cDNA Cloning and Characterization of acis-Retinol/3α-Hydroxysterol Short-chain Dehydrogenase*

We report a mouse cDNA that encodes a 317-amino acid short-chain dehydrogenase which recognizes as substrates 9-cis-retinol, 11-cis-retinol, 5α-androstan-3α,17β-diol, and 5α-androstan-3α-ol-17-one. Thiscis-retinol/androgen dehydrogenase (CRAD) shares closest amino acid similarity with mouse retinol dehydrogenase isozymes types 1 and 2 (86 and 91%, respectively). Recombinant CRAD uses NAD+ as its preferred cofactor and exhibits cooperative kinetics for cis-retinoids, but Michaelis-Menten kinetics for 3α-hydroxysterols. Unlike recombinant retinol dehydrogenase isozymes, recombinant CRAD was inhibited by 4-methylpyrazole, was not stimulated by ethanol, and did not require phosphatidylcholine for optimal activity. CRAD mRNA was expressed intensely in kidney and liver, in contrast to retinol dehydrogenase isozymes, which show strong mRNA expression only in liver. CRAD mRNA expression was widespread (relative abundance): kidney (100) > liver (92) > small intestine (9) = heart (9) > retinal pigment epithelium and sclera (4.5) > brain (2) > retina and vitreous (1.6) > spleen (0.7) > testis (0.6) > lung (0.4). CRAD may catalyze the first step in an enzymatic pathway from 9-cis-retinol to generate the retinoid X receptor ligand 9-cis-retinoic acid and/or may regenerate dihydrotestosterone from its catabolite 5α-androstan-3α,17β-diol. These data also illustrate the multifunctional nature of short-chain dehydrogenases and provide a potential mechanism for androgen-retinoid interactions.

The retinol (vitamin A) metabolite all-trans-retinoic acid modulates the transcription of multiple genes in diverse cells during embryogenesis and post-natally by activating three retinoic acid receptors, RAR␣, 1 RAR␤, and RAR␥ (1)(2)(3). An isomer of all-trans-retinoic acid, 9-cis-retinoic acid also binds with these three receptors with K d values in vitro similar to those of all-trans-retinoic acid. 9-cis-Retinoic acid, but not all-transretinoic acid, activates three other members of the steroid hormone/thyroid hormone/vitamin D/retinoid receptor superfamily, the RXRs ␣, ␤, and ␥ (4 -6). RARs and RXRs function as heterodimers. RXRs also serve as partners for other members of the superfamily, such as thyroid hormone, vitamin D, and peroxisome proliferator-activated receptors, and can modulate gene expression as homodimers (2,3). This multiplicity of receptors, receptor partners, and ligands suggests a mechanism for the pleiotropic effects of retinoids, but much depends on the loci of all-trans-retinoic acid and 9-cis-retinoic acid biosynthesis. Even though 9-cis-retinoic acid occurs in vivo, and has been identified as an endogenous ligand of RXR, its concentrations are lower than all-trans-retinoic acid, and it has been found only in a few tissues to date, compared with the ubiquitous distribution of all-trans-retinoic acid (7,8). Rapid conversion in vivo into the receptor inactive isomer 9,13-di-cis-RA most likely limits 9-cis-retinoic acid concentrations and effects (9,10).
All-trans-retinoic acid undergoes isomerization in cultured cells into 9-cis-retinoic acid (7) and pharmacological amounts of all-trans-retinoic acid dosed to rats produce small amounts of 9-cis-retinoic acid (11), but 9-cis-retinoic acid has not been detected in normal serum, and there is no evidence of enzymatic 9-cis-retinoic acid production from all-trans-retinoic acid (12,13). The availability of potential substrates, however, suggests that a route of 9-cis-retinoic acid biosynthesis could occur other than through all-trans-retinoic acid. Both 9-cis-retinol and 9-cis-␤-carotene are normal constituents of mammalian diets, including fresh and processed fruits and vegetables (14 -17) and accumulate in mammalian tissues, and the latter undergoes conversion into 9-cis-retinoids (18 -22). A 9-cis-RoDH could rely on these sources of 9-cis-retinol as a first step in the production of 9-cis-retinoic acid.
This work reports the cDNA cloning and expression, the mRNA tissue distribution, and the enzymatic characterization of a heretofore unknown SDR that catalyzes the conversion of 9-cis-and 11-cis-retinols into their respective aldehydes, but shows more efficient 3␣-hydroxyandrogen dehydrogenase activity than cis-retinoid catalytic activity. This SDR, herein referred to as CRAD, provides a means of generating 9-cis-* This work was supported by National Institutes of Health Grant DK47839. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF030513.
Isolation of a cDNA-Primers were designed from genomic sequences distinct from the corresponding sequences of RoDH-1: sense primer CCCAAGCTTGACCAGTAGTGCCAGAT (nucleotides 64 -80 of pBSK/ 1.1, i.e. 683-699 of the final cDNA ( Fig. 1); antisense primer CGCG-GATCCCCCAACTCTCCTAATTTC (nucleotides 576 -593 of pBSK/0.75, i.e. 1063-1080 of final cDNA) containing HindIII and BamHI, respectively (underlined). A probe was generated by PCR amplification of cDNA from a mouse liver gt10 library with these primers. The probe was labeled with 32 P by random priming and used to screen ϳ6 ϫ 10 5 plaques on nitrocellulose membranes at 42°C from the mouse cDNA library. A final wash was done at 65°C with 0.2 ϫ SCC. The first round of screening identified more than 200 plaques. The second and third rounds of screening were done by PCR at an annealing temperature of 55°C with the sense primer ATGTGGCTCTACCTGCTG, based on the nucleotide sequence of the first six amino acids of RoDH-1, and the antisense primer CGCGGATCCCCCAACTCTCCTAATTTC from nucleotides 576 to 593 of pBSK/0.75, containing a BamHI site (underlined). Of the 40 plaques screened, two were positive. Phage DNA from one was isolated and digested with EcoRI. Two fragments, 1.2 and 2 kb, were subcloned into pBSK to produce pBSK/CRAD1.2kb and pBSK/ CRAD2kb. These two fragments represent the cDNA, which has an internal EcoRI site. The inserts were sequenced in both directions.
Expression of CRAD-CHO-K1 cells (ATCC) were cultured in Ham's F-12 medium supplemented with 10% fetal calf serum. Cells were plated semiconfluently onto 100-mm tissue culture dishes 24 h before transfection. The coding region for CRAD was digested from pBSK/ CRAD1.2kb with EcoRI and ligated into pcDNA3 to produce pcDNA3/CRAD. The cells were transfected by using LipofectAMINE reagent (Life Technologies, Inc.) with pcDNA3/CRAD or with pcDNA3 (mock) and harvested 24 h later. The cell pellets were suspended in 10 mM Hepes, 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 (32).
Enzyme Assays-Unless noted otherwise, retinol dehydrogenase assays were done at 37°C in 0.25 ml of 50 mM Hepes, 150 mM KCl, and 1 mM EDTA, 1.6 mM NAD ϩ , pH 8. In some assays, egg yolk L-␣-phosphatidylcholine was added in 2 l of ethanol. Reactions were quenched with 0.1 ml of 0.1 M O-ethylhydroxylamine, incubated at room temperature for 10 min, and extracted with 2.5 ml of hexane. The retinoids were quantified by normal-phase high performance liquid chromatography eluted with hexane for 5 min followed by 15% tetrahydrofuran in hexane for 5 min, followed by 14% tetrahydrofuran in hexane for 2 min eluted at 2 ml/min, as described previously (23). Replicates averaged within 7% of their means. Recovery was quantitative, as determined by comparing the areas of graded concentrations of retinoid oximes injected directly onto high performance liquid chromatography to the areas of the same concentrations after processing through the extraction procedure. The limit of detection was ϳ1 pmol and linear detection ranged from ϳ1 pmol to at least 1200 pmol. Retinal isomers were detected at 370 nm; retinol isomers were detected at 325 nm. 11-cis-Retinoids undergo isomerization into more stable isomers during incubation and extraction; therefore the sum of the retinal isomers recovered from the 11-cis-retinol incubation was used to determine the rate of its conversion into 11-cis-retinal (15).
Ethanol and propanol dehydrogenase assays were done with 33 mM substrate in 0.1 M glycine/NaOH, pH 10, and also in the same buffer used to measure retinoid activity (pH 8.0), both in a total volume of 1 ml. The absorbance was read at 340 nm during a 10-min incubation with 3 mM NAD ϩ and 20 -32 g of protein from an 800 ϫ g supernatant of transfected CHO cells.
Kinetic data were obtained under initial velocity conditions (linear rates with protein and time; i.e. up to 6 and 2 g of 800 ϫ g supernatant protein, and for 20 and 10 min, for retinols and sterols, respectively) and were fit with the nonlinear regression program Enzfitter using simple weighing (33).
Northern Blot-Northern blots were done with the mouse Multiple Tissue Northern blot which provides 2 g of poly(A ϩ ) RNA per lane on a Nylon membrane (CLONTECH). The probe was generated by PCR amplification of pBSK/CRAD2kb with the sense primer TTCAGTTC-CTGGTGGTGA (nucleotides 1476 -1493 of the final cDNA) and the antisense primer GGAGTCAGGCATTTATGG (nucleotides 2082-2099). The probe was labeled with 32 P by random priming. Prehybridization was done in 10 ml of ExpressHyb TM Hybridization solution (CLON-TECH) at 68°C for 30 min. Hybridization was done for 1 h in the same solution containing 2 ϫ 10 6 cpm of probe/ml. The final wash was done at 50°C with 0.1 ϫ SSC, 0.1% SDS. Signals were visualized with a Bio-Rad GS-505 Molecular Imager System.
RNase Protection Assays-A CRAD-specific probe was amplified by PCR from pBSK/CRAD1.2kb with the sense primer ATGGAGCAT-GCTCTGACT (nucleotides 846 -863) and the T7 primer CGGGATAT-CACTCAGCATAATG of pBluescript-II SK(Ϯ). The PCR product was digested with EcoRI to produce a 341-bp probe (nucleotides 846 -1187), which was subcloned into pBluescript-II SK(Ϯ) and linearized with HindIII. A 32 P-labeled antisense probe was transcribed with T3 RNA polymerase (Promega) for 1 h at 37°C in 10 mM dithiothreitol, 0.5 mM each ATP, CTP, and GTP, 50 M UTP, and 50 Ci of UTP (800 Ci/ mmol). The DNA template was removed by DNase I digestion. The transcript was gel purified with a 5% polyacrylamide, 8 M urea gel. A 250-bp KpnI/XbaI fragment of mouse actin was used to generate an antisense probe for ␤-actin mRNA. RNase protection assays were done with the Hybspeed TM RPA kit (Ambion) following the manufacturers instructions with modifications as follows. For each experimental tube, labeled probes (9 ϫ 10 4 cpm) were mixed with 30 g of total RNA. The probe and sample RNAs were co-precipitated at Ϫ20°C for 30 min with 2.5 volumes of ethanol and 0.1 volume of 5 M ammonium acetate. The supernatants from a 15-min centrifugation with a microcentrifuge were removed and the pellets were air dried for 5 min at room temperature. To each pellet 10 l of hybridization buffer preheated to 95°C were added with good mixing. Samples were heated for 3.5 min at 95°C and then hybridized at 68°C for 30 min. A 100-l aliquot of RNase A/T1 mixture diluted 1/100 was allowed to digest the unhybridized probe 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 a polyacrylamide, 8 M urea gel and run at ϳ180 volts for 2 h in 1 ϫ TBE. The protected fragments were quantified as described above.

RESULTS
cDNA and Amino Acid Sequence-To isolate genomic clones of RoDH isozymes 1-3, an EMBL3 Sp6/T7 mouse genomic DNA library was screened with a probe consisting of nucleotides 1-788 of RoDH-1, i.e. a span homologous with RoDH-2 and RoDH-3. The deduced amino acid sequence from exons 2 through 4 had no better than 80 and 82% amino acid identities with rat RoDH-1 and rat RoDH-2, respectively. Because proteins that mediate retinoid function, such as the retinoid-binding proteins and the retinoid receptors, retain very high interspecies conservation (2, 34), we suspected that the clone represented a distinct SDR, rather than the mouse homolog of a known rat RoDH isozyme. To test this suspicion, a mouse liver cDNA library was used as template for PCR with primers from rat RoDH-1 (sense, nucleotides 295-313; antisense, nucleotides 1308 -1326). The mouse PCR product had 99% nucleotide and Ͼ98% amino acid identity with rat RoDH-1. Similarly, PCR with primers to rat RoDH-2 (sense, nucleotides 229 -246; antisense, nucleotides 1207-1224) amplified a product with 99% nucleotide and amino acid identity to rat RoDH-2 (data not shown). These two products encode the mouse homologs of RoDH-1 and RoDH-2, and therefore the genomic clone represented a heretofore unknown mouse SDR.
To obtain a cDNA clone of the mouse genomic clone, primers from the genomic sequence were used for PCR amplification of a probe from a mouse liver cDNA library. The plaques identified with this probe were further screened by PCR with a sense probe from RoDH-1 and an antisense probe from the genomic DNA. This identified a 3.2-kb cDNA which was subcloned in two parts, 1.2 (pBSK/CRAD1.2kb) and 2 kb (pBSK/CRAD2kb), because of an internal EcoRI site. The shorter fragment contained a single open reading frame with a complete coding region, whereas the longer fragment contained 3Ј-untranslated sequence. The open reading frame in pBSK/CRAD1.2kb predicts a polypeptide with a calculated molecular mass of ϳ35 kDa, i.e. in the 25-35 kDa range of SDR (Fig. 1). The amino acid sequence deduced from pBSK/CRAD1.2kb had other features consistent in SDR. Twenty-three amino acids are conserved in ϳ70% of SDR: 19 of these are conserved in the protein encoded by pBSK/CRAD1.2kb. This includes the cofactor binding residues, Gly 36 (X) 3 GlyXGly, the catalytic residues, Tyr 176 -(X) 3 -Lys, and the conserved sequence of unknown function, Leu 109 -X-Asn-Asn-Ala-Gly. One of the three non-identical amino acids represents a conservative substitution, I159V, also in the three RoDH isozymes and in 11-cis-RoDH. The two other substitutions are not conservative: A191R, also in the three RoDH isozymes and in 11-cis-RoDH; and D107W present in the RoDH isozymes, but not in 11-cis-RoDH, which has Phe 107 . Both of these substitutions occur in other SDR (30, 31, 35). The first 103 N-terminal amino acids of RoDH-1, -2, and -3 are identical. Compared with these three, the product of pBSK/CRAD2kb has only seven substitutions in these amino acids versus 44 for 11-cis-RoDH. The closest amino acid similarity between the pBSK/CRAD1.2kb encoded protein and other SDR occurs with RoDH isozymes, with less similarity to both bovine and human 11-cis-RoDH, which also recognizes 9-cis-retinol (Table I) (28,29,36,37). Even less similarity occurs with rat ␤-hydroxybutyrate and mouse 11␤-hydroxysteroid dehydrogenases (35,38). SDR usually have two or less cysteine residues, but some mammalian SDR have four: rat 11␤-hydroxysteroid dehydrogenase (38); human (R)-3-hydroxybutyrate dehydrogenase (39); human 17␤-hydroxysteroid dehydrogenase, type I (40); human 15-hydroxyprostaglandin dehydrogenase (41). The rat liver D-␤-hydroxybutyrate dehydrogenase (35), the RoDH isozymes, and CRAD represent SDRs with six cysteine residues, whereas bovine 11-cis-RoDH has seven (28,29).
Retinoid turnover in CHO cells was monitored to determine whether the different amounts of retinals isolated reflected the inherent activity of CRAD or endogenous activity of CHO cells or a combination. The 800 ϫ g supernatant of mock-transfected CHO cells had no endogenous activity that metabolized 9-cisretinol: the substrate was recovered unchanged over the course of a 15-min incubation (Fig. 2). 9-cis-Retinol also was recovered unchanged from the supernatant of cells transfected with pcDNA3/CRAD when NAD ϩ was omitted from the incubation. Addition of NAD ϩ to the CHO supernatant was obligatory for 9-cis-retinol metabolism. Under the same conditions, 5 M alltrans-retinol was recovered unchanged at the end of a 15-min incubation. 9-cis-Retinal (400 pmol) also was recovered unchanged. These data demonstrate that use of the 800 ϫ g supernatant of CHO cells transfected with pcDNA3/CRAD reflects the inherent activity of CRAD and not differential catalysis by enzymes endogenous to CHO cells.
CRAD exhibited cooperative kinetics for 9-cis-retinol with a K 0.5 value of 5.4 Ϯ 0.5 M (mean Ϯ S.D., n ϭ 3) and a Hill coefficient of 1.7 Ϯ 0.2 (mean Ϯ S.D., n ϭ 3) (Fig. 3, top panel). CRAD recognized 11-cis-retinol as substrate with a K 0.5 value of 7.6 and a Hill coefficient of 2.8 Ϯ 0.5. Thus, the V m /K 0.5 values indicate that CRAD shows little preference between 9-cis-retinol and 11-cis-retinol. These characteristics contrast with those of recombinant RoDH-1 and RoDH-2 which exhibited Michaelis-Menten kinetics for all-trans-retinol with K m values of 0.9 and 2 M, respectively (25, 26). Concentrations of 9-cis-retinol in liver have not been quantified, but are less than  those of the ϳ2-5 M of all-trans-retinol. 2 The impact of this K 0.5 on 9-cis-retinal synthesis, however, depends on the exact concentrations of 9-cis-retinol, their range during variable dietary and humoral conditions, and the potential for allosteric modulation of CRAD.
Subcellular Fractionation of CRAD Expressed in CHO Cells-Centrifugation of the 800 ϫ g supernatant of transfected CHO cells for 30 min at 10,000 ϫ g partitioned the recovered CRAD activity, measured with 9-cis-retinol, between the pellet (30%) and the supernatant (70%). Centrifugation of the 10,000 ϫ g supernatant at 100,000 ϫ g for 2 h distributed most of the recovered activity into the microsomal pellet (91%), with relatively little in the cytosolic fraction (9%), indicating that CRAD associates with membranes.
Modulators of CRAD Activity-Various agents affect the activities of recombinant RoDH-1 and -2 (23)(24)(25). For comparison, the effects of these agents were examined on CRAD activity (Table III). Ethanol (1 mM) enhances the activity of recombinant RoDH-1 and RoDH-2 by ϳ50 -60%, but has no significant effect on CRAD activity. Moreover, neither ethanol nor propanol was metabolized by CRAD either at pH 8 or pH 10 (data not shown). Recombinant RoDH-1 and -2 require 2 mM phosphatidylcholine for maximum activity, which stimulates RoDH-1 7-fold and RoDH-2 3-fold. In contrast, 2 mM phosphatidylcholine, added in 2 l of ethanol to give a final ethanol concentration of 137 mM, does not stimulate CRAD activity. Carbenoxolone, the steroidal aglycone of the licorice-derived compound glycyrrhizin, inhibits SDRs, such as rat 11␤-hydroxysteroid dehydrogenase, 15-hydroxyprostaglandin dehydrogenase, RoDH-1, and RoDH-2 (23)(24)(25)(42)(43)(44). It also inhibits CRAD potently. Phenylarsine oxide, which forms covalent adducts between spatially-proximal sulfhydryl groups (45), inhibits CRAD, as it does RoDH-1 and RoDH-2. 4-Methylpyrazole, a potent inhibitor of the medium-chain alcohol dehydrogenase ADH class I and a moderate inhibitor of alcohol dehydrogenase class II, inhibits CRAD. 4-Methylpyrazole has little effect on RoDH-1 and -2, but inhibits cytosolic NAD ϩ -supported alltrans-retinoic acid synthesis with the complex cellular retinolbinding protein/all-trans-retinol as substrate (23). The NAD-dependent, cytosolic holocellular retinol-binding protein recognizing activity has features of an SDR/RoDH, i.e. it recognizes holocellular retinol-binding protein as substrate and apocellular retinol-binding protein and carbenoxolone as inhibitors, and can be distinguished from liver alcohol dehydrogenase isozymes by pI values. It is, however, distinct from CRAD, because the later does not recognize the complex cellular retinol-binding protein/all-trans-retinol as substrate (data not shown) and because the two activities associate with different subcellular fractions. Finally, these data show the preference of CRAD for NAD ϩ as cofactor, rather than NADP ϩ .
Tissue Distribution of CRAD mRNA-Northern blot hybridization revealed intense expression of CRAD mRNA in kidney and liver (Fig. 4). Each tissue expressed multiple isoforms of mRNA (kidney, 4.4, 3.5, and 2.7 kb; liver, 4.4, 3.5, 3.0, and 2.7 kb). No signals were observed by Northern blotting in heart, brain, spleen, lung, skeletal muscle, or testis. The more sensitive RNase protection assays also were done to determine whether low mRNA expression occurs, not necessarily detectable by Northern blot hybridization. These assays showed CRAD mRNA expression in multiple mouse tissues (relative intensity normalized to the ␤-actin signal): kidney (100) Ͼ liver (92) Ͼ small intestine (9) ϭ heart (9) Ͼ retinal pigment epithelium ϩ sclera (4.5) Ͼ brain (2) Ͼ retina ϩ vitreous (1.6) Ͼ spleen (0.7) Ͼ testis (0.6) Ͼ lung (0.4). The anticipated 341-bp protected fragment was not observed in the iris, cornea, or lens. The intensity of expression of CRAD mRNA correlates well with the report that kidney has relatively high concentrations of 9-cis-retinoic acid (30 ng/g wet weight) and liver has somewhat lower concentrations (4 ng/g wet weight) (8), and with the Northern data which detected mRNA only in kidney and liver. A 210-bp fragment was noted in all tissues that showed the 2 X. Chai, Y. Zhai, and J. L. Napoli, unpublished results.

TABLE III
Characteristics of CRAD transiently-expressed in CHO cells CRAD activity was assayed with 5 M 9-cis-retinol and 2 g of protein for 10 min and with 0.1 M 3␣-adiol for 5 min with 50 ng of protein. The data for 9-cis-retinol are the means Ϯ S.E. of 6 -11 replicates from three separate transfections. The data for 3␣-adiol are the means Ϯ S.D. of triplicates from a fourth transfection. The two entrees without errors indicate averages of duplicates. NADP ϩ was added in the absence of NAD ϩ .  anticipated 341-bp fragment, but was absent from the three tissues that lacked the 341-bp fragment. In intestine and retina/vitreous the intensity of the 210-bp fragment was equal to or greater than the 341-bp fragment. This seems consistent with a CRAD isoform.

DISCUSSION
This article describes cloning and expression of a cDNA which encodes a novel SDR that recognizes both retinols and sterols as substrates. This SDR increases the total cloned SDRs reported which catalyze retinoid metabolism to include, in addition to the current mouse CRAD, the three rat and/or mouse RoDH isozymes (types 1, 2, and 3) (24 -26), a human "RoDH" (27), the human and bovine 11-cis-RoDHs (28,29,36,37). We have also cloned and expressed another distinct mouse SDR, which recognizes cis-retinols and sterols, with 82% amino acid identity with CRAD. 2 These observations indicate that SDRs have considerable roles in retinoid metabolism, just as they have in steroid hormone metabolism. Thus, steroids and retinoids seem to share families of metabolic enzymes, ligandactivated receptors (2,3,6), and cellular binding-proteins (46,47).
Dihydrotestosterone promotes prostate epithelial cell growth, whereas all-trans-retinoic acid and its isomers inhibit prostate epithelial cell growth (48,49). Each hormone seems to work through modulating the receptor(s) of the other. Androgens decrease the mRNA of RAR␣ ϳ5-fold in prostate epithelia and 15-20-fold in seminal vesicles (but increase it 2-fold in kidney) (50), whereas RA decreases androgen receptor binding 3-fold (48,51). In addition, all-trans-retinoic acid decreases serum concentrations of dihydrotestosterone, 3␣-adiol, and androsterone significantly, and seems to cause a metabolic deviation away from the 5␣-path in liver as well (48,52). Because of the dual androgen/retinoid activity of SDRs, they could be additional targets of retinoid-sterol interaction, either directly through substrate and/or allosteric or other interactions, or indirectly through any facet of gene expression or protein turnover via RARs, RXRs, and/or the androgen receptor. The observance of potentially dual androgen/retinoid substrate SDRs provides opportunity for additional insight into the physiological impact of retinoids on male reproduction, and extends the insights afforded by recognition of the requirement for functional retinoid receptors during spermatogenesis (53,54), and detection of a epididymal retinoid-binding protein with high affinity for both all-trans-retinoic acid and 9-cis-retinoic acid (55).
Two pathways of testosterone metabolism have been observed in prostate cells (56). Prostate epithelial cells make and maintain dihydrotestosterone as the predominant testosterone product. Dihydrotestosterone serves as the major biologically active androgen in prostate (testosterone itself serves this role in testis) (57). Prostate stromal cells convert dihydrotestosterone into the inactive 3␣-adiol and then into the inactive androsterone. The androgen 3␣-hydroxydehydrogenase activities of RoDH-1, as shown by Biswas and Russell (27), and CRAD, shown here, convert 3␣-adiol (an inactive intermediate between dihydrotestosterone and the inactive androsterone) back into dihydrotestosterone, suggesting that RoDH-1 and CRAD could help maintain adequate transcriptionally-active androgen levels in prostate epithelial cells. The potential importance of dihydrotestosterone synthesis from 3␣-adiol is illustrated by detection of dihydrotestosterone in castrated and functionallyhepatectomized rats and the retention of androgen activity in the presence of 5␣-reductase inhibitors (57). Indeed, although 3␣-adiol does not bind to the androgen receptor, it stimulates prostate growth in vivo and in prostate organ culture, consistent with metabolism into dihydrotestosterone (58 -60).
From its retinoid isomer recognition, CRAD also could function in either one or both of two retinoid metabolic paths. Its 9-cis-RoDH activity, along with its most intense mRNA expression in kidney and liver, but widespread expression throughout many tissues, implies that CRAD participates in an enzymatic pathway that generates 9-cis-retinoic acid from 9-cis-retinol (Fig. 5). 9-cis-Retinol and 9-cis-carotenoids occur in diets, 9-ciscarotenoids undergo conversion into 9-cis-retinoids, and both 9-cis-carotenoids and 9-cis-retinol accumulate in animal tissues (14 -21, 61). CRAD could initiate the pathway by catalyzing the conversion of 9-cis-retinol into 9-cis-retinal. The 9-cis-retinal produced would undergo conversion into 9-cis-retinoic acid by retinal dehydrogenase type I, which catalyzes dehydrogenation of all-trans-and 9-cis-retinal into their respective aldehydes with equivalent efficiencies (62,63). Like CRAD, many tissues express the mRNA of retinal dehydrogenase type I, including relatively intense expression in liver and kidney (64). The substrate specificity and the widespread expression of CRAD and retinal dehydrogenase type I mRNA, considered with the potential substrate availability, are con- FIG. 4. Distribution of CRAD in mouse tissues. Top, Northern blot hybridization of CRAD. The probe was hybridized at high stringency as detailed under "Materials and Methods" with a commerciallyavailable blot from a gel of mouse poly(A ϩ ) RNA: 1, heart; 2, brain; 3, spleen; 4, lung; 5, liver; 6, skeletal muscle; 7, kidney; 8, testis. Bottom, RNase protection assays were done as described under "Materials and Methods" on RNA prepared from the tissues of 1-month-old male mice: 1, probe; 2, yeast RNA; 3, brain; 4, heart; 5, intestine; 6, kidney; 7, liver; 8, lung; 9, spleen; 10, testis; 11, iris; 12, cornea; 13, retina/vitreous; 14, retinal pigment epithelium/sclera; 15, lens. These data were normalized to the signals produced by the mouse ␤-actin probe.
The human 11-cis-RoDH, or a very closely-related isozyme, has been re-cloned and shown to convert 9-cis-retinol into 9-cisretinal (37). This observation was based on a one-point enzyme assay with 10 M substrate for an unspecified amount of time. No non-retinoid substrates were tested, nor was 11-cis-retinol re-evaluated side by side with 9-cis-retinol. The assays included phosphatidylcholine, but no indication was given of its effect on reaction rates. Without more insight its efficiency (i.e. K 0.5 and rate values) for 9-cis-retinol and whether it recognizes other substrates, its potential contribution to 9-cis-retinoic acid biosynthesis is difficult to assess.
The 11-cis-retinol dehydrogenase activity of CRAD technically could provide 11-cis-retinal for use as a rhodopsin chromophore. At first consideration, expression of CRAD in the retinal pigment epithelium would seem to support this notion. The mechanism of generating 11-cis-retinoids argues against this probability, however. Concerted hydrolysis-isomerization of all-trans-retinyl esters in the retinal pigment epithelium generates 11-cis-retinol for use in rhodopsin (65). This reaction has not been demonstrated in other tissues. Moreover, 11-cisretinoids do not occur outside of the eye. 11-cis-Retinoids are unstable and rapidly undergo thermal, chemical, and lightcatalyzed isomerization into other geometric isomers. Nor does CRAD seem to be co-specific with either of the two known dehydrogenases associated with the visual cycle. The 11-cis-RoDH expressed in the retinal pigment epithelium recognizes both 11-cis-and 13-cis-retinol as substrates, but neither alltrans-nor 9-cis-retinol (66). The retinol dehydrogenase activity of the rod outer segments requires NADP ϩ as cofactor and recognizes only all-trans-retinol as substrate and not 9-cis-, 11-cis-, or 13-cis-retinol (67). All these considerations weaken the expectation that CRAD contributes to the production of 11-cis-retinal. Moreover, in the eye, retinoids do not function solely as cofactors in the visual cycle; they also regulate gene expression. The expression of CRAD mRNA in the eye may be related to widespread use of 9-cis-retinoic acid, rather than to the requirement for 11-cis-retinal in the visual cycle, but further research will have to address this issue.
In summary, CRAD could be serving as either a sterol 3␣hydroxydehydrogenase or a cis-retinol dehydrogenase or both. Likely, however, the exact function of CRAD will vary with its precise spatial-temporal expression loci, substrate availability, and humoral influences.