In vivo measurement of human somatic mutations may be a valuable biodosimeter of exposure to carcinogens and of cancer risk. We have surveyed translocations at the bcl2 locus in B lymphocytes, and mutations at hprt in T lymphocytes, in 120 individuals with varying exposure to radon and cigarette smoke. bcl2 t(14:18) translocation is the commonest chromosomal alteration observed in non-Hodgkins lymphoma (NHL). We observed a significantly larger range of bcl2 translocation frequency (range: 0-372 x 10(-6), median: 1.9 x 10(-6)) than of hprt mutation frequency (range: 0-76.4 x 10(-6), median: 11.1 x 10(-6)), which is likely the result of clonal proliferation of deathless B cell mutants. We observed that the frequencies of these two distinct lymphocytic mutations are significantly correlated. Although some of the correlated variation is explained by age, a significant correlation of bcl2 mutagenesis persists after age adjustment. Correlated mutagenesis at distinct loci in distinct cell types could be explained by the existence of a mutator phenotype or by variation in exposure to environmental mutagens. NHL is commoner in men than in women, and our data indicate a trend toward higher bcl2 mutagenesis in males than females. There is mounting epidemiological evidence for a worldwide increase in NHL, which may have an environmental basis; molecular epidemiological analysis of bcl2 mutagenesis in exposed populations might be especially relevant to the identification of putative environmental causes. Given the relative ease of the bcl2 assay versus the hprt assay, and the consistency with which data are reproduced from laboratory to laboratory, it is likely that the bcl2 assay will be soon added to the array of assays used in human mutational surveillance.

In this paper, we have compared mutant frequency data at the hprt locus in circulating T-lymphocytes from four large datasets obtained in the UK (Sussex), the USA (Vermont), France (Paris) and The Netherlands (Leiden). In total, data from > 500 non-exposed individuals ranging in age from newborns (cord blood samples) to > 80 years old have been included in the analysis. Based on raw data provided by the four laboratories, a model is presented for the analysis of mutant frequency estimations for population monitoring. For three of the laboratories, a considerable body of data was provided on replicate estimates of mutant frequency from single blood samples, as well as estimates from repeat blood samples obtained over a period of time from many of the individual subjects. This enabled us to analyse the sources of variation in the estimation of mutant frequency. Although some variation was apparent in the results from the four laboratories, overall the data were in general agreement. Thus, in all laboratories, cellular cloning efficiency of T-cells was generally high (> 30%), although in each laboratory considerable variation between experiments and subjects was seen. Mutant frequency per clonable T-cell was in general found to be inversely related to cloning efficiency. With the exception of a few outliers (which are to be expected), mutant frequencies at this locus were in the same range in each dataset; no effect of subject gender was found, but an overall clear age effect was apparent. When log mutant frequency was analysed vs log (age + 0.5) a consistent trend from birth to old age was seen. In contrast, the effect of the smoking habit did differ between the laboratories, there being an association of smoking with a significant increase in mutant frequency in the Sussex and Leiden datasets, but not in those from the Vermont or Paris datasets. Possible reasons for this are discussed. One of the objectives of population monitoring is an ability to detect the effect of accidental or environmental exposure to mutagens and carcinogens among exposed persons. The large body of data from non-exposed subjects we have analysed in this paper has enabled us to estimate the size of an effect that could be detected, and the number of individuals required to detect a significant effect, taking known sources of variation into account. We now suggest that, given the variation between subjects that is seen, large sample sizes (30–50 subjects per subject category) may be required to have a 90% chance of detecting increases 1.5-fold above the control level, while rather smaller samples (∼ 20 per group) would be required to detect a 2-fold increase.