Cloning of a cDNA for a Second Retinol Dehydrogenase Type II EXPRESSION OF ITS mRNA RELATIVE TO TYPE I*

A retinol dehydrogenase, RoDH(I), which recognizes holo-cellular retinol-binding protein (CRBP) as substrate, has been cloned, expressed, and identified as a short-chain dehydrogenase/reductase (Chai, X., Boerman, M. H. E. M., Zhai, Y., and Napoli, J. L. (1995) J. Biol. Chem. 270, 3900–3904). This work reports the cloning and expression of a cDNA encoding a RoDH isozyme, RoDH(II). The predicted amino acid sequence verifies RoDH(II) as a short-chain dehydrogenase/reductase, 82% identical with RoDH(I). RoDH(II) recognized the physiological form of retinol as substrate, CRBP, with a K m of 2 m M . Similar to microsomal RoDH and RoDH(I), RoDH(II) had higher activity with NADP rather than NAD, was stimulated by ethanol and phosphatidyl cho- line, was not inhibited by the medium-chain alcohol dehydrogenase inhibitor 4-methylpyrazole, but was in- hibited by phenylarsine oxide and the short-chain dehydrogenase/reductase inhibitor carbenoxolone. Northern blot analysis detected RoDH(I) and RoDH(II) mRNA only in rat liver, but RNase protection assays revealed RoDH(I) and RoDH(II) mRNA in kidney, lung, testis, and brain. These data indicate that short-chain dehydrogenases/reductase isozymes expressed tissue-distinctively catalyze the first step of retinoic acid biogenesis from

Metabolic activation of retinol provides the hormone alltrans-retinoic acid (RA). 1 RA produces a variety of biological responses by modulating the expression of genes that regulate the state of differentiation or entry into apoptosis of diverse cell types in numerous organs (1)(2)(3)(4)(5). A model of RA biosynthesis postulates that the enzyme(s) that catalyze RA synthesis physiologically recognize as substrate the predominant form of retinol in vivo, viz. retinol bound within CRBP (6). The CRBP concentration exceeds that of retinol ϳ(7 versus ϳ5 mM, respectively (7)), and CRBP has a high affinity interaction with retinol (K d ϳϭ 0.1-1 nM), much higher than substrate-enzyme interactions (8,9). Binding of retinol within CRBP shields the prohormone from the cellular environment and would confer specificity on RA biosynthesis by restricting access of retinol to enzymes capable of recognizing and interacting with the CRBP-retinol "cassette." This would prevent opportunistic oxidation by dehydrogenases/oxidases with inexact substrate tolerances, thereby contributing to precise spatial-temporal control over RA biogenesis. CRBP binding would also protect retinol from nonenzymatic oxidation and cells from the membrane-altering faculty of unbound retinol (10 -12). A pathway of RA biosynthesis consistent with this hypothesis involves as the first and rate-limiting step production of retinal in microsomes with holo-CRBP as substrate, i.e. RoDH (13,14). Retinal generated in microsomes from holo-CRBP by RoDH supports RA biosynthesis by cytosolic retinal dehydrogenases (15).
A microsomal RoDH has been partially purified, its active site has been identified with a 34-kDa polypeptide by chemical cross-linking with holo-CRBP, and its cDNA has been cloned and expressed (16 -18). This RoDH, hereafter termed RoDH(I), belongs to the family of short-chain dehydrogenase/reductase (19). By Northern blot analysis, mRNA expression of RoDH(I) was detected only in rat liver, despite the well established occurrence of CRBP-recognizing RoDH activity in multiple tissues. These results suggested occurrence of multiple RoDHs.
This work reports the cDNA cloning and expression of a second RoDH, RoDH(II), shows that it is a previously unknown short-chain dehydrogenase/reductase that can catalyze the first step in RA synthesis with CRBP as substrate, and compares the expression of RoDH(II) mRNA in rat tissues with that of RoDH(I) by RNase protection assays.

MATERIALS AND METHODS
Library Screening-A rat liver gt11 cDNA library (Clontech) was screened through three rounds with probe A (nucleotides 653-975 of RoDH(I)), as described (18). DNA from one of the three plaques obtained was cloned into P-Direct to provide p-DirectRo3, which contained a 1.5-kilobase cDNA insert distinct from RoDH(I) but with no initiation codon. Probe B (nucleotides 587-909 of the final product) was prepared from p-DirectRo3 by AvaII digestion and used to identify 36 plaques by screening the same library through three rounds. The inserts of these plaques were amplified by PCR and analyzed by Southern blot with a synthetic 32-base pair oligonucleotide (probe C, nucleotides 974-1005 of the final cDNA). Six PCR products hybridized at 42°C to probe C and were washed at 68°C in 0.5% SDS in 0.1 ϫ SCC (SSC ϭ 0.15 M NaCl and 15 mM sodium citrate). The longest was cloned into pBluescript to provide PBSK/RoDH(II). The insert in PBSK/RoDH(II) was sequenced by dideoxy chain termination with Taq DNA polymerase.
Expression of RoDH-The coding region of RoDH(II) of PBSK/ RoDH(II) was amplified with the sense primer 5Ј-CGCGGATCC-CCTCTGCTTGTCTTCTAC-3Ј (nucleotides 206 -223) and the antisense primer 5Ј-CGGAATTCCTCCCTAACACTGTACCA-3Ј (nucleotides 1233 to 1216) containing BamHI and EcoRI cleavage sites, respectively (underlined). The PCR product was ligated into pcDNA3 (Invitrogen) to yield pcDNA3/RoDH(II). pcDNA3/RoDH(II) was transfected by calcium phosphate/DNA precipitation into semi-confluent P19 cells as described (18). Mock transfections were done with pcDNA3. 24 h after transfection cells were harvested, the pellets were suspended in 10 mM Hepes, * This work was supported by National Institutes of Health Grant DK36870. 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 /EMBL Data Bank with accession number(s) RODHII 1 The abbreviations used are: RA, all-trans-retinoic acid; CRBP, cellular retinol-binding protein type I; PAO, phenylarsine oxide; PCR, polymerase chain reaction; 10Ksup, 10,000 ϫ g supernatant of a cell homogenate.
Northern Blot Analysis-Northern blot analysis was done as described (18) with 5 mg of poly(A) ϩ mRNA hybridized at 65°C for 16 h to a RoDH(II) probe of the 841 nucleotide FokI/SphI product (nucleotides 1148 -1988) from the 3Ј-untranslated region of PBSK/RoDH(II). The blot was reprobed with a glyceraldehyde dehydrogenase cDNA probe.
RNase Protection Assays-The RoDH(I)-specific probe was obtained by amplifying a 341-base pair fragment from p-DirectRo2 (18) by PCR with the sense primer 5Ј-CGCGGATCCCTTCTCAGACTCCCTCA-3Ј (nucleotides 849 -865) and the antisense primer 5Ј-CGGAATTCGTGG-GAAGGTAGCTCATG-3Ј (nucleotides 1190 -1173) containing BamHI and EcoRI cleavage sites, respectively (underlined). The RoDH(II)-specific probe was obtained by amplifying a 341-base pair fragment from PBSK/RoDH(II) by PCR with the same sense primer used for RoDH(I) and the antisense primer 5Ј-CGGAATTCTCCCGTAGGTGTTCTT-TCCA-3Ј (nucleotides 1140 -1123) containing an EcoRI cleavage site (underlined). The fragments were subcloned into pcDNA3 in the antisense orientations, and the plasmids were linearized with BamHI. 32 P-labeled antisense riboprobes were transcribed with SP6 RNA polymerase (Ambion) for 1 h at 37°C in 10 mM dithiothreitol, 0.5 mM each ATP, CTP, and GTP, 50 mM UTP, and 50 mCi [␣-32 P]UTP. DNA templates were removed with DNaseI. The 450-base pair transcripts, which included 109 base pairs of plasmid, were purified with a 5% polyacrylamide/8 M urea gel. A 250-base pair KpnI and XbaI fragment of mouse ␤-actin was used to generate an antisense probe for ␤-actin mRNA. RNase protection assays were done with a ribonuclease protection kit (Ambion). Probes (1.2 ϫ 10 6 cpm) were co-precipitated by 0.5 M ammonium acetate with total RNA, extracted with guanidinium thiocyanate/ phenol/chloroform from the tissues of adult male Sprague-Dawley rats (ϳ30 g for extra-hepatic tissues and for ␤-actin in all tissues and ϳ15 mg for liver with RoDH probes). The precipitates were resuspended in 20 ml of 80% deionized formamide, 100 mM sodium citrate, 300 mM sodium acetate, and 1 mM EDTA, pH 6.4, and were incubated at 45°C for 20 h. The same amounts of probes were hybridized with 10 mg of yeast RNA as control. After digestion with RNase A (2.5 units/ml) and RNase T1 (100 units/ml) for 30 min at 37°C, the protected fragments were resolved on 5% polyacrylamide/8 M urea gels and visualized at Ϫ80°C overnight with one intensifying screen. Quantification was done with a Bio-Rad Model GS-670 densitometer.

RESULTS AND DISCUSSION
cDNA and Amino Acid Sequence-During screening of a cDNA library with a probe from RoDH(I), a 1.5-kilobase partial cDNA was identified with sequence distinct from RoDH(I). A probe from this cDNA was used to re-screen the library. The longest clone identified ϳ(2 kilobase) was subcloned to create PBSK/RoDH(II) and was sequenced in both directions to reveal a cDNA with a nucleotide sequence distinct from that of RoDH (I). A single open reading frame in PBSK/RoDH(II) predicts an RoDH(II) of 317 amino acids with a calculated molecular mass of ϳ35 kDa; the same as RoDH(I) (Fig. 1). Of the 23 amino acid residues conserved in Ͼ70% of the extended short-chain dehydrogenase/reductase family, 21 have been conserved in RoDH(II) (19). These include the residues typical of an shortchain dehydrogenase/reductase: the G(X) 3 GXG cofactor binding site (Gly 36 ), N-terminal to the active site; the sequence LXNNAG (Leu 109 ); and the active site Y(X) 3 K (Tyr 176 ), C-terminal to the cofactor binding site. One of the two substitutions, D107W, represents a common substitution in short-chain dehydrogenase/reductase; the second, A191R, also occurs in the short-chain dehydrogenase/reductases 11␤-hydroxysteroid and D-␤-hydroxybutyrate dehydrogenases of rat (23,24). RoDH(II) differs from RoDH(I) in 57 amino acid residues; most are nonconservative substitutions. Seven of these had been verified in RoDH(I) by microsequencing (18). Yet the predicted primary sequence of RoDH(II) diverges less from RoDH(I) compared with a bovine 11-cis-retinol dehydrogenase, a short-chain dehydrogenase/reductase expressed only in retinal pigment epithelium (25), or to other rat short-chain dehydrogenase/reductases (Table I). A second Y(X) 3 K pattern 15 residues closer to the C terminus from the probable active site Y(X) 3 K, presents a curious feature of both RoDH(I) and RoDH(II). The first Y(X) 3 K most likely serves as the active site because of its position in the primary sequence and because it resembles the Y(C/G/S)(A/I/S)(T/S)K sequence of other short-chain dehydrogenase/reductase active sites (26).
Secondary Structure Predictions-RoDH(II) and RoDH(I) have identical N-terminal sequence through residue 103. The first 18 amino acids have an average hydropathy of ϳ1.6 (27), consist mostly of helix (28), and are juxtaposed to four of the most hydrophilic amino acids, RERK (Fig. 2). This probably serves a membrane-anchoring function, with the positioning of hydrophobic amino acids next to hydrophilic amino acids fastening RoDH in the membrane. Two other hydrophobic sections of RoDH(II) seem long enough to span membranes: the 22 residues 131 through 152 and the 18 residues 160 through 177. The former has an average hydropathy of ϳ1.2, and the latter, which includes the putative active site, has an average hydropathy of ϳ0.8. Presumably, because of insufficient hydrophobicity, neither spans a membrane. Thus, RoDH(II) seems to lack transmembrane helices similar to RoDH(I), a microsomal enzyme, and human heart (R)-3-hydroxybutyrate dehydrogenase, an inner mitochondrial membrane short-chain dehydrogenase/ reductase (29). These secondary structure predictions suggest anchoring of both RoDHs to the endoplasmic reticulum through N-terminal sequence, with the bulk of the polypeptide extending into the cytoplasm, accessible to CRBP.
Transient Transfection in P19 Cells-RoDH(II) was expressed transiently in P19 cells to determine its catalytic characteristics. In two different transfections, mock transfected cells had no RoDH activity in the absence of phosphatidyl choline and low activity in its presence (Fig. 3, bars 1-4).  7 and 9). Carbenoxolone and PAO inhibited RoDH(II) (compare lane 7 with lanes 10 and 11, respectively). PAO inhibits by forming covalent heterocyclic adducts between the reagent and spatially proximal sulfhydryl groups (30 -32). The six cysteine residues of RoDH(II) and RoDH(I) occur in the same positions, with Cys 37 in the putative cofactor binding site and Cys 177 in the putative active site. Binding of PAO to either one or both of these and another cysteine residue close in the secondary or tertiary structure could inhibit RoDH. The steroidal aglycone of glycyrrhizin, carbenoxolone, inhibits short-chain dehydrogenase/reductase besides RoDH(II) and RoDH(I), including 11␤-hydroxysteroid dehydrogenase (33,34).
The average K m value of RoDH expressed in the P19 cell 10Ksup for holo-CRBP was 2 mM. This value was obtained from two transfections: 1.6 Ϯ 0.2 and 2.4 Ϯ 0.4, each determined by fitting data with the nonlinear regression program Enzfitter (35) (Fig. 4). This K m compares well with the previously determined values for holo-CRBP of 1.6 mM for rat liver microsomal RoDH, 0.6 mM for partially purified RoDH(I), and 0.9 mM for recombinant RoDH (I) (13, 16, 18).
Tissue Distribution of RoDH mRNA-Northern blot analyses of RoDH(II) revealed a 1.8-kilobase mRNA in liver but did not detect mRNA in brain, kidney, lung, or testis (data not shown), just as had occurred with RoDH(I) (18). RNase protection as- The first line of amino acid sequence immediately below the nucleotide sequence shows the predicted amino acid sequence of RoDH(II). The underlined residues in this line identify the 21 amino acids in RoDH(II) conserved in Ͼ70% of the 57 members of the extended short-chain dehydrogenase/reductase superfamily (19). The second line of amino acid sequence represents RoDH(I). The blanks indicate residues identical to RoDH(II). The underlined residues in this second line are those that have been determined by microsequencing of RoDH(I) (18). Retinol Dehydrogenase Type II says of RoDH(I) and RoDH(II), however, showed wider tissue distribution of both mRNAs ( Fig. 5; Table II). Liver was the major site of expression of both RoDHs. Expression of RoDH(I) in the extra-hepatic tissues screened was Ͻ2% of liver. RoDH(II), in contrast, had relatively abundant expression in kidney and more abundant expression in brain and lung than RoDH(I). Testis had equivalent expression of RoDH(II) and RoDH(I). RNase protection assays under high stringency conditions revealed many protected fragments that could not be rationalized from the nucleotide sequences of RoDH(I) and RoDH(II) and the sequences of the probes, consistent with occurrence of closely related mRNAs, possibly from additional isozymes of RoDH. This expression of RoDH mRNA among multiple tissues reflects the ability of multiple tissues to biosynthesize RA (12, 13, 36, 37) and the widespread tissue distribution of CRBP (38 -40).
Concluding Summary-This work identifies a second RoDH as a another isozyme that catalyzes the first step in RA biosynthesis. Occurrence of this heretofore unknown short-chain dehydrogenase/reductase indicates the importance of tissuedistinctive expression of RoDHs to RA biogenesis. The Ͼ5 mM concentration of holo-CRBP in normal rat liver exceeds the K m value for RoDH(I) and (II), consistent with physiological roles for them in RA biogenesis (7). Their ability to recognize CRBP provides a mechanism for accessing the major pool of retinol in vivo while allowing CRBP to control the availability and distribution of retinol. Two other classes of proteins important to RoDH activity was measured from holo-CRBP composed of total CRBP/ retinol in the ratio 1.4 with 175 mg of protein from the 10Ksup of P19 cells transfected with pcDNA3/RoDH(II). The Michaelis-Menten data were fit with Enzfitter (35). The curve shown is from one of two experiments, each done with a separate transfection. retinoid function are expressed in definite temporal-spatial patterns, the receptors RAR and RXR and the binding proteins CRBP and cellular retinoic acid-binding protein. These proteins also belong to superfamilies of sterol/lipid hormone receptors (41)(42)(43) and sterol/lipid-binding proteins (44,45), respectively. Thus, RoDHs represent a third class of proteins potentially important to retinoid action that belong to a larger superfamily of steroid/lipid-specific proteins.