Tumor immunogenicity: how far can it be pushed?

One recent paper in the Proceedings (1) and one in this issue (2) demonstrate that tumor immunogenicity can be enhanced by engineering the tumor cells to produce any one of several molecules known to play a role in the immune mechanism, a molecule that is not expressed or is underexpressed in the parental tumor. An important feature of these studies is that the induction of increased immunogenicity in the engineered cells results in an enhanced reaction against the unaltered parental cells, thus possibly opening the door to successful immunotherapy. Of course, before the age of genetic engineering, many less sophisticated methods had been used to increase the immunogenicity of tumor cells, and it remains to be seen whether these more modem approaches are of greater practical significance than the methods of the past. If nothing else, they do promise to give insight into the mechanisms of the enhanced immunogenicity. It is now almost 40 years since it became apparent that many tumors are immunogenic, as judged by immunizationchallenge tests with transplanted tumors in syngeneic mice. During that time the immune mechanism has been extensively dissected, and, although much remains to be learned, the increase in understanding has been impressive. However, despite some limited indications of beneficial therapeutic effects, attempts to use the immune mechanism for therapy have been largely disappointing. It seems to me that the reasons for this disappointment reside in two separate features of the immune response to tumors, either or both of which may possibly be overcome by current studies such as those described in the two Proceedings papers (1, 2). The first feature that in my opinion may have helped to frustrate past attempts at immunotherapy was observed in the earliest modem studies of tumor immunity. Work in vitro and with syngeneic tumor implants in vivo has made it clear that many tumors possess antigens to which the host is reactive in transplantation and other tests, but, as Stutman (3) has pointed out, this immunogenicity is seldom realized during oncogenesis in a way that affects the incidence of de novo tumors; alteration of the immune capacities of the hosts seldom altered incidence. The juxtaposition of these facts points unmistakably to the critical role of the manner of antigen presentation. Excision of the primary de novo potentially immunogenic tumor and the subsequent challenge of that animal with that tumor showed little or no effect upon the growth of the challenge implant as compared with the growth of the same tumor in syngeneic controls. Thus, to show that a mouse could be immunized against implants of its own (i.e., autochthonous, methylcholanthrene-induced tumor), it was necessary to repeatedly immunize the animal with that tumor before giving the challenge implant (4); the temporary growth of the primary in situ untransplanted tumor did not of itself produce a detectable immunity to subsequent implants of the tumor. A number of experiments can be cited that make essentially the same point (5, 6). An animal (and by inference a human) is not appreciably immunized by the growth of its undisturbed de novo tumor even if the tumor is demonstrably immunogenic when transplanted to syngeneic hosts; furthermore, for one reason or another, the original host may be difficult to immunize as compared with secondary hosts-a point that needs further systematic study (4). Nonetheless, the mouse work does show that the host can be immunized, perhaps only with difficulty, against its own native tumor; therefore, immunotherapy of at least some tumors would seem to be a possibility. The second and (in the present context) probably more important feature of tumor immunity, one that may be fundamental in frustrating attempts at immunotherapy, was also observed at the very beginning of the modem era-namely, the lack of immunogenicity of so-called "spontaneous" tumors. It seems to be a general rule among animal tumors when the appropriate transplantation studies can be performed that the weaker an associated carcinogenic agent may be, the less immunogenicity on average the tumor will appear to possess. At the extremes, sarcomas induced with high dosages of the potent carcinogen 3-methylcholanthrene tended to be highly immunogenic with few exceptions, whereas those that arose spontaneously (i.e., without known contact with a carcinogenic agent) appeared to be nonimmunogenic (7). The effect seems to be independent of the immunodepressive effects of the carcinogen because similar results are obtained in vitro or when less than immunodepressive dosages of carcinogen are used (3, 8). In my laboratory, a reproducible positive correlation was observed between the dosage of the chemical carcinogen and the average immunogenicities of the resulting sarcomas (9, 10). Probably the most often cited paper illustrating the lack of immunogenicity of spontaneous rodent tumors is that of Hewitt, who studied 27 spontaneous mouse tumors and failed to detect immunogenicity in any one of them (11). If so-called spontaneous tumors were in truth devoid of moieties with immunogenic potential, the outlook for the immunotherapy of most human tumors might indeed be dim; the efficacy of genetic engineering in providing molecules instrumental in increasing immunogenicity presumably depends upon the prior existence in the tumor of an antigen whose effectiveness as an immunogen awaits only amplification. If this were not the case, the new immunogenicity of the engineered cells would not be crossreactive with the parental cells, and all would be for naught. It may be impossible to be sure what part of the spectrum of human tumors is analogous to the putatively nonimmunogenic spontaneous tumors of rodents. If the definition of "spontaneous" means sporadic appearance without the obvious aid of a carcinogenic agent, many human tumors seem to qualify. However, humans are not inbred, and it is possible that certain tumor types that seem to appear sporadically in low incidence might actually be appearing in 100% of humans of a particular susceptible genotype. In any event, for the further purposes of this commentary, I will assume that many human tumors are analogous to the spontaneous rodent tumors and discuss what this might or might not imply. The key question upon whose answer I believe the future of immunotherapy depends is whether or not most tumors, spontaneous as well as induced, in fact do possess antigens whose immunogenicity might be augmented by genetic engineering or other means. Although many tumors are completely nonimmunogenic by the transplantation-