Computer‐graphics interpretations of residue exchanges between the α, β and γ subunits of human‐liver alcohol dehydrogenase class I isozymes

Abstract
Three-dimensional models of human alcohol dehydrogenase subunits have been constructed, based on the homologous horse enzyme, with computer graphics. All types of class I subunits (.alpha.,.beta., and .gamma.) and the major allelic variants (.beta.1/.beta.2 and .gamma.1/.gamma.2) have been studied. Residue differences between the E-type subunit of the horse enzyme and any of the subunits of the human isozymes occur at 64 positions, about half of which are isozyme-specific. About two thirds of the substitutions are at the surface and all differences can be accommodated in highly conserved three-dimemsional structures. The model of the .gamma. isozyme is most similar to the crystallographically analyzed horse liver E-type alcohol dehydrogenase, and has all the functional residues identical to those of the E subunit except for one which is slightly smaller: Val-141 in the substrate pocket. The residues involved coenzyme binding are generally conserved between the horse enzyme and the .alpha., .beta., and .gamma. types of the human enzyme. In contrast, single exchanges of these residues are the ones involved in the major allelic differences (.beta.1 versus .beta.2 and versus .gamma.2), which affects the overall rate of alcohol oxidation since NADH dissociation is the rate-determining step. Residue 47 is His in .beta.2 and Arg in the .beta.1, .gamma.1, and .gamma.2 subunits, and in horse liver alcohol dehydrogenase. Both His and Arg can make a hydrogen bond to a phosphate oxygen atom of NAD; hence the lower turnover rate of .beta.1 apparently derives from a charge effect. The substitution to Gly in the .alpha. subunit results in one less hydrogen bond in NAD binding, and consequently in rapid dissociation. This may explain why the overall rate is an order of magnitude faster than that of .beta.1. The important differences between .gamma.1 and .gamma.2 is an exhange at position 271 from Arg to Gln which can give a hydrogen bond from Gln in .gamma.2 to the adenine of NAD. The tighter binding to .gamma.2 can account for the slower overall catalytic rate in this isozyme. The kinetics of interactions of cyclohexanol and benzyl alcohol with the isozymes were judged by docking experiments using an interactive fitting program. Most of the differences in Km values for alcohols of the .beta. and .gamma. subunits could be related to one amino acid difference, i.e. Thr-48 in the .beta. subunit, compared with Ser in the .gamma. and the horse liver enzyme subunits. The substrate pocket of our model of the .alpha. subunit differs most markedly both from those of the other subunits of human isozymes and from that of the horse enzyme. Substitution of Ala-93 in .alpha. for the Phe in the others makes the substrate cleft much wider close to the active-site zinc atom and its ligand Cys-174. Our model of the .alpha. subunit suggests that the substrate specificity of isozymes with this subunit differs significantly from that of other isozyme subunits.