The most serious problem in current gene therapy is that clinical applications have often led to unsatisfactory results. Here we show novel concepts and crucial factors that have been missing for successful cytokine gene therapy. A clinically-relevant mouse model of primary and micro-metastatic osteosarcoma was generated by subcutaneously and intravenously injecting murine osteosarcoma LM8 cells, in which adenoviral gene transduction efficiencies were extremely low; current therapies remain less effective for such disseminated micro-metastases. A single injection of adenoviral vector encoding interleukin-2 gene (Ad.IL-2) was given only into the established primary tumor. Notably, antitumoral immunity was successfully elicited by IL-2 secretion from connective tissues adjacent to the primary tumor, and this immunity not only suppressed primary tumor growth but also eradicated disseminated micro-metastases in distant organs. Most importantly, not only minimal side effects but also maximal therapeutic effects were exerted only in the case of injecting the optimal (i.e., not the highest) dose of Ad.IL-2, because spleen injuries caused by excessive levels of circulating IL-2 might diminish the therapeutic effect. Although the narrow range of the optimal therapeutic expression level of IL-2 may be crucial, it was feasibly determined by serum IL-2 levels. Thus, a crucial factor for successful cytokine gene therapy is not the high gene transduction efficiency in the tumor, which has been generally recommended, but the use of the optimal therapeutic expression level. In conclusion, just a single injection of Ad.IL-2 into a primary tumor lesion, which is feasible, not invasive and cost effective, is potently therapeutic for distant disseminated micro-metastases, as long as the optimal therapeutic level is monitored. These novel concepts, which contradict those of previous studies, warn researches about the possible problems with the ongoing clinical cytokine gene therapy.