Coenzyme interaction with horse liver alcohol dehydrogenase. Evidence for allosteric coenzyme binding sites from thermodynamic equilibrium studies.

The techniques of fluorescence enhancement, fluorescence quenching, fluorescence polarization, and equilibrium dialysis are utilized to study the binding properties of coenzyme to horse liver alcohol dehydrogenase. Polarization of fluorescence and equilibrium dialysis show that NADH binds to alcohol dehydrogenase with a stoichiometry of 6 mol per mol of enzyme, in contrast to the value of 2 determined from fluorescence enhancement measurements. NAD+ also binds with a stoichiometry of six as was determined by equilibrium dialysis. The two NADH sites which bind coenzyme more tightly and which are revealed by fluorescence enhancement measurements are designated the catalytic sites. Binding of coenzyme to the four ancillary sites does not alter the quantum yield of NADH but results in a 20% contribution to quenching of enzyme's tryptophan fluorescence. From the emission anisotropy of bound NADH of 24.0% for the additional sites and 28.1% for the catalytic sites and their relative fluorescence lifetimes at the same wavelengths of excitation and emmision, we conclude that the nicotinamide ring of NADH bound to the additional sites exhibits a freedom of motion independent of the macromolecule, while that bound to the catalytic sites is more rigidly held. Polarization of fluorescence yields negative intrinsic free energies of 9.2 and 7.5 Cal M-1 for NADH interaction with the catalytic and additional sites, respectively. Although these values are 1.3 to 2.0 Cal higher than those determined by fluorescence quenching and equilibrium dialysis, the mean Hill coefficient of 1.76 plus or minus 0.06, the titration span of 2.4 logarithmic units and coupling free energies (in magnitude and sign) are the same for all these techniques. The above difference in the intrinsic free energies are attributed largely to the different modes of interaction of excited and unexcited NADH molecules with alcohol dehydrogenase.

unit, nicotinamide adenine dinucleotide-dependent enzyme that catalyzes the interconversion of alcohols and aldehydes. Coenzyme interaction properties of this enzyme have been investigated previously by various authors (l-3).
By the technique of enhancement of fluorescence of bound NADH it was shown (1) that the stoichiometry of binding is two, as might be anticipated for a protein with two identical subunits (4). In the course of comparative studies of the zinc-free alcohol dehydrogenase and native enzyme, it was observed that the latter possesses additional coenzyme binding sites (5-7).
In this paper, we report coenzyme binding properties of alcohol dehydrogenase by the techniques of fluorescence enhancement, fluorescence quenching, fluorescence polarization, and equilibrium dialysis.
Our results indicate differences in and the complementary nature of these techniques.
Kinetic evidence shows that the binding of coenzyme to the allosteric sites inhibits catalysis (5). [3HlNADH was precipitated from this solution by the method of Lehninger (9 an intrinsic dissociation constant of 0.32 FM from the midpoint of the titration.
Although most of the data in Fig. 2A are confined to a region where J > 1, it can be assumed that the formation curve is symmetrical about the midpoint.
A span of 1.9 logarithmic units between P = 0.2 and 1.8 and the order of binding of 1.0 from the Hill plot (Fig. 2B) indicate a simple type of binding and that the sites arc identical in agreement with previous findings (2, 3). The synthesized [3H]NADH, which contains only 0.1% radioactive coenzyme, is bound with approximately the same dissociation constant as the commercial NADH.
Coenzyme Interaction with Alcohol Dehydrogenase by Polarization of Fluorescence and Equilibrium Dialysis- Fig.  3A shows the fluorescence polarization curve as concentrated NADH solution was added to a fixed concentration of alcohol dehydrogenase and Fig. 3B is a replot of the data with the ordinate in logarithmic form.' The solid cwws in both plots represent the theoretical computer-generated titration curves with the assumption of two identical and equivalent sites for NADH, an intrinsic dissociation constant of 0.19 PM and an emission anisotropy of 28.1% for bound enzyme. The logarithmic plot of the experimental data clearly indicates that there are two types of sites exhibiting different AI, values for NADH bound to alcohol dehydrogenase.
The values of the emission anisotropy of bound coenzymes can be calculated from the data presented in Fig. 3. From the upper ordinate intercept of the logarithmic plot, the fluorescence emission anisotropy of coenzyme bound to the sites that bind coenzyme more tightly can be calculated to be 28.1 y. in excellent agreement with that determined by titrating an excess of enzyme against a fixed concent,ration of NADH at the same wavelengths of excitation (340 nm) and emission (430 nm). The conditions are those described in Fig. 1. v is the ratio of molar concentration of bound ligand to molar concentration of enzyme. The ordinate is the negative of logarithm of the concentration of free NADH. R, Hill plot of the data in A. n = 2, the maximum value for v in A. The graph is a least square fit of the data. A* and A'6 are the respective values for the emission anisotropies of fluorescence of coenzyme bound to the first and second types of sites; nr and n2 are the corresponding number of coenzyme binding sites and nt is the total number of binding sites. A'b, which then turns out to be the only unknown parameter in Equation 5, can be evaluated. Fig. 4 presents the Bjerrum formation curves for the binding of NADH by polarization of fluorescence and equilibrium dialysis and of NAD+ by equilibrium dialysis. Each curve indicates a stoichiometry of 6 mol of coenzyme per mol of enzyme and can be resolved, albeit incompletely, into two titration curves containing two (nr) and four (nz) sites and corresponding dissociation constants, &r and Kd2, respectively.
For a complete resolution  I  I  I  I  I  8  TABLE   I  a, binding of NASH by polarization of fluorescence as was described in Fig. 3. The branch curve (broken line) is that for the uncorrected data computed with Equation 3. 6, [sH]NADH interaction with 2.0 MM enzyme by equilibrium dialysis at 4" in pH 7.5 phosphate buffer, p = 0.05. c, [3H]-NAD+ interaction with 2.0 pM enzyme by equilibrium dialysis at 4" in pH 7.5 phosphate buffer, p = 0.05. parallel to that of the free coenzyme, a fact which indicates (a) the equivalence of the quantum yield of the free coenzyme and that bound to the additional sites, (b) saturation of the additional sites does not affect binding to the sites that enhance coenzyme fluorescence.
The thermodynamic and other useful parameters obtained from Fig. 4 are summarized in Table I Fi/F = (1 + KAQIQl)(l + k,gJQl) (7) Fi and r,, are, respectively, the fluorescence yield and lifetime of t,he fluorophore in the absence of quencher; k, is the collision rate constant of the fluorophore and quencher and K Ac is the equilibrium constant for complex formation by the unexcited A and Q should be obtained. Fig. 5 gives the result of quenching by coenzyme of enzyme's tryptophan fluorescence. For either NADH or NAD+ as a quencher the curve is biphasic.
The concentration of NADH at the point at which linearity breaks down is that required to saturate all the sites that can enhance the fluorescence of bound NADH (Fig. 1). The second phase of each curve is therefore due to binding of coenzyme to the additional sites. From the breaking points in the curves 20% of the total quenching is calculated to be due to the binding of coenzyme to the  (20) gives kq = 1O-3 4noDN (9) N being the Avogadro's number, u the sum of the radii of the macromolecule and quencher and D the sum of their diffusion coefficients.
The diffusion coefficient of alcohol dehydrogenase was taken as 6.1 x lo-' cm2 s-r (21) and its radius as 3.62 x 1OV cm. The latter was calculated with Equation 10 r, = p;T 1 (10) 6NnDi n where ri and Di are, respectively, the radius and diffusion coefficient of a solute species i, R is gas constant, 2' and 7 are temperature and viscosity of solvent, respectively. From a molecular model built to scale and on the assumption of a folded conformation for coenzyme (22,23), the coenzyme has a molecular diameter of 12 A and a height of 5 A. This allows the molecular volume and hence the radius of an equivalent sphere to be computed as 5.1 x lo-* cm. Its diffusion coefficient (4.28 X 1OP cm2 s-1) was also computed by the use of Equation 10. The value of k, was calculated to be 2.55 x lo9 M-I s-i. If we assume that r0 = 6.0 x 1OV' s, the maximum experimental value reported for indole in some solvents (24), kq70 = 15.1 M-l.
It can be seen that kqrO << KQ values in Table II, a fact which allows the approximation to be made that KQ = KAQ.

Molecular Weight
Determinations--A molecular weight of 80,000 f 5,000 was obtained in the presence of low and high coenzyme concentrations, indicating no dissociation or association of enzyme dimer was caused by the binding of coenzyme.

DISCUSSION
The results presented in this paper indicate that while the technique of enhancement of fluorescence of bound coenzyme establishes a binding stoichiometry of 2 mol of coenzyme per mol of alcohol dehydrogenase as was previously reported (a), polarization of fluorescence and equilibrium dialysis techniques indicat)e a stoichiometry of 6 mol of coenzyme per mol of enzyme. The conclusion from electron spin resonance (25) and nuclear magnetic resonance (26) spectroscopies that 7 to 8 mol of a spinlabeled analog of coenzyme binds per mol of alcohol dehydrogenase is comparable to the above.
The use of absorption and fluorescence spectroscopies in determining the stoichiometric number of ligand binding sites in macromolecules is based on the implicit assumption that every ligand bound has its absorption, or fluorescence spectrum or intensity, or both, perturbed.
The reliability of the fluorometric technique in determining stoichiometry for dehydrogenases has been questioned formerly (27), since all the binding sites in some pyridine nucleotide-dependent dehydrogenases may not perturb the absorption or fluorescence spectrum of the coenzyme (to the same extent).
The earlier discrepancy as to whether n-glyceraldehyde 3-phosphate dehydrogenase possesses three or four NADf binding sites, as was determined by the perturbation of the absorption spectrum of the coenzyme, has been ascribed to the inability of the fourth subunit to alter the spectrum of the bound coenzyme (28). The limitation of the perturbation technique in determining the total number of binding sites for a ligand is also exemplified in the binding of a drug, a pyrazolidinedione analog, to human serum albumin.
This protein has been shown by fluorescence technique to have one binding site for a pyrazolidinedione analog, while equilibrium dialysis indicated two strongly binding and four weakly binding sites (29). A similar disparity is exhibited in the binding of dicoumarol to human serum albumin (30). The revelation of additional coenzyme binding sites in alcohol dehydrogenase by polarization of fluorescence and by equilibrium dialysis, techniques which do not necessarily impose the requirement that the chromophoric group in a ligand be perturbed, further underscores this limitation.
Yet the results yielded by the fluorescence technique have been utilized to complement those of fluorescence polarization to allow simple but meaningful interpretations to be made. It is in this perspecitve that we emphasize the usefulness of these techniques.
The Bjerrum formation curves (Fig. 4) can be given two alternative interpretations: (a) there exist two independent types of sites with overlapping titration curves; (b) the binding of coenzyme to the second type of site is conditional upon binding to the first type of site, the conditional or coupling free energies (A&) being -1.7 and -1.3 Cal M-' for NADH and NAD+, respectively.
The sign of the free energies indicate cooperativity of the first and second types of sites in their saturation by coenzymes. No distinctive choice can as yet be made between these two interpretations, which by themselves arc oversimplified as each type of site contains more than one coenzyme binding site. A more complex system could be envisaged if binding is considered in terms of statistical rather than intrinsic free energies. Approaches to the threoretical estimates of the statistical free energies for relatively simple systems and the compatibility of such estimates with experimental data have been described (31)(32)(33).
The constancy of the molecular weight at high and low coenzyme concentrations excludes the possibility that the observed cooperativity, as is indicated by the Hill coefficient of 1.7 for the second type of sites, is due to dissociation of subunits and subsequent creation of new sites, a condition that would have amounted to the phenomenon called relaxation effect. The conclusion is therefore apparent that the generalization that in dehydrogenases one coenzyme is bound per subunit (34) cannot be extended to liver alcohol dehydrogenase.
It has been well established that alcohol dehydrogenase has two catalytic sites for the oxidation of ethanol (35). The two coenzyme binding sites, as was determined by fluorescence enhancement, therefore, have been equated logically with the catalytic sites. The finding that horse liver alcohol dehydrogenase possesses additional coenzyme binding sites poses the problem of which really are the catalytic sites. We have shown previously that in the inactive zinc-free alcohol dehydrogenase only two coenzyme binding sites are detectable by the technique of fluorescence polarization and equilibrium dialysis (6,8). These sites were shown to be equivalent in the binding of NAD+ and have the same intrinsic dissociation constant as the two "tight" binding sites in the native enzyme. In addition a mole of the zinc-free enzyme-coenzyme complex binds either 2 mol of substrates or substrate analogs to form ternary complexes. The above two facts suggest that the two sites in the zinc-free enzyme and the corresponding "tight" binding sites in the native enzyme are indeed the active sites.
The equivalence of the quantum yield of the free coenzyme and that bound to the additional sites and lack of perturbation of the absorption spectrum of such bound coenzyme lead to t.he conclusion that the fluorescence lifetimes and the environment of the fluorophore are the same, and that possibly the nicotinamide ring is hanging out from the macromolecule into solution. A theoretical estimate of the emission anisotropy of such bound coenzyme can be made from the ratio of the emission anisotropies of the bound and free coenzymes.
Al, ~1, pl are, respectively, emission anisotropy, lifetime of the excited state, and mean rotational relaxation time of coenzyme bound to the additional sites, and AZ, 72, p2 are those for unbound coenzyme. pl was calculated to be 6.5 x lo-* s from Debye relationship (36) on the assumption of a partial specific volume of 0.75 for alcohol dehydrogenase Equation  12 represents the Debye relationship : A value of 4.1 x lo-i0 s was computed for pZ from the relation 3 p2 =%Fr being the molecular radius of NADH as a sphere and the other parameters have the meanings described above; r1 = ~2 = 3.8 x lo-i0 s from Spencer and Weber (37). The computed value for A'b is 30.8%. The experimental value of 24.0% indicates some degree of freedom of motion of the nicotinamide ring of coenzyme bound to allosteric sites. Thus the shortness of the lifetime and the assumption of the rotary diffusion constant of alcohol dehydrogenase provide adequate explanation for the high degree of polarization of NADH bound to the allosteric sites. The higher emission anisotropy of 28.1'%, found for NADH that is bound to the catalytic sites and whose fluorescence lifetime is several times larger, leads to the conclusion that the nicotinamide ring is more rigidly held in the catalytic sites.
We have previously presented extensive discussion on the possible mechanism of quenching of enzyme's tryptophan fluorescence by coenzyme (10). Since quenching could be due to the different modes of energy transfer or to a conformational change of the enzyme, and since energy transfer from tryptophan to NAD+ is ruled out, the observed 20% quenching of tryptophan fluorescence by coenzyme bound to the additional sites can be ascribed to a conformational change caused by such bound coenzyme. By definition, KAQ is the equilibrium constant for complex formation between alcohol dehydrogenase and coenzyme in their ground state configurations.
A comparison of Kd and corresponding KQP1 in Tables I and II indicates that for NADH Ka obtained by fluorescence and fluorescence polarization are smaller than KQwl. We therefore conclude that the interaction free energy between alcohol dehydrogenase and the excited NADH molecule differs from that of the unexcited molecule by 1.2 to 2.0 Cal. We attribute this difference to the different conformations and electron distribution among atoms of the nicotinamide ring in the excited and ground states of this molecule.
The nicotinamide ring in the ground state assumes essentially a puckered conformation with an out-of-plane amide group, while in the excited state the ring and the amide group are to some extent planar.
The compatibility of Kovl with corresponding Kd values determined by kinetics of oxidation-reduction of alcohols and coenzymes (5, 35) and by equilibrium dialysis, techniques which measure equilibria of the unexcited molecules, offers a reasonable explanation for why there are in dehydrogenases in general differences between kinetically determined dissociation constants and those determined by fluorescence enhancement. APPENDIX Assume that there are in alcohol dehydrogenase two types of sites differentiable by their capabilities to bind NADH with the enhancement of the coenzyme fluorescence by factors of X and X' and with emission anisotropies, Ab and A'b, respectively. Then Equations 14 and 15 can be written to relate the fraction of coenzyme bound in terms of fluorescence and emission anisotropy, respectively. where o( and LY' are the fractions of coenzyme bound to the two types of sites, respectively, at any point in the titration and F and Ff are the fluorescence intensities of the coenzyme in the presence and absence of enzyme, respectively. A, is the emission anisotropy of free NADH, and A is the measured emission anisotropy representing the average of the emission anisotropies of all bound and free forms weighted according to their fractional intensities.
For any species, i, in solution, the fluorescence intensity Fi = Flli + 2P1;. When X' = 1, as is the case for the additional sites in alcohol dehydrogenase (see Fig. I