Attempts to design an agent which would release cytotoxic nitrogen mustards within collagenase-producing tumors led to the synthesis of Cbz-L-Pro-L-Leu-Gly-L-Pro-Gly-NHC6H4N(CH2CH2Cl)2 (10). 10 was cleaved in vitro by bacterial and tumor-associated collagenase as expected at the peptide bond joining L-leucine and glycine to give Gly-L-Pro-Gly-NHC6H4N(CH2CH2Cl)2 which was over six times more toxic, on a molar basis, than 10. In vivo tests of 10 against well-advanced Sarcoma-180 gave disappointing results. The lack of specific antitumor activity may be accounted for by the presence of competing cleavage reactions by collagenases in certain normal tissues.