Conformational drug determinants of the sequence specificity of drug-stimulated topoisomerase II DNA cleavage.

To gain further knowledge of the molecular features of topoisomerase II inhibitors required for drug-receptor complex formation, we investigated the conformational drug determinants of the sequence specificities of drug-stimulated DNA cleavage by computer-aided molecular modeling techniques. DNA sequence specificities of bisantrene, genistein, piroxantrone and ellipticinium were determined by using simian virus 40 DNA and compared to those of mitoxantrone, 4-demethoxydaunorubicin, VM-26 and mAMSA. DNA cleavage intensity patterns of bisantrene and mAMSA were virtually identical in sequencing gels, although these drugs are of distinct chemical classes. Genistein and ellipticinium showed drug-specific DNA cleavage intensity patterns with no apparent similarity to other drugs or to each other. From 54 to 72 drug-stimulated sites were sequenced, and local base sequence specificities were established by statistical analyses. In complete agreement with mAMSA requirements, bisantrene required an adenine at position +1. Ellipticinium required a thymine and excluded a cytosine at position -1. Genistein was the only drug showing base requirements (thymines) at both positions -1 and +1. Piroxantrone (structurally related to mitoxantrone) required a pyrimidine at position -1. Since the common sequence specificity of bisantrene and mAMSA could not be simply explained by the nature of some chemical substituents, a comparative molecular modeling analysis of the drugs was carried out based on their steric and electronic attributes. Energy-minimized structures of mAMSA and bisantrene were very similar, since their planar aromatic domains and pendant side-chains overlapped to a very good approximation. In contrast, their most stable conformations were different from other drug structures. In particular, the planar system and pendant sugar moiety of doxorubicin, which also required an adenine but at position -1, was not superimposed to the corresponding moieties of mAMSA and bisantrene even when considering computer-generated conformations with higher energy contents. The most stable conformations of the other drugs studied revealed specific three-dimensional motifs. Therefore, since in a simple model of drug action each spatial region has a single chemical-pharmacological function, these results suggest that bisantrene and mAMSA share common steric and electronic features that may constitute a specific pharmacophore. We suggest that the molecular properties of this pharmacophore may be critical determinant of the +1 position specificity shown by mAMSA and bisantrene.