cDNA Cloning and Characterization of a New Human Microsomal NAD+-dependent Dehydrogenase that Oxidizes All-trans-retinol and 3α-Hydroxysteroids*

We report the cDNA sequence and catalytic properties of a new member of the short chain dehydrogenase/reductase superfamily. The 1134-base pair cDNA isolated from the human liver cDNA library encodes a 317-amino acid protein, retinol dehydrogenase 4 (RoDH-4), which exhibits the strongest similarity with rat all-trans-retinol dehydrogenases RoDH-1, RoDH-2, and RoDH-3, and mouse cis-retinol/androgen dehydrogenase (≤73% identity). The mRNA for RoDH-4 is abundant in adult liver, where it is translated into RoDH-4 protein, which is associated with microsomal membranes, as evidenced by Western blot analysis. Significant amounts of RoDH-4 message are detected in fetal liver and lung. Recombinant RoDH-4, expressed in microsomes of Sf9 insect cells using BacoluGold Baculovirus system, oxidizes all-trans-retinol and 13-cis-retinol to corresponding aldehydes and oxidizes the 3α-hydroxysteroids androstane-diol and androsterone to dihydrotestosterone and androstanedione, respectively. NAD+ and NADH are the preferred cofactors, with apparent K m values 250–1500 times lower than those for NADP+ and NADPH. All-trans-retinol and 13-cis-retinol inhibit RoDH-4 catalyzed oxidation of androsterone with apparentK i values of 5.8 and 3.5 μm, respectively. All-trans-retinol bound to cellular retinol-binding protein (type I) exhibits a similarK i value of 3.6 μm. Unliganded cellular retinol-binding protein has no effect on RoDH activity. Citral and acyclic isoprenoids also act as inhibitors of RoDH-4 activity. Ethanol is not inhibitory. Thus, we have identified and characterized a sterol/retinol-oxidizing short chain dehydrogenase/reductase that prefers NAD+ and recognizes all-trans-retinol as substrate. RoDH-4 can potentially contribute to the biosynthesis of two powerful modulators of gene expression: retinoic acid from retinol and dihydrotestosterone from 3α-androstane-diol.

Short chain alcohol dehydrogenases/reductases are either cytosolic or membrane-bound enzymes with a subunit molecular mass of 25-35 kDa that utilize a vast variety of substrates, including steroids and prostaglandins (1). Recently, this family of enzymes has expanded to include the retinol-oxidizing dehydrogenases (2)(3)(4)(5)(6). Retinol dehydrogenases are involved in the biosynthesis of all-trans-retinoic acid, the activating ligand for a family of nuclear receptors (7). All-trans-retinoic acid is produced from all-trans-retinol in two oxidative steps: all-transretinol is oxidized to all-trans-retinal and then further to alltrans-retinoic acid. Retinol dehydrogenases catalyze the ratelimiting step: the oxidation of retinol to retinaldehyde (8).
Although the effects of retinoic acid on gene transcription and regulation have been intensively studied during the last decade, the exact enzymes that synthesize this morphogen and the mechanisms that regulate its production in tissues are not fully understood. Enzymatic activity capable of oxidizing retinol to retinaldehyde is readily detected in the cytosolic and microsomal fractions of total cell homogenates (9). The cytosolic activity has been linked to the NAD ϩ -dependent medium-chain alcohol dehydrogenases (ADHs), 1 which, in addition to trans and cis forms of retinol, oxidize a variety of aliphatic and a number of cyclic alcohols (10). In the cells, most retinol is bound to the cellular retinol-binding protein (CRBP) (11). ADHs cannot oxidize the bound form of retinol. 2 The first enzyme purified by following its ability to oxidize CRBP-bound retinol, RoDH-1, turned out to be a microsomal NADP ϩ -dependent short chain dehydrogenase/reductase (3). The cDNA for RoDH-1 was initially isolated from a rat liver cDNA library, and later on, its mouse homolog was cloned and found to share 98% amino acid sequence identity with the rat enzyme (6). Two more closely related enzymes were subsequently cloned, RoDH-2 (4), with 82% sequence identity to RoDH-1, and RoDH-3 (12), with 95% sequence identity, establishing a multigene family of all-trans-retinol dehydrogenases.
The physiological role of these NADP ϩ -dependent enzymes in retinoic acid biosynthesis in vivo has been questioned because of the cellular ratios of the reduced and oxidized forms of NADP ϩ (13). In the liver cytosol, and presumably other tissues, the NADP ϩ /NADPH ratio is about 0.01, whereas the NAD ϩ / NADH ratio is about 1000 (14), suggesting that enzymes that prefer NADP ϩ will function in the reductive rather than oxidative direction.
We became interested in finding an isoenzyme of RoDH that would function efficiently in the oxidative direction (i.e. prefer NAD ϩ as cofactor) and would recognize all-trans-retinol as substrate, because it was clear that RoDH exists in multiple isoenzymic forms. Here, we report a cDNA sequence and catalytic properties of a new human short chain dehydrogenase/ reductase that shares more than 70% sequence identity with rat RoDHs, recognizes all-trans-retinol, and prefers NAD ϩ over NADP ϩ .

MATERIALS AND METHODS
cDNA Cloning-A human liver gt10 cDNA library (CLONTECH Inc., Palo Alto, CA) was screened with the [␣-32 P]dATP-labeled coding region of rat RoDH-1 prepared by PCR amplification of the rat liver mRNA with gene-specific primers designed according to a published sequence (3). The hybridization conditions were as follows: 25% formamide, 5ϫ Denhardt's solution (0.1% bovine serum albumin, 0.1% polyvinylpyrrolidone, 0.1% Ficoll 400), 5ϫ saline-sodium-phosphate-EDTA, 0.1 mg/ml salmon sperm DNA, and 0.1% SDS at 42°C overnight. After hybridization, the Nytran filters were washed several times in 6ϫ SSC (150 mM sodium chloride, 15 mM sodium citrate, pH 7.5), 0.1% SDS at room temperature, and the final wash was performed in 0.2ϫ SSC, 0.1% SDS. Positive recombinant phage plaques were purified, and phage DNA was isolated using a Qiagen Lambda kit (Qiagen, Chatsworth, CA). The DNA insert was obtained by digestion with EcoRI and subcloned into a M13mp19RF digested with EcoRI. Sense and antisense single-stranded M13 DNAs were each sequenced at least twice.
Northern Blot Analysis-The cDNA probe was prepared by EcoRI digestion of the cDNA clone in M13 vector and purified by electrophoresis in 1% agarose gel. The human adult and fetal multiple tissue Northern blots (CLONTECH) were hybridized with 32 P-labeled cDNA probe in ExpressHyb hybridization solution according to the manufacturer's instructions (CLONTECH). Briefly, the blots were prehybridized in ExpressHyb solution for 30 min at 68°C and transferred to a fresh solution containing 2 ϫ10 6 cpm/ml of the denatured radiolabeled cDNA. The hybridization was performed at 68°C for 1 h. The blots were rinsed in 2ϫ SSC, 0.05% SDS several times at room temperature and washed in 0.1ϫ SSC, 0.1% SDS for 40 min at 50°C. The blots were exposed to x-ray film at Ϫ70°C with two intensifying screens for 1 week.
Expression of RoDH-4 in Escherichia coli-The N-terminal fragment of RoDH-4 cDNA in M13 vector was amplified by PCR using primers Eco5Ј (sense, CGG GAA TTC CAG GTG CTG AGC CAC CTG; nucleotides at position 200 -217) and Hin3Ј (antisense, AGA AAG CTT TGT CTC TCA CGC ACT CC; nucleotides at position 430 -446). The primers carried recognition sites for restriction endonucleases EcoRI and Hin-dIII, respectively (underlined). The C-terminal fragment of RoDH-4 cDNA was amplified using primers Hin5Ј (sense, CCA AAG CTT GTG TGG TCA ACG TCT CCA GT; nucleotides at position 609 -628) and Xho3Ј (antisense, AGA CTC GAG TGG CAT CCC AGC CAG CTG; nucleotides at position 981-998), which carried recognition sites for restriction endonucleases HindIII and XhoI, respectively (underlined). Both PCRs were heated for 5 min at 94°C and cooled to 72°C, 2 l of Pfu polymerase (Stratagene) were added, and 30 cycles were run as follows: denaturing at 94°C for 45 s, annealing at 52°C for 45 s, and extension at 72°C for 6 min. The PCR fragments were purified by electrophoresis in 1% agarose gel and subcloned into pET32a vector (Novagen) digested with EcoRI/HindIII and HindIII/XhoI, respectively, using Rapid Ligation kit (Boehringer Mannheim, Indianapolis, IN). Competent BL21 E. coli cells were transfected with ligation mixtures and spread onto TY hard agar plates containing 200 g/ml ampicillin. The pET32a vectors that contained inserts were sequenced to verify the sequence of the inserts and were transfected into BL21(DE3) cells. The expression of recombinant proteins in BL21(DE3) cells was induced by 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside at A 600 of 0.8 -1.0. The cultures were incubated for 24 h at 30°C. Cell pellets were resuspended in ice-cold 1ϫ binding buffer (5 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9) and homogenized twice using a French press. The homogenate was centrifuged at 20,000 ϫ g for 15 min, the supernatant was discarded, and the pellet was resuspended in 1ϫ binding buffer, sonicated, and centrifuged to remove all soluble proteins. The pellet was resuspended in the 1ϫ binding buffer supplemented with 6 M urea and incubated on ice for 1 h shaking. The solubilized proteins were loaded onto a His Bind Resin charged with NiSO 4 at room temperature and equilibrated with 1ϫ binding buffer plus 6 M urea. Protein was eluted with 1ϫ elute buffer plus 6 M urea. From a 1-liter culture, 5.5 mg of the N-terminal fragment and 3.5 mg of the C-terminal fragment were obtained. The purified proteins were dialyzed against 10 mM Tris-HCl, pH 7.4, to remove urea. Rabbits were injected subcutaneously with 500-g portions of each protein mixed 1:1 with adjuvant five times at 3-week intervals. A 1:2000 dilution of each anti RoDH-4 antiserum detected ϳ1 ng of the corresponding recombinant protein.
The RoDH-4 cDNA was subcloned into the XbaI/BglII restriction sites of pVL1392 vector (Pharmingen, San Diego, CA). The expression construct was sequenced to verify the sequence of RoDH-4. The cotransfection of Sf9 cells with RoDH-4-pVL1392 and BaculoGold DNA was performed according to manufacturer's protocol (Pharmingen). Briefly, 2.5 million Sf9 cells were cotransfected with a mixture of 2 g of sterile RoDH-4-pVL1392 and 0.5 g of BaculoGold DNA. The infected cells were incubated for 5 days at 27°C, and the medium was collected and centrifuged for 10 min at 5000 ϫ g to remove detached cells. The supernatant was amplified two more times, and the titer of the amplified supernatant was determined by plaque assay. To produce recombinant RoDH-4, attached Sf9 cells were infected at virus:cell ratio of 10:1. Cells were collected after 3 days of incubation at 27°C; resuspended in 0.01 M potassium phosphate, pH 7.4, 0.25 M sucrose, 0.1 mM EDTA, 0.1 mM DTT; and homogenized using a Dounce homogenizer. The unbroken cells, cellular debris, and mitochondria were removed by centrifugation at 20,000 ϫ g for 15 min. Microsomes were pelleted by centrifugation at 105,000 ϫ g for 2 h and resuspended in 0.1 M potassium phosphate, pH 7.4, 0.1 mM EDTA, 0.1 mM DTT, 20% glycerol. Microsomal suspension was aliquoted into small portions and stored frozen at Ϫ80°C. Protein concentration was determined by Lowry (15) and by dye-binding assay (Bio-Rad) using bovine serum albumin as a standard.
Western Blot Analysis-A sample of frozen human liver was homogenized in 50 mM Hepes, pH 6.8, 0.5% Triton X-100, 2 mM DTT, 1 mM benzamidine, and 1 mM EDTA. The homogenate was centrifuged at 20,000 ϫ g for 30 min, and the supernatant was recentrifuged at 105,000 ϫ g for 2 h to isolate the membranes. Twenty-one micrograms of protein from each fraction, 105,000 ϫ g supernatant, and 105,000 ϫ g pellet were separated by 10% SDS-polyacrylamide gel electrophoresis and transferred to a nitrocellulose membrane. The membrane was blocked with 3% bovine serum albumin in PBS, washed several times with PBST, and incubated with a 1:2000 dilution of either anti-Nterminal or anti-C-terminal antiserum overnight. After washing with PBST, the membrane was incubated with a 1:2,000 dilution of 125 Ilabeled protein A. The bands were visualized by overnight exposure to x-ray film (Kodak X-OMAT AR).
Preparation of CRBP-bound Retinol-The coding region of CRBP (type I) cDNA (16) was amplified from rat liver total RNA by reverse transcription-PCR using Pfu polymerase. The gene-specific nucleotide primers carried restriction sites for BamHI and EcoRI endonucleases. The PCR product was subcloned into the corresponding sites in pGEX-2T expression vector and sequenced. The recombinant protein was produced at 30°C overnight in TG-1 E. coli cells in the presence of 0.2 mM isopropyl-1-thio-␤-D-galactopyranoside and 200 g/ml ampicillin as a fusion with glutathione S-transferase. The cell pellet was resuspended in ice-cold PBS, 2 mM EDTA, 0.1% ␤-mercaptoethanol and lysed using a French press. The fusion protein was purified to homogeneity using affinity chromatography on a glutathione-agarose column. The purified fusion protein was cleaved with thrombin in 50 mM Tris, pH 8.0, 0.1% ␤-mercaptoethanol, 150 mM NaCl, 2.5 mM CaCl 2 . CRBP was separated from glutathione S-transferase by elution with 0 -500 mM NaCl gradient in 10 mM Tris, pH 7.4, 1 mM DTT from a Q Sepharose column. The yield of CRBP was about 14 mg per liter of culture. The final preparation exhibited a single protein band of approximately 16 kDa by SDS-polyacrylamide gel electrophoresis. The amount of functional protein was determined from the fluorescence titration curve of apo-CRBP with retinol (17). Excitation was at 350 nm; emission was measured at 480 nm. The fluorescence values were corrected for contribution of free retinol.
The purified recombinant CRBP was incubated with excess of free retinol for 1-2 h in the dark. The free retinol was separated from the bound by elution of the S Superose column with 10 mM sodium phosphate, pH 7.4. The final preparation of CRBP-retinol had a ratio of A 350 /A 280 of 1.75.
Determination of Kinetic Constants-Steady-state kinetics were performed in 90 mM potassium phosphate, pH 7.3, and 40 mM KCl at 37°C in siliconized glass tubes. The radiolabeled steroids (NEN Life Science Products) 5␣-androstan-3␣,17␤-diol (3␣-adiol) (41 Ci/mmol), 5␣-androstan-3␣-ol-17one (androsterone) (45 Ci/mmol), and 5␣-androstan-17␤-ol-3-one (dihydrotestosterone) (43.5 Ci/mmol) were diluted with cold steroids (Sigma) to achieve the required specific radioactivity of each steroid. Aqueous solutions of the substrates were prepared by adding 100ϫ stock of the radiolabeled substrate in dimethyl sulfoxide (Me 2 SO), so that the final concentration of Me 2 SO in the reaction mixture did not exceed 1%. Equimolar amounts of bovine serum albumin were added to improve the solubility of steroids. The suspensions were sonicated for 10 min, and the concentration of the radiolabeled substrate in the aqueous phase was verified by counting an aliquot of the suspension. The 250-l reactions were started with the addition of cofactor and stopped after 15 min by addition of 3.5 ml of methylene chloride. The aqueous phase was removed and methylene chloride was evaporated under stream of N 2 . Steroids were dissolved in 50 l of ethanol; 10 l were spotted on alumina oxide thin layer chromatography plates (Sigma) and resolved by development in chloroform/ethyl acetate (3:1), according to Biswas and Russell (18). The R F values for steroids under these conditions were 0.26 for 3␣-adiol, 0.48 for androsterone, 0.53 for dihydrotestosterone, and 0.82 for 5␣-androstan-3,17-dione (androstanedione). The lanes were cut into pieces ϳ1 cm wide and counted in scintillation liquid (Bio-Safe II). For determination of apparent K m values, five concentrations between 0.6 and 1 M were used for androsterone, 0.25-4 M for dihydrotestosterone, and 0.1-2.5 M for adiol. Initial velocities (nmol of product formed/mg of protein) were obtained by linear regression. The amount of product formed was less than 10% within the 15-min reaction time and was linearly proportional to the amount of microsomes added. The K m values for oxidation of alcohols were determined at a fixed NAD ϩ (1 mM) concentration; for reduction of dihydrotestosterone, values were determined at a fixed NADH (0.5 mM) concentration. Each K m determination was repeated at least three times. A control without added cofactor was included with each experiment. The apparent K m values for cofactors were determined with six concentrations between 0.05-6.4 M for NAD ϩ , 0.25-8 mM for NADP ϩ , 15-1000 M for NADPH, and 0.15-10 M for NADH.
Retinol inhibition of steroid oxidation was evaluated by incubating RoDH-4-containing microsomes with 1 mM NAD ϩ , 4 concentrations of androsterone, and 3 concentrations of retinol in reaction buffer. All retinol solutions and reaction mixtures were kept in the dark. Each data set was evaluated for fit to different types of inhibition (19). Kinetics of initial velocities were evaluated by non-linear regression of inhibition equations using the method of Marquart (20).
Retinol Assays and HPLC Analysis of Reaction Products-Assays of RoDH-4-catalyzed oxidation and reduction of retinoids were performed in 90 mM potassium phosphate, pH 7.3, and 40 mM KCl at 37°C in siliconized glass tubes. Retinoid stock solutions in Me 2 SO were added to the reaction buffer along with equimolar bovine serum albumin and sonicated for 10 min. Experiments with tritiated retinol showed that this procedure improved solubilization of retinol. The concentration of Me 2 SO in the reaction mixture did not exceed 1%. The 500-l reactions were started with the addition of cofactor and stopped after 30 min by addition of an equal volume of cold ethanol supplemented with 100 g/ml butylated hydroxytoluene and an internal standard, retinol acetate. Reactions were placed on ice and extracted twice with 7 volumes of hexane. The aqueous phase was removed, and hexane was evaporated under a stream of N 2 . Retinoids were dissolved in 200 l of mobile phase, and an aliquot was analyzed by HPLC.
All HPLC procedures were performed using an automatic injector 710WIS from Waters and a Varian 9010 pump. Elution was monitored at 370 nm with a variable wavelength Varian 9050 detector connected to a Hewlett Packard P3390 integrator. The stationary phase was a Beckman ultrasphere ODS column (4.6 mm x 15 cm). The mobile phase consisted of 0.05 M ammonium acetate, pH 7.0:acetonitrile:tetrahydrofuran (70:168:12). The flow rate was 1 ml/min. Under these conditions, all-trans-retinol, all-trans-retinaldehyde, and retinol acetate eluted at 13, 17, and 34 min, respectively. All-trans-retinal was quantitated by comparing its peak height to a calibration curve of the amount of pure retinal injected onto the column versus the resulting peak heights.

RESULTS
To isolate human homologs of rat RoDH isoenzymes, we prepared a RoDH-1 cDNA probe by reverse transcription-PCR amplification of rat liver mRNA with gene-specific primers. The radiolabeled RoDH-1 cDNA was used to screen the human liver gt10 cDNA library. We applied low stringency conditions for screening the library, because our goal was to find an enzyme with similar (recognition of all-trans-retinol) but not identical properties (preference for NAD ϩ , not NADP ϩ ). Four positives were identified during the first round of screening. Two of these positive clones were purified and sequenced. The longest clone contained a 1308-base pair cDNA (Fig. 1), which exhibited 78% nucleotide sequence identity with RoDH-1 cDNA in the coding region. Similar to RoDH-1, the human cDNA encoded a 317-amino acid protein including a starting Met. The deduced protein product exhibited features characteristic of the short chain dehydrogenase/reductase family of enzymes, such as putative consensus sequence for the cofactor binding site, G(X) 3 GXG, at Gly-36, and the active site consensus sequence, Y(X) 3 K, at Tyr-176. The new human short chain dehydrogenase/reductase showed the highest sequence identity with the recently reported cis-retinol/androgen dehydrogenase from mouse (73%), rat RoDH forms 1 and 3 (72%), and rat RoDH-2 (71%). Because rat RoDHs share 98% sequence identity with their mouse homologs, we assumed that the human cDNA cloned in this study encoded a previously unknown form of short chain dehydrogenase/reductase, and we called the new isoenzyme RoDH-4.
The tissue distribution of RoDH-4 mRNA was analyzed by Northern blot analysis (Fig. 2). A very strong signal was observed in human liver, similar to the expression patterns of rat all-trans-retinol dehydrogenases (3,4,12). In addition, relatively high levels of hybridizing message were detected in fetal liver and lung (Fig. 2). The size of the message in fetal lung was somewhat smaller than in fetal liver. This could be due to cross-hybridization with a closely related gene product or to a different processing of the mRNA. The tissue distribution of RoDH-4 was distinctively different from that of the mouse cis-retinol/androgen dehydrogenase (6) and the two human cis-retinol dehydrogenases that are not active toward all-transretinol (2, 5).
Next, we tested whether the mRNA for RoDH-4 is translated into RoDH-4 protein in the human liver by Western blot analysis. Antibodies were raised against the N-terminal and the C-terminal fragments of RoDH-4 expressed in E. coli as described under "Materials and Methods." A sample of frozen human liver was homogenized and fractionated by centrifugation into cytosol and microsomes. RoDH-4 protein was detected by incubation with either anti-N-terminal or anti-C-terminal antiserum (Fig. 3). A single band of 35 kDa appeared in the 105,000 ϫ g membrane fraction using either antiserum (Fig. 3, lane P), indicating that RoDH-4 message is translated into a protein in human liver, and the protein is associated with membranes. This is similar to the subcellular localization of other RoDH isoenzymes.
To obtain a catalytically active enzyme, the cDNA for RoDH-4 was expressed in insect cells using the BaculoGold Baculovirus system. The full-length cDNA was subcloned into the pVL1392 transfer vector and cotransfected with linearized BaculoGold DNA into Sf9 cells. The recombinant virus was amplified and used to produce recombinant RoDH-4. The identity of the band was confirmed by Western blotting using polyclonal antiserum raised against partial RoDH-4. Our attempts to solubilize the recombinant enzyme and purify it from the membranes using Triton X-100 led to complete inactivation of the enzyme. Thus, intact membranes were used for kinetic characterization of RoDH-4.
The apparent K m values for oxidation of 3␣-adiol and androsterone were determined with saturating concentration of NAD ϩ , and the K m value for the reduction of dihydrotestosterone was determined with saturating NADH. All three steroids exhibited K m values of under 1 M (Table I), similar to RoDH-1 and CRAD. The V max values of RoDH-4 were in the range of nmol/ min/mg of microsomes (Table I) Numbers on the right correspond to nucleotide sequence, and numbers on the left correspond to amino acid sequence. The termination codon is indicated with an asterisk. The putative cofactor and substrate binding consensus sequences are underlined.

FIG. 2. Northern blot analysis of RoDH-4 in adult and fetal human tissues. Human Multiple Tissue Northern blots (CLONTECH)
containing a minimum of 2 g of poly(A) ϩ RNA per lane were hybridized with human RoDH-4 cDNA at high stringency conditions. The blot was exposed to an x-ray film for 1 week at Ϫ80°C.

FIG. 3. Western blot analysis of RoDH-4 in human liver.
Twenty-one micrograms of 105,000 ϫ g supernatant (S) and 105,000 ϫ g pellet (P) were separated by 15% SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose membrane. Western blot analysis was performed as described under "Materials and Methods." cofactors NAD ϩ and NADP ϩ were determined with saturating androsterone, and the K m values for NADH and NADPH were determined with saturating dihydrotestosterone (Table II).
RoDH-4 oxidized all-trans-retinol and 13-cis-retinol to corresponding retinaldehydes (Fig. 4). The reaction products were analyzed by HPLC. Microsomes that contained RoDH-4 oxidized all-trans-retinol to all-trans-retinal (Fig. 4, A and B) in the presence of NAD ϩ . No activity was detected in the absence of cofactor or with microsomes isolated from mock-transfected cells. The reaction rate increased linearly with the amount of enzyme in the reaction mixture (Fig. 4A). About 400 pmol of all-trans-retinaldehyde were produced from 10 M all-transretinol by 1.9 g of RoDH-4-containing microsomes in 30 min. More product was formed in the presence of NAD ϩ compared with NADP ϩ from both all-trans-retinol (not shown) and 13cis-retinol (Fig. 4C), consistent with the cofactor preference of RoDH-4 determined in the experiments with steroids. Thus, human RoDH-4 is similar to rat RoDH-1 in that it also recognizes all-trans-retinol as a substrate. However, unlike the rat enzyme, human RoDH-4 prefers NAD ϩ over NADP ϩ and is likely to function in the oxidative direction in vivo. Human RoDH-4 is also different from the NAD ϩ -dependent mouse CRAD, which is specific for cis-retinols and does not oxidize all-trans-retinol. Hence, RoDH-4 is clearly a new isoenzyme with a distinctively different primary structure and catalytic properties.
We tested whether stereoisomers of retinol can inhibit oxidation of androsterone, because both retinol and sterols were substrates for RoDH-4. Both all-trans-retinol and 13-cis-retinol acted as competitive inhibitors of androsterone oxidation (Table III). Most all-trans-retinol is bound to CRBP in liver cells. RoDH-1 recognizes CRBP-bound retinol as substrate (3); therefore, it was very interesting to see whether retinol bound to CRBP could act as efficiently as free retinol in inhibiting steroid oxidation. Surprisingly, the apparent K i value for CRBPretinol inhibition was almost the same as the K i value for free retinol (Table III). Apo-CRBP alone had no effect on the reaction. The almost identical K i values for the free and bound all-trans-retinol suggest that CRBP binding does not alter the affinity of RoDH-4 for all-trans-retinol.
Other inhibitors of RoDH-4 activity toward steroids included citral and acyclic isoprenoids. The concentration of androsterone in these experiments was at the K m value of 0.14 M. 20 M citral (3,7-dimethyl-2,6-octadienal) inhibited RoDH-4 activity with androsterone about 50% (Table IV). Some of the acyclic isoprenoids, perillyl alcohol, geraniol, farnesol, and geranyl geraniol, were even more potent inhibitors of steroid oxidation (Table IV). Ethanol at 100 mM had no effect on RoDH-4 activity with androsterone. DISCUSSION We have cloned and characterized a new short chain dehydrogenase/reductase, RoDH-4, which represents the first human microsomal enzyme capable of oxidizing all-trans-retinol to all-trans-retinaldehyde and the second human enzyme to oxidize 3␣-hydroxysteroids with NAD ϩ as the preferred cofactor. The primary structure of RoDH-4 is about 70% identical to the rodent retinol/sterol dehydrogenases.
Currently, four types of retinol-oxidizing short chain dehydrogenases/reductases can be distinguished. The first type are the NADP ϩ -preferring isoenzymes that utilize free and CRBPbound all-trans-retinol in rat (RoDH-1 and RoDH-2) (3, 4) and mouse (RoDH-1 and RoDH-2) (6). The second type is the NAD ϩdependent CRAD, recently cloned from mouse (6). CRAD has 80 -85% sequence identity to RoDHs, and it oxidizes 9-cisretinol and 11-cis-retinol but not all-trans-retinol. The third type is the eye-specific retinol dehydrogenase in retinal pigment epithelium that recognizes only 11-cis-retinol (2). The fourth type shares 95% amino acid sequence identity with the eye 11-cis-retinol dehydrogenase but apparently prefers 9-cisretinol as substrate and is found in multiple tissues (5). The last two cis-retinol dehydrogenases prefer NAD ϩ over NADP ϩ , and have only 50 to 55% identity with RoDHs and CRAD.
Recently, a cDNA for the first human RoDH-like short chain dehydrogenase/reductase has been isolated from the human prostate cDNA library by screening with rat RoDH-1 cDNA (18). Analysis of the substrate specificity of the prostate RoDHlike enzyme, transiently expressed in embryonic kidney 293 cells, showed that it oxidizes 3␣-hydroxysteroids efficiently and exhibits higher affinity for NAD ϩ than for NADP ϩ . Whereas reductive NADP ϩ -dependent 3␣-hydroxysteroid dehydrogenases have been described, the prostate RoDH-like dehydrogenase represents the first known oxidative NAD ϩ -dependent 3␣-hydroxysteroid dehydrogenase. Whether it is also active on retinols was not reported. This prostate 3␣-hydroxysteroid dehydrogenase has only 62% amino acid sequence identity with human RoDH-4 reported in this study and is less similar to retinol-oxidizing RoDH isoenzymes (61-63% amino acid sequence identity) than RoDH-4 (72% identity).
Similar to the prostate RoDH-like steroid dehydrogenase, RoDH-1 and CRAD were found to oxidize the 3␣-hydroxysteroids 3␣-adiol and androsterone to dihydrotestosterone and androstanedione, respectively (3,6). We tested whether the new human RoDH-4 was active against steroid alcohols. As expected, human RoDH-4 oxidized 3␣-adiol to dihydrotestosterone and androsterone to androstenedione in the presence of either NAD ϩ or NADP ϩ . Accordingly, 3␣-dihydrotestosterone was reduced to 3␣-adiol in the presence of NADH or NADPH. The catalytic efficiency of RoDH-4 in the reductive direction was about 2.5 times lower than in the oxidative direction.
The nonphosphorylated cofactors NAD ϩ and NADH were clearly preferred by RoDH-4: the apparent K m value for NAD ϩ was almost 1500 times lower than that for NADP ϩ (Table III). Similarly, the apparent K m value for NADH was 250 times lower than that for NADPH.
The   of each experiment; thus, no direct comparisons can be made. The relative catalytic efficiencies of all steroid-oxidizing isoenzymes toward each substrate, however, were similar. In general, the catalytic efficiencies were about twice higher for oxidation of 3␣-adiol than for oxidation of androsterone. Attempts to purify this and similar enzymes out of microsomal membranes resulted in either loss or significant decrease in enzymatic activity. The only retinol/sterol dehydrogenase ever purified directly from rat tissues, RoDH-1, had to be reconstituted with phosphatidylcholine to regain its activity (21). The specific activity reported for the partially purified RoDH-1 was 307 pmol/min⅐mg with free retinol and 96 pmol/ min⅐mg for CRBP-bound retinol in the presence of NADP ϩ (21).
RoDH-1 activity with free retinol was lower in the presence of NAD ϩ : 1.7 times with free retinol and 5.6 times if CRBP-retinol was a substrate.
Similar to RoDH-1, RoDH-4 converted all-trans-retinol to all-trans-retinal in the presence of either NAD ϩ or NADP ϩ ; however, the reaction rate was about five times higher with NAD ϩ . Enzymatic activity similar to RoDH-4 has been described in the liver microsomes of rat and ADH-negative deermouse (22). The rat liver enzyme in the above study was more active with NAD ϩ than with NADP ϩ , was similarly insensitive to ethanol inhibition, and was strongly inhibited by Triton X-100. Thus, it is likely that homologs of human RoDH-4 exist in rats and mice.
If the catalytic site of RoDH-4 faces the cytosol, then the cytosolic ratios of cofactors (NAD ϩ /NADH ϭ 1000) should drive the reaction in the oxidative direction. However, it is not known how RoDH and similar enzymes are inserted into the microsomal membrane. We analyzed the primary structures of retinoloxidizing short chain dehydrogenases/reductases for the presence of membrane-spanning domains using programs HELIXMEM (23), RAOARGOS (24), and SOAP (25,26). These programs predict the existence of membrane-associated ␣ helices in a protein sequence, evaluate the hydropathic index along the sequence, and predict whether a membrane protein is peripheral or integral. Although the number and exact location of the membrane-spanning domains predicted by the different programs differed somewhat, all retinol-oxidizing short chain dehydrogenases/reductases were classified as integral membrane proteins (Fig. 5A). Comparative analysis of the primary structures of the isoenzymes suggested that RoDH-4 contains four potential transmembrane segments: the N-terminal segment 1 (amino acids 1-21), the two closely positioned central segments 2 and 3 (amino acids 105-125 and 130 -150), and the C-terminal segment 4 (amino acids 288 -309) (Fig. 5, boxed). The hydrophobic N-terminal segment 1 is followed by the two highly conserved arginines at positions 19 and 21. The hydrophobic domain 3 is also followed by two conserved arginines at positions 156 and 158. The topological rules for membrane protein assembly in eukaryotic cells imply that the positively charged amino acids, such as arginine and lysine, immediately downstream of the first hydrophobic domain prevent translocation across the microsomal membrane and that there is a  4. Oxidation of retinols to retinaldehydes. A and B, oxidation of all-trans-retinol (atRol) to all-trans-retinal (atRal). A, 10 M all-trans-retinol was incubated with 2.5-40 g of RoDH-4-containing microsomes. Open circles, 10 M all-trans-retinol plus 1 mM NAD ϩ ; closed circles, 10 M all-trans-retinol minus cofactor. Rate is expressed in relative units (peak height of retinaldehyde divided by peak height of internal standard). B, a representative HPLC chromatogram of the above reaction products in the presence (plus NAD ϩ ) and absence (minus NAD ϩ ) of cofactor with 5 g of microsomes. C, oxidation of 13-cis-retinol (13c Rol) to 13-cis-retinaldehyde (13c Ral) in the presence of NAD ϩ versus NADP ϩ . RoDH-4-containing microsomes (0.6 g) were incubated with 50 M 13-cis-retinol and 1 mM of either NAD ϩ (left) or NADP ϩ (right) for 30 min. The reaction products were analyzed as described under "Materials and Methods." The minor peaks coeluting with all-trans-retinaldehyde and 13-cis-retinaldehyde were also present in the commercial standards and could be due to isomerization. clear tendency for highly charged internal loops to remain on the cytoplasmic side (27). The two neighboring hydrophobic segments, 2 and 3, which are connected by a short loop, can be inserted according to a helical hairpin mechanism (28).
Based on the analysis of the primary structure of RoDH-4, we propose a model of transmembrane insertion of RoDH-4 in the microsomal membrane (Fig. 5B). According to this model, the two loops between the hydrophobic segments will remain on the cytosolic side. Because the concentration of NAD ϩ in the cytosol greatly exceeds that of NADH, RoDH-4 will likely function as a dehydrogenase, not reductase, in the cells. Interestingly, the cofactor binding consensus sequence is located on the N-terminal loop and is highly conserved among short chain dehydrogenases/reductases. The substrate binding consensus sequence is located on the less conserved C-terminal loop, which is consistent with the different substrate specificity of the isoenzymes. Based on our model of transmembrane insertion of RoDH-4, the cofactor binding domain will face the cytosol, and RoDH-4 will function in the oxidative direction.
This model is consistent with our observation that CRBPretinol inhibits oxidation of androsterone catalyzed by RoDH-4 with the same efficiency as free retinol, suggesting that both forms of retinol have equal access to the active site. CRBP is a cytosolic protein, and this is consistent with the suggestion that the active site of RoDH-4 faces cytosol. Several microsomal enzymes were found to metabolize bound forms of retinoids with the same catalytic efficiency as free. For example, the intestinal retinal reductase does not distinguish between free and CRBP-II-bound all-trans-retinal (29). It exhibits similar affinity (0.78 M for free and 0.46 M for bound) and almost identical rates (ϳ300 pmol/min/mg of microsomal protein) with both forms of retinal. In the pigment epithelium of the eye, 11-cis-retinaldehyde reductase is about equally efficient in reducing free 11-cis-retinal and 11-cis-retinal complexed with cellular retinal-binding protein (30).
Citral is often used to inhibit oxidation of retinol and retinaldehyde to retinoic acid in cell culture experiments (31). Human keratinocytes incubated in the presence of 100 M citral for 4 h show 75% reduction in retinoic acid synthesis from all-trans-retinol (32). In our experiments, 20 M citral inhibited RoDH-4 45%. This also distinguishes RoDH-4 from RoDH-1, which is not inhibited by citral.
RoDH-4 can oxidize the carbon 3 alcohol group on 3␣-adiol, producing dihydrotestosterone, a powerful androgen with high affinity for androgen receptor. RoDH-4 is the second isoform of human NAD ϩ -dependent 3␣-hydroxysteroid dehydrogenases; it shares 62% sequence identity with the first isoform, human RoDH-like 3␣-hydroxysteroid dehydrogenase cloned from prostate. In prostate, dihydrotestosterone is produced mainly from testosterone by steroid 5␣-reductases (reviewed in Ref. 33). Dihydrotestosterone is inactivated by NADPH-dependent cytosolic 3␣-hydroxysteroid dehydrogenases, which reduce dihydrotestosterone to adiol (reviewed in Ref. 34). The presence of the NAD ϩ -dependent 3␣-hydroxysteroid dehydrogenase in prostate explains the so-called back reaction, in which adiol is oxidized back to dihydrotestosterone. Both RoDH-4 and RoDHlike steroid dehydrogenase are abundant in the liver. Dihydrotestosterone is inactivated in the liver cytosol by multiple reductive 3␣-hydroxysteroid dehydrogenases. The role of dihydrotestosterone producing RoDH-4 in the non-steroidogenic liver is not clear yet.
Dietary acyclic isoprenoids d-limonene, perillyl alcohol, geraniol, farnesol, and geranyl geraniol are currently undergoing clinical trials as chemopreventive therapeutical agents against mammary, liver, lung, stomach, skin, and pancreas cancers (35). Geraniol and farnesol are present in the essential oils of lemongrass, perillyl alcohol is present in cherry and spearmint, and d-limonene is present in orange peel oil, caraway, and dill. The primary metabolites of limonene, perillic acid and dihydroperillic acid, are more potent inhibitors of tumor cell proliferation than is perillyl alcohol (36). The enzymes that oxidize acyclic isoprenoids to their corresponding acids in vivo are not known. We hypothesized that acyclic isoprenoids may be recognized by RoDH-4 along with all-trans-retinol and steroid alcohols. In this study, we have established that acyclic alcohols can function as effective inhibitors of RoDH-4.
To our knowledge, RoDH-4 is the first human retinol-oxidizing short chain dehydrogenase/reductase that is active toward all-trans-retinol and prefers NAD ϩ over NADP ϩ . This implies that in vivo RoDH-4 functions in the oxidative direction and can contribute to the biosynthesis of all-trans-retinoic acid in the adult and fetal liver, and possibly in fetal lung. The contribution of this microsomal enzyme to retinoic acid biosynthesis relative to cytosolic ADH isoenzymes is a subject of future studies. It is clear, however, that microsomal RoDH isoenzymes may have a broader substrate specificity than was initially thought. Additional isoforms of all-trans-retinol dehydrogenases may be responsible for retinol oxidation in extrahepatic tissues.