Alcohol Dehydrogenase

Before I go over the mechanism of action of alcohol dehydrogenase, I'd like to describe the various types of enzyme which have been called alcohol dehydrogenases. Table 1 in the handout shows the major classes. What I tend to refer to as "classical" alcohol dehydrogenases are in this table denoted "medium chain", with subunit polypeptide around 375 amino acids, molecular weight about 40,000. The defining enzyme of the class is horse liver alcohol dehydrogenase, which is the main enzyme I shall talk about. It and related enzymes are dimeric and contain two zinc atoms per subunit, one essential for catalysis; the other is termed "structural", which is biochemistryese for "we don't know why it's there". However, the sequence of cysteine residues which bind that zinc atom in a loop, cys-97, 100, 103 and 111 in horse liver ADH, is very recognizable and immediately identifies a dehydrogenase as belonging to this class.
Yeast alcohol dehydrogenase is the next most studied. In yeast there are four different isozymes - the normal fermentative isozyme I, isozyme II which is important when it's growing on alcohol, isozyme III in mitochondria, and isozyme IV which is expressed only when the other three have been knocked out and which shows considerable homology to an ADH from the bacterium Zymomonas mobilis. The first three yeast enzymes are tetramers, subunit mol. wt. about 35,000, one zinc per subunit; they are about 20% homologous to liver ADH. There are other sequence-related proteins which have no zinc and even no known enzymatic activity - the crystallins, proteins of the lens of the eye, seem to be derived from various enzymes but be present in the lens and vitreous humor mainly to give the right refractive index.
Another major class is the short-chain dehydrogenases, about 250 amino acids long and containing no zinc. The first enzyme of the class to be described was the alcohol dehydrogenase of Drosophila, useful in eating fermenting fruit; but apart from other closely related insect alcohol dehydrogenases no other alcohol dehydrogenases, active on ethanol and other alkanols, have been found in this class, sometimes wrongly called the short chain ADH class.
Two other classes are so far restricted to microorganisms. Long-chain ADHs are 600 to 750 amino acids long, and use pyrroloquinoline quinone as a cofactor. Iron-activated ADHs include 1,2-propanediol dehydrogenase of E. coli, ADH II of the bacterium Zymomonas mobilis, which ferments glucose and sucrose to ethanol by the Entner-Doudoroff pathway, and ADH IV of yeast.
The second table lists dehydrogenases found in plants, mostly medium-chain, zinc-containing ADHs, though a scattering of short-chain dehydrogenases, or at least gene products homologous to short-chain dehydrogenases, have also been found. While the "classical", ethanol-active ADHs of plants, which act when the plant or its roots are under anaerobic conditions, are quite homologous to horse liver ADH - 52% identical in sequence, only two or three insertions of one amino acid plus a couple more at the C-terminus - they are believed by Jörnvall not to be directly derived from them, but to share a common ancestor in what were originally, in animals, called cc alcohol dehydrogenases or class III ADHs, but act on ethanol only at very high concentrations and are now recognized to have as principal substrate hydroxymethylglutathione, the thiohemiacetal of formaldehyde with glutathione, and thus in effect are formaldehyde dehydrogenases with formate as a product. The cinnamyl alcohol dehydrogenases designated Cad2 are related zinc medium-chain dehydrogenases, which use NADP+ as cofactor and are important in lignin biosynthesis. Cad1, which happened to be the first isozyme off a column when Boudet's group in Toulouse were purifying, is quite unrelated, monomeric, and belongs to a completely different group of dehydrogenases. Another group of dehydrogenases which seem to be most active on benzyl alcohols are designated Bad1; they are found in plants, and related enzymes are found in bacteria. They too are medium chain, dimeric enzymes with the signature zinc-binding cysteine residues. There are also NADP+ dependent alcohol dehydrogenases in brain and kidney reducing aldehydes generated from neurotransmitter amines, and also glucuronate; I haven't checked whether their sequences are related to any of the plant enzymes. They are often called aldose reductases.
Liver alcohol dehydrogenase, catalyzing the oxidation of alcohols to aldehydes with reduction of NAD+ to NADH - functionally, the reaction is usually in the other direction, aldehyde reduction, except when we go out drinking - is one of the most studied enzymes. This study has centered on ADH from horse liver, HLADH for short. When Professor and Nobel Laureate Hugo Theorell was here visiting Dr. Pietruszko in the Center for Alcohol Studies a few years ago, a bold graduate student asked him why he focused on horse liver ADH. Dr. Theorell replied, "When I began this work in Sweden in the late 1930s, there were not so many automobiles, and there were still very many horses." A large part of the work on this enzyme has continued to come from Sweden, by Brandén, Eklund, Jörnvall and their co-workers, including that graduate student John Hempel who went on to a post-doc with Jörnvall and sequenced human liver ADHs, of which there are 3 isozymes. (While I'm blowing the Rutgers horn, I'll mention that Joe Shore, who as you can see in the list of references has done a lot concerning the mechanism, is also a Rutgers Ph.D., with the late Dr. Walter Wainio.) Rat and mouse ADHs have also been sequenced, and a number of plant ADHs have been sequenced at the cDNA and/or genomic level.
As with all pyridine nucleotide-dependent dehydrogenases, the reaction is sequential. The liver ADH kinetic mechanism is highly ordered, coenzyme first on and last off; yeast ADH is largely random in mechanism, at least in the cuvette. The actual chemical reaction is termed hydride transfer, because formally it is transfer of a negatively charged hydrogen atom, although it is not clear exactly whether the atom and the charge are transferred together. In yeast ADH this step is largely rate-limiting for the overall reaction, especially with aromatic alcohols, where the effect of substituents on the benzene ring can be examined.
The mechanism of LADH was originally stated by Theorell and Chance in 1951 to have no significant central complexes, alcohol on, hydride transfer and aldehyde dissociation occurring as one concerted step, to the point that aldehyde is a competitive product inhibitor vs. alcohol (& vice versa). This so-called Theorell-Chance mechanism is an idealization, there are appreciable though low concentrations of the central complexes even with primary alkyl alcohols, and they are much increased with secondary and aromatic alcohols. With primary alkyl alcohols the apparent rate-limiting step is coenzyme dissociation, 3 s-1 = 4.5 µmoles/ min.mg protein; with some aromatic alcohols aldehyde dissociation is rate-limiting, with secondary alcohols hydride transfer is largely rate-limiting. With primary alcohols there is substrate inhibition due to formation of an E.NADH.alcohol complex which dissociates NADH more slowly than the binary E.NADH complex; with cyclohexanol such a complex forms but NADH dissociation is faster than from the binary complex, substrate activation occurs at high cyclohexanol concentration.
X-ray crystallography shows the enzyme to consist of a larger catalytic part and a smaller coenzyme binding part (see handout, Fig. 2); the coenzyme-binding portions are at the center of the molecule and have most of the intersubunit contacts, while the catalytic parts are the 'wings'. The gross structure, though not the sequence, of the coenzyme-binding portion is conserved among pyridine nucleotide-dependent dehydrogenases and indeed among many proteins binding adenine nucleotides; see Rossman et al. in vol. 3 of The Enzymes, 3rd ed. The triangles in the drawing indicate the positions of the introns in the maize ADH genomic sequence; it was hoped that they would indicate how enzymes have been assembled from independent domains by gene shuffling. Intron 4 does mark the division between the N-terminal catalytic portion, but intron 9 is within the small C-terminal part of the catalytic domain, not at its beginning.
The catalytic zinc atom is at the center of the structure, held by cys-46, cys-174 and his-67; the other, 'structural' zinc can be seen in a loop at the edge of the structure. Zinc-chelating compounds such as o-phenanthroline bind at the active site, competitive with both substrate and coenzyme. Iodoacetate reacts, fairly slowly, with cys-46 to inactivate the enzyme; modified enzyme, and metal-free enzyme, still binds coenzyme, but not substrates. Some other inactivators react with cys-174.
Coenzyme stretches in an open conformation from the coenzyme binding domain down toward the catalytic zinc. The adenine ring fits into a slit between ile-224 and ile-269; it makes no H bonds with the enzyme, being bound entirely hydrophobically (and many other planar aromatic compounds, such as the dye Cibacron Blue 3GA, bind there); the 6-amino group sticks out into solution. The adenosine ribose 2-hydroxyl H-bonds to asp-223, one of four invariant residues in the coenzyme-binding domain - the others are glycines at turns. The 3-OH is H-bonded to lys-228; modification of this lysine with imidate reagents increases the Vmax of the enzyme by increasing the off rate of coenzyme product, a chance observation which a contemporary of mine in graduate school has parlayed into a nice career. The pyrophosphate linkage, and inhibitors such as chloride and Pt(CN)4=, interact ionically with arg-47 and arg-369. In the human b isozyme arg-47 is a histidine, and the pH optimum is shifted because the histidine is not protonated at basic pH. The nicotinamide binds in a cleft down into the enzyme, near the catalytic zinc. One side interacts with thr-178, val-203 and val-294, the other faces the zinc. Thr-178 is conserved in known ADHs, and is in van der Waals contact with C-4 of the nicotinamide ring, where the hydride is picked up.
The substrate binds in a long narrow pocket, with the Zn at the bottom. The enzyme residues at either side closest to the catalytic site are phe-93 and ser-48; yeast and plant ADHs have a thr instead of ser-48 and are not active on secondary alcohols, but this is not a simple explanation, as two of the human isozymes also have thr but are still nearly as active on sec-butanols. Two types of inhibitors form tight ternary complexes with E.coenzyme - pyrazoles and imidazoles with E.NAD+, isobutyramide with E.NADH. Both can be used to titrate the enzyme , the former by absorbance at 290 nm, the latter by enhanced fluorescence of NADH in the complex. Pyrazole and isobutyramide are kinetically competitive with ethanol and acetaldehyde respectively. If the reaction E + NADH + aldehyde is run in presence of a high conc. of pyrazole, as soon as E.NAD+ is formed by dissociation of alcohol it binds pyrazole. allowing a single turnover to be observed. Under favorable conditions a single NADH oxidation can be observed by stopped-flow techniques; it has a rate of about 150 s-1 and a deuterium isotope effect of about 4, as expected.
A major feature of the enzyme is that a substantial conformational change occurs when coenzyme binds, as shown in Fig. 11 on the handout: the catalytic domain rotates about 10° in toward the coenzyme-binding domain, while the latter rotates about 1.5°. This held up crystallographic investigations for some time; it was realized as early as 1967 that adding NAD+ to crystals of the free enzyme caused them to fracture, because of conformational changes, and the binary and ternary complexes (using trifluoroethanol, imidazole or bromobenzyl alcohol) crystallize in a different crystal form and had to be solved independently. This conformational change is specific for LADH and closely related enzymes; it does not seem to happen for yeast ADH. It closes the catalytic site up considerably. It requires functional coenzyme; ADP- ribose, without the nicotinamide, doesn't do it. I mentioned earlier that the rate-limiting step of LADH appears to be coenzyme dissociation; actually, it is reversal of this conformational change, before coenzyme can dissociate, that is the slow step; the actual rates of association and dissociation of coenzyme, measured by nmr techniques, are much faster. When closely related ADHs differ in Vmax by a large amount, as with plant ADH1 and ADH2, the probable explanation is that the conformational change either is much faster or does not occur.
The overall reaction and the chemical mechanism are shown in Fig. 23. The reaction in the direction of alcohol oxidation requires release of a proton, which formally comes from the alcohol. In other dehydrogenases such as lactate dehydrogenase this occurs simultaneously with hydride transfer; in LADH proton release can be shown, by reaction of the proton with an indicator such as thymol blue or phenol red in stopped-flow spectrophotometry, to be faster than hydride transfer, 270 s-1 vs. 150 s-1, and unaffected by use of deuterated substrate, so it occurs before hydride transfer. Indeed, partial proton release, 0.63 H+ per E at pH 8.0 can be measured on binding of NAD+ to the enzyme, which also causes a conformational change which quenches the fluorescence and phosphorescence of a buried tryptophan, 314, perhaps due to a tyrosine OH moving closer to the trp. Proton release is completed by addition of trifluoroethanol, which is substrate-like but not oxidized.
A group with a pKa of at least 9.6 in the free enzyme - this is measured by the pH dependence of NAD+ binding, which requires the protonated form - appears to be lowered to 7.7 in E.NAD+ and still further in the ternary complex. The conventional wisdom is that this is the pKa of water bound to the zinc, which can have a pKa as low as 7.0 in carbonic anhydrase and thermolysin. However, the paper by Maret & Makinen says that the lower pKa affecting activity, pK1 - they find about 8.1 in E.NAD+ by pH dependence of kcat/KM, 6.5 in the ternary complex as measured by pH dependence of kcat - is not affected by substitution of Co++ for Zn++ at the active site, and not affected by running the reaction in D2O. On the other hand, a higher pK2, ˜ 10.7 for kcat of benzyl alcohol oxidation, 9.6 for kcat/KM, but above 11.5 for isopropanol oxidation, is lowered about 1 pH unit in the cobalt substituted enzyme and raised 0.6 to 0.8 unit in D2O. They conclude that this is the metal-bound water, and that it stays bound even when the alcohol substrate is also coordinated to the metal, a pentacoordinate complex, rather than dissociating as shown in Fig. 23 in the handout. The water would act as one step in a previously suggested proton relay system: a proton is passed from the alcohol to the water, it passes a proton on its other side to ser-48, which passes its previous proton to the 2'-OH of the nicotinamide ribose hydroxyl, which passes its previous proton to His-51, which dissociates the proton on the other side of its ring This is believed to be the group with pKa = 8.1 in E.NAD+, 6.5 in the ternary complex. This is supported by abolition of this pKa in yeast ADH in which the histidine has been mutated to glutamine and in human b1b1 ADH where this mutation occurs naturally. However, 8.1 is high for a histidine, and 9.6 in the free enzyme would be very high. Perhaps the latter pKa is actually Zn-bound water, and the histidine ionization doesn't affect NAD+ binding; Maret & Makinen didn't investigate the pKa in the free enzyme which affects NAD+.
Reduction of benzaldehyde by yeast ADH is actually faster in D2O than in H2O, kH/kD = 0.5; the explanation is that in the intermediate before the rate-limiting step (the ternary complex, since hydride transfer is rate-limiting) a species with a weak O-H bond, probably the zinc-bound water, becomes a species with a normal O-H bond, i.e. normal or free water. This suggestion doesn't integrate well with the Maret & Makinen results, but after all it's a different enzyme; it could be that water is displaced in the YADH reaction, but not in the LADH reaction. This finding should not be confused with the strong isotope effect, kH/kD = 3 to 5, seen when the substrate (NADH/D for aldehyde reduction) is deuterated.
Dunn and Hutchinson (Biochemistry12:4882 [1973]) observed that the substrate p-dimethylaminocinnamaldehyde forms an intensely colored complex with ADH in presence of NADH, which they suggest to be due to resonance stabilization of negative charge on the aldehyde oxygen by double bond shift and positive charge on the dimethylamino group, as well as by nearness to the Zn++. With NADH this has a half-life of 23 sec at pH 8.72, but in presence of 1,4,5,6-tetrahydroNAD, a good inhibitor, the colored complex is stable for months, and the crystal structure has been done. In this the substrate oxygen is unquestionably bound directly to the Zn, as earlier suggested, for instance, by the Hammett r constant of -0.85 for binding of substituted benzaldehydes to yeast ADH.NADH, and also seen in the crystal structure of the reactive complex with p-bromobenzyl alcohol and NAD+. The alcohol is believed to be bound as the alcoholate anion, having transferred its proton to Ser-48 and the charge relay system I just described. The suggestion of the above results is that this proton is lost when NAD+ binds at pH above 8, so that the system is ready to receive the alcohol's proton.
Hydride transfer then occurs, from the pro-R position of the substrate to C-4 of the NAD+, as shown in Fig. 27. The transfer is direct, in a water-free environment, as was known as long ago as 1951; it is what is called A-specific. Long before any 3D structures were available, it was realized that the two hydrogens at the 4 position in NADH are not equivalent, and that a given dehydrogenases removes only one of them. Consider your right hand as a nicotinamide, with the thumb as the CONH2 group; as you hold it out, a hydride may approach it from above or below its plane, from the back side or the palm side, and only one of these will available when the nicotinamide is bound to an enzyme. The nicotinamide binds in only one way - in LADH, the carboxamide would collide with residues 178 and 203 and the zinc if it tried to bind it the other way. The drawings make it clear that for LADH the right way has the hydride approaching from the palm side of the right hand. Originally there was no way to know absolutely which direction was which; it was simply known that there were two classes of dehydrogenases, and NADT made using a tritiated substrate would transfer the tritium to another substrate if the two enzymes were of the same class, not if they were of different classes; when the 3D structures became available they knew which was which. Glyceraldehyde-3-phosphate dehydrogenase, for example, is B-specific, and has the nicotinamide bound with the other side accessible.
The 3D structures don't tell much about hydride transfer itself, particularly since even the bromobenzyl alcohol/bromobenzaldehyde complex isn't exactly the transition state of hydride transfer. The zinc favors hydride transfer from the alcohol by stabilizing the anion, and to the aldehyde by making it electrophilic, attracting negative charge to the carbonyl O and away from the carbon which is to receive the hydride. The negative charge of the alcoholate anion is relieved by hydride transfer, since the positive charge of the pyridinium ring of NAD+ is then neutralized.
Further evidence comes from secondary isotope effects - the effect of substitution of a heavy atom in some position other than where actual bond breaking takes place. These effects are generally small, but can be measured by a technique known as equilibrium perturbation. A mixture of substrates and products at equilibrium concentrations is set up, containing a heavy atom in one substrate, but not in the corresponding molecule on the other side of the reaction. When enzyme is added, it acts on the light compound faster than on the heavy compound which would go the other way in the reaction, so that the equilibrium is temporarily displaced, until there are equal levels of heavy atom in both compounds. This method can measure isotope effects as low as 1.003.
For instance, a reaction is set up with NAD+ containing 15N in the pyridine ring, but not in the NADH present. Initially NADH would be oxidized faster than NAD+ being reduced, and the absorbance at 340 nm drops from the starting level, then increases back toward it as 15NADH accumulates and also is oxidized. The isotope effect, k14/k15, is 1.062 for NAD+ reduction, 1.018 for NADH oxidation, 1.044 on the equilibrium constant. The fact that the isotope effect is greater than 1 in both directions is taken to mean that the N must be more loosely bonded in the transition state than in either NAD+, where of course it is in a planar aromatic ring, or NADH, which still has some aromatic character since the amide carbonyl is conjugated to the 2-3 double bond. In the transition state the bond to the ribose goes down, those to C-2 and C-6 go up, yielding a boat form, which as a transition state for NAD+ has carbonium ion character at C-4, increasing reactivity there. Such a deformed ring is actually seen in the X-ray structure of dihydrofolate reductase with NADPH and methotrexate bound, the bond to the ribose is 11° out of the dihydropyridine ring plane. If only hydride transfer were sensitive to the 15N, the intrinsic effect would be 1.22, unbelievably large for a secondary effect; so they suppose that bending occurs rather at proton transfer and alkoxide formation, with 15Keq for that step about 1.07, 15k for the step about 1.06. Presumably a protein conformation change bends the coenzyme into this form, which both stabilizes the alkoxide and facilitates hydride transfer.
The rest of the reaction is then dissociation of the aldehyde, isomerization of the E.NADH complex back to the open conformation, and coenzyme dissociation. Although the conformational change is essentially the rate-limiting step for LADH acting on substrates like ethanol, I am not aware of studies as to why it is so slow. If I had succeeded in cloning tomato ADH1, which is very similar in structure to ADH2 but ten times faster acting, I would have had a good system for studying what makes the difference.
In lactate dehydrogenase, which has no zinc, the protonated form of his-195 plays the role of zinc in activating the alcohol or carbonyl; the active site is positively charged when pyruvate is bound, neutral when lactate is bound. Conformational changes occur only when the substrate binds, and the proton is lost from lactate only as the hydride transfer occurs.
The latest thing in ADH catalysis is 'tunneling', which refers to the quantum uncertainty of exactly where an atom is and the facilitation of transfer over distances of the order of magnitude of the uncertainty, without having to go through the usual transition state. This is expressed by the de Broglie equation l = ; the distance l depends inversely on the square root of the mass, so that it is longer for H, 0.63 Å, than for D and T, 0.45 Å and 0.35 Å respectively, and tunneling is more likely to occur with H than with D or T. However, it appears that the average position of the H on substrate and coenzyme isn't that close, but structural fluctuations can bring it closer, to the point where the H is aided in jumping to the other C by this quantum effect.
Tunneling is investigated by various aspects of isotope effects, expressed as kH/kD, kH/kT and kD/kT, ratios of the rates with H, D and T, either as the hydrogen being transferred - primary isotope effects - or next to the one transferred, secondary effects. It is large secondary isotope effects, larger than the effect of isotope on the overall equilibrium, which suggest tunneling.
The ratios are normally related by what are called Swain-Schaad relationships, kH/kT = (kH/kD)1.44, kH/kT = (kD/kT)3.26-3.34. When there is tunneling, the effect of substituting D or T for the secondary H is greater than expected, and the exponent, particularly in the latter case, is greater than 3.34. On the other hand, if there is what they call kinetic complexity, which simply means that some non-isotope-sensitive step is partially rate-limiting, less isotope effect is seen, and the exponent decreases below 3.26. In yeast ADH hydride transfer is purely rate-limiting, but with liver ADH benzaldehyde departure also affects the rate. Bahnson & Klinman therefore used mutants with more rapid aldehyde release to demonstrate the tunneling. The primary isotope effects kH/kT and kD/kT and the secondary kH/kT were unaffected by the mutations, but the secondary kD/kT was reduced, so that kH/kT = (kD/kT)6 to 8.5. See Table I on the last page of the handout.
Another way of looking for tunneling uses an Arrhenius plot of the temperature dependence of the isotope effect, ln(kHor D/kT) = DE*/RT + ln(AH or D/AT); the second term is the y intercept of a plot of ln(kHor D/kT) vs 1/T. If there is no tunneling plots of ln kH/kT and ln kD/kT vs 1/T will have the same intercept and AH or D/AT will be 1, but if there is tunneling the intercept will be different.