Risk estimates, and the resulting radiation-protection standards, are based on the "Linear No-Threshold (LNT)" hypothesis. The LNT model assumes that risk is linearly proportional to dose, without a threshold, and thus makes the assumption that every dose, no matter how low, carries with it some risk. This is a testable hypothesis, but such tests have, until recently, been hampered by our lack of sensitive techniques to measure or detect biological responses to low doses. We have therefore relied on human epidemiological evidence, which suffers from large uncertainties at these doses. While good evidence indicates that large doses, delivered at a high dose rate create a risk of cancer formation, there are no actual data to support this assumption at doses that are relevant to occupational and public exposures. Our work is testing the LNT hypothesis that all exposures, including low doses, can only be harmful. We are using molecular and cellular techniques to investigate the biological responses of cells and animals to low doses and low dose rates of low LET radiation.

Human and animal studies typically measure the frequency and latency of cancer from a radiation exposure. However, all cancers arise from genetic changes in a single cell and it is therefore appropriate to consider the effects of radiation on single cell. When DNA is damaged by radiation, the cell will attempt to repair that damage. There are three possible outcomes of those repair attempts (Fig.1).

Fig. 1 Potential outcomes of a radiation exposure in a normal cell

Successful, error-free DNA repair will restore the cell to its original state, With no resulting consequence to the cell and hence no resulting risk. Improper repair of the DNA damage should activate a genetically programmed cell death process, again producing no risk of carcinogenesis. Occasionally, the DNA damage is repaired in a way which avoids cell death but which is error-prone and this can produce a mutation. Very few mutations create the potential for cancer, but it is these few mutations, and therefore only this last outcome, which creates the cancer risk. Our experiments are examining the relative probabilities of these three possible outcomes, and how they vary with radiation exposure. Lack of variation in the biological outcome of radiation exposure is an inherent assumption of the LNT model.

This enhanced error-free repair of chromosomal breaks after low dose exposure has the characteristics of a DNA repair type known as "homologous recombination". Repair of chromosomal breaks after high doses is well known to produce deletions and translocations, the

products of an error-prone DNA repair type called illegitimate recombination. Thus the level of exposure determines whether the repair will be "good" or "bad".

In vivo experiments in mice support the concept that low doses reduce rather than increase risk. Pre-exposure of the skin of mice to b-radiation reduced by about five-fold the frequency of skin tumors in mice whose skin was subsequently exposed to a chemical carcinogen, indicating a protective effect on early, tumor initiation events in cells. Other experiments show that a low, whole body dose of g radiation extended the latent period for myeloid leukemia

Fig. 3 Life shortening in irradiated mice that developed myeloid leukemia

induced in the mice by a subsequent large radiation dose, indicating a protective effect of low doses against the late, secondary processes of carcinogenesis (Fig. 3).

The LNT hypothesis that radiation produces only harmful effects does not allow for cellular responses that vary with dose. However, cellular responses to low doses are different from their responses to high doses, which changes the biological outcome and therefore the risk.

The Seventh International Conference on Nuclear Engineering
Special Symposium

April 21, 1999
Concorde Ball Room, Keio Plaza inter-Continental Tokyo