Kinetic mechanism of human class IV alcohol dehydrogenase functioning as retinol dehydrogenase.

Molecular genetic studies have indicated that alcohol dehydrogenase may be involved in the synthesis of retinoic acid, a hormonal molecule regulating diverse cellular functions at the transcriptional level. Class IV alcohol dehydrogenase (ADH) has been reported to be the most efficient enzyme catalyzing oxidation of retinol in human ADH family. Initial velocity, product inhibition, and dead-end inhibition experiments were performed with the recombinant human class IV ADH to elucidate kinetic mechanism with all-trans-retinol and all-trans-retinal as natural substrates. Fluorescence quenching was titrated in formation of the binary and abortive ternary enzyme complexes. The minimal mechanism deduced from steady-state kinetic and equilibrium binding studies is best described as an asymmetric rapid equilibrium random mechanism with two dead-end ternary complexes for retinol oxidation and a rapid equilibrium ordered mechanism with one dead-end ternary complex for retinal reduction, a unique mechanistic form for zinc-containing ADHs in the medium chain dehydrogenase/reductase superfamily. Dissociation constants for the binary complexes as well as the productive and abortive ternary complexes determined from different experimental approaches are in reasonable agreement. Kinetic isotope effect studies suggest rate-limiting isomerization of the central ternary complexes in both reaction directions. The potential interference of retinol metabolism by ethanol through the ADH pathway may play a significant role in the pathogenesis of fetal alcohol syndrome and alcohol-related upper digestive tract cancer.

Human alcohol dehydrogenase (ADH) 1 constitutes a complex family with diverse catalytic properties (1,2). The family members have been classified into five distinct classes, designated ADH1-ADH5, 2 on the basis of the structural and kinetic char-acteristics (3). Three additional classes have been expanded in the vertebrates, i.e. ADH6 in rodents (4,5), ADH7 in chicken (6), and ADH8 in frog (7). Interclass positional identities of the amino acid/nucleotide sequences are ϳ60% (1,3). No single species has been found to contain all of the ADH classes. Human ADH gene family clustering on chromosome 4q22 spanning 367 kb and the mouse ADH gene complex on chromosome 3 spanning 250 kb shows that these mammalian species possess closely linked genes encoding ADH1-ADH5 all in the same transcriptional orientation (8,9).
ADH exhibits a wide substrate specificity. It has been implicated in many physiological processes (1,10). Traditionally, ethanol and steroids were considered to be the main substrates for the horse liver enzyme, thus designated ADH1E (horse class I alcohol dehydrogenase E (ethanol active) enzyme) and ADH1S (horse class I alcohol dehydrogenase S (steroid active) enzyme), respectively (3,11). Retinol has recently received considerable attention as a natural substrate for ADH (12), although it was first reported with horse enzyme over 50 years ago (13). Human ADH1, ADH2, and ADH4 (14 -16), rat ADH1 and ADH4 (17), mouse ADH4 (18), horse ADH1E (19), chicken ADH7 (6), and frog ADH8 (7) have been shown to have retinol activity with K m values in the micromolar range.
Conversion of retinol to retinal is a rate-limiting step in the synthesis of retinoic acid, a ligand controlling a nuclear receptor signaling pathway that regulates growth, development, and epithelial maintenance in vertebrates (12). Retinoic acid has been widely used in the treatment of various skin diseases, especially acne, and also as a cancer chemopreventive agent (20). In human ADH family, ADH4 is the most efficient form for retinol oxidation, exhibiting 10 -113-fold higher V max /K m compared with other members (16). Several lines of evidence support current notion that ADH4 plays a significant role in the synthesis of retinoic acid. ADH4 is predominantly expressed in the mucosa of the upper digestive tract (21)(22)(23), the endothelium/media layer of blood vessels (24), the basal layer of skin (25), testis, and epididymis (26,27). The tissue distribution is compatible with involvement of the local retinoic acid in epidermal differentiation and spermatogenesis. The spatiotemporal expression patterns of ADH4 appear to coincide with the synthesis of retinoic acid during embryogenesis in mice (28,29). The ADH4 gene knockout mice subjected to vitamin A deficiency during gestation show decreased newborn survival compared with wild-type mice (30). This implies that ADH4 may function in the metabolism of retinol to retinoic acid known to be needed for normal development.
It has been thought that interference with retinoic acid synthesis via the ADH pathway by ethanol may contribute to the pathogenesis of fetal alcohol syndrome, a birth defect characterized by craniofacial, limb, and brain malformations (31). The evidence in support of this concept includes the fact that eth-anol competitively inhibits retinol oxidation catalyzed by the human ADH family with apparent K i values ranging from 37 M to 11 mM, indicating that retinoic acid formation in the embryo can be effectively impeded by physiologically attainable levels of blood alcohol during moderate to heavy drinking (16,32). Correlation of expression of ADH4 and the synthesis of all-trans-retinoic acid in developing mouse embryos points to a potential perturbation with ethanol (33,34).
The kinetic mechanism of retinol for human ADH was not documented in the literature, although the horse enzyme with ethanol has long been established as a prototype of the ordered sequential mechanism (35)(36)(37). Structurally, human ADH4 is unique in having a deletion of residue 117, which shortens the loop comprised of residues 114 -120 and widens the outer part of the cylinder-shaped substrate-binding pocket as revealed in x-ray studies (38). Molecular dockings show that the ␤-ionone ring of retinol binds at the widened entrance of the substrate pocket such that an extended conformation of retinol can be accommodated (38). The recombinant ADH4 with an insertion of Gly 117 exhibits decreased catalytic efficiencies for oxidation of both all-trans-retinol and 9-cis-retinol (39). Interestingly, the x-ray structure of horse ADH1S also shows that compared with the ethanol-active ADH1E, a widening of the substrate channel that facilitate access of bulky steroid substrate to the activesite zinc can be attributed to the deletion of Asp 115 in addition to other amino acid exchanges lining the substrate pocket (40). To evaluate the functional role of ADH in the metabolism of retinoids as well as to assess the potential influence by ethanol consumption in the cell, it is important to understand complete steady-state kinetic mechanism with all-trans-retinol and alltrans-retinal. Both are major metabolites of ␤-carotene and precursors of all-trans-retinoic acid, a ligand for retinoic acid receptors in the retinoid signaling pathway (41).

EXPERIMENTAL PROCEDURES
Materials-The human recombinant ADH4 was expressed in Escherichia coli as described previously (38). The resulting enzyme was initially isolated from the lysate supernatant by DEAE-cellulose chromatography (DE52, Whatman; 10 ml of DEAE/liter of culture) in 50 mM Tris, pH 8.8, 0.1% (v/v) 2-mercaptoethanol at 4°C. The unbound proteins were eluted in a batch procedure. The eluate was concentrated and then dialyzed into 10 mM sodium phosphate, pH 6.5, 0.1% 2-mercaptoethanol and loaded onto a 1.6 ϫ 9-cm 5Ј-AMP-Sepharose (Amersham Biosciences) column equilibrated in same buffer. The protein was eluted with a linear gradient from 10 mM sodium phosphate, pH 6.5, 0.1% 2-mercaptoethanol to 100 mM Tris, pH 9.0, 0.1% 2-mercaptoethanol. The active fractions were dialyzed into 10 mM Tris, pH 8.5, 0.5 mM dithiothreitol and applied to a MonoQ column (Amersham Biosciences). The enzyme was eluted with a linear gradient of 0 -200 mM NaCl in same buffer and then dialyzed into 10 mM sodium phosphate, pH 7.5, in the absence of reducing agent for subsequent kinetic and equilibrium binding studies. The recombinant enzyme exhibited a single proteinstaining band with a molecular weight of 40,000 on sodium dodecyl sulfate-polyacrylamide gel electrophoresis by a PhastSystem according to the manufacturer's protocol (Amersham Biosciences). Total protein concentration was determined by a dye binding assay (Bio-Rad) using bovine serum albumin as standard (42). The enzyme activity was assayed in 0.1 M glycine, pH 10.0, containing 2.4 mM NAD ϩ and 100 mM ethanol at 25°C. The specific activity of the purified ADH4 was determined to be 68 mol min Ϫ1 mg Ϫ1 . The enzyme (Ͼ1 mg/ml) appeared stable for at least 1 week in 10 mM sodium phosphate, pH 7.5, when it was kept in an ice bath.
The sodium salt of (4R)[4-2 H]nicotinamide adenine dinucleotide was prepared from NAD ϩ and ethanol-d 5 as described previously (43), and the prepared coenzyme exhibited an A 260 /A 340 of 2.35. NAD ϩ (grade I) and NADH (grade I) were obtained from Roche Molecular Biochemicals, all-trans-retinol ( Ն 99% purity) was from Fluka (Buchs, Switzerland), all-trans-retinal and all-trans-retinoic acid were from Sigma, and ethanol, acetaldehyde, ethanol-d 6 (44). It has been reported that the detergent principally affects the dispersion of retinoid substrate rather than enzyme (44). The addition of acetonitrile (2%) and Tween 80 (0.02%) to the assay buffer appeared not to affect the K m and k cat values for ethanol oxidation for human ADH (16). To eliminate possible interference with kinetic measurements, the final concentrations of Tween 80 and acetonitrile in the assay mixtures remained fixed except that 3% acetonitrile was used for those experiments in which all-trans-retinol was added as a product inhibitor against either all-trans-retinal or NADH because of the higher inhibition constants and the limitation of retinol solubility. Steady-state kinetic data were analyzed by nonlinear regression using the programs of Cleland (45) or the formulating equations that were written according to the statistical package program of SigmaPlot (version 7.0). Initial velocity data were fit with either the HYPER program to the Michaelis-Menten equation in Equation 1 or with the SEQUEN program to the kinetic equation for a two-substrate sequential reaction in Equation 2.
where V is the maximum velocity; A and B are the coenzyme and substrate concentrations, respectively; K ma and K mb are the Michaelis constants for A and B, respectively; and K ia is the dissociation constant of A from the free enzyme E. Maximal activity is expressed as turnover number (min Ϫ1 ) based on a subunit molecular weight of 40,000. To assess the validity of the assumption of rapid equilibrium, the initial velocity data were fit with Equation 3 (the EQORD program) for an obligatory order sequential reaction.
where K ia and K b are the dissociation constants of A from the free enzyme E and of B from the binary EA complex, respectively. The data from dead-end and product inhibition studies were fit with one of the linear inhibition equations, the COMP program for competitive inhibition (Equation 4), the NONCOMP program for noncompetitive inhibition (Equation 5), or the UNCOMP program for uncompetitive inhibition (Equation 6).
where I is the inhibitor concentration, and K is and K ii in Equations 4 -6 are the dissociation constants of a dead-end inhibitor from the EI binary complex and EAI ternary complex, respectively. However, when Equations 4 -6 were used to fit the data from the product inhibition experiment, K is and K ii are the apparent slope inhibition constant and inter-cept inhibition constant of the product inhibitor, respectively. The best fit was determined by evaluating the standard errors of the parameters and the residual variance (45). All of the kinetic measurements were run in duplicate. The values represent the means Ϯ S.E. Equilibrium Binding Studies-The binding of coenzyme and substrate to the enzyme was examined by measuring the induced changes in intrinsic tryptophan fluorescence of ADH4 upon binding. The changes in the fluorescence were measured at an excitation wavelength of 290 nm and an emission wavelength between 330 and 370 nm for the quenching of the protein fluorescence or an emission wavelength of 460 nm for the enhancement of the NADH fluorescence, which is attributed to a resonance energy transfer from tryptophan to the bound reduced coenzyme (47). Similar to horse ADH1E, human ADH4 contains two tryptophan residues per subunit, Trp 15 and Trp 313 (equivalent to Trp 314 in the horse enzyme) (3). Trp 15 is at the surface of the protein, on the outer edge of the catalytic domain, and Trp 313 is part of the coenzymebinding domain, buried in the subunit interface (38). Previous studies with the horse enzyme indicate that Trp 314 is primarily responsible for the quenching/enhancement of fluorescence in the formation of the enzyme-coenzyme binary complexes or the enzyme-coenzyme-dead-end inhibitor ternary complexes (48,49).
All of the fluorescence measurements were conducted in 0.1 M sodium phosphate, pH 7.5, containing 2% (v/v) acetonitrile and 0.02% (v/v) Tween 80 at 25°C and using a PerkinElmer LS 50B luminescence spectrometer. The binding studies were conducted by adding coenzyme/ retinoid substrate in small increments to a 110 -2,000 nN enzyme solution in a volume of 1 ml. After the addition of ligand, the mixture was incubated for 2-3 min, and then the fluorescence spectrum was recorded from 300 -500 nm. The normality of the enzyme was calculated based on a subunit molecular weight of 40,000. The concentrations of the ligand employed in fluorescence titration ranged within 0.06 -6 K d . For determination of dissociation constants, the changes of the protein fluorescence quenching or NADH fluorescence enhancement were fit with a modified Stern-Volmer equation (50).
where F max is the maximal change in fluorescence when all of the free enzyme or the binary enzyme complex is bound by ligand, L is the ligand concentration, and K d is the dissociation constant.

RESULTS
Initial Velocity Patterns-The all-trans-retinol oxidation and all-trans-retinal reduction were studied by varying the substrate concentrations at different fixed corresponding coenzyme concentrations. Examples of the double-reciprocal plots are shown in Fig. 1. For oxidation of retinol, the double reciprocal plots for either NAD ϩ or retinol were linear and intersected above the abscissa and to the left of the ordinate, which is consistent with a sequential kinetic mechanism. The data for both the substrates fit well to the sequential bireactant mechanism (Equation 2). Table I shows the calculated kinetic constants K a , K b , K ia , and V 1 with standard errors less than 13%. The double reciprocal plots for NADH or retinal in the reverse reaction were linear, and the former intersected above the abscissa and to the left of the ordinate, whereas the latter intersected on the ordinate and above the abscissa (Fig. 1B). The data did not fit Equation 2 as revealed by a large standard error (Ͼ 100%) of the calculated K m for NADH. This is because of very small intercept effect observed in the double reciprocal plot. But the data fit well to Equation 3, which uniquely illustrates a rapid equilibrium ordered mechanism. The standard errors of the resulting kinetic constants K p , K iq , and V 2 were 16% or less (Table I). A rapid equilibrium ordered mechanism is expected to show a slope approaching 0 in double reciprocal plot of the first reactant if using a highly saturating concentration of the second reactant. This experiment was not carried out because of solubility limitation of retinal in the assay buffer.
Product and Dead-end Inhibitions-Product inhibition studies were performed to determine whether the kinetic mechanism was random or ordered. The studies were carried out by varying the concentrations of one of the substrates while keeping the other substrate level constant in the absence or presence of different amounts of one of the products. The NADH versus NAD ϩ and retinal versus retinol product inhibition patterns fit linear competitive inhibition (results not shown). The NADH versus retinol product inhibition pattern fit linear noncompetitive inhibition (Fig. 1C). These results indicate that the oxidation of retinol is consistent with either a rapid equilibrium random mechanism with abortive ternary enzyme-NAD ϩ -retinal and enzyme-NADH-retinol complexes or a Theorell-Chance ordered mechanism. The retinol versus retinal and retinol versus NADH product inhibition patterns fit linear competitive (results not shown) and linear noncompetitive inhibition ( 1D), respectively. The NAD ϩ versus NADH product inhibition pattern was obtained with unacceptable statistical significance because of a very low K m for NADH and the limitation of sensitivity of the spectrophotometer used. Retinal was not inhibited by the product NAD ϩ at saturating NADH (data not shown). These results suggest that the reduction of retinal conforms to either a rapid equilibrium ordered mechanism with an abortive enzyme-NADH-retinol complex or a Theorell-Chance ordered mechanism. However, the latter was precluded from the initial velocity studies (Fig. 1B). The calculated Michaelis constants or inhibition constants for the binary and ternary complexes from product inhibition studies displayed standard errors of 14% or less except for the K Iq value of 18% (Table I), indicating that the data fit well to the respective formulating equations.
Dead-end inhibition studies were performed to distinguish a random mechanism from an ordered mechanism for oxidation of retinol. Trifluoroethanol, which binds to substrate pocket of the enzyme, was found to be a competitive inhibitor against retinol (Fig. 1E) and a noncompetitive inhibitor against NAD ϩ (results not shown). The data fit well to linear noncompetitive inhibition against coenzyme with standard errors less than 13% for the slope and intercept inhibition constants (Table II). This result eliminated the possibility of a Theorell-Chance mechanism. Thus dead-end inhibition studies suggest that the forward reaction in the absence of products conforms to a rapid equilibrium random mechanism. It was attempted to utilize all-trans-retinoic acid as a potential substrate inhibitor. Preliminary observations showed that the slope inhibition constant could not be precisely determined as a result of relative weak inhibition by all-trans-retinoic acid and solubility limitation of the inhibitor in the assay buffer. The inhibition patterns of the substrate inhibitor isobutyramide were found to be competitive with retinal (results not shown) but uncompetitive with NADH (Fig. 1F). The data fit the corresponding linear inhibition equations with standard errors less than 10% for the resultant inhibition constants (Table II). These results support the possibility that the reverse reaction conforms to an ordered rather than a random mechanism.
Equilibrium Binding-Equilibrium binding studies were performed to assess the formation of binary and ternary complexes. Quenching of native enzyme tryptophan fluorescence was observed in formation of the enzyme complexes with NAD ϩ or retinal ( Fig. 2A) as well as with NADH or retinol (Fig. 2B). Further quenching of the fluorescence of the binary complexes was observed with formation of the abortive ternary complexes. The data of fluorescence titration fit well to a modified Stern-Volmer equation (Fig. 3). The standard errors of the resulting  Fig. 1. E, A, B, P, and Q represent the free enzyme, NAD ϩ , all-trans-retinol, all-trans-retinal, and NADH, respectively. The Michaelis constants K a , K b , and K p are dissociation constants for the ternary complexes of A, B, and P with EB, EA, and EQ, respectively. The inhibition constants K ia , K ib , and K iq represent dissociation constants for the binary complexes of E with A, B, and Q, respectively. The inhibition constants K Ib , K Ip , and K Iq are also dissociation constants, but for the ternary complexes of B, P, and Q with different forms of the enzyme, namely, EQ, EA, and EB, respectively. ␣ is the factor that represents synergism of the substrate and coenzyme binding to the enzyme. The values are the means Ϯ S.E.

Kinetic constants
Initial velocity Product inhibition From fitting the data to Equation 2 or 9 (see Fig. 1A). It is worth noting that Equations 2 and 9 are identical in algebraic form.  . 1C); the K b value was corrected using K b ϭ K m (1 ϩ K a /A)/(1 ϩ K ia /A), the K iq value was corrected by K iq ϭ K is /(1 ϩ A/K ia ), and the K Iq value was corrected by K Iq ϭ K ii /(1 ϩ A/K a ), where K a ϭ 280 M and K ia ϭ 620 M. e From fitting the data to Equation 3 (see Fig. 1B). . The K p value was corrected using K p ϭ K m /(1 ϩ K iq /Q), and the K Ib value was corrected by Because the nonvaried substrate NADH was almost saturating, the observed K m and K is values are nearly to the desired constants. g From fitting the data to Equation 5 (see Fig. 1D); the K Ib value was corrected using K Ib ϭ K ii /(1 ϩ P/K p ) where K p ϭ 68 M. The K ib value (K ib ϭ K is and hence no correction needed) was 185 Ϯ 54 M; this value was not precise as revealed by the large standard error that was mainly due to low solubility of all-trans-retinol (the highest concentration used in experiments, 100 M).
h The equilibrium constant was calculated from K eq ϭ V   . 1E); the K is value was corrected using K is ϭ K is TFE) and K ia ϭ 620 M.
b From dead-end inhibition by 2,2,2-trifluoroethanol with 0, 4, 8, and 16 mM trifluoroethanol against varied concentrations of NAD ϩ at 30 M all-trans-retinol. The data were fit to Equation 5. The K is value was corrected using K is ϭ K is app /(1 ϩ B/K ib ), and the K ii value was corrected by c From dead-end inhibition by isobutyramide with 0, 50, 100, and 200 mM isobutyramide against varied concentrations of all-trans-retinal at 200 M NADH. The data were fit to Equation 4. The K is value was corrected using K is ϭ K iis app /(1 ϩ K iq /Q), where K is ϭ K i(EQ,IBA) and K iq ϭ 2.3 M. Because the nonvaried substrate NADH was almost saturating, the observed K is value is nearly to the desired constant.
d From fitting the data to Equation 6 (see Fig. 1F); the K ii value was corrected using K ii ϭ K ii app /(1 ϩ P/K p ), where K ii ϭ K i(EQ,IBA) and K p ϭ 68 M. dissociation constants were 6 -11% for the binary complexes and 18% or less for the ternary complexes (Table III). Formation of a complex of binary enzyme-NAD ϩ with retinal was not quantitatively assessed because of the small fluorescence changes observed with titrating of the ternary complex. The dissociation constants for enzyme and NADH determined by fluorescence quenching (1.0 M) were similar to 1.6 M, which was determined by the fluorescence enhancement through resonance energy transfer from tryptophan to NADH (Table III). Both of the values are in good agreement with those by the product inhibition studies after correction for the concentration of the nonvaried substrate or coenzyme, i.e. K iq ϭ 1.9 -2.3 M. A random kinetic mechanism for the forward reaction predicted that the enzyme combines with either NAD ϩ or retinol. The dissociation constants for the binary enzyme-NAD ϩ and enzyme-retinol complexes from the equilibrium binding studies (750 and 65 M) were close to those from the steady-state kinetic studies (K ia ϭ 620 M and K ib ϭ 49 M) giving additional support to the theory that the oxidation of retinol conforms to a rapid equilibrium random mechanism.
Isotope Effects-A rapid equilibrium mechanism, either ordered or random, predicted that the central complexes occur as the predominant enzyme species relative to those of free enzyme and binary complexes. To assess whether conversion of the central complexes acts as a rate-determining step in reduction of retinal, kinetic isotope effect experiments were performed with stereospecifically deuterated coenzyme (4R)[4-2 H]nicotinamide adenine dinucleotide in 0.1 M sodium phosphate, pH 7.5, containing 2% (v/v) acetonitrile and 0.02% (v/v) Tween 80 at 25°C. For comparison, the isotope effects for acetaldehyde reduction were also measured in the same buffer but in the absence of acetonitrile and Tween 80. The isotope effects for V 2 and V 2 /K p for reduction of retinal and acetaldehyde were determined at 200 and 400 M reduced coenzyme, respectively, and for V 2 /K q at 60 M all-trans-retinal or 60 mM acetaldehyde. No significant deuterium isotope effect was observed for the maximum velocity with retinal ( D V 2 ϭ 1.0) or acetaldehyde ( D V 2 ϭ 1.4) as substrate. Thus hydride transfer appeared not to be rate-limiting in both the reactions. A more pronounced effect was observed for the D V 2 /K p relative to D V 2 /K q with acetaldehyde reduction ( D V 2 /K p ϭ 2.4 versus D V 2 /K q ϭ 1.1) compared with that with retinal reduction ( D V 2 /K p ϭ 1.0 versus D V 2 /K q ϭ 0.75). This is consistent with a steady-state ordered mechanism for acetaldehyde reduction that was reported in an earlier study with human ADH4 (51).

DISCUSSION
Deduction of Mechanism-It has long been established as a prototype that horse ADH1E conforms to a symmetric ordered sequential mechanism catalyzing the reversible oxidation of ethanol using NAD ϩ as a cofactor (36,37). Human ADH1B, ADH2, and ADH4 follow the same mechanism with the same substrates (51)(52)(53). In this report we have deduced from the studies of initial velocity, product, and dead-end inhibition as well as equilibrium binding that human ADH4, when functioning as a retinol dehydrogenase, conforms to an asymmetric mechanism, random for the oxidation of all-trans-retinol and ordered for the reduction of all-trans-retinal, under rapid equilibrium for binding the substrate and coenzyme for both reactions. To our knowledge, this is a first description of kinetic mechanism with retinol for the human ADH family, and it appears to be the first documentation of a rapid equilibrium ordered mechanism observed with zinc-containing ADHs across species in the medium chain dehydrogenase/reductase superfamily. Binding bireactants reaching equilibrium before conversion of the central ternary complex is consistent with the observation that as a retinol dehydrogenase human ADH4 reacts considerably slower compared with its activity as an ethanol dehydrogenase. The maximum velocity for the retinol dehydrogenase reaction was found to be 14-fold lower for oxidation and 830-fold lower for reduction than that for the ethanol dehydrogenase reaction (Table I and Ref. 51). In the latter reaction, the rate-limiting step appeared to be dissociation of the enzyme-NADH complex for the forward reaction and isomerization of the enzyme-NAD ϩ complex for the reverse reaction (51).
One of the most convincing pieces of evidence that reduction of retinal for human ADH4 conforms to a rapid equilibrium ordered mechanism came from the distinctive initial velocity pattern of intersecting on the ordinate (Fig. 1B). The uncompetitive inhibition pattern of substrate analog isobutyramide against coenzyme NADH in dead-end inhibition studies (Fig.   1F) has ruled out a random mechanism for the reduction reaction. The competitive product inhibition of retinol versus retinal indicates an abortive enzyme-NADH-retinol complex. With regard to oxidation of retinol, the competitive product inhibition pattern of retinal versus retinol has precluded either a steady-state simple ordered or a steady-state random mechanism, both of which should display a noncompetitive pattern. The noncompetitive inhibition pattern of substrate analog trifluoroethanol against NAD ϩ has precluded either a Theorell-Chance ordered mechanism or a simple ordered mechanism under steady state or rapid equilibrium. Taking together the linear double-reciprocal plot (Fig. 1A), the product inhibition and dead-end inhibition patterns, as well as results of the equilibrium binding, the oxidation of retinol for human ADH4 is best described as a rapid equilibrium random mechanism with abortive enzyme-NAD ϩ -retinal and enzyme-NADH-retinol complexes. Scheme 1 illustrates the proposed minimal mechanism for the overall reactions. The complete kinetic equation for the mechanism in Scheme 1 was derived using the method of Cha (55).
The expression is similar to the one for the simple symmetric ordered Bi Bi mechanism, except that it has different coefficients for terms in AP and BQ, and the terms in P, ABP, and BPQ are missing.
Quantitative Analysis-Quantitative analysis for steadystate kinetic constants predicted from the proposed mechanism where rapid equilibrium conditions apply in comparison with data from equilibrium binding studies lends additional support. As shown in Table I, the values for K a , K b , K p , and K iq are in good agreement with those determined from the initial velocity and product inhibition experiments. The values for K ia and K ib match well with those for K E,A and K E,B determined from the initial velocity and equilibrium binding experiments, respectively (Tables I and III). The K iq , K Ib , and K Iq values from the product inhibitions are reasonably close to those of K E,Q , K EQ,B , and K EB,Q from the equilibrium binding. Thus all kinetic data in Table I, except for K Ip , which had no comparison, agree from two different experimental approaches, providing a quantitative evidence for the proposed mechanism (Scheme 1). It should be noted that the constants from the product inhibition studies were corrected for the concentration of the fixed substrate according to the proposed mechanism (for the correction equations, see legend of Table I). All of the corrected values for the constants appeared to better match the corresponding initial velocity or equilibrium binding data. It is interesting to Alcohol Dehydrogenase as Retinol Dehydrogenase note that the rapid equilibrium ordered mechanism appeared to be a rare documentation in the literature. Schimerlik and Cleland (56) presented a similar mechanism with the rabbit creatine kinase at neutral pH. Two Haldane relationships were derived from Equation 8, which represents a ratio for the termolecular rate constant of forward reaction to that of reverse reaction. It is worth noting that the Michaelis constant in this case under rapid equilibrium condition represents a dissociation constant for the second reactant from the corresponding binary complex.
The overall equilibrium constant calculated from the Haldane relationships (1.5 nM) agrees well with the experimentally determined value (2.0 nM) (19), indicating that the kinetic constants are self-consistent. According to the rapid equilibrium random mechanism, the binding synergism of the substrates to the enzyme can be quantitatively analyzed by the following equation (46).
where ␣ is a factor representing interaction of the substrate and coenzyme binding, i.e. if the binding of one ligand increases the affinity for the second ligand (␣ Ͻ 1, i.e. binding synergism) and vice versa (i.e. binding antagonism) or if one ligand has no effect on the binding of the other, ␣ ϭ 1. The ␣ value was calculated to be 0.45 Ϯ 0.09, clearly indicating a synergism between NAD ϩ and retinol binding to the enzyme. The synergistic substrate binding can explain in part the randomness for the oxidation of retinol. It is worth noting that Equations 10 and 2 are identical in algebraic form but that the former is defined as ␣K ia ϭ K a and ␣K ib ϭ K b , where K ia and K ib are the dissociation constants of A and B from the free enzyme E, respectively; K a and K b are the dissociation constants of A and B from the binary EB and EA complexes, respectively. Trifluoroethanol (TFE) acts as a competitive inhibitor against retinol (Fig. 1E) but noncompetitive against NAD ϩ . Isobutyramide (IBA) competitively inhibits retinal but uncompetitively inhibits NADH (Fig. 1F). The deduced mechanism of the dead-end inhibition is illustrated in Scheme 2, where I 1 and I 2 (inhibitors 1 and 2) represent TFE and IBA, respectively. K i1 and K i2 are the corresponding inhibition constants. ␤ is the factor by which the dissociation constant for substrate is changed by the binding of inhibitor and also the factor by which the inhibitor dissociation constant is changed by the binding of the substrate. Under rapid equilibrium condition, the products of the dissociation constants through either binding path should be equal, i.e. (K i1 )(␤K ia ) ϭ (K ia )(␤K i1 ). Equations 11 and 12 are derived kinetic formulations for inhibitor 1 or inhibitor 2 in the absence of the reaction products, respectively.
The constants from dead-end inhibition studies were corrected for the concentration of the fixed substrate according to the proposed mechanism (for calculations, see the Table II footnotes). The dissociation constant for E-NAD ϩ and TFE determined from the experiment of TFE against retinol, i.e. K is ϭ ␤K i1 ϭ 8.0 mM (cf. 8.1 mM before correction), is found close to that determined from against NAD ϩ , i.e. K ii ϭ ␤K i1 ϭ 9.7 mM (cf. 23 mM before correction), giving additional support to the proposed rapid equilibrium random mechanism for the oxidation of retinol. In the latter experiment, the dissociation constant for the free enzyme and TFE, i.e. K is ϭ K i1 ϭ 9.3 mM (cf. 15 mM before correction), is also found close to the corrected K ii value of 9.7 mM, indicating a pure noncompetitive inhibition pattern (i.e. intersecting on the abscissa) instead of the apparent mixed type noncompetitive pattern (results not shown).
The K ia value calculated from the abscissa of the intersecting point of the double-reciprocal plots (470 M) agreed with those determined from the initial velocity and equilibrium binding experiments. The ␤ value (1.1 Ϯ 0.1) appeared to be close to 1, suggesting that there is no significant binding synergism with respect to NAD ϩ and TFE. The dissociation constant for the binary E-NADH complex and IBA determined from the experiment of IBA against NADH, i.e. K ii ϭ K i2 ϭ 290 mM (cf. 440 mM before correction), appeared to be greater than that determined from against retinal, i.e. K is ϭ K i2 ϭ 95 mM (the value was unchanged with correction). The slight discrepancy could be in part due to a potential nonspecific interaction of the inhibitor with hydrophobic regions of the enzyme at high IBA concentrations (up to 450 mM in measurements). Therefore, qualitative analysis of kinetic behavior patterns from the initial velocity, product, and dead-end inhibition studies, as well as quantitative analysis of the resulting kinetic constants in comparison with the equilibrium binding data supports the possibility that human ADH4 functioning as a retinol dehydrogenase conforms to a predominantly random order for oxidation with abortive EAP and EBQ complexes and a predominantly obligatory order for reduction with an abortive EQB complex that approximation of the rapid equilibrium condition can be applied to both the reactions (Scheme 1). It is interesting to note that horse ADH1E, a prototype enzyme that catalyzes oxidation of ethanol as well as other primary or secondary aliphatic alcohols, has been generally described as ordered but with some degree of randomness for alcohol oxidation, whereas it is strictly ordered for reduction of the corresponding aldehydes with rate limiting at E-NADH dissociation for the oxidation and E-NAD ϩ isomerization for the reduction as revealed from steady-state, pre-steady-state, and isotope exchange studies (35)(36)(37)(57)(58)(59)(60). Thus when functioning as a retinol dehydrogenase, human ADH4 appears to be an interesting special case in kinetic mechanisms documented in general for the zinc-containing ADHs in the medium chain dehydrogenase/reductase superfamily. Rate-limiting Step-A rapid equilibrium bireactant kinetic SCHEME 3 SCHEME 2 mechanism implies that the steady-state concentrations of the ternary complexes are predominant relative to other enzyme species in the catalytic cycle irrespective of the reactant binding order. The predominant central complex species could be created by slow step(s) in the chemical transformation, isomerization of the central complexes, or release of the first product.
To effectively reach equilibrium for substrate binding before conversion of the central complex to produce the product, the rates for dissociation of the reactant from either a binary or a ternary complex (e.g. k Ϫ1 to k Ϫ4 , or k 7 and k 8 ; see Scheme 3) should be at least an order of magnitude greater than V max for that reaction. Because both the forward and reverse reactions reach equilibrium and the difference between V 1 and V 2 is less than 4-fold (Table I), release of the first product seems unlikely to be the slowest step for either direction. The rates for dissociation of the binary enzyme-coenzyme complexes estimated from steady-state kinetic studies with ethanol and acetaldehyde for human ADH4 (51) are found to be at least 50-fold greater than the V max for the opposite direction as determined with retinol/retinal as the substrate. The comparison indicates that dissociation or isomerization of the enzyme-coenzyme complex cannot be rate-limiting in the latter case and that much faster rates for the coenzyme product release are consistent with a rapid equilibrium mechanism with the retinol and retinal. Hydride transfer could be rate-limiting in a rapid equilibrium mechanism. Because no deuterium isotope effect was observed for reduction of retinal, D V 2 ϭ 1.0 and D V 2 /K p ϭ 1.0, and equilibrium isotope effect for the primary alcohol (ethanol) has been reported close to unity, D K eq ϭ D (V 1 /K b K ia )/ D (V 2 / K p K iq ) ϭ 1.07 (61), a simple explanation would include an extra step as shown in Scheme 3 (for clarity the abortive complexes are omitted). As an internal commitment factor, the slow isomerization of central complexes can reduce the magnitude of the observed isotope effect (Ref. 62 and Scheme 3). Thus kinetic isotope effect studies suggest that k 6 and k Ϫ6 are rate-limiting for retinol oxidation and retinal reduction, respectively.
In conclusion, the bulky retinol molecule fitting well into the substrate pocket (38,63), which involves multiple hydrophobic forces as well as the extremely slow catalytic rate, may play important roles in the mechanism of rapid equilibrium random for oxidation reaction and in approximation of the rapid equilibrium condition for reduction reaction for human ADH4 acting as retinol dehydrogenase. The inhibition of retinol oxidation by ethanol through human ADH4 that results in reduction of retinoic acid synthesis in the target cell may achieve by combined effects of the competitive blockade of binding retinol to the enzyme substrate pocket (16) and the product inhibition of oxidation rate by coenzyme NADH (64) as revealed from the derived steady-state rate equation and the experimental values of the kinetic constants.