The Potential Roles of the Conserved Amino Acids in Human Liver Mitochondrial Aldehyde Dehydrogenase*

The sequence alignment of all known aldehyde dehydrogenases showed that only 23 residues were completely conserved (Hempel, J., Nicholas, H., and Lindahl, R. (1993) Protein Sci.2, 1890–1900). Of these 14 were glycines and prolines. Site-directed mutagenesis showed that Cys302 was the essential nucleophile and that Glu268 was the general base necessary to activate Cys302 for both the dehydrogenase and esterase reaction. Here we report the mutational analysis of other conserved residues possessing reactive side chains Arg84, Lys192, Thr384, Glu399, and Ser471, along with partially conserved Glu398and Lys489, to determine their involvement in the catalytic process and correlate these finding with the known structure of mitochondrial ALDH (Steinmetz, C. G., Xie, P.-G., Weiner, H., and Hurley, T. D. (1997) Structure 5, 701–711). No residue was found to be absolutely essential, but all the mutations caused a decrease in the specific activity of the enzyme. None of the mutations affected the K m for aldehyde significantly, although k 3, the rate constant calculated for aldehyde binding was decreased. TheK m and dissociation constant (K ia ) for NAD+ increased significantly for K192Q and S471A compared with the native enzyme. Mutations of only Lys192 and Glu399, both NAD+-ribose binding residues, led to a change in the rate-limiting step such that hydride transfer became rate-limiting, not deacylation. Esterase activity of all mutants decreased even though mutations affected different catalytic steps in the dehydrogenase reaction.

in the human mitochondrial ALDH, which possessed a functional side chain. Since completion of the study, the threedimensional structure of the corresponding beef liver enzyme has been determined to 2.65 Å (3).
Prior to the advent of molecular biological techniques, chemical modifications were used to identify the components of the active site of the enzyme. Classical sulfhydryl reagents inactivated the enzyme (4 -6). Iodoacetamide was shown to modify Cys 302 in human ALDH (4) while using protection studies; our laboratory reported that Cys 49 was a component of the active site of the horse liver enzyme (7). Two other residues were identified by chemical modifications as being possibly involved in the catalytic process. These were Glu 268 and Ser 74 in the human (8) and sheep liver (9) enzymes, respectively. It was shown later by site-directed mutagenesis studies that Cys 302 is a nucleophile (10) and Glu 268 (11) functions as a general base during catalysis. However, Ser 74 (12) and Cys 49 /Cys 162 (10) were found to be not essential for the catalytic reaction.
The kinetics of ALDH was found to follow an ordered sequential mechanism where NAD ϩ binds first followed by aldehyde (13). The reaction involves both acylation and deacylation steps during the oxidation of aldehyde to acid, or hydrolysis of nitrophenyl acetate. It was proposed that deacylation (k 7 ) ( Fig. 1) was rate-limiting for horse liver ALDH2 (13,14) for the dehydrogenase reaction, while acylation (k 3 ) was rate-limiting for the esterase reaction.
Here we report the properties of human ALDH2 mutant enzymes produced by replacing the conserved amino acids possessing a reactive side chain with different residues. The purpose of this study is to understand the potential role of the conserved residues in the catalytic mechanism of ALDH and to relate the effects to the now known structure. In the accompanying paper, a detailed analysis of the Lys 192 and Glu 399 mutants that caused a change in the rate-limiting step of the enzyme will be presented (15).

EXPERIMENTAL PROCEDURES
Materials-NAD ϩ and NADH were purchased from Sigma; Sequenase version 2.0 kit was obtained from United States Biochemical Corp.; propionaldehyde, chloroacetaldehyde, and p-nitrophenyl acetate were from Aldrich; Magic Minipreps DNA purification system and T4 DNA ligase were from Promega Corp.; alkaline phosphatase-conjugated goat anti-rabbit IgG and Muta-Gene in vitro mutagenesis kit were from Bio-Rad; GeneClean kit was from Bio 101, Inc.; [␣-35 S]dATP was from Amersham Corp.; and the restriction enzymes used were from either New England Biolabs or Promega Corp.
Oligonucleotide-directed Mutagenesis-To construct the human ALDH2 R84E/Q, K192Q, T384A/S, K489E/Q, S471A/T, E398K, and E399Q mutants, the oligonucleotide primers containing the mutation were used for site-directed mutagenesis with the Mutagene Kit following the manufacturer's instructions. The mutant colonies were selected by sequencing using the dideoxynucleotide chain-termination method (19). After the identification of mutant colonies, the cDNA fragments containing the mutants were exchanged with the corresponding fragments of the native ALDH2 cDNA from the pT7-7 plasmid and were transformed in BL21 (DE3) pLysS cells possessing the chloramphenicol resistance pLysS plasmid. The mutation was again confirmed by double-stranded DNA sequencing of the pT7-7 plasmid.
Expression and Purification of Native and Mutant Enzymes-All the enzymes were expressed and purified according to the methods described elsewhere (11). Recombinantly expressed enzymes were purified by protamine sulfate treatment (1.25 mg/ml) and DEAE-cellulose and 4-hydroxyacetophenone-based affinity chromatography (20). The purity of the enzymes was determined by SDS-PAGE using the Coomassie Blue staining procedure. Fractions containing only ALDH2 were pooled and concentrated using a Centricon unit (Amicon). The pure enzyme was stored at Ϫ20°C in presence of 50% glycerol.
Fluorescence Assay for the Dehydrogenase Activity-The dehydrogenase activity assays were performed by measuring the rate of increase in the fluorescence of NADH formation in 100 mM sodium phosphate (pH 7.4) at 25°C (21). The K m and V max values for NAD ϩ were determined in the presence of 14 M propionaldehyde, and K m and V max values for propionaldehyde were determined in the presence of 7 mM NAD ϩ for the K192Q mutant, 5 mM NAD ϩ for the S471A/T mutants, 2 mM for the E399Q mutant, 2 mM for the E398K mutant, and 1 mM NAD ϩ for the rest of the mutants and the native ALDH2 enzymes. Assays for the bisubstrate reaction kinetics for the native, K192Q, and S471A mutant enzymes were performed at different concentrations of propionaldehyde and NAD ϩ .
Spectrophotometric Assay for Esterase Activity-The esterase activities of recombinantly expressed native and mutant enzymes were determined by assaying the rate of p-nitrophenol formation at 400 nm in 100 mM sodium phosphate (pH 7.4) with 800 M p-nitrophenyl acetate as the substrate. A molar extinction coefficient of 18.3 ϫ 10 3 at 400 nm for p-nitrophenolate and a pK a of 7.1 for p-nitrophenol were used to calculate the rate of the p-nitrophenol formation (11,13).
Pre-steady State Burst of NADH Formation-The pre-steady state burst magnitude of NADH formation was determined with a Hitachi 2000 Fluorescence Spectrophotometer (12,21). Enzyme and NAD ϩ were incubated in 100 mM sodium phosphate (pH 7.4) to establish a fluorescence base line. Concentrations of NAD ϩ were 1-7 mM for the native and different mutant enzymes. At a time called zero, propionaldehyde (140 M) was added to initiate the reaction. The extrapolated line intersecting at time zero gave the magnitude of the burst of NADH formation. By calibrating the fluorometer with various concentrations of NADH, it was possible to calculate the moles of NADH produced prior to the steady state rate of NADH formation (12,14).
Determination of the Dissociation Constant (K d ) for NADH-The K d values for enzyme-NADH were determined for the native, S471A, and K192Q mutants in 100 mM sodium phosphate (pH 7.4) at 25°C. The fluorescence emission of unbound and bound NADH was measured at 450 nm with excitation at 340 nm (12, 22) with a Hitachi 2000 Fluorescence Spectrophotometer.
Determination of Protein Concentration-The protein concentration was determined with the Bio-Rad protein assay kit, using bovine serum albumin as a standard.

Expression and Purification of Human Native and Mutant
Forms of ALDH2-Recombinantly expressed native and the mutant forms of human liver ALDH2 were purified to homogeneity by established methods, as judged by SDS-PAGE followed by Coomassie Blue staining (11,18,21). The expression of the native and mutant enzymes was also verified by Western blotting with antibodies against the beef liver ALDH2. R84Q, K192Q, T384A/S, K489Q, S471A/T, E398K, and E399Q mutants were expressed at a level similar to the native ALDH2, but, R84E, K192E, and K489E mutants were present at a much lower level. The expression of the S471T mutant was similar to that of the native enzyme, but this mutant was less stable. After purification the dehydrogenase activity of S471T decreased to 50% if kept overnight while the others were stable for several days at 4°C in the presence of 0.1 mM dithiothreitol.
Kinetic Properties of the Human ALDH2 Mutants-All the mutants examined were found to have lower activity compared with the native enzyme. Replacement of conserved residues by neutral amino acids had less effect on the activity than when oppositely charged amino acids were used (Table I). No major changes in the K m for propionaldehyde was found with the mutant enzymes, when compared with the native enzyme, as was observed for other mutants we previously characterized (11,12,21). The exception was with C302S, where the K m for propionaldehyde increased several thousandfold (10). A significant variation, however, was observed when the K m for NAD ϩ was determined. In T384A, E398K, E399Q, and K489Q, the K m values obtained were within a factor of 3-8-fold higher than native enzyme. The values increased to approximately 100 and 50 times the native value for K192Q and S471A/T mutants, respectively. Other mutants, R84E/Q, T384S, and K489E had K m for NAD ϩ similar to that found with the native enzyme. However, by comparing the second order rate constant values (V max /K m(NAD ϩ ) ), it was observed that the rate constant was significantly lower with all the mutants except for R84Q, compared with the native enzyme.
NAD ϩ and NADH Binding to ALDH2 Mutants Possessing Increased K m for NAD ϩ -Earlier kinetic investigations suggested that the ALDH-catalyzed reaction followed a sequential  mechanism with NAD ϩ being the first substrate followed by aldehyde binding (13). The dissociation constant (K ia ) for NAD ϩ with ALDH2 was determined by bisubstrate kinetic analysis (11,12,21). Several concentrations of NAD ϩ and propionaldehyde were used for the analysis of the data. The kinetic constants were calculated by the Dalzeil graphical method (23). The value of K ia increased to more that 50-and 25-fold in K192Q and S471A, respectively, compared with the native enzyme (Table II). The value k cat /K m(NAD ϩ ) ) is k 1 (on velocity), which decreased 200-fold compared with the native enzyme. Similarly, k 2 , the off velocity for NAD ϩ dissociation, was also affected. The dissociation constant (K d ) for NADH was calculated by measuring the increase in fluorescence observed when NADH binds to the enzyme (22). Analysis of the NADH binding data showed a very small increase in the dissociation constant with the mutant enzymes and no difference in the binding stoichiometry. These results are similar to the previously characterized S74A and E487K mutants, where the K m and K ia for NAD ϩ were increased significantly but the binding for NADH was not affected (12,21).

Determination of the Rate-limiting
Step for the Mutant Enzymes-The mechanism of ALDH involves several intermediary steps, as shown in Fig. 1. It was observed that a pre-steady state burst of 2 mol of NADH/mol of tetrameric enzyme occurred with the recombinantly expressed human mitochondrial ALDH2 (11). In addition, the V max value was found to be dependent on the nature of the substrate (10, 24), suggesting deacylation (k 7 ) was the rate-limiting step for the native enzyme. For most mutants a burst magnitude of essentially 2 mol of NADH/mol of enzyme was found, except with K192Q and E399Q, where no burst could be measured (Table I).
It was shown that aldehydes possessing an electron withdrawing group were oxidized more rapidly than propionaldehyde by the liver enzymes (10,24). To determine the effect of a substrate having an electron withdrawing group on velocity, the V max for the oxidation of chloroacetaldehyde was measured for both native and mutant enzymes (Table I). Chloroacetaldehyde was oxidized at a much faster rate than propionaldehyde under V max conditions with the native ALDH2 and all the mutants except K192Q and E399Q. A faster rate is indicative that hydride transfer is not the rate-limiting step. Therefore, based on pre-steady state burst and increase in the rate of the reaction with the chloroacetaldehyde, it can be concluded that for R84Q, T384A, T384S, E398K, S471A, S471T, and K489Q the rate-limiting step remains deacylation (k 7 ). For both K192Q and E399Q, where no pre-steady state burst of NADH formation and a decreased rate of the reaction with chloroac-etaldehyde was found, hydride transfer appears to be the ratelimiting step (k 5 ).
Determination of Esterase Activity of the Mutant Enzymes-In addition to dehydrogenase activity, aldehyde dehydrogenase also possess esterase activity (13,25,26). The reaction scheme for the esterase reaction is shown in Fig. 1. The mechanism for both the esterase and dehydrogenase reaction remains essentially the same except that the hydride transfer step is not involved in the esterase reaction. Although the esterase activity was decreased for the mutants, a relatively higher esterase activity was found, compared with their respective dehydrogenase activity (Table III). In all mutants the esterase activity was stimulated 2-7-fold by NAD ϩ . Only Arg 84 mutants did not show any NAD ϩ stimulation. DISCUSSION When we initiated the project to elucidate the role of the conserved amino acids found in all aldehyde dehydrogenases, no structure of the enzyme was known. After we completed the mutational analysis, the structure of the dimeric class three enzyme (27) and the beef liver mitochondria enzyme (3) were determined in the presence and absence of NAD ϩ . From the latter structure, it is apparent that only residues Lys 192 and Glu 399 are in contact with the coenzyme, and that none of the other conserved residues make contact with the coenzyme or other components of the active site. Fig. 2 illustrates the location of each conserved residue in one subunit as well as a few others we found to affect activity. The enzyme is actually a pair of dimers, as shown elsewhere (3).
All the residues investigated were conserved or at least conserved in all mammalian ALDHs. Mutation of Ser 471 to a threonine reduced the k cat to less than 5% of the value of the native enzyme. Converting the residue to an alanine produced an enzyme which retained 15% of the specific activity. Most surprisingly was to find that mutation of the two residues that interacted directly with the coenzyme did not more drastically affect the specific activity of the enzyme. In fact, the K m for NAD ϩ was increased only when Lys 192 was converted to a glutamine (100-fold). Mutation of the other ribose binding ligand, Glu 399 , caused the K m for NAD ϩ to increase just 4-fold. In contrast, Ser 471 , which is located approximately 9 Å from the coenzyme, caused the K m for NAD ϩ to increase over 50-fold. The effect is related to binding, for the K ia values for NAD ϩ to K192Q and S471A increased over 50-and 25-fold, respectively, as shown in Table II. While the binding of NAD ϩ was affected by mutations to Lys 192 and Ser 471 , the interaction of NADH was not. Its value increased only 3-and 6-fold, respectively.
Although none of the charged residues investigated in this study can be deemed essential, changing a residue to one with  K192Q, and S471A mutant forms of human liver mitochondrial aldehyde dehydrogenase (ALDH2) K a is the K m for NAD ϩ ; K b is the K m for propionaldehyde; k 1 and k 2 refer to on and off velocity of NAD ϩ , respectively. k 3 is the rate constant for acylation; K ia represents the dissociation constant for NAD ϩ . k 1 was calculated from k cat /K a , and k 2 was calculated from K ia k 1  the opposite charge greatly affected the enzyme activity. This can be illustrated by the fact that K489E had only 3% the specific activity of the native enzyme while K489Q had nearly 40%. Similarly, R84E had just 2% activity while the Gln mutant had 32%. The reversal of the charge did not change the K m for NAD ϩ or aldehyde substantially. This was in contrast to what we found with E487K/Q (21), where the Lys mutant possessed a very high K m for NAD ϩ while the Gln mutant had native like K m values. Neither Glu 487 or Lys 489 are completely conserved, but a negative and a positive charge, respectively, are always found at these positions.
We can now offer some structural arguments for the effects of the various mutants we investigated. Prior to this study, we reported that Cys 302 is the active site nucleophile (10) and Glu 268 is the general base (11). It is possible that a water attached to Glu 268 may abstract the proton from Cys 302 (3). We previously showed that mutations to Ser 74 drastically affected the K m for NAD ϩ , while it did not affect K d for NADH (12). Ser 74 is located near the dimer interface and is associated with residues 69 and 71 which make contact between the pair of dimers. Similarly, Glu 487 makes contact with arginine residues in its own subunit as well as in the other dimer pair (3). When this residue was mutated to a lysine to mimic what has been found in many Oriental people (21), the K m for NAD ϩ increased and the specific activity decreased. We argued that disrupting the glutamate-arginine salt bonds was responsible for the observed effect. Lys 489 also binds to the opposed subunit in the dimer pair, capping the C terminus of ␣ helix between residues 436 and 445. Thus, its role appears to be in stabilizing subunit interactions (3).
Residue 384 (threonine) prefers to have a hydroxyl as the side chain in that the serine mutant is more native-like than is the alanine. This residue is located near the solvent surface but near a hydrophobic core that extends toward the coenzyme binding pocket. Potentially, of more important is that the hydroxyl group of Thr 384 binds to the carbonyl back bone of Pro 383 . This interaction appears to maintain the local structure of the conserved Pro 383 and Thr 384 residues found in all AL-DHs. Mutation to residue 471 (serine), as mentioned above, affects the K m for NAD ϩ . It is located such that it interacts with residues 269 and 270. Conceivably a disruption of that interaction would alter the position of Glu 268 , the conserved general base. This disruption could explain the decrease in specific activity. From the structure it appears that Glu 268 might have to move to accommodate the binding of NAD ϩ . The calculated k 1 term decreased 1500-fold in the S471A mutant and K ia increased 15-fold, consistent with the argument that an alteration near 268 could affect NAD ϩ interactions with the enzyme for Ile 269 is in van der Waal contact with the nicotinamide ring.
Arg 84 is part of a long helix between residues 81 and 110 and interacts through a water molecule with the C-terminal Ser 500 in a different subunit. Furthermore, it serves to cap a helix in its own subunit by binding to the peptide carbonyls of residues 183 and 184. Removing the charge did not alter drastically the catalytic properties. A negative charge at this position affected k cat without increasing K m for either substrate. The Arg 84 containing helix is located near the subunit interface. Lower concentrations of R84E and K489E were found after expressing the mutants in E. coli. Mutation to these residues could have affected assembly. It is not apparent why lower levels of K192E were also found. This residue appears to be exposed to solvent and might not be expected to affect assembly or stability. It will be necessary to study the stability of the mutants to determine if the point mutations did actually affect assembly. Both residues 192 and 399 bind to ribose hydroxyl groups and appear to be the only conserved residues that directly interacts with the coenzyme. It is surprising that mutation to only one drastically alters NAD ϩ binding. However, the nicotinamide ring appears to have to move during the catalytic process (3) and since Glu 399 is bound to the nicotinamide ribose, it may not be unexpected that this residue should be less important in the binding of NAD ϩ . The major affect of mutating Glu 399 is a change in the rate-limiting step, while mutating Lys 192 causes a change in both NAD ϩ binding and the rate-limiting step. Apparently the proper anchoring of the NAD ϩ is essential for the hydride transfer step.
Our laboratory has argued that since substrates with an electron withdrawing group, such as chloroacetaldehyde, are oxidized more rapidly than propionaldehyde, and a pre-steady state burst was observed, deacylation, k 7 in Fig. 1, is ratelimiting for mitochondrial aldehyde dehydrogenase. Only with K192Q and E399Q were these properties not found. For those,  2. The structure of one subunit of beef liver mitochondrial ALDH. The various conserved and partially conserved residues with a reactive side chain are shown along with NAD ϩ . Lys 489 caps an ␣ helix in another subunit, indicated by a. Glu 487 binds to an arginine in that subunit, indicated by b.
the rate of oxidation of chloroacetaldehyde was not greater than that of propionaldehyde, showing that the rate-limiting step did not involve the attack of the nucleophile, k 3 or k 7 . Further, as no burst was found, k 7 and k 9 were eliminated as the rate-limiting step. Therefore, we conclude that the ratelimiting step was changed to hydride transfer (k 5 ) for these two mutants. Detailed studies presented in the following paper will show that there is a primary isotope affect on aldehyde oxidation, verifying the fact that hydride transfer became rate-limiting for K192Q and E399Q mutants (15).
The term k cat /K m(propionaldehyde) is related to aldehyde interaction with the enzyme-NAD ϩ complex (10).
If it is assumed that the value of k 4 is small, due to the tight binding of aldehyde to enzyme-NAD ϩ complex, then k cat / K m(propionaldehyde) approximates to the value of k 3 . Mutations of native amino acids decreased this value by 2-40-fold. Mitochondrial ALDH can hydrolyze activated esters such as p-nitrophenylacetate; presumably, both esterase and dehydrogenase reaction take place at the same active site (10 -12, 28). All the mutants produced in this study still hydrolyzed the ester. The esterase reaction involves many of the same steps that the dehydrogenase reaction use, except it does not include the hydride transfer step (k 5 ) (Fig. 1). The acylation step (k 3 ) is thought to be the rate-limiting step for the esterase reaction with horse (14) and human liver ALDH2 (11). This decreased esterase activity was found with the mutants that had a change in k 7 and those that produced a change in k 5 of the dehydrogenase reaction.
The formation of the tetrahedral intermediate in the esterase reaction should be similar to that of the hemiacetal formation in the dehydrogenase reaction. Decrease in the esterase activity, irrespective of which step was involved in the ratelimiting step of the dehydrogenase reaction, indicates that the mutation also affected the k 3 and kЈ 3 steps, which are common to both esterase and dehydrogenase reaction (Reactions 1 and 2).
A correlation (r 2 Ϸ 0.74) existed between the calculated k 3 values for dehydrogenase and the esterase velocity as shown in Fig. 3. The esterase activity of K192Q enzyme appeared to be higher than expected. This might be related to the fact that this mutant had higher activity with the aromatic substrates, as will be discussed in the accompanying paper (15). For the esterase reaction an additional, nonhydride transfer step (kЈ 5 ) is required to form the acyl intermediate. Conceivably the mutations could affect this step as well as the kЈ 3 step needed to form the tetrahedral intermediate.
This study showed that of the conserved residues with reactive side chains, other than Cys 302 and Glu 268 , only Lys 192 and Glu 399 could be considered to be involved in the active site of the enzyme. The other residues appear to make long ranged contacts with residues which make it possible for the catalytic residues to be properly aligned, analogous to Asp 194 making a salt bond with Ile 16 in chymotrypsin (29). The two ribose-binding residues were essential in catalysis in that the rate-limiting step for the enzyme changed when these were mutated. However, mutation of any completely or partially conserved residue appeared to affect more than one step in the catalytic reaction. Preliminary results on the mutational analysis of the conserved Glu 377 in fatty aldehyde dehydrogenase from luminescent bacteria (Vibrio harveyi) were reported to be similar to the corresponding Glu 399 in mitochondrial ALDH2 (30). A more detailed analysis of the precise involvement of the Lys 192 and Glu 399 will be presented in the following paper (15).