cDNA Cloning, Tissue Distribution, and Substrate Characteristics of a cis-Retinol/3α-Hydroxysterol Short-chain Dehydrogenase Isozyme*

We report here a mouse cDNA that encodes a 316-amino acid short-chain dehydrogenase that prefers NAD+ as its cofactor and recognizes as substrates androgens and retinols, i.e. has steroid 3α- and 17β-dehydrogenase and cis/trans-retinol catalytic activities. This c is-retinol/androgendehydrogenase type 2 (CRAD2) shares close amino acid similarity with mouse retinol dehydrogenase isozyme types 1 and 2 and CRAD1 (86, 84, and 87%, respectively). CRAD2 exhibits cooperative kinetics with 3α-adiol (3α-hydroxysteroid dehydrogenase activity) and testosterone (17β-hydroxysteroid dehydrogenase activity), but Michaelis-Menten kinetics with androsterone (3α-hydroxysteroid dehydrogenase activity), 11-cis-retinol, all-trans-retinol, and 9-cis-retinol, withV/K 0.5 values of 1.6, 0.2, 0.1, 0.04, 0.005, and not saturated, respectively. Carbenoxolone (IC50 = 2 μm) and 4-methylpyrazole (IC50 = 5 mm) inhibited CRAD2, but neither ethanol nor phosphatidylcholine had marked effects on its activity. Liver expressed CRAD2 mRNA intensely, with expression in lung, eye, kidney, and brain (2.9, 2, 1.6, and 0.6% of liver mRNA, respectively). CRAD2 represents the fifth isozyme in a group of short-chain dehydrogenase/reductase isozymes (retinol dehydrogenases 1–3 and CRAD1), closely related in primary amino acid sequence (∼85%), that are expressed in different quantities in various tissues, have different substrate specificities, and may serve different physiological functions. CRAD2 may alter the amounts of active and inactive androgens and/or convert retinols into retinals. These data expand insight into the multifunctional nature of short-chain dehydrogenases/reductases and into the enzymology of steroid and retinoid metabolism.

The SDR 1 superfamily consists of ϳ100 bacterial, plant, and animal enzymes ranging in size from ϳ25 to 38 kDa that are related in terms of tertiary structure, including conserved cofactor-binding sites and catalytic residues (1)(2)(3). But the members of the SDR superfamily have relatively few strictly conserved residues, and indeed, different members do not always share substantial amino acid identity. SDRs tend to have a multifunctional nature, i.e. they catalyze dehydrogenations/ reductions of seemingly disparate substrates. In animals and bacteria, members of this superfamily catalyze the activation or inactivation of prostaglandins and many steroids. An apparent subgroup of the SDR superfamily, consisting of enzymes closely related to each other, catalyzes the metabolism of alltrans-retinol, cis-retinols, and androgens. This subfamily includes RoDH isozymes 1-3 and CRAD1 (4 -7). A related SDR also catalyzes 11-cis-and 9-cis-retinol dehydrogenations (8 -10).
Vertebrates require retinoid hormones, derived from the prohormone retinol (vitamin A), for vision, reproduction, embryogenesis, and maintenance of normal epithelial, bone, nerve, and immune system function (11). The retinol metabolite alltrans-retinoic acid satisfies all known retinoid functions in retinol-deficient animals, except for light transduction during vision, because it cannot undergo reduction into the opsin cofactor retinal, and spermatogenesis, because it cannot cross the mammalian blood-testis barrier in low concentrations. Alltrans-retinoic acid functions through the three ligand-dependent transcription factors known as RAR␣, RAR␤, and RAR␥ (12,13). An all-trans-retinoic acid isomer, 9-cis-retinoic acid, controls the in vitro activity of a distinct group of receptors known as RXR␣, RXR␤, and RXR␥. RXRs affect the function of several receptors in the steroid/retinoid/thyroid/vitamin D superfamily of ligand-activated transcription factors, including RARs, through heterodimerization. Because receptor function depends on ligand concentrations, pathways of all-trans-retinoic acid and 9-cis-retinoic acid biosynthesis require understanding. A pathway of all-trans-retinol conversion first into all-trans-retinal and then into all-trans-retinoic acid has been established, but pathways of 9-cis-retinoic acid biosynthesis have not been determined (14). One problem with the latter has been identifying processes for enzymatically generating the putative hormone. Thus, enzymes that produce 9-cis-retinoids incur much interest.
Here, we report the isolation of a cDNA that encodes a heretofore unknown SDR, CRAD2. Many tissues express CRAD2 mRNA, but liver is the quantitatively major site of expression. CRAD2 shows 3␣-and 17␤-hydroxysteroid dehydrogenase activities and catalyzes the dehydrogenation of retinols, including 9-cis-retinol. Expression of CRAD2 provides a means of altering the concentrations of active versus inactive androgens and of generating retinals from retinols.

MATERIALS AND METHODS
Production of a CRAD2-specific Probe-An 11-d.p.c. mouse embryonic gt11 cDNA library (CLONTECH) was screened through three rounds (final wash at 55°C) with a 32 P-labeled probe consisting of RoDH1 nucleotides 298 -673 (4). DNA inserts from six positive phages were digested with EcoRI and cloned into pBluescript II SK ϩ/Ϫ . One of the six clones, pBSK/E6, was completely sequenced by nested deletion. Despite the use of a cDNA library, the 2.2-kb insert represented a genomic clone that included exon 1 of a gene in the SDR family. A 367-base pair probe containing 55 base pairs of 5Ј-end untranslated region and 312 base pairs of the coding region in exon 1 of E6 was generated by polymerase chain reaction with the primers 5Ј-TTACTCTCTGAAAACGGGGC (sense) and 5Ј-TCTGTTCCCAACAC-GCTC (antisense).
CRAD2 cDNA Isolation-A mouse liver gt10 library (CLONTECH) was screened with a probe consisting of nucleotides 683-1080 of CRAD1 (7). The positive plaques were then screened through two more rounds with the 367-base pair probe isolated from pBSK/E6. Phage DNA from three of five plaques isolated was digested with EcoRI. The inserts were ligated into pBluescript II SK ϩ/Ϫ to produce pBSK/CRAD-39, pBSK/ CRAD-188, and pBSK/CRAD-218 and sequenced in both directions by dideoxy chain termination.
Expression of CRAD2-The coding region of an SDR was digested from gt10 phage DNA containing the insert of pBSK/CRAD-218 with EcoRI and ligated into pcDNA3 to produce pcDNA3/CRAD2. CHO-K1 cells were cultured and transfected using LipofectAMINE with pcDNA3/CRAD2 or with pcDNA3 (mock) as described (7). Cell pellets were suspended in 10 mM HEPES and 10% sucrose, pH 7.5, and homogenized with a Dounce homogenizer. The homogenate was centrifuged at 800 ϫ g for 20 min, and the supernatant protein was used for enzyme assays, unless noted otherwise. Protein concentrations were determined by the method of Bradford (21).
Enzyme Assays-Incubations and analysis of products have been described in detail (7). Briefly, retinoid and steroid dehydrogenase assays were done at 37°C in 0.25 ml of 50 mM HEPES, 150 mM KCl, 1 mM EDTA, and 1.6 mM NAD ϩ , pH 8, with the 800 ϫ g supernatant of mock-or pcDNA3/CRAD2-transfected CHO cells, unless noted otherwise. Retinoid dehydrogenase assays were quenched with 0.1 ml of 0.1 M O-ethylhydroxylamine and 0.35 ml of methanol, incubated at room temperature for 10 min, and extracted with 2.5 ml of hexane. The retinoids in the hexane extract were quantified by normal-phase highperformance liquid chromatography, with a detection limit ϳ1 pmol. 11-cis-Retinoids isomerize into more stable isomers during incubation and extraction; therefore, the sum of the retinal isomers recovered from incubating 11-cis-retinol was used to determine the rate of 11-cisretinal synthesis (22). Steroid dehydrogenase assays were done with 3 H-labeled steroids (40 -101 Ci/mmol, 20,000 dpm/reaction). Reactions were extracted with methylene chloride (4 ml); the extracts were analyzed by thin-layer chromatography. 3 H-Labeled steroids were detected by autoradiography. The radioactive zones were excised and counted with a liquid scintillation counter. Kinetic data were obtained under initial velocity conditions and were fit with Enzfitter using simple weighing (23).
Northern Blotting-Northern blots were done with the mouse multiple tissue Northern blot (CLONTECH), which provides 2 g of poly(A ϩ ) RNA/lane on a nylon membrane. The probe was a chemically synthesized 75-base-long oligonucleotide of nucleotides 1-75 of CRAD2. The probe was labeled with 32 P by random priming. Prehybridization was done in 10 ml of hybridization solution (50% formamide, 5ϫ Denhardt's solution, 0.1% SDS, 100 g/ml denatured salmon sperm DNA, and 5ϫ saline/sodium phosphate/EDTA) at 40°C for 4 h. Hybridization was done overnight in the same solution containing 2 ϫ 10 6 cpm of probe. The final wash was done at 55°C with 1ϫ SSC and 0.1% SDS. Signals were visualized with a Bio-Rad GS-505 Molecular Imager system.
RNase Protection Assay-A CRAD2-specific probe was amplified by polymerase chain reaction from pBSK/CRAD-39 with the sense primer 5Ј-GCTTCCATACTACTCAGA (nucleotides 1135-1152) and the antisense primer 5Ј-CAAGATTCTATCCCACCA (nucleotides 1463-1480). The polymerase chain reaction product was subcloned into pGEM-T (Promega) and linearized with SpeI. A 32 P-labeled antisense probe was transcribed with T7 RNA polymerase (Promega) for 1 h at 37°C in 10 mM dithiothreitol; 0.5 mM each ATP, CTP, and GTP; and 50 Ci of UTP (800 Ci/mmol). The 308-nucleotide antisense ␤-actin mRNA probe (nucleotides 51-358) used as an internal standard was transcribed from pTRI mouse ␤-actin (Ambion Inc.) under the same conditions. DNA templates were removed by DNase I digestion. Transcripts were purified with 5% polyacrylamide and 8 M urea gels. RNase protection assays were done with the Hybspeed TM RPA kit (Ambion Inc.) following the manufacturer's instructions. Total RNA (50 g) was extracted from mouse tissues with guanidinium thiocyanate/phenol/chloroform and coprecipitated with cRNA probes (1 ϫ 10 5 cpm for CRAD2 and 5 ϫ 10 4 cpm for mouse ␤-actin) by 0.5 M ammonium acetate and 70% ethanol. Pellets were resuspended in 10 l of hybridization buffer (Ambion Inc.) by four alternating 15-s periods of vigorous vortexing and incubation at 95°C for 3 min. Samples were hybridized at 68°C for 10 min. A 100-l aliquot of RNase A/T1 mixture diluted 1:100 was allowed to digest the unhybridized probes and RNA for 30 min at 37°C. Inactivation/precipitation mixture (150 l) was added, and the samples were kept at Ϫ20°C for 30 min. After centrifugation, the supernatants were removed, and the pellets were dissolved in 8 l of gel loading buffer for denaturing gels by heating at 95°C for 4 min. The samples were loaded onto 5% polyacrylamide and 8 M urea gels and run at ϳ180 V for 2 h in 1ϫ Tris borate/EDTA. Quantitative analysis was performed with a Bio-Rad Molecular Imager PC.

RESULTS
cDNA and Amino Acid Sequences-A mouse embryonic cDNA library was screened with a probe that encoded amino acids 2-126 of RoDH1, i.e. sequence highly conserved among the three known RoDH isozymes and CRAD1 (4 -7). A partial genomic clone was isolated, most likely from contamination during preparation of the commercial cDNA library. Because we have isolated genomic clones for RoDH/CRAD, 2 it was recognized as exon 1 of an unknown SDR with high sequence similarity to CRAD1. To obtain the complete coding region of the novel cDNA, a mouse liver cDNA library was first screened with a probe from CRAD1; the positive plaques were rescreened with a probe generated by polymerase chain reaction from the exon of the novel SDR. Two of the three positive plaques identified included complete coding regions, and all three had identical deduced amino acid sequences in areas of overlap. The amino acid sequences, however, were distinct from those of RoDH isozymes 1-3 and CRAD1. The new SDR was named CRAD2 because its substrate specificity resembled that of CRAD1 (see below).
The deduced amino acid sequence of CRAD2 contained 20 of the 23 amino acids conserved in ϳ70% of SDRs, including the cofactor-binding residues G 36 (X) 2 SGXG, the L 109 XNNAG sequence (unknown function), and the catalytic residues Y 175 (X) 3 K (Fig. 1). All three RoDH isozymes, CRAD1, and CRAD2 show high sequence conservation in their first 115 N-terminal amino acids. The largest difference occurs between CRAD2 and RoDH2, which differ in six amino acid residues, four of which represent nonconservative changes (T13N, Q22K, N71S, and R104T). CRAD2 differs from CRAD1 by only five amino acids of the first 115, four of which are nonconservative (E20V, N43T, Q68E, and A92T). RoDH1, identical to RoDH3 in this area, and CRAD1 also differ by only five amino acids of the first 115 N-terminal residues: three changes are nonconservative (T13N, Q22K, and N71S). This region contains the cofactor-binding site, the conserved SDR sequence LVNNAG, and a putative membrane-anchoring sequence.
Outside of this group of five SDRs (CRAD1, CRAD2, and RoDH1-3), rat 17␤-HSD6 has the closest amino acid similarity and identity to CRAD2 (Table I). The SDR that catalyzes 11cis-retinol dehydrogenation has less sequence similarity to CRAD2 than does 17␤-HSD6, suggesting that the CRADs and RoDHs belong to a distinct subgroup of SDRs that may not include the 11-cis-retinol dehydrogenase (8,9). The bovine 11cis-retinol dehydrogenase, for example, differs from CRAD2 by 50 amino acid residues in the first 115 residues.
Subcellular Fractionation of CRAD2-Association of CRAD2 with membrane fractions was demonstrated by differential centrifugation. Centrifugation of the 800 ϫ g supernatant of transfected CHO cells at 10,000 ϫ g for 30 min partitioned 75% of the CRAD2 activity into the supernatant, measured with 3␣-adiol. Centrifugation of this supernatant at 100,000 ϫ g for 2 h partitioned 89% of the recovered CRAD2 activity into the microsomal pellet.
CRAD2 mRNA Tissue Expression-Northern blot hybridization revealed intense expression of CRAD2 mRNA in mouse liver, with the most intense signal at 1.7 kb and two less intense signals at 3.5 and 2.9 kb (Fig. 4). Lung also showed the 1.7-kb mRNA, albeit at ϳ3% of the 1.7-kb liver signal. In comparison, CRAD1 was expressed very intensely in both liver and kidney, with major mRNA species at 3.5 and 2.7 kb in both tissues, a major 3.0-kb band in liver only, and a weaker 4.4-kb band in liver and kidney. RoDH isozymes 1-3 showed a single 1.7-kb band in liver by Northern blot analysis (4 -7). No signals for CRAD2 were observed by Northern blotting in heart, brain, spleen, skeletal muscle, or testis, as was the case for CRAD1 and RoDH isozymes 1-3. The more sensitive RNase protection assays detected CRAD2 mRNA expression in the following (relative intensity normalized to the ␤-actin signal): liver (100) Ͼ lung (3) Ͼ eye (2) Ͼ kidney (1.6) Ͼ brain (0.6) (Table IV). No signals were detected in testis or heart. RNase protection assays also revealed low levels of CRAD1, RoDH1, and RoDH2 in multiple tissues, but did not reveal expression of RoDH3 outside of the liver (4 -7). DISCUSSION cDNA cloning of CRAD2 reveals a fifth member of a subgroup of SDRs (CRAD1, CRAD2, and RoDH1-3) whose members share amino acid sequence identities of 80 -97%. Nearly 95% conservation of the first 115 N-terminal amino acid residues represents a distinguishing feature of these five members. Rat 17␤-HSD6, the SDR with the nearest amino acid identity to the group of five, shares only 86% identity in this area with the group of five. Human RoDH has only 83% identity in this region, and the SDR associated with 11-cis-retinol dehydrogenation has only 57% amino acid identity in this area, not markedly different from its overall identity to the group of five. Most likely, this conservation of the N terminus serves a unique, but as yet undetermined function. Similarly, amino acid residues 260 -304 share ϳ84 -98% identity among the group of five. In contrast, rat 17␤-HSD6 has only 76 -78% identity with the group of five from amino acid residues 260 to 304.
It appears that at least two distinct subgroups of SDRs occur with activity toward retinoids. The "group of five" enzymes possibly compose a subgroup distinct from the three other enzymes most closely related in sequence, namely 17␤-HSD6, human RoDH, and 11-cis-RoDH. For example, both the overall difference in identity and the difference in identity in the conserved N terminus indicate that the SDR candidate for a human homologue of an RoDH isozyme represents an SDR outside of the subgroup, given the generally high species conservation of proteins involved in retinoid metabolism and/or function. The large differences between the group of five and the SDR with activity toward 11-cis-retinol also suggest that the latter lies outside of the subgroup and perhaps constitutes a subgroup of its own. The recent cloning of a cDNA that encodes a mouse SDR (RDH4) supports this supposition. RDH4, which has 9-cis-retinol dehydrogenase activity, has ϳ87% amino acid identity to the 11-cis-retinol dehydrogenase, but only 53% identity to mouse RoDH1. 3 Moreover, the 11-cisretinol dehydrogenase also has potent 9-cis-retinol dehydrogenase activity. 2 Although the enzymes in the group of five have relatively high amino acid conservation, to the extent tested, they show substantial differences in enzymatic activity. CRAD2 discriminates between 9-cis-and 11-cis-retinols, in contrast to CRAD1, which shows similar activity (V/K 0.5 values) for these two cisretinols (7). Recombinant RoDH isozymes have not been tested with cis-retinols. The activities with androgens also differed among the three tested. CRAD2 was most efficient as a 3␣hydroxysteroid dehydrogenase, but had less efficient 3␣-hydroxysteroid dehydrogenase activity than either RoDH1 or CRAD1, which had K m values ϳ0.1-0.2 M for 3␣-adiol. CRAD1 had a V m value (27 nmol/min/mg of protein) with 3␣adiol that was 8-fold faster than that of CRAD2. Even though the experiments were done at different times and care must be exercised when comparing V m values produced from different transfections, these rates observed for each CRAD were consistent for several transfections and therefore likely reflect inherent differences between the two. CRAD2 also differs from CRAD1 and RoDH1 in its relatively high 17␤-hydroxysteroid dehydrogenase activity. Neither CRAD1 nor RoDH1 had activity with dihydrotestosterone, and both had negligible activity with testosterone.
The prostate epithelial cell steroid 5␣-reductase reduces the C 4 -ene of testosterone to produce the major biologically active androgen of prostate, dihydrotestosterone (15). The effects of dihydrotestosterone are, in turn, limited by 3␣-hydroxysteroid dehydrogenases (Fig. 5). 3␣-Hydroxysteroid dehydrogenases are members of the aldo-keto reductase superfamily (expressed in prostate, liver, kidney and several other tissues) that reduce the 3-oxo function of dihydrotestosterone to produce 3␣-adiol.
3␣-and 17␤-Hydroxysteroid dehydrogenases constitute a pathway of dihydrotestosterone inactivation. 3␣-Adiol binds to the androgen receptor with 5 orders of magnitude less affinity than dihydrotestosterone (31)(32)(33). 17␤-Hydroxysteroid dehydrogenases, members of the SDR superfamily, convert 3␣-adiol into androsterone, a steroid with even less androgen activity in vivo than its precursor (20). On the other hand, 3␣-hydroxysteroid dehydrogenases that function oxidatively would convert the relatively inactive 3␣-adiol and androsterone into dihydrotestosterone and androstanedione, respectively. The latter could then undergo activation into dihydrotestosterone through reduction of its 17-oxo function (Fig. 5). The three SDRs tested so far with androgens (RoDH1, CRAD1, and TABLE III CRAD2 activity with steroid and retinoid substrates K 0.5 values represent the averages of two separate measurements with 8 -12 substrate concentrations, except those for 3␣-adiol and testosterone, for which four and three independent determinations were made. Duplicate determinations were made for each concentration in each experiment. Steroid reactions were done with 1-15 g of protein for 10 min. Retinoid reactions were done with 100 g of protein for 30 min.  CRAD2) have the most efficient activity as 3␣-hydroxysteroid dehydrogenases of any enzymes known so far. They may contribute to androgen action by producing dihydrotestosterone from 3␣-adiol and androsterone, thereby "rescuing" dihydrotestosterone from inactivation and excretion. Such a role for RoDH/CRAD isozymes seems feasible in vivo because dihydrotestosterone has been detected in castrated and functionally hepatectomized rats, and 5␣-reductase inhibitors do not obliterate androgen activity (15). Also, despite low affinity for the androgen receptor, 3␣-adiol stimulates prostate growth in vivo and in organ culture, consistent with metabolism into dihydrotestosterone (19,34,35).
3␣-Adiol may not function solely as an androgen inactivation product. Pregnant mice with a null allele in the type 1 5␣reductase gene failed to deliver pups on time, but instead entered prolonged labor on days ϳ21-22 (36). About half resorbed their fetuses or expelled dead fetuses, and the other half either died during labor or suffered massive sepsis. The defect was seemingly related solely to parturition because the birth canal suffered no apparent developmental defect, and apparently normal pups were delivered on day 19.5 by cesarean section. Dosing with 3␣-adiol increased the incidence of normal parturition from 27 to 93%. An equivalent dose of dihydrotestosterone was less effective, raising the incidence to 57%. These data prompted the suggestion that 3␣-adiol may serve as a hormone required for parturition in mice. If so, then a second function of RoDH/CRAD could involve affecting the onset of parturition through altering the 3␣-adiol concentration.
Observance of dual androgen/retinoid substrate SDRs provides opportunity for providing insight into the physiological interactions between retinoids and steroids. Spermatogenesis requires functional steroid and retinoid receptors (37,38). Retinoic acids inhibit prostate epithelial cell growth (39,40), and inhibition of all-trans-retinoic acid metabolism in the rat Dunning prostate cancer model inhibits carcinoma relapse after castration by raising all-trans-retinoic acid plasma levels (41). All-trans-retinoic acid decreases concentrations of dihydrotestosterone, 3␣-adiol, and androsterone in serum and seems to cause a metabolic deviation away from the 5␣-path in liver (39,42). Other than causing a 3-fold decrease in androgen receptor binding (39,43), very little is known about the mechanisms of retinoid effects on androgen activity. Indeed, very little is known in general about the extent of the retinoid/androgen interaction. Conversely, androgens affect the actions of retinoids by decreasing the mRNA of RAR␣ ϳ5-fold in prostate epithelia and 15-20-fold in seminal vesicles, while increasing it 2-fold in kidney (44). The dual androgen and retinoid activities of RoDH1 and CRAD isozymes could position them as mediators of retinoid/androgen interactions. Mechanisms of such potential interactions might include direct competitive and/or allosteric effects or indirect effects through gene expression via RARs, RXRs, and/or the androgen receptor.
The 11-cis-retinol dehydrogenase activity of CRAD2 and its expression in the eye are consistent with CRAD2 contributing 11-cis-retinal for use as a rhodopsin chromophore. 11-cis-Retinoids, however, have not been demonstrated outside of the eye. Therefore, extraocular and perhaps intraocular CRAD2 may support 9-cis-retinoic acid biosynthesis by converting 9-cis-retinol, available from diet or from 9-cis-␤-carotene metabolism (45)(46)(47)(48)(49)(50)(51)(52)(53), into 9-cis-retinal. Although CRAD2 was not saturated kinetically with physiological levels of 9-cis-retinol, it showed sufficient activity with low 9-cis-retinol concentrations to contribute to the pool of 9-cis-retinal, which is in the low nM range. Two other SDRs have been reported with 9-cis-retinol dehydrogenase activity. CRAD1 showed a K 0.5 value of ϳ5 M and a V m of ϳ10 nmol/min/mg for 9-cis-retinol (7). The second, originally reported as 11-cis-retinol dehydrogenase (8,9), was shown subsequently also to recognize 10 M 9-cis-retinol as substrate in a one-point assay (10). With the data currently available, it is not possible to assess the relative contributions of these three enzymes to the production of 9-cis-retinoic acid in vivo.
The exact function of CRAD2, and CRAD1 as well, may depend on loci of expression, substrate availability, hormonal influences, and other as yet unappreciated factors. Future investigations will address these issues. Finally, the variable effects of agents such as carbenoxolone and 4-methylpyrazole on different SDRs and alcohol dehydrogenases suggest that caution should be exercised in interpreting experiments in vivo using such reagents. RNase protection assays were done as detailed under "Materials and Methods." Data for CRAD2 were normalized to its own mouse liver signal. Data for CRAD1 were normalized to its own mouse kidney signal. CRAD2 and CRAD1 data were not normalized to each other or to the data for RoDH isozymes. Data for all RoDH isozymes were normalized to the RoDH1 rat liver signal. FIG. 5. Possible roles of RoDH and CRAD isozymes in androgen metabolism. 5␣-Reductase converts testosterone into dihydrotestosterone irreversibly. The aldo-keto reductase 3␣-hydroxysteroid dehydrogenase (3␣-HSD) inactivates dihydrotestosterone by reduction into 3␣-adiol. 17␤-HSD converts 3␣-adiol into androsterone, which undergoes glucuronidation and elimination. RoDH/CRAD may regenerate dihydrotestosterone by oxidizing 3␣-adiol. Secondarily, RoDH/CRAD may oxidize androsterone into androstanedione, which may undergo reduction into dihydrotestosterone by an as yet unknown 17␤-HSD.