Molecular Basis for Differential Substrate Specificity in Class IV Alcohol Dehydrogenases

Mammalian class IV alcohol dehydrogenase enzymes are characteristic of epithelial tissues, exhibit moderate to high K m values for ethanol, and are very active in retinol oxidation. The human enzyme shows aK m value for ethanol which is 2 orders of magnitude lower than that of rat class IV. The uniquely significant difference in the substrate-binding pocket between the two enzymes appears to be at position 294, Val in the human enzyme and Ala in the rat enzyme. Moreover, a deletion at position 117 (Gly in class I) has been pointed out as probably responsible for class IV specificity toward retinoids. With the aim of establishing the role of these residues, we have studied the kinetics of the recombinant human and rat wild-type enzymes, the human G117ins and V294A mutants, and the rat A294V mutant toward aliphatic alcohols and retinoids. 9-cis-Retinol was the best retinoid substrate for both human and rat class IV, strongly supporting a role of class IV in the generation of 9-cis-retinoic acid. In contrast, 13-cisretinoids were not substrates. The G117ins mutant showed a decreased catalytic efficiency toward retinoids and toward three-carbon and longer primary aliphatic alcohols, a behavior that resembles that of the human class I enzyme, which has Gly117. TheK m values for ethanol dramatically changed in the 294 mutants, where the human V294A mutant showed a 280-fold increase, and the rat A294V mutant a 50-fold decrease, compared with those of the respective wild-type enzymes. This demonstrates that the Val/Ala exchange at position 294 is mostly responsible for the kinetic differences with ethanol between the human and rat class IV. In contrast, the kinetics toward retinoids was only slightly affected by the mutations at position 294, compatible with a more conserved function of mammalian class IV alcohol dehydrogenase in retinoid metabolism.

Many alcohol dehydrogenase (ADH, 1 EC 1.1.1.1) forms exist in mammals. These forms have been grouped into at least six classes according to their structural and enzymatic properties (1). The physiological function of each class can be estimated from its kinetic characteristics and organ distribution. Thus, classes I and II, with low or moderate K m for ethanol, are those mostly responsible for hepatic ethanol metabolism in humans. Class III exhibits activity with long chain primary alcohols, but its major function is the elimination of formaldehyde, acting as a glutathione-dependent formaldehyde dehydrogenase (2). Classes V and VI are little defined and only detected at the mRNA level.
Recently, much effort has been devoted to the functional and molecular characterization of class IV. Its localization in the stomach and a moderate K m for ethanol of the human enzyme make it suitable for a contribution to the first-pass metabolism of ethanol (3,4). Moreover, class IV is characteristic of epithelial tissues, like the mucosa of the upper gastrointestinal tract, the cornea, and the blood vessel endothelium (5)(6)(7). It is very active in retinol oxidation (8 -11), strongly suggesting a role of the class IV enzyme in the formation of retinoic acid. The latter compound is known to regulate morphogenesis and epithelial cell differentiation. A function of class IV in development was supported by the findings that retinoic acid synthesis in embryonic tissues correlates spatiotemporally with the expression of the class IV ADH gene (12). Moreover, class IV knock-out mice have a decreased production of retinoic acid from retinol (13) and have an increased risk of embryonic lethality during vitamin A starvation (14).
Although class I ADH is also active with retinoids, class IV, or , is the most active ADH form in humans (9). Structural features responsible for class IV specificity have been investigated using molecular models, in which docking simulations with retinoid isomers indicate a better binding of these molecules to the class IV active site in comparison to that of class I, suggesting that deletion at position 117 facilitates their binding (15)(16)(17). Recently, the x-ray structure of human class IV ADH has revealed an active site more suitable for oxidation of long chain aliphatic alcohols, such as retinols, than for short chain molecules, e.g. ethanol, in contrast with the opposite behavior of the class I ␤ 1 ␤ 1 -isozyme (18,19).
An intriguing feature of class IV enzymes is the strong variation of kinetic properties between enzymes of different mammalian species. Thus, the human enzyme exhibits a K m for * This work was supported by Grant BIO4-CT97-2123 from the Commission of the European Union, Grants PM96-0069 and PB98-0855 from the Spanish Dirección General de Enseñ anza Superior e Investigación Científica, Projects 13X-3532 and 03P-11312 from the Swedish Medical Research Council, and the Magnus Bergvall Foundation. 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  ethanol ϳ100-fold lower than that of the rat form, at pH 7.5 (20). However, the two enzymes show 88% identity in amino acid sequence, and exchanges at the substrate-binding site are scarce, the most interesting being at position 294. This position is a Val in most animal ADHs, but an Ala in the rat and mouse class IV enzymes, both with high K m values for ethanol (21,22). The exchange to a smaller and less hydrophobic residue in this area is compatible with low affinity for small substrate molecules, and we hypothesized that this substitution was responsible for the dramatic kinetic differences between the rat and human class IV enzymes (20,23).
In the present work, the influence of the Val/Ala exchange on the class IV kinetics with retinoids and ethanol has been assessed from the inversely complementary properties of the human class IV V294A and the rat A294V mutants. Moreover, we have investigated the role in substrate specificity of the deletion at position 117, a specific feature of class IV, by the construction of a human class IV mutant with Gly 117 (G117ins), as found in class I. The differences in the substratebinding pocket between the x-ray structure of human class IV (18), the models of rat enzyme, and the respective 294 mutants have been studied. Substrate docking with several retinol isomers also highlighted the importance of positions 117 and 294.

EXPERIMENTAL PROCEDURES
Screening of a Rat Lung cDNA Library-A rat class IV (20) ADHspecific probe (50 ng) was labeled with 50 Ci of [␣-32 P]dCTP (3000 Ci/mmol, Amersham Pharmacia Biotech) by the random-priming method (Prime-a-Gene labeling system, Promega). The labeled probe was used in a plaque hybridization procedure to screen a "5Ј-stretch" rat (adult Sprague-Dawley male) lung cDNA library in gt11 (CLON-TECH). One positive cDNA clone was isolated and DNA was prepared by a plate lysate method (24). The cDNA insert was excised by EcoRI digestion and analyzed by 1% agarose gel electrophoresis. The insert was subcloned into the EcoRI site of pBluescript II SK(ϩ) (Stratagene). DNA sequence determination was performed by the dideoxynucleotide chain termination method (25), using fluorescently labeled T7 and T3 primers with the Auto Read Sequencing Kit, in an ALF Sequencer (Amersham Pharmacia Biotech).
Isolation of Clones Containing the 5Ј-and 3Ј-Ends of the cDNA-The 5Ј-and 3Ј-ends of the rat class IV ADH cDNA were obtained by direct amplification from the gt11 cDNA library (26) combined with the use of two sets of nested gt11 and class IV-specific primers.
Expression of the Human and Rat Class IV ADH cDNA in Escherichia coli-The cDNA containing the region coding for -ADH was obtained by the PCR method from a gt11 human stomach cDNA library. The sense HUM1 (5Ј-GATCCAAGACAAAGACCATATGGGCA-CTGCTGG-3Ј) and antisense HUM2 (5Ј-GTGAGACAGATACATGAAT-TCAGATGAGGGAAC-3Ј) primers were designed according to the known cDNA sequence (20), including recognition sites (underlined) for the restriction endonucleases NdeI (for which ATG initiates the trans-lation) and EcoRI, respectively. The complete cDNA encoding rat class IV ADH was amplified by PCR from the 5Ј-stretch gt11 rat lung cDNA library with the sense RAT7 (5Ј-CATGAAGCTTCAGGAAAGCCCATA-TGGACAGTC-3Ј) and the antisense RAT8 (5Ј-CATGGGATCCTCAAA-ATGTCAGGACAGTCCGAA-3Ј) primers, containing recognition sequences (underlined) for the restriction endonucleases NdeI and BamHI, respectively. The PCR products were cloned into pBluescript II SK(ϩ) and sequenced. After NdeI/EcoRI or BamHI excision, the cDNAs were subcloned into the expression vector pET5a (Promega). The E. coli strain BL21(DE3)/plysS was transformed with the final constructs pET/ HUM12 and pET/RAT78, and grown in 5 liters of liquid Terrific Broth medium. When an OD at 595 nm of 0.65 was reached, a 2-h induction of protein expression was performed by the addition of 0.6 mM isopropyl-1-thio-␤-D-galactopyranoside to the culture.
The Human V294A, G117ins, and Rat A294V Mutants-To obtain the V294A mutant, four primers were used in three PCR reactions. The primers HUM1 and HUM3 (5Ј-CATCTTGGCTGATGATGGAGGAGCT-CCTACAACCAC-3Ј), containing the antisense mutation (underlined), were used in the PCR 1 construct. The primers HUM4 (5Ј-TTGTAGG-AGCTCCTCCATCAGCCAAGATGCTCA-3Ј), containing the sense mutation (underlined), and HUM2 were used in the PCR2 construct. The products from PCR1 and PCR2 were mixed with HUM1 and HUM2 primers in a third PCR reaction, obtaining the -ADH coding region, including the V294A mutation. The same procedure was followed for the G117ins mutant but using HUM5 (5Ј-GATCTCCACGACCGGTAC-CAATATCGCTCCTAA-3Ј) and HUM6 (5Ј-ATTAGGAGCGATATTGGT-ACCGGTCGTGGAGTA-3Ј) as the mutagenic primers. The mutated PCR products were cloned into pET5a. To construct the A294V mutant, the RAT9 (5Ј-TTGCCATATGAACTATGGGACCAGCGTGGTGGTCGG-GGTACCTCC-3Ј), containing the A294V substitution, and RAT8 primers were used in a PCR amplification. The mutated PCR product was cloned into pET/RAT78 by endonuclease excision and substitution for the corresponding wild-type DNA fragment. All mutated DNA fragments were checked by sequencing prior to expression.

Recombinant Wild-type Human Class IV ADH and V294A and G117ins Mutants-Isopropyl-1-thio-␤-D-galactopyranoside-induced
cells from a 5-liter culture (18 g wet weight) were harvested by centrifugation at 5000 ϫ g for 20 min. The pellets were frozen at Ϫ80°C to facilitate cell lysis and resuspended in 50 mM Tris-HCl, 1 mM EDTA, pH 7.0, 10% sucrose, 0.05% sodium azide, and 2 mM ␤-mercaptoethanol. Glass beads (diameter 0.5 m) were added, and the mixture was stirred at 4°C for 20 min and incubated with DNase (18 mg/g wet weight) at room temperature for 30 min to reduce sample viscosity. The homogenate was centrifuged at 11,000 ϫ g for 20 min, the supernatant was treated with 0.6% protamine sulfate and incubated on ice for 30 min, and it was centrifuged at 20,000 ϫ g for 15 min. The supernatant resulting from the centrifugation was dialyzed against 20 mM Tris-HCl, 0.5 mM DTT, pH 8.9, and applied to a DEAE-Sepharose (Amersham Pharmacia Biotech) column (2.5 ϫ 40 cm) equilibrated with the same buffer (3). The enzyme was eluted with a 0 -150 mM NaCl linear gradient (1000 ml), and the active fractions were pooled, concentrated, dialyzed against the initial buffer, and applied to a AMP-Sepharose (Amersham Pharmacia Biotech) column (1 ϫ 15 cm). The enzyme was eluted with a linear gradient (500 ml) of 20 -100 mM Tris-HCl, 0.5 mM DTT, pH 8.2, and the active fractions were pooled, concentrated, and stored at Ϫ80°C. Starch gel electrophoresis, followed by activity staining with ethanol or 2-buten-1-ol as a substrate (28), and electrophoresis in SDS-polyacrylamide gel with subsequent silver staining were used to assess the degree of purity after each step of purification. For the V294A mutant enzyme the same purification procedure was used. For the G117ins mutant, an additional gradient from 100 to 400 mM Tris-HCl, 0.5 mM DTT, pH 8.2 (500 ml) was necessary to elute the enzyme from the AMP-Sepharose column.
Recombinant Wild-type Rat Class IV ADH and the A294V Mutant-Only the differences with respect to the purification procedure for the human enzyme are detailed. The complete cell lysis was obtained by shaking the cell resuspension with glass beads (diameter 0.5 m) in a bead beater (Biospec Products). The buffer used during the purification was 20 mM Tris-HCl, 0.5 mM DTT, pH 7.9. The dialyzed supernatant was applied to a DEAE-Sepharose (Amersham Pharmacia Biotech) column (2.5 ϫ 40 cm) equilibrated with the same buffer, and the enzyme was eluted with a 0 -150 mM NaCl linear gradient (500 ml). The active fractions were pooled, concentrated, and applied to a Blue-Sepharose (Amersham Pharmacia Biotech) column (1 ϫ 15 cm). The enzyme was eluted with a linear gradient (400 ml) of 20 -300 mM Tris-HCl, 0.5 mM DTT, pH 7.9. The active fractions were pooled, concentrated, dialyzed against the initial buffer, and applied to a Q-column (Waters). Elution of the enzyme was performed with a 0 -250 mM sodium acetate linear gradient.
Enzyme Assays-Alcohol dehydrogenase activity with alcohols was determined by monitoring the formation of NADH at 25°C in a Varian Cary 219 spectrophotometer by measurements at 340 nm. Alcohol oxidation was measured in 0.1 M sodium phosphate, pH 7.5, or in 0.1 M glycine-NaOH, pH 10. Kinetic parameters were obtained from activity measurements, with substrate concentrations that ranged from 0.1ϫ K m to 10ϫ K m , except for ethanol with the V294A mutant and rat wild-type enzyme, where saturation with this substrate could not be reached. Each individual rate measurement was run in duplicate. Three determinations were performed for each kinetic constant. The dissociation constant for NAD ϩ (K ia ), and the K m values for NAD ϩ and ethanol were calculated by fitting the kinetic data to a sequential Bi Bi mechanism (29). Kinetics with retinoids (Sigma) were performed according to published procedures (8,10,27), using the following molar absorbances of ⑀ 328 ϭ 39500 M Ϫ1 cm Ϫ1 for all-trans-retinol, ⑀ 326 ϭ 39,733 M Ϫ1 cm Ϫ1 for 9-cis-retinol, ⑀ 330 ϭ 43751 M Ϫ1 cm Ϫ1 for 13-cis-retinol, ⑀ 400 ϭ 29,500 M Ϫ1 cm Ϫ1 for all-trans-retinal, and ⑀ 367 ϭ 26,700 M Ϫ1 cm Ϫ1 for 9-cis-retinal. 9-cis-Retinol was synthesized by reduction of 9-cis-retinal with sodium borohydride (30), and the purity was checked by high pressure liquid chromatography. Tween 80 (0.02%) was used to solubilize retinoids according to established protocols (8,10,28). The kinetic constants were obtained by means of the ENZFITTER (Elsevier Biosoft) and SEQUEN or COMP (29) programs. The values were expressed as the mean Ϯ S.D. Statistical significance was determined using the unpaired Student's t test. Differences were considered statistically significant when the p value was Ͻ0.01.
Rat Class IV Alcohol Dehydrogenase Modeling and Substrate Docking-A three-dimensional model of rat class IV ADH was constructed by adopting its amino acid sequence into the known fold of the human class IV ADH dimer (PDB code, 1AGN; Ref. 18) using the ICM program (version 2.7, Molsoft LLC, 1997), as described (16). The atomic positions of the coenzyme were tethered to ones corresponding to the crystallographic coordinates, to avoid undesired displacements. The quality of the model was assessed with the PROCHECK program (31). To study interactions between the enzyme and different substrates, a nonrigid docking was used based upon a Monte Carlo procedure (32) with 500,000 iterative cycles, allowing free movement of the substrate, its rotable bonds, and the angles of the residues inside a 5 Å radius from the docked substrate. Additional distance restraints were imposed: 2.0 -2.4 Å between the alcohol oxygen of retinol and the catalytic zinc ion, 2.0 -4.0 Å between C15 of retinol and C4 of NAD, and 2.0 -4.0 Å between the alcohol oxygen and the oxygen of the side chain hydroxyl group of Thr 48 . Energy values were calculated with the ICM program, which includes the REBEL method for estimation of electrostatic free energy and the constant surface tension method with 20 cal/Å 2 for calculation of hydrophobic energy. As a binding score, a docking energy value was calculated by subtracting the energy values of the isolated molecules from the energy value of the enzyme-retinoid complex.

Cloning and Sequence Analysis of the cDNA Coding for Rat
Class IV ADH-A composite full-length sequence for rat class IV cDNA encompassed 2052 bp (Fig. 1) and was obtained via screening of a rat lung cDNA library and using reverse transcription-PCR of rat liver RNA. The sequences isolated from the lung cDNA library included nucleotides 1-1685, whereas an overlapping PCR fragment from liver RNA extended the 3Ј-sequence up to position 2052. A part of the 3Ј-untranslated region including nucleotides 1182-2017 was also amplified from rat genomic DNA and sequenced, matching the sequence isolated from liver RNA and showing that this was not a cloning artifact. Both the 5Ј-and 3Ј-noncoding regions showed a notable sequence identity with those corresponding to the human (20) and mouse (33)  The amino acid sequence deduced is identical to that obtained from sequence analysis of the protein (23), except for six residue differences. Notably, the first four N-terminal residues (Met-Asp-Thr-Ala) are different from those previously reported (Ser-Asn-Arg-Val), which had been considered tentative because of the fact that the N-terminal residue was blocked by an acetyl group, as in all other mammalian ADHs. Indeed, the present rat N-terminal sequence resembles more closely the human (20) and mouse (33) class IV ADH sequences except for the presence of an Asp (instead of Gly) following the initiator Met. Asp is quite an unusual residue at this position (mostly Ser, Ala, or Gly) in other ADHs. Human and mouse class IV have a Gly, which implies loosing Met and the subsequent acetylation of the Gly residue. Interestingly, when Met is retained and acetylated, Asp is the predominant penultimate residue found in eukaryotic proteins (36). Because the rat protein analysis had revealed that the N terminus contained one additional residue compared with the human class IV enzyme (20,23), it is tempting to speculate that in rat class IV the initiator Met may be retained and acetylated. The other two observed differences, a Glu for Gly at position 109 and a Val for Ile at position 208, can be explained by single-nucleotide exchanges and may reflect a polymorphism without functional consequences.
Expression in E. coli and Site-directed Mutagenesis-To produce recombinant human and rat class IV ADHs, their corresponding entire coding sequences were cloned into appropriate expression vectors. No extra amino acid residues were added to the N termini, but in the products prokaryotic formyl-Met is likely to be present, rather than the acetyl-Gly of the human or acetyl-Met of the rat enzymes. The recombinant wild-type and mutant class IV subunits were expressed to similar levels, as assessed from the amount of immunoreactive protein seen in cell homogenates by Western blot analysis. Approximately 1 mg of either enzyme was expressed/10 g of wet weight cells. Both of the two recombinant proteins were soluble, enzymatically active, and were purified to homogeneity following a modification of the methods previously used for the enzymes isolated from stomach tissue. Purity was assessed by SDSpolyacrylamide gel electrophoresis with silver staining, and a unique band in the pure fractions was observed.
PCR-based site-directed mutagenesis was used to construct several mutant class IV enzymes: human V294A, rat A294V, and human G117ins. The mutants were purified to homogeneity and their kinetic properties were compared with those of the wild-type enzymes (Tables I-III).
Kinetic Properties of the Mutant and Wild-type Enzymes with Aliphatic Alcohols-Bisubstrate kinetics were performed by covariation of NAD ϩ and ethanol concentrations. The experimental data could be fitted to the equation for the sequential ordered Bi Bi mechanism (29), supporting the conclusion that the wild-type and mutant enzymes follow this mechanism. The kinetic constants K a (K m for NAD ϩ ), K ia (dissociation constant for NAD ϩ ), K b (K m for ethanol), and k cat could be determined (Table I). In addition, K iq (dissociation constant for NADH) and K i for 4Ϫmethylpyrazole were calculated by fitting the data to the equation for competitive inhibition with NAD ϩ and ethanol, respectively.
The kinetic constants for the G117ins mutant were similar to those of the human wild-type enzyme. No significant change was observed between the K ia values, although the K iq and k cat values were lower for the mutant enzyme, and there was a 2-3-fold decrease in the K m value for ethanol. Consistent with this finding, the K i for the competitive inhibitor 4-methylpyrazole was also decreased.
The kinetic constants with ethanol of the human V294A and rat A294V mutants compared with those of the corresponding wild-type enzymes were quite altered, although less so for the rat enzymes (Table I). The V294A mutant exhibited a K m value for ethanol, which was 280-fold that of the wild-type human class IV enzyme. Accordingly, the inhibition constant for 4-methylpyrazole increased significantly in this mutant. Although the K m value for NAD ϩ increased notably in the V294A mutant, the K ia and K iq values did not vary substantially. Thus, the marked increase in the k cat value with respect to the wild-type human class IV enzyme cannot be correlated with dissociation of the coenzyme. Because dissociation of coenzyme is likely to be the limiting step (k cat remains constant for different substrates, Table II), variation of k off cannot be excluded. A reciprocal tendency was observed in the A294V mu-  tant, in which the K m value for ethanol was 50-fold lower than that with the wild-type rat class IV enzyme. No change in the NAD ϩ constants and a 2-fold decrease in the k cat value were also observed for the A294V mutant. Overall, the kinetic properties of each pair of enzymes containing the same residue at position 294 were very much alike. These results indicate that the single Val/Ala exchange at position 294 fully explains the differences in the K m values for ethanol found in the human and rat wild-type class IV enzymes. With human wild-type and mutant enzymes, the K m values for primary aliphatic alcohols decreased with increasing chain length (Table II). Because the k cat values remained essentially constant, the catalytic efficiency, k cat /K m , increased with the substrate chain length. Notably, for both the wild-type and V294A mutant, the highest increment in catalytic efficiency was observed between propanol and butanol, which resulted from a dramatic decrease in the K m value for butanol. The insertion of Gly at position 117 produced a slight decrease or no change in the K m values for ethanol and propanol, with respect to the wild-type enzyme. In contrast, the K m values increased about 3-5-fold for substrates containing four or more carbons. Overall, for the G117ins mutant, the catalytic efficiency increased more steadily with increases in the substrate hydrophobicity than for the wild-type enzyme. The V294A mutation increased not only the K m value for ethanol but also those for primary aliphatic alcohols from 3 to 8 carbons.
Kinetic Properties of the Mutant and Wild-type Enzymes with Retinoids-The kinetic constants for all-trans, 9-cis, and 13-cis retinoids of the human and rat wild type and the three mutant enzymes were determined (Table III). The human and rat wild-type enzymes showed high activity with both all-trans and 9-cis isomers as substrates, whereas no activity was detected with the 13-cis isomers in both the oxidative and reductive reactions, even though a high enzyme concentration was used in each assay. In general, the wild-type human class IV enzyme was more efficient toward retinoids than the rat enzyme. This is because of the higher k cat values with the human enzyme. The k cat and k cat /K m values for the 9-cis isomer were higher than those for the all-trans compound, and oxidation of 9-cisretinol was found to be the most efficient reaction.
The kinetic constants for retinoids showed some differences between the mutants and the wild-type enzymes. With retinoids, the k cat values of the G117ins mutant were 3-8-fold lower than for the human wild-type enzyme. Interestingly, the A294V mutant showed k cat and k cat /K m values moderately higher (2-5-fold) than those for the rat wild-type enzyme and similar to those exhibited by the human class IV. Only small changes were observed in the K m values of the V294A and the A294V mutants with respect to those of their corresponding wild-type enzymes, suggesting that the exchange at position 294 did not disturb retinoid binding.
Structural Model of Rat Class IV ADH and Substrate Docking-The high sequence identity (88%) between human and rat class IV ADH suggested that the protein fold was conserved between the two forms. Significantly, the root mean square deviation between the model and the crystal structure (18) was 1.6 Å, and this value decreased to 0.3 Å when only the coenzyme-binding domain was considered. Ramachandran plots of the main chain conformational angles showed that 87% of the residues were in the most favored regions, and 13% were in allowed regions. Like in the human class IV crystal, no residues were positioned in disallowed regions. Special attention was paid to the R47G and A102P exchanges, two nonconservative substitutions involving ␣-helices. However, all the secondary structural elements were conserved in the model. No severe clashes were observed between side chains, which is reflected in the total energy value (Ϫ3700 kcal/mol).
When the x-ray structure of the human class IV enzyme (18) and the model of the rat class IV were compared, atomic dis-  tances between the oppositely positioned residues 116 and 294 in the substrate-binding pocket could be measured. Whereas in the human structure, Ile 116 and Val 294 were separated by only 4.8 Å, in the rat enzyme the atomic distance between Leu 116 and Ala 294 was 6.8 Å (Fig. 2). In the human enzyme, this distance is close to the threshold value for water accessibility (2.8 Å), when the contribution of van der Waals radii is consid-ered. Consistent with the kinetic results (Table I), each of the 294 mutations reverts the narrowness of the bottleneck; in the V294A mutant the atomic distance was 5.9 Å, and in the A294V it was 5.5 Å.
Docking of all-trans-, 9-cis-and 13-cis-retinol isomers to the x-ray structure of human class IV and to the model of the rat class IV ADH revealed that the atomic distance between the catalytic zinc and the oxygen of the hydroxyl group of the all-trans-retinol and 9-cis-retinol isomers was 2.4 Å in both structures. However, the distance was 3.1-3.3 Å for 13-cisretinol, a value excessively large for catalysis to be productive. Moreover, 13-cis-retinol bound in an orientation different from that of the other two isomers, with its ␤-ionone ring much closer to the 114 -120 loop. The docking energy values for the three isomers studied ranged from Ϫ14 to Ϫ27 kcal/mol (Fig. 3). DISCUSSION Distinct class IV structural features are observed in the middle region of the substrate-binding pocket, where all members of the class exhibit a deletion at position 117 (Gly in class I), and the rodent enzyme has Ala 294 instead of Val. Both the deletion at 117 (15,16,18,19) and the exchange at 294 (20,23) have been suggested to explain the distinct kinetic properties of the class IV enzymes. Interestingly, these residues are located within the variable segments V2 I and V3 I of ADH structures (37).
Deletion of residue 117 in class IV ADH shortens the loop including residues 114 -120, and thus it has been reported to widen the entrance to the substrate-binding pocket permitting the efficient binding of retinol (15,16,18). Here we have shown that by reverting the deletion at position 117 (G117ins mutation), kinetics with primary aliphatic alcohols (Table II) and with retinoids (Table III) are affected. In terms of the kinetic properties toward primary aliphatic alcohols, the results obtained with wild-type class IV versus those with the G117ins mutant (Table II) are compatible with previous studies on class I mutations: human ␤ 1 ␤ 1 L116A (38) and horse EE D115del (39), which likewise diminish the size of loop 114 -120 (either by substitution by a smaller side chain or by deletion). In all cases, when the size of the loop decreased, the K m values for short-chain aliphatic alcohols (ethanol and propanol) increased, whereas at the same time the enzymes became more specific for alcohols with four or more carbons.
Residues 116 and 294 define a narrow bottleneck in the middle region of the substrate-binding pocket that restricts the access of bulky substrates to the bottom of the active site ( Fig.  2; 18 , 40). The presence of Met 57 , Met 141 , and Phe 309 also contributes to narrowing the middle region of the substrate pocket compared with the relationships in class I ADH (18,19). It is tempting to speculate that the deletion at position 117 in class IV and the V294A mutation could create additional space for water molecules to enter the active site. This effect increases the number of nonproductive conformations and it would impair the correct binding of short-chain alcohols, such as ethanol, resulting in increased K m values. In fact, it is possible to establish a good correlation between the exchange Val/Ala in the middle region of the substrate-binding pocket, the atomic distance between residues 116 and 294, and the K m values for ethanol (Table I). The V294A exchange also increased the K m values for longer aliphatic alcohols, although to a lesser extent, probably because of their ability to establish interactions with several residues in the active site pocket.
This work provides an in depth kinetic characterization of rat class IV ADH regarding retinoid metabolism. No data have ever been reported for 9-cis-and 13-cis-retinol oxidation with this enzyme. The present kinetic values toward the all-trans compounds do not differ significantly from those previously reported (8), with the exception of a higher k cat value for alltrans-retinol, which may be due in part to a somewhat higher specific activity obtained in the preparation of purified recombinant rat class IV.
Among the isomers assayed, 9-cis-retinol was the best substrate in terms of catalytic efficiency. The pathway that leads to the bioactive compound, 9-cis-retinoic acid, has not been clearly established yet. As suggested in previous works (10,41), isomerization at the level of 9-cis-retinol and subsequent oxidation to 9-cis-retinal is one possibility, supported now by favorable kinetics of both human and rat class IV ADHs. No activity was found with 13-cis-retinol, which is a competitive inhibitor of all-trans-retinol and 9-cis-retinol oxidation in human class IV (10). In good agreement with its inhibitory behavior, our docking simulations show that 13-cis-retinol could bind to human and rat class IV. However, the rigidity provided by the double bond in cis conformation at C13 appears to prevent the hydroxyl group from reaching a catalytically productive distance or proper orientation with respect to the zinc atom. In contrast, all-trans-and 9-cis-retinol docking simulations showed a binding effective for catalysis, in good agreement with other docking studies (15)(16)(17)(18).
The G117ins mutant showed k cat values for retinoids that were below those of the wild-type human class IV enzyme but far above those of the class I enzymes (9,11). This implies that deletion of residue 117 contributes to the high specificity of class IV toward retinoids but also that this is not the only amino acid exchange involved in the creation of such a specificity. It is conceivable that other substitutions, such as the smaller Leu 110 (in class IV) for Tyr (in class I) could contribute to differences in substrate specificity. Interestingly, the two 294 mutations altered minimally the kinetic constants toward retinoids, despite the fact that they profoundly altered ethanol oxidation.
From the present results, it can be concluded that human and rat class IV ADH are enzymes similarly efficient toward retinoids, although they differ dramatically with respect to ethanol oxidation. This difference is extremely relevant when drawing conclusions from alcohol toxicity studies using rodents as model animals. Ala 294 in rodent class IV, in substitution of a Val 294 in human class IV (a residue highly conserved in vertebrate ADH), is clearly the change responsible for the physiological inefficiency of the rat enzyme toward ethanol (K m ϭ 2.4 M) in contrast to the moderate activity of the human class IV enzyme (K m ϭ 42 mM). Evolution has allowed the presence of a residue in the rodent class IV that makes the enzyme unsuitable for ethanol elimination but still permitting retinoid metabolism. Together with the enzyme localization in epithelial tissues but not in liver (5), this is a strong evidence of a role of class IV in a specific metabolic step such as the conversion of retinol to retinal in the crucial pathway of retinoic acid synthesis, rather than in an unspecific alcohol detoxification.