Fifty-eight cell lines in the National Cancer Institute drug screen were analyzed for their ability to effiux the fluorescent dye rhodamine 123 as a functional assay for P-glycoprotein(Pgp). Using flow cytometry, the rhodamine fluorescence was meas-ured for each cell line under four incubation conditions, ie, after accumulation in the presence or absence of the Pgp antagonist cyclosponn A and after efflux in rhodamine-free medium in the presence or absence of cyclosponn A. The results in some cell lines were compatible with Pgp-mediated efflux. There was a significant correlation between mdr-1 expression and rhodamine efflux in the 58 cell lines (r= 0.788, p= 0.0001). Using the rhodamine effiux data as a seed for COMPARE analysis with the cytotoxicity data on> 30,000 compounds in the National Cancer Institute drug screen database, hundreds of compounds with high correlation coefficients were identified. Selected compounds were tested for reversal of cross-resistance in a multidrugresistant cell line. A high degree of reversibility, up to 10,000-fold, for some of the compounds was noted in the presence of the Pgp antagonist PSC 833. This finding suggested that compounds with predominately Pgp-mediated resistance were being identified. Using these compounds as seeds for COMPARE analysis against a more restricted database of 187 standard agents, a series of standard compounds were repeatedly identifled as having high correlation coefficients with the newly identified Pgp substrates. These standard agents, including phyllanthoside, bisantrene, and homoharnngtonine, constitute an mdr-1 profile. New agents identified as being highly correlated with these compounds may benefit from clinical trials with Pgp antagonists.

The goal of the drug evaluation program of the NCI is to discover new anticancer agents through large-scale, semiautomated screening of compounds in human tumor cell lines. To achieve the goal of new drug discovery, approximately 10,000 compounds are screened yearly, by cytotoxicity assay, using 60 human tumor cell lines in disease-oriented panels (1). More than 30,000 compounds have been screened using this approach and a number of agents have been readied for clinical trial. Data from the cytotoxicity assays are stored in the COMPARE database. The differing sensitivities of the 60 cell lines allow a unique profile, termed a fingerprint, to be generated for each compound when the IC values are plotted on a graph. Using the drug sensitivity data, the COMPARE analysis program computes correlation coefficients that describe the degree of similarity of two compounds(2, 3). It has been reported that compounds with similar mechanisms of action, such as those binding tubulin, have similar profiles and high correlation coefficients(4, 5). In contrast, compounds with unique mech-anisms of action have unique profiles and may be identified as