Molecular and Enzymatic Characterizations of Novel Bifunctional 3β-Hydroxysteroid Dehydrogenases/C-4 Decarboxylases from Arabidopsis thaliana*

We have isolated two cDNAs from Arabidopsis thaliana encoding bifunctional 3β-hydroxysteroid dehydrogenase/C-4 decarboxylases (3βHSD/D) involved in sterol synthesis, termed At3βHSD/D1 and At3βHSD/D2. Transformation of the yeast ergosterol auxotroph erg26 mutant, which lacks 3βHSD/D activity, with the At3βHSD/D1 isoform or with At3βHSD/D2 isoform containing a C-terminal At3βHSD/D1 endoplasmic reticulum-retrieval sequence restored growth and ergosterol synthesis in erg26. An in vitro enzymatic assay revealed high 3βHSD/D activity for both isoenzymes in the corresponding microsomal extracts. The two At3βHSD/D isoenzymes showed similar substrate specificities that required free 3β-hydroxyl and C-4-carboxyl groups but were quite tolerant in terms of variations of the sterol nucleus and side chain structures. Data obtained with 4α-carboxy-cholest-7-en-3β-ol and its 3α-deuterated analog revealed that 3α-hydrogen-carbon bond cleavage is not the rate-limiting step of the reaction. In planta reduction on the expression of the 3βHSD/D gene as a consequence of VIGS-mediated gene silencing in Nicotiana benthamiana led to a substantial accumulation of 3β-hydroxy-4β,14-dimethyl-5α-ergosta-9β,19-cyclo-24(241)-en-4α-carboxylic acid, consistent with a decrease in 3βHSD/D activity. These two novel oxidative decarboxylases constitute the first molecularly and functionally characterized HSDs from a short chain dehydrogenase/reductase family in plants.

terized the activities of a sterol C-4 methyl oxidase (SMO), 3 a 4␣-carboxysterol-3␤-hydroxysteroid dehydrogenase/C-4 decarboxylase (3␤HSD/D), and an NADPH-dependent 3-oxosteroid reductase from partially purified preparations (4 -6) in order to define the steps involved in C-4 demethylation in plants. The first step is initiated by the SMO, whereby this enzyme converts the C-4␣ methyl group to produce a 4␣-carboxysterol derivative that is subsequently oxidatively decarboxylated by the 3␤HSD/D to produce a C-4-monodemethylated 3-oxosteroid, which is then stereospecifically reduced by the 3-ketoreductase. In contrast to animals and yeast where the SMO is encoded by a single gene (7), we biochemically characterized two distinct microsomal SMOs in Zea mays (4), and we identified two distinct families of SMO genes in Arabidopsis thaliana (8). Until now, the gene(s) coding the single bifunctional protein 3␤HSD/D ( Fig. 1) have not been characterized.
Furthermore, little is known about plant hydroxysteroid dehydrogenases that are either members of the short chain dehydrogenases/reductases (SDR) (9 -11) or the aldo-keto reductase family (12), which include soluble and membranebound HSDs. Identification and characterization of gene(s)encoding plant HSD of the SDR family have not been reported thus far.
To further our knowledge about hydroxysteroid dehydrogenases and the sterol C-4 demethylation multienzymatic complex in plants, in particular, we have identified and characterized at the molecular and enzymatic levels two novel 3␤HSD/D isoforms from A. thaliana. In this study, we present identification and cloning of these 3␤HSD/Ds cDNAs, in vivo heterologous and functional expression in yeast, in vitro enzymological characterization of the recombinant isoenzymes, and in planta down-regulation by virus-induced gene silencing. erg26⌬::TRP1 trp1::hisG ⌬hem1) used in the present study, has been described previously (13). Sterol auxotrophs were grown aerobically at 30°C on solid enriched medium (YPG: 1% yeast extract, 2% peptone, 2% glucose) supplemented with 2% ergosterol or cholesterol dissolved in ethanol/Tween 80 (1:1, v/v) or on minimal medium (YNB: 0.67% yeast nitrogen base, 2% glucose) containing suitable supplements (50 mg/liter each), casamino acids (1 g/liter) and 2% of ergosterol or cholesterol. In the case of liquid medium, the concentration of sterol used was 0.5%. Sterol prototrophic strains were grown aerobically at 30°C on solid or liquid YNB medium containing suitable supplements (50 mg/liter each) or enriched medium (YPG) in the presence of ␦-aminolevulinic acid (50 mg/liter).
The pVT102U (14) S. cerevisiae shuttle vector optimized for expressing recombinant proteins in yeast was used for cloning, sequencing, and transformation of the erg26 strain. This plasmid contains an Escherichia coli origin of replication, a yeast 2-m origin of replication, an E. coli ampicillin resistance gene, and the yeast URA3 gene. It contains an expression cassette, including the alcohol dehydrogenase promotor and terminator. TTO vector is an RNA viral vector and has been described previously (15). It consists of sequences from tobacco mosaic virus strain U1 (TMV-U1) and tomato mosaic virus (fruit necrosis strain F; ToMV-F).
Plants-Nicotiana benthamiana was grown in a greenhouse at 24°C with a 16-h light/8-h dark cycle. A. thaliana plantlets were from the Wassilewskija ecotype.
cDNAs Cloning-We searched the Arabidopsis genome for genes encoding putative orthologs of the yeast ERG26 gene and also belonging to the hydroxysteroid dehydrogenase (HSD) family. Two genes were selected as follows: At1g47290 and At2g26260, and their corresponding cDNAs (At3␤HSD/D1 and At3␤HSD/D2, respectively) were cloned by PCR using reverse-transcribed mRNAs from the Wassilewskija ecotype. The amplified cDNA fragments were cloned into the XbaI and XhoI sites of the pVT102U shuttle vector and placed under the control of the constitutive alcohol dehydrogenase promoter. Because the lack of a C-terminal ER retrieval signal KKXX in At3␤HSD/D2 might result in inefficient localization in the ER yeast C-4 demethylation complex, we replaced the seven terminal amino acids of At3␤HSD/D2 with the KKID sequence to produce ⌬-At3␤HSD/D2, which was also cloned into the pVT102U vector.
Identity Scores and Phylogenetic Analysis-Protein sequences of Arabidopsis 3␤HSD/D and their homologs were identified by BLAST searches and aligned using the ClustalW algorithm. A neighbor-joining tree was constructed based on the sequence alignment and tested further with 1000 bootstraps resampling by using the MacVector package.
Transformations-S. cerevisiae transformations were performed using the lithium acetate procedure as described previously (18). The transformed erg26 yeast strain was plated on minimal YNB medium containing suitable supplements (adenine, 50 g ml Ϫ1 ) without uracil and 2% of ergosterol, as well as on the same minimal YNB medium without uracil but also containing ␦-aminolevulinic acid (50 mg/liter). Cells were grown aerobically at 30°C.
Sterol Analysis-Lyophilized yeast cells (10 -30 mg) were sonicated in the presence of KOH/methanol (6%, w/v) (2 ml) for 10 min and heated in the same medium at 70°C under reflux conditions for 2 h. The mixture was diluted with 1 volume of water, and after acidification to pH 3, total sterols were extracted three times with 3 volumes of ethyl acetate. The extract was dried on Na 2 SO 4 , evaporated to dryness, treated with 500 l of a 0.4 M solution of diazomethane in diethyl ether for 1 h at 0°C and 1 h at room temperature, and then evaporated to dryness. Sterols were analyzed by gas chromatography. GC analysis was carried out with a Varian GC model 8300 (Les Ulis, France) equipped with a flame ionization detector at 300°C, column injector at 250°C, and a fused capillary column (WCOT, 30 m ϫ 0.25 mm inner diameter) coated with DB1(H 2 flow rate of 2 ml/min). The temperature program used included a 30°C/min increase from 60 to 240°C and followed by a 2°C/ min increase from 240 to 280°C. Relative retention times (t R ) are given with respect to cholesterol (t R ϭ 1). Identification of individual sterols was performed using a GC-MS spectrometer (Agilent 5973N) equipped with an "on column" injector and a capillary column (30 m ϫ 0.25 mm inner diameter) coated with DB5. Sterols were unequivocally identified by retention times and an electron impact spectrum identical to that of authentic standards (19).
Lyophilized plantlets of N. benthamiana (0.9 -2.0 g) were homogenized with an Ultra-Turrax homogenizer in the presence of methanol/methylene chloride (2:1, v/v) and heated in the same medium at 70°C under reflux conditions for 3 h. The mixture was filtered, and the extract was evaporated to dryness. Acidic derivatives (R F ϭ 0.0 -0.12), desmethylsterols (R F ϭ 0.27), 4␣-methylsterols (R F ϭ 0.38), 4,4-dimethylsterols (R F ϭ 0.44), and sterol esters (R F ϭ 0.70) were purified twice by TLC on Silica Gel 60F254 plates (Merck), using methylene chloride as developing solvent, and various fractions were eluted from the silica gel. Sterol esters were saponified under standard conditions. The polar fraction (R F ϭ 0 -0.12) was treated with diazomethane at 0°C and subsequently acetylated under standard conditions. The derivatized polar fraction was further purified by TLC on Silica Gel 60F254 plates (Merck), using methylene chloride as developing solvent. The fraction with an R F ϭ 0.07-0.20 and migrating with the standard of 4-carbomethoxysteryl acetate was eluted from the silica gel. A known amount of cholesterol was added as internal standard for GC quantification in the case of the 4␣-methylsterol, 4,4-dimethylsterol, and 4-carboxymethylsteryl acetate fractions and of coprostane in the case of the desmethylsterol fraction. All fractions were then analyzed by GC and GC-MS under the same conditions as described above.
Chemical Details-Melting points are uncorrected. Proton magnetic resonance was monitored in a [ 2 H]chloroform solution with a Brucker 400-or 500-MHz spectrometer. Chemical shifts (␦) (ppm) were determined relative to tetramethylsilane. Coupling constants (J) were in Hertz.
Preparation of Microsomes-Yeast microsomes were prepared as described previously (20). The corresponding 100,000 ϫ g supernatants corresponding to the cytosolic extracts were concentrated 5-8-fold by dialysis over carboxymethylcellulose sodium salt (Fluka) for 16 h at 4°C.
Standard Assay for Recombinant 4␣-Carboxysterol-3-dehydrogenase-C-4 Decarboxylases -Microsomes (0.8 mg of protein) were incubated in the presence of exogenous synthetic 4␣-carboxy-5␣-cholest-7-en-3␤-ol (1) (20 -200 M) emulsified with Tween 80 (final concentration 1.5 g/liter) and NAD ϩ at a 400 M concentration (which is 50-fold the K m value). Incubations were continued aerobically at 30°C with gentle stirring for 30 -45 min. During this period the progression of the reaction was linear. The reaction was stopped by adding 0.5 ml of EtOH. After addition of a known amount of coprostanone (1-4 g) as internal standard, sterols and sterones were extracted from the incubation mixture three times with a total volume of 15 ml of n-hexane, and after drying with Na 2 SO 4 , the extract was concentrated to dryness. Under these conditions, the residual carboxysterol substrate was not extracted. The extract was analyzed by TLC on silica gel eluted with CH 2 Cl 2 (developed twice). The fraction migrating as authentic standards of coprostanone and cholest-7-en-3-one (6) and containing the enzymatically produced cholest-7-en-3-one (R F ϭ 0.50) was eluted and analyzed by GC (DB1) (supplemental Fig. A). The amount of cholest-7-en-3-one (6) produced (t R ϭ 1.074) was calculated by comparison of the integrated peak areas with a known amount of added coprostanone (t R ϭ 1.000), which also allowed the reaction rate of cholest-7-en-3-one to be determined. No endogenous component having the same t R as (6) was present in the inactivated control. The ketone metabolite produced by the reaction was unequivocally identified by its retention time on GC and by an electron impact mass spectrum identical to that of an authentic synthetic standard (Table 1 and supplemental Fig. A). Moreover, in control experiments, GC-MS analysis of the complete extract before TLC analysis indicated the absence of formation of cholest-7-en-3␤-ol during the course of the incubation, confirming that in the absence of exogenously added NADPH to the microsomal extract, the C-4demethylated-3-oxo-derivative was not reduced and further metabolized in accord with the cofactor requirements for this reduction step (6). Under these conditions, the estimated limit of detection of the 3␤HSD/D activity was 0.1-0.2 nmol ϫ h Ϫ1 ⅐mg Ϫ1 . Incubation of the 3␣-deuterated substrate (11) was performed under the same standard conditions. For control experiments, the untransformed substrate (11) was extracted from the incubation mixture with ethyl acetate and methylated with diazomethane. GC-MS analysis of the 4-carbomethoxy derivative revealed no lost of deuterium, thus excluding a possible washout of the deuterium label at C-3 during the incubation procedure.
Apparent maximum velocity V and V/K values were determined by fitting the data to the Michaelis-Menten equation using the nonlinear regression program DNRP-EASY derived by Duggleby which corresponds to Ref. 21. Computer-assisted linear regression analysis gave correlation coefficients greater than 0.98 (n ϭ 5-6). The primary deuterium kinetic isotope effects were defined as D V ϭ H V max / D V max and D (V/K ) ϭ H (V max /K m )/ D (V max /K m ) according to the conventions of Northrop (22,23).
Incubation of 4␣-Carboxysterol Analogs and Identification of Enzyme-generated Products-The apparent K m and V max of analogs (1, 5, and 2) (Tables 3 and 4) were determined by incubating them for 30 -45 min at 30°C under standard assay conditions in a microsomal preparation containing either 3␤HSD/D1 or ⌬3␤HSD/D2. The concentration of substrate was 20 -150 M, and the concentration of NAD ϩ was 0.4 mM (50-fold the measured K m value). As indicated previously, in the absence of NADPH addition to the microsomal extract, the expected C-4-demethylated 3-oxo-derivative metabolites were  (33); 299(66) 4␣-Carboxy-stigmasta-7,24(24 1 )-dien-3␤-ol (4) Stigmasta-7,24(24 1 )-dien-3-one (9) 1.306 M ϩ ϭ 410 (3); 312 (20); 297(10); 260(100); 229 (8) not further metabolized. In the case of substrates (3 and 4) for which apparent K m and V max values could not be determined because of difficulties in obtaining sufficient quantities of these two substrates, the reaction rates of (3 and 4) were compared with those of (1, 2, and 5) in microsomal extracts containing either 3␤HSD/D1 or ⌬3␤HSD/D2. Low substrate concentrations (40 M) were used to determine enzyme specificity; the concentration of NAD ϩ was 0.4 mM, and the reaction rates measured for (2, 3, 4, and 5) were normalized to the rate obtained with the standard substrate (1) (taken as 100%, Fig. 4,  A and B). In the case of the carboxysterols (2, 3, 4, and 5) after addition of coprostanone as internal standard, extraction, and TLC analysis, the fraction migrating between a standard of 4-demethyl-sterone (R F ϭ 0.50) and 4,4-dimethyl-sterone (R F ϭ 0.70) was eluted and analyzed by GC and GC-MS ( Table  1). The 3␤HSD/D products were unequivocally identified by their retention time and an electron impact spectrum identical to that of authentic standards ( Table 1). The data were compared with those obtained in the case of a control where the microsomes were inactivated. In addition, in the case of incubation of radiolabeled carboxysterols (2, 3, and 4), the newly labeled sterone was directly visualized by radioscanning of the TLC plate, in accord with the migration of the corresponding nonlabeled standards. The areas of the GC peak of coprostanone and of the product peak formed, corrected from endogenous components of the same t R (if present) determined in the corresponding control, allowed the rate of transformation of these analogs to be measured. Incubations of labeled sterols (12, 13, and 14) and corresponding control incubation with inactivated enzyme were extracted according to the same procedure. Radioscanning of the TLC analysis indicated the complete recovery of the added substrate and absence of metabolic alteration of these compounds by any of the two recombinant At3␤HSD/Ds microsomal preparations. In the case of the carboxymethylated analog (10), the hexane extracts of the incubation and control samples were directly analyzed by GC and gave identical profiles, indicating the absence of detectable At3␤HSD/D activity with this analog. Similarly, the extracts of incubations of 3␤-hydroxysteroids (16, 17, 18, and 19) and ursolic acid (15) and corresponding controls were directly analyzed by GC and GC-MS. Comparison to the corresponding controls performed with inactivated microsomes revealed the absence of dehydrogenated product and complete recovery of these compounds, which was confirmed by ion monitoring that corresponded to the masses of the substrate and the expected C-3 dehydrogenated product.
Miscellaneous-Membrane protein was determined as described by Bradford (24).

RESULTS
Cloning of 3␤HSD/D in Arabidopsis-We searched the Arabidopsis genome for genes encoding putative orthologs of the yeast ERG26 gene and belonging to the HSD family. Among the best candidates were At1g47290 and At2g26260, two genes that shared the highest sequence identity with the yeast 3␤HSD/D protein Erg26p (30 and 29% respectively). Corresponding cDNAs (named At3␤HSD/D1 for A. thaliana 3␤hydroxy-steroid dehydrogenase/decarboxylase isoform 1 and At3␤HSD/D2, respectively) were cloned by PCR using reversetranscribed mRNAs from the Wassilewskija ecotype. Sequencing of At3␤HSD/D1 (AY957470) indicated an open reading frame of 1143 bp with two nucleotide changes from the At1g47290 gene transcript (ecotype Columbia, NM_179448) leading to two amino acid changes. Sequencing of At3␤HSD/D2 (DQ302749) indicated an open reading frame of 1173 bp that was identical with the annotated At2g26260 gene transcript (NM_128183.1). While this study was in progress, an additional possible splicing site of the At2g26260 gene leading to a 3Ј-extended sequence (1695 bp) was annotated (NM_128183.2); however, we were not able to amplify the corresponding cDNA from our RNAs preparation.
Sequences Analysis-At3␤HSD/D1 (381 amino acids) and At3␤HSD/D2 (391 amino acids) proteins share 80% identity (supplemental Fig. B). A sequence comparison analysis revealed significant homology between the two At3␤HSD/Ds to the 3␤-hydroxysteroid dehydrogenase/isomerase protein family (Pfam 01073), a member of the SDR superfamily (FAD/ NAD(P)-binding Rossmann fold superfamily clan) (9 -11, 25, 26). However, there is little sequence identity (only 15-30%) between different SDR enzymes. The two sequences include N-terminal conserved glycine and aspartic residues, TGGXGXXAX 18 D, which are required to form the coenzymebinding site. In addition, the At3␤HSD/D1 and At3␤HSD/D2 proteins possess the YX 3 K motif ( 159 YX 3 K 163 for At3␤HSD/D1) conserved in the active site of most of the members of the SDR family (10) (supplemental Fig. B). Finally a conserved serine residue, found at position 130 or 131 for At3␤HSD/D1, might constitute the third residue of the catalytic triad proposed for a variety of members of the SDR family (9,10,27). In the At3␤HSD/D1 and At3␤HSD/D2 proteins, there is a single putative membrane spanning domain near the C terminus (supplemental Fig. B). Finally, At3␤HSD/D1 possesses a C-terminal ER retrieval signal KKXX, which is absent in At3␤HSD/D2.
Phylogenetic Analysis of the At3␤HSD/D-A molecular phylogenetic tree of the amino acid sequences of a variety of characterized HSD proteins from different organisms related to the putative At3␤HSD/Ds was designed (Fig. 3). The protein sequences of At3␤HSD/D1 and At3␤HSD/D2 show 30% identity with the yeast ERG26 protein (13) and 37% identity with the NAD(P)H steroid dehydrogenase-like protein from animals (28). These clones show 22% identity with the human 3␤-hydroxysteroid dehydrogenase/isomerase (29 -30) and 26% identity with the cholesterol dehydrogenase from Nocardia (31), all enzymes catalyzing dehydrogenation of 3␤-hydroxysteroids. The similarity of the two At3␤HSD/Ds to Erg26p and Nsdhl is higher than with any other known 3␤HSD. Together with the present 3␤HSD/Ds from plant, these orthologs from yeasts and animals form a differentiated cluster distinct from other HSDs, particularly those catalyzing dehydrogenations at positions other than the 3␤ of the steroid nucleus or displaying an opposite stereochemistry for C-3 hydride abstraction. These data are consistent with the suggestion that At3␤HSD/D1 and At3␤HSD/D2 code two plant 3␤-hydroxysteroid dehydrogenase/C-4 decarboxylases isoenzymes.
At3␤HSD/D1, and ⌬-At3␤HSD/D2 Containing the At3␤HSD/D1 ER Retrieval Signal, Can Complement a Yeast Strain Deficient in Erg26p-To further characterize the function of At3␤HSD/D1 and At3␤HSD/D2, we performed a yeast complementation assay in the ERG26-deficient strain SDG200 deficient in 3␤HSD/D activity (13). In S. cerevisiae, the ERG26 product is an essential enzyme because of the fact that disruption of ERG26 is lethal, and the erg26 strain requires ergosterol or cholesterol supplementation for viability (13). Despite one putative membrane-embedded domain, the lack of the C-terminal ER retrieval signal KKXX in At3␤HSD/D2 might result in inefficient localization in the ER yeast C-4 demethylation complex. Thus, we replaced the seven terminal amino acids of At3␤HSD/D2 with the KKID sequence to produce ⌬-At3␤HSD/D2 (387 amino acids). The At3␤HSD/D1 and ⌬-At3␤HSD/D2 cDNAs were cloned into the pVT102U shuttle vector under the control of the constitutive alcohol dehydrogenase promoter.
At3␤HSD/D1 and ⌬-At3␤HSD/D2 Have 3␤-Hydroxysteroid Dehydrogenase/C-4 Decarboxylases Activity in Vitro-We next performed an enzymatic assay to test whether the two recombinant Arabidopsis putative 3␤HSD/D proteins in the transformed erg26 strain indeed possess 3-hydroxysteroid dehydrogenase/C-4 decarboxylase activity. Overexpression and purification of plant membrane proteins for functional analysis are still relatively unexplored fields for yeast or bacteria with little documentation in the literature. Additionally, in the case of an enzyme that is part of a membrane-bound multienzymatic complex, interactions with other components of the complex may be necessary for optimum enzymatic activity. Thus, 3␤HSD/D activity was assayed in the microsomal extracts and corresponding cytosolic fractions prepared from erg26-pVT-VOID, erg26-pVT-At3␤HSD/D1, and erg26-pVT-⌬-At3␤HSD/D2 by using the standard assay conditions for recombinant 3␤HSD/D described under "Experimental Procedures." The results from these studies revealed that microsomal extracts obtained from erg26-pVT-At3␤HSD/D1 and erg26-pVT-⌬-At3␤HSD/D2 were able to oxidatively decarboxylate the 3␤HSD/D substrate, 4␣-carboxycholest-7-en-3␤-ol (1), with a high efficiency in the presence of NAD ϩ , to produce a single 3-keto-4-decarboxylated metabolite, cholest-7-en-3-one (6), which was unequivocally identified by GC-MS analysis (Table 1 and supplemental Fig. A). 3␤HSD/D activity was undetectable in reactions with microsomal extracts of erg26/pVT-VOID. All corresponding concentrated cytosolic extracts were also inactive. The observed high catalytic competence of At3␤HSD/D1 and ⌬-At3␤HSD/D2 revealed that they indeed encode two membrane-bound plant 3␤HSD/Ds.
Determination of the Apparent Kinetic Parameters of At3␤HSD/D1 and ⌬-At3␤HSD/D2 for 4␣-Carboxy-cholest-7en-3␤-ol (1) and Its 3␣-Deuterated Analog (11)-In two separate series of experiments, the apparent kinetic parameters of each 3␤HSD/D with substrate 4␣-carboxy-cholest-7-en-3␤-ol (1) were determined by varying the concentration of (1) by using a constant 50-fold K m concentration of NAD ϩ . However, the V max for NAD ϩ could not be measured due to the inability of using saturating concentrations of carboxysterol (1), because it has limited solubility in aqueous media. Under our standard assay conditions, the velocity/ substrate concentration curves obey simple Michaelis-Menten kinetics with respect to (1) and NAD ϩ cofactor (supplemental Fig.  C). The obtained kinetics data for 3␤HSD/D1 and -2 are summarized in Tables 3 and 4, respectively.
To gain insight into the enzyme mechanism of the two 3␤HSD/Ds, we compared V max and V/K values measured for 3␣-protonated 4␣-carboxy-cholest-7-en-3␤-ol (1) with those obtained with its synthetic 3␣-deuterated analog (11) in the same microsomal preparation. We used a noncompetitive assay, in which the protonated and deuterated substrates are measured separately. The advantage of the direct comparison method is that it is the only means to determine the primary kinetic deuterium isotope effect on V max ( D V). For both At3␤HSD/D1 and ⌬-At3␤HSD/D2, the measured primary

TABLE 3 Apparent kinetics parameters for selected substrates of recombinant Arabidopsis 3␤HSD/D1
For each experiment, kinetic parameters of the different substrates were measured in the same microsomal preparation.

TABLE 4 Apparent kinetics parameters for selected substrates of recombinant Arabidopsis ⌬3␤HSD/D2
For each experiment, kinetic parameters of the different substrates were measured in the same microsomal preparation. deuterium kinetic isotopic effect for (11) was near unity, both for V and V/K, within experimental error (Tables 3 and 4 and supplemental Fig. D). Moreover, recovery and GC-MS analysis of the untransformed substrate (11) revealed no lost of deuterium, thus excluding any washout of the deuterium label at C-3 during the incubation procedure.

Substrate
Substrate Screening of the At3␤HSD/Ds-We used a series of natural or synthetic substrate analogs to determine the structural requirements of the 3␤HSD/Ds. The steroid substrate screen used our standard 3␤HSD/D assay conditions (see "Experimental Procedures") at saturating levels of NAD ϩ cofactor (50-fold the K m value). The assay conditions were complemented by analysis of products and residual substrates using GC-MS.
First, a series of 3␤-hydroxy-4␣-carboxysterols with distinct nucleus or side chain structures were assayed with 3␤HSD/D1. These included (3) and (4), the respective products of the two plant sterol C-4-methyl oxidases (SMO) (4), and thus the most probable substrates for decarboxylation in plants, and (2) the physiological substrate of the yeast 3␤HSD/D, which accumulates in the erg26 mutant (13,32), was also assayed. Because of the difficulty of obtaining sufficient quantities of (3) and (4) that could only be produced enzymatically in vitro, we could not determine the kinetic constants of these two analogs but were able to compare their reaction rates, in the same 3␤HSD/D1 enzymatic preparation, with that of the other substrates used at the same concentration. The data indicate that all four 3␤-hydroxy-4-carboxysterols (1-4) were dehydrogenated and decarboxylated by At3␤HSD/D1 (Fig. 4A and Table 3) and that this isoenzyme did not show a marked preference for any of these four 3␤-hydroxy-4-carboxysterols. The corresponding enzy-matic 3-oxo-C-4-decarboxylated products of reaction were unequivocally identified as (6 -9), respectively, by GC-MS analysis (Table 1).
To probe the stereoselectivity of the plant 3␤HSD/D1 for the 3␤-hydroxy group of the substrate, the epimer of (1) with the 3␣-configuration (5) was synthesized. It was dehydrogenated by the recombinant 3␤HSD/D1, but with a 50-fold lower activity, to also produce (6), which was unequivocally identified by GC-MS (Table 3 and Fig. 4A).
Next, we examined the requirement for the C-4-carboxylic substituent. Ursolic acid (15), a triterpene derivative possessing a 3␤-hydroxyl group and a distal C-17-carboxyl function, was not metabolized. Converting the carboxyl group in (1) to generate the C-4-carboxymethylated derivative (10), or substitution of the carboxyl group by a methyl group as in the 4,4-dimethylsterols (12) and (13) or the 4␣-methylsterol (14), totally abolished the dehydrogenase activity of 3␤HSD/D1 in the presence of NAD ϩ cofactor (Fig. 4A). Indeed, following incubation of cold or radioactive labeled samples of these analogs, TLC and GC-MS analysis of the reaction excluded the formation of the corresponding 3-oxo products. Moreover, by using a series of 3␤-hydroxysteroids, pregnenolone (16), trans-androstenone (17), 3␤-androstanediol (18), and dehydroepiandrosterone (19), which are known substrates of a number of animal 3␤HSDs, we could not detect any oxidative conversion of the 3␤-OH by the plant 3␤HSD/D1.
Silencing of Endogenous N. benthamiana 3␤HSD/D-In order to learn more about the function of the 3␤HSD/D, a cDNA fragment of 500 bp termed Nb3␤HSD/D, homologous to the N terminus of At3␤HSD/D1 and At3␤HSD/D2 (or ⌬-At3␤HSD/D2), was isolated in N. benthamiana and cloned into the viral TTO vector. This sequence was placed under the control of the tobacco mosaic virus (TMV-U1) coat protein subgenomic promoter in the antisense orientation. Nb3␤HSD/D shares 59 -57% identity with At3␤HSD/D1 and At3␤HSD/D2 and thus falls clearly within the 3␤HSD/D family ( Fig. 3 and supplemental Fig. B).
N. benthamiana plants were inoculated with infectious TTO-Nb3␤HSD/D mRNAs (Fig. 5C), and noninfected plants (Fig. 5A) as well as plants infected with the previously described TTO-DXR (Fig. 5B) construct (17) were used as controls. Silencing of Nb3␤HSD/D strongly reduced the growth of young leaves (Fig. 5C) compared with uninfected plants and plants infected with TTO-NbDXR (Fig. 5,  A and B). To measure the VIGS of 3␤HSD/D, semiquantitive RT-PCR analysis was performed using 3␤HSD/D primers annealing outside of the region used for the silencing. We observed a 70% reduction of the transcripts in Nb3␤HSD/D silenced plants compared with uninfected plants and the TTO-NbDXR control (Fig. 5D).
Approximately 3 weeks after infection, total lipids from young plants, containing sterols and expected carboxysterol derivatives, were extracted and purified by TLC allowing the separation of steryl esters, 4,4-dimethyl-, 4␣-methyl-, 4-demethylsterols, and a more polar fraction which was treated with diazomethane to stabilize and decrease the polarity of the carboxysterol derivatives. The sterols were quantified by GC and compared with authentic standards by GC-MS electron impact mass spectra.
Results indicate that the most noteworthy difference between the control and the TTO-Nb3␤HSD/D infected plants appeared to be the addition of a novel sterol derivative that represents 2-3% of the total sterol content and is not detectable in extracts from control plants (Fig. 6). This compound was identified as 4␣-carboxy-4␤,14␣-dimethyl-9␤, 19-cyclo-ergost-24(24 1 )en-3␤-ol (3) by GC-MS analysis after derivatization and comparison with the MS spectrum of an authentic enzymatically produced sample (4). In addition, a 2-fold accumulation of 4,4-dimethylsterols was observed in the plants infected with TTO-Nb3␤HSD/D, whereas the bulk of the sterol profile remained unchanged within the limit of experimental error (data not shown).

DISCUSSION
The SDR family includes more than 2000 annotated enzymes that contain a 250 -350-amino acid residue core structure, catalyzing NAD(P)(H)-dependent oxidoreductions of great functional diversity (9 -11, 33). Despite a sequence identity level of only 15-30% between different SDR members, the three-dimensional folds are quite similar, except for the C-terminal regions. According to the Sanger Institute data base, most of the 3␤-hydroxysteroid dehydrogenases show a common 3␤HSD architecture (Pfam 01073) found in the present At3␤HSD/D isoforms, and a small number (5 of 198) possess an additional C-terminal 200-amino acid residue segment homologous to the reticulon containing two large hydrophobic regions that are likely embedded in the ER membrane (34,35). Whereas the 3␤HSD/D1 cDNA corresponding to the sticky substrate that reacts to give products as fast or faster than it dissociates from the enzyme, i.e. the enzyme-substrate complex has a high commitment to catalysis (47,48,50).
Extensive studies using multiple isotope effects have been carried out with a number of enzymes catalyzing oxidative decarboxylations with NADP ϩ as cosubstrate, exemplified by the malic enzyme (51)(52)(53) or the 6-phosphogluconate dehydrogenase (54). In these cases, a stepwise mechanism was shown, and a small isotope effect for the hydride transfer step was measured, comparable with that found with the 3␤HSD/Ds studied here. However, measurement of a single isotope effect is not sufficient to elucidate the mechanism of the 3␤HSD/Ds, although it would seem logical that 3C-H bond cleavage and decarboxylation are separate events. Determination of the mechanism of the 3␤HSD/Ds will require further studies, including in particular the 13 C isotope effect for C-C bond cleavage (53,54), and variation with pH of primary isotope effects (47,(55)(56)(57).
To complete our catalytic analysis, we studied the substrate requirements of each of the two 3␤HSD/Ds in a separate series of experiments incubating various substrate analogs with enzymatic preparations. For both 3␤HSD/Ds isoenzymes, the data showed a substrate specificity that required free 3␤-hydroxyl and C-4 carboxyl groups indicating that both recombinant At3␤HSD/Ds are highly specific for 3␤-hydroxy-4␣-carboxysterols. In contrast, substrate requirements were quite tolerant in terms of variations of the sterol nucleus and side chain structures. Remarkably, the data did not reveal for either of the two 3␤HSD/Ds a clear preference between substrates possessing one or two substituents at C-4. In addition, the substrate preferences of the two 3␤HSD/Ds were found to be very similar. This is in contrast with the strict and distinct structural requirements observed for the two distinct SMOs metabolizing 4,4-dimethyl-and 4␣-methylsterols, respectively, to produce the corresponding carboxysterols that are the physiological substrates of the present 3␤HSD/Ds (4). Furthermore, (2), the typical substrate of the yeast 3␤HSD/D, was efficiently oxidized by both At3␤HSD/Ds, whereas 4,4-dimethyl-zymosterol (12), the substrate of the yeast SMO, was not metabolized by any of the plant SMOs (20). These data suggest that a single type of 3␤HSD/D is metabolizing the products of the two distinct SMOs inside the plant C-4 demethylation complex.
The VIGS experiments carried out in this study led to young N. benthamiana plants exhibiting a clear biochemical phenotype. As this biochemical phenotype was observed in several distinct infected plants but not in plants infected with the TTO-DXR construct or noninfected plants, it can clearly be ascribed to the presence of the Nb3␤HSD/D cDNA. These biochemical changes obtained after infection with TTO-Nb3␤HSD/D, correlating with the specific reductions in Nb3␤HSD/D mRNA levels, confirm the genetic silencing of the corresponding Nb3␤HSD/D endogene.
Silencing of Nb3␤HSD/D resulted in a substantial accumulation of a novel 4␣-carboxysterol, 4␣-carboxy-4␤,14␣-dimethyl-9␤,19-cyclo-ergost-24(24 1 )-en-3␤-ol (3), involved in the pathway for removing the first methyl group at C-4, thereby confirming in planta the function of the present 3␤HSD/Ds. Recent studies in A. thaliana indicate that plants compromised early in the sterol biosynthetic pathway, i.e. upstream of the removal of the second C-4 methyl group, such as the fackel mutant defective in C-14 reductase (58,59), show severe defects in development and embryogenesis, whereas those affected later in the pathway, such as dwf7/ste1 (60) or dwf1/ DIM (61), do not. Thus, in the event of an impaired expression of a gene upstream of the second C-4 demethylation step, plants might develop regulatory processes to reduce these defects, thereby limiting as in the present case the accumulation of (3). In these lines, a limited accumulation of the substrates of the SMOs in SMO-silenced plants was also observed previously by using the same VIGS approach in N. benthamiana (8). In this respect, 3␤HSD/D-silenced plants, albeit accumulating modest amounts of (3), present substantial developmental alterations that will need further work to be precisely characterized.
The present combination of molecular and biological chemical approaches allowed a thorough identification and functional characterization of 3␤HSD/D as one further step in the completion of the molecular inventory of sterols synthesis in higher plants. Genetic redundancy or lethality may account for the failure to isolate plant 3␤HSD/D genes by screening of phenotypes by using a genetic approach. Given that plant mutants affected in 3␤HSD/D have not been reported thus far, the present study provides important clues for the physiological roles of C-4-substituted sterols in photosynthetic eukaryotes. Finally, these two novel oxidative decarboxylases constitute the first plant hydroxysteroid dehydrogenase genes of the SDR superfamily to be molecularly and enzymatically characterized.