Tumor necrosis factor-alpha (TNF-alpha) enhances X-ray killing of human tumor cells in vitro and enhances tumor control when combined with radiotherapy (RT) in animal tumor models. In multiple Phase I studies, intravenous injection of TNF-alpha appeared to have severe systemic side effects. To overcome these limitations, we used a bispecific antibody (BAb) directed against carcinoembryonic antigen and human TNF-alpha to target this cytokine in human digestive carcinoma treated with simultaneous RT. We used human digestive carcinoma cell lines (colon cancer, LS174T, and pancreatic cancer, BxPC-3) to determine the interaction of TNF-alpha and RT on clonogenic cytotoxicity. Isobolograms were established to confirm additive or supra-additive effects between both treatments. LS174T and BxPC-3 cells were grafted subcutaneously at Day 0 into female nude mice (7-8 weeks old). When the tumors reached a volume of about 80 mm(3), the mice were randomly assigned to treatment: Group 1, normal saline i.v. injection (control group); Group 2, TNF-alpha at 1 microg/i.v. injection; Group 3, BAb at 25 microg/i.v. injection; Group 4, BAb plus TNF-alpha (ratio 25 microg to 1 microg) i.v. injection; Group 5, local RT plus normal saline (0.5 Gy. min(-1)) at a total dose of 30 Gy delivered in five fractions; Group 6, local RT plus TNF-alpha injections 3 h before RT; Group 7, local RT plus BAb plus TNF-alpha co-injected 24 h before RT. Tumor growth delay was used as the end point for all groups. In the LS174T experiments, TNF-alpha added 12 h before RT showed a statistically significant decrease in the survival fraction at 2 Gy compared with RT alone (0.23 vs. 0.42 Gy, p = 0.0017). These results were largely confirmed with the BxPC-3 cell lines (0.29 vs. 0.72, p <0.00001). Isobolograms confirmed the additivity between TNF-alpha and RT in both cell lines. At 50% survival, the data points were within the envelope of additivity. In the LS174T and BxPC-3 xenografts, RT as a single agent (Group 5) slowed tumor progression compared with Group 1 (p <0.027 and p = 0.00001, respectively). TNF-alpha alone, BAb alone, or BAb plus TNF-alpha (Groups 2, 3, and 4) had no effect. In the LS174T model, TNF-alpha plus RT enhanced the delay to reach 2000 mm(3) compared with RT alone but without statistical significance. This delay was significantly longer when BAb was added (p = 0.0033, for Group 6 vs. Group 7). In the BxPC-3 experiments, the median delay to reach 2000 mm(3) was similar between the RT and TNF-alpha plus RT groups (93 days). The use of our BAb in combination with TNF-alpha and RT dramatically enhanced this median delay (177 days, p = 0.0013). No body weight loss was observed in any group. Our data could be used as a solid preclinical rationale on which to base a clinical study of locally advanced pancreatic or rectal cancers in the near future.