the genomic DNA (from ATG to TGA), including the coding sequence (CDS) of the BoDFR1 gene, were cloned by PCR using Ex Taq (Takara, Kyoto, Japan) and sequence-specific primers designed on the basis of BoDFR1 (Bol035269) sequence obtained from the Brassica database (http://brassicadb.org/brad/index.php); primers DFR-F (5 -ATGGTAGCTCACAAAGAGACCGT-3 ) and DFR-R (5 -CTAAGCACAGATCTGCTGTGCCG) were used to amplify 1.6 kb BoDFR1 genomic DNA from ATG to TGA, and primers pDFR-F1 (5 -CCTCTCACTTCGAAGCCTTCCT) and pDFR-R (5 -CACGGTCTCTTTGTGAGCTACCAT) were used to amplify the BoDFR1 promoter region. PCR conditions were as follows: initial denaturation at 94 • C for 5 min, followed by 30 cycles of denaturation at 94 • C for 30 s, annealing at 58 • C for 30 s, and synthesis at 72 • C for 1 min and a final extension at 72 • C for 7 min. PCR products were electrophoresed on 1% agarose gels and purified using LaBoPass Gel Extraction Kit (COSMOGENETECH, Seoul, Korea). The purified DNA fragments were cloned into the T&A cloning vector (RBC, Banqiao, Taiwan) and transformed into Escherichia coli strain DH5α. Plasmid DNA was isolated from the transformed bacteria using LaBoPass Plasmid Mini Kit (COSMOGENETECH, Seoul, Korea). To avoid PCR and sequencing errors and to identify different BoDFR1 alleles, at least ten clones were sequenced for each genotype. The obtained sequences were compared and analyzed using the multiple sequence alignment program CLUSTAL Omega (EMBL-EBI, UK), and cis-elements in the promoter region were identified using web-based promoter analysis tools, PLACE (http://www.dna.affrc.go.jp/htdocs/PLACE/) and PlantPAN (http://PlantPAN2.itps.ncku.edu.tw). To validate T/A SNP, 1177 bp fragment spanning exons 1-5 was amplified using the forward primer, 5 -TGCCACTGTTCGCGATCCTG-3 , and reverse primer, 5 -CATCCACTCCTTCAAACCTGTTCCA-3 , and ca. 30 bp were used for an alignment.

Expression Analysis of BoDFR1
Total RNA was extracted from leaf samples with TRIzol Reagent (Thermo Scientific, Waltham, USA), and 1 µg of total RNA was treated with RQ1 RNase-free DNase (Promaga, Madison, USA) at 65 • C for 10 min. First-strand cDNA was synthesized using oligo dT primers and Revertra Ace Kit (TOYOBO, Osaka, Japan). Primers used to analyze BoDFR expression are listed in Table S2. Semi-quantitative RT-PCR was performed as follows: initial denaturation at 94 • C for 5 min, followed by 27 cycles of 94 • C for 30 s, 60 • C for 30 s, and 72 • C for 30 s, and a final extension at 72 • C for 7 min. DNA fragments amplified by semi-quantitative RT-PCR were separated on 1.5% agarose gels and stained with ethidium bromide. Quantitative real-time PCR (qRT-PCR) was performed on the MiniOpticon system (Bio-Rad, Hercules, California, USA) using SYBR Green Realtime Master Mix (TOYOBO, Osaka, Japan). Primers for BoDFR1 were designed to amplify 275 nucleotides between exon 4 and exon 6: forward primer, 5 -CTCTCTCCTATCACTCGTAACGAGGC and reverse primer, 5 -TCCCATATCAATCAGCTTCTTGGAACTG. PCR conditions were as follows: denaturation at 95 • C for 10 min, followed by 40 cycles of 94 • C for 30 s, 58 • C for 30 s, and 72 • C for 30 s. Fluorescence values were measured at the last step of each cycle. All analyses were performed with three biological replicates. Transcript levels of target genes were normalized relative to the BoActin2 gene (internal reference) and analyzed using the 2 −∆∆C T method [68].

Expression Analysis of Anthocyanin Biosynthesis-Related Genes
Total RNA extraction and cDNA synthesis were performed as described above. Expression analysis of anthocyanin biosynthesis-related genes (two representative genes for EBGs and LBGs and eight regulatory genes) was performed by qRT-PCR using primers listed in Table S3. Relative transcript levels were quantified as described above.

Anthocyanin Extraction and Quantification
Anthocyanin content was measured as described previously [69]. Briefly, 1 mL of 1% (v/v) HCl-methanol was added to 0.3 g of the powdered sample. The sample was shaken gently in darkness at room temperature for 24 h and then centrifuged at 13,000 × g. The upper aqueous phase was collected, and its absorbance was measured at 530, 620, and 650 nm. Relative anthocyanin content was calculated according to the following equation: where A530, A620, and A650 represent the absorbance of the aqueous layer at 530, 620, and 650 nm, respectively. Total anthocyanin content (mg anthocyanin/100 g fresh weight) was then calculated on the basis of the reference, cyanidin-3-oglucoside [70], according to the following equation: where MW is the molecular weight of cyanidin-3-glucoside (449.2 g/mol); DF is the dilution factor; and is the cuvette optical path length (1 cm).

Establishment of PCR-Based Assay
To discriminate between BoDFR1 and BoDFR1v, an allele-dependent PCR assay was developed, which utilized a common reverse primer and two forward allele-dependent primers in a single reaction. Forward primers possessed different tails containing a destabilizing mismatch (or artificial mismatch) ( Table 2) [71,72], while the reverse primer was complementary to a common reference sequence located 200-300 bp away from the forward primer-binding site. Several PCRs were conducted to select primer combinations with high SNP discriminating power and to optimize template concentration, annealing temperature, and amplification cycle number. The selected primer combinations and PCR reaction conditions are summarized in Table 2. PCR reactions contained 5-10 ng of genomic DNA template, 4 µL of Taq Premix (Elpisbiotech, Daejoen, Korea), and 1 µL (10 pmol) each of forward and reverse primer in a total volume of 20 µL.

High Resolution Melt (HRM) Analysis
To detect SNPs in BoDFR1 rapidly and efficiently, HRM analysis was carried out on a LightCycler 96 (Roche, Basel, Switzerland) instrument post-PCR in a volume of 20 µL containing 10 ng of template DNA, 1 µL (10 pmol) each of forward, reverse, and probe primer, and 10 µL of the LightCycler 480 High Resolution Melting Master Mix (Roche, Basel, Switzerland). The reaction was pre-incubated for 10 min at 95 • C, followed by 50 cycles of amplification including denaturation for 10 s at 95 • C, annealing for 20 s between 63 and 55 • C under touchdown command, and synthesis for 20 s at 72 • C. The amplified PCR products were then subjected to the following HRM cycle: 60 s at 95 • C, 120 s at 40 • C and 1 s at 83 • C. The fluorescence value (-dF/dT) was recorded during the final step with six readings per • C. The LightCycler 96 software (Roche, Basel, Switzerland) was used to analyze the HRM curves. Each of the samples was appropriately labeled in sample editor. Both delta Tm discrimination and curve shape discrimination were set at 75%, where a 0.2 positive/negative threshold level was set to obtain the final output. Primers used for HRM analysis are listed in Table 2.

Recombinant DFR1 Protein Expression
To generate N-terminal His-tagged fusions of BoDFR1 and BoDFR1v proteins, the BoDFR1 gene was amplified from the cDNA of HAA and LAA inbred lines 154 and 09WH45, respectively, by PCR using Ex Taq (Takara, Kyoto, Japan), forward primer BoDFR-poxF (5 -ACAGAATTCATGGTAG CTCACAAAGAG-3 ) containing EcoRI restriction site (underlined), and reverse primer BoDFR-poxR (5 -GCAAGCTTGAGCACAGATCTGCTGTG-3 ) containing HindIII restriction site (underlined). PCR products were digested with EcoRI and HindIII restriction enzymes and cloned into the pET28a expression vector (Sigma-Aldrich, St. Louis, Missouri, USA) previously digested with the same enzymes. The resulting plasmids were transformed into E. coli DH5α cells, purified and sequenced. The sequence confirmed plasmids were transformed into E. coli strain BL21 cells (RBC, Banquiao, Taiwan), and protein expression was induced by isopropyl-thio-β-D-galactoside (IPTG; 0.3 mM) treatment at 23 • C for 16 h. Cells were harvested, resuspended in phosphate-buffered saline (PBS), and lysed by sonication. The His-tagged BoDFR1 and BoDFR1v proteins were purified by Ni Sepharose High Performance His-tagged protein purification resin (GE Healthcare, Chicago, Illinois, USA), and the purified proteins were dialyzed using MOPS (3-(N-morpholino)propanesulfonic acid) buffer. Protein concentrations were measured by the Bradford assay and confirmed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) on 10% polyacrylamide gels.

Enzyme Activity Assay
Enzyme activity of BoDFR1 and BoDFR1v was examined as described previously [73] using 10 mg/ml of DHK (AdipoGen, San Diego, California, USA), DHQ (Sigma-Aldrich, St. Louis, Missouri, USA), and DHM (Sigma-Aldrich, St. Louis, Missouri, USA) substrates. A reaction (500 µL total volume) containing 100 mM Tris-HCl buffer (pH 7.0), 35 µg of enzyme extract, 100 µg of substrate, and 2 mM NADPH (Reduced nicotinamide adenine dinucleotide phosphate) was incubated at 30 • C for 1 h. The activity of DFR1 was determined by measuring the rate of NADPH oxidation. To determine the content of NADPH, absorbance at 340 nm was measured using a spectrophotometer and calculated according to Beer's law: where A is the absorbance of a 1 M solution of NADPH at 340 nm; ε is the extinction coefficient of NADPH (6.22 mM −1 cm −1 ); is the path length of the cuvette in cm; and C is the molar concentration.

Statistical Analysis
Anthocyanin contents, BoDFR1 activity, temperature, and gene expression levels were analyzed using one-way analysis of variance (ANOVA) and Tukey's pair-wise comparisons in Minitab v.17 (Minitab Inc., State College, PA, USA). The data were presented as the mean of three biological replicates ± standard deviation (±SD).

Identification of BoDFR1 as the Key Gene Responsible for Anthocyanin Accumulation under LT Conditions
To identify the gene that regulates anthocyanin accumulation under LT conditions, we examined the expression profiles of two early anthocyanin biosynthetic genes (BoCHS and BoF3H) and two late anthocyanin biosynthetic genes (BoDFR1 and BoLDOX) encoding structural proteins in the LT tolerant inbred line BN106. The temperature was decreased to 0 • C, and gene expression was analyzed by qRT-PCR ( Figure 1A). The expression of BoF3H and BoDFR1 increased in response to the decrease in temperature, whereas the expression of BoCHS and BoLDOX decreased or remained constant. Since the increase in BoDFR1 expression in response to LT was more notable than that of BoF3H, we selected BoDFR1 for further analysis. Plants were grown in the greenhouse for 4 weeks and then placed either in the greenhouse (G) or in the field (L) (under LT) for 4 weeks. Average temperature varied from 20 • C (low) to 30 • C (high) in the greenhouse, and from 6 • C (low) to 18 • C (high) in the field. Putative BoDFRs were selected on the basis of B. rapa DFR sequences [49].
The presence of several DFR genes with differential expression under different abiotic stresses is suggested to have functional implications [49]. Based on the DFR sequences of Chinese cabbage, we identified eight putative DFR genes in cabbage (http://brassicadb.org/brad/) (Table S4) and investigated their expression by semi-quantitative RT-PCR in three HAA genotypes (337, 154, and 2437) under LT ( Figure 1B). In these genotypes, three putative BoDFR genes (Bol006592, Bol015141, and Bol024018) were expressed constitutively, whereas four genes showed no expression; only one gene (Bol035269) showed LT-dependent expression ( Figure 1B). An ortholog of Bol035269 (Bo9g058630) was identified in B. oleracea var. oleracea (http://plants.ensembl.org/Brassica_oleracea/Info/Index), which is hereafter referred to as BoDFR1. The expression of BoDFR1 was then studied in various cabbage genotypes exhibiting differential anthocyanin accumulation under LT.
Initially, we focused on six inbred lines, three of which accumulated anthocyanins to high levels under LT (HAAs: 337, 154, and 2437) but three to show low levels (LAAs: 09WH45, 842, and 2409). Based on RT-PCR results (data not shown), we selected four inbred lines, including two HAAs (337 and 154) and two LAAs (09WH45 and 2409), for further analyses. First, we examined the level of anthocyanins and BoDFR1 expression in these lines grown at LT (<10 • C) for at least 30 days in a growth chamber. Upon LT exposure, anthocyanin content increased 4-10-fold in both HAA genotypes, and this increase was proportional to the increase in BoDFR1 expression (5.5-10-fold) ( Figure 2). However, anthocyanin accumulation and BoDFR1 expression levels were not remarkable in the LAA genotypes. Next, to examine the relationship between anthocyanin accumulation and BoDFR1 expression under natural environmental conditions, we grew more than 40 cabbage genotypes, including previously examined inbred lines, either in the greenhouse or field at LT (<10 • C) for 1 month. Four HAAs (154, T-523, T-530, and CT415) and four LAAs (09WH45, 2409, Golden Cross, and Speed King) were selected from field grown genotypes (Table 1, Figure 3) for further expression analyses and marker development and validation. Under LT, the anthocyanin content of HAAs varied from 30-53 mg/100 g fresh weight (FW), whereas that of LAAs varied from 4-16 mg/100 g FW ( Figure 4A). The expression level of BoDFR1 was more than 5-fold higher in HAAs than in LAAs ( Figure 4B), indicating that BoDFR1 expression is closely associated with anthocyanin content.

Analysis of the BoDFR1 Gene Sequence Isolated from HAAs and LAAs
To investigate any relationship between genomic sequences and the expression levels of BoDFR1 in HAAs and LAAs, we isolated the BoDFR1 gene, including its promoter, from six HAAs (337, 154, 2437, T-523, T-530, and CT-415) and five LAAs (09WH45, 842, 2409, Golden Cross, and Speed King) ( Table 1) using primers designed against Bol035269 (http://brassicadb.org/brad/) and Bo9g058630 (http://plants.ensembl.org/Brassica_oleracea/Info/Index) sequences and sequence information obtained from the National Center for Biotechnology Information (NCBI) website (https://blast.ncbi.nlm.nih. gov/Blast.cgi) (Table S2). To avoid any errors due to PCR and sequencing, we analyzed at least ten clones for each sample. The promoter region of BoDFR1 Next, analysis of the number and the distribution of cis-acting elements revealed two MYB-binding sites (MBS1 or MYB-core; CNGTTR) [74], three bHLH-binding sites (E/G-BX; CANNTG/CACGTG) [75,76], and two MYB recognition sites (MYB1AT; A/TAACCA) [77] in the BoDFR1 promoter region, and their sequences were highly similar between HAAs and LAAs ( Figure 5). By contrast, the MYBD-binding site (MYBST1; GGATA) [78] at −581 bp was present in reverse orientation in BoDFR1 promoter. However, the presence of MYBST1 might not have affected BoDFR1 expression and anthocyanin accumulation because the BoDFR1 gene in red (or purple) cabbage genotypes 7S4-63 and 7S4-51 (GenBank accession numbers: MN537971 and MN537972, respectively), which exhibited high anthocyanin accumulation and high BoDFR1 expression, does not contain MYBST1 in its promoter.

In Vitro Enzyme Activity Assay of BoDFR1 and BoDFR1v
Previously, Johnson et al. (2001) [79] reported that minor changes in the DFR gene affect the activity and the substrate specificity of the encoded protein. Additionally, we identified one amino acid substitution in BoDFR1v (L261H). Therefore, we examined the enzyme activity of BoDFR1 and BoDFR1v by expressing His-tagged recombinant proteins in E. coli. Based on the crystal structures of both these proteins ( Figure 6A), we hypothesized no differences in their enzyme activity or substrate specificity. The L261H substitution in BoDFR1v is located in the α10 helix of the small C-terminal domain, distant from the ball-and-stick structure, which is important for substrate-and NADP + -binding [80]. We generated recombinant BoDFR1 and BoDFR1v constructs ( Figure 6B) and purified the His-tagged recombinant DFR proteins ( Figure 6C). Enzyme activity, expressed as the decline in the oxidation rate of NADPH, differed among the substrates DHK, DHQ, and DHM but not between BoDFR1 and BoDFR1v ( Figure 6D). These results suggest that the L261H amino acid substitution does not affect the activity or substrate specificity of the DFR protein.

Relationship between BoDFR1 and Other Anthocyanin Biosynthesis-Related Genes
BoDFR1 showed allele-dependent expression (Figure 4), and anthocyanin content was proportional to the expression level of BoDFR1 under LT treatment. However, the sequence of the promoter, which controls gene expression levels, was highly similar among HAAs (337, 154, 2437, T-523, T-530, and CT-415) and LAAs (09WH45, 842, 2409, Golden Cross, and Speed King) ( Figure 5, Figure S1). Additionally, both BoDFR1 and BoDFR1v showed the same substrate specificity and enzyme activity ( Figure 6). These results led us to hypothesize that the function of upstream regulatory gene(s) is associated with the DFR1 expression level. To test this hypothesis, we examined the expression levels of anthocyanin biosynthetic genes (encoding structural and regulatory proteins) under normal and LT conditions in the same samples as those used in Figure 4, and BoDFR1 expression data served as a reference (Figure 7, Figure 8). The expression of EBGs, BoCHS and BoF3H, was similar between HAAs and LAAs; however, the expression of LBGs, BoLDOX and BoDFR1, was lower in LAAs than in HAAs (Figure 7). This similar expression pattern of BoLDOX and BoDFR1 may reflect that BoLDOX levels were dependent to DFR activity in directly.  Figure 3). Gene expression was analyzed by qRT-PCR. Expression level of BoActin2 was set to one and used for data normalization. Data represent mean ± SD (n = 3).
We also examined the expression of anthocyanin biosynthesis regulatory genes ( Figure 8). Expression of BoHY5 and BoMYBH1 (a homolog of MYBD) was slightly increased by LT, but no significant difference was detected between HAAs and LAAs. The expression of BoMYBH2 showed no significant differences between HAAs and LAAs or between normal and LT conditions. However, the expression of genes comprising the MBW complex, BoMYB114 (representative MYB in cabbage), BobHLH, and BoTTG1 (WD-repeat gene), seemed to provide some clues about the expression of BoDFR1. Expression levels of BoMYB114 and BobHLH were higher in HAAs than LAAs, regardless of the temperature (normal vs. LT). However, the expression of the genes did not show a significant increase upon exposure to LT; instead, the expression of BobHLH in LAAs was decreased by LT treatment. Although the expression of BoTTG1 showed no significant difference between HAAs and LAAs, it increased significantly upon LT treatment. These results reveal a novel phenomenon in cabbage, which explains why LAAs exhibit low anthocyanin content and low BoDFR1v expression under LT condition. Moreover, expression of the repressor of MBW complex, BoMYBL2, was low in HAAs and greatly increased in LAAs in response to LT. The expression of BoMYBL2 might be negatively associated with BoMYBD expression suggested by a previous study [81]. Thus, the expression pattern of BoMYBL2 may also contribute to BoDFR1 expression and anthocyanin accumulation in cabbage.  Figure 3). Gene expression was analyzed by qRT-PCR. Expression level of BoActin2 was set to one and used for data normalization. Data represent mean ± SD (n = 3). BoDFR1 data were same as that in Figure 4B and used as a reference.
In Arabidopsis, anthocyanin accumulation is associated with freezing tolerance, and freezing tolerant accessions show higher DFR expression levels than freezing sensitive accessions [8,9,82]. In apple, a MYB transcription factor binds to both the CBF and the DFR promoters, thereby activating their expression [13]. However, cabbage genotypes (HAAs and LAAs) used in this study seemed to be freezing tolerant, as some of them survived the Korean winter (data not shown). Therefore, any association of BoDFR1 or BoDFR1v expression with cold tolerance (or anthocyanin accumulation and cold tolerance) must be elusive. In cabbage genotypes used in this study, all four CBF genes were upregulated upon LT treatment with no difference in expression between HAAs and LAAs ( Figure 9). This suggests that anthocyanin accumulation is not related to cold tolerance at least in the genotypes used in this study.  Figure 3). Expression level of BoActin2 was set to one and used for data normalization. Data represent mean ± SD (n = 3).

Development of a Molecular Marker to Distinguish between BoDFR1 and BoDFR1v
We designed PCR primers to distinguish between BoDFR1 and BoDFR1v based on cytosine (C)/adenine (A), cytosine (C)/thymine (T), and thymine (T)/adenine (A) SNPs at nucleotide positions 423, 628, and 1118, respectively ( Figure 5, Table 2). First, PCR was performed using these primers and genomic DNA extracted from six inbred lines provided by Asian Seed Co. ( Figure 10A) and eight inbred lines and cultivars selected in Figure 3 for further analysis ( Figure 10A,C, Table 2). All primer pairs could discriminate the six inbred lines based on the three SNPs ( Figure 10A), but only the primer pair specific to the T/A SNP at position 1118 correctly discriminated among the eight inbred lines and cultivars ( Figure 10C). Cabbage cultivars T-523, T-530, and CT-415 were heterozygous for SNPs at nucleotide positions 423 and 628 ( Figure S1, Figure 5). Then, based on these PCR results, we designed HRM primers targeting the T/A SNP at nucleotide position 1,118. The results showed that this SNP distinguished between BoDFR1 and BoDFR1v ( Figure 10B,D). Therefore, we conclude that the T/A SNP could be used to discriminate between HAAs and LAAs under LT conditions. To validate this marker, we used the HRM primer set to genotype 120 cabbage inbred lines and cultivars ( Figure 11A,B). Among these 120 samples, 70 (58%) were homozygous for the TT allele (HAA homozygote), 32 (27%) were homozygous for the AA allele (LAA homozygote), and 18 (15%) were heterozygous (TA allele). Interestingly, all Chinese cultivars tested in this study carried the AA type (BoDFR1v) ( Table S1). Some of the heterozygotes were confirmed by sequencing ( Figure S2), and association of the T/A SNP with anthocyanin accumulation was confirmed by examining the level of anthocyanins and BoDFR1 expression ( Figure 11C,D). The phenotype of heterozygotes (T/A allele) resembled that of HAA homozygotes (TT allele), indicating dominance of the T allele (HAA genotype was TT or TA). These results indicate that the HRM primer pair developed in this study is useful for distinguishing between BoDFR1 and BoDFR1v genotypes. Notably, HAAs were more abundant than LAAs or heterozygotes, and BoDFR1 was dominant to the BoDFR1v allele.
We also investigated the association between anthocyanin accumulation and BoDFR1 SNP in B. oleracea, Chinese cabbage, and mustard genotypes, which showed high anthocyanin accumulation and very low accumulation. As shown in Figure S3, the BoDFR1v allele was found in cabbage and ornamental kale (B. oleracea var. acephala) but not in cauliflower (B. oleracea var. botrytis), kohlrabi, Chinese cabbage, and mustard. Interestingly, F1 cultivars of ornamental or flower kale including M23 (Benitaka-red) and M24 (Shirotaka-white) were heterozygous and homozygous for the BoDFR1v allele, respectively. Recently, Guo et al. (2019) [83] reported that BoDFR1 expression is associated with anthocyanin accumulation under LT, but the authors did not discuss allele-dependent expression. Overall, we conclude that genotypes heterozygous at the 1,118 SNP show HAA phenotype.

Discussion
Anthocyanin accumulation in plants is closely related to abiotic stress tolerance, particularly LT tolerance, in plants and is proportional to DFR gene expression. Despite the positive effects of anthocyanins on human health, one of the breeding objectives of green cabbage is to maintain the deep green color of leaves at harvest, regardless of the season, especially under LT conditions. In this study, we investigated cabbage cultivars and inbred lines that show low anthocyanin accumulation under LT but retain their LT tolerance. Our results showed that BoDFR1 expression is related to anthocyanin accumulation in cabbage. Analysis of the BoDFR1 gene in contrasting cabbage genotypes (HAAs vs. LAAs) revealed that an SNP at nucleotide position 1118 in 1580 bp BoDFR1 genomic DNA is closely related to its expression levels. We used this SNP to develop a PCR-based marker as well as HRM marker for the identification of HAAs and LAAs under LT conditions.

Selection of HAA and LAA Lines
Analysis of six inbred lines (three HAAs and three LAAs) and five cultivars (three HAAs and two LAAs) of cabbage revealed that BoDFR1 expression was associated with anthocyanin accumulation under LT conditions (Figures 1-4). This association study was performed using plants grown either in an environmentally controlled growth chamber or in the field, where light intensities vary from 140 to 1300 µmol m −2 s -1 . Since light intensity influences anthocyanin biosynthesis [84], we examined gene expression under both conditions. Previous reports imply that the DFR gene is a good candidate for anthocyanin accumulation. The BoPr gene, which confers purple leaf color in ornamental kale (Brassica oleracea L. var. acephala), has been mapped and identified as Bo9g058630 (orthologous to Bol035269 in cabbage) and shows high homology to the gene encoding dihydroflavonol reductase in Arabidopsis (AT5G42800) [85]. In crabapple (Malus crabapple) leaves and apple peels, upregulation of McDFR or MdDFR expression, respectively, is accompanied by a proportional increase in the expression of other genes involved in anthocyanin biosynthesis [45]. This result and our data together imply that the BoDFR gene is a good candidate for the investigation of anthocyanin accumulation in cabbage.

BoDFR1 Sequence Analysis and SNP Detection
The genomic sequence of BoDFR1 was analyzed in 13 cabbage genotypes including six HAAs, five LAAs, and two red (or purple) cabbage cultivars (Table 1). However, sequences of BoDFR1 from two red cabbage cultivars were excluded from the alignment shown in Figure S1. The genomic DNA of BoDFR1 analyzed in this study comprised the promoter region (1380 bp upstream of the ATG start codon) and the 1580 bp genomic DNA from ATG to TAG stop codon, which includes six exons and five introns ( Figure 5). Sequences of BoDFR1 in 13 cabbage genotypes were highly similar. Only three SNPs were identified in the 1580 bp genomic DNA. Among these, only the T/A SNP at the 1118 nt was predicted to cause an amino acid substitution (Leu to His), and this SNP could be used to discriminate between HAA-type DFR1 (BoDFR1) and LAA-type DFR1 (BoDFR1v).
In the promoter region, the number and the distribution of cis-acting elements, such as MYB1AT, MYB-core (or MYBS1), and E/G-BX, were highly similar between HAAs and LAAs ( Figure 5); only MYBST1 was found in reverse orientation at -581 bp in BoDFR1 promoter compared with BoDFR1v promoter. HY5-induced MYBD binds to MYBST1 and the MYB-binding motif (A/TAACCA) in the promoter of the repressor MYBL2, resulting in the inhibition of MYBL2 expression [81]. Thus, the presence of the MYBST1 motif in the BoDFR1 promoter affects gene expression. However, the MYBST1 motif was not detected in the BoDFR1 gene isolated from purple (or red) cabbage, which is highly rich in anthocyanins. Moreover, the expression pattern of BoMYBD was not similar to BoMYBL2 expression ( Figure 8). These data suggest the possibility that the MYBST1 motif found in the BoDFR1 promoter does not affect the difference in anthocyanin levels between HAAs and LAAs.
A mutation in DFR promoter in onion [61] and two insertion/deletion mutations (InDels) in DFR CDS in kale [85] is associated with anthocyanin accumulation. Rearrangement of a 23 bp sequence in the promoter regions of MdMYB10 and McMYB10 is responsible for the difference in gene expression between white-and red-fleshed apples and red-leaf crabapples [86]. The bHLH transcriptional regulator binds to the CANNTG site (E/G-BX) present in the DFR promoter [50]. Additionally, a MYB transcription factor recognizes the MYBS1, thus regulating the expression of DFR [87]. Together, these data suggest that differential expression of the DFR allele results from differences in promoter sequences or cis-acting sequences. However, promoter sequences of BoDFR1 and BoDFR1v were highly similar, implying that differential expression of BoDFR1 between HAAs and LAAs probably resulted from differences in upstream regulatory proteins.

Regulation of BoDFR1 Expression
DFR regulates the essential step of anthocyanin biosynthesis and possesses a preference for a precursor of the three different types of anthocyanins [35,36]. However, BoDFR1 and BoDFR1v, which differed by one amino acid substitution, showed identical enzyme activity and substrate specificity ( Figure 6) because the substituted amino acid was distant from the active site of DFR [79]. This suggested that the difference in anthocyanin accumulation between HAAs and LAAs could be associated with the difference in BoDFR1 expression levels. However, no significant differences were detected in the promoter and the intron sequences between HAAs and LAAs ( Figure S1), which could control gene expression, suggesting the possibility that upstream regulatory transcription factors are involved in the regulation of BoDFR1 expression.
Data shown in Figure 8 provided hints as to why LAAs accumulate less anthocyanins than HAAs under LT conditions. The expression of BoTTG1 increased upon exposure to LT, but that of BoMYB114 and BobHLH did not, especially in LAAs. Instead, BoMYB114 and BobHLH expression levels were high in HAAs. These expression patterns represent a novel characteristic of cabbage plants because the expression of MYB and bHLH genes increases in purple kale in response to LT, whereas that of TT1 shows no change [33]. Similar results have been reported in Arabidopsis [88]. This difference in gene expression levels might explain why prolonged LT treatment (lasting at least one month) is required for increasing anthocyanin accumulation or inducing DFR expression in cabbage (data not shown). In addition, the expression of BoMYBL2 (a repressor) might also affect the formation of the MBW complex. BoMYBL2 expression was low in LAAs but greatly increased upon LT treatment ( Figure 8). The expression of BoMYBL2 might be slightly repressed by HY5-dependent MYBD (BoMYBD), as mentioned above [81], although this would not be critical. Taken together, these data suggest that the level of WD-repeat proteins would be sufficient in all cabbage types under LT, but the levels of BoMYB114 and BobHLH were high in HAAs and low in LAAs, regardless of the temperature. The level of BoMYBL2 increased in LAAs upon LT exposure, resulting in the inhibition of MBW complex formation. The MBW complex regulates the expression of LBGs, including DFR, at the transcriptional level [28,89,90]. Moreover, the MBW complex regulates bHLH (TT8) expression in a positive feedback loop, which ensures high level accumulation of proanthocyanins [91]. Our results suggest that formation of the MBW complex, which induced BoDFR1 expression, would be greatly reduced in LAAs but unaffected in HAAs, leading to differential anthocyanin accumulation. However, how the BoDFR1 SNP is associated with the expression of upstream regulatory components remains elusive.
CBF transcription factors play a key role in the cold acclimation of plants by regulating the expression of cold-regulated (COR) genes [92,93]. High level of DFR expression is associated with high anthocyanin accumulation, which confers cold (or freezing) tolerance in Arabidopsis [8,9,82], Chinese cabbage [49], tobacco (Nicotiana tabacum) [12], and purple kale [33,83]. In apple, a MYB transcription factor regulates both cold tolerance and anthocyanin accumulation by binding to the promoters of CBF and DFR genes, thus activating their expression [13]. However, four LAAs used in this study were freezing tolerant (data not shown). Additionally, we did not observe differences in the induction of BoCBFs between HAAs and LAAs under LT treatment (Figure 9), indicating that the mechanisms of anthocyanin accumulation and cold tolerance in cabbage were independent.

Molecular Marker Development and Application
In cabbage breeding programs, one of the desired horticultural traits is the green color of leaves at the harvest time, regardless of the season. To maintain green leaf color, anthocyanin accumulation should be repressed, which can be achieved by inhibiting BoDFR1 expression. Fortunately, cabbage genotypes harboring the BoDFR1v allele accumulate anthocyanins to a very low level not only under normal conditions but also under LT. The BoDFR1v allele harbors an A at nucleotide position 1118, whereas the BoDFR1 allele harbors a T. This SNP has also been detected in ornamental kale genotypes, M23 (Benitaka-red) and M24 (Shirotaka-white), which are heterozygous and homozygous for BoDFR1v, respectively ( Figure S3). Based on this T/A SNP, we developed and validated a PCR-based and an HRM marker, which perfectly discriminated between LAAs and HAAs (Figures 10 and 11). Although we could not explain why LAAs express BoDFR1v at low levels while maintaining cold tolerance, both PCR-based and HRM markers will be very useful for cabbage breeding.

Conclusions
Cabbage and some ornamental kale genotypes carried the BoDFR1v allele, which contains a single amino acid substitution (Leu to His) due to a single nucleotide change (T to A). However, both BoDFR1v and BoDFR1 showed identical enzyme activity and substrate specificity. BoDFR1v homozygotes exhibited green leaf color, regardless of the temperature, and a very low level of BoDFR1 expression. Using the T/A SNP identified between BoDFR1 and BoDFR1v, molecular markers were developed and validated to distinguish between HAAs and LAAs. Further investigation is needed to identify upstream elements or key regulators that control the expression of anthocyanin-related genes in LAAs, which could be further utilized in cabbage breeding programs.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4395/10/4/602/s1. Figure S1: Nucleotide sequence alignment of BoDFR1 promoter and genomic DNA from ATG to TGA isolated from 11 cabbage genotypes, Figure S2: Validation of the T/A SNP at nucleotide position 1,118 in BoDFR1 genomic DNA, Figure S3: Genotyping various B. oleracea, B. rapa, B. nuapus and B. juncea accessions using the 1118T/A SNP marker, Table S1: List of cabbage genotypes used in HRM analysis, Table S2: List of primers used for expression analysis and cloning of BoDFR genes, Table S3: List of primers used for qRT-PCR experiments, Table S4: Putative BoDRF genes identified in B. oleracea using B. rapa DFR (BrDFR) genes [49].