Our recent studies have determined some of the structural features that confer selectivity for bioreductive activation of lavendamycin analogs by NAD(P)H:quinone oxidoreductase (NQO1) in NQO1-rich cancer cells. To accomplish NQO1-directed lavendamycin antitumor agent development, novel analogs were rationally designed utilizing information obtained from our structure-activity studies and in silico general model of the NQO1 active site. The rational design criteria for substituent features at key positions were as follows: 1) Quinolinedione-7-position (R1)- Small to medium size substituents that do not produce steric interactions and are capable of hydrogen bond formation in the NQO1 active site, and/or can form Van der Waals interactions with Trp-105/Phe-106 mini-pocket, 2) Quinolinedione-6-position (R2)- No substituent, and 3) Indolopyridine-2′-position (R3)- Small to medium size substituents that are capable of hydrogen bond formation in the NQO1 active site or medium to large substituents that contain aliphatic chains that are capable of formation of Van der Waals interactions in the NQO1 active site. Several assumed poor and good designed substrates were docked into the model using the FlexX module of SYBYL 7.0 to be computationally ranked in regards to the docking score (CSCORE) and post-docking analysis data. Synthesized analogs were then tested biologically to determine the practicality of the design criteria, predictive power of the model and whether the computational data were consistent with biological results. Cytochrome c and MTT assays were used to determine the reduction rate by NQO1 and cytotoxicity of the analogs toward NQO1-deficient BE-WT and NQO1-rich BE-NQ human colon adenocarcinoma cells. Modeling studies ranked the designed analogs 2′-CONH(CH2)3CH3-7-NHCOCH3 (MB-116), 2′- CON[(CH2)2O(CH2)2′-7-NH2 (MB-115) and 2′-CONHCH(CH3)C2H5-7-NHCOCH3 (MB-137) as good, good and poor substrates, respectively. Reduction rates for the analogs were 142.6 ± 10.9, 102.7 ± 8.4 and 4.9 ± 2.9 μmol/min/mg NQO1, respectively, and selectivity ratios [IC50 (BE-WT) / IC50 (BE-NQ)] were 29 for MB-116 and 9 for MB-115 and no measurable cytotoxicity (IC50 > 50 μM) was determined for MB-137. Although MB-116 and MB-137 possessed similar chemical structures with only a minor difference in the substituent at R3, the model predicted very different substrate efficiencies for them, which were consistent with the biological data. This study determined that the rational design criteria were instrumental, resulting in the design of analogs with high selectivity ratios. It also indicated that the in silico model could be utilized as a predictive, time- and cost-efficient tool in NQO1-directed rational design of lavendamycins since the predicted, computational data were in agreement with biological results. Supported by NIH grants CA74245, RR15583 and RR17670.

[Proc Amer Assoc Cancer Res, Volume 47, 2006]