cDNA Sequence and Catalytic Properties of a Chick Embryo Alcohol Dehydrogenase That Oxidizes Retinol and 3β,5α-Hydroxysteroids

This study was undertaken to identify the cytosolic 40-kDa zinc-containing alcohol dehydrogenases that oxidize all-trans-retinol and steroid alcohols in fetal tissues. Degenerate oligonucleotide primers were used to amplify by polymerase chain reaction 500-base pair fragments of alcohol dehydrogenase cDNAs from chick embryo limb buds and heart. cDNA fragments that encode an unknown putative alcohol dehydrogenase as well as the class III alcohol dehydrogenase were identified. The new cDNA hybridized with two messages of ∼2 and 3 kilobase pairs in the adult chicken liver but not in the adult heart, muscle, testis, or brain. The corresponding complete cDNA clones with a total length of 1390 base pairs were isolated from a chicken liver λgt11 cDNA library. The open reading frame encoded a 375-amino acid polypeptide that exhibited 67 and 68% sequence identity with chicken class I and III alcohol dehydrogenases, respectively, and had lower identity with mammalian class II (55-58%) and IV (62%) isozymes. Expression of the new cDNA in Escherichia coli yielded an active alcohol dehydrogenase (ADH-F) with subunit molecular mass of ∼40 kDa. The specific activity of the recombinant enzyme, calculated from active site titration of NADH binding, was 3.4 min−1 for ethanol at pH 7.4 and 25°C. ADH-F was stereospecific for the 3β,5α- versus 3β,5β-hydroxysteroids. The Km value for ethanol at pH 7.4 was 17 mM compared with 56 μM for all-trans-retinol and 31 μM for epiandrosterone. Antiserum against ADH-F recognized corresponding protein in the chicken liver homogenate. We suggest that ADH-F represents a new class of alcohol dehydrogenase, class VII, based on its primary structure and catalytic properties.

Cytosolic zinc-containing alcohol dehydrogenases (ADH) 1 with 40-kDa subunits are capable of oxidizing a variety of primary, secondary, and aliphatic alcohols and a limited number of cyclic alcohols (1). Six classes of dimeric ADH isozymes have been identified in mammals (1). Except for class I, all other classes of ADH have high K m values for ethanol and oxidize medium-chain and long-chain alcohols most effectively.
Potential physiological substrates for ADH isozymes include retinoid and steroid alcohols. Human class IV is the most efficient retinol dehydrogenase, followed by class II and the class I ␣␣ ADH (2). Class IV and II ADH are not active with steroid alcohols, whereas class I isozymes oxidize both retinoid and steroid substrates with relatively low catalytic efficiency. Class I isozymes 2 exhibit stereospecificity toward alcohol substrates. For example, horse SS and human ␥␥ ADH oxidize 3␤-hydroxysteroids but not 3␣-hydroxysteroids (3). ADH isozymes vary in their tissue distribution; class IV ADH is expressed in the epithelial tissues of mammals and is present in the human stomach mucosa and esophagus (1), whereas class II is found in fetal and adult liver. Class I isozymes ␤ 1 ␤ 1 , ␤ 2 ␤ 2 , ␤ 3 ␤ 3 , ␥ 1 ␥ 1 , ␥ 2 ␥ 2 , ␣␣, and their heterodimers, as well as class II , are the predominant forms responsible for ethanol oxidation in the human adult liver. Class I isozymes are also expressed to a lesser extent in certain adult and fetal tissues, such as kidney, skin, gastrointestinal tract, and lung.
The physiological significance of the cytosolic ADHs for steroid and retinoid metabolism is not clear. The retinoid and steroid hormones play a major role in fetal development and are detected in the embryonal tissues during the early developmental stages. Since ADH isozymes appear at different stages of embryogenesis, it is important to determine which isozymes are present in the embryo during the stages when retinoid and steroid hormones are synthesized. The chick embryo is used as a model to study the effects of various hormones on gene expression during development. In this study, we analyzed the mRNA isolated from the chick fetal heart and limb buds at stage 21 for the presence of messages encoding cytosolic ADH.

EXPERIMENTAL PROCEDURES
PCR Amplification of ADH Isozymes-Degenerate oligonucleotide primers were synthesized based on the peptide sequences E(D/E)(I/ V)EVAP and FGLGGVG, which are conserved in all animal alcohol dehydrogenases (4). The first region corresponds to amino acids 24 -30 (sense primer) and the second region to amino acids 198 -204 (antisense primer) of the human class I ␤ 1 ADH. Four oligonucleotides were synthesized for the sense orientation: 2), TT(T/C)GGITT(A/G)GGIGGIGTIGG. Inosines were incorporated in * This work was supported by National Institute on Alcohol Abuse and Alcoholism Grants K08 AA00221-01 (to N. Y. K.), R37-AA02342 (to T. K. L.), and R37-AA07117 (to W. F. B.). 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM 1 The abbreviations used are: ADH, alcohol dehydrogenase; ADH-F, the new ADH in fetal chick described in this study; bp, base pairs; GST, glutathione S-transferase; PCR, polymerase chain reaction. all positions that required a degeneracy of 4. Limb buds and hearts were dissected from 30 chick stage-21 embryos (3 days old). Total RNA was isolated from the pooled limb buds and from the pooled hearts (RNAzol, Cinna/Biotecx Laboratories, Inc., Houston, TX). The total RNA from each pool was reverse transcribed and used for PCR amplification of ADH isozymes. PCR was performed with various combinations of 4 sense and 2 antisense primers for 30 cycles with annealing at 50°C (1 min), extension at 72°C (2 min), and denaturing at 94°C (1 min). Several combinations of primers produced a ϳ500-bp product. Each ϳ500-bp band was isolated, reamplified, and subcloned into M13mp19RF (Life Technologies, Inc.) vector for sequencing (U. S. Biochemical Corp.).
Northern Blot Analysis and Screening of the cDNA Library-The liver, kidney, lung, heart, brain, skeletal muscle, and bladder tissues were dissected from a 7-week-old chicken and frozen immediately in liquid nitrogen. Total RNA was isolated from each tissue with RNAzol according to the manufacturer's protocol. Twenty micrograms of each RNA preparation were loaded onto a formaldehyde-agarose gel and separated by electrophoresis. After transfer to the Nytran filter (Shleicher & Schuell), the separated mRNAs were hybridized with the [␣-32 P]dATP-labeled ϳ500-bp PCR product in 50% formamide, 5 ϫ Denhardt's solution, 5 ϫ saline/sodium/phosphate/EDTA, 0.1 mg/ml salmon sperm DNA, and 0.1% SDS at 42°C overnight. After hybridization, the filter was washed several times in 2 ϫ SSC, 0.1% SDS at room temperature, and the final wash was performed in 0.1 ϫ SSC, 0.1% SDS at 65°C for 30 min.
A chicken liver gt11 cDNA library (Clontech) was screened with the radiolabeled ϳ500-bp PCR product. The hybridization and washing conditions were the same as those described for the Northern blot analysis. Positive plaques were purified through three more rounds of screening. The purified phage was cleaved with EcoRI restriction endonuclease, and the cDNA insert was isolated and subcloned into M13mp19RF digested with EcoRI. Sense and antisense single-stranded M13 DNA were prepared, and each was sequenced at least three times.
Expression in Escherichia coli-The coding region of the new ADH cDNA was amplified by PCR with the following primers: CTCAGGATC-CATGGCCACTTCTGGAAAAGTT for the sense strand and TGGGAAT-TCTCAGAAGAGCATCACGGTGC for the antisense strand. The sense and antisense primers contained recognition sequences for the restriction endonucleases BamHI and EcoRI (underlined in the nucleotide sequence above), respectively. The amplified coding region of the new cDNA was subcloned into the expression vector pGEX-2T (Pharmacia Biotech Inc.). The final construct encoded a 375-amino acid polypeptide fused with glutathione S-transferase (GST). The expression of the fusion protein in the E. coli TG-1 cells was performed as described for human stomach -ADH (5). Cells were harvested by centrifugation and suspended in phosphate-buffered saline with Tween 80 (138 mM NaCl, 2.7 mM KCl, 1.2 mM KH 2 PO 4 , 8.1 mM Na 2 HPO 4 , pH 7.5, and 0.05% (w/v) Tween 80 (PBST)) containing 0.1% ␤-mercaptoethanol, 10 M ZnSO 4 , and the protease inhibitors phenylmethylsulfonyl fluoride (50 g/ml) and benzamidine (5 mM). The cells were homogenized using a French press, and the insoluble fraction was separated by centrifugation. The fusion protein was purified by glutathione-agarose affinity chromatography. The alcohol dehydrogenase activity of the recombinant protein was determined in a standard assay containing 4.7 mM cinnamyl alcohol, 2.5 mM NAD ϩ in 0.1 M sodium phosphate, pH 7.4, at 25°C. The GST domain was separated from chick ADH by cleavage with human thrombin (Sigma). The efficiency of the cleavage was monitored by the appearance of separate 40-(ADH) and 26-kDa (GST) protein bands in SDS-polyacrylamide gel electrophoresis. ADH was purified from GST by chromatography over S Sepharose and eluted with a NaCl gradient in 10 mM sodium phosphate buffer, pH 6.5, 10% glycerol, 2 mM dithiothreitol. Chick ADH eluted at 100 mM NaCl. GST did not bind to the resin under these conditions. Glycerol and dithiothreitol were found to stabilize enzyme activity. Therefore, purified ADH was stored in 10 mM sodium phosphate, pH 7.4, 50% glycerol, and 2 mM dithiothreitol at Ϫ20°C. The concentration of glycerol was reduced to 10% before each experiment. Glycerol never exceeded 0.5% in the assay mixture, and this concentration did not alter ADH-F activity measurements. The concentration of ADH active sites was determined by observing fluorescence (excitation wavelength at 328 nm and emission at 425 nm) while titrating enzyme (1-2 mg/ml) with NADH in the presence of 99 mM isobutyramide in 10 mM sodium phosphate at pH 7.4. The concentration of NADH binding sites was evaluated from the intersection point of the linear regression of the fluorescence titration above and below NADH saturation (6). The specific activity of the chick ADH-F was calculated based on the concentration of NADH binding sites. Total protein concentration was determined by a dye-binding assay (Bio-Rad) using bovine serum albumin as a standard.
The kinetic constants for retinol were determined by monitoring the production of all-trans-retinal at 400 nm (⑀ ϭ 29.5 mM Ϫ1 cm Ϫ1 ) (7). The retinol stock solution was prepared in acetone, and aqueous retinol solutions were prepared by dissolving the calculated amount of retinol stock solution in 0.1 M sodium phosphate, pH 7.5, and 0.02% Tween 80. The addition of 0.02% Tween 80 did not inhibit the enzyme activity. The concentration of retinol in aqueous solution was determined by measuring the absorbance at 328 nm (⑀ ϭ 39.5 mM Ϫ1 cm Ϫ1 ), and solutions were used immediately.
All kinetic studies were performed in 0.1 M sodium phosphate, pH 7.4, at 25°C with 2.4 mM NAD ϩ or 0.2 mM NADH. The kinetic constants for alcohols other than retinol were obtained by monitoring the production of NADH at 340 nm (⑀ ϭ 6.22 mM Ϫ1 cm Ϫ1 ). Reaction mixtures with steroid substrates contained 0.02% Tween 80. Steroid stock solutions were prepared in methanol, and concentration of methanol in the assay mixtures was kept constant at 0.3 M. Chick ADH-F was neither active toward nor inhibited by methanol up to 3 M at pH 7.5. The V max and K m values for alcohol substrates (at 2.4 mM NAD ϩ ) were calculated from a fit of the kinetic data to the Michaelis-Menten equation, where A is the concentration of the varied substrate. The k cat (min Ϫ1 ) was obtained by dividing V max by the concentration of active sites assuming a subunit M r of 40,000. The apparent K m values for NAD ϩ and NADH were determined with 1 mM cinnamyl alcohol and 100 M cinnamyl aldehyde, respectively. The inhibition constant for 4-methylpyrazole was determined with butanol as a substrate by varying both butanol (68 -200 M) and 4-methylpyrazole (75-350 M) concentrations. The K i of 4-methylpyrazole was calculated from a fit of the kinetic data to the equation for competitive inhibition, where B and I are butanol and 4-methylpyrazole concentrations, respectively (8). The K i value for NADH was determined by varying NAD ϩ (15-60 M) at 1 mM cinnamyl alcohol, using 0 -10 M NADH as the inhibitor.
The rabbit antiserum was raised against recombinant ADH-F. A 1:5,000 dilution of this antiserum detected 10 ng of purified ADH-F. Frozen chicken liver was homogenized in 10 mM Tris-HCl, pH 7.4, plus 5 mM benzamidine and 1 mM dithiothreitol. The homogenate was centrifuged at 10,000 ϫ g for 15 min, and the supernatant was concentrated twice. Glycerol was added to 50% concentration, and the liver extract was stored at Ϫ20°C. The proteins in the chicken liver homogenate were separated by isoelectic focusing using 3-10 pH gradient isoelectic focusing agarose plates (FMC Bioproducts, Inc.). After focusing, the separated proteins were transferred to nitrocellulose membrane, blocked with 3% bovine serum albumin in PBST, and incubated with a 1:5,000 dilution of antiserum. The binding of anti-ADH-F antibodies was visualized with 125 I-protein A.
The amino acid substitutions occurring in chick ADH-F were modelbuilt into the human ␤ 1 -structure using the molecular graphics program QUANTA (Molecular Simulations, Inc.). Following substitution of all amino acid side chains in the dimer, the model structure was subjected to 100 cycles of energy minimization using X-PLOR 3.1 with the x-ray energy term omitted (9). The position for the epiandrosterone molecule in the human ␤ 1 -structure was found by manually adjusting its position to minimize close contacts between the enzyme active site and the substrate molecule.
Sequences of human class I ␤, ␣, and ␥ ADH; class I ADHs from the alligator, cod, frog, horse E and S, mouse, ostrich, quail, rabbit, and rat; class II ADHs from human and rat; class III ADHs from human, horse, mouse, and rat; class IV ADHs from human and mouse; human ADH6; and class VI from deer mouse were aligned with ADH-F by a progressive alignment method according to Feng and Doolittle (10). Sequences of ADHs were obtained from the GenBank TM .

RESULTS
The pool of ϳ500-bp PCR products obtained with ADH-specific primers from the limb buds and heart mRNA of chick embryos at stage 21 was subcloned in M13 vector, and 48 individual clones were sequenced. Two of the clones from heart mRNA were found to have a novel sequence with a high resemblance to ADH sequences, and 6 of the clones from limb bud mRNA encoded a fragment that exhibited 87% protein sequence identity with human -ADH (11,12). The rest of the clones contained cDNA sequences that were not related to ADH. Since human class III ADH is not active with retinol, we did not pursue the cloning and characterization of this -ADHlike chick isozyme further.
The deduced protein sequence of the two novel identical PCR clones had a high resemblance to ADH sequences but was different from that of the -ADH-like chick ADH and the chick class I ADH (13). A Northern blot analysis of adult chicken tissues demonstrated that this partial PCR product from embryonal heart hybridized with two messages of approximately 2 and 3 kilobase pairs in adult chicken liver (Fig. 1). Other tissues (brain, testis, skeletal muscle, and heart) did not show a detectable hybridization signal after 24 h of exposure. A chicken liver gt11 cDNA library was used to isolate a fulllength cDNA. Three independent clones hybridizing with the partial cDNA were isolated. Two clones encoded a complete cDNA, and one lacked the N terminus. The total composite cDNA sequence was 1408 bp long with the ATG starting codon at nucleotide 74 and the TGA stop codon at nucleotide 1202 (Fig. 2). The open reading frame encoded a 375-amino acid mature polypeptide with predicted M r of 40,016.
The relationships of this presumed new ADH (ADH-F) with other ADH isozymes were analyzed by progressive alignment (10). Table I shows percentage identity of the new chick enzyme (ADH-F) with other ADH classes (11) as well as the range of percentage identity of the isozymes from different species within the same ADH class. The identity of the new ADH was highest with class I isozymes (69%). Class II and VI ADH were the least similar (about 60% identity) ( Table I).
To characterize the catalytic properties of the new isozyme, the ADH-F cDNA was expressed in E. coli as fusion protein with GST (14). The recombinant enzyme separated from GST by thrombin cleavage had an apparent subunit M r of 40,000 on SDS-polyacrylamide gel electrophoresis. 1 to 2 units of activity were obtained from a 1-liter culture, which corresponded to 12-24 mg of active enzyme. The specific activity of the ADH after thrombin cleavage was the same as that for the ADH-GST fusion protein. Specific activity was determined utilizing fluorescence active site titration by directly measuring the concen-  2. Nucleotide sequence and deduced protein sequence of chick ADH-F. Numbers on the right correspond to nucleotide sequence, and numbers on the left correspond to amino acid sequence. The peptide regions that were used to design degenerate oligonucleotides are underlined. The starting Met is present at nucleotide 74 (MET). The amino acid sequence is numbered from the Ala following the initiating Met codon in accordance with numbering of other ADH isozymes. The termination codon is indicated with an asterisk. The residues discussed in the text are shown in reversed color (white on black background). The insertion of N at amino acid 56 is shown in italic.
tration of NADH binding sites. The K m value of the new ADH for ethanol was relatively high, 17 mM, and the k cat value was 3.4 min Ϫ1 (Table II). The K m values were several orders of magnitude lower for long-chain and large hydrophobic alcohols than for ethanol (Table II). The K m for cinnamyl alcohol was 8.4 M, and the k cat /K m value was 580 min Ϫ1 mM Ϫ1 . The apparent K m values for NAD ϩ and NADH were 5.4 and 5.3 M, respectively (Table II). Inhibition of NAD ϩ reduction by NADH with cinnamyl alcohol held constant at 1 mM was consistent with competitive inhibition. The K i for NADH was 4.0 Ϯ 0.5 M. Inhibition of butanol oxidation by 4-methylpyrazole also fitted best the competitive inhibition model. The K i of 4-methylpyrazole was 300 Ϯ 50 M. These data are consistent with results for horse liver ADH (15,16) and an Ordered Bi Bi mechanism.
The protein corresponding to the wild-type ADH-F was detected in the chicken liver homogenate with the rabbit antiserum raised against recombinant ADH-F (Fig. 3). This antiserum cross-reacted with 100 ng of human class I, II, and IV but not class III ADH proteins (not shown), all of which have similar subunit molecular weights and cannot be separated by SDS-polyacrylamide gel electrophoresis. Thus, isoelectric focusing was employed to separate the 80-kDa dimers of chick ADH isozymes according to their isoelectric points. Recombinant ADH-F appeared as a smear of multiple protein bands with pI values ranging from 7.1 to 8.0 (Fig. 3, lane 1). All bands exhibited activity with 100 mM ethanol, 5 mM trans-2-hexen-1ol, and 100 M 3␤-hydroxysteroid alcohols. Protein bands binding anti-ADH-F antibodies also appeared in the chicken liver homogenate separated by isoelectric focusing in the same range of pI values as the recombinant ADH-F (Fig. 3, lanes 2 and 3). DISCUSSION This study was undertaken to identify the isoforms of cytosolic 40-kDa (subunit) ADH in chick fetal tissues that oxidize the retinoid and steroid alcohols. The only ADH-related PCR product amplified from limb bud mRNA was identified as chick class III ADH based on the 87% protein sequence identity with human class III -ADH (12). Because human class III ADH is not active with retinol, we were not interested in characterizing  the corresponding chick enzyme. However, PCR amplification of embryonal heart mRNA yielded a novel ADH-related cDNA. This new cDNA encoded an active enzyme named ADH-F. Northern blot analysis of tissues from a 7-week-old chicken showed that among seven tissues analyzed, only the liver contained an mRNA hybridizing with the partial PCR product from the fetal heart. The adult heart mRNA did not contain a message that hybridized with this PCR product. A change in the tissue-specific expression of ADH gene between fetal and adult organism was also observed for the class I ADH mRNA in rat (17) and may reflect different metabolic needs of the tissues at different stages of development. The complete cDNA isolated from the chicken liver cDNA library encoded an ADH (ADH-F) with the polypeptide size similar to those of other ADH isozymes (375 amino acids without the starting methionine) (Fig. 2). ADH-F contained 13 residues (Fig. 2) that are conserved in the 47 members of the zinc-containing ADH family excluding -crystallin (4). Sequence alignment with other animal ADHs indicated a single amino acid insertion after position 55. Therefore, the numbers of the conserved amino acid residues after Gly-55 were shifted by one position in chick ADH-F (Fig. 2) when compared with class I ADH (18). The conserved glycines and the valine of the substrate binding domain were present at positions 67, 72, 78, 87, and 81 (Fig. 2). The four glycines of the coenzyme binding domain were in positions 193, 202, 205, and 237. Conserved ligands to the catalytic zinc, Cys-46 and His-68, were also present. The new ADH-F sequence had Asp-224, which has been suggested to determine the specificity for NAD ϩ versus NADP ϩ as a coenzyme, and Thr-48, which is thought to form a hydrogen bond with the alcohol hydroxyl group bound to the catalytic zinc (19). Cysteines 98, 101, 104, and 112, which are responsible for binding the noncatalytic zinc (20), were conserved in the new ADH. However, the sequence of this ADH had less than 68% identity with any of the known ADH isozymes (Table I); hence, we conclude that ADH-F belongs to a separate class in the family of ADHs, class VII.
The new ADH gene encoded an active enzyme when produced as a recombinant protein in E. coli. Antiserum against this recombinant ADH-F recognized protein bands in the chicken liver homogenate with the same range of isoelectric points as the multiple ADH-F forms (Fig. 3). The slightly more basic pI of the recombinant protein is consistent with the lack of N-acetylation in E. coli-expressed proteins (21).
The functional and kinetic properties of the new recombinant ADH-F were compared with those of other ADHs. The yield of ADH-F protein was high (up to 14 mg/L of E. coli culture), but the specific activity with saturating ethanol at pH 7.4 was relatively low (0.08 unit/mg), a value that is similar to that of the human class I ␤ 1 -ADH (0.1 unit/mg) (22). The K m value for ethanol (17 mM) was close to that of human stomach class IV ADH (29 mM) (5) and human liver class II ADH (34 mM) (23) (Table II). NADH inhibited NAD ϩ reduction competitively, and 4-methylpyrazole was a competitive inhibitor of butanol. These inhibition results are consistent with the Ordered Bi Bi mechanism suggested for other ADHs. ADH-F sensitivity to 4-methylpyrazole inhibition was similar to that of the human class IV ADH (K i ϭ 350 M) (5). The K i value (300 M) was greater than that of class I and less than that of class II ADH.
The catalytic efficiency of ADH-F toward the secondary al-   Kedishvili et al. (5), and kinetic constants of class I and II ADH for retinol oxidation are from Yang et al. (2). Remaining kinetic data were reported as follows from the sources listed in the footnotes. All kinetic data were obtained in 0.1 M sodium phosphate at pH 7.4 and 25°C. K m values for alcohols were determined with saturating 2.4 mM NAD ϩ , for NAD ϩ with saturating 1 mM cinnamyl alcohol, and for NADH with saturating 100 M cinnamyl aldehyde.   cohol (R)-(Ϫ)-2-butanol was seven times that of (S)-(ϩ)-2-butanol. This specificity appeared to be similar to that of the human ␣␣ isozyme, where its catalytic efficiency was about four times higher with (R)-(Ϫ)-2-butanol (136 min Ϫ1 mM Ϫ1 ) than with (S)-(ϩ)-2-butanol (37.6 min Ϫ1 mM Ϫ1 ) (24). It has been suggested that the specificity of ADHs toward secondary alcohols is affected by amino acids at positions 48 and 93. Chick ADH-F has Thr-48 as in ␣␣ ADH and a unique Pro at the position homologous to residue 93 (94 in ADH-F). Modeling of the amino acid substitutions present in chick ADH shows that Pro can easily be accommodated at position 93 (Fig. 4) and that the region of the active site occupied by the secondary alcohol more closely resembles ␣␣, with Ala at position 93, than ␤ 1 ␤ 1 , with Phe at position 93. Thus, it is not surprising that this new ADH isozyme possesses a stereospecificity for small secondary alcohols that is more similar to ␣␣ than ␤ 1 ␤ 1 (or horse ADH). The catalytic efficiency (k cat /K m ) of ADH-F was higher for oxidation of large alcohols. For example, ADH-F was 5 ϫ 10 3 times more efficient with trans-2-hexen-1-ol than with ethanol. Chick ADH-F was similar to human class I ␥␥ isozyme in that it oxidized both retinoid and steroid alcohols (Tables III   and IV). It was different from the other two retinol-oxidizing ADH isozymes, class IV and class II , which were not active with steroids (5,25). ADH-F oxidized epiandrosterone about 88 times more slowly than horse SS ADH (0.51 min Ϫ1 versus 44 min Ϫ1 ) (26). The catalytic properties of ADH-F suggest that it may function as a steroid/retinoid dehydrogenase in chick. However, the physiological significance of chick ADH-F for steroid and retinoid metabolism will be clarified once the tissue-specific expression pattern during development and the amount of the active enzyme in tissues are determined.
In general, no single amino acid difference appears to be responsible for the unique kinetic properties of the new chick ADH. The ability to oxidize large hydrophobic alcohols, such as retinol and 3␤,5␣-hydroxysteroids, appears to be the result of several amino acid substitutions near the active site zinc atom and at the entrance to the alcohol binding pocket. With current knowledge, the ability of ADH isozymes to oxidize 3␤-hydroxysteroids has been limited to those that possess a Ser at position 48 (human ␥␥ and horse SS ADH). ADH-F has a Thr at position 48. Its ability to bind steroids productively may be due to substitutions in the vicinity of position 93. In sterol-oxidizing The position for epiandrosterone was found by minimizing the close contacts between the enzyme active site and epiandrosterone. All contact distances are greater than 2.6 Å, except for the distance between the 3␤-hydroxyl group and the catalytic zinc atom, which was modeled at 2.3 Å. B, aligned active site structures of human ␤ 1 ␤ 1 ADH (thin lines) and chick ADH-F (thick lines). The modeled position of epiandrosterone is shown with heavy lines. The unfavorable van der Waals contact (a distance of 2.0 Å) generated between Phe-93 (F93) in human ␤ 1 ␤ 1 ADH and epiandrosterone is shown as a dashed line.
horse SS and human ␥␥, which have a Ser at position 48, there is also a Phe at position 93, which is usually preceded by a Pro-Leu sequence. The sequence Leu-Phe-Pro in chick ADH-F, instead of Pro-Leu-Phe as in most class I isozymes, may account for the difference in steroid alcohol specificity. In addition, there are unusual substitutions at positions 318 and 319, where Leu and Ala substitute for Ile and Phe, respectively. Thus, the "floor" of the alcohol binding pocket appears to be more open in ADH-F compared with class I ␤ 1 ␤ 1 , and this could make the site more accessible to large substrates (Fig. 4). The enzyme appears to be sensitive to the configuration at the 5-position of steroid alcohol, since the 5␤-hydroxysteroid alcohols are inactive (Table IV). Molecular modeling suggests that the stereospecificity at the 5␣-position of the sterol may be due to the presence of the extra methyl group of Thr-48 and the rearranged floor of the substrate binding pocket in the ADH-F compared with ␤ 1 ␤ 1 (Fig. 4).
The productive association of large hydrophobic substrates leading to efficient oxidation is usually associated with rearrangements near the entrance to the alcohol binding pocket. It was shown that both the horse SS and the human class IV isozyme efficiently oxidize sterols and retinol, respectively, due to alterations in the loop at the entrance to the alcohol binding site comprising residues 112-119 (5,27). Both of these isozymes possess single amino acid deletions that appear to widen the mouth of the substrate binding site, permitting easier access for these large substrates. Although chick ADH-F isozyme does not possess such a deletion, the presence of His and Trp in positions 115 and 142, respectively, may affect the conformation of this loop. A neutral or acidic residue at position 115 helps to correctly position this loop by hydrogen bonding with the peptide nitrogen of residue 118 in most class I ADH crystal structures. The His at position 115 will not perform the same function to anchor this loop structure in place, and a conformational change in the structure of this loop could create a more open substrate binding site. The substitution of Asp-115 by Trp in the cod ADH crystal structure appears to be the primary reason for the conformation of this loop to adopt an ␣-helical structure (28). It is not clear whether the insertion of one amino acid in the region between positions 55 and 60 will also affect the structure at the entrance to the substrate binding site. Mutagenesis in class III -ADH, which also has an insertion in this region, strongly implicates a role for Asp-57 in binding of the substrate hydroxymethyl glutathione (29). Our modeling of chick ADH would suggest that Phe-57 could form favorable van der Waals contacts with the hydrophobic face of hydroxysteroids.
Another interesting substitution occurs at position 173. In most ADH isozymes the catalytic zinc ligand Cys-174 is surrounded by two glycines. These glycines may provide the necessary flexible linkage between the catalytic and coenzyme binding domains to allow the large conformational change observed upon coenzyme binding. The presence of Ala at this position may impair the ability of this isozyme to undergo rapid conformational shifts in its structure and may explain, at least in part, the relatively low turnover rate of this isozyme.
Thus, ADH-F appears to be unique in terms of its structurefunction relationships. This enzyme has low specific activity; it is active with 3␤,5␣-hydroxysteroids but not with 3␤,5␤-hydroxysteroids; it is active toward steroid substrates in the absence of Ser-48; and it is active toward retinol in the absence of deletion in the loop between amino acids 115 and 120. Several amino acid substitutions discussed above suggest an explanation for some of its properties. X-ray structure determination of the enzyme will provide more complete insight into the structural basis of its substrate specificity.