Molecular Characterization of a Novel Short-chain Dehydrogenase/Reductase That Reduces All-trans-retinal*

The reduction of all-trans-retinal in photoreceptor outer segments is the first step in the regeneration of bleached visual pigments. We report here the cloning of a dehydrogenase, retSDR1, that belongs to the short-chain dehydrogenase/reductase superfamily and localizes predominantly in cone photoreceptors. retSDR1 expressed in insect cells displayed substrate specificities of the photoreceptor all-trans-retinol dehydrogenase. Homology modeling of retSDR1 using the carbonyl reductase structure as a scaffold predicted a classical Rossmann fold for the nucleotide binding, and an N-terminal extension that could facilitate binding of the enzyme to the cell membranes. The presence of retSDR1 in a subset of inner retinal neurons and in other tissues suggests that the enzyme may also be involved in retinol metabolism outside of photoreceptors.

The reduction of all-trans-retinal in photoreceptor outer segments is the first step in the regeneration of bleached visual pigments. We report here the cloning of a dehydrogenase, retSDR1, that belongs to the shortchain dehydrogenase/reductase superfamily and localizes predominantly in cone photoreceptors. retSDR1 expressed in insect cells displayed substrate specificities of the photoreceptor all-trans-retinol dehydrogenase. Homology modeling of retSDR1 using the carbonyl reductase structure as a scaffold predicted a classical Rossmann fold for the nucleotide binding, and an Nterminal extension that could facilitate binding of the enzyme to the cell membranes. The presence of retSDR1 in a subset of inner retinal neurons and in other tissues suggests that the enzyme may also be involved in retinol metabolism outside of photoreceptors.
Vitamin A and its metabolites are active participants in a number of important physiological processes (see Fig. 1). Alltrans-retinol is the precursor of other naturally occurring retinoids and appears to be indispensable in reproduction (1). All-trans-retinal is an intermediate in the production of alltrans-and perhaps 9-cis-retinoic acids, which act as hormones, affecting important biological processes such as morphogenesis and differentiation through their interaction with ligand-gated transcription factors (2). In most tissues, retinals do not accumulate in high concentrations, perhaps because of their chemical reactivity and potential toxicity. However, the visual system is unique in that its function is based on retinals and as a consequence, they accumulate in extraordinarily high concentrations, up to 3 mM in the outer segment of photoreceptor cells (3). 11-cis-Retinal is the chromophore for all known visual pigments; however, this isomer is not found in other tissues in significant amounts. Thus, metabolism of 11-cis-retinal and its photoproduct all-trans-retinal, in particular, are of prime importance in the visual system.
In the retina, light isomerizes 11-cis-retinal, activates rhodopsin, and initiates the process of phototransduction by which the visual sensation is produced. The product of photoisomerization, all-trans-retinal, enters into a series of reactions that regenerate the 11-cis-configuration and the native visual pigment. At any given level of illumination, a steady state is established in which the visual pigment photoisomerization rate is equal to and opposed by the rate of visual pigment regeneration (4,5). Thus, phototransduction and regeneration reactions are irrevocably linked in a cyclical process called the visual cycle (6).
The regeneration reactions begin with the NADPH-dependent reduction of all-trans-retinal. The enzyme catalyzing this reaction, all-trans-retinol dehydrogenase (RDH), 1 plays an important role in photoreceptor physiology that is only beginning to be understood. RDH catalyzes the rate-limiting reaction of the visual cycle in rodent rod photoreceptors (7) and plays an important role in phototransduction as the final step in the quenching of photoactivated rhodopsin (8,9). Furthermore, the activity of RDH controls the level of all-trans-retinal in the retina. This retinoid is responsible, in part, for setting the level of sensitivity of the visual system (10) and may play a role in retinal pathology. All-trans-retinal has been shown to be a constituent of the major fluorescent component of lipofuscin ( Fig. 1), a pigment that accumulates in retinal pigment epithelium (RPE) during aging and in pathological conditions (12).
The reactions and enzymes of phototransduction have been thoroughly characterized at a molecular level (12); however, the molecular details of the visual cycle remain poorly characterized. The amino acid sequence is known for only one enzyme of the cycle, 11-cis-retinol dehydrogenase (13). The remaining reactions have been characterized only as enzymatic activities in membrane preparations (14,15). In the present study, we report the molecular cloning of a cDNA expressing a shortchain dehydrogenase/reductase (retSDR1), and demonstrate that this enzyme is highly abundant in cone outer segments. Expression of retSDR1 in other cells of the retina, and in other tissues, suggests that it may be involved more generally in retinoid metabolism. The molecular characterization of retSDR1 will open the way for further studies of this visual cycle enzyme, and of its potential role in retinal cone dystrophies.

EXPERIMENTAL PROCEDURES
Materials-All-trans retinal was obtained from Sigma, [ 3 H]NaBH 4 was from NEN Life Science Products, and 11-cis-retinal was a gift from the National Eye Institute. Retinoids were purified by HPLC (16).
DNA Sequence Analysis-A search of the EST data base with primer FH28 (5Ј-GGCCTGGTCAACAATGCTGG-3Ј) was performed in the GenBank data base with FASTA from the GCG package. Amino acid sequence alignments were generated with PILEUP.
Cloning of Human retSDR1-Total RNA was isolated from human retinal tissue, obtained from the Lions Eye Bank at the University of Washington, using the Ultraspec RNA Isolation system (Biotecx, Inc.) and reverse transcribed with oligo(dT) (Life Technologies, Inc.). Rapid amplification of cDNA ends (RACE) was performed to amplify the 3Ј-end of the selected EST cDNA using the Marathon cDNA amplification kit (CLONTECH) and the Expand high fidelity PCR system (Boehringer Mannheim). 3Ј-RACE was primed with an internal gene-specific primer (FH42: 5Ј-ATGGCGTGGAAACGGCTGGG-3Ј) and the Marathon adaptor primer (AP1) (CLONTECH). Samples were heated at 95°C for 5 min and amplified for 40 cycles at 94°C for 30 s, and 68°C for 4 min. A secondary PCR reaction was carried out using the AP2 Marathon adapter primer and a nested gene-specific primer (FH41: 5Ј-CGGNGGCGGGAGAGGNATCGGG-3Ј) as described above. The 5Јend of the selected EST cDNA was amplified from a human retina cDNA library with primers FH43 (5Ј-TCCAAGAACTGGCCCAGGGT-GTTG-3Ј, a gene-specific primer) and gt10S. After heating at 95°C for 5 min, the reactions were cycled 5 times through 94°C for 30 s, 72°C for 4 min; 5 times through 94°C for 30 s, 70°C for 4 min; and 30 times through 94°C for 5 s, 68°C for 4 min. Two amplification products for each PCR were cloned into pCR TM 2.1 vector (TA cloning kit, Invitrogen) and sequenced by dyedeoxy-terminator sequencing (ABI-Prism, Perkin-Elmer).
Cloning of Bovine retSDR1-Total RNA was isolated from bovine retinas using the Ultraspec RNA isolation system (Biotecx, Inc.). cDNA used in the PCR was prepared by reverse transcription with oligo(dT) from 20 g of total RNA in a 20-l reaction (Life Technologies, Inc.). 5Ј-RACE PCR was carried out using 0.5 l of cDNA and the Expand high fidelity PCR system (Boehringer Mannheim) with a gene-specific primer FH43 (5Ј-TCCAAGAACTGGCCCAGGGTGTTG-3Ј) and the Marathon adaptor primer (AP1) (CLONTECH). Samples were heated at 95°C for 5 min and amplified for 5 cycles at 94°C for 30 s, 60°C for 30 s, and 68°C for 4 min and for 35 cycles at 94°C for 30 s, and 68°C for 4 min. A secondary PCR amplification was carried out using the AP2 Marathon adapter primer and primer FH43, for 5 cycles at 94°C for 30 s, and 72°C for 3 min; 5 cycles at 94°C for 30 s, and 70°C for 3 min; and 25 cycles at 94°C for 30 s, and 68°C for 3 min. 3Ј-RACE was performed with primers FH42 (5Ј-ATGGCGTGGAAACGGCTGGG-3Ј) and the Marathon adaptor primer (AP1) (CLONTECH) followed by a secondary PCR amplification carried out using the AP2 Marathon adapter primer and a nested gene-specific primer FH41 (5Ј-CGGNG-GCGGGAGAGGNATCGGG-3Ј) using the same PCR conditions described for 5Ј-RACE. Two amplification products for each PCR were cloned in pCR TM 2.1 vector (TA cloning kit, Invitrogen) and sequenced by dyedeoxy-terminator sequencing (ABI-Prism, Perkin-Elmer).
Cloning of Mouse retSDR1-The mouse cDNA sequence was amplified from a ZAPII both oligo(dT)-and random-primed mouse retinal cDNA library (obtained from Dr. W. Baehr, University of Utah) in two overlapping fragments. The 5Ј-end was amplified with primers FH59 (5Ј-TCATCGTCACTGTCCATCAAGCTT-3Ј, gene-specific primer) and FH100 (5Ј-TTGTAATACGACTCACTATAGGGCG-3Ј, covering T7 primer) for 5 cycles at 94°C for 30 s, and 72°C for 3 min; for 5 cycles at 94°C for 30 s, and 70°C for 3 min; and for 30 cycles at 94°C for 30 s, and 68°C for 3 min. A secondary nested PCR amplification was carried out with primers FH100 and FH40 (5Ј-CACGGCGGCATTGTTCAC-CAG-3Ј) for 5 cycles at 94°C for 30 s, 64°C for 30 s, and 68°C for 1 min and for 25 cycles at 94°C for 30 s, and 68°C for 1 min. The 3Јend was amplified with primers FH100 and FH54 (5Ј-TCGGGACTTGTCGCGG-GAGTCA-3Ј, gene-specific primer) for 5 cycles at 94°C for 30 s, and 72°C for 3 min; for 5 cycles at 94°C for 30 s, and 70°C for 3 min; and for 30 cycles at 94°C for 30 s, and 68°C for 3 min. A secondary PCR amplification was performed with primers FH100 and FH58 (5Ј-ATCG-GACGCCACCTCGCTCGGG-3Ј, nested gene-specific primer) for 25 cycles at 94°C for 30 s, and 72°C for 2 min. Two amplification products for each PCR were sequenced by dyedeoxy-terminator sequencing (ABI-Prism, Perkin-Elmer) after subcloning into pCR TM 2.1 vector (TA cloning kit, Invitrogen).
Northern Blot Analysis-Bovine retina poly(A) ϩ RNA was purified from total RNA using the mRNA purification kit (Amersham Pharmacia Biotech), resolved by agarose gel electrophoresis in the presence of 0.66 M formaldehyde and transferred to nylon membranes. Hybridization with a 1.2-kb bovine dehydrogenase cDNA 32 P-labeled with the Megaprime DNA labeling systems (Amersham Pharmacia Biotech) was performed in 40% formamide, 10% dextran sulfate, 1% SDS, 1 M NaCl, 50 mM Tris, pH 7.4, 25 g/ml herring sperm DNA and washed in 0.1ϫ SSC at 58°C. A human multiple tissue Northern blot containing 2 g of poly(A) ϩ RNA from various human tissue (CLONTECH) was hybridized with the dehydrogenase or glyceraldehyde-3-phosphate dehydrogenase 32 P-labeled cDNA according to the manufacturer's instructions.
Expression of Human retSDR1 in Escherichia coli-The full-length 1.4-kb human retSDR1 cDNA was generated by cloning the 5Ј and 3Ј ends using the NarI restriction site present in the overlapping region. The coding sequence for retSDR1 was amplified from this plasmid by PCR with primers FH48 (5Ј-CATATGGCGTGGAAACGGCTGGGC-3Ј), which placed a NdeI restriction site on the ATG, and FH49 (5Ј-CTAGT-GATGGTGATGGTGATGTGTCCGCCCTTTGAAAGTGTT-3Ј), which placed a 6-His tag at the 3Ј-end and using a denaturing temperature of 94°C for 30 s, and an annealing and extension temperature of 68°C for 2.5 min. The purified fragment was cloned into pCR TM 2.1 vector (TA cloning kit, Invitrogen). The insert was transferred as a NdeI-BamHI fragment into NdeI and BamHI sites of pET-3b (Novagen) and expressed in BL21(DE3) pLysS after induction with 0.1 mM isopropyl-1thio-␤-D-galactopyranoside. Proteins insoluble in bacteria were purified on Ni-NTA resin (Qiagen) under denaturing conditions following the manufacturer's instructions. The sequences of all constructs presented in this study were verified by DNA sequencing.
Expression of Human retSDR1 and Bovine 11-cis-Retinol Dehydrogenase in Insect Cells-The coding sequence for retSDR1 was amplified from the retSDR1 plasmid by PCR with primers FH48 (5Ј-CATATG-GCGTGGAAACGGCTGGGC-3Ј), which placed a NdeI restriction site on the ATG, and FH50 (5Ј-CTATGTCCGCCCTTTGAAAGTGTT-3Ј) using a denaturing temperature of 94°C for 30 s, and an annealing and extension temperature of 68°C for 2.5 min. The purified fragment was cloned into pCR TM 2.1 vector (TA cloning kit, Invitrogen), designated pFR415. A fragment XbaI-HindIII from pFR415 covering the human retSDR1 coding sequence was cloned between the XbaI and HindIII sites of pFastBac1 expression vector (Life Technologies, Inc.). The expression cassette was then transferred into the baculovirus shuttle vector (bacmid) by transposition. Sf9 insect cells were transfected with the recombinant bacmid using cationic liposome-mediated transfection (CellFECTIN reagent, Life Technologies, Inc.) according to the manufacturer's protocol. The coding sequence for 11-cis-retinol dehydrogenase was amplified from bovine retina cDNA by PCR with primers FH51 (5Ј-CATATGTGGCTGCCTCTGCTGCTG-3Ј), which placed a NdeI restriction site on the ATG, and FH52 (5Ј-TTAGTAGACTGTCTGGG-CAGG-3Ј) for 32 cycles at 94°C for 30 s, and 68°C for 2.5 min and cloned into pCR TM 2.1 vector (TA cloning kit, Invitrogen) (designated pFR425) and then transferred into insect cells following the same procedure as for human retSDR1. For the expression of recombinant proteins, cells cultured at 27°C in Sf-900 II SFM (Life Technologies, Inc.) were harvested by centrifugation at 1200 ϫ g, 72-96 h after infection. Variable expression and activity levels were observed between different preparations. The reasons for these differences were not investigated.
Expression of Truncated Human retSDR1 in Insect Cells-An ATG and NdeI restriction site were introduced upstream of the amino acid 36 of retSDR1 by PCR on pFR415 plasmid with primers FH64 (5Ј-CATAT-GCTGTCGCGGGAGAACGTCC-3Ј) and FH50 (5Ј-CTATGTCCGCCC-TTTGAAAGTGTT-3Ј) through 32 cycles using a denaturing temperature of 94°C for 30 s and an annealing and extension temperature of 68°C for 2.5 min. The purified fragment was cloned into pCR TM 2.1 vector (TA cloning kit, Invitrogen) and sequenced by dyedeoxy-terminator sequencing (ABI-Prism, Perkin-Elmer) and then transferred into insect cells following the same procedure as for the full-length retSDR1.
Measurement of RDH Activity-RDH activity was measured by following the transfer of 3 (18), and purified on a Nucleosil NB 10 column (Macherey-Nagel) (17). Cells were homogenized in water containing 2 mM benzamidine, 0.1 mM NADP, 0.5 mM dithiothreitol. Cell membranes were washed in the same solution, pelleted by centrifugation at 40,000 rpm for 5 min, and resuspended in 5 times the volume of the pellet. The reaction mixture contained 10 mM MES, pH 5.5, 14.2 M [ 3 H]NADPH (83,000 dpm/nmol, both pro-4S and pro-4R isomers), and membranes from insect cells (2.5-3.0 mg/ml protein), in a final volume of 60 l. All-trans-retinal or 11-cis-retinal in ethanol was added at a concentration of 10 M (final concentration of ethanol Ͻ1%, v/v). Reactions were quenched with 400 l of methanol, 50 l of 10 mM NH 2 OH, pH 7.0, 50 l of 0.1 M NaCl and extracted with 500 l of petroleum ether. Radioactivity was determined in 250 l of the organic phase.
Preparation of Anti-retSDR1 Monoclonal Antibody-BALB/c mice were immunized with his-tagged human retSDR1 expressed in E. coli retSDR1. A hybridoma cell line producing specific antibodies was prepared by fusion of BALB/c mouse myeloma cells with spleen cells from the immunized mouse according to the procedure described by Harlow and Lane (20). Supernatants from culture of the hydridomas were screened by immunocytochemistry and by immunoblotting. Subsequent cloning yielded mAb A 11 , which was an IgG 1 . The epitope recognized by mAb A 11 has not been characterized in detail; however, it must be present in the C-terminal half of the protein because mAb A 11 recognized a truncated protein comprising amino acid residues 169 -303 (data not shown). An anti-11-cis-retinol dehydrogenase rabbit polyclonal antibody was generated using a peptide whose sequence (SKFL-GLEAFSDSLRRDV) encompassed the catalytic region of the enzyme. The peptide was coupled to a carrier protein and used for immunization as described previously (21). Anti-retSDR1 and anti-11-cis-retinol dehydrogenase reacted specifically with the dehydrogenase used for immunization.
In Situ Hybridization-Samples of bovine, monkey and human retina were processed with sense and antisense digoxygenin-labeled riboprobes as described previously (21).
Immunocytochemistry-The anterior segments of bovine, monkey, and human eyes were removed and the eye cups immersed in 4% paraformaldehyde, 0.13 M sodium phosphate, pH 7.4, at 4°C for 6 h. For immunoperoxidase staining, the whole mount retinal sections were processed as described by Milam et al. (22). For confocal microscopy, agarose-embedded retinal sections (100 m) were processed as described previously (22,23).
Fluorescence in Situ Hybridization-A 1.5-kb cDNA probe covering human retSDR1 was labeled with biotin-11-dUTP by nick translation (Life Technologies, Inc.). Metaphase chromosome preparations from lymphocytes of a human male were obtained using 75 mM KCl as a hypotonic buffer and methanol:acetic acid (3:1, v/v) as fixative. The hybridization was carried out as described previously by Edelhoff et al. (24). Hybridization signals were detected using a detection system from Vector Laboratories. After incubation with goat anti-biotin antibody, the slides were rinsed in modified 2ϫ SSC (2ϫ SSC, 0.1% Tween 20, 0.15% bovine serum albumin). A second incubation with fluoresceinlabeled anti-goat IgG and a rinse in modified 2ϫ SSC followed. The chromosomes were banded using Hoechst 33258-actinomycin D staining and counterstained with propidium iodide. The chromosomes and hybridization signals were visualized by fluorescence microscopy, using a dual band pass filter (Omega).
Modeling-The model of retinol retSDR1 was constructed using the two crystal structures of ternary complexes of short-chain dehydrogenases with NADP and substrates. The structure of residues 35-286 is based on that of mouse carbonyl reductase, which has 26% sequence identity and 34% similarity (26). Inserts (1, 4, 3, 1, 1, and 1 amino acids) and omissions (1 amino acid) were modeled using the Biopolymer (Tripos) and Chain (Baylor University) software. The conformation of the inserted loops was modeled to resemble that of the human 17␤-hydroxysteroid dehydrogenase (17␤-HSD), which has 26% sequence identity and 33% similarity (26). The C-terminal part of retSDR1, residues 287-302, was modeled based on the fold of 17␤-HSD. The model was allowed to relax from possible steric overlaps of the side chains and then was optimized by energy minimization using the Biopolymer software. The model of retinol was constructed using the coordinates of retinoic acid available in the Cambridge Structural Database. The retinol molecule was positioned in the active site of retSDR1 so that its oxygen atom corresponded to the oxygen atom of the propanol present in the structure of carbonyl reductase and the ligand position was optimized with the docking option of the Biopolymer software.

Molecular
Cloning of Human retSDR1-The enzymatic properties of retinol dehydrogenase in ROS extracts, such as molecular weight, sensitivity to thiol-reactive reagents, and solubility, suggested that the enzyme was likely to be a member of the short-chain dehydrogenase/reductase (SDR) superfamily. A DNA sequence (ϳ40 base pairs) (see "Experimental Procedures") corresponding to a conserved domain among retinol dehydrogenases was used to search for homology in an expressed sequence tag (EST) data base. One EST obtained from a human retinal cDNA library (W22782 generated by J. Macke, P. Smallwood, and J. Nathans) was similar to 11-cis-retinol dehydrogenase, and its translation product had amino acid sequence motifs conserved in the SDR superfamily (27). PCR with primers designed from this EST sequence amplified a product of 250 base pairs, which provided a probe to identify the putative retSDR1 cDNA among RACE products. Overlapping cDNA 5Ј and 3Ј ends obtained by RACE were combined to construct a 1.4-kb human cDNA. This full-length cDNA contains an open reading frame encoding a basic protein (pI ϭ 9.07) of 302 amino acids with a predicted molecular mass of 33,520 Da ( Fig. 2A). The first ATG (55-57 nucleotides) matched the Kozak consensus sequence for translation initiation (28). Bovine and mouse cDNAs were cloned following the same strategy, and sequences are shown in Fig. 2A. The amino acid sequences of human, bovine and mouse retSDR1s are 94 -98% identical (Table I). This high homology is absent in the 5Ј-and 3Ј-untranslated regions of bovine, human, and mouse retSDR1 (data not shown).
Tissue Distribution of retSDR1 mRNA-Tissue distribution of retSDR1 mRNA was assessed by Northern blot analysis of several human tissues and bovine retina. A transcript of ϳ1.8 kb is present in placenta, lung, liver, kidney, pancreas, and retina but was not detected in brain. The size of the human retSDR1 detected by blotting is similar to that of the partial (perhaps lacking fragments of untranslated regions) isolated cDNAs (1.401 for human, 1.478 for mouse, and 1.460 for bovine retSDR1). A transcript of smaller size (ϳ1.4 kb) was observed in heart and skeletal muscle (Fig. 3). This transcript could arise from tissue-specific polyadenylation of the retSDR1 mRNA or splicing of the retSDR1 pre-mRNA, because a similar hybridization pattern was observed with the probe derived from the 3Ј-untranslated region of retSDR1 (data not shown). The presence of the retSDR1 mRNA, and not a cross-reacting transcript, in other tissues was also supported by a homology search of EST data bases, which revealed sequences identical to retSDR1 in several mouse and human tissues. retSDR1 was not detected, however, by Western blotting with specific monoclonal anti-retSDR1 antibody (described below) at 10 g of membranous proteins loaded per lane from bovine brain, mammary gland, heart, skeletal muscle, liver, kidney, testis, adrenal gland, RPE, retina, or rod outer segments (data not shown).The failure of mAb to react with membrane components of outer segments (OS) preparations is not surprising, consid-ering that only about 10% of the photoreceptors in mammalian retinas are cones, that the method of isolation of OS is preferential for rods, and that retSDR1 is a minor component related to cone opsin. Even antibodies to cone opsins fail to detect the proteins in preparations of mammalian OS (data not shown). Thus, the expression level must be low in these tissues or restricted to a subset of cells or cellular compartments.
Immunocytochemical Localization of retSDR1-A monoclonal antibody was raised against bacterially expressed retSDR1-His 6 protein, and a hybridoma cell producing mAb A 11 was cloned. The mAb A 11 reacted with retSDR1 expressed in E. coli and in insect cells and did not cross-react with other SDRs, such as expressed 11-RDH (results not shown). Immunostaining with mAb A 11 was used to determine the cellular localization of retSDR1 in bovine retina. The antibody reacted intensely with cone but not rod outer segments when examined by immunoperoxidase (Fig. 4A). A similar result was obtained by indirect immunofluorescence (Fig. 4B), which revealed intense labeling of cone outer segments, diffuse labeling of the remainder of the cone photoreceptors, and no labeling of rod outer segments. All cone types were immunoreactive as determined by double labeling immunofluorescence with cone-type specific markers (data not shown). Both methods also showed immunoreactivity with a subset of somata localized in the inner nuclear layer. Similar results were obtained with sections of human and monkey retina. The restricted distribution of retSDR1 to cone and not rod photoreceptors is a strong indication of the specificity of the antibody because rods contain an RDH activity (10,34,36,37) and outnumber cones in bovine retina by about 10 to 1. At present, we do not know whether reactivity of the mAb A 11 with a subset of neurons within inner nuclear layer represents the presence of the retSDR1 within these cells or cross-reactivity with another SDR. These cells have not been identified.
In Situ Hybridization-Specific hybridization with digoxygenin-labeled sense and antisense riboprobes was used to provide  3. Northern blot analysis of retSDR1 expression in human tissues. Poly(A) ϩ -enriched RNA (2 g) from various human tissues and from bovine retina were probed with 32 P-labeled human SDR1. Control hybridization with a 32 P-labeled glyceraldehyde-3-phosphate dehydrogenase probe is shown at the bottom of the panel. The migrations of molecular size standards (in kilobases) are shown on both sides. The blots were exposed for 3 days. further indication of the bovine retinal cell type expressing retSDR1. The antisense riboprobe produced an intense hybridization signal in a band corresponding to photoreceptor inner segments (Fig. 4E). Cone ellipsoids, which are regions of the photoreceptor rich in mitochondria, appeared as negative images within the layer of intense hybridization. Somata located in the inner nuclear layer also displayed an intense hybridization signal, similar to the pattern seen with antibody staining. A third type of hybridization signal was seen in the ganglion cell layer; however, this signal was also produced by the sense riboprobe and is likely to be an artifact (30). No specific staining of the photoreceptor layer was observed with the sense riboprobe (Fig. 4F). Similar results were obtained with monkey and human retinas.
Enzymatic Activity of Expressed retSDR1-Human retSDR1 and 11-cis-retinol dehydrogenase were expressed in insect cells and assayed for RDH activity. The expression of both enzymes was confirmed by Western blot analysis with mAb A 11 (anti-retSDR1) and polyclonal anti-serum specific for a peptide derived from the 11-cis-retinol dehydrogenase sequence (data not shown). Membranes from insect cells infected with virus containing the retSDR1 sequence reduced all-trans-retinal but not 11-cis-retinal in the presence of chemically prepared NADPH, with tritium in pro-4S and pro-4R positions (Fig. 5A). In con-trast, in control experiments, membranes from insect cells infected with virus containing 11-cis-retinol dehydrogenase reduced 11-cis-retinal, but not all-trans-retinal (Fig. 5A). Membranes from insect cells infected with virus without RDH sequences (bacmid) displayed little activity (Fig. 5A), however, above that found in the absence of retinal. Similar studies revealed that NADH could not substitute for NADPH in the reduction of all-trans-retinal (results not shown).
ROS and RPE membranes utilized the pro-4S proton of NADPH (Fig. 5B) and, after 20 min, more than 75% of alltrans-retinal generated from bleached rhodopsin and 11-cisretinal added to the reaction mixture were converted to alltrans-retinol and 11-cis-retinol, respectively. 11-cis-Retinol dehydrogenase and retSDR1, 2 expressed in insect cells, also displayed pro-4S specificities (Fig. 5B), yielding 11-cis-retinol (not shown) and all-trans-retinol (Fig. 5C), respectively. The basal activity in the bacmid control was abolished, suggesting that this low activity in the insect cell membranes resulted FIG. 4. Localization of retSDR1 protein and mRNA in bovine retina. A and B, immunoperoxidase. Blocks of fixed tissue were processed with specific mAb (A) or specific mAb preabsorbed with antigen (B), and then stained with immunoperoxidase reagents and sectioned at 5 m. Note the intense immunoreactivity of the cone outer segments (open arrows) and the lack of staining of rod outer segments. Note also the staining of a subset of somata within the inner nuclear layer (arrowheads). C and D, confocal microscopy of retina stained with specific mAb (C) or mAb preabsorbed with antigen (D). Note the intense labeling of cone outer segments (open arrows), and the lighter staining of the rest of the cone photoreceptors, including the synaptic triad. Labeled somata are also evident within the inner nuclear layer (arrowheads). Sections of 100 m were employed. E and F, in situ hybridization with digoxygenin-labeled antisense (E) and sense (F) riboprobes. Cone ellipsoids (arrows) appear as negative images within the heavily labeled inner segment layer. Staining of somata within the ganglion cell layer was noted with both the sense and antisense probes and is likely to be an artifact of the procedure. Sections of 5 m were used. OS, photoreceptor outer segments; IS, photoreceptor inner segments; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar ϭ 50 m. from a dehydrogenase activity with pro-4R specificity.
Potential 17␤-HSD, 11␤-HSD, and 3␣-HSD activities were assayed with selected steroids using membranes from insect cells infected with virus containing the retSDR1 sequence or control cells. Membranes containing retSDR1 did not exhibit HSD activities above those of the control cells toward these selected steroids (Fig. 5D). In contrast, in control experiments, liver microsomes displayed 5 times higher 17␤-HSD activity than insect cell membranes (data not shown).
Chromosomal Location of the retSDR1 Gene-The retSDR1 gene was mapped to human metaphase chromosomes by FISH using a 1.4-kb human cDNA probe. Of 117 cells examined, 79 (68%) showed signals on both chromatids of chromosome 1 at band p36.1 (Fig. 6). The localization of a STS (WI-16826) assigned to position 54.6 cR (a measure of distance that is analogous to centrimorgans but depends on the radiation dose) from the top of Chr1 linkage group that covers a part of the human retSDR1 gene (bases 1143-1369 of human retSDR1 cDNA), further confirmed the FISH chromosomal localization. Preliminary characterization of the human retSDR1 gene suggests a complex structure spanning Ͼ7 kb of the genomic DNA (data not shown). DISCUSSION Over a century ago, Boll and Kü hne (31, 32) reported the progressive change in colors from red to yellow to colorless that occurred in a frog retina exposed to light. Decades later, Wald and colleagues (6,33) established the molecular foundations for these color changes as the photoisomerization of opsin-bound 11-cis-retinal to all-trans-retinal (red to yellow) and reduction of all-trans-retinal to all-trans-retinol (yellow to colorless). Subsequently, it was established that the reduction of all-transretinal is catalyzed by a membrane bound dehydrogenase of ROS (34), whose required cofactor is NADPH (35), and that the enzyme is highly specific for all-trans-retinal (36,10). Although several detergents were reported to solubilize the activity, reports of purification of the enzyme did not lead to its molecular characterization (37). In retrospect, it is apparent that the low abundance of RDH relative to opsin and peripherin/ROM, proteins present in orders of magnitude greater amounts, confounded the identification of RDH in enriched preparations. The difficulties are magnified even more with biochemical approaches to identify cone enzymes. However, identification of an SDR sequence motif in an EST data base from human retina led to cloning of a full-length cDNA encompassing this EST and generation of molecular probes (antibodies and riboprobes) and allowed demonstration of RDH activity by the expressed protein, localization of its mRNA to photoreceptor inner segments, and localization of the protein to cone photoreceptor outer segments.
Amino Acid Sequence and Structure of retSDR1-A number was used to determine the nucleotide specificities. The initial reduction rate was determined from measurements of the dehydrogenase activity at 0, 3, 5, and 15 min. For ROS membranes, all-trans-retinal was generated by bleaching rhodopsin (50 M); for RPE membranes, exogenous 11-cis-retinal (100 M) was added; for 11-cis-retinol dehydrogenase (11-cis-RDH) and retSDR1 expressed in insect cells, 11-cis-retinal (100 M) or all-trans-retinal (100 M) was added, respectively; and for bacmid control, all-trans-retinal (100 M) was added. C, HPLC analysis of the product generated by expressed retSDR1 in the presence of all-trans-retinal. Retinols were separated using 10% ethyl acetate in hexane on a Silica 5U column (Alltech; 250 mm ϫ 2.1 mm) at a flow rate of 0.3 ml/min, and retinols were detected at 325 nm. The radioactive profile (dotted line) represents a sample produced in the RDH assay by retSDR1 expressed in insect cells, and the elution profile of standard retinols is shown as a solid line. D, steroids as substrates for retSDR1.The assay was carried out as described under "Experimental Procedures." 17␤-HSDactivity was assayed by following the reduction of ␤-estradiol to estrone, 4-androstene-3,17-dione to testosterone, and 5␣-androstan-17␤-ol-3-one to 5␣-androstane-3␣, 17␤-diol; 11␤-HSD activity was assayed by following the reduction of 11-dehydrocorticosterone to corticosterone, and 3␣-HSD activity was assayed by following the reduction of 5␣-androstane-3,17-dione to androsterone. Controls included reduced forms of steroids and sample without steroids. The data are an average of three independent measurements. retSDR1 does not exhibit significant enzymatic activity toward these selected steroids.
of enzymes have been postulated to be members of the SDR superfamily based on the occurrence of several invariant sequence motifs rather than on their overall amino acid sequence homology, which is low (25-40%) (27,38). These SDRs catalyze diverse oxidation/reduction reactions of selected stereoisomers of hydrophobic aldehydes and ketones. Crystal structures have been determined for members of this superfamily, and a similar core structure is clearly evident in each. The sequence of retSDR1, translated from the cDNA sequence, contains the sequence motifs shared by members of this superfamily, including the invariant YXXXK active site motif, which is not present in the medium-chain alcohol dehydrogenase or aldo-keto reductase superfamilies, and the highly conserved nucleotide binding motif (TGXXXGXG). Other dehydrogenases catalyzing the interconversion of retinol and retinal have been shown to be members of the SDR superfamily, including three isozymes of liver all-trans-retinol dehydrogenase (39,40), and 11-cis-retinol dehydrogenase of RPE (13). retSDR1 is a member of the subfamily (as defined in Fig. 2B) of SDRs that includes 11-cis-RDH, liver retinol dehydrogenases I-III, CRAD, the steroid metabolizing dehydrogenases, and 15-hydroxyprostaglandin dehydrogenase. Members of the SDR family transfer the pro-4S hydrogen atom of NAD(P)H, while aldo-keto reductases, and medium-chain alcohol dehydrogenases transfer the pro-4R atom (41,42), and retSDR1 transfers the pro-4S hydrogen (Fig.  5B). The specificity of hydrogen transfer of this enzyme had been addressed previously (43) with [ 3 H]NADH; however, this dinucleotide is not a substrate for RDH (10).
A model of retSDR1 was constructed (Fig. 7A) employing the two crystal structures of ternary complexes of SDR with NADP(H) (see "Experimental Procedures"). The region 35-302 (shown in red) comprises two domains of different topology. The N-terminal region (residues 35-204) of retSDR1 has some amino acid identity with other ␣/␤ doubly wound enzymes that bind NADP (shown as a gray stick-and-ball model). Our simulations showed that the N-terminal ␣/␤ unit has the characteristics of the dinucleotide binding fold ("Rossmann fold") of lactate dehydrogenase (44). The phosphate moiety of NADP has hydrogen bonding contacts with side chains of Arg-56, Arg-71, and Thr-72. These residues correspond to Lys-17, Arg-39, and Thr-40 in mouse carbonyl reductase (25), on which structure the current model has been built. Thus, the environment of the phosphate moiety is conserved between the two enzymes. The conservation in this region is not observed between retSDR1/carbonyl reductase and 17␤-hydroxysteroid dehydrogenase (26), even though all of these enzymes utilize NADP as a substrate. 11-cis-Retinol dehydrogenase utilizes either NAD or NADP and lacks the Arg-71-Thr-72 motif in corresponding positions.
The C-terminal ␣-helical region is responsible for binding all-trans-retinol (shown as a black stick-and-ball model). Carbonyl reductase is specific for a secondary alcohol while alltrans-retinol is a primary alcohol. It appears that replacement of V190 by M230 fills the cavity in which the secondary ␤ carbon is placed and provides tight contacts between the retinol and the enzyme. Other residues in retSDR1 involved in the contacts with the retinol molecule are: Leu-177 (the corresponding positions in carbonyl reductase are given in parentheses; Val-138), Ile-182 (Phe-143), His-221 (Val-181), Leu-241 (Leu-200), Glu-244 (Arg-203), and Leu-282 (Tyr-241). The environment of the retinol molecule is shown in Fig. 7B.
The model of the N-terminal extension (amino acids 1-34) (shown in yellow on one side and gray on the other) is highly speculative, but of great interest, and provides a working hypothesis for further experimental studies of how the enzyme is anchored to the membranes or how it interacts with itself or other proteins. The N terminus is unique among SDRs, and does not correspond to any known signal peptide. Its predominantly hydrophobic character (except for Lys-22) suggests that it interacts with cellular membranes. This region appears to be also necessary for the stability of retSDR1 in the insect cells. The two positively charged regions Lys-4-Arg-5, and Lys-32-Leu-33-Arg-34 are likely to be involved in the interaction with the phosphate rich layer of the membranes. The N-terminal extension shows a strong preference for a ␤-conformation (as depicted in Fig. 7A), in which the side chain of Lys-22 could serve as the transmembrane anchor, ϳ43Å from the Arg/Lysrich region on the other side of the membranes. Within the ␤-structure, Pro-12 and Pro-30 would be close to the membrane face and adjacent on different strands. In these positions, the strands form a twist, because Pro residues cannot form Hbonds. Although it is generally thought that transmembrane segments of proteins are in ␣-helical conformations due to the requirement for main chain hydrogen bonding (45), the alternative model presented here is worth experimental scrutiny. The main chain hydrogen bonding in the ␤-structure may be formed as a result of oligomerization (tetramerization) of retSDR1 or interaction with other proteins anchoring the dehydrogenase to specific regions of the cone cell.
Physiological Relevance-The signal transduction cascade in photoreceptor cells is characterized by expression of enzymes and proteins specific to photoreceptors or to a small number of related neurons. From Northern analysis, retSDR1 mRNA is present not only in the retina (Fig. 4), but also is clearly evident in a number of other tissues. This widespread distribution of retSDR1 is borne out by EST data bases from several tissues in which fragments of the retSDR1 sequence are present. It is possible that retSDR1 acts as a generic all-trans-retinol dehydrogenase in many tissues, ensuring that deleterious levels of all-trans-retinal do not accumulate. Perhaps expression in photoreceptor cells evolved hand-in-hand with the evolution of a phototransduction cascade based on production of all-transretinal. In that light, it is not unexpected that several tissues express the message for retSDR1. retSDR1 may also be involved in other reduction/oxidation reactions throughout the body (including the retina). A similar situation is evident for 11-cis-retinol dehydrogenase (13) in that Northern analysis provides evidence for high level expression of mRNA encoding the enzyme in RPE, whereas a search of EST data bases reveals fragments of the sequence in many tissues (Ref. 46 and data not shown), even though 11-cis-retinal is unique to the visual system.
The relatively broad substrate specificity of SDRs has generated uncertainty regarding their physiological substrates in some tissues. For instance, rat and human hepatic retinol dehydrogenases (RODH1) have recently been shown to oxidize 5␣-androstane-3␣,17␤-diol to dihydrotestosterone as efficiently as they oxidize retinol (47). However, bleaching of the visual pigment in photoreceptor cells is known to produce all-trans-retinal within the outer segments in concentrations approaching 3 mM (48,49). Thus, the conclusion that retSDR1 is involved in reduction of all-trans-retinal in vivo is relatively secure given the large amounts of this substrate generated within the compartment in which the enzyme is localized. retSDR1 appears to lack HSD activity, as tested with the model steroid compounds.
Cone photoreceptors are known to regenerate bleached visual pigments more rapidly than rods. For instance, the time constant for regeneration of human cone visual pigments is 150 s and that of human rods is 400 s (4 -5). We observed that cone outer segments label heavily with anti-retSDR1 (mAb A 11 ), and that the enzyme is not detectable in rod outer segments, suggesting that the amount of retSDR1 is much greater in cones. This raises the intriguing hypothesis that the difference in the regeneration rates of rods and cones may be related, in part, to the amount of retSDR1 in the outer segments. However, another idea, which we favor, is that a homologous enzyme is responsible for the reduction in all-trans-retinal in rods.
In summary, we have demonstrated that a novel member of the short-chain dehydrogenase/reductase superfamily is abundant in cone outer segments and that the enzyme is possibly responsible for reduction of all-trans-retinal in the visual cycle. The amount of retSDR1 in cone outer segments appears to be much greater than in rod outer segments, suggesting an explanation for the faster regeneration of bleached visual pigments observed in cone vision. These results provide the basis for a more detailed study of the control, mechanisms, and role in retinal pathologies of this important enzyme. So far, no retinal disease-specific mutations have been found that map to chromosome 1 at band p36.1, where the retSDR1 gene is located. However, due to the potential involvement of the dehydrogenase in cone retinoid metabolism, retinal diseases characterized by retarded (cone) pigment regeneration appear to be worth testing for mutations in the retSDR1 gene. In addition, agerelated macular degeneration and diseases characterized by accumulation of lipofuscin and drusen, may result from the mutations in the retSDR1 gene.  (25,26). A, structural representation of retSDR1; the modeled enzyme is represented as a ribbon structure. In red, the binding domains of NADP (gray) and retinol (black); in yellow, hypothetical N-terminal fragment possibly involved in membrane association and/or oligomerization. Within the N-terminal domain, invariant Lys-23, at the bottom of the model, is shown using a ball-and-stick model. B, space-filling model representing the binding site for all-trans-retinol (ball-and-stick model in green) in the active site cavity of retSDR1. In black, carbon atoms with their van der Walls radii; in red, oxygen; in blue, nitrogen; in yellow, sulfur (from Met); and in gray, NADP. The environment of the binding site is highly hydrophobic.