Therapeutic cancer vaccination is based on the finding that tumors in both humans and experimental animals, such as mice, express potential immunological targets, some of which have high selectivity for cancer cells. In contrast to the successful vaccination against some infectious diseases, where most vaccines induce neutralizing antibodies that act prophylactically, the aim of therapeutic cancer vaccines is to treat established tumors (primarily micrometastases). Since most tumor-destructive immune responses are cell-mediated, therapeutic cancer vaccination needs to induce and expand such responses and also to overcome "escape" mechanisms that allow tumors to evade immunological destruction. Tumor antigens (as with other antigens) are presented by "professional" antigen-presenting cells, most notably dendritic cells (DC). Therefore DC that have been transfected or "pulsed" to present antigen provide a logical source of tumor vaccines, and some encouraging results have been obtained clinically as well as in preclinical models. An alternative and more physiological approach is to develop vaccines that deliver tumor antigen for in vivo uptake and presentation by the DC. Vaccines of the latter type include tumor cells that have been modified to produce certain lymphokines or express costimulatory molecules, as well as cDNAs, recombinant viruses, proteins, peptides and glycolipids which are often given together with an adjuvant. Several studies over the past 5 years have demonstrated dramatic therapeutic responses against established mouse tumors as a result of repeated injections of agonistic monoclonal antibodies (MAbs) to the costimulatory molecule CD137 (4-1BB). However, the clinical use of such MAbs may be problematic since they depress antibody formation, for example, to infectious agents. The alternative approach to transfect tumor cells to express the CD137 ligand (CD137L) increases their immunogenicity, but vaccination with tumor cells expressing CD137L is ineffective in several systems where injection of anti-CD137 MAb produces tumor regression. Recent findings indicate that a more effective way to engage CD137 towards tumor destruction is to transfect tumor cells to express a cell-bound form of anti-CD137 single-chain Fv fragments (scFv). Notably, tumors from melanoma K1735, growing either subcutaneously or in the lung, could be eradicated following vaccination with K1735 cells that expressed anti-CD137 scFv. This was in spite of the fact that K1735, as with many human neoplasms, expresses very low levels of MHC class I and has low immunogenicity. Similar results were subsequently obtained with other tumors of low immunogenicity, including sarcoma Ag104. We hypothesize that the concomitant expression of tumor antigen and anti-CD137 scFv effectively engages NK cells, monocytes and dendritic cells, as well as activated CD4(+) and CD8(+) T cells (all of which express CD137) so as to induce and expand a tumor-destructive Th1 response. While vaccines in the form of transfected tumor cells can be effective, at least in mouse models, the logical next step is to construct vaccines that combine genes that encode molecularly defined tumor antigens with a gene that encodes anti-CD137 scFv. Before planning any clinical trials, vaccines that engage CD137 via scFv need to be compared in demanding mouse models for efficacy and side effects with vaccines that are already being tested clinically, including transfected DC and tumor cells producing granulocyte-macrophage colony-stimulating factor.