The specificity of alcohol dehydrogenase with cis-retinoids

Activity with 11-cis-retinol and localization in retina


X. Parés, Department of Biochemistry and Molecular Biology, Faculty of Sciences, Universitat Autònoma de Barcelona, E-08193 Bellaterra, Barcelona, Spain.
Fax: + 34 93 5811264, Tel.: + 34 93 5813026,


Studies in knockout mice support the involvement of alcohol dehydrogenases ADH1 and ADH4 in retinoid metabolism, although kinetics with retinoids are not known for the mouse enzymes. Moreover, a role of alcohol dehydrogenase (ADH) in the eye retinoid interconversions cannot be ascertained due to the lack of information on the kinetics with 11-cis-retinoids. We report here the kinetics of human ADH1B1, ADH1B2, ADH4, and mouse ADH1 and ADH4 with all-trans-, 7-cis-, 9-cis-, 11-cis- and 13-cis-isomers of retinol and retinal. These retinoids are substrates for all enzymes tested, except the 13-cis isomers which are not used by ADH1. In general, human and mouse ADH4 exhibit similar activity, higher than that of ADH1, while mouse ADH1 is more efficient than the homologous human enzymes. All tested ADHs use 11-cis-retinoids efficiently. ADH4 shows much higher kcat/Km values for 11-cis-retinol oxidation than for 11-cis-retinal reduction, a unique property among mammalian ADHs for any alcohol/aldehyde substrate pair. Docking simulations and the kinetic properties of the human ADH4 M141L mutant demonstrated that residue 141, in the middle region of the active site, is essential for such ADH4 specificity. The distinct kinetics of ADH4 with 11-cis-retinol, its wide specificity with retinol isomers and its immunolocalization in several retinal cell layers, including pigment epithelium, support a role of this enzyme in the various retinol oxidations that occur in the retina. Cytosolic ADH4 activity may complement the isomer-specific microsomal enzymes involved in photopigment regeneration and retinoic acid synthesis.


alcohol dehydrogenase


cellular retinaldehyde binding protein


cellular retinol binding protein


3,3′-diaminobenzidine tetrahydrochloride


retinal pigment epithelium


short-chain dehydrogenase/reductase

Retinoids are essential in several physiological processes such as development, growth and cellular maintenance [1,2]. The active forms of retinol are its oxidized derivatives all-trans- and 9-cis-retinoic acid which perform their function through the binding to specific nuclear receptors [3,4]. Retinoic acids are synthesized by two enzymatic reactions which include retinol oxidation to retinal, and oxidation of retinal to retinoic acid. Two types of enzymes have been implicated in the first reaction: the alcohol dehydrogenases (ADH) of the medium-chain dehydrogensase/reductase family and the retinol dehydrogenases of the short-chain dehydrogenase/reductase (SDR) family [5]. In mammals, ADH is a cytosolic NAD+-dependent enzyme formed by two subunits of 40 kDa, with two zinc atoms per subunit [6]. Genomic studies indicate that five ADH classes (ADH1–ADH5) exist in mammals [7]. It is well established that ADH1 and ADH4 [5,8], and to a lesser extent ADH2 [9], are involved in retinoid metabolism. Recently, it has been proposed that ADH3, the ubiquitous enzyme responsible for formaldehyde elimination, could also have a role in retinoic acid generation in vivo[10]. Nevertheless, the high activity toward retinoids and the spatiotemporal colocalization of ADH1 and ADH4 with retinoic acid during embryogenesis and in adult tissues [11,12], suggest a major role of these two enzymes in retinoid metabolism. Null-mutant mice to ADH1 or ADH4 show a normal development, but a reduced retinol oxidation, and indicate that each enzyme plays a distinct role in vivo[8].

Retinol dehydrogenases of the SDR family are enzymes of 25–38 kDa per subunit and, in contrast to ADH, do not require a metal ion in the active site [13]; they are microsomal enzymes and use NAD(H) or NADP(H) [5]. Some retinol dehydrogenases can oxidize retinol bound to cellular retinoid binding protein (CRBP), which constitutes the major form of retinol within the cell [5,14,15]. However, disruption of the CRBPI gene has shown that the CRBP protein is essential for retinyl ester storage, but not for retinoic acid synthesis [16], supporting the notion that enzymes which do not use CRBP-retinol, such as ADH [17], could contribute to retinoid metabolism.

11-cis-retinal bound to opsin is the chromophore of the retina. The absorption of one photon produces the isomerization of 11-cis-retinal to all-trans-retinal, which constitutes the first step of the vision process [18]. A series of reactions, known as the visual cycle, will then regenerate 11-cis-retinal. All-trans-retinol dehydrogenase, an SDR enzyme, reduces the all-trans-retinal formed to all-trans-retinol in the rod outer segments. The retinoid is then transported to the retinal pigment epithelium (RPE) where the visual cycle is completed. All-trans-retinol could be stored there as retinyl esters, isomerized to 11-cis-retinol, and finally oxidized to 11-cis-retinal [18]. In addition, 11-cis-retinal has to be produced in the retina to generate the photopigments of cones [19] and of the photosensitive ganglion cells [20]. Finally, oxidation of all-trans-retinol is also required for the synthesis of retinoic acid, necessary for retina functions [21–23]. Although different microsomal SDR retinol dehydrogenases have been proposed to play an essential role in each oxidation step [18,19], cytosolic ADH4 has been purified from retina [24] and its activity has been detected in RPE [25].

In the present report, we have determined the kinetic constants of ADH1 and ADH4 toward cis-retinol and cis-retinal isomers. We used the human enzymes, because of their biomedical interest, but also the mouse ADHs as much is known on the involvement of ADH in retinol metabolism from the knockout experiments, but little information was available on the mouse ADH kinetics with retinoids [26]. An important finding has been the demonstration of 11-cis-retinol dehydrogenase activity in both ADH1 and ADH4 enzymes, which suggests a contribution of ADH in the photopigment regeneration. This has been further supported by the immunolocalization of ADH4 in the RPE and in several retinal cell layers. We have also explored the molecular basis of the ADH4 specificity with retinoids by docking simulations on the crystallographic structures.

Experimental procedures

Preparation of full-length cDNAs for human and mouse ADHs

Human ADH1B1 cDNA, cloned in the vector pT4 [27], was a gift from J.-O. Höög (Karolinska Institute, Stockholm). We designed two primers to amplify the full-length cDNA by polymerase chain reaction and introduced restriction sites (underlined) for BamHI at the 5′ end (5′-CTATCGGATCCATGAGCACAGCAGGAAAAG-3′) and for EcoRI at the 3′ end (5′-CCACTTGAATTCTCAAAACGTCAGGACGGT-3′). Double digestion with BamHI and EcoRI allowed the cloning in the expression vector pGEX-4T-2 (Amersham Pharmacia Biotech). Human ADH1B2 cDNA was prepared from ADH1B1 cDNA using the ADH1B1 cDNA cloned in the expression vector pGEX-4T-2 as follows. Based on the Quickchange™ Site-Directed Mutagenesis Kit method (Stratagene), we designed two primers (5′-GAATCTGTCACACAGATGACCACGTGG-3′, amino acid positions 44–52 and 5′-GTCATCTGTGTGACAGATTCCTACAGCC-3′, amino acid positions 42–50) to introduce the mutation R47H by PCR. Mutated nucleotides are underlined.

The cDNA encoding for human ADH4 was amplified by PCR using as a template the full-length cDNA cloned into the vector pET-5a [28], and two primers to introduce the same restriction sites as in the case of ADH1B1 (5′-CTATCGGATCCATGGGCACTGTTGGAAAAG-3′ and 5′-CCACTTGAATTCTCAAAACGTCAGGACCGT-3′), for the cloning into pGEX-4T-2. The same mutagenesis protocol as that used to prepare human ADH1B2, was followed for the ADH4 M141L mutant, using two specific primers (5′-CACCACTTCCTGAACACCAGTACATTTAC-3′, amino acid positions 138–146, and 5′-CTGGTG TTCAGGAAGTGGTGGACTGGTTTG-3′, amino acid positions 134–144), and the ADH4 cDNA cloned in the expression vector pGEX-4T-2 as a template. Mouse ADH1 and ADH4 cDNA, both cloned in the vector pGEX-4T-2, were prepared as reported by Deltour et al. [29]. Full-length cDNAs were sequenced by Oswel Research Products Ltd (University of Southampton, UK).

Expression and purification of ADH proteins

Escherichia coli BL21 cells, containing human ADH1B1, ADH1B2, ADH4, ADH4 M141L mutant, or mouse ADH1 or ADH4 cDNA, cloned in pGEX-4T-2, were grown in 2 L of 2× YT medium until stationary phase, at 25 °C. Zinc sulfate (10 µm) was added prior to induction with 0.1 mm isopropyl thio-β-d-galactoside (Roche Molecular Biochemicals), for 15 h at 22 °C. Cells were centrifuged at 2800 g, for 15 min at 4 °C, and pellets were frozen at −80 °C to facilitate cell lysis. Pellets were thawed and resuspended in 100 mm Tris/HCl, pH 7.0, 2.5 mm dithiothreitol (Sigma), 10% glycerol, 0.2 m sodium chloride, 10 µm zinc sulfate, and incubated with lysozyme (1 mg·mL−1, Sigma), for 30 min in an ice bath. The suspension was sonicated and the resulting homogenate was incubated with 1% (v/v) Triton X-100 for 30 min, and then treated with DNase (1 µg·mL−1, Roche Molecular Biochemicals) for 30 min at room temperature, to reduce sample viscosity. The homogenate was then centrifuged at 16 000 g for 30 min. The supernatant, containing the ADH-glutathione-S-transferase fusion protein, was incubated with Glutathione-Sepharose 4B (Amersham Pharmacia Biotech), for 15 h at room temperature, and after washing with 100 mm Tris/HCl, pH 7.0, 2.5 mm dithiothreitol, 10% glycerol, 0.2 m sodium chloride, 10 µm zinc sulfate, the elution of the ADH was performed by thrombin digestion (10 U·mg−1 protein, Amersham Pharmacia Biotech), for 15 h at room temperature. Protein homogeneity was checked by electrophoresis on SDS/PAGE followed by the Coommassie® Brilliant Blue (Sigma) stain technique. Protein concentration was determined by a dye binding assay (Bio-Rad) using bovine serum albumin as standard [30].

Enzyme kinetics

Standard ADH activity was determined by measuring the change in NADH absorbance at 340 nm (inline image =6220 m−1·min−1) in a Varian Cary 400 spectrophotometer, at 25 °C. One unit (U) of ADH activity is defined as the amount of enzyme required to produce 1 µmol NADH per min at 25 °C. Activity was determined in 0.1 m glycine/NaOH, pH 10.5, for all ADHs except for ADH1B2 that was determined in 0.1 m glycine/NaOH, pH 8.5. The following cofactor and substrate concentrations were used: 2.4 mm NAD+ (Sigma) and 30 mm ethanol for human ADH1B1 and ADH1B2; 2.4 mm NAD+ and 100 mm ethanol for human ADH4; 0.3 mm NAD+ and 10 mm ethanol for mouse ADH1; 2.4 mm NAD+ and 2.5 m ethanol for mouse ADH4.

Commercially available retinoids were obtained from Sigma. 7-cis-retinal was prepared from the corresponding methyl 7-cis-retinoate, obtained by Suzuki cross-coupling, as described by Alvarez et al. [32]. 11-cis-retinol resulted from a highly stereoselective Wittig reaction [33], and it was used to prepare 11-cis-retinal by oxidation with MnO2[34]. For the synthesized retinoids, the retinals were the forms of storage. Synthesis of 7-cis-retinol, 9-cis-retinol and 11-cis-retinol were performed by reduction of the corresponding aldehydes with sodium borohydride immediately before use. The purity of the products was checked by reverse-phase HPLC [28]. The calculated molar absorption coefficients in the assay buffer were ε329 = 25 800 m−1·min−1 for 11-cis-retinol, ε380 = 19 000 m−1·min−1 and ε400 = 15 600 m−1·min−1 for 11-cis-retinal and ε376 = 25 100 m−1·min−1 and ε400 = 17 800 m−1·min−1 for 7-cis-retinal. Because molar absorption coefficients for 7-cis-retinol in any organic solvent were not found in the literature, we determined a value in ethanol of ε315 = 42 000 m−1·min−1, which served to calculate an ε318 = 40 900 m−1·min−1 in the assay buffer.

Activity with retinoids was determined by following the change in absorbance at 400 nm, using the molar absorption coefficients described above and those previously published [31]. Retinoid (3 mg) was dissolved in 250 µL acetone, and 175 µL of this solution was diluted in 25 mL of 0.1 m sodium phosphate, pH 7.5, 0.02% Tween-80. Retinoid solutions were prepared under dim red light and were kept protected from light at 4 °C, to prevent degradation. The final acetone concentration in the assay was lower than 0.12 mm.

Retinol oxidation was measured with 2.4 mm NAD+ or 0.3 mm NAD+ (mouse ADH1) using 1 cm pathlength cuvettes, while retinal reduction was measured with 1 mm NADH (Sigma) or 0.77 mm NADH (human ADH4) in 0.2 cm pathlength cuvettes. Retinoid concentration ranged from 0.1× Km to 10× Km. Activities were measured from the initial slope of the progress curves, registered for 3 min. During this time, the activity rate was linear. No photoisomerization of 11-cis-retinal to all-trans-retinal was detected during the assay, as assessed by the UV visible absorption spectra. Kinetic constants were calculated using the grafit program (version 5.0, Erithacus Software Limited, Horley, Surrey, UK), and the reported results were expressed as the mean ± S.E.M of at least three independent determinations. Catalytic constant (kcat) values were calculated using an Mr of 80 000 for the ADH dimer.

Substrate-docking simulations

Docking simulations were performed in a Silicon Graphics Indigo 2 R10000 workstation, using the icm program (version 2.7, Molsoft LLC, 1997; La Jolla, CA, USA). Crystallographic coordinates of human ADH4 [35] were used to simulate its interaction with all-trans, 9-cis and 11-cis isomers of retinol and retinal. Crystallographic coordinates of human ADH1B1 [36] and of the mutant M141L [37] were used to simulate their interaction with 11-cis-retinal. In all cases, a nonrigid docking based on a Monte Carlo procedure was employed with 500 000 iterative cycles, allowing free movement of the rotatable bonds of the substrate and of the χ angles of the residues inside a 5 Å radius from the docked substrate, and using distance restraints as described previously [31].


Adult Sprague–Dawley rats were used. Animal protocols were approved by the Ethical Committee of the Universitat Autònoma de Barcelona. After decapitation, eyes were immediately dissected and washed in NaCl/Pi (10 mm Na2HPO4, 2 mm KH2PO4, pH 7.3, 0.14 m NaCl, 2.7 mm KCl). Lens were removed and the eye samples were immersed in 4% (w/v) paraformaldehyde (freshly prepared in NaCl/Pi) for 12 h. Eyes were embedded in paraffin and sliced into serial 8 µm sections using a Leica microtome, attached to coated microscope slides. Sections were dried at 37 °C for at least 12 h. Eye sections were dewed with xylene and hydrated through a graded series of decreasing ethanol concentrations (100% to 30%), followed by treatment with 0.5% (v/v) H2O2 in methanol for 20 min to eliminate endogenous peroxidase activity. Then, the sections were incubated with purified polyclonal antibodies against mouse ADH4 (1 : 100 dilution) [12], for 1 h. The ADH4 antibodies were highly specific for ADH4; they did not recognize ADH1 or ADH3, and only the ADH4 band was observed in a Western blot of eye homogenate [12]. The bound primary antibody was visualized by the Vectastain Elite ABC kit (Vector Laboratories, Inc.), using biotinylated antirabbit IgG as a second antibody and a complex avidin-biotin conjugated with peroxidase. 3,3′-diaminobenzidine tetrahydrochloride (Sigma) was used as a chromogenic reagent. Sections were incubated, for 10 min, in NaCl/Tris (0.25 mm Tris/HCl, pH 7.4, 0.14 m NaCl, 2.7 mm KCl) containing 0.05% (w/v) 3,3′-diaminobenzidine tetrahydrochloride and 0.033% (v/v) H2O2. Tissues were then rinsed in NaCl/Tris, dehydrated and mounted using a xylene-based medium (ENTELLAN® neu, Merck).

Negative immunostaining controls were made by the preadsorption of the ADH4 antibody with an excess of purified recombinant ADH4, or by the omission of the primary antibody. Slides containing adjacent sections were stained with hematoxylin (Vector Laboratories, Inc.), dehydrated through a graded series of ethanol concentrations, followed by two xylene washes, and cover-slipped with ENTELLAN® neu. Examination of eye sections and image acquisition of immunohistochemical results were performed as reported previously [38].


Expression and purification of ADHs

Human ADH1B1, ADH1B2, ADH4 and ADH4 M141L, and mouse ADH1 and ADH4 have been expressed at high levels in E. coli BL21 cells and purified to homogeneity. The usual yield of pure protein obtained, ranged from 0.1 mg·L−1 culture for mouse ADH4 to 4–5 mg·L−1 culture for human ADH4 and mouse ADH1. Specific activities, measured under standard conditions, were 0.2 U·mg−1 for human ADH1B1 and 15 U·mg−1 for human ADH1B2, values comparable with those reported elsewhere [39,40]. In contrast, specific activities for mouse ADH1 (3.1 U·mg−1) and mouse ADH4 (130 U·mg−1) were higher than those reported previously for enzymes purified from mouse tissues [41]. The specific activities for human ADH4 and ADH4 M141L were 67 U·mg−1 and 36 U·mg−1, respectively.

Kinetic constants of mouse enzymes toward aliphatic alcohols

The kinetic constants with ethanol and hexanol, for the recombinant ADH1 and ADH4, were determined at pH 7.5 and 10.5 (Table 1). For both enzymes, the Km values with hexanol were much lower than those with ethanol, resulting in a higher catalytic efficiency for the substrate with the longer carbon chain; a general property of mammalian ADH. Mouse ADH1 showed similar kinetic properties to rat ADH1 [42] and to human ADH1C [40]. Mouse ADH4 showed similar kinetic constants to rat ADH4 but it exhibited much higher Km values for ethanol, at pH 7.5, than the human enzyme [31,42].

Table 1. Kinetic constants for recombinant mouse alcohol dehydrogenases. Activities were determined in 0.1 m sodium phosphate (pH 7.5) or 0.1 m glycine (pH 10.5), using 0.3 mm NAD+ for ADH1 or 2.4 mm NAD+ for ADH4, at 25 °C.
SubstrateConstantUnitspH 7.5pH 10.5
EthanolKm(mm)0.48 ± 0.091625 ± 3700.83 ± 0.06255 ± 60
kcat(min−1)115 ± 52480 ± 225265 ± 512900 ± 905
kcat/Km(mm−1·min−1)240 ± 451.5 ± 0.4320 ± 2551 ± 13
HexanolKm(mm)0.085 ± 0.0031.9 ± 0.10.006 ± 0.0010.63 ± 0.03
kcat(min−1)27 ± 11850 ± 500230 ± 55190 ± 105
kcat/Km(mm−1·min−1)315 ± 15970 ± 9036200 ± 70008230 ± 400

Kinetic constants with retinoids

Kinetic constants with all-trans-retinol and all-trans-retinal, and with the cis-isomers of retinol and retinal (7-cis-, 9-cis-, 11-cis- and 13-cis-), were determined for human and mouse ADH1 and ADH4 enzymes (Tables 2 and 3, respectively). Except for the 13-cis isomers, all enzymes showed similar Km values for all retinoids assayed, ranging from 8 to 35 µm for retinols and from 4 to 28 µm for retinals. However, ADH4 exhibited, in general, higher kcat values than ADH1, thus having higher catalytic efficiencies (kcat/Km). Mouse ADH1 was the best class I ADH tested, in terms of catalytic efficiency, followed by human ADH1B2. Human ADH1B1 was a poor enzyme toward retinoids, with catalytic constants being lower than 2 min−1. The ADH4 enzymes from the two species showed similar kinetic properties.

Table 2. Kinetic constants of alcohol dehydrogenases with retinol isomers. Activities were determined in 0.1 m sodium phosphate, pH 7.5, 0.02% Tween-80, using 2.4 mm NAD+ (0.3 mm for mouse ADH1), at 25 °C. NA, no activity up to 150 µm substrate; ND, not determined.
  1. a Data taken from [28].

all-trans-retinolKmm)30 ± 833 ± 915 ± 4a31 ± 523 ± 4
kcat(min−1)1.7 ± 0.115 ± 167 ± 10a55 ± 3125 ± 5
kcat/Km(mm−1·min−1)57 ± 16455 ± 1304500 ± 1370a1775 ± 3005480 ± 990
7-cis-retinolKmm)NDND27 ± 3NDND
kcat(min−1)  425 ± 15  
kcat/Km(mm−1·min−1)  15740 ± 1815  
9-cis-retinolKmm)11 ± 323 ± 536 ± 4a8 ± 1.121 ± 3
kcat(min−1)0.89 ± 0.063.6 ± 0.4475 ± 44a23 ± 1340 ± 15
kcat/Km(mm−1·min−1)81 ± 23160 ± 4013200 ± 2020a2900 ± 45516100 ± 2400
11-cis-retinolKmm)35 ± 718 ± 428 ± 534 ± 923 ± 6
kcat(min−1)0.85 ± 0.049.5 ± 1.0190 ± 2025 ± 2225 ± 5
kcat/Km(mm−1·min−1)24 ± 5530 ± 556790 ± 1400735 ± 2059740 ± 2550
13-cis-retinolKmm)NANA26 ± 3NDND
kcat(min−1)  0.76 ± 0.02  
kcat/Km(mm−1·min−1)  29 ± 3  
Table 3. Kinetic constants of alcohol dehydrogenases with retinal isomers. Activities were determined in 0.1 m sodium phosphate, pH 7.5, 0.02% Tween-80, using 1 mm NADH (0.77 mm for human ADH4), at 25 °C. NA, no activity up to 150 µm substrate; ND, not determined.
  1. a Data taken from [28].

all-trans-retinalKmm)11 ± 212 ± 334 ± 6a9.3 ± 1.111 ± 2
kcat(min−1)1.1 ± 0.133 ± 3110 ± 25a19 ± 133 ± 2
kcat/Km(mm−1·min−1)100 ± 202750 ± 7303300 ± 960a2045 ± 2653000 ± 525
7-cis-retinalKmm)NDND8 ± 2NDND
kcat(min−1)  80 ± 5  
kcat/Km(mm−1·min−1)  10000 ± 2610  
9-cis-retinalKmm)11 ± 24.1 ± 0.921 ± 5a13 ± 215 ± 2
kcat(min−1)1.8 ± 0.22.7 ± 0.1190 ± 25a17 ± 1190 ± 5
kcat/Km(mm−1·min−1)165 ± 35660 ± 1458980 ± 2350a1310 ± 21512800 ± 1740
11-cis-retinalKmm)15 ± 316 ± 421 ± 326 ± 728 ± 3
kcat(min−1)0.31 ± 0.023.4 ± 0.118 ± 27.4 ± 0.534 ± 1
kcat/Km(mm−1·min−1)21 ± 4215 ± 55860 ± 155285 ± 801215 ± 135
13-cis-retinalKmm)NANA25 ± 6NDND
kcat(min−1)  11 ± 1  
kcat/Km(mm−1·min−1)  440 ± 115  

All tested ADH1 and ADH4 enzymes used 11-cis- retinoids. Human and mouse ADH4 efficiently oxidized 11-cis-retinol, while the ADH1 enzymes showed lower activity (Table 2). All enzymes exhibited comparable activity for the two reaction directions with any retinol/retinal pair, except ADH4 with 11-cis-retinoids. Interestingly, the two ADH4 enzymes showed an 8-fold higher kcat/Km value with 11-cis-retinol than with 11-cis-retinal (Tables 2 and 3), while the Km values were comparable. ADH4 therefore exhibits a strong and unique specificity for the 11-cis-retinol oxidation over the 11-cis-retinal reduction.

Previously we reported that ADH4 had no activity toward 13-cis-isomers [28,31]. However, by using a higher enzyme concentration (above 30 µg·mL−1) in the assay, we show here that human ADH4 is in fact also active with 13-cis-retinoids, although with low kcat values (Tables 2 and 3). Human ADH1 enzymes were not found to be active with 13-cis-retinoids, although a low activity had been previously reported with 13-cis-retinal [43].

7-cis-retinoids have not been described physiologically, but their kinetic study gives an estimate of the effect of the cis-bond position on the substrate specificity of human ADH4. The 7-cis- and 9-cis-retinol and retinal isomers were the most active substrates, in terms of kcat/Km, for ADH4, followed by 11-cis-retinol (Tables 2 and 3). In contrast, ADH1 generally exhibited more activity toward all-trans-retinoids.

The specificity of human ADH4 with retinoids

The structural basis for the retinoid specificity of ADH4, was studied by docking all-trans-, 9-cis- and 11-cis- isomers of retinol and retinal into human ADH4-NAD(H) binary complex (Fig. 1). In all cases, except for 11-cis-retinal, retinoids are properly placed in the substrate-binding pocket, with an atomic distance between the functional oxygen atom and the catalytic Zn shorter than 3.16 Å. Moreover, the distance between the O atom of retinoids and the C4 of the nicotinamide ring, involved in the hydride transfer, is lower than 4.83 Å. In contrast, both distances are notably increased in the docked 11-cis-retinal, suggesting that the distinct location of the substrate in the binding pocket of ADH4 is the reason for the low activity observed with this retinal isomer.

Figure 1.

Docking of retinol and retinal isomers to human ADH4. Schematic representation of human ADH4 bound to different isomers of retinol [(A) all-trans-; (C) 9-cis-; (E) 11-cis-] and retinal [(B) all-trans-; (D) 9-cis-; (F) 11-cis-] is shown with the simultaneous binding of NAD+ for docked retinol, or NADH for retinal. Dashed lines represent atomic distances (in Å) from the oxygen atom of the retinoid functional group to the catalytic Zn and to the C4 of the coenzyme.

The interaction of 11-cis-retinal with ADH1B1 was also studied and compared with that of ADH4 (Fig. 2A–D). 11-cis-retinal was well placed in the ADH1B1 substrate-binding pocket, as suggested by the short distance to the catalytic Zn (Figs 2C,D), in contrast to what is observed in ADH4 (Figs 2A,B). The middle region of the substrate-binding pocket of ADH4 is characterized by two Met residues at positions 57 and 141, resulting in a narrow space in comparison to ADH1B1, where these two residues are Leu. On the other hand, docking studies showed that the cis bond of 11-cis-retinoids is facing residues 57 and 141, indicating that they could have a key role in the interaction with 11-cis-retinoids. To check this possibility, 11-cis-retinal was docked to the human ADH4 M141L crystallographic structure (Figs 2E,F). The M141L substitution widens the middle part of the hydrophobic tunnel. As a result, the reactive group of 11-cis-retinal was found best oriented, and placed at a productive distance from the catalytic Zn.

Figure 2.

Docking of 11-cis-retinal to human ADH4, ADH1B1 and ADH4 M141L. (A), (C) and (E), representation of 11-cis-retinal in the substrate-binding pocket of ADH4, ADH1B1 and ADH4 M141L, respectively, viewed from the outer part to the inner part of the hydrophobic tunnel, where the catalytic Zn atom is found (shown as van der Waals radius sphere). Residues 57 and 141 are visualized with their accessible surface. The wideness of the hydrophobic tunnel, measured as the atomic distance between the two residues is high for ADH1B1 (C), low for ADH4 (A) and intermediate for the mutant (E). (B), (D) and (F), schematic representation of an almost lateral view of the substrate-binding pocket with 11-cis-retinal docked to human ADH4, ADH1B1 and ADH4 M141L, respectively, showing the atomic distances (in Å) from the Zn atom to the oxygen of the substrate, and between residues 57 and 141.

To examine the influence of residue 141 on the kinetics of ADH4 with retinoids, the human ADH4 M141L mutant was prepared, purified to homogeneity and characterized. The kinetic constants toward ethanol and hexanol (Table 4) were comparable to those previously reported for this mutant [37]. Thus, it showed half of the catalytic efficiency of the wild-type enzyme, while the Km values did not change. Kinetic constants toward different retinoid isomers were also determined (Table 4). ADH4 M141L showed high catalytic efficiency toward all-trans- and 9-cis-retinoids and, in contrast to ADH4, it had similar catalytic efficiencies toward 11-cis-retinol and 11-cis-retinal. Thus, while ADH4 showed a strong preference for 11-cis-retinol oxidation over 11-cis-retinal reduction, this was not observed in ADH1 enzymes or in the ADH4 M141L mutant. The middle region of the substrate-binding pocket (namely position 141) is therefore essential to define the higher specificity of ADH4 for the oxidation direction, in the interconversion of 11-cis-retinoids.

Table 4. Kinetic constants of human ADH4 M141L. Activities were determined in 0.1 m sodium phosphate, pH 7.5, using 2.4 mm NAD+ for alcohol oxidation or 0.77 mm NADH for aldehyde reduction, at 25 °C. 0.02% Tween-80 was present in the assay with retinoids.
Ethanol40000 ± 40001105 ± 2528 ± 3
Hexanol48 ± 7440 ± 109145 ± 1355
all-trans-retinol9 ± 220 ± 22220 ± 540
9-cis-retinol29 ± 3100 ± 53450 ± 395
11-cis-retinol24 ± 440 ± 21670 ± 290
all-trans-retinal17 ± 331 ± 21825 ± 345
9-cis-retinal22 ± 345 ± 12045 ± 280
11-cis-retinal8 ± 118 ± 12250 ± 310
13-cis-retinal27 ± 77.1 ± 0.7265 ± 75

Localization of ADH4 in retina

ADH4 has been immunolocalized in rat eye sections, using mouse-ADH4 polyclonal antibodies. The enzyme was detected in the RPE and it was widely distributed in the inner layers of the retina (Fig. 3). ADH4 was present in the outer nuclear, inner nuclear, inner plexiform and ganglion cell layers. The signal was absent in the choroid and outer plexiform layer, and in the outer and inner segments of the photoreceptor cells.

Figure 3.

Localization of ADH4 in rat retina by immunohistochemistry. (A) Retina section stained with hematoxylin. (B) Immunolocalization of ADH4 in retina. ADH4 is detected in the retinal pigment epithelium (RPE) and it is extensively distributed in the retina. ADH4 is found in the outer (ONL) and the inner (INL) nuclear layers, in the inner plexiform layer (IPL) and in the ganglion cell layer (GCL), but it is not detected in the choroid (Ch), in the outer plexiform layer (OPL) and in the photoreceptor outer (OS) and inner (IS) segments. (C) No signal is found when sections are incubated with the biotinylated rabbit IgG antibody without preincubation with the ADH4 antibody. Calibration bar (50 µm) shown in (B) applies to all panels.


We have here presented a complete kinetic characterization of recombinant human and mouse ADH1 and ADH4 with retinoids. From these results and previous reports on the human [28,43] and rat [31,42] enzymes, it can be concluded that in mammals ADH4 uses retinoids more efficiently than ADH1. In contrast, activity with ethanol is lower for ADH4. The remarkable difference in Km values for ethanol showed by rodent ADH4 (approximately 2 m) ([31] and this work) and human ADH4 (40 mm) [44], has been related to a single residue exchange (Val294 in human vs. Ala294 in the rat and mouse ADH4), which makes the active site wider in the rodent ADH4, resulting in a decreased affinity toward ethanol [31]. This substitution has apparently not affected the activity with retinoids, because human and rat ADH4 [31,42], and now also the mouse enzyme, show high catalytic efficiencies with these substrates, which supports a physiological role more related to the redox transformations of large substrates, like retinoids, rather than the metabolism of short-chain alcohols.

The involvement of ADH4 in specific retinoid metabolism is supported by the kinetic studies (present work and [28,43,45,46]), by its presence in several epithelial cells that require retinoic acid for differentiation [47], by its colocalization with retinoic acid during development [11,12] and by the decrease of retinoic acid production in the ADH4 knockout mice [8]. Nevertheless, although ADH4 is usually more efficient, the activity of ADH1 with retinoids should not be neglected, particularly for human ADH1B2 and mouse ADH1. This is consistent with a role of ADH1 in the clearance of retinoid excess as proposed from knockout studies in mouse [48].

Activity of ADH with 11-cis-retinoids had not been reported before. ADH1 and ADH4 reversibly transform 11-cis-retinol to 11-cis-retinal with high efficiency. This is a relevant result because it provides the possibility for ADH of being involved in the photopigment regeneration. In this regard emphasis will be put on ADH4 in the present discussion, because this is the major ADH form in the mammalian eye tissues [25]. ADH4 efficiently uses 11-cis-retinol, but it shows a comparatively poor reductase activity with 11-cis-retinal. The enzyme exhibits an 8-fold higher catalytic efficiency for 11-cis-retinol oxidation than for 11-cis-retinal reduction while it shows only about 1.5 times more activity for retinol oxidation with other isomers, and ADH1 catalytic efficiency is similar in the two directions with all retinoids tested. In fact, ADH4 is the only reported case among mammalian ADHs, and with any alcohol/aldehyde pair, in which a strong preference for the oxidation reaction is observed at physiological pH. Two factors can contribute to this specificity of ADH4: the structure of 11-cis-retinal and the distinct ADH4 substrate-binding pocket. 11-cis-retinal is unique among retinal isomers in that it shows a helical geometry in the region C11 to C13 which might, in part, be responsible for its fast photoisomerization, thus explaining its selection as the chromophore of the visual pigments [49,50]. This special conformation is not a limiting feature for the binding to ADH1, with a wide hydrophobic tunnel in the active site, but 11-cis-retinal cannot interact with ADH4 in a highly productive manner. Docking studies show that the 11-cis position is placed between the residues 57 and 141 of the pocket. In ADH4 these two residues are Met, defining a narrow region in comparison to ADH1, where these two residues are Leu. The substitution of Met141 by a Leu, results in a wider substrate-binding pocket, which allows proper binding of 11-cis-retinal, as kinetic and docking studies with ADH4 M141L have demonstrated. Thus, the region defined by position 141 is essential for conferring the specificity of 11-cis-retinol oxidation over 11-cis-retinal reduction in ADH4. This specificity provides additional support for the involvement of ADH4 in the physiological 11-cis-retinol oxidation in the eye.

In the RPE an isomerohydrolase catalyzes the formation of 11-cis-retinol from all-trans-retinyl ester [51]. An 11-cis-retinol dehydrogenase (RDH5) is then believed to be essential in the production of 11-cis-retinal, as mutations in its gene are associated with the eye disorder fundus albipunctatus [18,52,53], while knockout mice for this gene accumulate 11-cis-retinyl esters in the eye [54,55]. However, the knockout animals have normal vision indicating that other enzymes must exist in the RPE, capable of oxidizing 11-cis-retinol, and thus completing the visual cycle. We have localized ADH4 protein in the RPE by immunohistochemistry, consistent with the ADH4 activity previously found in this epithelium [25]. Thus, the presence in the RPE, the high activity with 11-cis-retinol, and the specificity for the oxidation direction of the reaction, suggest a participation of ADH4 in the rhodopsin regeneration pathway. With respect to the relative contribution of each enzyme, the microsomal RDH5 seems to play a major role because of its low Km (2.5–7.5 µm[56]) and its capacity of using 11-cis-retinol bound to cellular retinaldehyde-binding protein (CRALBP) [56,57]. Comparatively, the cytosolic ADH4 shows a higher Km (28 µm) and uses less efficiently the retinoid bound to CRALBP [57]. However, this could be in part compensated by a 40-fold higher kcat for ADH4 (200 min−1 vs. 5 min−1 for RDH5 [56]). Preliminary results on both, ADH4 knockout mice and ADH4/RDH5 double knockout mice, indicate mild effects on vision, suggesting the existence of several enzymes with a redundant function [58].

ADH4 may also be involved in other retinoid metabolism steps in RPE. Thus, acting as an all-trans-retinol dehydrogenase, it could provide the all-trans-retinal to the retinal G protein-coupled receptor opsin, an isomerase which can convert all-trans-retinal to the cis isomer by photoisomerization [59].

Retinoic acid is important in the function of neural retina. It has been related to eye development [21] and it has been proposed to act as a neuromodulator [23]. The localization of ADH4 in almost all parts of the neural retina, together with the presence of receptors and other proteins related to retinoic acid [21,60,61], indicate a complex retinoid metabolism and signaling in retina, with the probable participation of ADH4 in mammals. Moreover, as RDH5 is not present in neural retina [53], ADH4 could contribute to the 11-cis-retinol dehydrogenase activity responsible for the regeneration of cone photopigments, in addition to a specific microsomal enzyme [19], and finally, ADH4 may be involved in providing 11-cis-retinal to the photopigments of the photosensitive retinal ganglion cells that set the circadian clock [20].

In conclusion, human and rodent ADH1 and ADH4 show a wide specificity toward retinoids, using efficiently the all-trans and most of the cis isomers of retinol and retinal, including the 11-cis-retinoids involved in photosensitivity. Kinetic properties and its localization in many retinal cell layers support the involvement of ADH4 in the retinol oxidation reactions of retina as a cytosolic activity, complementary to the more specific and membrane-bound SDR enzymes.


This work was supported by grants from the Spanish Dirección General de Investigación (BMC2002-02659, BMC2003-09606 and SAF2001-3288), Generalitat de Catalunya (2001SGR 00198) and National Institutes of Health (EY13969).

Supplementary material

The following material is available from

Fig. S1. Absorbance spectra of 11-cis-retinal (A) and all-trans-retinal (B) dissolved in 0.1 m sodium phosphate, pH 7.0, 0.02% Tween-80, at 25 °C.