Cellular retinol-binding protein-supported retinoic acid synthesis. Relative roles of microsomes and cytosol.

This study shows that microsomal retinol dehydrogenases, versus cytosolic retinol dehydrogenases, provide the quantitatively major share of retinal for retinoic acid (RA) biogenesis in rat tissues from the predominant substrate available physiologically, holo-cellular retinol-binding protein, type I (CRBP). With holo-CRBP as substrate in the absence of apo-CRBP microsomal retinol dehydrogenases have the higher specific activity and capacity to generate retinal used for RA synthesis by cytosolic retinal dehydrogenases. In the presence of apo-CRBP, a potent inhibitor of cytosolic retinol dehydrogenases (IC50 = approximately 1 microM), liver microsomes provide 93% of the total retinal synthesized in a combination of microsomes and cytosol. Cytosolic retinol dehydrogenase(s) and the isozymes of alcohol dehydrogenase expressed in rat liver had distinct enzymatic properties; yet ethanol inhibited cytosolic retinol dehydrogenase(s) (IC50 = 20 microM) while stimulating RA synthesis in a combination of microsomes and cytosol. At least two discrete forms of cytosolic retinol dehydrogenase were observed: NAD- and NADP-dependent forms. Multiple retinal dehydrogenases also were observed and were inhibited partially by apo-CRBP. These results provide new insights into pathways of RA biogenesis and provide further evidence that they consist of multiple enzymes that recognize both liganded and nonliganded states of CRBP.

This study shows that microsomal retinol dehydrogenases, versus cytosolic retinol dehydrogenases, provide the quantitatively major share of retinal for retinoic acid (RA) biogenesis in rat tissues from the predominant substrate available physiologically, holo-cellular retinol-binding protein, type I (CRBP). With holo-CRBP as substrate in the absence of apo-CRBP microsomal retinol dehydrogenases have the higher specific activity and capacity to generate retinal used for RA synthesis by cytosolic retinal dehydrogenases. In the presence of apo-CRBP, a potent inhibitor of cytosolic retinol dehydrogenases (IC 50 ‫؍‬ ϳ1 M), liver microsomes provide 93% of the total retinal synthesized in a combination of microsomes and cytosol. Cytosolic retinol dehydrogenase(s) and the isozymes of alcohol dehydrogenase expressed in rat liver had distinct enzymatic properties; yet ethanol inhibited cytosolic retinol dehydrogenase(s) (IC 50 ‫؍‬ 20 M) while stimulating RA synthesis in a combination of microsomes and cytosol. At least two discrete forms of cytosolic retinol dehydrogenase were observed: NAD-and NADP-dependent forms. Multiple retinal dehydrogenases also were observed and were inhibited partially by apo-CRBP. These results provide new insights into pathways of RA biogenesis and provide further evidence that they consist of multiple enzymes that recognize both liganded and nonliganded states of CRBP.
The metabolism of retinol (vitamin A) generates RA, 1 a humoral factor critical to vertebrate development (1,2). In the developing embryo, RA transcriptionally regulates genes that specify body axis pattern and may help program limb formation (3). In mature vertebrates, RA maintains epithelial tissues (prevents squamous cell metaplasia), contributes to bone remodeling, and sustains reproductive processes, including the estrus cycle, spermatogenesis, and placental growth. RA aberrant in concentration, locus, or developmental stage causes teratism and/or toxicity of the central nervous system and skeleton (4 -7). RA acts through ligand-activated receptors that comprise two distinct subfamilies of the steroid hormone superfamily of receptors (8 -11). These receptors modify transcription as homodimers or by modifying the effects of other receptors through heterodimerization. These pervasive and fundamental effects of RA, as well as the consequences of its aberrant distribution, imply that its biosynthesis must be regulated closely.
In many tissues, unesterified retinol occurs bound to CRBP. Thus, holo-CRBP may provide the most abundant substrate for RA biosynthesis. Direct transfer of retinol between holo-CRBP and enzymes that catalyze RA synthesis would circumvent uncontrolled diffusion of retinol through the aqueous phase and would participate in controlling the pathways of RA synthesis by protecting retinol from opportunistic reactions catalyzed by enzymes that do not recognize the high affinity, high specificity CRBP. Recent work has outlined a pathway of RA biosynthesis with the first step catalyzed by a NADP-dependent microsomal retinol DH, expressed in liver and in extrahepatic tissues, that recognizes holo-CRBP as substrate (12)(13)(14)(15). Microsomes have low retinal DH activity and do not convert the retinal produced from holo-CRBP into RA at high rates. Rather, microsome-produced retinal undergoes conversion into RA by cytosolic retinal DHs, which can interact with CRBPretinal complexes (16).
During initial characterization of microsomal retinol DH, we also observed cytosolic retinol DH activity that was inhibited markedly by apo-CRBP and subsequently showed that the inhibition might not result solely from sequestering free retinol, because of ambiguities in the K d value for CRBP (12,17). Inhibition could result from apo-CRBP interacting with a cytosolic enzyme that recognized holo-CRBP as substrate. This suggested the presence of at least two pathways for generating RA from holo-CRBP: one involving retinal synthesis in microsomes and the other involving retinal synthesis in cytosol. In either pathway, subsequent conversion of the retinal into RA occurs in the cytosol. Ottonello and co-workers have analyzed cytosolic retinol DH activity further and concluded that it catalyzes considerable RA synthesis from holo-CRBP in the absence of apo-CRBP (18). The relative contributions, however, of microsomes and cytosol to overall cellular RA synthesis remain uncertain, as well as the interactions between these two pathways and/or their effects on each other. This work compares the contributions of microsomal and cytosolic retinol DH activities to RA synthesis and characterizes interactions that occur during RA biosynthesis between cytosol and microsomes. With holo-CRBP as substrate in the absence of apo-CRBP, microsomal retinol DH has the higher specific activity and capacity (60 -83% of the total of cytosol plus microsomes, individually) to generate retinal for RA synthesis in the four tissues assayed. In the presence of apo-CRBP, the more likely condition in vivo, the microsomal contribution increased to 80 -94% of the total retinal synthesized. In separating the activity of rat liver cytosolic retinol DH from that of the ADH isozymes expressed in rat liver, ADH-2 and ADH-3, we demonstrate further that ethanol inhibits cytosolic retinol * 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.
‡ To whom correspondence should be addressed: 140 Farber Hall, School of Medicine and Biomedical Sciences, SUNY-Buffalo, Buffalo, NY 14214. Tel.: 716-829-2032; Fax: 716-829-2725. 1 The abbreviations used are: RA, retinoic acid; ADH, alcohol dehydrogenase; CRBP, cellular retinol-binding protein, type I; DH, dehydrogenase; HPLC, high performance liquid chromatography; 10kS, 10,000 ϫ g supernatant of a rat liver homogenate. DH activity. We also show that at least two forms of cytosolic retinol DH occur and that apo-CRBP inhibits not only cytosolic retinol DH activity but also conversion of retinal into RA. Thus, we establish here that multiple paths and complex interactions contribute to RA biosynthesis.

EXPERIMENTAL PROCEDURES
General-Retinoids were purchased from Eastman Kodak and were purified by HPLC (19,20). Subcellular fractions from rat liver, kidney, testis, and lung were prepared from male Sprague-Dawley rats (ϳ150 g) as described previously (12). Protein was determined by the dyebinding method with bovine serum albumin as standard (21). Kinetic data were fit to standard kinetic equations with the microcomputer program "Enzfitter" (22).
CRBP-CRBP saturated with all trans-retinol was prepared and purified from CRBP generated in Escherichia coli with the vector pMONCRBP (23) as described previously (12,16). Purified holo-CRBP had an A 350 /A 280 ratio of 1.4 and was stored in 20 mM Hepes, 150 mM KCl, 2 mM dithiothreitol, and 1 mM EDTA, pH 7.5 (100 nmol/ml) at Ϫ70°C. Apo-CRBP was prepared and purified just as holo-CRBP, omitting saturation with retinol. The concentration of functional apo-CRBP was determined by saturating an aliquot with retinol, separating free and bound retinol by size exclusion chromatography and determining the A 350 /A 280 ratio.
Retinol DH Assays-Assays were done in duplicate (duplicates were within 10% of their averages) at 37°C in 250 l of 10 mM Hepes, 150 mM KCl, and 2 mM EDTA, pH 8, with 2 mM NADP for microsomes and 2 mM each of NAD and NADP for cytosol, the combination of microsomes and cytosol, and for the 10,000 ϫ g supernatant (10kS) for the times specified in the figure legends. The concentrations of holo-CRBP and apo-CRBP were 5 and 2 M, respectively (these represent physiological concentrations in rat liver), unless noted otherwise. Reactions were quenched with 1 ml of 0.025 M KOH/ethanol. Retinal was extracted with 2.5 ml of hexane (hex-1). To obtain RA, the alkaline phase was acidified to pH 1 with 4N HCl and extracted with 2.5 ml of hexane (hex-2). The solvents from hex-1 and hex-2 were evaporated under a stream of nitrogen. Each residue was dissolved in 100 l of hexane and analyzed by HPLC (16,20). Retinal in hex-1 and RA in hex-2 were quantified independently with a DuPont Zorbax-Sil Reliance cartridge column (0.4 ϫ 4 cm). For retinal, a linear gradient was run from 4 to 15% tetrahydrofuran in hexane for 5 min, followed by an additional 5 min of 4% tetrahydrofuran/hexane at 2 ml/min. Retinal eluted in 2 min and was detected at 370 nm. For RA, the mobile phase was dichloroethane/ hexane/acetic acid (9.65/90/0.35) eluted at 2 ml/min. RA eluted in 3.3 min and was monitored at 340 nm.
Inhibitor Studies-IC 50 values were determined in 30-min assays with 5 M holo-CRBP and 100 g of protein for microsomal retinol DH or 200 g of protein for cytosolic retinol DH. Ethanol, pentanol, and octanol were tested up to 10 mM with microsomal retinol DH and as shown in Fig. 7 for cytosolic retinol DH. Me 2 SO and dimethylformamide were tested up to 4% (v/v) in the assay mixture (10 l in 250 l). 4-Methylpyrazole was tested up to 500 mM with microsomal retinol DH and with 8 concentrations between 0.1 M and 1 M for cytosolic retinol DH. Carbenoxolone was tested with 7-9 concentrations between 5 M and 1 mM for both microsomal and cytosolic retinol DH.

Effects of apo-CRBP on Retinol DH Activities in Cytosol and
in the 10kS-apo-CRBP markedly inhibited cytosolic retinol DH activity supported by holo-CRBP, measured by RA synthesis ( Fig. 1). 2 As little as 0.1 M apo-CRBP caused 30% inhibition of a reaction supported by 5 M holo-CRBP; Ͻ1 M apo-CRBP caused 50% inhibition and 5 M caused Ͼ90% inhibition. Apo-CRBP inhibition of cytosolic RA synthesis was noncompetitive in nature (data not shown), with both decreased apparent In contrast to cytosol, 1 M apo-CRBP did not inhibit RA synthesis by the 10kS, which includes cytosol and microsomes, and 50% inhibition of the 10kS required 5 M apo-CRBP, a concentration equal to that of the holo-CRBP substrate. Lack of inhibition of RA synthesis in the 10kS by 1 M apo-CRBP, a concentration that inhibited cytosolic RA synthesis Ͼ50%, suggests that cytosolic retinol DH activity was not contributing substantially to RA synthesis in the 10kS, i.e. in the presence of microsomes. The relatively apo-CRBP-resistant microsomal retinol DH, therefore, must have been providing retinal for RA synthesis by the cytosolic retinal DH component of the 10kS. Inhibition of RA synthesis in the 10kS by higher concentrations of apo-CRBP may result from a combination of inhibition of the microsomal retinol DH (12) and a decrease in the conversion of retinal into RA by cytosol (see below).
Comparison of Microsomal and Cytosolic Retinol DHs-Retinal (Fig. 2, top panel) and RA (Fig. 2, bottom panel) syntheses were measured in cytosol, microsomes, and a combination of cytosol and microsomes with respect to time. In the absence of apo-CRBP (open symbols), the amounts of retinal produced by microsomes alone and RA produced by cytosol alone increased throughout the two hr of the reaction and were about equivalent. Because the cytosolic protein concentration was 5-fold greater than that of microsomes, however, the rate of microsomal retinal synthesis exceeded that of cytosolic RA synthesis. In the combination of cytosol and microsomes, retinal concentrations reached a plateau in 15 min of ϳ40 pmol and declined after 60 min (data not shown), and RA production exceeded retinal at all times. This illustrates use of the retinal generated from holo-CRBP by microsomes as substrate for cytosolic RA synthesis. Also, after 30 min, RA production in the combination exceeded RA production by cytosol alone.
In the presence of 2 M apo-CRBP (Fig. 2, top panel, filled symbols), the amounts of retinal produced by microsomes alone were comparable quantitatively throughout the incubation to the retinal produced by microsomes in the absence of apo-CRBP. RA production by cytosol, however, was below detection limits before 30 min, when it reached a plateau, which on the average was only 13% of the amount of RA observed in the absence of apo-CRBP (Fig. 2, bottom panel). In the combination of microsomes and cytosol, retinal concentrations reached a plateau of ϳ30 pmol and declined after 60 min (data not shown). RA production by the combination increased throughout the incubation and at all times exceeded RA production by cytosol alone, even though the microsomal protein added to the combination was 5-fold less than that of the cytosolic protein present.
The specific activities of microsomal and cytosolic retinol DHs were also compared in the absence (open symbols) and the presence (filled symbols) of apo-CRBP (Fig. 3). Microsomal retinol DH was linear to 0.1 mg of protein and generated retinal from 5 M holo-CRBP at rates of ϳ23 pmol/min/mg protein in the absence of apo-CRBP and ϳ15 pmol/min/mg protein in the presence of 2 M apo-CRBP. In the absence of apo-CRBP, cytosolic retinol DH activity was linear to 0.25 mg of protein and functioned at a rate of ϳ4 pmol/min/mg of protein, i.e. ϳ6-fold less than the rate due to microsomal retinol DH. In the presence of 2 M apo-CRBP, cytosolic retinol DH activity was not observed below 0.25 mg of protein but was linear between 0.25 and 1 mg of protein. The rate was ϳ0.33 pmol/min/mg of protein or ϳ50-fold less than that of microsomal retinol DH in the presence of apo-CRBP.
RA Synthesis during Titration of Microsomes into Cytosol-Titration of microsomes into cytosol was examined for its im-pact on RA synthesis from holo-CRBP (Fig. 4, top panel). In the absence of apo-CRBP (open circles), cytosol alone produced 20 pmol of RA at a rate of ϳ3 pmol/min/mg of protein. 3 The addition of microsomes (10 g of protein) nearly doubled the quantity of RA produced to 39 pmol. The rate of RA production supported by the added microsomal retinol DH, however, was ϳ63 pmol/min/mg of protein, Ͼ20-fold higher than that of cytosolic retinol DH. In the presence of apo-CRBP (filled circles), RA production was not detected in cytosol alone. Furthermore, at least 50 g of microsomal protein were required before RA synthesis was observed, which occurred at a lower rate (ϳ4 pmol/min/mg of microsomal protein) than in the absence of apo-CRBP.
Inhibition of RA Production from Retinal by apo-CRBP-During the titration of microsomes into cytosol, in the absence of CRBP, retinal was not observed until the microsomal protein reached 0.1 mg, (Fig. 4, bottom panel, open circles). From 0.1 mg of microsomal protein to 0.4 mg of protein, retinal steadily increased, indicating that microsomal retinal production outpaced its cytosolic conversion into RA. In the presence of apo-CRBP (filled circles), the addition of microsomes resulted in retinal accumulation before RA was observed (compare the bottom panel with the top panel, filled circles), the opposite of 3 The rate calculation used for cytosolic retinol DH was 20 pmol of RA/30 min/0.25 mg of cytosolic protein. This amount of protein was used in the calculation rather than the 0.35 mg actually present in the experiment because 0.25 mg of cytosolic protein ends the linear range of cytosolic retinol DH (see Fig. 2). The rate calculation used for microsomal retinol DH was 19 pmol/30 min/0.01 mg. what had occurred in the absence of apo-CRBP. Moreover, markedly more retinal accumulated than in the absence of apo-CRBP. Overall, the data of Fig. 4 suggest that apo-CRBP not only inhibits cytosol retinol DH but also diminishes the conversion by cytosol of retinal into RA.
To test the possibility that apo-CRBP inhibits the conversion of retinal into RA, cytosolic RA production was monitored from retinal bound to CRBP in the presence of increasing concentrations of apo-CRBP (Fig. 5). An apo-CRBP concentration 50% of the CRBP/retinal concentration inhibited Ͻ20% (filled circles). An apo-CRBP concentration equal to the CRBP/retinal concentration of 0.5 M caused the maximum inhibition of 50%. Apo-CRBP up to 9-fold greater than CRBP/retinal resulted in no additional inhibition. These data indicate that CRBP/retinal acted as substrate for RA synthesis, because the concentrations of unbound retinal and the rate of RA synthesis did not coincide, i.e. the rate of retinal synthesis was not dependent on the concentration of unbound retinal. The decrease in RA synthesis with increasing apo-CRBP, therefore, was caused by inhibition of retinal dehydrogenation by apo-CRBP. This phenomenon probably contributes to the inhibition of RA synthesis in the 10kS by the higher concentrations of apo-CRBP (see above).
Relative Contributions of Retinol DHs to RA Synthesis-The specific activities of microsomal retinol DH, assayed in the absence of apo-CRBP ranged from 5-to 22-fold greater than those of cytosolic retinol DH in liver, kidney, lung, and testis (Table I). At least 80% of the total retinol DH units were microsomal in these tissues, except in testis, where 60% of the total retinol DH units were microsomal. In the presence of apo-CRBP, the specific activity of microsomal retinol DH ranged from 13-to 60-fold greater than cytosolic retinol DH.
Under these conditions, greater than 90% of the units were microsomal, except in testis where 80% of the units were microsomal.
Because ethanol inhibits cytosolic retinol DH but not microsomal retinol DH (see below), it was used to provide further insight into the contribution of retinal generated by cytosol to RA synthesis (Fig. 6). Ethanol (hatched bars) inhibited by 90% cytosolic RA synthesis supported by holo-CRBP (compare bars 1 and 2) but stimulated ϳ40% retinal synthesis in microsomes (bars 3 and 4) and in a combination of microsomes and cytosol (bars 5 and 6). These data indicate that ethanol inhibits the cytosolic retinol DH, but neither ethanol nor its metabolite acetaldehyde inhibits the cytosolic retinal DHs and also that cytosolic retinal DHs do not discriminate against cytosolic or microsomally produced retinal. Most importantly, these data show that in the combination of cytosol and microsomes, inhibition of cytosolic retinol DH does not diminish RA synthesis.
Inhibition of Retinol DHs-Microsomal and cytosolic retinol DHs were affected differently by inhibitors (Fig. 7, Table II). Ethanol, octanol, and pentanol, substrates for various medium chain (or classical) alcohol dehydrogenase isozymes, were potent inhibitors of cytosolic retinol DH activity with IC 50 values Ͻ20 M; even at 10 mM, however, these alcohols did not inhibit microsomal retinol DH. The nonalcoholic solvents Me 2 SO and dimethylformamide used in these assays were evaluated to test whether inhibition of cytosolic retinol DH resulted from a solvent effect. Although both solvents did inhibit cytosolic retinol DH, inhibition was never Ͼ50% and required a much higher concentration than ethanol inhibition: 0.4% (v/v) ethanol caused Ͼ90% inhibition; 2% (v/v) of Me 2 SO or dimethylformamide caused 50% inhibition. Neither Me 2 SO nor dimethylformamide inhibited microsomal retinol DH. In contrast to the alcohol substrates, carbenoxolone, an inhibitor of short chain alcohol dehydrogenases (26,27), was an equally potent inhibitor of both microsomal and cytosolic DHs.
The interactions of these compounds with ADH-2 and ADH-3, the two isozymes of the medium chain alcohol dehy- drogenases expressed in rat liver cytosol, were compared with their effects on microsomal and cytosolic retinol DH activities (Table II). ADH-3 is far more sensitive to 4-methylpyrazole than is either retinol DH. Even though the sensitivities of cytosolic retinol DH and ADH-2 to 4-methylpyrazole are similar, carbenoxolone had no impact on ADH-2, thereby distinguishing cytosolic retinol DH and ADH-2. Carbenoxolone did not inhibit ADH-3 either. Unlike the case with retinol DHs, apo-CRBP also did not depress ADH-2 or ADH-3 activity in concentrations up to 10 M. The differences in the K m values of ethanol, pentanol, and octanol for ADHs and their IC 50 values for cytosolic retinol DH further indicate that the structurefunction properties of cytosolic retinol DH and ADH are readily distinguishable.
Multiple Cytosolic Retinol and Retinal DH activities-Cytosolic retinol DH activity proceeded at equivalent rates in the presence of saturating amounts of NAD or NADP (18). To determine whether each cofactor supported catalysis by distinct enzymes, four different batches of rat liver cytosol, prepared routinely during the course of this work, were tested for RA synthesis from holo-CRBP, supported by either NAD or NADP (Table III). With NAD alone, all four synthesized RA at In the assays with pentanol and octanol added in Me 2 SO, 1 l of Me 2 SO was added to each tube. Me 2 SO was used because octanol was not miscible with water. Because Me 2 SO itself causes some inhibition, pentanol was tested in the presence and the absence of Me 2 SO so that the octanol data could be related to the results with pentanol and ethanol. Each point is the average of duplicate experiments.

TABLE I Relative amounts of microsomal and cytosolic retinol DH activities
Rates (pmol/min/mg of protein) of microsomal and cytosolic retinol DH activities were calculated from linear amounts of protein and time with 5 M holo-CRBP in the absence and the presence of 2 M apo-CRBP. Units were calculated from total protein/subcellular fraction of pooled tissues from 20 male Sprague-Dawley rats (ϳ150 g each) and are expressed as pmol/min. The relative contribution of each retinol DH to total retinal production is listed under %.  . c NS, not saturable. d,e Inhibition was determined with ethanol or octanol as substrate, respectively, as described under "Experimental Procedures." similar rates. Although two preparations produced RA with NADP alone with rates 61 and 97% of those supported by NAD, two of the pools had greatly reduced RA production, i.e. 15 and 9% of the rates supported by NAD. Furthermore, retinal did not accumulate in these pools that had low NADP-supported activity, indicating the decreased level of an NADP-dependent cytosolic retinol DH. This inference is consistent with the occurrence of at least two cytosolic retinol DHs with different pyridine nucleotide specificities. Moreover, the cytosolic retinol DH activity supported by NAD was inhibited to a greater extent by ethanol (82% inhibition) than was the NADP-supported activity (30% inhibition). This result provides further support for the presence of two retinol DHs.
The generation of RA in cytosol from holo-CRBP with NADP as the sole cofactor indicated that NADP alone supports cytosolic retinal synthesis. Consistent with this, cytosolic preparations without the NADP-supported cytosolic retinol DH activity, as well as cytosol with the NADP-supported cytosolic retinol DH activity, converted retinal into RA with NADP as sole cofactor (Table IV). Ethanol had a modest or no effect on the rates of RA synthesis from CRBP/retinal.

DISCUSSION
This work demonstrates the primary contribution quantitatively of microsomal retinol DH to RA biosynthesis relative to cytosolic retinol DH activity. Not only was microsomal retinol DH severalfold greater in specific activity than cytosolic retinol DH activity in a comparison of subcellular fractions from four rat tissues, it had a greater capacity to generate retinal (the majority of enzyme units) and was relatively resistant to inhibition by apo-CRBP. Most revealing were the data showing that cytosolic retinol DH could be inhibited in a mixture of cytosolic retinol DH and microsomal retinol DH without decreasing the production of RA (Figs. 1 and 6). For example, 1 M apo-CRBP inhibited cytosolic retinol DH 50% but had no effect on the amount of RA produced by the 10kS fraction, whereas ethanol caused 90% inhibition of cytosolic retinol DH but enhanced RA production in a combination of cytosol and microsomes, similar to its effect on retinal synthesis in microsomes. These results are consistent with either cytosolic retinol DH not functioning in the presence of microsomal retinol DH or microsomal retinol DH compensating for the quantitatively minor production of retinal by cytosolic retinol DH in cases of cytosolic retinol DH inhibition. Notably, under conditions likely to prevail physiologically, i.e. a mixture of holo-CRBP and apo-CRBP with concentrations in the range of ϳ5 and ϳ2 M, respectively, microsomal retinol DH accounted for Ͼ90% of the retinal-generating capacity in three of the four tissues screened.
Because there is a RA response element in the CRBP gene (28) and RA induces CRBP expression in vivo (29), Ottonello et al. have proposed that generation of CRBP by RA may provide apo-CRBP as a signal in a feedback loop to inhibit RA synthesis from cytosolic retinol DH (18). This hypothesis has limitations. Firstly, CRBP may affect the amount of retinol sequestered by cells; as CRBP increases, the concentration of holo-CRBP could increase as long as plasma retinol were available (30). Generation of CRBP, therefore, does not necessarily result in elevated concentrations of apo-CRBP. It seems more reasonable that RA acts as an on/off signal inducing constitutive expression of CRBP rather than acutely regulating relatively modest changes in the concentration of CRBP. The "feedback" loop hypothesis encounters another problem during vitamin A depletion. During times of less than optimal blood and cell retinol concentrations, RA synthesis would need to continue efficiently to generate the active humoral agent, even though substrate concentrations were diminishing. Potent inhibition of cytosolic retinol DH during vitamin A depletion, when apo-CRBP concentrations would increase, appears to be counterproductive to RA generation at a time when RA generation would be needed.
Lack of an obvious role for cytosolic retinol DH in RA generation, because it would be inhibited under normal conditions by the apo-CRBP present, doesn't imply cytosolic retinol DH has no role; perhaps cytosolic retinol DH reduces retinal generated by ␤-carotene metabolism into retinol. Such a function would make sense with respect to the sensitivity of cytosolic retinol DH to inhibition by apo-CRBP. If cytosolic retinol DH were a reductase, inhibiting it during vitamin A depletion would provide increased retinal from ␤-carotene for conversion into RA. Additional work will ultimately address these issues.
This and previous works have demonstrated unequivocally that microsomal retinol DH does not belong to the medium chain ADH family (12)(13)(14)(15). This work has also distinguished between the two known rat liver ADH isozymes and cytosolic retinol DH. Differences in affinities for apo-CRBP, carbenoxolone, and 4-methylpyrazole distinguish cytosolic retinol DH and ADH-3, whereas differences in affinities for apo-CRBP, carbenoxolone, and the short chain alcohols (ethanol, pentanol, and octanol) distinguish cytosolic retinol DH and ADH-2. These results are consistent with our previous results obtained using 4-methylpyrazole to study RA synthesis from retinol not bound with CRBP in cytosols prepared from rat tissues and from the tissues of ADH Ϫ and ADH ϩ deermice (31-33). 4-Methylpyrazole inhibited potently (Ͼ94%) the conversion of unbound retinol into RA catalyzed by liver cytosol from rat or ADH ϩ deermouse. In cytosol from the ADH Ϫ deermouse, however, which showed only 13% of the retinol DH activity of the ADH ϩ deermouse, 4-methylpyrazole was a much less effective inhibitor. Thus, these previous data had demonstrated the presence of a quantitatively minor retinol DH activity in liver cytosol that was not affected markedly by 4-methylpyrazole. The present data suggest that the cytosolic retinol DH activity that recognizes holo-CRBP as substrate substantially accounts for this 4-methylpyrazole-insensitive DH. The earlier report by Ottonello et al. (18) had concluded that cytosolic retinol DH was distinct from ADH, in part because 1 mM pyrazole did not  inhibit cytosolic retinol DH; however, 1 mM pyrazole does not inhibit ADH-2 (24). On the other hand, carbenoxolone inhibition characterizes several short chain dehydrogenases/reductases but not the medium chain ADHs (26,27). Whether cytosolic retinol DH, like microsomal retinol DH, belongs to the short chain superfamily remains to be established. There has been some speculation that an ADH isozyme may contribute to RA biosynthesis, in addition to metabolizing various xenobiotic and endogenous long and short chain alcohols. The proposed isozyme has been revised, however, as evidence has eliminated candidates. There are convincing arguments against this hypothesis, however, that include the requirements for control and specificity in RA biogenesis; the occurrence of holo-CRBP as the predominant substrate available in vivo; the ability of cytosol of the ADH Ϫ deermouse mutant to convert retinol into RA (32,33), and the expression of distinct enzymes (microsomal and cytosolic retinol DHs) with extraordinary built in specificity, i.e. they recognize holo-CRBP as substrate (12)(13)(14)(15).
Ethanol inhibits cytosolic retinol DH potently despite differences between cytosolic retinol DH and the known liver ADH isozymes. Thus, this is the first demonstration of ethanol inhibition of a retinol DH, identified by its affinity for holo-CRBP (0.4 M K m , rat liver cytosolic retinol DH, this work; 0.8 M K m , calf liver cytosolic retinol DH, Ref. 18). The impact, if any, however, of ethanol consumption on the retinoid humoral system cannot be predicted from this observation, especially because ethanol stimulates RA biogenesis in microsomes and in a combination of microsomes and cytosol. Certainly chronic excessive ethanol consumption decreases hepatic vitamin A stores (34,35), and alcoholism leads to testicular atrophy and loss of testes function (36 -38), similar to vitamin A deficiency (39,40). Because of these observations, the possibility has been proposed that the medium chain ADHs catalyze retinol metabolism in vivo and thereby ethanol may competitively inhibit retinol metabolism (34,35). For the reasons discussed above, the potential involvement physiologically of an ADH isozyme in retinol metabolism remains in question. But our data suggest an alternative model, namely, that a cytosolic retinol DH, which is not a known ADH, is present and is inhibited by low concentrations of ethanol.
Another significant result was the variability among different batches of rat liver cytosol in the NADP-supported cytosolic retinol DH-catalyzed reaction in contrast to the constancy of the NAD-supported reaction. This suggests two or more cytosolic retinol DHs occur, an NAD-dependent enzyme and a NADPdependent enzyme. More potent ethanol inhibition of the NAD than the NADP-dependent activity (80 versus 30% inhibition, respectively, with 2% ethanol) supports the occurrence of two distinct enzymes.
Our data also show that either NAD or NADP can support the conversion of retinal into RA. Three isozymes with retinal dehydrogenase activity have been separated by anion exchange chromatography of cytosol from rat liver. The major one, representing ϳ67% of the total liver cytosolic NAD-dependent units, has been purified and does not use NADP as cofactor (16). The other isozymes, and additional isozymes resolved chromatographically from kidney and testis, have not been tested with NADP and may not differ from the NADP-dependent activity observed here. Nevertheless, this result, along with the observation of cytosolic retinol DH multiplicity, clearly broadens the scope of enzymes and/or pathways involved in RA biogenesis.
In summary, several new insights have resulted from this work. The quantitatively dominant contribution has been demonstrated of microsomal versus cytosolic retinol DH to RA biogenesis, including the ability of microsomal retinol DH to supplant retinal produced by cytosolic retinol DH. At least two cytosolic retinol DHs have been demonstrated: a NAD-dependent DH and a NADP-dependent DH. Intriguingly, ethanol in micromolar concentrations inhibits cytosolic retinol DH-catalyzed RA synthesis, despite the differences between cytosolic retinol DH and the known ADHs, and stimulates RA synthesis by microsomes or a combination of microsomes and cytosol. Finally, the complex nature of retinal DHs has been confirmed and extended. It is becoming more apparent that RA biogenesis involves intricate metabolic pathways that involve multiple enzymes that recognize both liganded and nonliganded forms of CRBP.