The makings of a tumor rejection antigen

The demonstration that naturally induced tumors in rodents were largely nonimmunogenic and disappointing results from clinical studies were responsible for the notion that tumors are not sufficiently distinct from normal tissue to activate the immune system and led to the inevitable conclusion that immunological intervention in cancer is futile (Hewitt et al., 1976). In a seminal work, van Pel and Boon have shown that a protective immune response can be generated against a "nonimmunogenic" murine tumor, providing the first experimental evidence that lack of immunogenicity could be due to the tumor's inability to activate the immune system rather then the absence of tumor antigens (van Pel and Boon, 1982). This observation, subsequently confirmed and extended to other rodent nonimmunogenic tumor models, has shown that by proper manipulation-otherwise called vaccination-the tumor antigens present in nonimmunogenic tumors can be "exposed" to the immune system to generate an immune response capable of eradicating the tumor. If this conclusion can be extrapolated to human cancer-and I see no reason why it cannot-all forms of cancer should be susceptible to immunological intervention; namely, all forms of cancer contain tumor antigens that can be targeted for immunotherapy. The recognition that tumors could after all be sufficiently "foreign" to be recognized by the immune system has reinvigorated the efforts to identify and isolate tumor antigens (Boon and van der Bruggen, 1996; Rosenberg, 1999). This review will focus on what makes a tumor antigen a good or not-so-good target for immunotherapy.