The use of tumor necrosis factor alpha (TNFalpha) in cancer therapy is limited by its short circulatory half-life and its severe systemic side effects. To overcome these limitations, we evaluated the capability of a bispecific antibody (BAb) directed against carcinoembryonic antigen (CEA) and human TNFalpha to target this cytokine in tumors. A BAb was constructed by coupling the Fab' fragments from an anti-CEA monoclonal antibody (MAb) to the Fab' fragments from an anti-TNFalpha MAb via a stable thioether linkage. The double specificity of the BAb for CEA and TNFalpha was demonstrated using a BIAcoreTM two-step analysis. The affinity constants of the BAb for CEA immobilized on a sensor chip and for soluble TNFalpha added to the CEA-BAb complex were as high as those of the parental MAbs (1.7 x 10(9) M-1 and 6.6 x 10(8) M-1, respectively). The radiolabeled 125I-labeled BAb retained high immunoreactivity with both CEA and TNFalpha immobilized on a solid phase. In nude mice xenografted with the human colorectal carcinoma T380, the 125I-labeled BAb showed a tumor localization and biodistribution comparable to that of 131I-labeled anti-CEA parental F(ab')2 with 25-30% of the injected dose (ID)/g tumor at 24 h and 20% ID/g tumor at 48 h. To target TNFalpha to the tumor, a two-step i.v. injection protocol was used first, in which a variable dose of 125I-labeled BAb was injected, followed 24 or 48 h later by a constant dose of 131I-labeled TNFalpha (1 microg). Mice pretreated with 3 microg of BAb and sacrificed 2, 4, 6, or 8 h after the injection of TNFalpha showed a 1.5- to 2-fold increased concentration of 131I-labeled TNFalpha in the tumor as compared to control mice, which received TNFalpha alone. With a higher dose of BAb (25 microg), mice showed a better targeting of TNFalpha with a 3.2-fold increased concentration of 131I-labeled TNFalpha in the tumor: 9.3% versus 2.9% ID/g in control mice 6 h after TNFa injection. In a one-step injection protocol using a premixed BAb-TNFalpha preparation, similar results were obtained 6 h postinjection (3.5-fold increased TNFalpha tumor concentration). A longer retention time of TNFalpha was observed leading to an 8.1-fold increased concentration of TNFalpha in the tumor 14 h postinjection (4.4 versus 0.5% ID/g tumor for BAb-treated and control mice, respectively). These results show that our BAb is able, first, to localize in a human colon carcinoma and, there, to immunoabsorb the i.v.-injected TNFalpha, leading to its increased concentration at the tumor site.