Geldanamycin (GA) is a potent anticancer antibiotic that inhibits Hsp90. Its potential clinical utility is hampered by its severe toxicity. To alleviate this problem, we synthesized a series of carbohydrate-geldanamycin conjugates for enzyme-specific activation to increase tumor selectivity. The conjugation was carried out at the C-17-position of GA. Their anticancer activity was tested in a number of cancer cell lines. The enzyme-specific activation of these conjugates was evaluated with beta-galactosidase and beta-glucosidase. Evidently, glycosylation of C-17-position converted GA to an inactive prodrug before enzyme cleavage. Glucose-GA, as positive control, showed anticancer activity with IC(50) of 70.2-380.9 nM in various cancer cells by beta-glucosidase activation inside of the tumor cells, which was confirmed by 3-fold inhibition using beta-glucosidase specific inhibitor [2,5-dihydroxymethy-3,4-dihydroxypyrrolidine (DMDP)]. Compared to glucose-GA, galactose- and lactose-GA conjugates exhibited much less activity with IC(50) greater than 8000-25 000 nM. However, when galactose- and lactose-GA were incubated with beta-galactosidase in the cells, their anticancer activity was enhanced by 3- to 40-fold. The results suggest that GA can be inactivated by glycosylation of C-17-position and reactivated for anticancer activity by beta-galactosidase. Therefore, galactose-GA can be exploited in antibody-directed enzyme prodrug therapy (ADEPT) with beta-galactosidase for enzyme-specific activation in tumors to increase tumor selectivity.

A general methodology has been established for rapid generation and screening of combinatorial glycopeptide library and subsequent mass spectrometric sequencing to identify the mimetics of Galalpha(1,3)Gal epitopes. Using this approach, several active glycopeptide sequences were recognized and found to inhibit the binding of human natural anti-Gal antibodies with comparable IC(50)s to synthetic Galalpha(1,3)Gal oligosaccharides. The most active glycopeptides detected from the library included Gal-Tyr-Trp-Arg-Tyr, Gal-Thr-Trp-Arg-Tyr, and Gal-Arg-Trp-Arg-Tyr. These glycopeptides showed higher affinities to anti-Gal antibodies than known Galalpha(1,3)Gal peptide mimetics, such as DAHWESWL and SSLRGF. Our results suggest that, by combining a peptide sequence (the "functional" mimic part) with a terminal alpha-linked galactose moiety (the "structural" mimic part), the resulting glycopeptide could be a very good Galalpha(1,3)Gal mimetic. Analysis of these active glycopeptides provided first-hand information regarding the binding site of anti-Gal antibodies to facilitate the structurally based design of more potent and stable inhibitors.