Molecular basis for the DNA sequence selectivity of ecteinascidin 736 and 743: Evidence for the dominant role of direct readout via hydrogen bonding

The marine natural product ecteinascidin 743 (Et 743) is currently in phase II clinical trials. We have undertaken parallel structural and modeling studies of an Et 743-(N2-guanine) 12-mer DNA adduct and an adduct involving the structurally related Et 736 of the same sequence in order to ascertain the structural basis for the ecteinascidin-DNA sequence selectivity. In contrast to the C-subunit differences found in Et 736 and Et 743, they have identical A-B-subunit scaffolds, which are the principal sites of interaction with DNA bases. These identical scaffolds generate parallel networks of drug-DNA hydrogen bonds that associate the drugs with the three base pairs at the recognition site. We propose that these parallel hydrogen bonding networks stabilize the Et 736 and Et 743 A- and B-subunit prealkylation binding complex with the three base pairs and are the major factors governing sequence recognition and reactivity. The possibility that a unique hydrogen-bonding network directs the course of sequence recognition was examined by first characterizing the hydrogen-bonding substituents using H-1 NMR properties of the exchangeable protons attached to the hydrogen-bond donor and other protons near the proposed acceptor. Using these experimental findings as indicators of hydrogen bonding, Et 736-12-mer duplex adduct models (binding and covalent forms) containing the favored sequences 5'-AGC and 5'-CGG were examined by molecular dynamics (MD) in order to evaluate the stability of the hydrogen bonds in the resulting conformations. The MD-generated models of these favored sequences display optimal donor/acceptor positions for maximizing the number of drug-DNA hydrogen bonds prior to covalent reaction. The results of MD analysis of the carbinolamine (binding) forms of the sequences 5'-GGG (moderately reactive) and 5'-AGT (poorly reactive) suggested reasons for their diminished hydrogen-bonding capability. These experimental and modeling results provide the structural basis for the following sequence specificity rules: For the target sequence 5'-XGY, the favored base to the 3'-side, Y, is either G or C. When Y is G, then a pyrimidine base (T or C) is favored for X. When Y is C, a purine (A or G) is favored for X.