Human solid tumors contain hypoxic regions that have considerably lower oxygen tension than normal tissues. These impart resistance to radiotherapy and anticancer chemotherapy, as well as predisposing to increased tumor metastases. To develop a potentially therapeutic protein drug highly specific for solid tumors, we constructed fusion proteins selectively stabilized in hypoxic tumor cells. A model fusion protein, oxygen-dependent degradation (ODD)-beta-galactosidase (beta-Gal), composed of a part of the ODD domain of hypoxia-inducible factor-1alpha fused to beta-Gal, showed increased stability in cultured cells under a hypoxia-mimic condition. When ODD-beta-Gal was further fused to the HIV-TAT protein transduction domain (TAT(47-57)) and i.p. injected to a tumor-bearing mouse, the biologically active fusion protein was specifically stabilized in solid tumors but was hardly detected in the normal tissue. Furthermore, when wild-type (WT) caspase-3 (Casp3(WT)) or its catalytically inactive mutant was fused to TAT-ODD and i.p. injected to a tumor-bearing mouse, the size of tumors was reduced by the administration of TAT-ODD-Casp3(WT) but not by TAT-ODD-mutant Casp3. TAT-ODD-Casp3(WT) did not cause any obvious side effects on tumor-bearing mice, suggesting specific stabilization and activation of the fusion protein in the hypoxic tumor cells. These results suggest that the combination of protein therapy using a cytotoxic TAT-ODD fusion protein with radiotherapy and chemotherapy may provide a new strategy for annihilating solid tumors.