We have synthesized peptidyl prodrugs of doxorubicin (Dox) designed to be selective substrates of plasmin. Such prodrugs might be locally activated by the elevated levels of plasmin produced near many solid tumors under the action of tumor-associated plasminogen activators. One such prodrug, 3'-(D-Val-Leu-Lys)-Dox, was obtained via a mixed-anhydride coupling with isobutyl chloroformate between the protected peptide Fmoc-D-Val-Leu-N epsilon-Fmoc-Lys-OH and doxorubicin, followed by removal of the Fmoc groups with anhydrous ammonia. Compared to doxorubicin, the prodrug showed about a 7-fold improved selective cytotoxicity against chicken embryo fibroblasts transformed with the Rous sarcoma virus (which produce high levels of plasminogen activator) compared to normal cells (which produce low levels of plasminogen activator). However, the prodrug was a very poor plasmin substrate, and although in vivo tests against the murine B16 melanoma showed that the prodrug was active, the maximum T/C obtained was less than that achieved by doxorubicin even at 25 times the molar concentration of prodrug. Qualitatively similar results were obtained for a far more hydrophobic prodrug, 3'-(Boc-Val-Leu-Lys)-Dox. These results demonstrate that peptidyl prodrugs of doxorubicin designed as plasmin substrates are more selective anticancer agents in vitro than doxorubicin itself but that the bulky anthracycline moiety probably prevents efficient plasmin-catalyzed conversion to the active parent drug, so that, in their present form, these drugs are not potent enough to allow a determination as to whether or not they are more selective in vivo.