Cytochrome P450 CYP79A2 from Arabidopsis thaliana L. Catalyzes the conversion of L-phenylalanine to phenylacetaldoxime in the biosynthesis of benzylglucosinolate.

Glucosinolates are natural plant products gaining increasing interest as cancer-preventing agents and crop protectants. Similar to cyanogenic glucosides, glucosinolates are derived from amino acids and have aldoximes as intermediates. We report cloning and characterization of cytochrome P450 CYP79A2 involved in aldoxime formation in the glucosinolate-producing Arabidopsis thaliana L. The CYP79A2 cDNA was cloned by polymerase chain reaction, and CYP79A2 was functionally expressed in Escherichia coli. Characterization of the recombinant protein shows that CYP79A2 is an N-hydroxylase converting L-phenylalanine into phenylacetaldoxime, the precursor of benzylglucosinolate. Transgenic A. thaliana constitutively expressing CYP79A2 accumulate high levels of benzylglucosinolate. CYP79A2 expressed in E. coli has a K(m) of 6.7 micromol liter(-1) for L-phenylalanine. Neither L-tyrosine, L-tryptophan, L-methionine, nor DL-homophenylalanine are metabolized by CYP79A2, indicating that the enzyme has a narrow substrate specificity. CYP79A2 is the first enzyme shown to catalyze the conversion of an amino acid to the aldoxime in the biosynthesis of glucosinolates. Our data provide the first conclusive evidence that evolutionarily conserved cytochromes P450 catalyze this step common for the biosynthetic pathways of glucosinolates and cyanogenic glucosides. This strongly indicates that the biosynthesis of glucosinolates has evolved based on a cyanogenic predisposition.

Glucosinolates are amino acid-derived, secondary plant products containing a sulfate and a thioglucose moiety. Glucosinolates are found throughout the order Capparales, which includes agriculturally important crops of the Brassicaceae family such as oilseed rape and Brassica forages and vegetables, and the model plant Arabidopsis thaliana L. Upon tissue damage, glucosinolates are rapidly hydrolyzed to biologically active degradation products by the thioglucosidase myrosinase (EC 3.2.3.1). Glucosinolates or rather their degradation products defend plants against insect and fungal attack (1,2) and serve as attractants to insects that are specialized feeders on Brassicaceae (3). The degradation products have toxic as well as protective effects in higher animals and humans (1). Anti-nutritional effects such as growth retardation caused by consumption of large amounts of rape seed meal have an economic impact, as they restrict the use of this protein-rich animal feed. Anticarcinogenic activity has been documented by pharmacological studies for several degradation products of glucosinolates, e.g. for sulforaphane, a degradation product of 4-methylsulfinylbutylglucosinolate from broccoli sprouts (4). Metabolic engineering of the biosynthetic pathways of glucosinolates would open the possibility to tissue-specifically regulate and optimize the level of individual glucosinolates to improve the nutritional value of a given crop.
To date, more than 100 different glucosinolates have been identified. They are grouped into aliphatic, aromatic, and indolyl glucosinolates, depending on whether they are derived from aliphatic amino acids, phenylalanine and tyrosine, or tryptophan. The amino acid often undergoes a series of chain elongations prior to entering the biosynthetic pathway, and the glucosinolate product is often subject to secondary modifications such as hydroxylations, methylations, and oxidations, giving rise to the structural diversity of glucosinolates. A. thaliana cv. Columbia has been shown to contain 23 different glucosinolates derived from tryptophan, the chain-elongated phenylalanine homologue homophenylalanine, and several chainelongated methionine homologues (5).
Although not many genes of the glucosinolate biosynthetic pathway have been identified, many of the intermediates and some of the enzymes involved are known. In vivo biosynthetic studies have previously shown that N-hydroxyamino acids, aldoximes, thiohydroximates, and desulfoglucosinolates are precursors of glucosinolates (for review, see Ref. 6). The enzymes catalyzing the last two steps in the pathway, UDPG: thiohydroximate glucosyltransferase (EC 2.4.1.-) and 3Ј-phosphoadenosine 5Ј-phosphosulfate:desulfoglucosinolate sulfotransferase (EC 2.8.2.-), have been purified and shown to be nonspecific with respect to the nature of the side chain (7)(8)(9)(10). The sulfur-donating enzyme has not been characterized, but feeding experiments suggest that cysteine is the sulfur donor (11). The nature of the enzymes catalyzing the conversion of amino acids to aldoximes has been the subject of many discussions. Independent biochemical studies have indicated that three different enzyme systems are involved in this step, namely cytochrome P450-dependent monooxygenases, flavincontaining monooxygenases, and peroxidases (12). The aromatic amino acids (tyrosine and phenylalanine) are converted to the corresponding aldoximes by cytochrome P450-dependent monooxygenases in microsomes isolated from Sinapis alba L., Tropaeolum majus L., and Carica papaya L. (13)(14)(15)(16). Conversion of homophenylalanine and chain elongated methionine homologues into the corresponding aldoximes by microsomal preparations of Brassica spp. and S. alba has been shown to be cytochrome P450-independent and suggested to be catalyzed by flavin-containing monooxygenases (16 -19). The formation of indole-3-acetaldoxime from tryptophan in the biosynthesis of indoleglucosinolates has been shown to be catalyzed by a plasma membrane-bound peroxidase in microsomes from Chinese cabbage (Brassica campestris ssp. pekinensis) (20).
In the biosynthesis of cyanogenic glucosides, cytochromes P450 of the CYP79 family catalyze the formation of aldoximes from amino acids (21)(22)(23)(24). Several CYP79 homologues have previously been identified in glucosinolate-producing plants (25), but their function has never been determined. In the present paper, we report cloning and functional expression of the cytochrome P450 CYP79A2 from A. thaliana. We show that CYP79A2 catalyzes the conversion of L-phenylalanine to phenylacetaldoxime and that transgenic A. thaliana expressing CYP79A2 under control of CaMV35S 1 promoter accumulate high levels of benzylglucosinolate. Our data are consistent with the involvement of CYP79A2 in the biosynthesis of benzylglucosinolate in A. thaliana.
cDNA Cloning-PCR was performed on phage DNA representing 2.5 ϫ 10 7 plaque-forming units of the A. thaliana L. (cv. Wassilewskija) silique cDNA library CD4 -12 (kindly provided by Dr. Linda A. Castle and Dr. David W. Meinke, Department of Botany, Oklohoma State University, Stillwater, OK, and ABRC) using primers A2F1/A2R1. PCR reactions were set up in a total volume of 50 l in Expand HF buffer with 1.5 mM MgCl 2 (Roche Molecular Biochemicals) supplemented with 200 M dNTPs, 50 pmol of each primer, and 5% (v/v) Me 2 SO. After incubation of the reactions at 97°C for 3 min, 2.6 units of Expand High Fidelity PCR system (Roche Molecular Biochemicals) were added and 35 cycles of 90 s at 95°C, 60 s at 65°C, 120 s at 70°C were run. 0.5 l of the reaction were subjected to nested PCR with primers A2F2/A2R2 using the same PCR conditions. PCR fragments of the expected size were excised from an agarose gel and cloned into EcoRI/HindIII-digested pYX223 (R&D Systems), and the inserts of 10 clones derived from two nested PCR reactions were sequenced.
DNA Sequencing and Computer Analysis-Sequencing was performed using a Thermo Sequence fluorescent-labeled primer cycle sequencing kit (7-deaza-dGTP) (Amersham Pharmacia Biotech) and analyzed on an ALF-Express DNA sequencer (Amersham Pharmacia Biotech). Sequence computer analysis was done with programs of the GCG Wisconsin sequence analysis package. The GAP program was used with a gap creation penalty of 8 and a gap extension penalty of 2 to compare pairs of sequences. The splice site prediction was done using NetPlantGene (26).
Generation of Expression Constructs-Expression constructs were derived from a CYP79A2 cDNA, which had been obtained by fusion of the two CYP79A2 exons generated from genomic DNA of Arabidopsis thaliana L. The two exons were amplified by PCR with primers A2F2/ A2R3 and A2F3/A2R2, respectively, using 1.25 units of Pwo polymerase (Roche Molecular Biochemicals) and 4 g of template DNA. PCR reactions were set up in a total volume of 50 l in Pwo polymerase PCR buffer with 2 mM MgSO 4 (Roche Molecular Biochemicals) supplemented with 200 M dNTPs, 50 pmol of each primer, and 5 (v/v) % Me 2 SO. After incubation of the reactions at 94°C for 3 min, 30 PCR cycles of 20 s at 94°C, 10 s at 60°C, and 30 s at 72°C were run. After digestion of the PCR fragments with EcoRI (exon 1) and HindIII (exon 2), the blunt ends generated with primers A2R3 and A2F3 and Pwo polymerase were phosphorylated with T4 polynucleotide kinase (New England Biolabs). The two exons were then ligated into EcoRI/HindIII-digested pYX223. The cloned cDNA was sequenced to exclude incorporation of PCR errors.
Expression in Escherichia coli-E. coli cells of strain JM109 transformed with the expression constructs were grown overnight in LB medium supplemented with 100 g ml Ϫ1 ampicillin and used to inoculate 100 ml of modified TB medium containing 50 g ml Ϫ1 ampicillin, 1 mM thiamine, 75 g ml Ϫ1 ␦-aminolevulinic acid, and 1 mM isopropyl-␤-D-thiogalactoside (28). The cells were grown at 28°C for 65 h at 125 rpm. Cells from 75 ml of culture were pelleted and resuspended in buffer composed of 0.1 M Tris-HCl, pH 7.6, 0.5 mM EDTA, 250 mM sucrose, and 250 M phenylmethylsulfonyl fluoride. Lysozyme was added to a final concentration of 100 g ml Ϫ1 . After incubation for 30 min at 4°C, magnesium acetate was added to a final concentration of 10 mM. Spheroplasts were pelleted, resuspended in 5 ml of buffer composed of 10 mM Tris-HCl, pH 7.5, 14 mM magnesium acetate, and 60 mM potassium acetate, pH 7.4, and homogenized in a Potter-Elvehjem homogenizer. After DNase and RNase treatment, glycerol was added to a final concentration of 29%. Temperature-induced Triton X-114 phase partitioning was performed as described previously (29). The Triton X-114-rich phase was analyzed by SDS-polyacrylamide gel electrophoresis.
CO Difference Spectroscopy-Fe 2ϩ ⅐CO versus Fe 2ϩ difference spectroscopy (30) was performed on 100 l of E. coli spheroplasts resuspended in 900 l of buffer containing 50 mM KP i , pH 7.5, 2 mM EDTA, 20% (v/v) glycerol, 0.2% (v/v) Triton X-100, and a few grains of sodium dithionite. The solubilizate was distributed between two cuvettes, and a base line was recorded between 400 and 500 nm on a SLM Aminco DW-2000 TM spectrophotometer (SLM Instruments, Urbana, IL). The sample cuvette was flushed with CO for 1 min, and the difference spectrum was recorded. The amount of functional cytochrome P450 was estimated, based on an absorption coefficient of 91 l mmol Ϫ1 cm Ϫ1 .
Measurements of Enzyme Activities-The activity of CYP79A2 was measured in E. coli spheroplasts reconstituted with NADPH:cytochrome P450 oxidoreductase purified from Sorghum bicolor (L.) Moench as described previously (21). In a typical enzyme assay, 5 l of spheroplasts and 4 l of NADPH:cytochrome P450 reductase (equivalent to 0.04 units defined as 1 mol of cytochrome c min Ϫ1 ) were incubated with 3. C]phenylalanine was linear with time within the first 2 h of incubation as determined using time points 30 min, 1 h, 2 h, and 6 h. For estimation of K m and V max values, reaction mixtures were incubated for 2 h at 26°C. For GC-MS analysis, 450 l of reaction mixture containing 33 M L-phenylalanine (Sigma) or 33 M homophenylalanine were incubated for 4 h at 26°C and extracted twice with a total volume of 600 l of chloroform (Normapur, May and Baker). The organic phases were combined and evaporated to dryness. The residue was dissolved in 15 l of chloroform and analyzed by GC-MS. GC-MS analysis was performed on an HP5890 Series II gas chromatograph directly coupled to a Jeol JMS-AX505W mass spectrometer. An SGE column (BPX5, 25 m ϫ 0.25 mm, 0.25-m film thickness) was used (head pressure 100 kilopascals, splitless injection). The oven temperature program was as follows: 80°C for 3 min, 80°C to 180°C at 5°C min Ϫ1 , 180°C to 300°C at 20°C min Ϫ1 , 300°C for 10 min. The ion source was run in EI mode (70 eV For expression of CYP79A2 under control of the CaMV35S promoter in A. thaliana, the native full-length CYP79A2 cDNA was introduced into EcoRI/KpnI-digested pRT101 (31) via several subcloning steps. The expression cassette was excised by HindIII digestion and transferred to pPZP111 (32). Agrobacterium tumefaciens strain C58 (33) transformed with this construct was used for plant transformation by floral dip (34) using 0.005% (v/v) Silwet L-77 and 5% (w/v) sucrose in 10 mM MgCl 2 . Seeds were germinated on Murashige and Skoog medium supplemented with 50 g ml Ϫ1 kanamycin, 2% (w/v) sucrose, and 0.9% (w/v) agar. Transformants were selected after 2 weeks and transferred to soil.
HPLC Analysis of Plant Extracts-Rosette leaves (five to eight leaves of different age from each plant) were harvested from 6-week-old plants (nine transgenic plants and three wild-type plants), immediately frozen in liquid nitrogen and freeze-dried for 48 h. Desulfoglucosinolates were analyzed as described in Ref. 35. Briefly, freeze-dried material (2-5 mg) was homogenized in 3.5 ml of boiling 70% (v/v) methanol by a Polytron homogenizer for 1 min, 10 l of internal standard (5 mM p-hydroxybenzylglucosinolate; Bioraf Denmark) were added, and homogenization was continued for another 1 min. Plant material was pelleted, and the pellet was re-extracted with 3.5 ml of boiling 70% (v/v) methanol for 1 min using a Polytron homogenizer. Plant material was pelleted, washed in 3.5 ml of 70% (v/v) methanol, and centrifuged. The supernatants were pooled and loaded on a DEAE-Sephadex A-25 (Amersham Pharmacia Biotech) column equilibrated as follows; 25 mg of DEAE-Sephadex A-25 were swollen overnight in 1 ml of 0.5 M acetate buffer, pH 5, packed into a 5-ml pipette tip, and washed with 1 ml of water. The plant extract was loaded, and the column was washed with 2 ml of 70% (v/v) methanol, 2 ml of water, and 0.5 ml of 0.02 M acetate buffer, pH 5. Helix pomatia sulfatase (type H-1, Sigma; 0.1 ml, 2.5 mg ml Ϫ1 in 0.02 M acetate buffer, pH 5) were applied, and the column was left at room temperature for 16 h. Elution was carried out with 2 ml of water. The eluate was dried in vacuo, the residue dissolved in 150 l of water, and 100 l were subjected to HPLC on a Shimadzu LC-10A Tvp equipped with a Supelcosil LC-ABZ 59142 C 18 column (25 cm ϫ 4.6 mm, 5 mm; Supelco) and a SPD-M10AVP photodiode array detector (Shimadzu). The flow rate was 1 ml min Ϫ1 . Elution with water for 2 min was followed by elution with a linear gradient from 0% to 60% methanol in water (48 min), a linear gradient from 60% to 100% methanol in water (3 min), and with 100% methanol (3 min). The assignment of peaks was based on retention times and UV spectra compared with standard compounds. Glucosinolates were quantified in relation to the internal standard and by use of the response factors as described previously (36,37). In the analysis of rosette leaves, the term "total glucosinolate content" refers to the molar amount of the five major glucosinolates (4-methylsulfinylbutylglucosinolate, 4-methylthiobutylglucosinolate, 8-methylsulfinyloctylglucosinolate, indol-3-ylmethylglucosinolate, and 4-methoxyindol-3-ylglucosinolate), which account for 85% of the glucosinolate content in rosette leaves of wild-type A. thaliana (37) and benzylglucosinolate. The glucosinolate content of seeds harvested from T1 plants 10, 13, and 14 was analyzed and compared with the glucosinolate content of wild-type seeds. Twelve to thirty milligrams of seeds were extracted and subjected to HPLC analysis as described above (with the exception that lyophilization of the tissue was omitted). In the analysis of seeds, the term total glucosinolate content refers to the molar amount of the 10 major glucosinolates (3-hydroxypropylglucosinolate, 4-hydroxybutylglucosinolate, 4-methylsulfinylbutylglucosinolate, 4-methylthiobutylglucosinolate, 8-methylsulfinyloctylglucosinolate, 7-methylthioheptylglucosinolate, 8-methylthiooctylglucosinolate, indol-3-ylmethylglucosinolate, 3-benzoyloxypropylglucosinolate, and 4-benzoyloxybutylglucosinolate), which account for more than 90% of the glucosinolate content in seeds of wild-type A. thaliana (37) and benzylglucosinolate.

RESULTS
Cloning of the CYP79A2 cDNA-CYP79A2 (GenBank accession no. AB010692 (11,000 -13,200 region)) is one of several CYP79 homologues identified in the genome of A. thaliana. According to a computer-aided splice site prediction (26), CYP79A2 contains one intron. This intron, which is characteristic for A-type cytochromes P450 (38), is shared by all members of the CYP79 family known to date. Although it is the only intron in CYP79A2, CYP79B2 and CYP79B3, other members of the CYP79 family have one or two additional introns. This may suggest that CYP79A2 is one of the ancient members of the family. Using a PCR approach, we have isolated a full-length CYP79A2 cDNA from an A. thaliana silique cDNA library. The sequence of the cDNA confirmed the splice site prediction (Fig.  1). The reading frame of the CYP79A2 cDNA has two potential ATG start codons, one positioned 15 base pairs downstream of a stop codon in the 5Ј-untranslated region and another one 15 base pairs further downstream. We have used a cDNA starting with the second ATG codon for all further studies. This cDNA encodes a protein of 523 amino acids, which has 64% similarity and 53% identity to CYP79A1 involved in the biosynthesis of the cyanogenic glucoside dhurrin (22).
PCR amplification of a full-length CYP79A2 cDNA from the silique cDNA library required a total of 70 PCR cycles resulting in incorporation of PCR errors. PCR errors at different positions were verified by sequencing of 10 different cloned PCR products, each of which contained at least a single PCR error. Consequently, these cDNAs were not suitable for expression of the protein. PCR amplification of fragments of the CYP79A2 cDNA from the A. thaliana PRL2 cDNA library (CD4 -7 (Ref. 39), kindly provided by ABRC) and an A. thaliana whole plant library (Stratagene, 937010) was unsuccessful. This may indicate that the CYP79A2 mRNA is expressed at very low levels. An alternative approach, in which the two exons of the CYP79A2 gene were directly PCR-amplified from genomic DNA by relatively few PCR cycles, was applied to clone the CYP79A2 cDNA for protein expression. The property of Pwo polymerase not to add nucleotides to the ends of the PCR product enabled blunt end ligation of the two amplified exons to yield the cDNA. This cDNA was used for all further studies.
Expression of CYP79A2 in E. coli-CYP79A2 was expressed in E. coli strain JM109 by use of the expression vector pSP19g10L and cDNAs encoding the native CYP79A2 or three N-terminally modified CYP79A2 (Fig. 2). The expression vector pSP19g10L contains the lacZ promoter fused with the short leader sequence (g10L) of gene 10 from T7 bacteriophage. The g10L sequence has been documented as an excellent leader sequence for the expression of various heterologous proteins (40). N-terminal modifications of CYP79A2 were designed to apply previously obtained expression results for high-level expression of eukaryotic cytochromes P450 in E. coli (28,29). Two constructs were made to introduce the eight N-terminal amino acids of the bovine cytochrome P450 CYP17A in front of the N terminus of CYP79A2 (yielding modified CYP79A2) and a trun-cated CYP79A2 (yielding truncated-modified CYP79A2), respectively. Previous experiments with CYP79A1 from S. bicolor had shown that expression levels as high as 900 nmol of cytochrome P450 (liter of culture) Ϫ1 were obtained with a protein, in which the first 24 amino acids of CYP79A1 were replaced by the eight N-terminal amino acids of bovine CYP17A (CYP79A1⌬(1-24) bov ; Ref. 29). Thus the N terminus of this cytochrome P450 seems to be especially suitable for expression in E. coli. We have therefore generated a fourth construct (chimeric CYP79A2; Fig. 2) in which we fused the cDNA encoding the N terminus of CYP79A1⌬(1-24) bov (50 amino acids) with the cDNA encoding the catalytic domain of CYP79A2 (amino acids 41-523).
Protein bands migrating with an apparent molecular mass of about 60 kDa on SDS-polyacrylamide gels were detected in the detergent-rich phase obtained by temperature-induced Triton X-114 phase partitioning of E. coli spheroplasts harboring expression constructs for the native, the truncated-modified, and the chimeric CYP79A2. As expected, the chimeric CYP79A2 migrated with a slightly higher molecular mass than the native and the truncated-modified CYP79A2. The band was not present in the detergent-rich phase from cells harboring the modified CYP79A2 expression construct or the empty vector. Spectral analysis of the different spheroplast preparations showed that the chimeric CYP79A2 and, to a lesser extent, the truncated-modified CYP79A2 produced a CO difference spectrum with the characteristic peak at 452 nm, indicating the presence of a functional cytochrome P450 (Fig. 3). A peak at 415 nm was found for all spheroplast preparations. This peak may arise from E. coli derived heme protein, unattached heme groups produced in the presence of ␦-aminolevulinic acid in the medium, or cytochrome P450 in a non-functional conformation (29,41). Based on the peak at 452 nm, the expression level of chimeric CYP79A2 was estimated to be 50 nmol of cytochrome P450 (liter of culture) Ϫ1 . When incubated with L-[ 14 C]phenylalanine, spheroplasts of E. coli transformed with the native, the truncated-modified, or the chimeric CYP79A2 expression construct and reconstituted with the purified NADPH:cytochrome P450 oxidoreductase from S. bicolor produced two radiolabeled compounds, which comigrated with the (E)-and (Z)-isomers of phenylacetaldoxime in thin layer chromatography (Fig. 4). These products were not detected in assay mixtures containing E. coli spheroplasts harboring either the modified CYP79A2 expression construct or the empty vector. GC-MS analysis showed that two compounds with identical fragmentation patterns were present in the reaction mixture with chimeric CYP79A2, but not in the control reaction (Fig. 5). The retention times and the fragmentation pattern identified these compounds as the (E)-and (Z)-isomers of phenylacetaldoxime (Fig. 5). Administration of L-[ 14 C]tyrosine, L-[ 14 C]methionine, or L-[ 3 H]tryptophan to spheroplasts of E. coli expressing the native or the chimeric CYP79A2 did not result in production of detectable amounts of the respective aldoximes. The ability of CYP79A2 to metabolize DL-homophenylalanine was investigated in spheroplasts of E. coli expressing chimeric CYP79A2. GC-MS analysis of the reaction mixture showed the absence of detectable amounts of the homophenylalanine-derived aldoxime. A K m value of 6.7 mol l Ϫ1 and a V max value of 16.6 pmol min Ϫ1 (mg protein) Ϫ1 were determined for CYP79A2 using spheroplasts of E. coli expressing native CYP79A2 with L-[ 14 C]phenylalanine as the substrate. As no CO spectrum was obtained with native CYP79A2, we were not able to estimate the amount of functional native CYP79A2. However, based on the expression level of functional chimeric CYP79A2, we can estimate a turnover number of 0.24 min Ϫ1 for native CYP79A2.
The activity of recombinant CYP79A2 was strongly dependent on the pH of the reaction mixture and, to a lesser extent, on several other factors. Compared with the activity at pH 7.5, the activity of chimeric CYP79A2 was 25% at pH 6, 50% at pH 6.5, 80% at pH 7.0, and 70% at pH 7.9. Addition of Tween 80 to a final concentration of 0.083% (v/v) resulted in a 1.5-fold increase in aldoxime production. Addition of reduced glutathione Primers A2F1/A2R1 and A2F2/A2R2 were used for PCR amplification of the cDNA from the A. thaliana CD4 -12 cDNA library. Primers A2F2/A2R3 and A2F3/A2R2 were used for PCR amplification of exons 1 and 2, respectively, from genomic DNA. Primers A2FX1, A2FX2, and A2FX3 were used in combination with primer A2R4 to generate constructs for expression of N-terminally modified CYP79A2. The numbers below the sequence refer to AB010692 (11,000 -13,200 region reversed-complemented). Numbers of the first nucleotide, the first ATG codon in exon 1, the first nucleotide of the intron, the first nucleotide of exon 2, and the first nucleotide after the stop codon are given.

FIG. 2. N-terminal amino acid sequences deduced from the CYP79A2 expression constructs.
Sequence of CYP17A is shown in white letters on black background, and sequence of CYP79A1 is shown in black letters on gray background. Dots above the sequences refer to the amino acid numbering of native CYP79A2.
to a final concentration of 3 mM stimulated aldoxime production, but to a lesser extent.
Constitutive Expression of CYP79A2 in Transgenic A. thaliana-We have expressed CYP79A2 under control of the CaMV35S promoter in A. thaliana cv. Columbia. The appearance of the transgenic plants was comparable to wild-type plants. The glucosinolate content of rosette leaves of the T1 generation was investigated by HPLC of the desulfoglucosinolates. All the transgenic plants analyzed accumulated benzylglucosinolate, whereas benzylglucosinolate was not detected in simultaneously grown wild-type plants (Fig. 6). The content of benzylglucosinolate varied between different transgenic plants. In the three plants with highest accumulation, benzylglucosinolate accounted for 38% (plant 10), 5% (plant 14), and 2% (plant 13), respectively, of the total glucosinolate content of the leaves. HPLC analysis of seeds of these plants showed that benzylglucosinolate accounted for 35% (plant 10), 12% (plant 14), and 3% (plant 13) of the total glucosinolate content of the seeds. In seeds of wild-type plants (cv. Columbia and Wassilewskija), minute amounts of benzylglucosinolate were detected (in cv. Columbia, 0.034 mol (g fresh weight) Ϫ1 corresponding to 0.05% of the total glucosinolate content). The content of the homophenylalanine-derived 2-phenylethylglucosinolate was unaffected in leaves and seeds of the transgenic plants compared with wild-type plants. DISCUSSION In the present study, we report cloning and characterization of CYP79A2, a cytochrome P450-dependent monooxygenase, which catalyzes the conversion of L-phenylalanine to phenylacetaldoxime in the biosynthesis of benzylglucosinolate in A. thaliana. Characterization of recombinant CYP79A2 and transgenic A. thaliana constitutively expressing CYP79A2 provide evidence for previous biochemical data, which have shown that cytochrome P450-dependent monooxygenases catalyze the N-hydroxylation of aromatic amino acids to aldoximes in glucosinolate biosynthesis.
A. thaliana represents an ideal model plant for the study of glucosinolate biosynthesis, as the knowledge rapidly emerging from the Arabidopsis Genome Sequencing Initiative can directly be used to identify candidate genes involved in glucosinolate metabolism. Glucosinolates are related to cyanogenic glucosides as the biosynthesis of both classes of secondary plant products proceeds from amino acids via aldoximes. Cyanogenic glucosides are widely distributed in the plant kingdom. They are present in angiosperms, gymnosperms, and ferns, indicating that cyanogenesis has developed early in evolution (42). The occurrence of glucosinolates is restricted to the order Capparales and the genus Drypetes (Euphorbiales) (43). C. papaya is the only known example of a plant containing both glucosinolates and cyanogenic glucosides (15). It has been suggested that glucosinolate biosynthesis might have evolved based on a "cyanogenic predisposition" (43), implicating that homologous enzymes catalyze related reactions in the biosynthesis of both groups of compounds. CYP79A2 is one of several CYP79 homologues in A. thaliana. We have cloned the corresponding cDNA and characterized the recombinant protein expressed in E. coli. Our data show that CYP79A2 is an N-hydroxylase, which catalyzes the conversion of L-phenylalanine to phenylacetaldoxime. This provides the first evidence that CYP79 homologues are involved in the biosynthesis of glucosinolates. The substrate specificity of CYP79A2 seems to be rather narrow, as neither L-tyrosine, DL-homophenylalanine, L-tryptophan, nor L-methionine are metabolized by the enzyme, even though the former two amino acids are structurally similar to phenylalanine and function as precursors of glucosinolates. The high substrate specificity is in agreement with results obtained with CYP79 homologues involved in the biosynthesis of cyanogenic glucosides (23,29,44). The kinetic data for recombinant E. coli does not contain endogenous cytochromes P450 or NADPH:cytochrome P450 reductase. However, two soluble E. coli flavoproteins, flavodoxin and NADPH-flavodoxin reductase, have been shown to support the catalytic activity by donating reducing equivalents to heterologously expressed cytochromes P450 (28,45). Our experiments with CYP79A2 showed that the E. coli flavoproteins do not support the catalytic activity of recombinant CYP79A2. Detectable amounts of phenylacetaldoxime were only obtained in the presence of ex-ogenously supplied NADPH:cytochrome P450 reductase. The stimulation of the enzyme activity upon addition of glutathione or Tween 80 to the reaction mixture might reflect an improved interaction of CYP79A2 with NADPH:cytochrome P450 reductase. Glutathione has been reported previously to stimulate cytochrome P450 activity in reconstitution experiments with purified CYP3A4 (46) and has been suggested to affect the reductase-cytochrome P450 interaction (29). Detergents, e.g. Tween 20 and Triton X-100, have been demonstrated to cause conformational changes of cytochrome P450 CYP1A2, leading to increased catalytic activity (47).
Additional evidence for the involvement of CYP79A2 in glucosinolate biosynthesis is provided by accumulation of the phenylalanine-derived glucosinolate benzylglucosinolate (also referred to as glucotropaeolin) in transgenic A. thaliana expressing CYP79A2 under the control of the CaMV35S promoter. All transgenic plants analyzed in the present study contained benzylglucosinolate in the rosette leaves, whereas rosette leaves of wild-type plants did not contain detectable amounts of this glucosinolate. Although seeds of A. thaliana cv. Columbia are known to contain the homophenylalanine-derived 2-phenylethylglucosinolate, the occurrence of benzylglucosinolate has never been reported for A. thaliana. However, we have detected minute amounts of benzylglucosinolate in seeds of A. thaliana cv. Columbia and cv. Wassilewskija. Seeds of transgenic plants accumulated high levels of benzylglucosinolate. The content of 2-phenylethylglucosinolate was unchanged in seeds of transgenic plants compared with seeds of wild-type plants. This supports the data obtained with CYP79A2 expressed in E. coli and shows that CYP79A2 converts specifically phenylalanine, but not homophenylalanine to the corresponding aldoxime. As indicated by the accumulation of high levels of benzylglucosinolate in several transgenic plants, the formation of phenylacetaldoxime is the rate-limiting step in the biosynthesis of benzylglucosinolate in A. thaliana.
Benzylglucosinolate is only sporadically observed in roots and cauline leaves of wild-type A. thaliana cv. Columbia and may be induced by environmental conditions. 2 The sporadic occurrence of benzylglucosinolate corresponds well with our observation that the CYP79A2 mRNA is a low abundant transcript. We were not able to detect the CYP79A2 mRNA in seedlings, rosette leaves of different developmental stages, and cauline leaves of A. thaliana cv. Columbia by Northern blotting and RT-PCR (data not shown). Initial experiments with wildtype A. thaliana cv. Columbia seedlings in liquid culture indicate that no induction of benzylglucosinolate biosynthesis occurs after treatments with salicylic acid, methyl jasmonate, or the ethylene precursor 1-aminocyclopropane-1-carboxylic acid. The elucidation of environmental factors which lead to induction of benzylglucosinolates in A. thaliana must await further studies.
The nature of the enzymes involved in the conversion of amino acids to aldoximes in the biosynthesis of glucosinolates has been studied in different plant species (for review, see Ref. 12). In microsomal enzyme systems isolated from S. alba, T. majus, and C. papaya, tyrosine or phenylalanine are converted to the corresponding aldoximes by cytochrome P450-dependent monooxygenases, as evidenced by photoreversible carbon monoxide inhibition and other inhibitor studies (13)(14)(15)(16). Biochemical studies with microsomal preparations from Brassica napus L. have indicated that the conversion of homophenylalanine, dihomo-, trihomo-, and tetrahomomethionine to the corresponding aldoximes is not affected by carbon monoxide and 2 P. Brown and J. Gershenzon, personal communication. other cytochrome P450 inhibitors (17,18). Instead, the enzymes showed characteristics of flavin-containing monooxygenases, which have subsequently been suggested to catalyze these reactions as well as the N-hydroxylation of phenylalanine in other members of the Brassicaceae family (16,19). A third enzyme system, a plasma membrane-bound peroxidase first characterized in Chinese cabbage (B. campestris ssp. pekinensis), has been shown to convert tryptophan to indole-3-acetaldoxime in the biosynthesis of indoleglucosinolates (20). It has been proposed that the involvement of cytochrome P450-dependent monooxygenase may be restricted to species that do not belong to the Brassicaceae family, implicating that the cytochrome P450-dependent formation of p-hydroxyphenylacetaldoxime in S. alba has to be regarded as a unique exception from the rule or an experimental artifact (16). Our data strongly suggest that aldoxime formation from aromatic amino acids is dependent on cytochrome P450 enzymes in members of the Brassicaceae as well as in other families.
In conclusion, we have identified a cytochrome P450-depend-ent monooxygenase, CYP79A2 from A. thaliana, which catalyzes the conversion of L-phenylalanine to phenylacetaldoxime, the first step in the biosynthesis of benzylglucosinolate. CYP79A2 is the first enzyme shown to catalyze aldoxime formation in glucosinolate biosynthesis. Our data provide the first evidence for the involvement of CYP79 homologues in this step, confirming that the aldoxime formation is catalyzed by evolutionarily conserved cytochromes P450 in the biosynthesis of glucosinolates and cyanogenic glucosides. This indicates that the biosynthesis of glucosinolates has evolved based on a cyanogenic predisposition. Several glucosinolates, including benzylglucosinolate (48 -50), have been shown to be degraded to biologically active compounds with anticarcinogenic, fungitoxic, and herbicidal activity. Considering the increasing interest in glucosinolate-containing vegetables as foods possessing cancer-preventing properties (51,52) and in the use of glucosinolates as biopesticides and herbicides (53), the identification of enzymes involved in the biosynthesis of glucosinolates has great potential for biotechnological uses and the development FIG. 6. HPLC analysis of A. thaliana expressing CYP79A2 under control of the CaMV35S promoter. Methanol extracts of rosette leaves of CYP79A2 sense plant 10 (A) and a wild-type plant (B) were treated with H. pomatia sulfatase, and the content of desulfoglucosinolates was analyzed by HPLC. Peak 1, p-hydroxybenzyldesulfoglucosinolate (internal standard); peak 2, benzyldesulfoglucosinolate; peak 3, 4-methylthiobutyldesulfoglucosinolate; peak 4, indol-3-ylmethyldesulfoglucosinolate; peak 5, 8-methylsulfinyloctyldesulfoglucosinolate; peak 6, 4-methoxyindol-3-ylmethyldesulfoglucosinolate. of crops with improved pest resistance and increased nutritional value.