Isolation of Dihydroflavonol 4-Reductase cDNA Clones from Angelonia x angustifolia and Heterologous Expression as GST Fusion Protein in Escherichia coli

Blue Angelonia × angustifolia flowers can show spontaneous mutations resulting in white/blue and white flower colourations. In such a white line, a loss of dihydroflavonol 4-reductase (DFR) activity was observed whereas chalcone synthase and flavanone 3-hydroxylase activity remained unchanged. Thus, cloning and characterization of a DFR of Angelonia flowers was carried out for the first time. Two full length DFR cDNA clones, Ang.DFR1 and Ang.DFR2, were obtained from a diploid chimeral white/blue Angelonia × angustifolia which demonstrated a 99% identity in their translated amino acid sequence. In comparison to Ang.DFR2, Ang.DFR1 was shown to contain an extra proline in a proline-rich region at the N-terminus along with two exchanges at the amino acids 12 and 26 in the translated amino acid sequence. The recombinant Ang.DFR2 obtained by heterologous expression in yeast was functionally active catalyzing the NADPH dependent reduction of dihydroquercetin (DHQ) and dihydromyricetin (DHM) to leucocyanidin and leucomyricetin, respectively. Dihydrokaempferol (DHK) in contrast was not accepted as a substrate despite the presence of asparagine in a position assumed to determine DHK acceptance. We show that substrate acceptance testing of DFRs provides biased results for DHM conversion if products are extracted with ethyl acetate. Recombinant Ang.DFR1 was inactive and functional activity could only be restored via exchanges of the amino acids in position 12 and 26 as well as the deletion of the extra proline. E. coli transformation of the pGEX-6P-1 vector harbouring the Ang.DFR2 and heterologous expression in E. coli resulted in functionally active enzymes before and after GST tag removal. Both the GST fusion protein and purified DFR minus the GST tag could be stored at −80°C for several months without loss of enzyme activity and demonstrated identical substrate specificity as the recombinant enzyme obtained from heterologous expression in yeast.


Introduction
The genus Angelonia is a perennial plant which originates from South America. It belongs to the family Plantaginaceae (formerly Scrophulariaceae) [1], which comprises about 30 species, displaying a great diversity in form and colour ranging from blue to violet, white and pink ( Figure 1) [2,3]. The background of existing cultivar species is neither correctly reported nor investigated so far. Angelonia salicariifolia Humb. et Bonpl. (syn. Angelonia salicariaefolia Humb. et Bonpl., syn. Angelonia gardnerii Hook.), Angelonia angustifolia Benth., Angelonia grandiflora C. Morr., Angelonia integerrima Spreng. are currently considered as ancestral species [4][5][6]. We therefore use the term Angelonia 6 angustifolia. A. 6 angustifolia is sold as a flowering pot and bedding plant. It also has a high potential as a cut flower [7]. Spontaneous mutations occur which result in bicoloured flowers segregating in white and blue genotypes [2]. The bicolourness in Angelonia flowers is caused histogenetically (chimeral flower patterns) [2]. This makes it an interesting object for studies on the underlying molecular mechanisms.
In species which usually accumulate anthocyanins as plant pigments, white flowering genotypes result from a deficiency in at least one step of the anthocyanin pathway [8]. Such blockage of the anthocyanin pathway leading to white flower colour may occur at several levels. Deficiency in the activities of chalcone synthase (CHS), flavanone 3-hydroxylase (FHT), dihydroflavonol 4-reductase (DFR) and anthocyanidin synthase (ANS) results in the accumulation of their colourless upstream intermediates ( Figure 2). Frequently DFR and/or FHT are affected [9][10][11][12][13][14][15][16]. In contrast, absence of flavonoid 39-hydroxylase (F39H) and flavonoid 39,59-hydroxylase (F395H) activities only results in a shift of flower colour as this changes the anthocyanin spectrum in the flower ( Figure 2). A lack of enzyme activity can be based on mutations in the structural gene leading to a loss of functional activity or regulatory processes at the transcriptional or post-transcriptional level [17,18]. Post-transcriptional regulation processes are less understood, but micro RNAs seem to be largely involved [19][20][21][22][23][24]. At the transcriptional level, a variety of factors may result in lacking gene expression including mutations or transposable element insertions in structural genes and promoter regions, differentially expressed transcriptional factors and gene methylation.
The establishment of flower colour can also be influenced by the substrate specificity of the enzymes, particularly by DFR [25,26]. DFR is an oxidoreductase (EC 1.1.1.219) catalyzing the NADPH dependent stereospecific reduction of (+)-(2R,3R)-dihydroflavonols to the respective (2R,3S,4S)-flavan-2,3-trans-3,4-cis-diols (leucoanthocyanidins), as well as the reverse reaction in the presence of NADP + [27,28]. DFR is the first of the so-called 'late' enzymes of the flavonoid pathway which shows a major impact on the formation of anthocyanins, flavan 3-ols and flavonols. As a rule, DFRs accept dihydroflavonols independently from their hydroxylation pattern in the B-ring. Dihydrokaempferol (DHK, one hydroxyl group), dihydroquercetin (DHQ, two hydroxyl groups) and dihydromyricetin (DHM, 3 hydroxyl groups) are thus converted into leucopelargonidin, leucocyanidin and leucomyricetin, respectively. This provides bright-red pelargonidin, darkred/pink cyanidin and blue/violet delphinidin possessing one, two and three hydroxyl groups, respectively in the B-ring ( Figure 2). The DFR of some plants, however, show substrate specificity by accepting only dihydroflavonols showing at least two hydroxyl groups in the B-ring for example DFR from Petunia x hybrida, Nicotiana tabacum and Cymbidium hybrida [16,29].
We analyzed cyanic and acyanic flowers of Angelonia for the absence of enzymes from the anthocyanin pathway and demonstrated that the acyanic line is DFR deficient. For future studies into the molecular basis behind flower colour variegation in Angelonia we isolated for the first time DFR cDNA clones of this plant species. This study provides important clues for the investigation of DFR characteristics in general.

Results and Discussion
Enzyme activity in Angelonia 6 angustifolia with divergent flower colour Three key enzymes of the flavonoid pathway were tested in chimeral pink/blue, chimeral white/blue and homohistic white flowers of A. 6 angustifolia: CHS/CHI, FHT and DFR. ANS could not be included because demonstrations in crude plant preparations failed so far [10] (and unpublished results) and were, therefore, only studied with recombinant and/or purified enzymes [30]. Two different protocols [31] for the preparation of enzyme extracts from the flowers were tested. All three enzymes could be detected with preparations from flowers which were obtained according to a relatively simple extraction procedure with quartz sand and buffer, although with only moderate activity (Table 1). CHS and DFR activity were slightly higher when the preparations were obtained via a protocol particularly developed for polyphenol rich tissues [31,32] (data not shown). Whereas CHS/CHI and FHT activity was comparable in all three flower colour types, no DFR activity could be measured in white flowers in contrast to the white/blue and pink/blue flowers (Table 1). Heat inactivated enzyme preparations did not show any enzyme activity.

Cloning of DFR from Angelonia 6 angustifolia
Based on the DFR sequence information for the most closely related species available in the NCBI database, Antirrhinum pallida, Mimulus aurantiacus, Forsythia 6 intermedia and Solenostemon scutellarioides, degenerated primers (Table S1) were designed in conserved regions, which were used to isolate a fragment of a putative DFR cDNA clone from blue flowering petals of A. 6 angustifolia. Full sequence information was obtained by 59 and 39 RACE techniques (rapid amplification of cDNA ends) and specific primers were created (A-DFRfullF, A-DFRfullR, Table S1) which resulted in the isolation of two full size cDNA clones from the same plant material. Ang.DFR1 The two cDNA clones shared a sequence identity of 99% at the nucleotide level. In comparison to Ang.DFR1, Ang.DFR2 showed 4 exchanges and a deletion of 3 bp, which resulted in an exchange of two amino acids and a deletion of a proline in a proline-rich region at the N-terminus ( Figure 3). Compared to many other DFRs, the cDNA clones from Angelonia were approximately 50 to 100 amino acids longer, but the length was comparable to those of the DFRs from the closely related species Antirrhinum majus (Accession X15536) and Perilla frutescens (Accession AB002817). In the phylogenetic analysis, the Angelonia DFRs clustered together with DFRs from Torenia, Perilla, Solenostemon and Sinningia, which all belong to the order Lamiales ( Figure 4).

Heterologous expression in yeast and characterization of the recombinant enzymes
The cDNA clones were transferred into the pYES2.1 vector and heterologously expressed in yeast (Saccharomyces cerevisiae). The recombinant Ang.DFR2 was functionally active catalyzing the NADPH dependent conversion of dihydroflavonols into leucoanthocyanidins ( Table 2, grey-shaded). The enzyme demonstrated substrate specificity as described for a number of plants [25,26] and accepted DHQ and DHM as substrates but did not efficiently convert DHK. Johnson et al. [25] demonstrated that substrate specificity of DFR is determined in the amino acid region 132-157 in the Gerbera hybrida DFR, which corresponds to amino acids 141-168 in Ang.DFR2. The presence of an aspartic acid in position 134 (position 143 of Ang.DFR2) of Petunia hybrida contrasts with the highly conserved asparagine found in many plant species and is suggested to be essential for the substrate specificity [25]. However, Ang.DFR2 showed the same substrate specificity as the P. hybrida DFR despite the presence of an asparagine in this position. Ang.DFR1 in contrast did not show functional activity with any dihydroflavonol tested ( Table 2, greyshaded). The recombinant Ang.DFR2 was characterized in detail and the result is summarized as a standard enzyme assay in the Material and Method section and in Table 3.

Testing DFR activities
Substrate specificity of DFR has been intensively studied in various plants [25,26,28,[33][34][35][36][37][38][39][40][41]. Substrate acceptance of DFR is either studied in vitro using recombinant enzymes, using enzyme preparations from plant sources, or in planta by monitoring changes in the flower colour of genetically modified plants. When substrate acceptance is compared in vitro, several sources of error exist. Firstly, it is essential to consider the stereospecificity of DFR, which accepts only (+)-(2R,3R)-dihydroflavonols as substrates   [42]. Commercially available dihydroflavonols are frequently racemic mixtures and may occur in various qualities. The presence of impurities or the wrong stereoform may significantly hamper the reaction. Incubation of Ang.DFR2 with equimolar amounts of (+)-DHQ and a racemic mixture ((6)-taxifolin hydrate), respectively, reduced conversion rates of (6)-taxifolin by approximately 72% compared to (+)-DHQ. This is significantly higher than the 50% one would have expected from the fact that only one of the two stereoforms is converted. For comparison of different dihydroflavonols, it is therefore essential to use all of the same quality, stereoform and in equimolar amounts (same nmol, not mg). Another problem stems from the instability of leucoanthocyanidins, which are rapidly oxidised and polymerized and may react with proteins [43][44][45][46][47][48][49]. With an increasing number of hydroxyl groups, compounds show a higher sensitivity towards oxidation and polymerization. Losses may thus vary for the different substrates and this could bias conclusions on substrate preference. Long incubation times (hours) should therefore be avoided. As a rule, substrates and products are extracted with ethyl acetate and conversion rates are determined after separation on TLC plates [37,38,50] or HPLC [33]. However solubility of leucoanthocyanidins in organic solvents varies [45] with their polarity, which increases with the number of hydroxyl groups. Compared to leucoanthocyanidins, dihydroflavonols are more easily extracted by ethyl acetate (Table S2). For DHK and DHQ this causes an acceptable error in conversion rates of approximately 5%. Solubility of leucodelphinidin decreases dramatically however when compared to DHM and the other leucoanthocyanidins (Table S2). As a result biased results are obtained for the conversion of DHM (Table 4). When leucodelphinidin was extracted with ethyl acetate, recovery rates were low in both organic and aqueous solvents (Table S2). We assume that this is a result of unspecific interactions with proteins in the interlayer after oxidation of the phenolic groups. Thus we recommend the separation of whole assays on paper stripes as described for glycosyltransferases [51]. This enables exhaustive recovery and reliable conversion rates for DHM.
Some authors demonstrate functional activity of DFRs by conversion of the formed leucoanthocyanidins into anthocyanidins via acidic treatment and subsequent anthocyanin quantification with HPLC or via photometer [40,41,52]. However, the   Table 2. Functional activity, substrate specificity and influence of mutations in the N-terminus on the functional activity of DFR from Angelonia 6 angustifolia, varying positions between the two DFR types are marked in bold. conversion is not quantitative and strongly depends on various factors [45]. Measurements of the converted anthocyanidin are therefore an acceptable tool for demonstrating functional activity of DFRs. It remains questionable however whether this provides reliable results for the comparison of substrates with divergent hydroxylation patterns.

Heterologous expression as glutathione S-transferase (GST) fusion protein in E. coli
Heterologous expression of DFR is frequently performed in S. cerevisiae [50], however, this is time-consuming in comparison with expression in E. coli. Only a few recombinant DFRs were successfully expressed in E. coli thus far [28,33,40,41,53] with instability after His-tag removal reported. We therefore tested whether heterologous expression as a GST fusion protein in E. coli could be a viable option to rapidly produce highly purified, active recombinant DFR. The DFRs were subcloned from pYES2.1 into the pGEX-6P-1 vector and heterologously expressed in E. coli. Purification of the recombinant proteins was performed by using glutathione Sepharose TM 4B in either a batch procedure or by using GST Spin Traps. When the GST Spin Traps were used, a rapid loss of enzyme activity within two days was observed unless the enzyme was stabilized with BSA (2 mg/ml) and kept in Protein LoBind tubes. Both the resulting GST fusion protein and the purified enzyme after GST removal showed functional activity and no difference was observed in either the substrate acceptance of the fusion protein, the purified protein after GST removal or the recombinant enzyme produced in yeast. The purified recombinant DFR was stable and could be kept at 280uC for several months without loss of activity.

Restoring the functional activity of Ang.DFR1 by site-directed mutagenesis
Although the two cDNA clones differed only in three amino acids, no functional activity could be observed with recombinant Ang.DFR1. To identify the decisive residues, we mutated the two DFR cDNA clones by site-directed mutagenesis and investigated the resulting recombinant proteins for functional activity. Neither the deletion of the extra proline, along with an exchange of serine to proline in the Ang.DFR1 sequence nor an exchange of the arginine 26 with glycine alone was able to restore the functional activity of Ang.DFR1. Only simultaneous mutation of all three positions resulted in a highly active recombinant Ang.DFR1. In the same way, exchange of glycine to arginine in Ang.DFR2 or insertion and exchange of an additional proline with serine at position 12 led to a complete loss of Ang.DFR2 activity ( Table 2). The proline-rich area in the Angelonia DFR is closely located to the N-terminus and contains a unique structural feature ( Figure  S1), which is also not found in the Vitis DFR (Swiss-Prot Accession Number P93799_VITVI) and therefore not covered by the crystal structure [28]. It is thus difficult to explain why differences in the three amino acids are essential for functional activity. The region however, is closely located to a glycine-rich motif constituting the NADPH binding site [28]. As proline often acts as a structural disruptor [54] we theorize that the presence of two additional prolines and the large arginine residue in position 26 of  The substrates were converted with recombinant dahlia DFR (Accession FJ216425). Substrates and products were separated on cellulose in chloroform/acetic acid/water (10:9:1, v:v:v). n.d: not detected. *total amount of radioactivity was very low. doi:10.1371/journal.pone.0107755.t004 Ang.DFR1, which contrasts with the small glycine in Ang.DFR2, has an negative impact on protein folding or disturbs proper NADPH binding.

Conclusions
This study focused on the isolation and characterization of DFR from Angelonia x angustifolia which seems to play an essential role in the formation of chimeral patterns resulting in the formation of white/blue flowers. The DFR was used as model cDNA clone to demonstrate the suitability for heterologous expression as a GST fusion protein in E. coli. In addition, we demonstrated possible sources for biased results in substrate acceptance studies with DFRs.

Plant material
The studies were performed on pink/blue and white/blue flowering chimeras and white genotypes (arisen after somatic segregation of the white/blue chimera type) of Angelonia 6angustifolia plants (Figure 1) which were previously described [2]. The plants were grown in the greenhouse at the Julius Kühn-Institut in Quedlinburg (Germany). Flowers were harvested, shock-frozen in liquid nitrogen and stored at 280uC until use.

Enzyme preparation
Enzyme preparations were obtained using protocols 1 and 2 as described [31]. To remove low molecular compounds, crude enzyme preparations were passed through a gel chromatography column (Sephadex G25, GE Healthcare, Freiburg, Germany). Protein content was determined by a modified Lowry procedure [55] using crystalline bovine serum albumin as a standard.

Enzyme assays
Assays with enzyme preparations from Angelonia petals were performed according to [31] using naringenin as a substrate for FHT and DHQ as a substrate for DFR. CHS cannot be measured independently from CHI because the latter does not need any cofactor and rapidly converts naringenin chalcone formed by CHS into naringenin. Therefore, a combined assay was used as described previously [31]. For the standard enzyme assay with the recombinant DFR obtained from yeast, the reaction contained in a final volume of 50 mL: 15 mL enzyme preparation (22 mg), 0.048 nmol ( 14 C)-dihydroflavonol, 0.25 nmol NADPH, 30 mL 0.1 M KH 2 PO 4 /K 2 HPO 4 buffer pH 6.5 containing 0.4% Na ascorbate. The amount of enzyme was chosen within a range to ensure that the maximum conversion rate of the best substrate was around 50% (linear range of the reaction). The reaction mixture was incubated for 15 min at 25uC. Assays using DHK and DHQ as substrates were stopped by addition of 10 mL 100% acetic acid with products and substrates being extracted twice with 50 mL ethyl acetate. The combined organic phases were transferred to a pre-coated thin-layer cellulose plates without fluorescence indicator (Merck, Darmstadt, Germany) and developed in chloroform/ acetic acid/water (10:9:1, v:v:v). The evaluation was carried out on a Berthold LB 2842 TLC Linear Analyzer (Wilbad, Germany) by integration of the peak areas. Assays using DHM as substrates were stopped with the addition of 10 mL 100% acetic acid and 50 mL methanol. The mixture was chromatographed on 20 cm61 cm stripes of paper (Schleicher Schuell, 2043 b, Germany) in chloroform/acetic acid/water (10:9:1, v:v:v). Conversion rates were determined using a Typhoon 8600 and the software ImageQuant 5.1 (GE Healthcare, Munich, Germany).

Characterization of the recombinant DFR
All data represents an average of at least three independent experiments. Determination of the pH optimum was carried out as described for the standard DFR assay, but using 0.2 M McIlvaine buffers with pH values between 4.5 and 9.0.
Cloning of DFR cDNAs from Angelonia 6 angustifolia mRNA was isolated from Angelonia petals using the mMACS mRNA Isolation Kit (Miltenyi Biotec, Germany). cDNA was prepared using the SuperScript II Reverse Transcriptase (Invitrogen, Carlsbad, CA) and the oligo(-dT) anchor primer GAC-CACGCGTATCGATGTCGAC(T)16V. Based on information available in the NCBI-GenBank, nucleotide or amino acid sequences of DFRs from other plants from the order Lamiales, were aligned (Accessions X15536; EU305680, Y09127, EF522156) and conserved regions in the N-terminal region adopting the Rossmann fold [28] were used for the design of degenerated primers. The obtained cDNA fragments were isolated, ligated into the vector pCR2.1-TOPO (Invitrogen, Carlsbad, CA) and transformed in E. coli strain TOP10 (Invitrogen, Carlsbad, CA). Fragments were sequenced by a commercial supplier (Microsynth Austria AG, Vienna, Austria) and sequences were used for the design of specific 59-and 39primers for the amplification of the ends of the DFR by RACE techniques, using the SMART RACE cDNA amplification kit (Clontech, Takara Bio Europe, Saint-Germain-en-Laye, France) according to the manufacturer's instructions. Proofreading amplification of the complete open reading frame was carried out using specific forward and reverse primers (Table S1) and the Taq/Pwo Expand High Fidelity PCR System (Roche, Mannheim, Germany).

Phylogenetic analysis
Multiple alignments were performed with ClustalW [56,57]. The phylogenetic tree was conducted and bootstrapped with MEGA version 5 [58] using the neighbor-joining method and 1000 replicates.

Heterologous expression in Saccharomyces cerevisiae
Proofreading cDNA amplicons were ligated into the pYES2.1/ V5-His-TOPO vector (Invitrogen, Paisley, UK). Sense constructs were isolated and confirmed by sequencing. The vectors harbouring the DFR cDNAs Ang.DFR2 and Ang.DFR1 were transformed into yeast strain INVSc1 using the Sc. EasyComp Transformation Kit (Invitrogen, Carlsbad, CA). Preparation of the protein fractions was performed using a modified protocol according to Pompon et al. [59]. Protein fractions were shock frozen in liquid nitrogen and stored at 280uC.

Heterologous expression in E.coli
The plasmid pGEX-6P-1 (GE Healthcare, Munich, Germany) was used for the overexpression of the A. 6 angustifolia DFRs as GST fusion proteins. Plasmid pGEX-6P-1 was linearized by double digestion with BamHI and EcoRI (Fermentas, Germany). Sticky end DFR inserts were generated using sticky end PCR [60] with Pfu DNA polymerase (Fermentas, Germany) and the primers listed in Table S1. For each of the DFRs, two PCR reactions (PCR 1: A-DFR-FL, A-DFR-RS; PCR2: A-DFR-FS, A-DFR-RL) were performed with diluted pYES2.1 plasmids harbouring the Ang.DFR2 and Ang.DFR1 as templates. The PCR products were eluted and purified from the agarose gel after electrophoresis. Denaturation and reannealing of an equimolare mixture of the PCR products from the PCR 1 and 2 resulted theoretically in 25% double stranded dihydroflavonol 4-reductase sequences with sticky BamHI (GATC) and EcoRI (AATT) ends for direct ligation with T4 DNA ligase (Promega) into linearized pGEX-6P-1. After transformation into E. coli strain TOP 10 and isolation of the plasmids, integrity of the vector construct harbouring the dihydroflavonol 4-reductases (pGEX-6P-1::Ang.DFR2 and pGEX-6P-1::Ang.DFR1) was confirmed by sequencing. Single colonies harbouring parent DFRs and mutants were used for heterologous expression in E. coli as described [61].
Purification of GST fusion proteins and removal of GST tag by enzymatic cleavage GST fusion proteins were purified from the cell extract using Sepharose 4B (GE Healthcare, Munich, Germany) in either a batch procedure or by using GST Spin Traps according to the manufacturer's instructions The DFR was liberated from GST by PreScission protease according to the GST Gene Fusion System Manual (GE Healthcare, Munich, Germany).

Site-directed mutagenesis
Mutants were generated from pGEX-6P-1 vector constructs harbouring the DFRs by use of Q5 Site-Directed Mutagenesis Kit (NewEngland Bioloabs, Vienna Austria). Primer were designed using the NEBase Changer TM v 1.01 provided at http:// nebasechanger.neb.com. The sequences are given in table S1. The integrity of the constructs was confirmed by commercial sequencing (Microsynth Austria AG, Vienna Austria).