Cytochrome P450 CYP79F1 from Arabidopsis Catalyzes the Conversion of Dihomomethionine and Trihomomethionine to the Corresponding Aldoximes in the Biosynthesis of Aliphatic Glucosinolates*

, Glucosinolates are natural plant products that have received rising attention due to their role in interac-tions between pests and crop plants and as chemical protectors against cancer. Glucosinolates are derived from amino acids and have aldoximes as intermediates. We report that cytochrome P450 CYP79F1 catalyzes aldoxime formation in the biosynthesis of aliphatic glucosinolates in Arabidopsis thaliana . Using recombinant CYP79F1 functionally expressed in Escherichia coli , we show that both dihomomethionine and trihomomethionine are metabolized by CYP79F1 resulting in the formation of 5-methylthiopentanaldoxime and 6-methyl-thiohexanaldoxime, respectively. 5-methylthiopenta-naldoxime is the precursor of the major glucosinolates in leaves of A. thaliana , i.e. 4-methylthiobutylgluc-osinolate and 4-methylsulfinylbutylglucosinolate,

myrosinase system is believed to be involved in plant defense. In addition, it has been shown that glucosinolates or rather the isothiocyanates, particularly sulforaphane, the isothiocyanate of 4-methylsulfinylbutylglucosinolate, have anticarcinogenic properties (3,4). There is a rising interest in being able to control the level of specific glucosinolates in crops to improve nutritional value and pest resistance.
Glucosinolates are grouped into aliphatic, aromatic, and indole glucosinolates, depending on whether they are derived from aliphatic amino acids, phenylalanine and tyrosine, or tryptophan (for review, see Ref. 5). 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. Biosynthetic intermediates common to all glucosinolates are aldoximes, thiohydroximates, and desulfoglucosinolates. The glucosinolates found in A. thaliana ecotype Columbia are derived from tryptophan, several chainelongated methionine homologues, chain-elongated phenylalanine (6), and phenylalanine (7). However, the dihomomethionine-derived glucosinolates 4-methylthiobutylglucosinolate and 4-methylsulfinylglucosinolate account for more than 50% of the total glucosinolate content in the rosette leaves of A. thaliana (8).
A key step in the biosynthesis of glucosinolates is the Nhydroxylation of the precursor amino acids to the corresponding aldoximes. In the biosynthesis of cyanogenic glucosides, a group of natural plant products closely related to glucosinolates, cytochromes P450 belonging to the CYP79 family catalyze the conversion of amino acids to aldoximes (9 -11). The nature of the enzymes catalyzing the formation of aldoximes in glucosinolate biosynthesis has been discussed controversially (12). Recently, evidence has been provided for the involvement of CYP79 homologues in the biosynthesis of aromatic (7) and indole glucosinolates (13,14). Regarding the biosynthesis of aliphatic glucosinolates, extensive biochemical studies with preparations of Brassica sp. and chain-elongated methionine homologues as substrates have suggested that aldoxime formation from these amino acids is not catalyzed by cytochromes P450, but by flavin-dependent monooxygenases (15)(16)(17).
In the present paper, we report the identification of a cytochrome P450 of the CYP79 family, CYP79F1, which catalyzes the conversion of dihomomethionine and trihomomethionine to 5-methylthiopentanaldoxime and 6-methylthiohexanaldoxime, respectively. The reduced levels of aliphatic glucosinolates and the accumulation of the chain-elongated precursor amino acids dihomomethionine and trihomomethionine in transgenic A. thaliana with CYP79F1 cosuppression is consistent with the involvement of CYP79F1 in the biosynthesis of aliphatic glucosinolates.

EXPERIMENTAL PROCEDURES
Plant Material-A. thaliana ecotype Columbia was used for all experiments. The plants were grown in a controlled-environment Arabidopsis Chamber (Ar-60 I, Percival, Boone, IA) at a photosynthetic flux of 100 -200 mol photons m Ϫ2 s Ϫ1 , 20°C, and 70% humidity. Unless otherwise stated, the photoperiod was 12-h light, 12-h dark.
Generation of the Construct for Escherichia coli Expression-The expression construct was derived from the EST 1 ATTS5112 (Arabidopsis Biological Resource Center, Columbus, OH), which contains the full-length sequence of CYP79F1. The CYP79F1 coding region was amplified from the EST by PCR with primer 1 (sense direction; 5Ј-CTCTAGATTCGAACATATGGCTAGCTTTACAACATCATTACC) and primer 2 (antisense direction; 5Ј-CGGGATCCTTAAGGACGGAACTTT-GGATA). Primer 1 introduces an XbaI site upstream of the start codon and an NdeI restriction site at the start codon. To optimize the construct for E. coli expression (18) primer 1 changes the second codon from ATG to GCT and introduces a silent mutation in codon 5. Primer 2 introduces a BamHI restriction site immediately after the stop codon. The PCR reaction was set up in a total volume of 50 l in Pwo polymerase PCR buffer with 2 mM MgSO 4 using 2.5 units of Pwo polymerase (Roche Molecular Biochemicals), 0.1 g of template DNA, 200 M dNTPs, and 50 pmol of each primer. After incubation of the reaction at 94°C for 5 min, 20 PCR cycles of 15 s at 94°C, 30 s at 58°C, and 2 min at 72°C were run. The PCR fragment was digested with XbaI and BamHI and ligated into an XbaI/BamHI-digested pBluescript II SK (Stratagene). The cDNA was sequenced to exclude PCR errors and transferred from pBluescript II SK to an NdeI/BamHI-digested pSP19g10L expression vector (18).
DNA Sequencing and Computer Analysis-Sequencing was performed on an ALF-Express (Amersham Pharmacia Biotech) using a Thermo Sequence Fluorescent-labeled primer cycle sequencing kit (7deaza-dGTP) (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 eight and a gap extension penalty of two to compare pairs of sequences.
Expression in E. coli-E. coli cells of strain JM109 (Stratagene) and strain C43(DE3) (19) transformed with the expression construct were grown overnight in LB medium supplemented with 100 g ml Ϫ1 ampicillin and used to inoculate 40 ml of modified TB medium containing 50 g ml Ϫ1 ampicillin, 1 mM thiamine, 75 g ml Ϫ1 ␦-aminolevulinic acid, 1 g ml Ϫ1 chloramphenicol, and 1 mM isopropyl-␤-D-thiogalactoside. The cultures were grown at 28°C for 60 h at 125 rpm. The cells were pelleted and resuspended in buffer composed of 0.2 M Tris-HCl, pH 7.5, 1 mM EDTA, 0.5 M sucrose, and 0.5 mM phenylmethylsulfonyl fluoride. Lysozyme was added to a final concentration of 100 g ml Ϫ1 . After incubation for 30 min at 4°C, Mg(OAc) 2 was added to a final concentration of 10 mM. Spheroplasts were pelleted, resuspended in 3.2 ml of buffer composed of 10 mM Tris-HCl, pH 7.5, 14 mM Mg(OAc) 2 , and 60 mM KOAc, pH 7.4, and homogenized in a Potter-Elvehjem homogenizer. After DNase treatment, glycerol was added to a final concentration of 30%. Temperature-induced Triton X-114 phase partitioning resulted in the formation of a detergent-rich phase containing the majority of the cytochrome P450 and a detergent-poor phase (9). Functional expression of CYP79F1 was monitored by Fe 2ϩ ⅐CO versus Fe 2ϩ difference spectroscopy (20) performed on an SLM Aminco DW-2000 TM spectrophotometer (SLM Instruments, Urbana, IL) using 10 l of Triton X-114-rich phase in 990 l of buffer containing 50 mM KP i , pH 7.5, 2 mM EDTA, 20% glycerol, 0.2% Triton X-100, and a few grains of sodium dithionite. The amount of functional CYP79F1 was estimated based on an absorption coefficient of 91 liters mmol Ϫ1 cm Ϫ1 .
Measurements of Enzyme Activities-The activity of CYP79F1 was measured in E. coli spheroplasts reconstituted with NADPH:cytochrome P450 oxidoreductase purified from Sorghum bicolor (L.) Moench as described earlier (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) were incubated with substrate in buffer containing 30 mM KP i , pH 7.5, 3 mM NADPH, 3 mM reduced glutathione, 0.042% Tween 80, 1 mg ml Ϫ1 L-␣-dilauroylphosphatidylcholine in a total volume of 30 l. Reaction mixtures containing spheroplasts of E. coli C43(DE3) transformed with empty vector were used as controls in all assays. 3 PerkinElmer Life Sciences) were tested as potential substrates. After incubation at 28°C for 1 h, half of the reaction mixture was analyzed by TLC on Silica Gel 60 F 254 sheets (Merck) using toluene/ethyl acetate 5:1 (v/v) as eluent. Radiolabeled bands were visualized and quantified by STORM 840 PhosphorImager (Amersham Pharmacia Biotech). For GC-MS analysis, 450 l of reaction mixture containing 3.3 mM L-methionine (Sigma), 3.3 mM DL-dihomomethionine or 3.3 mM DL-trihomomethionine, respectively, were incubated for 4 h at 25°C and extracted with a total volume of 600 l CHCl 3 . The organic phase was collected and evaporated, and the residue was dissolved in 15 l of CHCl 3 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-kPa, splitless injection). The oven temperature program was as follows: 80°C for 3 min, 80 to 180°C at 5°C/min, 180 to 300°C at 20°C/min, 300°C for 10 min. The ion source was run in EI mode (70 eV) at 200°C. The retention times of the E-and Z-isomer of authentic 5-methylthiopentanaldoxime were 14.3 and 14.8 min. The two isomers had identical fragmentation patterns with m/z 130, 129, 113, 82, 61, and 55 as the most prominent peaks. The retention times of the E-and Z-isomer of authentic 6-methylthiohexanaldoxime were 17.1 and 17.6 min. The two isomers had identical fragmentation patterns with m/z 144, 143, 98, 96, 69, 61, and 55 as the most prominent peaks. DL-Dihomomethionine, DL-trihomomethionine, 5-methylthiopentanaldoxime, and 6-methylthiohexanaldoxime were synthesized as described previously (22) and authenticated by NMR spectroscopy.
For kinetic measurements of the conversion of DL-dihomomethionine and DL-trihomomethionine to the respective aldoximes by CYP79F1, enzymatic reactions were made essentially as described above. 5 nmol of phenylacetaldoxime was added to the reaction mixtures as internal standard. Product formation was determined after 0, 5, 10, 20, 40, 70, and 120 min of incubation and quantified based on authentic standards. The reaction mixtures were extracted with a total volume of 600 l of CHCl 3 . The organic phase was collected and evaporated, and the residue was dissolved in 15 l of 50% ethanol and analyzed by LC-MS. LC-MS was done on an HP1100 LC coupled to a Bruker Esquire-LC ion trap mass spectrometer. The LC was performed on a C 18 column (Chrompack Inertsil 3 ODS-3 S15 ϫ 3 COL CP 29126) using mixtures of water (A) and acetonitrile (Fischer, "Far UV grade") (B) as mobile phase at a flow rate of 0.3 ml min Ϫ1 . The elution program was as follows: 30% B (2 min), linear gradient 30 -70% B (28 min), then linear gradient 70 -100% (5 min). The electrospray ionization was done in positive ion mode. Each of the three aldoximes gave nicely separated peaks for the E-and Z-forms. The 210-nm UV trace was detected by the diode array detector and used for quantitation with the genuine software.
Generation of Transgenic Plants-To construct plants, which express the CYP79F1 cDNA under control of the CaMV35S promoter (35S: CYP79F1 plants), the CYP79F1 cDNA was PCR-amplified from the EST ATTS5112 (Arabidopsis Biological Resource Center) using primer 3 (sense direction; 5Ј-AACTGCAGCATGATGAGCTTTACCACATC) and primer 4 (antisense direction; 5Ј-CGGGATCCTTAATGGTGGTGAT-GAGGACGGAACTTTGGATAA). Primer 3 is tailed with a PstI restriction site. Primer 4 introduces four codons coding for His before the stop codon and a BamHI restriction site after the stop codon. The PCR fragment containing the CYP79F1 cDNA was digested with PstI and BamHI, ligated into a PstI/BamHI-digested pBluescript II SK and sequenced to exclude PCR errors. The CYP79F1 cDNA was placed under control of the CaMV35S promoter by ligation into a PstI/BamHIdigested pSP48 (Danisco Biotechnology, Denmark). The expression cassette was excised by XbaI digestion and transferred to pPZP111 (23). Agrobacterium tumefaciens strain C58 (24) transformed with this construct was used for plant transformation by floral dip (25) using 0.005% Silwet L-77 and 5% sucrose in 10 mM MgCl 2 . Seeds were germinated on MS medium supplemented with 50 g ml Ϫ1 kanamycin, 2% sucrose, and 0.9% agar. Transformants were selected after 2 weeks and transferred to soil. Seeds of these plants were harvested, and kanamycinresistant T 2 plants were selected on the same medium. The procedure was repeated to obtain T 3 plants.
HPLC Analysis of the Glucosinolate Content of Plant Extracts-The analysis was performed with tissue harvested from 9-week-old primary transformants and 7-week-old T 3 plants. Simultaneously grown wildtype plants of the same ages were used as controls. The tissue (two to three rosette leaves from each plant) was freeze-dried for 48 h. Glucosinolates were analyzed as desulfoglucosinolates as follows: 3.5 ml of boiling 70% (v/v) methanol were added to 20 mg of freeze-dried material, 75 l of internal standard (0.52 mM p-hydroxybenzylglucosinolate; Bioraf, Denmark) were added, and the sample was incubated in a boiling water bath for 4 min. Plant material was pelleted, and the pellet was re-extracted with 3.5 ml of 70% (v/v) methanol and centrifuged. The supernatants were pooled and analyzed by HPLC after sulfatase treatment as described previously (7). The assignment of peaks was based on retention times, and UV spectra were compared with standard compounds. Glucosinolates were quantified in relation to the internal standard and by use of response factors (8,26).
Analysis of the Amino Acid Content of Plant Extracts-Rosette leaves of 7-week-old plants (250 mg from each plant) were frozen in liquid nitrogen and homogenized using mortar and pestle. The tissue was extracted in 2.5 ml of 80% methanol. The plant material was pelleted (20,000 ϫ g, 10 min) and re-extracted in 2.5 ml of 80% methanol. The methanol phases were combined and dried in vacuo, and the residue was dissolved in 100 l of water. The individual protein amino acids in the sample were identified and quantified on an Ultropac 8 Resin reverse phase HPLC column (200 ϫ 4.6 mm) on a Biochrom 20 amino acid analyzer (Amersham Pharmacia Biotech) according to the manufacturer. The elution program was modified to identify and quantify the chain-elongated homologues of methionine, dihomomethionine, and trihomomethionine. Each plant was analyzed in triplicate.
For quantification of dihomomethionine in the plant material, the sample was subjected to two elution programs. Program 1 was as follows: 53°C for 7 min, buffer A; 50°C for 35 min, buffer A; 95°C for 34 min, buffer A. Program 2 was as follows: 53°C for 7 min, buffer A; 58°C for 12 min, buffer B; 95°C for 25 min, buffer C. Buffer A was 0.2 M sodium citrate, pH 3.25, buffer B was 0.2 M sodium citrate, pH 4.25, and buffer C was 1.2 M sodium citrate, pH 6.25. In program 1, phenylalanine and dihomomethionine coeluted at 63.6 min. In program 2, tyrosine and dihomomethionine coeluted at 25.3 min. Dihomomethionine was quantified as the difference between the peak area corresponding to phenylalanine and dihomomethionine in program 1 and the peak area corresponding to phenylalanine in program 2, and as the difference between the peak area corresponding to tyrosine and dihomomethionine in program 2 and the peak area corresponding to tyrosine in program 1. The response factor for dihomomethionine was determined using an authentic standard.
For quantification of trihomomethionine in the plant material, the sample was subjected to program 3, which was as follows: 53°C for 7 min, buffer A; 58°C for 5 min, buffer B; 95°C for 7 min, buffer B; 95°C for 25 min, buffer C. Trihomomethionine eluted at 29.0 min and was quantified as the peak area using a response factor determined with an authentic standard.
Synthesis of Control RNA-RNA was synthesized from pBluescript II SK vector (Stratagene) linearized by digestion with ScaI. The synthesis reaction was set up in a total volume of 100 l in Transcription Optimized buffer (Promega) supplemented with 500 M rNTPs, 10 mM dithiothreitol, 100 units of RNasin ribonuclease inhibitor (Promega), 3 g of linearized pBluescript II SK, and 50 units of T3 RNA polymerase (Promega). After incubation at 37°C for 2 h, 20 units of RNase-free DNase was added, and the reaction was incubated at 37°C for another 1 h. Following phenol-CHCl 3 extraction and precipitation with ethanol (27), the RNA was dissolved in diethylpyrocarbonate-treated water. The control RNA was used to check for inhibition of RT reactions by components of RNA preparations obtained from different plant tissues.
Expression Analysis by RT-PCR-To study the expression pattern of CYP79F1 in wild-type A. thaliana, the following tissues were investigated: 1) total plant tissue of 4-week-old plants (grown in 8-h light/16-h dark); 2) rosette leaves (without petioles) and 3) above-ground parts of 5-week-old plants (before onset of floral transition; grown at 8-h light/ 16-h dark); 4) rosette leaves (without petioles); and 5) cauline leaves of flowering plants (9 weeks old; grown at 12-h light/12-h dark to induce flowering). To study transcript levels in 35S:CYP79F1 plants, rosette leaves of 7-week-old 35S:CYP79F1 plants and of simultaneously grown 7-week-old wild-type plants were analyzed.
Total RNA was isolated from plant tissue using TRIzol reagent (Life Technologies, Inc.). The RNA was quantified spectrophotometrically and used to synthesize first-strand cDNA. To ensure linearity of the RT-PCR, first-strand cDNA synthesis was performed on 1, 0.

RESULTS
Expression of CYP79F1 in E. coli and Identification of Substrates-Cytochromes P450 of the CYP79 family have previously been shown to be involved in the biosynthesis of cyanogenic glucosides and glucosinolates. CYP79F1 is one of several CYP79 homologues identified in the genome of A. thaliana. The deduced amino acid sequence of CYP79F1 has 88% identity with the deduced amino acid sequence of CYP79F2 and 39 -46% identity with other CYP79 homologues from glucosinolate and cyanogenic glucoside-containing species. A full-length EST of CYP79F1 (ATTS5112) was identified by a data base search. The cDNA obtained from the EST clone was used for expression of CYP79F1 in E. coli using the vector pSP19g10L, which is optimized for expression of cytochromes P450 (18). The CYP79F1 expression construct was transformed into two different E. coli strains, C43(DE3) and JM109. A CO difference spectrum with the characteristic peak at 450 nm was obtained for CYP79F1 expressed in strain C43(DE3), indicating the presence of functional cytochrome P450 (Fig. 1). Based on the peak at 450 nm, the expression level of CYP79F1 in E. coli C43(DE3) was estimated to be 110 nmol of cytochrome P450 (liters of culture) Ϫ1 . A peak at 418 nm was detected in all preparations independent of whether E. coli was transformed with a cytochrome P450 expression construct or the empty vector. The origin of the peak at 418 nm is unknown, but could possibly be derived from endogenous heme proteins (28). The absence of a peak at 450 nm in the CO difference spectrum obtained with a preparation of E. coli strain JM109 harboring the CYP79F1 expression construct indicates low expression level or failure of expression of functional protein. Recombinant CYP79F1 was therefore obtained by use of strain C43(DE3) for all further experiments.
To identify substrates of CYP79F1, activity measurements were carried out using spheroplasts of E. coli C43(DE3) reconstituted with NADPH:cytochrome P450 reductase from S. bicolor. When the reaction mixture containing CYP79F1 was incubated with DL-dihomomethionine, two compounds, which were not present in the control reactions, were detected by GC-MS (Fig. 2). The retention times and the mass spectral fragmentation patterns of these compounds were identical with those of the E/Z-isomers of the authentic standard of 5-methylthiopentanaldoxime. When DL-trihomomethionine was administered to the reaction mixture containing CYP79F1, two compounds with retention times and fragmentation pattern identical to those of the E/Z-isomers of the authentic standard of 6-methylthiohexanaldoxime were detected by GC-MS. The formation of both 5-methylthiopentanaldoxime and 6-methylthiohexanaldoxime was linear with time within 40 min. Product formation combined with cytochrome P450 quantification allow the estimation of turnover number of 0.23 Ϯ 0.01 min Ϫ1 for dihomomethionine and 0.15 Ϯ 0.01 min Ϫ1 for trihomomethionine. No aldoximes were produced using boiled enzyme preparation. Administration of L-methionine, L-phenylalanine, L-tyrosine, and L-tryptophan to the reaction mixtures containing recombinant CYP79F1 did not result in the formation of detectable amounts of the corresponding aldoximes.
Analysis of 35S:CYP79F1 Plants-We have produced transgenic A. thaliana expressing the CYP79F1 cDNA under the control of the CaMV35S promoter to study the effect of altered expression levels of CYP79F1 on the content and composition of glucosinolates. Nine independent primary 35S:CYP79F1 transformants were investigated, and four were selected for analysis through the following two generations. The four primary transformants had dramatically reduced levels of short-chain aliphatic glucosinolates (plants S7 and S9) or slightly increased levels of these glucosinolates (plants S3 and S5) (Fig. 3). The four plants had a morphological phenotype characterized by reduced growth rates, reduced apical dominance, and production of multiple axillary shoots at the time of floral transition resulting in bushy plants. Analysis of plants of the T 2 and T 3 generations showed that the observed phenotype of changed glucosinolate profile and bushy appearance was not stable. However, T 3 plants of the primary transformant S3 (S3.8.1-S3.8.4) had the characteristic bushy phenotype (Fig. 4) and a dramatically reduced content of aliphatic glucosinolates ( Table  I). The effect was very pronounced for the glucosinolates derived from short-chain methionine homologues, i.e. homomethionine and dihomomethionine. The levels of 3-methyl-sulfinylpropylglucosinolate (derived from homomethionine), 4-methylthiobutylglucosinolate and 4-methylsulfinylbutylglucosinolate (both derived from dihomomethionine) were reduced to 9 -13% compared with wild-type. The level of glucosinolates derived from methionine elongated by three to six methylene groups was reduced to about 30 to 50% compared with wildtype. The total content of indole glucosinolates was increased to about the double of the level in wild-type plants. Plants S3.8.2 and S3.8.3 had the most pronounced phenotype and were selected for analysis of their content of the biochemically identified substrates of CYP79F1, dihomomethionine and trihomomethionine (Fig. 5). Both plants accumulated high amounts of the CYP79F1 substrates. Plant S3.8.2 contained as much as 50 times more dihomomethionine and ten times more trihomomethionine than wild-type plants. RT-PCR analysis of plant S3. 8.4 showed that the levels of CYP79F1 and CYP79F2 transcripts were strongly reduced compared with the level in wildtype plants, suggesting that introduction of the transgene had led to cosuppression (Fig. 6). The transcript level of CYP79B2 was slightly increased compared with the level in wild-type plants (Fig. 6).
S3.8.1-S3.8.4 had normal growth rates, but the edges of the leaves were curling upwards (Fig. 4). Before floral transition became apparent, reduced apical dominance resulted in production of multiple axillary shoots, which later developed into lateral influorescences. The plants had reduced fertility and produced only a few normal siliques and many short siliques with no or only few seeds. CYP79F1 Expression Analysis-The level of CYP79F1 transcript was investigated in rosette leaves of plants of different developmental stages and in cauline leaves (Fig. 7). The transcript was detected in all tissues examined. The transcript level increased with maturation of the plants. The expression level was approximately four times higher in rosette leaves of 9-week-old flowering plants than in rosette leaves of 5-week-old plants. When the above-ground parts of 5-week-old plants were analyzed, less CYP79F1 transcript was detected than when only rosette leaves of the same plants were analyzed. This indicates that CYP79F1 is expressed at higher levels in rosette leaves than in petioles. DISCUSSION The CYP79 family comprises multifunctional cytochromes P450 that catalyze two consecutive N-hydroxylations of amino acids followed by dehydration and decarboxylation resulting in the formation of aldoximes (9). CYP79F1 is one of several CYP79 homologues identified in the genome of A. thaliana (available on the Web). In the present paper we report that CYP79F1 is an N-hydroxylase catalyzing the conversion of dihomomethionine and trihomomethionine to 5-methylthiopentanaldoxime and 6-methylthiohexanaldoxime, respectively, a key step in the biosynthesis of aliphatic glucosinolates. CYP79F1 is the first aldoxime-forming enzyme in the biosynthesis of aliphatic glucosinolates to be heterologously expressed and characterized.
Using an EST clone containing the CYP79F1 cDNA we have expressed CYP79F1 in E. coli. The recombinant protein has the spectral characteristics of a cytochrome P450 enzyme. Both dihomomethionine and trihomomethionine are metabolized by CYP79F1 resulting in the formation of 5-methylthiopentanaldoxime and 6-methylthiohexanaldoxime as proven by GC-MS analysis and comparison with authentic standards. Neither methionine nor the other protein amino acids tested are substrates of CYP79F1. Thus CYP79F1 seems to convert specifically chain-elongated methionine homologues to the corresponding aldoximes. Previously, aliphatic amino acids with different chain lengths have been shown to be substrates of the same CYP79 as demonstrated for CYP79D1 and CYP79D2 from cassava (Manihot esculenta Crantz), which converts both valine and isoleucine to the corresponding aldoximes in the biosynthesis of the cyanogenic glucosides linamarin and lotaustralin (10). In contrast, CYP79A2 from A. thaliana has been shown to convert specifically phenylalanine, but not homophenylalanine, to the corresponding aldoxime (7).
Heterologous expression in E. coli of CYP79F1 was accomplished in the strain C43(DE3). Strain C43(DE3) is a mutant E. coli strain, which has been selected for its ability to accommodate high levels of heterologous membrane proteins (19). C43(DE3) has been used successfully for cytochrome P450 expression (14). In the present study, the use of strain C43(DE3) enabled sufficient expression levels for spectral characterization of CYP79F1, whereas this could not be accomplished in the strain JM109.
Two lines of evidence for the involvement of CYP79F1 in glucosinolate biosynthesis are provided by the analysis of transgenic A. thaliana containing the CYP79F1 cDNA under control of the CaMV35S promoter. First, several independent 35S:CYP79F1 transformants have either reduced or increased levels of aliphatic glucosinolates, and reduced CYP79F1 transcript levels in plants of the T 3 generation are accompanied by a dramatically reduced content of aliphatic glucosinolates. Second, the substrates dihomomethionine and trihomomethionine of CYP79F1 accumulate in the plants with reduced content of aliphatic glucosinolates as would be expected upon down-regulation of CYP79F1. The accumulation of the chain-elongated methionines indicates that the enzymes catalyzing the chain elongation of methionine (29) are not subject to feed-back inhibition by the chain-elongated product. Furthermore, it suggests that the enzymes that catalyze additional chain elongation cycles are rate-limiting in the biosynthesis of longer chain methionine homologues.
Comparison of the biochemical data for recombinant CYP79F1 and the glucosinolate profiles of the 35S:CYP79F1 plants raises the question whether CYP79F1 metabolizes not only dihomo-and trihomomethionine, but also other chainelongated methionine homologues produced in A. thaliana. Although the content of all aliphatic glucosinolates is reduced in the transgenic plants compared with wild-type, the effect is most pronounced for the glucosinolates derived from homo-and dihomomethionine. The decrease in the levels of other aliphatic glucosinolates than the dihomo-and trihomomethionine-derived ones might be explained by a broad substrate specificity of CYP79F1 for chain-elongated methionine homologues or by cosuppression not only of the CYP79F1 transcript but also of transcripts of other CYP79 homologues involved in the biosynthesis of aliphatic glucosinolates. As demonstrated by RT-PCR, introduction of the transgene resulted in down-regulation of both CYP79F1 and CYP79F2. CYP79F1 is 88% identical at the amino acid level to CYP79F2, and the two genes are located on the same chromosome, only separated by 1638 bp. This suggests that the two genes have been formed by gene duplication and that they might catalyze similar reactions. Because further investigations of the substrate specificity of CYP79F1 are limited by the unavailability of substrates, isolation of A. thaliana knock-out mutants of CYP79 homologues may facilitate such investigations. From the results of the RT-PCR, it appears that CYP79B2 is up-regulated approximately 2-fold in the 35S: CYP79F1 plants, resulting in an increased content of indole glucosinolates.
The transgenic A. thaliana plants with altered content of aliphatic glucosinolates possess a characteristic morphological phenotype characterized by production of multiple axillary shoots. A. thaliana has been reported to be able to tolerate overexpression of cytochromes P450 of the CYP79 family leading to a 2-to 5-fold increase in glucosinolate content (7,30) without similar changes in the appearance of the plants. There-fore it seems unlikely that the morphological changes result from the presence or absence of specific glucosinolates. The accumulation of very high levels of chain-elongated methionine homologues in the transgenic plants suggests that the morphological phenotype may be a pleiotropic effect caused by disturbance of the plant's sulfur metabolism, in which methionine plays a central role. Repression of cystathionine-␥-synthase, a   key enzyme in methionine biosynthesis, results in plants characterized by formation of a cluster of apical shoots at the time of floral transition and inability to produce flowers (31). 80% of the methionine synthesized by a plant is incorporated into S-adenosyl-methionine (32), which plays a central role in many biosynthetic processes, e.g. methylation reactions such as cytosine methyltransferase-catalyzed DNA methylation (33). One of the morphological changes seen in cytosine methyltransferase antisense plants is production of multiple axillary shoots (34). The onset of the morphological changes in CYP79F1cosuppressed plants at the time of floral transition may be due to the requirement for methionine to support flower development. Alternatively, it coincides with an increase in the level of CYP79F1 expression in wild-type plants. The relation between CYP79F1 down-regulation and the morphological phenotype is the subject of future investigations.
Based on biochemical studies using microsomal enzyme preparations from species of the Brassicaceae, it has previously been proposed that the conversion of dihomo-, trihomo-, and tetrahomomethionine to their corresponding aldoximes is catalyzed by flavin-containing monooxygenases and not by cytochromes P450 (15)(16)(17). However, the conversion of chain-elongated methionine derivatives by flavin-containing monooxygenases was measured by indirect enzyme assays (15), in which the release of CO 2 was used as a measure for enzyme activity. Despite high enzyme activity, the corresponding aldoximes have never been documented in these assays. Our study provides unequivocal evidence that a cytochrome P450 of the CYP79 family, CYP79F1, catalyzes the aldoxime formation in the biosynthesis of the dihomo-and trihomomethioninederived glucosinolates in A. thaliana.
Glucosinolates are related to cyanogenic glucosides, because both groups of natural products are derived from amino acids and have aldoximes as intermediates. Because cyanogenic glucosides are found throughout the plant kingdom and the occurrence of glucosinolates is limited to the order Capparales and the genus Drypetes in the order Euphorbiales, it has been hypothesized that glucosinolates have evolved from cyanogenic glucosides and that homologous enzymes catalyze the common aldoxime-forming step (35,36). Although biochemical data have indicated that other enzyme systems may catalyze this reaction (12), our data provide the final evidence that CYP79 homologues catalyze not only the conversion of aromatic amino acids and tryptophan to their corresponding aldoximes but also the conversion of aliphatic amino acids. This strongly indicates that the evolution of glucosinolates is based on a "cyanogenic predisposition" (35). Furthermore, it suggests that the CYP79 homologues in glucosinolate biosynthesis developed new substrate specificities (e.g. toward tryptophan and chain-elongated methionines) after having diverged away from the CYP79 homologues involved in the biosynthesis of cyanogenic glucosides.
Degradation of 4-methylsulfinylbutylglucosinolate by myrosinase in broccoli sprouts leads to the formation of the isothiocyanate product sulforaphane. As demonstrated in rats, extracts of broccoli sprouts have a pronounced protective effect against breast cancer (4), and sulforaphane has been identified as the principle active agent (3). The identification of a gene involved in the biosynthesis of 4-methylsulfinylbutylglucosinolate is an important step in the development of functional foods that release elevated levels of sulforaphane. 5-Methylthiopentanaldoxime is not only the precursor of 4-methylthiobutylglucosinolate and 4-methylsulfinylbutylglucosinolate but also the precursor of a number of glucosinolates with secondary modifications of the side chain. Besides their occurrence in A. thaliana, such glucosinolates are important constituents of Brassica crops and vegetables. For example, the major glucosinolate in B. napus, the goitrogenic 2-hydroxy-3-butenylglucosinolate, is formed by side-chain modification of 4-methylthiobutylglucosinolate (37). The occurrence of 2-hydroxy-3-butenylglucosinolate in B. napus restricts the use of the protein-rich seed cake as animal feed. Thus availability of biosynthetic genes has great potential for the development of crops with reduced levels of such toxic glucosinolates while retaining glucosinolates with desirable effects, e.g. for pest resistance.
In conclusion, we have shown that CYP79F1 catalyzes the conversion of dihomomethionine and trihomomethionine to 5-methylthiopentanaldoxime and 6-methylthiohexanaldoxime, respectively, in the biosynthesis of aliphatic glucosinolates in A. thaliana. The identification of CYP79F1 provides an important tool for tissue-specific alterations of the level of aliphatic glucosinolates to improve the nutritional value of crop plants and vegetables as well as pest resistance. In addition, the availability of a biosynthetic gene for aliphatic glucosinolates is a valuable means for studying the physiological role of these glucosinolates in plants, e.g. in plant-insect interactions.