Potential antitumor agents. 29. Quantitative structure-activity relationships for the antileukemic bisquaternary...

Quantitative relationships between physicochemical drug properties and antileukemic (L1210) efficacy have been examined for a series of bisquaternary ammonium heterocycles employing multiple variable regression analysis. Three measures of biologic response were examined: ILSmax, the percentage increase in mean life span of leukemic animals at the LD10 dose; D40, the drug dose necessary to provide 40% increase in life span; and CI (=LD 10/D40), the chemotherapeutic index. A cross correlation matrix between these three measures and the LD10 values demonstrates ILSmax and CI to be independent of toxicity. D40 is highly inversely correlated with LD10 and positively correlated with ILSmax, suggesting that this parameter measures a composite of both drug selectivity and toxicity. Superior regression equations resulted at all stages employing ILSmax as a measure of antitumor selectivity. Acceptable equations modeling LD10 could not be obtained. There was a parabolic relationship between agent lipophilic-hydrophilic balance, measured as chromatographic Rm values, and ILSmax. To reduce residual variance in the L1210 screening data, not accepted by this parabolic equation, measures of agent-DNA interaction were investigated as possible indices of site fit. Relative levels of drug-DNA interaction were obtained by spectrofluorimetric quantitation of drug displacement of DNA-bound ethidium. Addition to regression equations of agent C50 values for calf thymus DNA, those micromolar drug concentrations necessary to displace 50% of the ethidium bound to that DNA, provided a significant reduction in the screening data variance. C50 values for drug interactions with poly[d(A-T)] and poly[d(G-C)] were also investigated as possible indicators of drug selectivity towards different DNA sites. Marked differences were observed in the C50 values for the two synthetic nucleic acids, with those for calf thymus DNA and poly[d(G-C)] proving highly covariant. A regression equation containing a positive term in C50[poly[d(G-C)]] and a negative term in C50-[poly[d(A-T)]] provided the greatest acceptance of the variation in the biologic data. The term in C50[poly[d(A-T)]] is the single most important equation term, alone accounting for 35% of the variance in the data. This best equation provides both an adequate summary of the L1210 screening data for the 174 active compounds considered and a physical explanation for the antitumor selectivity for these agents. The major factor influencing antitumor selectivity in these drugs is their ability to distinguish alternating adenine-thymine sequences in DNA.