Human alpha 1-acid glycoprotein (AAG), a plasma drug transport protein, has three main genetic variants, the A variant and the F1 and S variants, which are encoded by two different genes. The binding of disopyramide, methadone, dipyridamole, chlorpromazine, lignocaine and progesterone to the two main gene products of AAG-the A variant and a mixture of the F1 and S variants (60% F1 and 40% S)-separated by chromatography from native commercial AAG, a mixture of almost equal proportions of the F1, S and A variants, was studied by equilibrium dialysis. A selective binding of disopyramide and methadone to the A variant and a preferential binding of dipyridamole to the F1S variant mixture were found. Lignocaine and chlorpromazine had a slight preference for binding to the A variant and to the F1S mixture, respectively, but progesterone showed no selectivity with regard to any of the variants of AAG. The differences in drug-binding demonstrated between the A variant and the F1S mixture confirmed those of a previous study, in which a selective binding of imipramine to the A variant and of warfarin and mifepristone to the F1S mixture have been found. These results indicate specific drug transport roles for each AAG variant, according to its separate genetic origin. The results of control binding experiments performed with (unfractionated) commercial AAG and the series of tested ligands concurred with that for the separate AAG variants, with respect to the proportion of the A variant (27%) and that of the F1 and S variants (73%) in the commercial protein. In addition, disopyramide, methadone, dipyridamole, chlorpromazine, lignocaine and progesterone were used in equilibrium dialysis displacement experiments to study interactions on binding sites labelled with imipramine for the A variant and with warfarin for the F1S variant mixture. The four latter ligands were found to competitively inhibit the binding of warfarin to the F1S variant mixture and all of them that of imipramine to the A variant. The ligands association constants to each AAG variant obtained from such inhibitory experiments were comparable to those determined in the direct binding studies. As the stochlometry of the interactions of the A variant and the F1S variants, respectively, with their specific ligands was approximately one (1), it was concluded that these ligands bind to each of these variants via a single common binding site. These results indicate that the AAG molecule would have for its ligands at least two separate binding sites, showing different specificity and localization, and not one site, as it is generally assumed. The possible pharmacological and clinical consequences of the binding results with the separate AAG variants are discussed.