There have been some attempts to develop immunotoxins utilizing human RNase as a cytotoxic domain of antitumor agents. We have recently shown that only human RNase 3 (eosinophil cationic protein, ECP) among five human pancreatic-type RNases excels in binding to the cell surface and has a growth inhibition effect on several cancer cell lines, even though the RNase activity of RNase 3 is completely inhibited by the ubiquitously expressed cytosolic RNase inhibitor. This phenomenon may be explained by that RNase 3 is very stable against proteolytic degradation because RNase 3 internalized through endocytosis could have a longer life time in the cytosol, resulting in the accumulation of enough of it to exceed the concentration of RNase inhibitor, which allows the degradation of cytosolic RNA molecules. Thus, we compared the stabilities of human pancreatic-type RNases (RNases 1-5) and bovine RNase A by means of guanidium chloride-induced denaturation experiments based on the assumption of a two-state transition for unfolding. It was demonstrated that RNase 3 is extraordinarily stabler than either RNase A or the other human RNases (by more than 25 kJ/mol). Thus, our data suggest that in addition to its specific affinity for certain cancer cell lines, the stability of RNase 3 contributes to its unique cytotoxic effect and that it is important to stabilize a human RNase moiety through protein engineering for the design of human RNase-based immunotoxins.