Functional expression and mutational analysis of flavonol synthase from Citrus unshiu Frank Wellmann 1, *, Richard Lukac ˇ in 1, *, Takaya Moriguchi 2 , Lothar Britsch 3 , Emile Schiltz 4 and Ulrich Matern 1 1 Institut fu ¨ r Pharmazeutische Biologie, Philipps-Universita ¨ t Marburg, Germany; 2 National Institute of Fruit Tree Science, Ibaraki, Japan; 3 Merck kgaA, Scientific Laboratory Products, Darmstadt, Germany; 4 Institut fu ¨ r Organische Chemie und Biochemie, Universita ¨ t Freiburg, Germany Flavonols are produced by the desaturation of flavanols catalyzed by flavonol synthase. The enzyme belongs to the class of intermolecular dioxygenases which depend on molecular oxygen and Fe II /2-oxoglutarate for activity, and have been in focus of structural studies recently. Flavonol synthase cDNAs were cloned from six plant species, but none of the enzymes had been studied in detail. Therefore, a cDNA from Citrus unshiu (Satsuma mandarin) designated as flavonol synthase was expressed in Escherichia coli,and the purified recombinant enzyme was subjected to kinetic and mutational chacterizations. The integrity of the recom- binant synthase was revealed by a molecular ion from MALDI-TOF mass spectrometry at m/z 37888 ± 40 (as compared to 37899 Da calculated for the translated poly- peptide), and by partial N-terminal sequencing. Maximal flavonol synthase activity was observed in the range of pH 5–6 with dihydroquercetin as substrate and a tempera- ture optimum at about 37 °C. K m values of 272, 11 and 36 l M were determined for dihydroquercetin, Fe II and 2-oxoglutarate, respectively, with a sixfold higher affinity to dihydrokaempferol (K m 45 l M ). Flavonol synthase polypeptides share an overall sequence similarity of 85% (47% identity), whereas only 30–60% similarity were apparent with other dioxygenases. Like the other dioxy- genases of this class, Citrus flavonol synthase cDNA encodes eight strictly conserved amino-acid residues which include two histidines (His221, His277) and one acidic amino acid (Asp223) residue for Fe II -coordination, an arginine (Arg287) proposed to bind 2-oxoglutarate, and four amino acids (Gly68, His75, Gly261, Pro207) with no obvious function- ality. Replacements of Gly68 and Gly261 by alanine reduced the catalytic activity by 95%, while the exchange of these Gly residues for proline completely abolished the enzyme activ- ity. Alternatively, the substitution of Pro207 by glycine hardly affected the activity. The data suggest that Gly68 and Gly261, at least, are required for proper folding of the flavonol synthase polypeptide. Keywords: Citrus unshiu (Rutaceae); flavonoid biosyn- thesis; flavonol synthase; functional expression; site-directed mutagenesis. Flavonoids fulfill vital functions in many plants beyond the scope of pigmentation and ultraviolet screening, e.g. in reproduction [1], in the defense against microbial pathogens and insects or in auxin transport [2], and are accumulated ubiquitously in flower and green tissues [1]. Their biosyn- thesis proceeds from 4-coumaroyl- and malonyl-CoAs to form naringenin chalcone [3] which is cyclized stereospeci- fically to the flavanone (2S)-naringenin [3]. Naringenin may be oxidized by flavone synthase (FNS) to yield the flavone apigenin [4–6] or hydroxylated by flavanone 3b-hydroxylase (FHT) to form a flavanol (syn. dihydroflavonol) [7–10], i.e. dihydrokaempferol, which might be reduced subsequently to a leucoanthocyanidin along the branch leading to catechins and anthocyanidins [3] (Fig. 1). Alternatively, flavonol synthase (FLS) catalyzes the oxidation of the flavanol to a flavonol (Fig. 1). FLS had been reported initially from irradiated parsley cells as a soluble dioxygen- ase requiring 2-oxoglutarate and Fe II /ascorbate for full activity [11]. The activity was subsequently detected in flower tissues of Matthiola incana [12], Petunia hybrida [13] or Dianthus caryophyllus [14]. The first FLS cDNA was cloned in 1993 from Petunia hybrida [15] and identified by functional expression in yeast, while the FLS-antisense transformation of petunia or tobacco intensified the red flower pigmentation [15]. Further FLS cDNAs were isolated later from Arabidopsis thaliana [16], Eustoma grandiflorum, Solanum tuberosum [17], Malus domestica and Matthiola incana, and approximately 85% similarity was determined for the translated polypeptides, mostly in the C-terminal 40% region based on total length of 335 residues. None of these enzymes has been satisfactorily expressed and characterized. Correspondence to U. Matern, Institut fu ¨ r Pharmazeutische Biologie, Philipps-Universita ¨ t Marburg, Deutschhausstrasse 17A, 35037 Marburg, Germany. Fax: + 49 6421 282 6678, Tel.: + 49 6421 282 2461, E-mail: matern@mailer.uni-marburg.de Abbreviations: ACC, aminocyclopropane-1-carboxylic acid; DAOCS, deacetoxcephalosporin C synthase; FHT, flavanone 3b-hydroxylase; FLS, flavonol synthase; FNS, flavone synthase; IPNS, isopenicillin N synthase. *Note: these authors contributed equally to the work presented. Note: flavonol synthase NCBI database accession numbers: Citrus unshiu, AB011796; Eustoma grandiflorum, AAF64168; Malus domes- tica, AAD26261; Matthiola incana, O04395. (Received 17 April 2002, revised 4 July 2002, accepted 11 July 2002) Eur. J. Biochem. 269, 4134–4142 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.03108.x Several intermolecular dioxygenases, particularly those of microbial or human origin, catalyze reactions of medicinal and industrial relevance, and their spatial organization and mode of action are under investigation. The reactions are diverse, such as the desaturation of aliphatic chains or the oxidative cyclization and the hydroxylation of substrates [18–22], and depend on the one-, two- or four-electron reduction of molecular oxygen. Most of these dioxygenases rely on the concomitant oxidation of 2-oxoglutarate. Deacetoxycephalosporin C synthase (DAOCS) serves as a precedent enzyme in comparison to the 2-oxoglutarate- independent isopenicillin N-synthase (IPNS), as the modu- lar composition and spatial configuration of IPNS and DAOCS appear to be rather similar [23] irrespective of only 19% primary structure identity. Both enzymes form a b-strand core folded into a distorted jelly roll motif [23,24] known also from viral capsid proteins and other enzymes [24,25]. During the preparation of this manuscript, an equivalent configuration was proposed for a putative anthocyanidin synthase from Arabidiopsis thaliana [26], although this enzyme has not been functionally proved and the lability of substrate and product rule out any cocrys- tallization. The setting of a helices and b-strands causes very similar CD spectra for this class of enzymes [27–29], which was confirmed recently also for Petunia FHT [9] sharing 30% sequence similarity with the DAOCS and IPNS polypeptides. The IPNS, DAOCS or FHT sequences also range in the order of 30% similarity to the translated FLS sequences, but the biochemical characterization of FLSs has remained very preliminary. In the course of our investigations on the related dioxygenases FHT [7–10] and flavone synthase I [4–6] we became interested in the molecular evolution of diversity concerning the enzymes of flavonoid biosynthesis. We report the expression of highly active Citrus unshiu FLS in Escherichia coli and the purification of the labile enzyme by a modified protocol devised for the isolation of Petunia hybrida FHT [8]. This enzyme served to deter- mine for the first time the kinetic parameters of an FLS. The relevance of three amino-acid residues which appear to be highly conserved in all plant intermolecular dioxygenases for FLS activity was examined by site-direc- ted mutagenesis. MATERIALS AND METHODS Expression vector The FLS cDNA from satsuma mandarin fruits, C. unshiu [30], was excised with EcoRI and XhoI from the Bluescript vector (Stratagene) and subcloned in pTZ19R [31]. The construct was used for the transformation of E. coli RZ1032 (Stratagene), ssDNA was isolated by the addition of phage M13K07 and used for site-directed mutagenesis by the site- elimination technique according to Zakour [32]. Hybridiza- tion of the mismatch primer 5¢-CTCCACCTCCATG GATTTTATTTTCC-3¢ to the FLS 5¢ coding region introduced a unique NcoI site at the start of translation, which was verified by DNA sequencing [33]. The DNA encoding FLS was subsequently isolated by digestion with NcoI and PstI and subcloned into the expression vector pQE6 (FLS-pQE6) as described previously [7,9]. Recombinant FLS E. coli strain M15 harboring the plasmid pRep4 was transformed with the FLS-pQE6 constructs containing the coding sequence of the wild-type or mutant enzymes and subcultured subsequently to a density of 0.8 in Luria– Bertani medium (typically 400 mL in 2 L flasks) containing ampicillin (100 lgÆmL )1 ) and kanamycin (25 lgÆmL )1 ). Isopropyl thio-b- D -galactoside (1 m M ) was added for the induction of FLS expression, and the bacteria were harvested after another 3 h at 37 °C and stored frozen at ) 70 °C as described previously [7–9]. Wild-type FLS was purified from the crude bacterial extracts by a modified procedure devised for Petunia FHT [8]. Briefly, the bacterial pellet (up to 6 g wet mass) was suspended in 20 mL of 50 m M potassium phosphate buffer pH 5.5, 10 m M EDTA, 5m M dithiothreitol and 15 m M MgCl 2 , and the suspensions were sonicated for 2 min and cleared by centrifugation (30 000 g,10min,5°C). Solid ammonium sulfate was Fig. 1. Reaction catalyzed by FLS. (2S)-Naringenin, formed from 4-coumaroyl- CoA and malonyl-CoAs by the action of chalcone synthase and chalcone isomerase, is 3b-hydroxylated by flavanone 3b-hydroxylase (FHT) to furnish the substrate (2R,3R)- dihydrokaempferol. Alternatively, the B-ring hydroxylation of naringenin to (2S)- eriodictyol preceeding the 3b-hydroxylation yields the substrate (2R,3R)-dihydroquercetin. Both dihydrokaempferol and dihydroquerce- tin are accepted by the FLS to produce kaempferol and quercetin, respectively. The flavanones naringenin and eriodictyol might also be oxidized by FNS to the flavones apigenin or luteolin. Ó FEBS 2002 Flavonol synthase (Eur. J. Biochem. 269) 4135 added to the clear supernatant, and the protein precipitating at 40–50% saturation was redissolved in 50 m M potassium phosphate buffer pH 5.5, containing 5 m M dithiothreitol (1 mL) for successive size exclusion and anion exchange chromatographies on Fractogel EMD BioSEC (S) (Merck, Darmstadt, Germany) and Fractogel EMD DEAE 650 (S) (Merck) as described previously [8]. The purification of wild-type FLS was monitored by enzyme assays and SDS/ PAGE. Site-directed mutagenesis Site-directed mutagenesis was accomplished by site-elimin- ation using the oligonucleotide- directed in vitro mutagenesis technique [32]. Oligonucleotides were synthesized (G. Igloi, Institut fu ¨ r Biologie III, Universita ¨ t Freiburg, Germany) as required for the substitution of glycine by alanine and proline, respectively, or of proline by glycine (Table 1). Each individual mutation was verified by DNA sequencing using the dideoxynucleotide chain-termination method [33] with the universal and reverse sequencing primers. Following the confirmation of successful mutation the mutated genes were ligated into the expression vector pQE6 in the same way as described for the wild-type cDNA. Data base retrieval Data base searches and sequence alignments were carried out with the ENTREZ and BLAST software (National Library of Medicine and National Institute of Health, Bethesda, MD, USA). The PROSIS package (Hitachi, San Bruno, CA, USA) was used for the analysis of multiple alignments and consensus sequences with minor adjustment the computer alignments. Circular dichroism spectroscopy Circular dichroism measurements of homogeneous FLS were performed on a Jasco-720 spectropolarimeter (Tokyo, Japan) interfaced to an 486/33 PC and controlled by Jasco software. The spectropolarimeter was equipped with a cylindrical quartz cuvette with a pathlength of 0.05 cm. The temperature of the cell holder was maintained at 5 °Cbya circulating water thermostat and the instrument was calibrated with 0.06% ammonium D –10-camphor sulfonate. FLS spectra were recorded in potassium phosphate buffer, pH 6.8, as described previously for FHT [9], and the protein concentration was adjusted to 0.371 mgÆmL )1 in the sample. The documented spectra show the accumulation of 10 scans with 50 nmÆmin )1 . The CD spectra of the FLS sample were analyzed for the secondary structure content by the self- consistent method [34] included in the program package DICHROPROT V2.4 by Delage and Geourjon [35]. Protein analysis and immunoassay Partial N-terminal sequencing was carried out by Edman degradation in a pulsed liquid sequencer (Model 477 A, Applied Biosystems Inc.) with a Model 120 A PTH-analyzer for on-line identification, following the supplier’s guidelines. Mass spectra were recorded on a Bruker Reflex II MALDI- TOF mass spectrometer in the linear mode. The protein solution (100 lg per 300 lL20m M potassium phosphate buffer) was diluted with an equal volume of 0.1% trifluo- roacetic acid/acetonitrile (1 : 1, v/v), and this acidified solution was then mixed with an equal volume of a saturated solution of sinapic acid in 0.1% trifluoroacetic acid/aceto- nitrile (1 : 1, v/v) and applied on the SCOUT MTP TM MALDI-TOF target plate target in 0.5 lL portions [8]. SDS/PAGE separation of protein extracts was performed on 5% stacking and 12.5% separation gels in a Mini- Protean II cell (Bio-Rad, Mu ¨ nchen) according to Laemmli [36]. Antibodies were raised in rabbit by repeated injection of the recombinant homogeneous FLS (1 mg total), and the antiserum was diluted 10,000-fold for Western blotting [37]. Enzyme assays FLS activity of filtered (PD10 column, Pharmacia, Frei- burg) plant or bacterial extracts was routinely assayed at 37 °C and pH 5.0. The assay mixture (total volume 360 lL) contained 100 l M dihydroquercetin as a substrate, 83 l M 2-oxoglutarate, 42 l M ammonium iron(II) sulfate, 2.5 m M sodium ascorbate, and 2 mgÆmL )1 bovine catalase, and the incubation was carried out in open vials under gentle shaking. The reaction linearity was assessed by proper choice of protein amounts (4.5–22 lg) and incubation periods (0.5–10 min). The reaction was stopped by the addition of 15 lL saturated aqueous EDTA solution. The flavonoids were isolated by repeated extraction with Table 1. Oligonucleotides for site-directed mutagenesis. The Citrus FLS coding sequence flanking the desired site of mutation (top) is aligned with the complementary oligonucleotide used to create the mutation (bottom). The triplets encoding glycine and proline are bold-printed, and the base changes are underlined. Mutant FLS Oligonucleotide Codon change Gly68Ala 5¢-CGGGAGTGGGGGATTTTCCAG-3¢ GGGfiGCG 3¢-GCCCTCACCCGCTAAAAGGTC-5¢ Gly68Pro 5¢-CGGGAGTGGGGGATTTTCCAG-3¢ GGGfiCCG 3¢-GCCCTCACCGGCTAAAAGGTC-5¢ Gly261Ala 5¢-CATCCACATCGGGGACCAGATC-3¢ GGGfiGCG 3¢-GTAGGTGTAGCGCCTGGTCTAG-5¢ Gly261Pro 5¢-CATCCACATCGGGGACCAGATC-3¢ GGGfiCCG 3¢-GTAGGTGTAGGGCCTGGTCTAG-5¢ Pro207Gly 5¢-GATTAATTATTATCCGCCATGCCC-3¢ CCGfiGGG 3¢-CTAATTAATAATACCCGGTACGGG-5¢ 4136 F. Wellmann et al. (Eur. J. Biochem. 269) Ó FEBS 2002 ethylacetate (twice, 75 lL) and reversed-phase HPLC (Shimadzu, Tokyo, Japan) on a Nucleosil C18-column (125 · 4 mm, 5 lm; Machery and Nagel, Du ¨ ren, Germany). The column was equilibrated with solvent A (20% aqueous methanol), and quercetin or kaempferol and dihydroqu- ercetin or dihydrokaempferol were eluted in a linear gradient of solvent A and solvent B (100% methanol) at 0.5 mLÆmin )1 over 3 min, followed by solvent B for 5 min [38]. The elution was monitored by the absorption profile at 290 nm (dihydroquercetin, dihydrokaempferol) or 368 nm (quercetin, kaempferol), and authentic flavonoid samples were employed as references for calibration. The reaction catalyzed by FLS follows second order kinetics, and the apparent Michaelis constants for the wild-type enzyme were determined with 11 lg of the homogeneous enzyme [7,9]. Protein concentrations were determined according to Lowry [39] following the precipitation with trichloroacetic acid in the presence of deoxycholate [40] and using bovine serum albumin as a standard. Mass spectrometry The FLS assay, routinely carried out in Tris/HCl buffer pH 7.5 prior to the final assessment of pH optima, was scaled up for preparative purposes to a volume of 40 mL (20 incubations of 2 mL each). The assay contained 6 mg HPLC-purified dihydroquercetin total, 170 l M sodium ascorbate, 35 l M ammonium iron(II) sulfate, 70 l M 2-oxoglutarate, 2 mgÆmL )1 bovine catalase and 1.1 mg (55 lg per 2 mL incubation) of the homogeneous FLS. The mixture was incubated for two hours at 37 °C on a rotary shaker (300 r.p.m.), and the flavonoids were extracted subsequently with ethylacetate (twice 500 lLper2mL incubation) and isolated by successive cellulose thin-layer chromatography in 15% aqueous acetic acid (solvent system I) and trichloromethane/acetic acid/water 10 : 9 : 1 (v/v/v) (solvent system II). The developed cellulose plates were dried for 2 h in a cold air stream, the substrate (dihydroquercetin) and product (quercetin) were spotted by absorbance at 366 nm, and the product was extracted with methanol. The solution was filtered and concentrated for EI-MS and MALDI-TOF-MS analyses. The EI-MS were recorded on a Finnigan MAT 70S mass spectrometer by direct inlet and an accelerating voltage of 6 kV at injection temperatures of 130 °C, 250 °Cor280 °C. Product identification was also accomplished on a Bruker Reflex MALDI-TOF-MS in the positive ion reflectron mode using an accelerating voltage of 23 kV. The mass spectra were analyzed over a range of m/z 50–750, and the [M + H] + ions of a-cyano-4-hydroxycinnamic acid (a-HCCA) and 2,5-dihydroxybenzoic acid (DHB) were employed for the internal calibration across the mass range. RESULTS Heterologous expression of Citrus FLS The unequivocal assignment of genes requires the functional characterization of the corresponding polypeptides, which has been occasionally neglected in the process of recent gene bank accessions. FLS cDNAs were accessed from several plant species (Petunia hybrida, Arabidopsis thaliana, Solanum tuberosum, Eustoma grandiflorum, Malus domestica and Matthiola incana), but functionality was only verified in case of Petunia and Arabidopsis [15,16]. In the course of our research on Rutaceae [41–43], we became interested in a clone from C. unshiu recently designated as FLS [30]. Accordingly, this cDNA containing an ORF of 1005 bp was ligated into the pQE6 vector (FLS-pQE6 construct), expressed in E. coli, and the product was purified under conditions that had been successfully employed for the isolation of another labile 2-oxoglutarate-dependent diox- ygenase from Petunia [8]. The cDNA from Citrus encoded a 335-residue polypeptide, which was extracted from the induced, recombinant bacteria with potassium phosphate buffer at pH 5.5, and the crude extract was fractionated by successive size exclusion and ion exchange chromatogra- phies at pH 7.5. This rapid protocol was very efficient and yielded elution profiles for the recombinant Citrus enzyme resembling those of the Petunia FHT [8]. The apparently homogeneous FLS eluted from the anion exchanger was used for spectroscopy, the generation of antibodies and activity assays. Polypeptide analysis The homogeneous Citrus enzyme revealed only one band of about 38 kDa on SDS/PAGE separation (Fig. 2), which correlated to the molecular mass of 37 899 Da calculated for the translated polypeptide. Furthermore, partial N-ter- minal sequencing of the polypeptide yielded a sequence, Met-Glu-Val-Glu-Arg-Val-Gln-Ala-Ile-Ala-Ser-Leu-Ser-His, identical to the N-terminal 14 amino acids translated from the FLS-pQE6 construct. Moreover, the pure polypep- tide was subjected to MALDI-TOF-MS which revealed a molecular ion at 37888 ± 40 fully matching the mass Fig. 2. SDS/PAGE separation of recombinant Citrus FLS. Crude extracts in phosphate buffer at pH 5.5 of E. coli expressing the FLS (lane 1) were subjected to 40–50% ammonium sulfate fractionation followed by size-exclusion (lane 2) and anion-exchange (lane 3) chro- matographies. The proteins were separated in 5% stacking and 12.5% separation gels and stained with Coomassie Brilliant Blue R250. Commercial molecular mass markers (lane M) served for calibration. Ó FEBS 2002 Flavonol synthase (Eur. J. Biochem. 269) 4137 calculated for the translation product. These data corro- borated the integrity of the recombinantly expressed polypeptide, an essential prerequisite for further structural investigations. A polyclonal antiserum to the pure recom- binant polypeptide was raised in rabbit, which recognized one protein band of 38 kDa in Western blots of crude enzyme extracts. The homogeneous Citrus FLS was subjected to CD spectroscopy in order to substantiate its structural relationship with mechanistically related enzymes. The CD profile revealed a characteristic double minimum at 222 nm and 208–210 nm and a maximum at 191–193 nm, which indicated the presence of extended a helical regions interrupted by b sheet strands. Very similar profiles had been recorded for Streptomyces IPNS [28,29] and Petunia FHT [9]. Catalytic activity The enzymatic activity of the recombinant protein was examined in FLS incubations employing dihydroquercetin or dihydrokaempferol as a substrate. Both these flavanols were accepted, and the reaction products were identified as the flavonols quercetin and kaempferol, respectively, by their mobility on RP-HPLC and cellulose-TLC in compar- ison to authentic reference samples. Thoroughly purified dihydroquercetin was additionally employed for preparative incubations, and the product was identified as quercetin by EI-MS and MALDI-TOF-MS showing the molecular ion at m/z 302. The conversion of flavanols to flavonols depended on the presence of ferrous iron and 2-oxogluta- rate, thus establishing that the clone from Citrus unshiu encoded the 2-oxoglutarate-dependent dioxygenase FLS. Dihydroquercetin is commercially available and seems to be the predominant substrate for Citrus unshiu flavonol biosynthesis [30]. Therefore, this substrate was employed in order to define the optimal assay conditions. At aerobic and saturating conditions for ferrous iron and 2-oxoglutarate, the rate of conversion depended exclusively on the substrate concentration. Maximal activity was observed at pH 5.0 (Fig. 3) and over a temperature range from 35 to 40 °C. Under these conditions, the apparent K m values were determined at 272, 11 and 36 l M for dihydroquercetin, Fe II and 2-oxoglutarate, respectively. Reexamination of the conversion rate of dihydrokaempferol under same condi- tions, however, revealed an apparent K m at 45 l M and, thus, dihydrokaempferol as the preferred Citrus FLS substrate in vitro. Nevertheless, rutin (quercetin 3-O-rutinoside) was identified as the major flavonol in satsuma mandarin plants [30]. Sequence analysis and mutagenesis The alignment of the polypeptide sequences of 2-oxogluta- rate-dependent dioxygenases and related enzymes retrieved from data banks (59 sequences total) revealed only 8 strictly conserved amino-acid residues which cluster in three regions of high overall similarity (Fig. 4). Three of these residues (His221, His277 and Asp223; the numbering refers to the Citrus FLS sequence) are essential for the coordination of ferrous iron as had been demonstrated with Petunia FHT by kinetic and mutational studies [7] as well as with Aspergillus IPNS by X-ray diffraction of the Fe II -IPNS complex [24,25]. A further residue (Arg287) is involved in 2-oxoglutarate binding as had been proved with Petunia FHT [7,9] and by X-ray diffraction of the Streptomyces clavuligerus DAOCS complexed with Fe II and oxoglutarate [23]. However, no particular function was assigned to the additional four conserved residues (Gly68, His75, Pro207, Gly261). This is compatible with the observation that the mutation of His78 in P. hybrida FHT, corresponding to His75 in FLS, only had a marginal effect on the enzyme activity [7]. On the assumption that Gly68, Pro207 or Gly261 might be required for structural integrity of the active FLS, point mutations were initiated aiming at the substitution of glycine residues by alanine or proline and the exchange of proline by glycine. Glycine residues are frequently found at the C-terminal ends of a helices providing the conforma- tional flexibility required at these sites of the polypeptide backbone. Accordingly, at least the substitution of glycine residues by proline was expected to interfere with the folding process of the FLS polypeptide. Conversely, the Pro207fi Gly mutation should increase the degree of structural freedom. Five mutants (Gly68fiAla, Gly68fiPro, Pro207fiGly, Gly261fiAla, Gly261fiPro) were generated, cloned into the expression vector pQE6 and expressed in E. coli strain M15prep4 as described for the wild-type FLS cDNA. Examination of crude cell-free extracts or of the solubi- lized pellet of E. coli expressing wild-type and the mutant Gly68fiAla, Pro207fiGly or Gly261fiAla FLSs by Fig. 3. Relative activity of recombinant Ci t rus FLS depending on the pH of the assay. The enzyme assays were performed in 200 m M buffers composed of glycine/HCl pH 2.0–3.5, sodium acetate pH 4.5–5.5, potassium phos- phate pH 5.0–7.5, Bis-Tris/HCl pH 6.5–7.0, Tris/HCl pH 7.0–8.5 or sodium glycinate pH 8.5–10.0. 4138 F. Wellmann et al. (Eur. J. Biochem. 269) Ó FEBS 2002 SDS/PAGE and Western blotting revealed no differences in the mobilities of the wild-type and mutant FLS polypeptides (Fig. 5). Replacement of either glycine resi- due by alanine reduced the enzyme activity below 10% of control, while the substitution in Pro207fiGly did not affect the FLS activity to a significant extent (Table 2). Extraction of the mutants Gly68fiPro and Gly261fiPro, however, failed to yield immunoreactive FLS polypeptide in the soluble supernatant. Considerable amounts of the immunopositive polypeptide were recovered from the solubilized bacterial pellets, which showed no change in relative mobility on SDS/PAGE separation (Fig. 5). Nevertheless, this fraction had completely lost the FLS activity, presumably as the result of major structural changes. DISCUSSION Plants of the Rutaceae family are a rich source of flavonol glycosides such as the abundant rutin (quercetin rutinoside) which had been described initially from Ruta species. Flavonols originate from flavanones, i.e. (2S)-naringenin, by the consecutive action of FHT and FLS (Fig. 1), and both of these enzymes use molecular oxygen for catalysis and are referred to as 2-oxoglutarate-dependent dioxygen- ases [18–21]. These types of dioxygenases are encoded by genes of moderate to high sequence identity (from 19% to 75%), which might catalyse very diverse reactions irrespect- ive of their relative degree of sequence conservation. Nevertheless, many of these enzymes expressed so far adopt a highly homologous jelly roll topology [44]. Therefore, Fig. 4. Alignment of the FLS polypeptide from C. unshiu (FLS-Cit) with the FLS consensus sequence derived from the FLS polypeptide sequences of P. hybrida, E. gr andiflorum, M. d omestica, S. tube rosum and A. tha liana. The consensus sequence is composed of the identical (marked by asterisks) or the most abundant amino acids with conservative exchanges (marked by dots). Residues of equivalent hydropathy (STPAG or NDEQ) were replaced by an x, and gaps were introduced for maximal alignment. Three regions of high similarity were inferred from 59 data base accessions (National Library of Medicine, NIH or EMBL library) of 2-oxoglutarate-dependent and related enzymes, which include five FLSs as above, 18 FHTs (Petunia hybrida, Zea mays, Hordeum vulgare, Malus sp., Matthiola incana, Vitis vinifera, Citrus sinensis, Daucus carota, Dianthus caryo- phyllus, Callistephus chinensis, Chrysanthemum morifolium, Anthirrhinum majus, Bromhaedia finlaysonia, Arabidopsis thaliana, Persea americana, Ipomea purpurea, Ipomea nil and Medicago sativa), three anthocyanidin synthases (Zea mays, Anthirrhinum majus and Oryza sativa), five gibberellin C20 oxidases (Arabidopsis thaliana, Cucurbita maxima, Pisum sativum, Spinacia oleracea and Marah macrocarpa), hyoscyamine 6b-hydroxylase from Hyoscyamus niger, the iron-deficiency-specific proteins 2 and 3 from Hordeum vulgare; desacetoxyvindoline-4-hydroxylase from Catharanthus roseus, 11 aminocyclopropane-1-carboxylic-acid oxidases (Actinidia deliciosa, Arabidopsis thaliana, Brassica juncea, Dianthus caryophyllus, Lyco- persicum esculentum I and II, Malus domestica, Nicotiana tabacum, Persea americana, Petunia hybrida and Pisum sativum), 11 isopenicillin N-synthases (Acremonium chrysogenum, Flavobacterium sp., Lysobacter lactamgenus, Nocardia lactamdurans, Streptomyces cattleya, Streptomyces clavuligerus, Streptomyces jumonjinensis, Streptomyces lipmanii, Aspergillus nidulans, Cephalosporium acremonium and Penicillium chrysogenum), deacetoxycephalosporin C synthase and deacetylcephalosporin C synthase from Streptomyces clavuligerus. These regions are underlined, and four amino acids shown to ligate the ACV substrate in IPNS [25] and Fe II in IPNS [24] and FHT [7] as well as 2-oxoglutarate in FHT [7] are highlighted in red and green. The additional four conserved amino acids (Gly68, His75, P207, Gly261) of unknown function [7] are shaded. Ó FEBS 2002 Flavonol synthase (Eur. J. Biochem. 269) 4139 sequence alignments per se support only a preliminary functional assignment. In the course of our studies on Ruta graveolens secondary metabolism [41–43], the report of a cDNA from C. unshiu [30] assigned to FLS appeared relevant and led us to express and characterize this enzyme for comparison with dioxygenases from other sources [18–21]. Plant 2-oxoglutarate-dependent dioxygenases, unfortu- nately, are commonly rather labile enzymes which might be digested partially after heterologous expression in E. coli [8], and this hampers their functional characterization. Based on our previous experience [7–10], the full size Citrus FLS was expressed and rapidly purified to a homogeneous polypeptide of 38 kDa (Fig. 2) corresponding to 335 amino-acid residues. The identity of the recombinant enzyme was verified by FLS assays employing dihydroqu- ercetin or dihydrokaempferol as a substrate (Table 2), and antibodies raised to the pure polypeptide detected exclu- sively the FLS polypeptide in crude extracts of recombinant E. coli (Fig. 5). FLS cDNAs had been reported before from P. hybrida [13], A. thaliana [16], E. grandiflorum, S. tubero- sum [17], M. domestica and M. incana, and the translated polypeptides share about 85% sequence similarity with the Citrus FLS with 158 identical residues (47%). The differ- ences in the FLS sequences were mostly confined to the N-terminal portion (approx. 38% identity), while 62% identity was observed for the C-terminus (amino-acid residues 200–335). Surprisingly, the kinetic data revealed a much higher affinity of the recombinant enzyme to dihydro- kaempferol as compared to dihydroquercetin, although the satsuma mandarin accumulates mainly the quercetin 3-O-rutinoside (rutin) [30]. This discrepancy might suggest the expression of more than one FLS in C. unshiu. Polypeptide alignments of FLSs with plant 2-oxogluta- rate-dependent dioxygenases of different functionality (FHT, anthocyanidin synthase, gibberellin C20-oxidase, hyoscyamine 6b-hydroxylase, barley Ids2, Ids3 and Cath- aranthus desacetoxyvindoline 4-hydroxylase) as well as with related 2-oxoglutarate-independent enzymes (ACC oxidase and microbial DAOCS and IPNS) revealed eight strictly conserved amino acids in three regions (Fig. 4). These enzymes probably evolved from a common ancestral gene, and the essence of most of the conserved amino acids has been further substantiated by site-directed mutagenesis of FHT [7,9] and by documentation of ligand binding in crystalline DAOCS and IPNS complexes [23–25]. The coordination of Fe II is commonly mediated by two histidine and one aspartate residues (correspondingly His221, His277 Fig. 5. Western blotting of Citr us wild-type and mutant FLSs in crude extracts of recombinant E. co li. The extracts were filtered through PD10 columns, the proteins (15 lg per lane) were separated subse- quently by SDS/PAGE on 5% stacking and 12.5% separation gels and transferred to polyvinylidene difluoride membranes for immuno- staining [7]. Soluble extracts of the bacteria expressing the wild-type FLS (lane 1) or the mutant enzyme Pro207fiGly (lane 3), Gly261fiAla (lane 5), Gly68fiAla (lane 7), Gly68fiPro (lane 9) and Gly261fiPro (lane 11), respectively, as well as the solubilized pellet fractions of these wild-type (lane 2) and mutant bacteria (lanes 4, 6, 8, 10 and 12) were subjected to Western blotting with reference to mo- lecular mass markers (indicated in the left margin). The Western blots were developed with goat anti-rabbit IgG conjugated to alkaline phosphatase and 5-bromo-4-chloro-3-indolyl phosphate as described elsewhere [7,9]. The relative FLS contents of the supernatant and pellet of wild-type and Pro207fiGly, Gly261fiAla, Gly68Ala mutant ex- tracts (lanes 1–8) as well as the Gly68fiPro mutant membrane extract (lane 10) were comparable, while the amount in the solubilized pellet of the Gly261fiPro mutant (lane 12) was negligible and the band could be hardly recognized in the soluble Gly68fiPro and Gly261fiPro mutant extracts (lanes 9 and 11). Table 2. Specific activities of the wild-type and the Gly68fiAla, Gly68fiPro, Gly261fiAla, Gly261fiPro and Pro207fiGly mutant flavonol synthases. Soluble extracts of E. coli expressing wild-type or mutant FLS were filtered through PD10 columns, and the specific activities were examined under standard assay conditions (360 lL total) using dihydroquercetin or dihydrokaempferol as a substrate. The wild-type activity reached 0.5 mkatÆkg )1 on average with either substrate, and the level of expression was equivalent for the wild-type and mutant FLSs except for the Gly68fiPro and Gly261fiPro mutants as determined by Western blotting (Fig. 5). FLS Relative specific FLS activities with Dihydroquercetin (%) Dihydrokaempferol (%) Amount of protein in the standard assay (lg) Wild-type 100 100 288 Gly68Ala 3.8 6 396 Gly68Pro 0 a 0 a 360 Gly261Ala 7.5 10 252 Gly261Pro 0 a 0 a 288 Pro207Gly 83.7 70 360 a The solubilized bacterial pellet of these mutants also lacked FLS activity up to 1 mgÆmL )1 protein. 4140 F. Wellmann et al. (Eur. J. Biochem. 269) Ó FEBS 2002 and Asp223 in Citrus FLS), whereas only in clavaminate synthase the aspartate is replaced by a glutamate residue [44], and an arginine residue (Arg287 in Citrus FLS) can be ascribed to 2-oxoglutarate-binding [7,23]. This arginine is also conserved in IPNS and ACC oxidase with slightly different functionalities, binding the substrate carboxylate, d-( L -a-aminoadipoyl)- L -cysteinyl- D -valine, in IPNS [25] and presumably ascorbate in ACC oxidase. This left two glycine and one proline residues unaccounted for (Gly68, Pro207 and Gly261 in Citrus FLS), but these amino acids are presumed to be required for proper folding of the enzyme polypeptide. The data obtained by site-directed mutagenesis supported this assumption, as the substitution of either glycine residue (Gly68fiAla or Gly261fiAla) reduced the enzyme activity to only about 5% and the Gly68fiPro or Gly261fiPro substitution completely abolished the activity. It is conceivable that these mutations greatly affected the tertiary structure of the FLS, because upon expression in E. coli the polypeptides accumulated in inclusion bodies. The CD spectroscopy of the wild-type FLS revealed an overall composition of helices and b sheets very similar to that recorded for FHT [9] or IPNS [28,29]. Unfortunately, considerable losses of activity occurred on purification of the enzyme mutants Gly68fiAla or Gly261fiAla, and the yields were too low for reliable CD spectroscopy. Further comparison of the Gly68fiPro and Gly261fiPro mutants was not reasonable, because these FLSs had to be partially renatured from the membraneous bacterial pellet. Albeit not absolutely required for activity, the data assign a role to Gly68 and Gly261 in the FLS functionality. ACKNOWLEDGEMENTS The work was supported by the Deutsche Forschungsgemeins- chaft and Fonds der Chemischen Industrie. We are grateful to Drs R. Zimmermann and H. Mu ¨ ller (Merck KGaA, Darmstadt) for EI-MS and MALDI-TOF-MS measurements, to Dr U. Pieper (Institut fu ¨ r Biochemie, Universita ¨ t Giessen) for CD spectroscopy, and to Prof E. Wellmann (Institut fu ¨ r Biologie II, Universita ¨ t Freiburg) for HPLC assays. We would also like to thank Michaela Mu ¨ ller and Olga Lezrich for their excellent technical assistance. REFERENCES 1. Deboo, G.B., Albertsen, M.C. & Taylor, L.T. (1995) Flavanone 3-hydroxylase transcripts and flavonol accumulation are tempo- rally coordinate in maize anthers. Plant J. 7, 703–713. 2. Koes, R.E., Quattrocchio, F. & Mol, J.N.M. (1994) The flavonoid biosynthetic pathway in plants: function and evolution. Bioessays 16, 123–132. 3. Heller, W. & Forkmann, G. (1994) Biosynthesis of the flavonoids. In The Flavonoids, Advances in Research Since 1986 (Harborne, J.B., ed.), pp. 499–536. Chapman & Hall, London. 4. Lukac ˇ in,R.,Matern,U.,Junghanns,K.T.,Heskamp,M L., Britsch, L., Forkmann, G. & Martens, S. (2001) Purification and antigenicity of flavone synthase I from irradiated parsley cells. Arch. Biochem. Biophys. 393, 177–183. 5. Martens, S., Forkmann, G., Matern, U. & Lukac ˇ in, R. (2001) Cloning of parsley flavone synthase I. Phytochemistry 58, 43–46. 6. Martens, S. & Forkmann, G. (1999) Cloning and expression of flavone synthase II from Gerbera hybrids. Plant J. 20, 611–618. 7. Lukac ˇ in, R. & Britsch, L. (1997) Identification of strictly con- served histidine and arginine residues as part of the active site in Petunia hybrida flavanone 3b-hydroxylase. Eur. J. Biochem. 249, 748–757. 8. Lukac ˇ in, R., Gro ¨ ning,I.,Schiltz,E.,Britsch,L.&Matern,U. (2000) Purification of recombinant flavanone 3b-hydroxylase from Petunia hybrida and assignment of the primary site of proteolytic degradation. Arch. Biochem. Biophys. 375, 364–370. 9. Lukac ˇ in, R., Gro ¨ ning,I.,Pieper,U.&Matern,U.(2000)Site- directed mutagenesis of the active site serine 290 in flavanone 3b-hydroxylase from Petunia hybrida. Eur. J. Biochem. 267, 853– 860. 10. Lukac ˇ in, R., Urbanke, C., Gro ¨ ning, I. & Matern, U. (2000) The monomeric polypeptide comprises the functional flavanone 3b-hydroxylase from Petunia hybrida. FEBS-Letters 467, 353–358. 11. Britsch, L., Heller, W. & Grisebach, H. (1981) Conversion of flavanone to flavone, dihydroflavonol and flavonol with an enzyme system from cell cultures of parsley. Z. Naturforsch. 36c, 742–750. 12. Spribille, R. & Forkmann, G. (1984) Conversion of dihydro- flavonols to flavonols with enzyme extracts from flower buds of Matthiola incana. Z. Naturforsch. 39c, 714–719. 13.Forkmann,G.,deVlaming,P.,Spribille,R.,Wiering,H.& Schramm, A.W. (1986) Genetic and biochemical studies on the conversion of dihydroflavonols to flavonols in flowers of Petunia hybrida. Z. Naturforsch. 41c, 179–186. 14. Forkmann, G. (1991) Flavonoids as flower pigments: the forma- tion of the natural spectrum and ist extension by genetic engineering. Plant Breed 106, 1–26. 15. Holton,T.A.,Brugliera,F.&Tanaka,Y.(1993)Cloningand expression of flavonol synthase from Petunia hybrida. Plant. J. 4, 1003–1010. 16. Wisman, E., Hartmann, U., Sagasser, M., Baumann, E., Palme, K.,Hahlbrock,K.,Saedler,H.&Weisshaar,B.(1998)Knock-out mutants from an En-1 mutagenized Arabidopsis thaliana popula- tion generate phenylpropanoid biosynthesis phenotypes. Proc. Natl Acad. Sci. USA 95, 12432–12437. 17. Van Eldik, G.J., Reijnen, W.H., Ruiter, R.K., Van Herpen, M.M., Scrauwen, J.A. & Wullems, G.J. (1997) Regulation of flavonol biosynthesis during anther and pistil development, and during pollen tube growth in Solanum tuberosum. Plant J. 11, 105–113. 18. Prescott, A.G. (1993) A dilemma of dioxygenases (or where biochemistry and molecular biochemistry fail to meet). J. Exp. Bot. 44, 849–861. 19. DeCarolis, E. & DeLuca, V. (1994) 2-oxoglutarate-dependent dioxygenase and related enzymes: biochemical characterization. Phytochem. 36, 1093–1107. 20. Prescott, A.G. & John, P. (1996) Dioxygenases: Molecular Structure and role in plant metabolism. Annu. Rev. Plant Physiol. Plant Mol. Biol. 47, 245–271. 21. Prescott, A.G. & Lloyd, M.D. (2000) The iron (II) and 2-oxoacid- dependent dioxygenases and their role in metabolism. Nat. Prod. Report 17, 367–383. 22. Britsch, L., Dedio, J., Saedler, H. & Forkmann, G. (1993) Mole- cular characterization of flavanone 3 b-hydroxylases. Consensus sequence, comparison with related enzymes and the role of con- served histidine residues. Eur. J. Biochem. 217, 745–754. 23. Valega ˚ rd, K., Van Schelting, T., Lloyd, M.D., Hara, T., Ramaswamy,S.,Perrakis,A.,Thompson,A.,Lee,H J.,Baldwin, J.E., Schofield, C.J., Hajdu, J. & Andersson, I. (1998) Structure of a cephalosporin synthase. Nature 394, 805–809. 24. Roach, P.L., Clifton, I.J., Fu ¨ lo ¨ p,V.,Harlos,K.,Barton,G.J., Hajdu, J., Andersson, I., Schofield, C.J. & Baldwin, J.E. (1995) Crystal structure of isopenicillin N synthase is the first from a new structural family of enzymes. Nature 375, 700–704. 25. Roach, P.L., Clifton, I.J., Hensgens, C.M.H., Shibata, N., Schofield, C.J., Hajdu, J. & Baldwin, J.E. (1997) Structure of isopenicillin N synthase complexed with substrate and the mechanism of penicillin formation. Nature 387, 827–830. 26. Turnbull, J.J., Prescott, A.G., Schofield, C.J. & Wilmouth, R.C. (2001) Purification, crystallization and preliminary X-ray Ó FEBS 2002 Flavonol synthase (Eur. J. Biochem. 269) 4141 diffraction of anthocyanidin synthase from Arabidopsis thaliana. Acta Crystallogr. D 57, 425–427. 27. Zhang, Z., Barlow, J.N., Baldwin, J.E. & Schofield, C.J. (1997) Metal-catalyzed oxidation and mutagenesis studies on the iron (II) binding site of 1-aminocyclopropane-1-carboxylate oxidase. Biochemistry 36, 15999–16007. 28. Borovok, I., Landman, O., Kreisberg-Zakarin, R., Aharonowitz, Y. & Cohen, G. (1996) Ferrous active site of isopenicillin N syn- thase: genetic and sequence analysis of the endogenous ligands. Biochemistry 35, 1981–1987. 29.Durairaj,M.,Leskiw,B.K.&Jensen,S.E.(1996)Geneticand biochemical analysis of the cysteinyl residues of isopenicillin N synthase from Streptomyces clavuligerus. Can. J. Microbiol. 42, 870–875. 30. Moriguchi,T.,Kita,M.,Ogawa,K.,Tomono,Y.,Endo,T.& Omura, M. (2002) Flavonol synthase gene expression during citrus fruit development. Physiol. Plant. 114, 251–258. 31. Zagursky, R.J. & Berman, M.L. (1984) Cloning vectors that yield high levels of single-stranded DNA for rapid DNA-sequencing. Gene 27, 183–191. 32. Kunkel, T.A., Roberts, J.D. & Zakour, R.A. (1987) Rapid and efficient site-specific mutagenesis without phenotypic selection. Methods Enzymol. 154, 367–382. 33. Sanger, F., Nicklen, S. & Coulsen, A.R. (1977) DNA sequencing with chain terminating inhibitors. Proc. Natl Acad. Sci. USA 74, 5463–5467. 34. Sreerama, N. & Woody, R.W. (1993) A self-consistent method for the analysis of protein secondary structure from circular dichro- ism. Anal. Biochem. 209, 32–44. 35. Deleage, G. & Geourjon, C. (1993) An interactive graphic pro- gram for calculating the secondary structure content of proteins from circular dichroism spectrum. Comput. Appl. Biosci. 9,197– 199. 36. Laemmli, U.K. (1970) Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 227, 680–685. 37. Towbin, H., Staehelin, T. & Gordon, J. (1979) Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: procedure and some applications. Proc. Natl Acad. Sci. USA 76, 4350–4354. 38. Lukac ˇ in, R. (1997) Molekulare und Strukturelle Charakterisuring der Flavanon 3b-Hydroxylase aus Petunia hybrida, einer 2-Oxoglutarrat-abhongigen Dioxygenase der Flavanoidbio- synthase. Doctoral Thesis, Universita ¨ t Freiburg im Breisgau, Germany. 39. Sandermann, H. & Strominger, L. (1972) Purification and properties of C 55 -isoprenoid alcohol phosphokinase from Staphylococcus aureus. J. Biol. Chem. 247, 5123–5131. 40. Bensadoun, A. & Weinstein, B. (1976) Assay of proteins in the presence of interfering material. Anal. Biochem. 70, 241–250. 41. Lukac ˇ in, R., Springob, K., Urbanke, K., Ernwein, C., Schro ¨ der, G., Schro ¨ der, J. & Matern, U. (1999) Native acridone synthases I and II from Ruta graveolens L. form homodimers. FEBS Letters 448, 135–140. 42. Springob, K., Lukac ˇ in, R., Ernwein, C., Gro ¨ ning,I.&Matern,U. (2000) Specificities of functionally expressed chalcone and acri- done synthases from Ruta graveolens. Eur. J. Biochem. 267, 6552– 6559. 43. Lukac ˇ in, R., Schreiner, S. & Matern, U. (2001) Transformation of acridone synthase to chalcone synthase. FEBS Letters 508,413– 417. 44. Schofield, C.J. & Zhang, Z. (1999) Structural and mechanistic studies on 2-oxoglutarate-dependent oxygenases and related enzymes. Curr. Op. Struct. Biol. 9, 722–731. 4142 F. Wellmann et al. (Eur. J. Biochem. 269) Ó FEBS 2002 . Functional expression and mutational analysis of flavonol synthase from Citrus unshiu Frank Wellmann 1, *, Richard Lukac ˇ in 1, *,. detail. Therefore, a cDNA from Citrus unshiu (Satsuma mandarin) designated as flavonol synthase was expressed in Escherichia coli ,and the purified recombinant