Many types of malignant cells and human tumors display increased concentrations of the protease plasminogen activator that converts plasminogen to the highly active protease, plasmin. Because plasmin rapidly cleaves various low molecular weight compounds coupled to appropriate peptide specifiers, we hypothesized that coupling of such peptide specifiers to anticancer drugs might create "prodrugs" which would be locally activated by tumor-associated plasmin and consequently would be less toxic to normal cells. To provide an initial test ofthis concept we have synthesized peptidyl prodrugs of the structure D-Val-Leu-Lys-X in which the peptidyl portion has been designed to allow the prodrug to serve as an excellent plasmin substrate and X is an anticancer drug-either the glutamine analog (aS,5S) a-amino-3-chloro-4,5-dihydro-5-isoxazoleacetic acid (AT-125) or the alkylating agent NN-bis(2-chloroethyl)-p-phenylenediamine (phenylenediamine mustard).Treatment of these prodrugs with plasmin generated the free peptide and the free drug, demonstrating that these prodrugs are plasmin substrates. The prodrugs and free drugs were tested in an in vitro system against either normal chicken embryo fibroblasts, which display a low level of plasminogen activator, or their virally transformed counterparts, which produce high levels of plasminogen activator. In each case the peptidyI prodrugs displayed at least a 5-fold increase in selectivity for the transformed cells compared to the free drug. The greater selectivity of action of the peptidyl prodrugs against transformed cell cultures suggests that these or similar prodrugs that are substrates for tumor-associated proteases may show increased therapeutic effectiveness in the treatment of tumors that produce sufficiently increased amounts of plasminogen activator.A persistent problem in cancer chemotherapy is the low therapeutic index of most anticancer drugs-i.e., the concentrations needed to kill tumor cells are close to those that produce severe toxicity to the host. One often-proposed approach to overcoming this problem has been to design anticancer "prodrugs" which are inactive until locally activated by some tumor-associated enzyme (1-5). Although this approach has heretofore shown little success, this may be due to incomplete characterization of tumor-associated enzymes and their substrate specificity as well as to the admittedly severe difficulty that no enzyme unique to tumor cells is known, so that the prodrugs are probably activated by some normal cells as well as tumor cells.One characteristic of many cancer cells-including cells transformed in vitro by viral (6-9), chemical (10), or physical (11) agents and cell lines established from various human neoplasms (12-16)-is an increased production of the serine protease plasminogen activator. This enzyme cleaves the serum zymogen plasminogen to generate the serine protease plasmin. The plasmin production associated with this increased plasminogen activator activity of many tumors offers an attractive possibility for...