Kinetic Cooperativity of Human Liver Alcohol Dehydrogenase γ2

Previous studies showed that natural human liver alcohol dehydrogenase γ exhibits negative cooperativity (substrate activation) with ethanol. Studies with the recombinant γ2 isoenzyme now confirm that observation and show that the saturation kinetics with other alcohols are also nonhyperbolic, whereas the kinetics for reactions with NAD+, NADH, and acetaldehyde are hyperbolic. The substrate activation with ethanol and 1-butanol are explained by an ordered mechanism with an abortive enzyme-NADH-alcohol complex that releases NADH more rapidly than does the enzyme-NADH complex. In contrast, high concentrations of cyclohexanol produce noncompetitive substrate inhibition against varied concentrations of NAD+ and decrease the maximum velocity to 25% of the value that is observed at optimal concentrations of cyclohexanol. Transient kinetics experiments show that cyclohexanol inhibition is due to a slower rate of dissociation of NADH from the abortive enzyme-NADH-cyclohexanol complex than from the enzyme-NADH complex. Fluorescence quenching experiments confirm that the alcohols bind to the enzyme-NADH complex. The nonhyperbolic saturation kinetics for oxidation of ethanol, cyclohexanol, and 1-butanol are quantitatively explained with the abortive complex mechanism. Physiologically relevant concentrations of ethanol would be oxidized predominantly by the abortive complex pathway.


and
HsADH␥ have high catalytic efficiencies on ethanol and contribute significantly to its metabolism (1). Polymorphisms in the ADH3 gene lead to the isoenzymes HsADH␥ 1 , which has Gln-271 and Val-349, and HsADH␥ 2 , with Arg-271 and Ile-349 (3). Genotyping indicates the allele frequency for HsADH␥ 2 is about 10% in East Asians and 43% in Europeans (4,5). The V max for ethanol oxidation by HsADH␥ 1 is 2.5 times higher than that for HsADH␥ 2 (1,6), and it was suggested that susceptibility to alcoholism and cirrhosis may be associated with the presence of HsADH␥ 2 (1,7). However, extensive studies have not established a correlation (4,5,8). Further insights into the possible roles of alcohol dehydrogenases in alcoholism require quantitative descriptions of the kinetics of the various enzymes involved, but the properties of HsADH␥ are a challenge because both isoenzymes exhibit negative cooperativity for ethanol oxidation (6), and the mechanism has not yet been described.
The negative cooperativity could arise from different mechanisms (9). Subunit interactions or "half-of-the-sites" reactivity (an extreme case of negative cooperativity) were reported for the horse liver E (ethanol active) enzyme (10,11) but were later disputed (12)(13)(14). Other studies have found nonadditivity in the heterodimers of horse liver enzymes (15) and human liver enzymes (16), suggesting that the constituent subunits do not act independently. Nonhyperbolic kinetics may also involve mechanisms that do not include subunit interactions, such as a random mechanism (17), a mixture of isoenzymes (18), or an Ordered Bi Bi mechanism with an abortive complex pathway (19,20). Negative cooperativity for ethanol oxidation was observed for a purified human liver alcohol dehydrogenase (18), which may have resulted from a mixture of isoenzymes in the preparation. Oxidation of cyclohexanol by horse liver alcohol dehydrogenase exhibits negative cooperativity (19), and oxidation of ethanol and benzyl alcohol show substrate inhibition (21)(22)(23)(24). These results are explained by an ordered mechanism with alternative pathways, including abortive enzyme-NADHalcohol and binary enzyme-NADH complexes, which differ in the rate of dissociation of NADH. Substrate activation and inhibition observed for recombinant HsADH␥ 2 with different alcohols are explained by a similar mechanism.

EXPERIMENTAL PROCEDURES
Materials-The plasmid for the expression of HsADH␥ 2 (3) was obtained from Dr. Jan-Olov Höög (Karolinska Institutet, Stockholm, Sweden). HsADH␥ 2 was expressed and purified as described previously (25). Protein homogeneity was confirmed by polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulfate. The concentration of active sites was measured by titration with NAD ϩ in the presence of 10 mM pyrazole (26). The turnover number of the purified enzyme in a standard assay at 25°C (27) was 0.5 s Ϫ1 . LiNAD ϩ (grade I) and Na 2 NADH (grade I) were obtained from Roche Molecular Biochemicals. Alcohols and carbonyl compounds were redistilled before use.
Kinetic Studies-Kinetic measurements were made in 50 mM sodium phosphate and 0.25 mM EDTA buffer, pH 7.5, at 25°C. Initial velocities were determined by monitoring the formation or oxidation of NADH with a Cary 118C spectrophotometer (⑀ 340 ϭ 6.22 mM Ϫ1 cm Ϫ1 ) or an SLM Aminco 4800 fluorometer ( ex ϭ 340 nm; em ϭ 460 nm) and fitting the progress curves to a line or a parabola, which calculates the initial slope. Very wide ranges of concentrations of alcohols were used as substrates in some experiments, and velocities at the lowest concentrations were used only if the background rates in the presence of NAD ϩ and enzyme and in the absence of alcohol were less than 10% of the measured rate. Steady-state kinetic data were analyzed using Cleland's programs (28). Data for the nonhyperbolic saturation curves were fit to the TWOONE equation, Equation 1, which is the general form of the equation that describes either activation or partial inhibition by substrates.
V is the velocity at saturating concentrations of the varied substrate, B, and b, c, and d represent collections of rate constants from which K b may be calculated (28). Initial velocity data were fit with the SEQUEN program to the kinetic equation for a sequential Bi reaction in Equation 2, where V 1 is the maximal velocity, A and B are coenzyme and substrate concentrations, K a and K b are the Michaelis constants, and K ia is the dissociation constant for coenzyme. Data for inhibition by high concentrations of alcohol were fit with NONLIN (29) to the hyperbolic replot equation in Equation 3, where V max and V min are the uninhibited and inhibited maximal velocities, respectively, and K d is the dissociation constant for the alcohol, B, from the inhibited form of the enzyme. Steady-state kinetics were simulated with KINSIM (30). Transient kinetic experiments used a BioLogic SFM3 stopped-flow instrument with a dead time of 2.5 ms. Data were analyzed with the BioKine Software. The rate constant for association of NAD ϩ was obtained as described previously (31) using pyrazole to trap the enzyme-NAD ϩ complex and form the ternary complex (26), which absorbs at 294 nm (⌬⑀ 294 ϭ 8400 M Ϫ1 cm Ϫ1 ). The transients were measured at varied concentrations of NAD ϩ (0.09 -0.27 mM) and pyrazole (0.5-5 mM) in 50 mM sodium phosphate, 0.25 mM EDTA, pH 7.5, at 25°C. The rate constants obtained from the exponential phase of each of the traces were fit to the SEQUEN equation (Equation 2). The rate constant for binding of NADH to free enzyme was determined by following quenching of protein fluorescence with ex ϭ 294 nm and em Ͼ 330 nm when 2 N enzyme was mixed with 2-10 M NADH in 50 mM sodium phosphate, 0.25 mM EDTA, pH 7.5, at 25°C. The rate constant for binding of NADH to the enzyme-alcohol binary complex was determined as described above in the presence of either ethanol (100 -400 mM) or cyclohexanol (1-100 mM). The rate of dissociation of NADH from enzyme was measured by trapping free enzyme with 40 mM AMP (a concentration determined to be saturating) and monitoring the absorbance increase at 355 nm (⌬⑀ 355 ϭ 3800 M Ϫ1 cm Ϫ1 ) that results from the shift in spectra for free and enzyme-bound NADH (32).

RESULTS
Steady-state Kinetics-The earlier report of negative cooperativity (substrate activation) for ethanol saturation kinetics with natural HsADH␥ 2 isolated from human liver (6) was confirmed with homogeneous recombinant enzyme (Fig. 1A). Ethanol saturation fit well to the TWOONE equation (Equation 1), and the "concave-up" curvature of the Eadie-Hofstee plot (9) clearly indicates the negative cooperativity (Fig. 1A). Because these results were obtained for the recombinant enzyme, isoenzyme heterogeneity is not likely to be the origin of the observed phenomenon. When the data from Fig. 1A are replotted as in Fig. 1B, some substrate inhibition is also apparent with ethanol concentrations above 100 mM. The dashed and solid lines in Fig. 1 will be discussed later.
Saturation kinetics of HsADH␥ 2 with other alcohols were also measured (Table I). Substrate inhibition, rather than activation, was observed for cyclohexanol ( Fig. 1C) and 1-hexanol, and mechanisms for these alcohols seem to be similar. The saturation data for cyclohexanol fit well to the TWOONE equation. The kinetic data for 1-hexanol (0.034 -8.5 mM) saturation could also be described by the TWOONE equation but were best fit to the equation describing classical substrate inhibition (SUBIN, 28), although the highest concentration of alcohol decreased the velocity only to 80% of the maximum. Interestingly, 1-butanol (Fig. 1D) and R-2-butanol kinetics exhibit both activation and inhibition. Oxidation of 1-butanol and R-2-butanol reaches maximal rates of 0.72 s Ϫ1 and 0.19 s Ϫ1 , respectively, before the inhibition is observed. The steady-state kinetics of methanol oxidation were hyperbolic, exhibiting no indication of activation or inhibition.
The initial velocity kinetics were studied in more detail by collecting data for the forward and reverse reactions with concentrations of the coenzymes and substrates varied over wide ranges. Kinetics for ethanol and cyclohexanol saturation showed nonhyperbolic behavior, even at saturating concentrations of NAD ϩ , which suggests that the activation and inhibition by the alcohols are not likely to be the result of a random Bi Bi mechanism (19). In contrast, the saturation kinetics for NAD ϩ , acetaldehyde, cyclohexanone, and NADH were hyperbolic. Initial velocity data collected for physiologically relevant concentrations of ethanol fit well to the SEQUEN equation (Equation 2), and good estimates of the steady-state kinetic constants could be obtained (Table II). Similarly, by using the noninhibitory concentrations of cyclohexanol, a good fit to the SEQUEN equation was obtained. The K eq values, calculated by the Haldane relationship from the steady-state kinetic constants for both ethanol and acetaldehyde or cyclohexanol and cyclohexanone reactions, are in good agreement with other reported values (33,34), which indicates that the kinetic constants obtained from these initial velocity studies are selfconsistent. The values for the Michaelis and inhibition constants for the ethanol and acetaldehyde reactions are somewhat smaller (3-4-fold) than those obtained previously for HsADH␥ 2 and clearly different from the values for the other human liver class I enzymes (6). Interestingly, the kinetic constants for the respective substrates are similar to those for EqADHE acting on ethanol (15) or cyclohexanol (19,34), except that V 1 and V 2 for the horse liver enzyme are larger than those of HsADH␥ 2 . For ethanol oxidation by HsADH␥ 2 , the values for V 1 and K b in Table II agree well with the values obtained from the TWOONE analysis (Table I). Thus, the values obtained from the initial velocity studies describe the activity in the activation region of the ethanol saturation curve.
The Abortive Complex Pathway-On the basis of the experimental results described herein, the participation of an abortive complex pathway (Scheme 1, A ϭ NAD ϩ , B ϭ alcohol, P ϭ  Table III. The concentrations of the enzyme used in the simulations are 66 -78% of those determined from the NAD ϩ -pyrazole titration. The dashed lines in A, B, and C represent the simulations of the alcohol saturation curves using the Ordered Bi Bi mechanism without the abortive complex pathway. aldehyde or ketone, and Q ϭ NADH) is the best explanation for the nonhyperbolic kinetics observed for the steady-state kinetics with ethanol, cyclohexanol, and 1-butanol.
Steps 1-5 represent the rate constants that describe the Ordered Bi Bi mechanism. For the class I alcohol dehydrogenases, the rate-limiting step for the reaction mechanism is often the release of NADH (k 5 ) for the oxidation of alcohols or release of NAD ϩ (k Ϫ1 ) for the reductive reaction. The abortive complex pathway bypasses step 5 and includes steps 6 -8. With saturating concentrations of NAD ϩ and concentrations of alcohol sufficient to form the enzyme-alcohol-NADH complex (EBQ), release of NADH from this complex (k 7 ) may become ratelimiting. The presence of such a pathway can lead to either substrate activation or inhibition (19,(22)(23)(24). Substrate activation would be observed when k 7 Ͼ k 5 , and substrate inhibition would result when k 7 Ͻ k 5 . Substrate inhibition could also occur with low concentrations of NAD ϩ and concentrations of alcohol sufficient to bind to free enzyme (Scheme 1, step 8; Refs. 19 and 21). The rate equation for the abortive complex pathway describing alcohol saturation kinetics under conditions of saturating NAD ϩ takes the form of the TWOONE equation (Equation 1) when k Ϫ6 Ͼ Ͼ k 7 and k 4 Ͼ Ͼ k 5 or k 7 is assumed (Equation where K b is the Michaelis constant for the alcohol, K qb is the dissociation constant of the alcohol from the enzyme-alcohol-NADH abortive complex, k Ϫ6 /k 6 , and V is k 7 .
The oxidation of various alcohols may be expected to differ in the relative rate constants for the steps in the abortive complex pathway and produce either substrate activation or inhibition that would be diagnostic for the abortive complex pathway. The following experimental results provide evidence for the abortive complex pathway and estimates for the rate constants for the mechanisms of reactions with three alcohols.
Substrate Inhibition and Abortive Complex Formation-The kinetics of inhibition by concentrations of cyclohexanol exceeding those required for maximum velocity was analyzed by an initial velocity study with varied concentrations of NAD ϩ at different fixed concentrations of cyclohexanol ( Fig. 2A). The inhibition is described by both hyperbolic intercept (1/V) and linear slope (K/V) effects (Fig. 2B). The intercept effect suggests that cyclohexanol is exerting its inhibitory effect by binding to a form of the enzyme distinct from that to which NAD ϩ binds, such as enzyme-NADH, to form an abortive complex. That the intercept effect is hyperbolic indicates that the inhibited rate reaches a lower limit, which in the context of the abortive complex pathway would be limited by k 7 . The fit to the hyperbolic replot equation of the apparent V max (from each NAD ϩ saturation curve) versus cyclohexanol concentration (Fig. 2B) gives values for V max , V min , and K d , which can be assigned to k 5 , k 7 , and K qb in Equation 4, respectively. The presence of a slope effect for the inhibition also indicates that cyclohexanol and NAD ϩ can compete for free enzyme. The enzyme-cyclohexanol complex apparently does not readily bind NAD ϩ to form the active ternary complex. Transient kinetic experiments also showed that 400 mM 2,2,2-trifluoroethanol, an analogue of ethanol that does not react with the NAD ϩ , decreases the rate constant for the association of NAD ϩ to 28% of the value in the absence of alcohol. These results are consistent with an ordered mechanism that requires NAD ϩ to bind before alcohol.
A similar inhibition study was also performed with high concentrations of 1-butanol (Fig. 2C). Substrate inhibition by 1-butanol is described primarily by a slope (K/V) effect, which indicates that 1-butanol inhibits HsADH␥ 2 by competing SCHEME 1 TABLE I Alcohol saturation kinetics of human liver alcohol dehydrogenase ␥ 2 Initial velocities were measured in 50 mM sodium phosphate and 0.25 mM EDTA buffer, pH 7.5, at 25°C with varied concentrations of alcohols and a saturating level of 2 mM NAD ϩ . Values for V and K b were obtained from fits to the TWOONE equation (Equation 1), except for the data for methanol, which were fitted to the Michaelis-Menten equation, and for 1-hexanol, which were fitted to the SUBIN equation, ϭ VB/(K b ϩ B ϩ B 2 /K I ), yielding K I ϭ 28 mM. The value of V for cyclohexanol represents the value extrapolated to high, inhibitory concentrations of alcohol. The errors were less than 15% of the values, which indicates that the fits are good (28). alcohol dehydrogenase ␥ 2 Initial velocity and product inhibition studies were performed in 50 mM sodium phosphate and 0.25 mM EDTA buffer at pH 7.5 and 25°C. K a , K b , K p , and K q are Michaelis constants for NAD ϩ , ethanol (or cyclohexanol), acetaldehyde (or cyclohexanone), and NADH, respectively. K ia and K iq are the dissociation constants for NAD ϩ and NADH, respectively. V 1 and V 2 are the turnover numbers for alcohol oxidation and aldehyde or ketone reduction, respectively. The error of the fitted value for K a is 50% for cyclohexanol; all other errors are Յ15% of the fitted values.
b Concentration ranges were 10 -220 M for cyclohexanol, 3-20 M for NAD ϩ , 0.5-10 mM for cyclohexanone, and 3-10 M for NADH. c Calculated from the Haldane relationship, K eq ϭ ( against NAD ϩ for binding to free enzyme (Fig. 2D). The effects on intercepts (1/V) are relatively small and do not permit a good fit to the hyperbolic replot equation. Because 1-butanol saturation kinetics exhibit substrate activation with low concentrations of alcohol and inhibition by high concentrations, it appears that the activation arises from the abortive complex pathway, whereas the inhibition involves competition between NAD ϩ and 1-butanol for free enzyme.
Evidence that the enzyme-NADH-alcohol abortive complex can form was obtained by measuring the quenching of NADH fluorescence during titration of the enzyme-NADH complex with cyclohexanol and 2,2,2-trifluoroethanol (Fig. 3). The dissociation constants of the alcohols from the abortive complexes were 11 and 12 mM for cyclohexanol and 2,2,2-trifluoroethanol, respectively.
The effects of the abortive complex on the kinetics were studied further by a quantitative analysis of the inhibition by cyclohexanol. The cyclohexanol saturation data (Fig. 1C) were fitted by the TWOONE equation (Table I). This analysis shows that the inhibition is partial and the maximally inhibited rate is 0.12 s Ϫ1 . The ratio of the value for V from the TWOONE fit to that of V 1 from the initial velocity study (Table II) indicates that high concentrations of cyclohexanol decrease the activity of HsADH␥ 2 to ϳ25% of the maximum velocity observed at optimal concentrations of cyclohexanol. The partial substrate inhibition was further investigated by determining rates at inhibitory concentrations of cyclohexanol (2.5-200 mM) and fitting the data (Fig. 4A) with a hyperbolic replot equation (Equation 3). This analysis determined that the maximum (V max ) and minimum (V min ) rates were 0.38 and 0.08 s Ϫ1 , respectively, values that are in good agreement with those obtained from the initial velocity and TWOONE analyses discussed above and the values obtained from the analysis described in Fig. 2B. The K d (34 mM) from the fit represents the dissociation constant of cyclohexanol from the enzyme-cyclohexanol-NADH ternary complex (K qb , Equation 4). This value is similar to the K d estimated from the NADH fluorescence quenching experiments (Fig. 3). These results support inhibition by an abortive complex pathway (Scheme 1), where k 7 Ͻ k 5 .
Transient Kinetics of Coenzyme Binding-As measured directly with a stopped flow spectrophotometer, the rate constant for dissociation of NADH from the enzyme-NADH-cyclohexanol complex was 33% of the rate constant for the enzyme-NADH complex (Fig. 4B). Although some estimates of the kinetic values from the hyperbolic replot fit had high errors, the values for k max (0.51 s Ϫ1 ), k min (0.17 s Ϫ1 ), and K d (40 mM) agree well with those obtained from the steady-state kinetics (Figs. 2B and 4A). k max corresponds to the rate constant for dissociation of the enzyme-NADH complex (k 5 ), and k min is an estimate of the rate constant for dissociation of NADH from the enzyme-NADHalcohol complex (k 7 ). These results confirm that cyclohexanol inhibits the overall rate of turnover by decreasing the rate of NADH release, as would be expected for inhibition by the abortive complex pathway where k 7 Ͻ k 5 . Similar experiments tested the effects of ethanol on the rate of dissociation of NADH, but the changes in the apparent rate constants were Only k max was well determined with less than 10% error. k min and K d were not well determined because the highest obtainable cyclohexanol concentration was limited by solubility.
within the experimental error.
NAD ϩ association was measured with varied concentrations of NAD ϩ and pyrazole, and the observed rate constants were fitted to the sequential Bi equation with SEQUEN to yield values for a limiting rate constant of 300 s Ϫ1 and a bimolecular association rate constant for NAD ϩ of 600 mM Ϫ1 s Ϫ1 . The limiting unimolecular step is likely to be due to an isomerization in the NAD ϩ association step, such as a conformational change upon coenzyme binding as is observed for EqADHE (31). The rate constant for NADH association to enzyme, as measured by protein fluorescence quenching, is 2 M Ϫ1 s Ϫ1 . The same value was obtained for NADH association in the presence of ethanol, but a value of 1 M Ϫ1 s Ϫ1 was obtained in the presence of saturating concentrations of cyclohexanol.
Simulation of the Abortive Complex Pathway-Because a complete description of the substrate activation and inhibition by the various alcohols requires assignment of rate constants for each step in the mechanism, the steady-state and transient data were combined for a quantitative analysis using KINSIM. The steady-state kinetics for the oxidation of ethanol, cyclohexanol, and 1-butanol at a fixed concentration of NAD ϩ (Fig.  1) were simulated using determined or estimated values for the rate constants (Table III) for each step of the abortive complex pathway in Scheme 1. The values for constants k 2 , k 3 , k Ϫ3, k 4 , and k Ϫ4 were taken from studies with horse liver enzyme for each of the respective alcohols (35,36). Rate constants k 1 , k 5 , k Ϫ5 , k Ϫ6 /k 6 (for reactions with ethanol and cyclohexanol), and k Ϫ7 were measured by the experiments described above, and an estimate for k Ϫ1 was calculated from K ia (Table II) and the value measured for k 1 . The initial estimate for k 7 for cyclohexanol was in the range of the measured rate constants in Figs. 2B, 4A, and 4B (0.08 to 0.17 s Ϫ1 ) but was refined to produce the best simulation of the data. The value for k 7 required by the simulation was 0.23 s Ϫ1 , which is in good agreement with the measured constants. The values for the remaining steps were adjusted to make the simulations agree with the data.
As would be predicted for activation by ethanol and 1-buta-nol, k 7 is greater than k 5 , whereas for the inhibition by cyclohexanol, k 7 is less than k 5 (Table III). The ratio of k 7 /k 5 is a measure of the extent of the activation or inhibition that is observed. The simulation of the inhibition by cyclohexanol gives a ratio of 0.45, which is in reasonable agreement with the steady-state analysis of 0.25. The activation ratio (k 7 /k 5 ) by ethanol and 1-butanol is 1.1 and 1.7, respectively. An important result from the simulations is that k 7 only needs to be modestly larger than k 5 in order for negative cooperativity to be observed, in particular for ethanol. This finding is consistent with the observation that ethanol did not detectably alter the rate of NADH dissociation. Furthermore, this analysis shows that physiologically relevant concentrations of ethanol would be oxidized predominantly through the abortive complex pathway, as would 1-butanol. 1-Butanol and cyclohexanol, at high concentrations, can also bind appreciably to free enzyme and prevent NAD ϩ binding, as k 8 (Scheme 1) is relatively small. This kind of inhibitory effect is also observed in the slope (K/V) effects observed for both 1-butanol and cyclohexanol inhibition against NAD ϩ (Figs. 2, B and D). Sensitivity analysis was used to determine which rate constants in the simulation made major contributions to the overall velocity at different concentrations of alcohol (Table IV). At higher concentrations of alcohol, k 5 and/or k 7 , the NADH release steps are important in controlling the rates. Simulations of the data using only the Ordered Bi Bi mechanism (without the abortive complex pathway) did not adequately describe the data (dashed lines in Fig. 1, A, B, and D).

DISCUSSION
The Abortive Complex Pathway-The nonhyperbolic kinetics of HsADH␥ 2 (both substrate activation and inhibition) appear to be attributable to the presence of an abortive complex pathway as an alternative to the classical Ordered Bi Bi pathway (Scheme 1). This pathway involves the formation of the ternary enzyme-NADH-alcohol abortive complex from which the rate constant for dissociation of NADH (k 7 ) is different from that for the binary enzyme-NADH complex (k 5 ). When k 7 Ͼ k 5 , substrate activation is observed, and when k 7 Ͻ k 5 , substrate inhibition is observed. Several lines of direct evidence support this explanation. It was shown that cyclohexanol and 2,2,2trifluoroethanol could bind to the enzyme-NADH complex and that cyclohexanol decreased the rate of NADH dissociation to a level that is consistent with the level of inhibition observed in the steady-state kinetics. Effects of ethanol on NADH release were not detectable, but simulation of the steady-state kinetics a The bimolecular association rate constant for NAD ϩ was obtained from stopped-flow experiments as described under "Experimental Procedures." b Calculated from the K ia from the initial velocity studies and the value for k 1 , see Table II.
c Values obtained for EqADHE (35). d Values obtained for EqADHE (36). e Values were set to obtain the best simulation. f Value from k max in Fig. 4B. g The rate constants for binding of NADH to free enzyme and to the enzyme-alcohol binary complex were determined by protein fluorescence quenching as described under "Experimental Procedures." h The values used for k 6 and k Ϫ6 were set so the k Ϫ6 /k 6 ratio (K qb ) is consistent with the measured values (Figs. 2B, 3, and 4, A and B).
i Value from measurement with ethanol was used.  Fig. 1), and each rate constant was set to either 0.5 or 2 times the estimated values found in Table III of ethanol revealed that NADH release from the abortive complex would only have to be 10% faster than that from the enzyme-NADH complex to account for the observed substrate activation. Further simulations also demonstrated that the pathway outlined in Scheme 1 could also explain the observed kinetics for cyclohexanol and 1-butanol. The results also indicate that binding of alcohol to free enzyme can inhibit the binding of NAD ϩ and decrease velocities of alcohol oxidation. The abortive complex pathway is the simplest mechanism that explains all of the results. Previous studies on horse liver alcohol dehydrogenase also invoked the abortive complex pathway mechanism to account for substrate activation and inhibition and included a random pathway for binding of NAD ϩ and alcohol (19,(22)(23)(24). In contrast to the kinetics observed with HsADH␥ 2 , the saturation kinetics for EqADHE showed substrate inhibition by ethanol and substrate activation by cyclohexanol. Ethanol inhibits EqADHE by decreasing the rate constant for NADH dissociation by 4-fold (24).
An abortive complex pathway has also been proposed to account for cooperativity in other enzymes. Sheep liver cytosolic aldehyde dehydrogenase forms complexes with NADH and aldehyde that either lead to substrate activation or inhibition because of rate-limiting NADH release that is either faster or slower (depending on the aldehyde) from the abortive complex than from the binary enzyme-NADH complex (37). The negative cooperativity for glucose saturation observed for monomeric wheat germ hexokinase type L 1 and the substrate inhibition by glucose seen for rat liver glucokinase have been attributed to the presence of an abortive complex pathway that involves the formation of an enzyme-MgADP-substrate complex (38). Dihydrofolate reductase from Escherichia coli also shows substrate activation with NADPH as the varied substrate, and rate constants determined for each step in the mechanism demonstrate that with saturating concentrations of NADPH and dihydrofolate, the predominant pathway for turnover includes an abortive enzyme-NADPH-tetrahydrofolate complex (39,40). It is clear that abortive complex pathways can generate the appearance of cooperativity and should therefore be considered when attempting to identify the mechanisms of cooperativity in enzymes (20).
The abortive complex pathway in HsADH␥ 2 appears to be physiologically relevant. The saturation curves (Fig. 1B), initial velocity studies (Table II), and the sensitivity analysis (Table  IV) show that low concentrations of ethanol are sufficient to promote substantial flux through the pathway involving the enzyme-NADH-ethanol complex. The amount of flux through the abortive complex pathway can be estimated by simulations with k 5 (the rate constant for NADH release from the enzyme-NADH complex, Scheme 1) set to zero. The velocity calculated with 12 mM ethanol (the approximate concentration achieved in a 70-kg person after drinking 3 oz of 40% ethanol) for the abortive complex pathway was 55% of the total velocity.
Because alcohol dehydrogenase has been demonstrated to catalyze a rate-limiting step in alcohol metabolism (1,41,42) and HsADH␥ 2 has significant activity on ethanol, the abortive complex pathway should be considered in studies that model the metabolism of ethanol in man. HsADH␥ 1 also exhibits negative cooperativity for ethanol oxidation (6), and it probably also results from an abortive complex pathway because the differences in amino acid sequences of HsADH␥ 1 and HsADH␥ 2 are not in the substrate binding site. Microscopic rate constants for each step in the mechanism under physiological conditions need to be determined in order to accurately simulate ethanol metabolism in man. It is also important to consider the effects of the abortive complex pathway with other alcohols that might be physiologically relevant.
Structural Considerations-HsADH␥ isoenzymes possess quite different properties than the other class I human and horse liver E isoenzymes, but the differences are not easily explained by comparisons of modeled three-dimensional structures (43). The formation of the abortive complex and its effects on the rates of oxidation of alcohols are observable at alcohol concentration ranges where the other human liver enzymes appear to have hyperbolic saturation kinetics (6). In addition to the ready formation of the enzyme-NADH-alcohol complex, HsADH␥ 2 is unusual among the class I human liver alcohol dehydrogenases in that it has good activity on 3-␤-hydroxy-5-␤-steroids (44) and can bind with high affinity to some bulky inhibitors, such as N-1-methylheptylformamide and N-cyclopentyl-N-cyclopropylformamide (45). HsADH␥ 2 is also the best of the human class I enzymes at catalyzing aldehyde dismutation (46). Perhaps this property is related to an increased ability to form the enzyme-NAD ϩ -aldehyde ternary complex, which is analogous to the enzyme-NADH-alcohol abortive complex. All of these properties suggest that the active site of HsADH␥ differs critically from the other class I enzymes.
HsADH␥ has Val-141 and Val-143, whereas the other enzymes have Leu, Phe, or Met at position 141 and Thr or Ile at 143 (47). Modeling of the HsADH␥ 2 structure from the known horse liver enzyme suggests that small adjustments to the C␣ backbone may be required to accommodate the valine residues. Perhaps the differences at positions 141 and 143 can account for the properties of HsADH␥, but other residues, distant from the active site, may also play a role.