The Future of These Studies
We have emphasized in the preceding two chapters that there is considerable error to be attached to the estimate that the doubling dose of chronic or intermittent, low level, low-LET ionizing radiation for both humans and mice is in the neighborhood of 4.0 Sv. Given the continuing concern with the genetic effects of exposures to ionizing radiation, further data bearing on both the estimate and its limits are highly desirable.
CURRENT RESEARCH PROJECTIONS
The additional studies of the F1 that seem to be indicated are of two types. On the one hand, an extension of some of the present studies, plus a reanalysis in due time of some of the data already at hand, should prove fruitful. On the other hand, new genetic technologies that in principle have the potential to yield additional insights are becoming available.
With respect to an extension of present programs, the studies of cancer incidence and mortality among the F1 continue. Furthermore, developments in clinical genetics will surely result in a refinement of the estimate of the contribution of spontaneous mutation to congenital defects and pre-reproductive-age deaths, and hence in an improvement in that aspect of the current doubling dose estimate which depends on these indicators. In this connection, we emphasize again that the estimate that we have generated of the contribution of spontaneous mutation in the parental generation to congenital defects and prereproductive death, namely, between 0.33% and 0.53%, is more-or-less specific to a Japanese population in which during the time span of the investigation some 10% of all pregnancies completing the fifth lunar month of gestation terminated in one of the end points concerned. For instance, unlike the current situation, in post-war Japan Down's
syndrome carried a relatively high mortality. Thus, this figure of 0.33%–0.53% cannot be automatically applied to other populations. Continuing advances, in Japan and elsewhere, in understanding the genetic basis of congenital defect should result in a substantial refinement of any estimate of this type. In particular, progress may be expected in defining additional chromosomal microdeletion syndromes, comparable to such already recognized entities as the Langer-Gideon, Prader-Willi-Angelman, Miller-Dieker, thalassemia/mental retardation, choroideremia/deafness/mental retardation, and DiGeorge Syndromes.
With respect to new technologies that can be brought to bear upon the question, the most obvious direction in which to proceed involves an attempt to extend the genetic studies to the DNA level. For a development of this type, as for any other study of mutation, both parents as well as offspring must be available. At this writing, the modal age of a living parent whose child is in the various study cohorts is 60–70, i.e., the parents are entering a period of high mortality. Recognizing the problem the passage of time was posing for genetic investigations, in 1985 the RERF began to develop as a resource for future genetic studies, a bank of transformed B-lymphocyte cell lines “immortalized” by infection with the Epstein-Barr virus. The current projection is to establish approximately 500 constellations comprised of one or more children and their father and mother for which one or both parents received relatively high radiation exposures ATB, and a matching number of similar constellations not involving parental exposure. An aliquot of untransformed cells is held in reserve for each individual. The donors for these cell lines are being selected for birth years which reflect proportionately the distribution of births to survivors since 1946. This major undertaking should be completed in 1993.
This decision in part dictates the DNA strategy to be pursued. With respect to any particular “gene” (however that be defined), there will be approximately 1,000 copies to scrutinize for mutational events in the children of exposed, and another 1,000 in the children of unexposed. Elsewhere, we have presented calculations to the effect that the spontaneous nucleotide germinal mutation rate is of the order of 1 per 108 nucleotides per generation; we assume a similar rate for relatively small insertion/deletion/rearrangement events (Neel et al., 1986). Given these baselines and the estimate of the human genetic doubling dose employed at that time, and accepting a type I error of 0.05 and a type II error of 0.20, we further estimated that a demonstration of a significant difference between the children of proximally exposed and the children of non-exposed parents would require two samples of approximately 7×109 nucleotides each (Satoh et al., 1990). This was a very approximate calculation, but it did serve to indicate the order of magnitude of any study expected to yield the decisiveness of a statistically significant difference between the two groups of children. With current technology and staffing, these are formidable numbers. It is important to recognize, however, that any substantial body of new data, whether or not it yields statistically significant differences between the two sets of children, can be folded in with the existing data to refine further the understanding of the genetic effects of the atomic bombs, both with respect to the minimal doubling dose at specified probability levels and the most probable dose.
The foregoing calculation suggests the need to scrutinize a relatively large number of nucleotide sequences from each of the 1,000 children represented in the cell line panels if the DNA studies are to make a meaningful contribution to the question of genetic effects. The choice of exactly which sequences should be examined from among the estimated 6× 109 nucleotides of a (diploid) human cell line is a critical decision. Two considerations seem important in this selection. First, since it is highly desirable to be able to translate the results of any new study into the criteria by which the public judges genetic risk—
morbidity and mortality —the genetic probes should predominantly involve functional genes, preferably genes whose normal products can be related to human well-being, although genes associated with presently anonymous polypeptides, such as are visualized by two-dimensional polyacrylamide gel electrophoresis of cellular contents, are acceptable, inasmuch as in time the functions of these polypeptides will become known. Data on mutations in pseudogenes, highly repetitive DNA sequences, or nonsense sequences are of relatively little value in the context of evaluating the phenotypic impact of radiation-induced genetic damage.
Second, since it is now apparent that the genes of higher eukaryotes, including man, vary widely in spontaneous and induced mutation rates (e.g., Searle, 1974; Chakraborty and Neel, 1989; Neel, 1990; Neel and Lewis, 1990), an effort must be made to ensure that the genes whose spontaneous and induced mutation rates are being studied are “representative.” On the assumption that the number of different alleles associated with any given locus is a rough index of locus mutability, “representativeness” translates into selecting probes exhibiting the range of genetic variability of probes in general. The need for many different probes imposed by the study sample sizes discussed earlier will to some extent protect against over-reliance on a few, perhaps atypical probes, but even so, as knowledge of the occurrence of variation at the genome level unfolds, it will be important to develop the battery of probes to be used in any study with an appreciation of probe variability in mutation rate.
With these numerical demands and the limitations on the personnel to be committed to the project, it seems likely that the generation of a substantial data set at RERF will require an effort extending over at least a decade. The DNA technologies are evolving so rapidly that any experimental design adopted today will probably be superseded by a more efficient procedure in the course of the next 3 or 4 years. Nevertheless, it has seemed desirable that a start be made, to provide a standard against which to judge future developments. At this writing the three most efficient techniques for the detection of nucleotide substitutions and/or chromosomal insertions/deletions/rearrangements (I/D/R) appear to be those described by Fischer and Lerman (1979, 1983), by Cotton et al. (1988), and by Mohrenweiser et al. (1989).
The experience of the personnel of RERF engaged in applying the DNA methodologies suggests that with appropriate modifications, the technique of Fischer and Lerman (1979, 1983) should permit the screening of approximately 2.7×108 nucleotide equivalents per technician year (Satoh et al., 1990; Takahashi et al., 1990; Satoh, personal communication). This technique is especially suitable for the detection of nucleotide substitutions but will also detect small I/D/R events (Takahashi et al., in press). On the other hand, an extension of the restriction-enzyme-site-mapping approach of Mohrenweiser et al. (1989), especially oriented toward the detection of larger I/D/R events (but detecting some nucleotide substitutions), should permit the screening of some 109 nucleotide equivalents per technician year (Mohrenweiser, personal communication). Thus on the basis of the foregoing, present technology is such that screening 108 nucleotide equivalents for mutational events with these two techniques requires 5.6 technician months, an estimate which, rough though it has to be at this stage in the development of the technologies, can be used to plan the approximate magnitude of a study designed to reach a given endpoint with current technologies. This estimate, however, does not include the investment of time that was required to collect and immortalize the samples, nor the time to verify variants or to characterize and perform family studies on putative mutations, and to manage the data.
THE LIMITATIONS OF THE NEW TECHNOLOGIES
As currently practiced, the DNA technologies of course will not usually detect the chromosomal aneuploidies that constitute about half of the mutational burden each generation as we have defined it. Furthermore, since as no effort is made at RERF to collect blood samples from the offspring of survivors prior to age 13, many of the larger chromosomal I/D/R types of mutation associated with impaired survival will presumably have been lost from the population. Thus the principal finding in a DNA-oriented program directed at surviving children will be nucleotide substitutions and relatively small I/D/R events. These DNA data, while they have the potential to extend the understanding of the genetic effects of radiation, cannot replace the earlier observations. In this connection, an interesting possibility created by the polymerase chain reaction technique is a return to preserved samples from the 717 infants autopsied in Hiroshima during the period 1948– 1953, to search for transmitted major DNA damage.
It seems important to emphasize that although investigations at the DNA level bring greater precision into the studies of genetic effects than previously, there may be real difficulties in translating the results of such studies into a phenotypic equivalent. Many of the mutations that may be detected in DNA may have no biomedical significance—even mutations in genes encoding for structural, regulatory, or enzymatic proteins. Given the current evidence for genetic regulation of mutation rates, it is entirely possible that noncoding sequences have higher spontaneous and induced mutation rates than the coding sequences; the relevance to human disease of the results of studies directed at the former will be particularly difficult to interpret. While all mutations are of interest to the geneticist, mutations with no fundamental impact are of little concern to a society worried about the deleterious effects of radiation exposure. In our view, then, the full potential of DNA studies will not be realized without a major effort to understand the functional impact of the mutations detected by this methodology. Developing this understanding promises to be difficult.
PARALLEL ANIMAL STUDIES
The issue of whether there should be a study involving an animal model, presumably the mouse, to parallel the human DNA studies, invites extensive discussion. In one sense, the focus has now shifted to human studies, and it is no longer necessary to extrapolate to humans from an animal model. On the other hand, the ability to manipulate the animal model should result in additional insights that cannot be obtained from the human material. At the DNA level, the use of homologous probes in parallel studies on mice could at long last provide truly comparable indicators of a radiation effect for the two species. It would be highly desirable for any study on mice to employ at least one dose rate schedule as nearly comparable as possible to the Nagasaki-Hiroshima experience. Unfortunately, the use of such low doses, low relative to the 3.0, 6.0, or 10.0 Gy gonadal exposures employed in so many of the earlier murine experiments, will require either sample sizes or end-points per mouse far beyond the experiments of the past.
STAYING THE COURSE
It is difficult in the 45th year of this study not to stray into philosophical considerations. The reprints which constitute the bulk of this volume reflect the most major effort to date directed at a single question in the history of human genetics. At what point can it be said, ‘mission accomplished'? Are further extensive studies of any type justified?
The proposed DNA studies will bring a precision into the genetic enterprise that many experimental geneticists will find more satisfying intellectually than the studies of the past. But DNA studies are not the only desirable aspect of future genetic studies. Continuing surveillance of the morbidity and mortality patterns of the F1 are also in order. These latter studies are important not only in their own right, but as a frame of reference through which to view the results of epidemiological studies on radiation effects in other populations.
The most recent example of this latter role for the RERF genetic data is provided by the report by Gardner et al. (1990) that the parents of children developing leukemia prior to age 25 in the West Cumbria district of the United Kingdom are characterized by occupational exposures to radiation in the Sellafield nuclear plant located in this district which are not observed in the parents of matched controls. The sample is small; for instance, the much discussed relative risk of leukemia to their children of 6.4 when the fathers had received an average total preconceptual occupational (chronic) exposure to ionizing radiation of 100 mSv or more, is based upon a total of four fathers. The investigators suggest a genetic explanation for the finding.
This report finds no support in our studies on malignancies in the F1 summarized in Chapter 11, although we recognize that the circumstances of exposure differ in the two studies. In this connection, we again note that, commentary to the contrary notwithstanding (Roberts, 1990), the study on malignancy in the F1 includes the first children conceived following the bombings. The suggestion that the studies in Japan may have “missed” a sensitive period in gametogenesis has no basis. The birth dates of the children comprising the cohorts studied for malignancy extended from May 1946 through 1982, i.e., include the first children conceived following the bombings. During 1946, 263 children were born to parents whose combined exposures were =0.01 Sv, and 1189 to parents receiving <0.01 Sv. In the former group, the mean paternal gonad dose was 0.26 Sv and the mean maternal dose, 0.12 Sv. No cases of leukemia were detected in this group, but one case is on record among the children whose parents received <0.01 Sv. Although we find it difficult to compare precisely the total gonadal dose for the Sellafield cohort of fathers with that of the Japanese cohort, it is clear that the total dose of acute radiation is definitely much higher for the latter than the dose of chronic radiation for the former. Little (1990), equating film badge exposures to gonadal doses, concludes that the Sellafield data imply genetic sensitivities 50 to 80 times higher than the Japanese data. He further concludes that the apparent Sellafield sensitivities are statistically incompatible with the Japanese findings. We believe the gonadal doses are so poorly defined for Sellafield that it is inappropriate to apply precise statistical tests to the situation, but certainly accept the thrust of his analysis. As pointed out by Abrahamson (1990), the Sellafield findings also imply much higher genetic sensitivities than the studies on mice reviewed in Chapter 14.
But the role of the RERF data in providing a continuing perspective to such reports as that of Gardner et al. (1990) goes beyond the specific disease entities that may be involved. These data provide a broad framework of reference within which to view the results of any ad hoc epidemiological study. Radiation is not specific in its genetic effects: the damage that it produces is distributed throughout the genome. The childhood leukemias, unlike Wilms' tumor and retinoblastoma, have not in the past appeared to be malignancies in whose etiology mutation in a parent plays a prominent etiological role (rev. in Chapter 11). Genetic damage sufficient to result in this increase in leukemia should thus have had other effects, reflected in morbidity and mortality, of a magnitude that could not have gone unnoticed in the West Cumbria health district. The totality of the Japanese data can and should be brought to bear in the evaluation of the results of disease-oriented epidemiological studies.
It seems likely at this writing that in the wake of the report by Gardner et al. (1990) there will be a series of epidemiological studies on the potential genetic effects of low-level radiation on humans. While our own experience leads us to entertain many reservations concerning the scientific value of such low-level effects studies, we recognize the political and sociological imperatives that may dictate such undertakings. These additional studies, should they appear to differ in outcome from the investigations summarized in this book, will be perceived by some as challenging the conclusions stemming from the studies in Japan. Accordingly, a clear statement a priori of certain elementary principles might be helpful.
Let us assume that a future study is directed at certain categories of congenital defect or malignancy in children. We suggest that if the broad category (congenital defect or malignancy) reveals no radiation effect, little purpose is served by extending the analysis to subtypes (and this is the principle followed in the analysis of the Japanese material). The emergence of an apparently “significant” finding with respect to a subtype under these circumstances implies a compensatory decrease (below normative expectation) in some other subtype or subtypes, but the latter phenomenon is seldom noted. Furthermore, the likelihood of false positives must always be borne in mind. Forty years ago, Tippett (1952; see also Wakeford et al., 1989) pointed out that if the significance level is set to a for each of n independent tests, and if the test statistics have continuous distributions, then under the null hypothesis the probability (τ) of encountering at least one outcome that is “significant” at the a level is:
Thus, if 20 tests are run on a set of data for which a has been set at 0.05, then in addition to the detection of “true” associations there is also a 0.64 probability of finding one or more “significant” results by chance alone. If a series of independent epidemiological studies were to be undertaken, this expectation applies to each study.
An example of this principle is presented by the study of Sever et al. (1988) on congenital defects in two counties in the southeastern portion of the State of Washington, where the Hanford nuclear facility has been a major employer. There was no significant association of occurrence of defect with either paternal employment at Hanford or the magnitude of the estimated (very low) radiation exposures incident to that employment. However, the analyses were extended to 12 subtypes of defect. With a one-tailed probability test, two of these showed a statistically significant association with employment of parents at Hanford, and one showed an association with the (very low) estimated paternal exposure levels. Different malformations were involved in the three associations. These two analyses cannot be regarded as statistically independent. From Tippett's formula we calculate a two-tailed probability of one or more “significant” findings as 0.46 in either series, so that the observations are well within null hypothesis expectations. Nevertheless, the authors discuss at some length why, in view of the studies in Japan, the associations must be regarded as spurious. It is true that in a question of this importance, one does not wish summarily to discard possible leads (see recent discussion in Rothman, 1990). We suggest that particular consideration in evaluating the outcome of the multiple studies of the future should be given to the consistency of any findings. Specifically, if several independent studies do yield findings of borderline significance, it would lend credence to the “findings” if they involved the same category of malformation or cancer. These same
considerations apply to any epidemiological study of the genetic effects of a chemical mutagen.
Periodically the question surfaces of an extension of these genetic studies to the next generation (F2), and even subsequent generations, on the premise that the recessively inherited effects of radiation would not be manifest in the first filial generation. The biochemical studies described in Chapter 9 were a first approach to the question of recessive damage. The variants detected by electrophoresis and enzyme activity studies, which would not be detected by the kinds of gross phenotypic classification that originally led to the terms “dominant” and “recessive,” fit the classical definitions of recessively inherited traits. The DNA studies would now be the definitive approach to the question of induced “recessive” damage. Carried out on a sufficient scale, the results of such studies, combined with the other data, should render unnecessary any studies on subsequent generations of Japanese. By extrapolation, the results of these studies should provide the data with which to deal in a factual fashion with concerns over the hidden genetic effects of lesser exposures in occupational and other settings.
Abrahamson, S. 1990. Childhood leukemia at Sellafield. Rad. Res. 123:237–238.
Chakraborty, R. and Neel, J.V. 1989. Description and validation of a method for simultaneous estimation of effective population size and mutation rate from human population data. Proc. Natl. Acad. Sci. USA 86:9407–9411.
Cotton, R.G.H., Rodrigues, N.R., and Campbell, R.D. 1988. Reactivity of cytosine and thymine in single-base-pair mismatches with hydroxylamine and osmium tetroxide and its application to the study of mutations. Proc. Natl. Acad. Sci. USA 85:4397– 4401.
Fischer, S.G. and Lerman, L. 1979. Length-independent separation of DNA restriction fragments in two-dimensional gel electrophoresis. Cell 16:191– 200.
Fischer, S.G. and Lerman, L.S. 1983. DNA fragments differing by single base-pair substitutions are separated in denaturing gradient gels: Correspondence with melting theory. Proc. Natl. Acad. Sci. USA 80:1579–1583.
Gardner, M.J., Snee, M.P., Hall, A.J., Powell, C.A., Downes, S., and Terrell, J.D. 1990. Results of case-control study of leukemia and lymphoma among young people near Sellafield nuclear plant in West Cumbria. Brit. Med. J. 300:423–429.
Little, M.P. 1990. A comparison between the risks of childhood leukemia from parental exposure to radiation in the Sellafield work force and those displayed among the Japanese bomb survivors. J. Radiol.Prot. 10:185–198.
Mohrenweiser, H.W., Larsen, R.D., and Neel, J.V. 1989. Development of molecular approaches to estimating germinal mutation rates. I. Detection of insertion/deletion/rearrangement variants in the human genome. Mut. Res. 212:241–252.
Neel, J.V. 1990. Average locus differences in mutability related to protein “class”: a hypothesis. Proc. Natl. Acad. Sci. USA 87:2062–2066.
Neel, J.V. and Lewis, S.E. 1990. The comparative radiation genetics of humans and mice. Annu. Rev. Genet. 24:327–362.
Neel, J.V., Satoh, C., Goriki, K., Fujita, M., Takahashi, N., Asakawa, J., and Hazama, R. 1986. The rate with which spontaneous mutation alters the electrophoretic mobility of polypeptides. Proc. Natl. Acad. Sci. USA 83:389–393.
Roberts, L. 1990. British radiation study throws experts into tizzy. Science 248:24–25.
Rothman, K. 1990. No adjustments are needed for multiple comparisons. Epidemiology 1:43–46.
Satoh, C., Hiyama, K., Takahashi, N., Kodaira, M., and Neel, J.V. 1990. Approaches to DNA methods for the detection of heritable mutations in humans. In: M.L.Mendelsohn and R.J.Albertini (eds.), Mutation and the Environment, Part C (Proc. Fifth Intn. Congr. on Environ. Mut.). New York: Wiley-Liss, Inc.pp. 197–206.
Searle, A.G. 1974. Mutation induction in mice. Ad vances in Radiation Biology 4:131–207. New York: Academic Press.
Sever, L.E., Gilbert, E.S., Hessol, N.A., and McIntyre, J.M. 1988. A case-control study of congenital malformations and occupational exposure to low-level ionizing radiation. Am. J.Epidem. 127:226– 242.
Takahashi, N., Hiyama, K., Kodaira, M., and Satoh, C. 1990. An improved method for the detection of genetic variations in DNA with denaturing gradient gel electrophoresis. Mut. Res. 234:61–70.
Takahashi, N., Hiyama, K., Kodaira, M., and Satoh, C.Detection of the length polymorphism in the 51 flanking region of the human b-globin gene with denaturing gradient gel electrophoresis in a Japanese population. Hum. Genet. , in press.
Tippett, L.H.C. 1952. The Methods of Statistics , 4th Edition . New York: Wiley. pp. 395.
Wakeford, R., Binks, K., and Wilkie, D. 1989. Childhood leukemia and nuclear installations. J.Roy. Stat. Soc. , Series A, 152:61–86.
Yoshimoto, Y., Neel, J.V., Schull, W.J., Kato, H., Mabuchi, K., Soda, M., and Eto, R. 1990. Malignant tumors during the first 2 decades of life in the offspring of atomic bomb survivors. Am. J.Hum. Genet. 46:1041–1052.