Lung cancer presents a major therapeutic problem. The mortality rate from lung cancer approaches 90%, and treatment of inoperable non-small-eell lung cancer only occasionally results in cures (7). In this setting, a new therapy yielding a significant clinical partial tumor response rate or any apparently complete tumor regressions is all that it takes to warrant large-scale clinical trials. Different therapeutic approaches are clearly needed, especially rationally designed approaches based on our emerging understanding of molecular abnormalities in cancer cells. It is now thought that cancer is the end result of an accumulation of genetic lesions in key regulatory molecules, and the identity and function of these molecules are the subjects of intense ongoing research (2). With the discovery of tumor suppressor genes whose function is missing in cancer cells came the idea of gene-replacement therapy: putting back into cancer cells a gene that is missing or damaged to return growth control. It is entirely possible, however, that several of the multiple lesions would have to be corrected before any antitumor effect could be seen. Because p53 (also known as TP53) is one such tumor suppressor gene that is missing or defective in the majority of human lung cancers as well as in other human cancers, it was a prime candidate for gene therapy (5-5). As a first step, several groups were able to show that the stable transfection (6) or retroviral transduction (7) of wild-type p53 into p53 mutant or null lung cancer cells dramatically inhibits cell growth in vitro despite the presence of multiple other genetic lesions. The immediate problem with application of this approach in patients is the in vivo delivery of the therapeutic gene to sufficient numbers of tumor cells to produce a clinically observable effect. In theory, nearly every clonogenic tumor cell would have to be effectively transduced (receive, integrate, and express the gene) for us to see a significant antitumor response. With a variety of gene transfer methods under optimal conditions in tissue culture, a 30%-50% rate of transduction and expression of a tumor suppressor gene in human cancer cells would be considered extremely good. This rate, however, would probably be of limited clinical signficance and barely detectable in vivo, where logs of tumor cell kill are required. Despite all of these negative hypotheticals, in this issue of the Journal, Fujiwara et al. (8) report the in vivo retroviral transduction of wild-type p53 into human lung cancer cells in an orthotopic nude mouse model. They convincingly demonstrate that intratracheal human lung cancer cell growth can be eliminated or greatly reduced by intratracheal treatment, beginning 3 days after tumor cell inoculation, with a wild-type p53 recombinant retrovirus (but not with a retrovirus containing a mutant p53 or the retrovirus vector alone). The biological effectiveness in this model system of in vivo gene transfer is surprising, especially given the high level of expression of endogenous mutant p53 protein that has been shown to act in a dominant negative fashion to inhibit the transcriptional function of the wild-type p53 protein (9). As is true with all interesting studies, these data raise a series of questions for future experiments. Which human clinical situations would be best approached with this treatment? In those mice cured by this treatment, was there a bystander effect (70) whereby neighboring tumor cells that were not transduced with wild-type p53 somehow had their growth suppressed? Did the growth suppression in transfected cells result from the induction of apoptosis (programmed cell death) reported in other systems (77) and, if so, did the bystander cells undergo the same fate? Is there a bystander effect in more distant, metastatic tumor cells? Does this therapeutic approach work at other sites in the body, for other p53 mutations, and for other human tumors? Why did some of the tumors grow even with the wild-type p53 incorporated? Fundamental practical problems are still apparent as well. Clinically evident human lung cancers are seldom only a few cell layers thick and confined to a closed surface, as is the case in this intratracheal orthotopic model. In addition, these experiments do not attempt to address the major clinical problem of metastases in lung cancer. However, these studies suggest several potential clinical applications. One application could be the treatment of the pleural space and/or the bronchial tree after surgical removal of the primary tumor in an attempt to reduce the incidence of pleural effusions and local recurrences. Another application could be the treatment of patients with small lesions involving the trachea or carina; these lesions could be debulked