Biosynthesis of 9- cis -Retinoic Acid in Vivo THE ROLES OF DIFFERENT RETINOL DEHYDROGENASES AND A STRUCTURE-ACTIVITY ANALYSIS OF MICROSOMAL RETINOL DEHYDROGENASES*

Retinoic acid is generated by a two-step mechanism. First, retinol is converted into retinal by a retinol dehydrogenase, and, subsequently, retinoic acid is formed by a retinal dehydrogenase. In vitro , several enzymes are suggested to act in this metabolic pathway. However, little is known regarding their capacity to contribute to retinoic acid biosynthesis in vivo . We have developed a versatile cell reporter system to analyze the role of several of these enzymes in 9- cis -retinoic acid biosynthesis in vivo . Using a Gal4-retinoid X receptor fusion protein-based luciferase reporter assay, the formation of 9- cis retinoic acid from 9- cis -retinol was measured in cells transfected with expression plasmids encoding different combinations of retinol and retinal dehydrogenases. The results suggested that efficient formation of 9- cis retinoic acid required co-expression of retinol and retinal dehydrogenases. Interestingly, the cytosolic alcohol dehydrogenase 4 failed to efficiently catalyze 9- cis -reti-nol oxidation. A structure-activity analysis showed that mutants of two retinol dehydrogenases, devoid of the carboxyl-terminal cytoplasmic tails, displayed greatly reduced enzymatic activities in vivo , but were active in vitro. The cytoplasmic tails mediate efficient endoplasmic reticulum localization of the enzymes, suggesting that the unique milieu in the endoplasmic reticulum

Retinol (vitamin A) and its derivatives are essential dietary compounds needed in a variety of physiological processes, e.g.embryonic development, reproduction, cell differentiation, postnatal growth, maintenance of the immune system, and vision (1)(2)(3).The metabolically active retinoids are 9-cis-retinoic acid (9cRA) 1 and all-trans-retinoic acid (atRA) in extraocular tissues, acting as ligands for two classes of nuclear retinoid receptors, the retinoic acid receptors (RARs) and the retinoid X receptors (RXRs) (4).In ocular tissues, 11-cis-retinal serves as the chromophore of the visual pigments (2).
Most cells obtain all-trans-retinol as the main source of retinoids, and it is known that activation of retinol into the active metabolites occur via tissue-specific isomerization and oxidation reactions.First, 9-cis-, 11-cis-, or all-trans-retinols are oxidized by retinol dehydrogenases (RDHs) into the corresponding retinals.For generation of the two isomers of RA, 9-cis-and all-trans-retinal are then further oxidized by a class of retinal dehydrogenases (Raldhs) (for reviews, see Refs. 5 and 6).It has been suggested that oxidation of the two isomers of retinol are the rate-limiting factors in the pathways generating 9cRA and atRA.
Two classes of RDHs have been implicated in oxidation of the different stereo isomers of retinol, i.e. microsomal members of the short chain alcohol dehydrogenase/reductases, and the cytosolic medium chain alcohol dehydrogenases (ADHs).Several members of both classes of enzymes can oxidize different stereo isomers of retinol in vitro.However, the physiological roles of these enzymes in retinoid biosynthesis are not well understood.
The first identified member of the microsomal RDHs was RDH5, an enzyme abundantly expressed in the retinal pigment epithelium with ability to convert 11-cis-retinol (7), and other cis-retinols (8,9), into the corresponding retinals.Subsequently, several other closely related members of the short chain alcohol dehydrogenase/reductase family have been identified as RDHs (10 -15).RDH5 is a homodimeric integral membrane protein anchored to the lipid bilayer via an aminoterminal signal anchor peptide and a carboxyl-terminal transmembrane segment.The catalytic domain of the enzyme is lumenal and only the extreme carboxyl-terminal 6 -8 amino acid residues reside in the cytosol (16,17) (the RDH4 enzyme has been renamed to RDH5 to indicate that it is the murine counterpart of RDH5).Genetic evidence for a role of the microsomal RDHs in ocular retinoid metabolism comes from the identification of mutations in RDH5 in patients suffering from fundus albipunctatus (19,20).These patients have a reduced rate of synthesis of 11-cis-retinal resulting in accumulation of white spots in the retina and stationary night blindness.These results suggest a physiological role of RDH5 in oxidation of 11-cis-retinol into 11-cis-retinal in the visual cycle.Surprisingly, mice carrying targeted deletion of the RDH5 gene do not display similar symptoms (21).Studies on mice carrying target deletions in the genes encoding ADH1 and ADH4 do not show any obvious phenotypes related to altered metabolism of endogenous retinoids (22,23).Instead, these data provide compelling evidence suggesting that ADH1 and ADH4 may not be critically required in generation of RA in vivo.
In this study we have analyzed the ability of various enzymes, suggested to play a role in RA biosynthesis, to contribute to generation of 9cRA from 9cROL in a cell reporter assay.Furthermore, we provide evidence suggesting that the unique milieu in the ER is necessary for efficient generation of 9cRA in vivo.

EXPERIMENTAL PROCEDURES
Co-transfection Reporter Assay-JEG-3 cells, COS-1 cells, and 293 cells were maintained in either minimum essential medium (for JEG-3; Life Technologies, Inc.) or Dulbecco's modified Eagle's medium (for 293 and COS-1; Life Technologies, Inc.) with 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, and 2 mM L-glutamine.The cells were transfected using calcium phosphate precipitation as described previously (24).The transfections were performed in 24-well plates, each experiment as triplicates using either minimum essential medium or Dulbecco's modified Eagle's medium with 10% charcoalstripped fetal bovine serum.The cDNAs encoding the enzymes were cloned into the expression plasmid CMX-PL1 (a kind gift from Dr. T. Perlmann).Each well was transfected with 100 ng of receptor Gal4-RXR plasmid, 100 ng of reporter MH100-tk-luc plasmid (24), and 200 ng of internal control plasmid encoding ␤-galactosidase.In addition, 100 ng of each plasmid encoding murine CRAD1 (13), CRAD2 ( 14), a CRADlike enzyme, CRADL,2 RDH5 (17,25) or murine ADH4 (a kind gift from Dr. Gregg Duester) were co-transfected with 100 ng of expression plasmid encoding retinal dehydrogenase 1 (Raldh1) or Raldh2 in various combinations.Original retinal dehydrogenase plasmids were kind gifts from Dr. Joseph Napoli.Unless otherwise indicated, medium with 1 M 9-cis-retinol (9cROL) or 9-cis-retinal (9cRAL) was separately added to the cells after 6 h of incubation with the calcium phosphate precipitate, and 1 M SR11237 was used as a positive control (26).For the titration experiment, 0.1-50 M 9cROL, 9cRAL, all-trans-retinol (atROL), or all-trans-retinal (atRAL) were used.Medium without added ligand served as a negative control.After an additional 18 h, the cells were lysed and their luciferase and ␤-galactosidase activities were measured with a luminometer (Lucy-1, Anthos, Saltzburg, Austria) as described previously (24).All luciferase values were normalized to the internal ␤-galactosidase control, and the maximal relative activation was set to 100%.The assays were performed several times with similar results, and each data point represents the mean of triplicate analyses.
Site-specific Mutagenesis of CRAD1 and RDH5-Mouse CRAD1 and RDH5 cDNAs in pBluescriptSKϪ (Stratagene) were used to produce single-stranded DNA with the helper phage M13KO7 (27).All mutations were made with oligonucleotide-directed site-specific mutagenesis without phenotypic selection (28).The oligonucleotides, with mutated nucleotide(s) underlined, were as follows; 5Ј-TTT GGT GGT GGT TTC TGC ATC TCT AAG (CRAD1, changing the tyrosine in the active site to phenylalanine, Y175F), 5Ј-TGG GAT GCT AAG TGA TTC TAC CTC CCC (CRAD1, inserting a premature stop codon before the COOHterminal transmembrane region, L289Stop), 5Ј-TGG ACT TCC CTG TGA CCT GAA AAA GCC (CRAD1, inserting a premature stop codon 6 amino acids before the normal stop codon thus deleting the cytosolic COOH-terminal R312Stop), and 5Ј-ACC TGG ATC CTT CCC TGA CCC GCC CAG TCA GTC (RDH5, inserting a premature stop codon 7 amino acids before the normal stop codon thus deleting the cytosolic COOHterminal tail, R312Stop).The mutations were verified by sequencing.Mutated plasmids were digested with EcoRI, the inserts were cloned into the eucaryotic expression vector CMX-PL1, and the enzymatic activities of the mutants were analyzed.

Expression of Normal and Mutant CRAD1 and RDH5 in COS-1 Cells and Measurement of Enzymatic Activity in Vitro by HPLC Analysis-
Wild-type and mutant forms of CRAD1 and RDH5 were expressed in transfected COS-1 cells essentially as described (7).Isolated microsomes from the transfected cells, and from mock-transfected cells, were collected and used in further analyses.The in vitro activity analyses were done essentially as described previously (19) using 9cROL as a substrate.A Waters Alliance 2690 HPLC system connected to a C 18 reversed-phase column (Supelco; 4.6 ϫ 250 mm) was used.9cROL was a kind gift from Dr. Michael Klaus (Hoffman-La Roche AG, Basel, Switzerland).
Proteinase K Assay, SDS-PAGE, Immunoblotting, RT-PCR Analysis, and Immunofluorescence Localization of CRAD1 and RDH5-The membrane topology of CRAD1 was determined using limited proteolysis as described previously (16).Essentially, microsomes (total 10 g of protein) were incubated with 20 g/l proteinase K, with or without 1% Triton X-100 treatment.Reactions were stopped after 5, 15, 25, and 45 min with the addition of phenylmethylsulfonyl fluoride to a final con-centration of 2 mM.Samples were prepared and analyzed on 12.5% SDS-PAGE gels and immunoblotted onto nylon membranes (Hybond ECL, Amersham Pharmacia Biotech).CRAD1 was detected using an affinity-purified polyclonal rabbit Ig fraction (see below).After extensive washings, bound Ig was detected by ECL using Hyperfilm ECL chemiluminescence film (Amersham Pharmacia Biotech) and developed.Cellular expression of RDH5 and ADH4 was verified by immunoblotting essentially as outlined above.Expression of CRAD2 and CRADL was verified by RT-PCR of total RNA fractions using oligo(dT) priming.The oligonucleotides 5Ј-TCT CTC TGC TTG ACT TCT ACA A and 5Ј-TTG GTG ACA TCC AGG ATC were used to amplify a 282-base pair fragment corresponding to CRAD2, and the oligonucleotides 5Ј-TGG GGT GAA GGT GGC TAT TAT and 5Ј-TCT GCA CCC TCA CAC AGC ACT were used to amplify a 418-base pair fragment corresponding to CRADL.As negative controls aliquots of the isolated total RNA fractions, without oligo(dT)-primed reverse transcription, were subjected to PCR amplifications using the above primer pairs.As positive controls, the corresponding expression plasmids were including as templates.The amplified products were analyzed on 1% agarose gels.
Polyclonal rabbit antisera to CRAD1 were generated by immunizing rabbits with a glutathione S-transferase fusion protein containing amino acids 19 -288 of CRAD1, essentially as described for the production of rabbit antisera to RDH5 (17).For affinity purification, the rabbit antisera were passed over columns containing the CRAD1-glutathione S-transferase fusion protein immobilized on CNBr-activated Sepharose, and bound Ig was eluted and characterized.The rabbit antibodies to RDH5 have been described previously (17), and the rabbit antiserum to human ADH4 was a kind gift from Dr. Jan-Olov Ho ¨o ¨g.
Immunofluorescence localization of CRAD1 and RDH5 was done using transfected CHO cells as described previously (17).Texas Redlabeled concanavalin A, Texas Red-labeled wheat germ agglutinin, and rhodamine-labeled phalloidin were used as suggested by the supplier (Molecular Probes).

A Reporter Assay Able to Measure the in Vivo Enzymatic
Conversion of 9cROL to 9cRA in Transfected Cells-Three different cell lines, JEG-3 cells, COS-1 cells, and 293 cells, were separately transfected with expression plasmids encoding the cis-retinol-specific dehydrogenase CRAD1 and the retinal dehydrogenase Raldh2 in an attempt to reconstruct a metabolic pathway able to generate 9cRA from 9cROL.Formation of 9cRA was monitored by co-transfecting the cells with plasmids encoding a Gal4-RXR-based luciferase reporter system (24).The induction of luciferase activity was subsequently measured following addition of 9cROL to a final concentration of 1 M.The results showed that all three cell lines efficiently converted 9cROL to 9cRA when co-expressing CRAD1 and Raldh2 (data shown for JEG-3 cells only) (Fig. 1A).In cells expressing only one of the two types of dehydrogenases, addition of 9cROL did not generate significant amounts of 9cRA.Addition of 9cRAL induced reporter gene activation in cells expressing Raldh2, either alone or in combination with CRAD1.Mock-transfected cells failed to efficiently generate 9cRA upon addition of 9cROL or 9cRAL (data not shown).The ability of the synthetic RXR agonist SR11237 to induce reporter gene activation in cells expressing all combinations of the two types of dehydrogenases confirmed that the transfected cells were able to respond to the presence of ligand.These data suggest that cellular components necessary for activity of the involved enzymes in this coupled enzyme-reporter system are present in several cell types and may be ubiquitously expressed.
The dose-response characteristics of transfected JEG-3 cells, expressing both enzymes, were measured by addition of different concentrations of 9cROL, 9cRAL, atROL, and atRA (ranging between 0.1 and 50 M for the 9-cis-retinoids and between 0.1 and 10 M for the all-trans-retinoids).The results showed that maximal reporter gene activation was achieved between 1 and 10 M 9cROL, and between 0.5 and 5 M 9cRAL.At higher retinoid concentrations, reporter gene activation was less effective probably due to the toxic effects of high concentrations of retinoids.All tested concentrations of atROL and atRAL failed to induce efficient reporter gene activation (Fig. 1B).The results are consistent with the notion that neither of the alltrans-retinoids are efficiently converted into the corresponding 9-cis-retinoids in vivo.
To analyze the capacity of different retinol dehydrogenases with ability to oxidize 9cROL in vitro, RDH5 (17,25), CRAD1 (13), CRAD2 (14), a novel CRAD-like enzyme "CRADL" 3 and the medium chain alcohol dehydrogenase ADH4 were separately expressed in combinations with Raldh2 in JEG-3 cells.Upon addition of 9cROL, the reporter gene activation analysis showed that efficient formation of 9cRA in vivo was achieved in cells expressing RDH5 and CRAD1, whereas CRAD2 and CRADL were less efficient (Fig. 1C).Surprisingly, cells coexpressing ADH4 and Raldh2 displayed a low capacity to generate 9cRA from 9cROL.Addition of 9cRAL efficiently induced reporter gene activation in transfected cells expressing all combinations of enzymes.No obvious differences in the capacity to oxidize 9-cis-retinal were found between Raldh1 and Raldh2 (data not shown).Immunoblot analysis of cell extracts from cells overexpressing ADH4 and CRAD1 verified that these enzymes were produced (Fig. 1D).In addition, RT-PCR analyses of total RNA fractions from the transfected cells demonstrated that transcripts encoding the enzymes were properly expressed.We conclude from these experiments that the in vivo capacities of the different retinol dehydrogenases to convert 9cROL into 9cRAL differ significantly, with CRAD1 being ϳ2 times as efficient in vivo as RDH5 using this cell reporter system.In line with their in vitro activities, 9-cis-retinol was a poor substrate for CRAD2 (14) and CRADL. 3 Membrane Topology of CRAD1 and a Structure-Activity Analysis of CRAD1 and RDH5-Previous analysis of RDH5, closely related in primary structures to CRAD1, has suggested a lumenal orientation of the catalytic domain of the enzyme and that it is membrane-anchored by a NH 2 -terminal signal anchor and a transmembrane segment close to the cytosolic COOH terminus (16,17).To investigate the membrane topology of CRAD1, microsomes from COS-1 cells expressing CRAD1 were isolated and subjected to limited proteolysis using proteinase K.In intact microsomes, CRAD1 was well protected from degradation, whereas in Triton X-100-treated microsomes, CRAD1 was rapidly degraded by the protease (Fig. 2A).These data are consistent with the notion that the membrane topology of CRAD1 is similar to that of RDH5 with a lumenal catalytic domain and the COOH terminus located in the cytosol.A schematic illustration of the membrane topology of CRAD1 is outlined in Fig. 2B.
To analyze structure-activity relationships in CRAD1, we constructed several mutants of the enzyme that could affect the membrane topology and mode of membrane interaction.The mutations were introduced to generate premature stop codons before the transmembrane segment (L289Stop), and before the COOH-terminal cytosolic tail (R312Stop), respectively.In addition, the active site tyrosine in CRAD1 was mutated into a phenylalanine (Y175F) (Fig. 2B).
Expression of the CRAD1 mutants in COS-1 cells and immunoblotting analysis of isolated microsomes from the transfected cells showed that significant expression levels of the mutant protein was found only in the mutant affecting the COOH-terminal tail (R312Stop).Other mutations failed to generate detectable levels of protein, probably due to decreased protein stability and subsequent degradation (Fig. 3A).An identical carboxyl-terminal tail deletion was also introduced in RDH5 (R312Stop) and the expression of wild-type RDH5 and mutant RDH5 R312Stop in transfected COS-1 cells was verified by immunoblotting (Fig. 3B).
Activity analysis of the two truncated versions of CRAD1 and RDH5 in the cell reporter assay showed that both mutants displayed low capacities to catalyze the conversion of the substrate in vivo compared with the wild-type protein (Fig. 3, C  and D).Addition of 2 mM NAD to the culture medium did not alleviate the lack of enzymatic activity in vivo (data not shown).Separate analysis showed that the carboxyl-terminal truncated enzymes present in isolated microsomes displayed almost full enzymatic activity in vitro when compared with the wild-type enzyme (Fig. 3 (E-G) for CRAD1, and data not shown for RDH5).These results suggested that the COOH-terminal cytosolic tails of the two enzymes are not directly involved in the catalytic mechanism, a result consistent with the membrane topologies of the enzymes.Instead, it may be involved in controlling intracellular compartmentalization of the enzymes, which may indirectly effect the ability of the enzyme to oxidize 9cROL in vivo.
To test this hypothesis, wild-type CRAD1, CRAD1 R312Stop mutant, wild-type RDH5, and RDH5 R312Stop mutant were separately expressed in CHO cells, and intracellular localization of the enzymes was determined by indirect immunofluo-rescence microscopy.As a marker of the ER, the cells were also stained with Texas Red-labeled concanavalin A. The results showed that wild-type CRAD1 and RDH5 displayed typical ER staining patterns (Fig. 4, A-C and I-K).In contrast, the CRAD1 R312Stop and RDH5 R312Stop mutants displayed strikingly different staining patterns.The CRAD1 mutant was localized in strongly stained elongated perinuclear fiber-like structures, while the RDH5 mutant was localized in unevenly distributed patchlike structures (Fig. 4, E-G and L-N).Staining with rhodamine-labeled phalloidin, to visualize the actin filaments, showed no obvious overlapping localization of the actin fibers with the truncated mutant CRAD1 (Fig. 4, D and H).Similarly, staining with fluorescent wheat germ agglutinin, to visualize the nuclear envelope and the Golgi apparatus, showed only marginal co-localization of this marker with the mutant enzymes (data not shown).All together, these results suggested that the COOH-terminal tails in CRAD1 and RDH5 have a role in compartmentalization of the enzymes and that ER localization of the enzymes might be necessary for enzymatic activity in vivo.

FIG. 2. Membrane topology of CRAD1.
A, limited proteolysis of CRAD1 expressed in transfected COS-1 cells.Proteinase K was added to microsomes permeabilized with 0.1% Triton X-100 (ϩ Triton-X) and to intact microsomes (Ϫ Triton-X).Following the different digestions (0 -45 min of incubation), the microsomes were subjected to SDS-PAGE and immunoblotting using antibodies to CRAD1.B, a schematic illustration of the membrane topology of CRAD1 and with an indication of the sites where mutations were introduced (arrows; see text for details).

FIG. 3. Expression and enzymatic activity of wild-type and mutant CRAD1 and RDH5 in vivo and in vitro.
A, immunoblotting analysis of wild-type CRAD1 and CRAD1 mutants Y175F, L289Stop, and R312Stop expressed in transfected COS-1 cells.B, immunoblotting analysis of wild-type RDH5 and mutant R312Stop expressed in transfected COS-1 cells.C, in vivo enzymatic activity measured using the reporter system in JEG-3 cells transfected with an expression vector encoding wild-type CRAD1 (black bar), the R312Stop mutant lacking the COOH-terminal cytosolic tail (white bar), and background activity in mock-transfected cells (gray bar).D, in vivo enzymatic activity measured using the reporter system in JEG-3 cells transfected with expression vectors encoding wild-type RDH5 (black bar), encoding the R312Stop mutant (white bar), and background activity in mock-transfected cells (gray bar).In C and D, the formation of 9cRA was assayed in culture medium containing 1 M 9cROL.All cell cultures were co-transfected with plasmids encoding Raldh2 and the RXR-based reporter system.The results were normalized to 100% activity for the wild-type enzymes.E-G, representative reverse phase HPLC profiles of 9cROL dehydrogenase activities in vitro in isolated microsomes from transfected COS-1 cells expressing wild-type CRAD1 (E), mutant CRAD1 R312Stop, lacking the COOH-terminal cytosolic tail (F), or microsomes from mock-transfected COS-1 cells (G).

DISCUSSION
At present, three major classes of enzymes have been implicated in synthesis of RA: the cytosolic medium chain alcohol dehydrogenases and microsomal members of the short chain dehydrogenase superfamily acting as retinol dehydrogenases, and the cytosolic aldehyde dehydrogenases acting as retinal dehydrogenases.The proposed roles of these enzymes in retinoid metabolism are mainly based on in vitro enzymatic activities, and in some cases also supported by the expression patterns of the enzymes.Genetic evidences for the involvement of some of these enzymes in retinoid biosynthesis come from the identification of patients carrying mutations in RDH5 which effects the synthesis of 11-cis-retinal in the eye, and from the analysis of a gene-targeted mouse strain lacking expression of Raldh2, which leads to embryonic lethality (19,29).
Despite the identification of several enzymes putatively involved in the pathways generating RA, no attempts to reconstitute a metabolic pathway able to generate RA from retinol have been reported.In the present work we demonstrate that efficient formation of 9cRA from 9cROL in vivo requires the concerted action of retinol dehydrogenases and retinal dehydrogenases.This finding implicates that efficient generation of RA in other cell types, i.e. cells in adult and embryonic tissues, may also require both enzymatic activities.We anticipate that modified versions of the coupled enzyme/reporter cell system can be useful in analyzing the structural and functional requirements for efficient synthesis of atRA in transfected reporter cells.
The microsomal CRAD1 and RDH5 appeared to be the most efficient enzymes converting 9cROL into 9cRAL of the retinol dehydrogenases tested.The subsequent formation of 9cRA in cells co-expressing retinal dehydrogenases shows that the two enzymatic reactions underlying formation of 9cRA from 9cROL are coupled, despite the different locations of the two reactions (lumenal versus cytosolic).In contrast, the cytosolic ADH4, previously reported to efficiently convert retinol into retinal in vitro, appeared as a less efficient retinol dehydrogenase in the transfected cells.The apparent lack of coupling of the enzymatic activities of ADH4 and the retinal dehydrogenases is surprising considering that both types of enzymes are cytosolic.Explanations to this observation might be that the cytosol is strongly reducing which would not favor retinol oxidation in vivo, or that sufficient concentrations of 9cROL or cofactor are not available to the enzyme.The latter explanation appears, however, less likely since 9cROL is hydrophobic and would easily cross cellular membranes and the cytosolic concentration of NAD is high.The efficiency of some of the microsomal retinol dehydrogenases contrasts the poor ability of ADH4 to oxidize 9cROL.Both CRAD1 and RDH5 display a unique membrane topology with the catalytic domains in lumenal compartments of the cells.The ER lumen is oxidative (30), and the milieu would favor retinol oxidation provided that sufficient concentrations of substrate and oxidized cofactors are available.CRAD1 and RDH5 have a preference for NAD as the cofactor (7,13,25); our results imply that sufficient concentrations of NAD are indeed present in the ER lumen.These conclusions are in line with previous measurements of cofactor concentrations in isolated microsomes (31).The origin of the lumenal NAD is currently unknown, but it may be generated inside the ER lumen by yet uncharacterized mechanisms, or it may be accumulated from the cytosol.In the latter case, we would predict the existence of a specific transport mechanism for the hydrophilic pyridine nucleotides, possibly involving a specific transmembrane transporter system.
The low in vivo activity and subcellular mislocalization of the mutants of CRAD1 and RDH5, lacking the extreme COOHterminal regions of 6 or 7 amino acids, respectively, implies that correct subcellular compartmentalization may be important for the activities of these enzymes.Under the experimental conditions used by us, the substrate 9cROL is unlikely to be the rate-limiting factor.Instead, the mislocalized mutant enzymes may end up in a subcellular compartment with a general milieu that does not favor retinol oxidation, or in a compartment with insufficient concentrations of NAD, or that formed 9cRAL generated by the mutant enzyme is unable to cross the ER membrane to reach the cytosolic Raldh2 for further oxidation to 9cRA.At present, it is not possible to distinguish between these possibilities.
The COOH-terminal cytosolic regions of the microsomal RDHs, including CRAD1 and RDH5, are well conserved (Fig. 5).In fact, the extreme COOH-terminal amino acids are -A-L in all members of the enzyme family, except in RDH5, which has an extension of one additional amino acid and ends with a hydroxyl-containing amino acid (Y in bovine and human RDH5, and S in mouse RDH5).Furthermore, the COOH-terminal tails of all the enzymes have a conserved pattern of several other residues, including the invariant proline residues at position -5 (-6 in RDH5).The conserved features of the cytosolic tails in the RDHs, and the fact that deletion of the cytosolic tails in CRAD1 and RDH5 leads to mislocalization of the enzymes and loss of enzymatic activities in vivo, argue for a role of the cytosolic tails in interactions with cellular components involved in retaining, or retrieving of the enzymes in the ER.Interestingly, several fundus albipunctatus patients recently identified in Japan have mutations in, or close to, the COOH-terminal cytosolic tail of RDH5 (A294P (Ref.20) and delC1085/ins GAAG (Ref.32)), suggesting that an intact COOH-terminal tail might be necessary for proper function of the enzyme in vivo.The identification of putative proteins interacting with the COOH-terminal tails of the RDHs might provide further clues to the roles of these enzymes in retinoid homeostasis in adult and embryonic tissues.

FIG. 1 .
FIG. 1. Synthesis of 9cRA in transfected JEG3 cells co-expressing retinol dehydrogenases and retinal dehydrogenases.A, 9cROL and 9cRAL were separately added to JEG-3 cells expressing CRAD1 (black bars), Raldh2 (white bars), and both CRAD1 and Ralhd2 (gray bars), and the activation of the RXR-based luciferase gene reporter system was used to monitor formation of 9cRA.The synthetic RXR ligand SR11237 served as a positive control.B, dose-response curves for 9cROL (f), 9cRAL (Ⅺ), atROL (q), and atRAL (E) added to JEG-3 cells co-expressing CRAD1 and Raldh2.C, synthesis of 9cRA by transfected JEG-3 cells expressing Raldh2 in combination with CRAD1 (black bars), CRAD2 (white bars), CRADL (dark gray bars), RDH5 (light gray bars), ADH4 (striped bars), monitored using the RXR-based reporter system.D, expression of several retinol dehydrogenases in the transfected JEG-3 cells.Expression of ADH4 (upper panel, left) and CRAD1 (upper panel, right) were detected by immunoblotting using specific antibodies.Mock-transfected cells (Mock) served as a negative control.The migration of molecular size markers is indicated to the right.Transcripts encoding CRAD2 (lower panel, left) and CRADL (lower panel, right) were detected by RT-PCR using aliquots of the total RNA fractions isolated from the transfected cells (RT-PCR).As positive controls the expression plasmids were used as templates (CRAD2 and CRADL).As negative controls aliquots of the isolated RNA fractions, without being subjected to reverse-transcription, were used (RNA).The sizes of the amplified fragments are shown to the right.

FIG. 4 .
FIG. 4. Localization of wild-type and mutant retinol dehydrogenases in transfected CHO cells by indirect immunofluorescence.A-D, expression of wild-type CRAD1; E-H, expression of mutant CRAD1 R312Stop lacking the COOH-terminal cytosolic tail; I-K, expression of wild-type RDH5; L-N, expression of mutant RDH5 R312Stop lacking the COOH-terminal cytosolic tail.The wild-type and mutant enzymes were localized using specific antibodies to CRAD1 or RDH5, respectively, using fluorescein isothiocyanate-labeled secondary antibodies (panels A, C-E, G, H and panels I, K, L, and N, respectively, in green).The ER was localized using Texas Red-labeled concanavalin A (B, C, F, G, J, K, M, and N, in red), and actin filaments were visualized using rhodamine-labeled phalloidin (D and H, in red).Merged images showing the wild-type and mutant retinol dehydrogenases with the ER marker are shown in C, G, K, and N (yellow color indicates co-localization), and wild-type and mutant CRAD1 with the actin filament marker are shown in D and H. Original magnification, ϫ630.