Appendix A
Workshop Summary
TWO-STAGE MODEL OF CARCINOGENESIS
The goals of this workshop were (1) to assess the scientific basis for the two-stage model of carcinogenesis and (2) to evaluate the possible applications of the two-stage model to the health risk assessment process. Two-stage models are based on the assumptions that carcinogenesis is a multistage process, and that in its simplest form, two critical events are sufficient to convert normal cells to cancer cells (e.g., retinoblastoma in children).
The workshop was opened by the vice-chair, D. Mattison, who welcomed the participants and provided perspective on the relation of this workshop to the overall activities of the Committee on Risk Assessment Methodology (CRAM). The workshop chair, R. Griesemer, emphasized that the workshop is one mechanism through which CRAM obtains information and urged the participants to share additional ideas or information with CRAM after the workshop.
BIOLOGICAL FACTORS IN TWO-STAGE MODELS
A.J. Knudson, who first proposed the concept of two-stage models, presented a keynote address on the evidence from studies of heritable cancers in humans that supports the concept of two-stage models.
If cancer is related to somatic mutations, there should be some background incidence for all cancers. One would anticipate that there would be an increase in incidence upon exposure to agents that affect this process and that there would be specific targets for those mutations with tissue specificity.
At present we know of two classes of targets, proto-oncogenes and anti-oncogenes (suppressor genes). Where oncogene mutations are found in human tumors, the evidence indicates they may not be the initial events; in some instances, specific translocations seem to be the only identifiable event in the origin of a cancer. The translocations seem to be dominant in the sense that activation of one copy of an oncogene confers malignancy on a cell. In the case of suppressor genes, with release of control of cell growth, two copies must become inactivated and the events can be hereditary or nonhereditary.
Hereditary cancers have provided useful information about the genetic events in carcinogenesis. Virtually every cancer type has a dominantly inherited subgroup. The hereditary fraction for retinoblastoma is rather large (about 40%). The probability of hereditary retinoblastoma in children with an inherited abnormal rb gene is 100,000 times greater than that for the nonhereditary form. The relation of incidence to age suggests that one event is necessary in somatic cells of carriers and two events in nonhereditary cases. Now that the gene has been isolated and mapped on chromosome 13, this suggestion has been supported by genetic analyses of cells from affected children. Somatic mutations depend on the mutation rate per cell division and the number of cell divisions per unit time; retinoblastomas do not develop in adults because the retinal cells have differentiated and no longer divide.
In embryonal tissues such as retina, there is no conditional cell division; a mutation results in a clone of cells carrying the mutation. One can imagine initiation as a loss of one retinoblastoma suppressor gene and promotion as the proliferation that normally occurs in retinoblasts.
Survivors of the hereditary form of retinoblastoma are at risk for other cancers. About 15% of gene carriers develop osteosarcoma. About 95% of patients with osteosarcoma have mutations in both the rb and p53 genes. These two genes are involved in virtually all small cell lung tumors but a third presumed gene on chromosome 3p is also involved in 90% of those tumor cases. A comparison of these three tumors that appear to involve a different number of genes (one for retinoblastoma, two for osteosarcoma, three for small cell lung tumors) sug-
gests that as the number of genes involved goes up, the relative risk goes down because the inherited gene is a smaller and smaller fraction of the total number of important events.
For adult carcinomas, the picture is less clear. Renewal tissues in which stem cells replicate may have some properties like embryonal tissues and a number of gene changes are being found in a variety of human tumor types (e.g., colon). Still unknown, however, is how many genetic events are required for a particular cancer and what is the meaning of the various events.
R. Maronpot's discussion dealt with oncogenes and cell proliferation from the perspective of an experimentalist. He suggested that the one-hit model may be more appropriate for cancers that arise from exposure to ionizing radiation, potent alkylating agents, or in transgenic mice developing lung cancer, for example, but he admitted that the multiclonal nature of the response indicated that a second event would have to be postulated. The mouse skin tumor model is a well known example of two-stage tumors induced by xenobiotics. He cautioned that the data to which models may be applied may have implied but unwarranted precision.
Illustrations of the importance of the ras-oncogene in the B6C3F1 mouse followed. Liver tumors from mice exposed to vinyl chloride have a high frequency of ras-oncogene activation. Those associated with methylene chloride, trichloroethylene, and dichloroacetic acid are similar in ras activation patterns to nonexposed controls. Ras-oncogene activation is not detected in liver tumors associated with the administration of tetrachloroethylene, chloroform, or phenobarbital. Furan produces some novel ras mutations. Maronpot suggests if specific types of oncogene mutation and activation are found in both animal and human tumors that those findings would be important for risk assessment.
Characterization of the cell proliferative response has a number of pitfalls and limitations. Examples were given where cellular proliferation appears to be an important consideration such as in the kidney with respect to d-limonene and unleaded gas and in the bladder for saccharin, but the hepatic-cell proliferation after methylene chloride increased only slightly at the 12-month interval and not at all at the 3-,6-, and 18-month intervals. There is an important temporal relationship between cancer and cellular proliferation but that by itself is not evidence of a causal association.
During the ensuing discussion, S.H. Moolgavkar assured the audience