Unlike monoclonal antibodies, clinical application of bispecific antibodies has so far lagged behind because of difficult, low-yield production techniques as well as considerable toxicity attributed to bispecific antibody preparations containing immunoglobulin-Fc parts and anti-CD3 homodimers. These difficulties were overcome by recombinant generation of a bispecific single-chain antibody (bscAb) joining two single-chain Fv fragments via a five-amino-acid glycine-serine linker. The anti-CD3 specificity directed against human T cells was combined with another specificity against the epithelial 17-1A antigen. The following domain arrangement was critical in this individual case to obtain a fully functional bscAb: VL17-1A-VH17-1A-VHCD3-VLCD3. The bscAb was expressed in chinese hamster ovary cells with a yield of 15 mg/l culture supernatant whereas numerous attempts to obtain a functional protein expression in Escherichia coli failed. The low-molecular-mass bispecific construct (60 kDa) could easily be purified by its C-terminal histidine tail. The antigen-binding properties are indistinguishable from those of the corresponding univalent single-chain Fv fragments as shown by enzyme immunoassay and flow cytometry. We could show that the bscAb, which lacks Fc parts and anti-CD3 homodimers is highly cytotoxic for 17-1A positive tumor cells in nanomolar concentrations and in the presence of human T cells. Most remarkable, it does not stimulate T lymphocyte proliferation in the absence of tumor cells and, moreover, does not induce CD25 up-regulation and the secretion of potentially toxic lymphokines such as tumor necrosis factor alpha, interleukin-6 and interferon gamma. Maximal cytotoxicity (51Cr release) was achieved without notable costimulation and was not further enhanced by adding costimulatory signals such as those delivered by anti-CD28 antibodies. CD8+ as well as CD4+ T cell subpopulations were recruited to exert cytotoxicity against tumor cells with different kinetics. CD8+ cells induced high 51Cr release within 4 h while CD4+ cells required a 20-h incubation. The systemic application of the 17-1A/CD3-bscAb could be a major improvement in therapy against disseminated micrometastatic tumor cells. A prospective, randomized clinical trial showing that an anti-17-1A monoclonal antibody could prolong survival of colorectal cancer patients after 5 and 7 years, warrants an assessment of the clinical efficacy of this bscAb exhibiting a 1000-fold higher specific cytotoxicity against tumor cells in vitro.