| ||||||||||||||||||||||||||||||||
|
|
|||||||||||||||||||||||||||||||
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 91
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
4
Heritable Genetic Effects of Radiation in Human Populations
INTRODUCTION AND BRIEF HISTORY
Naturally occurring mutations in somatic and germ cells contribute respectively to cancers and heritable genetic diseases (i.e., hereditary diseases). The discoveries by Muller (1927) of the mutagenic effects of X-rays in fruit flies (Drosophila) and by Stadler (1928a, 1928b) of similar effects in barley and maize, and the subsequent extension of these findings to other types of ionizing radiation (and also to ultraviolet) and other organisms, conclusively established the genetic damage-inducing effects of radiation. However, widespread and serious concern over the possible adverse genetic effects of exposure of large numbers of people to low levels of radiation first arose in the aftermath of the detonation of atomic bombs over Hiroshima and Nagasaki in World War II, some 20 years after the discoveries of the mutagenic effects of X-rays. In June 1947, at the meeting of the Conference on Genetics convened by the Committee on Atomic Casualties of the U.S. National Research Council to assess the program of research on the heritable effects of radiation to be undertaken in Japan, the leading geneticists voted unanimously to record the following expression of their attitude toward the program: “Although there is every reason to infer that genetic effects can be produced and have been produced in man by atomic radiation, nevertheless the conference wishes to make it clear that it cannot guarantee significant results from this or any other study on the Japanese material. In contrast to laboratory data, this material is too much influenced by extraneous variables and too little adapted to disclosing genetic effects. In spite of these facts, the conference feels that this unique possibility for demonstrating genetic effects caused by atomic radiation should not be lost …” (NRC 1947). Thus came into existence the genetics program in Hiroshima and Nagasaki under the auspices of the Atomic Bomb Casualty Commission (ABCC), the newly formed joint agency of the Japanese Ministry of Health and Welfare and the U.S. National Academy of Sciences. The ABCC was renamed the Radiation Effects Research Foundation in 1976. In the late 1940s, the mouse was chosen as the primary surrogate for assessing the genetic radiosensitivity of humans, and extensive studies were initiated in different research centers in the United States, England, and Japan.
In the mid-1950s, one major international and several national scientific bodies came into existence, including the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR), the Committee on the Biological Effects of Atomic Radiation (the BEAR committee; renamed the Committee on the Biological Effects of Ionizing Radiation [BEIR] in 1972) set up by the U.S. National Academy of Sciences, and the Committee of the British Medical Research Council. The UNSCEAR and the BEIR committees have continued their work up to the present, periodically reviewing the levels of radiation to which human populations are exposed and improving assessment of the somatic and genetic risks of radiation exposure (NRC 1972, 1980, 1988, 1990, 1999; UNSCEAR 1993, 2000b, 2001).
From the beginning of these efforts, it was obvious that in the absence of direct human data on radiation-induced germ cell mutations, quantitative estimates of genetic risk could be derived only through a knowledge of the prevalence of naturally occurring hereditary ill health in the population, the role of spontaneous mutations in supporting this burden, and plausible assumptions on the rates of induced germ cell mutations in humans. The methods developed and used by the above committees for risk estimation, therefore, were necessarily indirect. All were geared toward using human data on genetic diseases as a frame of reference, together with mouse data on radiation-induced mutations, to predict the radiation risk of genetic disease in humans. Both the UNSCEAR and the BEIR committees are cognizant of the need to make assumptions given the consequent uncertainties in extrapolating from mouse data on induced mutation rates to the risk of genetic disease in humans.
Details of the genetics program that evolved in Japan and the vast body of data that emerged from these studies have
OCR for page 92
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
been published in a series of articles. The most relevant ones have now been compiled in a single volume (Neel and Schull 1991). The most important finding of these studies is that there are no statistically demonstrable adverse genetic effects attributable to radiation exposures sustained by the survivors. Although cited and discussed in the UNSCEAR and BEIR reports over the years, these results did not constitute part of the “mainstream thinking” of genetic risk estimators and therefore were not used in risk estimation.
During the past few years, estimates of the baseline frequencies of Mendelian diseases have been revised and mathematical methods have been developed to estimate the impact of an increase in mutation rate (as a result of radiation exposures) on the frequencies of different classes of genetic diseases in the population. Additionally, there have been several advances in our understanding of the molecular basis and mechanisms of origin of human genetic diseases and of radiation-induced mutations in experimental systems. As a result of these developments, it now is possible to reexamine the conceptual basis of risk estimation, reformulate some of the critical questions in the field, and address some of the problems that could not be addressed earlier.
This chapter summarizes the general framework and the methods and assumptions used in risk estimation until the publication of BEIR V (NRC 1990). This is followed by a discussion of the advances in knowledge since that time, their impact on the concepts used in risk estimation, and how they can be employed to revise the risk estimates. Throughout this chapter, the terms “genetic diseases,” “genetic effects,” and “genetic risks” are used exclusively to mean “heritable genetic diseases,” “heritable genetic effects,” and “heritable genetic risks,” respectively.
GENERAL FRAMEWORK
Goal of Genetic Risk Estimation
The goal of genetic risk estimation, at least as envisioned and pursued by UNSCEAR and the BEIR committees, remains prediction of the additional risk of genetic diseases in human populations exposed to ionizing radiation, over and above that which occurs naturally as a result of spontaneous mutations. The concept of “radiation-inducible genetic diseases,” which emerged early on in the field, is based on two established facts and an inference. The facts are that (1) hereditary diseases result from mutations that occur in germ cells and (2) ionizing radiation is capable of inducing similar changes in all experimental systems adequately investigated. The inference, therefore, has been that radiation exposure of human germ cells can result in an increase in the frequency of genetic diseases in the population. Worth noting is the fact that although there is a vast amount of evidence for radiation-induced mutations in diverse biological systems, there is no evidence for radiation-induced germ cell mutations that cause genetic disease in humans.
Germ Cell Stages and Radiation Conditions of Relevance
From the standpoint of genetic risks, the effects of radiation on two germ cell stages are particularly important. In the male, these are the stem cell spermatogonia, which constitute a permanent germ cell population in the testes and continue to multiply throughout the reproductive life span of the individual. In the female, the corresponding cell stages are the oocytes, primarily the immature ones. The latter constitute the predominant germ cell population in the female. Female mammals are born with a finite number of oocytes formed during fetal development. These primordial oocytes, as they are called, grow, and a sequence of nuclear changes comprising meiosis takes place in them. The latter however are arrested at a particular stage until just before ovulation. Because oocytes are not replenished by mitosis during adult life and immature oocytes are the predominant germ cell population in the female, these are clearly the cell stages whose irradiation has great significance for genetic risks.
The radiation exposures sustained by germ cells in human populations are generally in the form of low-LET (linear energy transfer) irradiation (e.g., X-rays and γ-rays) delivered as small doses at high dose rates (e.g., in diagnostic radiology) or are greatly protracted (e.g., continuous exposures from natural and man-made sources). In estimating genetic risks to the population therefore, the relevant radiation conditions are low or chronic doses of low-LET irradiation. As discussed later, most mouse data used for estimating the rates of induced mutations have been collected at high doses and high dose rates. Consequently, assumptions have to be made to convert the rates of induced mutations at high doses and dose rates into mutation rates for radiation conditions applicable for risk estimation in humans.
GENETIC DISEASES
Since the aim of genetic risk estimation is to predict the additional risk of genetic diseases relative to the baseline frequency of such diseases in the population, the concept of genetic diseases and their classification and attributes are considered in this section. The term genetic diseases refers to those that arise as a result of spontaneous mutations in germ cells and are transmitted to the progeny.
Mendelian Diseases
Diseases caused by mutations in single genes are known as Mendelian diseases and are further divided into autosomal dominant, autosomal recessive, and X-linked, depending on the chromosomal location (autosomes or the X chromosome) and transmission patterns of the mutant genes. In an autosomal dominant disease, a single mutant gene (i.e., in the heterozygous state) is sufficient to cause disease. Examples include achondroplasia, neurofibromatosis, Marfan syndrome, and myotonic dystrophy. Autosomal recessive
OCR for page 93
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
diseases require homozygosity (i.e., two mutant genes at the same locus, one from each parent) for disease manifestation. Examples include cystic fibrosis, phenylketonuria, hemochromatosis, Bloom’s syndrome, and ataxia-telangietasia.
The X-linked recessive diseases are due to mutations in genes located on the X chromosome and include Duchenne’s muscular dystrophy, Fabry’s disease, steroid sulfatase deficiency, and ocular albinism. Some X-linked dominant diseases are known, but for most of them, no data on incidence estimates are currently available. Therefore, these diseases are not considered further in this report. The general point with respect to Mendelian diseases is that the relationship between mutation and disease is simple and predictable.
Multifactorial Diseases
The major burden of naturally occurring genetic diseases in human populations, however, is not constituted by Mendelian diseases, which are rare, but by those that have a complex etiology. The term “multifactorial” is used to designate these diseases to emphasize the fact that there are multiple genetic and environmental determinants in their etiology. Their transmission patterns do not fit Mendelian expectations. Examples of multifactorial diseases include the common congenital abnormalities such as neural tube defects, cleft lip with or without cleft palate, and congenital heart defects that are present at birth, and chronic diseases of adults (i.e., with onset in middle and later years of life) such as coronary heart disease, essential hypertension, and diabetes mellitus.
Evidence for a genetic component in their etiology comes from family and twin studies. For example, first-degree relatives of patients affected with coronary heart disease have a two- to sixfold higher risk of the disease than those of matched controls, and the concordance rates of disease for monozygotic twins are higher (but never 100%) than those for dizygotic twins (Motulsky and Brunzell 1992; Sankaranarayanan and others 1999).
As mentioned earlier, multifactorial diseases are presumed to originate from the joint action of multiple genetic and environmental factors; consequently, the presence of a mutant allele is not equivalent to having the disease. For these diseases, the interrelated concepts of genetic susceptibility and risk factors are more appropriate. The genetic basis of a common multifactorial disease is the presence of a genetically susceptible individual, who may or may not develop the disease depending on the interaction with other genetic and environmental factors. These concepts are discussed further in Annex 4A. The important general point is that unlike the situation with Mendelian diseases, the relationships between mutations and disease are complex in the case of multifactorial diseases. For most of them, knowledge of the genes involved, the types of mutational alterations, and the nature of environmental factors remains limited. Among the models used to explain the inheritance patterns of multifactorial diseases and to estimate the recurrence risks in relatives is the multifactorial threshold model (MTM) of disease liability. The MTM, its properties, and its predictions are discussed in Annex 4A.
Chromosomal Diseases
Historically, both UNSCEAR and the BEIR committees have always had an additional class of genetic diseases—“chromosomal diseases”—in their lists that included those that had long been known to arise as a result of gross (i.e., microscopically detectable), numerical (e.g., Down’s syndrome, which is due to trisomy of chromosome 21), or structural abnormalities of chromosomes (e.g., cri du chat syndrome, due to deletion of part or the whole short arm of chromosome 5 [5p-]). As discussed later, this is really not an etiological category, and deletions (microscopically detectable or not) are now known to contribute to a number of constitutional genetic diseases grouped under autosomal dominant, autosomal recessive, and X-linked diseases.
RISK ESTIMATION METHODS
In the absence of data on radiation-induced germ cell mutations that can cause genetic disease in humans, all of the methods developed and used for predicting the risk of genetic disease from the mid-1950s to the present are indirect. Their strengths and weaknesses are reviewed in BEIR V (NRC 1990). One such indirect method is the doubling dose method, on which attention is focused in this section. It has been in use since the early 1970s (NRC 1972, 1990; UNSCEAR 1977, 1982, 1986, 1988) and is used in the recent UNSCEAR (2001) report.
The Doubling Dose Method
The doubling dose method enables expressing of the expected increase in disease frequency per unit dose of radiation in terms of the baseline frequency of the disease class. The doubling dose (DD) is the amount of radiation required to produce in a generation as many mutations as those that arise spontaneously. Ideally, it is estimated as a ratio of the average rates of spontaneous and induced mutations in a given set of genes:
(4-1)
The reciprocal of the DD (i.e., 1/DD) is the relative mutation risk (RMR) per unit dose. Since RMR is the reciprocal of DD, the smaller the DD, the higher is the RMR and vice versa. With the doubling dose method, until recently, risk was estimated as a product of two quantities—namely, the baseline disease frequency, P, and 1/DD:
(4-2)
OCR for page 94
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
The population genetic theory that underlies the use of Equation (4-2) is the equilibrium theory that population geneticists use to explain the dynamics of mutant genes in populations. The theory assumes that the stability of mutant gene frequencies (and thus disease frequencies) in a population is the result of the existence of a balance between the rates at which spontaneous mutations enter the gene pool in every generation and the rate at which they are eliminated by natural selection (i.e., through failure of survival or reproduction).
When the mutation rate is increased as a result of radiation in every generation, this balance between mutation and selection is disturbed by the influx of induced mutations, but the prediction is that the population will attain a new equilibrium (over a number of generations) between mutation and selection. The amount of increase in mutation frequency, the time it takes for the population to reach the new equilibrium, and the rate of approach to equilibrium are all dependent on induced mutation rates, the intensity of selection, the type of genetic disease, and whether the radiation exposure occurs in one generation only or generation after generation. It should be noted that since the starting population (before radiation exposure) is assumed to be in equilibrium between mutation and selection, the quantity P in Equation (4-2) represents the equilibrium incidence of the disease, and the product of P and 1/DD is the expected increase in disease frequency at the new equilibrium.
Risk Estimation for Different Classes of Genetic Disease
The application of Equation (4-2) to risk estimation is straightforward for autosomal dominant diseases since the relationship between mutation and disease is simple for this class of diseases. Population genetic theory predicts that for these diseases, if there is an x% increase in mutation rate in every generation, at the new equilibrium this increase will be reflected as an x% increase in the frequency of these diseases. Until recently, estimates of risk for the first, second, or any postradiation generation of interest were obtained through “back calculation” from the predicted new equilibrium incidence using certain assumptions. If the population sustains radiation exposure in one generation only, there will be a transient increase in the mutant frequency in the first postradiation generation, followed by a progressive decline to the “old” equilibrium value.
The method used to predict the risk of X-linked diseases is approximately similar to that for autosomal dominant diseases discussed above. For autosomal recessive diseases, the risk calculation is more involved because when recessive mutations first arise (or are induced), they are present in the heterozygous state and do not precipitate disease in children of the first few postradiation generations. For multifactorial diseases, the situation is complex in that there is no simple relationship between mutation and disease, and as discussed later, the estimate of risk will depend on the model used for their maintenance in the population.
The Concept of Mutation Component
The concept of mutation component and the statistic MC, which is derived using this concept, help to unify attempts at predicting how the frequencies of different classes of genetic diseases in the population will change as a result of increases in mutation rate. The mutation component is defined as the relative increase in disease frequency (i.e., relative to the baseline frequency) per unit relative increase in mutation rate (i.e., relative to the spontaneous mutation rate). First introduced in BEIR I (NRC 1972) to address the problem of the impact of the radiation risk of multifactorial diseases in the population, and subsequently elaborated by Crow and Denniston (1981, 1985) and Denniston (1983), the concept can be used for all classes of genetic disease as done in BEIR V (NRC 1990). During the past few years, the concept has been developed further with the necessary algebraic formulations, that permit a direct evaluation of the impact of an increase in mutation rate for all classes of genetic disease in any postradiation generation of interest following exposure to radiation in either one generation only or generation after generation (Chakraborty and others 1998a; Denniston and others 1998). These advances are considered in a later section. Suffice to note here that the inclusion of MC in Equation (4-2) yields the revised equation:
(4-3)
where MC is the disease class and postradiation generation-specific mutation component and the other two quantities are as defined earlier.
RECENT ADVANCES WITH RESPECT TO THE THREE QUANTITIES USED WITH THE DD METHOD OF RISK ESTIMATION
The BEIR V report (NRC 1990) reviewed the advances that occurred from the mid-1950s to 1990 with respect to P, the baseline frequency of genetic disease, DD, and MC, the three quantities considered relevant for risk estimation with the DD method thus far. In the material that follows, attention is focused on progress made since 1990.
Baseline Frequencies of Genetic Diseases
Mendelian Diseases
Estimates of the baseline frequencies of Mendelian diseases used by UNSCEAR since its 1977 report and by the BEIR III and BEIR V committees (NRC 1980, 1990) have been based on the compilations and analysis of Carter (1976a, 1976b) primarily for Western European and Western
OCR for page 95
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
European-derived populations. These are the following (all in live births): autosomal dominants, 0.95%; X-linked, 0.05%; and autosomal recessive, 0.25%. Advances in human genetics during the past two decades now permit an upward revision of the above estimates to 1.5% for autosomal dominant diseases, 0.15% for X-linked diseases, and 0.75% for autosomal recessive diseases (Sankaranarayanan 1998). Note that the revised total frequency of Mendelian diseases is thus 2.4%, which is about twice the earlier figure of 1.25%.
Multifactorial Diseases
For multifactorial diseases (which include congenital abnormalities present at birth and chronic diseases), the estimates used by UNSCEAR (1986, 1988, 1993, 2001) derive from data obtained for the population of Hungary (Czeizel and Sankaranarayanan 1984; Czeizel and others 1988). These estimates are 6% of live births for congenital abnormalities and 65% of the population affected by chronic diseases (excluding cancers). Since most chronic diseases have their onset in middle and late ages (published figures pertain to these age groups), data on the distribution of the population in various age intervals (i.e., ages 0, 1, 2, 3–4, 5–9, 10–14, … 80–84, 85+, etc.; a total of 21 age intervals) for 1977 to 1981 were used to obtain estimates applicable to the population as a whole. For example, if the published estimate for a given disease pertains to the adult population (i.e., above age 14), the figure was reduced by 21% since the 0–14 year age group constituted 21% of the total population of 10.7 million (Czeizel and others 1988).
For the BEIR V committee (NRC 1990), the starting point for congenital abnormalities was the published data of Czeizel and Sankaranarayanan (1984) and Czeizel and others (1988), which gave an incidence estimate of 6%. This figure was reduced to 2–3% by noting that the 6% figure is “… so high, in part, because of the unusually high frequency of congenital dislocation of the hip in Hungary” (Czeizel and Sankaranaryanan 1984). For chronic diseases, the starting point was the estimate of about 60% based on preliminary data of Czeizel and colleagues made available to and used by UNSCEAR in its 1988 report. The BEIR V committee reduced the figure of 60% to 30% by (1) subtracting the estimates for essential hypertension, acute myocardial infarction, other acute and subacute forms of ischemic heart disease, and varicose veins of the lower extremities (together about 25%) and (2) reducing the figure for juvenile osteochondrosis of the spine from 11% (based on radiographic screening) to about 0.5% (on the assumption that only about 5% of the cases identified by radiographic screening may be deemed to be of clinical significance). The resulting adjusted figure of about 30% was given as the estimate for the “selected others” subgroup of “other diseases of complex etiology.” Together with the earlier committee’s figures for heart disease (60%) and cancer (30%; which were termed “round number approximations” for all varieties of the above diseases), the total became 120%. Footnote f to Table 2-5 of the BEIR V report (NRC 1990) offers the following explanation for the 120% figure: “Includes heart disease, cancer, and other selected disorders…. Note that the total exceeds 100%. The genetic component in many of these traits is unknown. To the extent that genetic influences are important, the effects are through genes that have small individual effects but that act cumulatively among themselves and in combination with environment factors to increase susceptibility.”
Estimates of Baseline Frequency of Multifactorial Diseases Used in This Report
In examining what would be considered a reasonable estimate of baseline frequency of congenital abnormalities for use in risk estimation, the BEIR VII committee took note of the vast body of data on their prevalence in different parts of the world, including some large-scale studies carried out in North America (Myrianthopoulos and Chung 1974; Trimble and Doughty 1974; Baird and others 1988). The estimates vary over a wide range, from about 1% in live births to a high of about 8.5% in total births (i.e., still- and live births), depending on, among other things, the definition, classification, and diagnostic criteria; entities included; method of ascertainment; duration of follow-up of live-born children; and sample sizes. In one of the largest U.S. studies (Myrianthopoulos and Chung 1974), the overall frequency of major abnormalities was 8.3% (53,257 deliveries of known outcome), which compares favorably with the estimate of about 6% from British Columbia (Baird and others 1988) and of about 6% from Hungary mentioned earlier. This documents the premise that under conditions of good ascertainment, the overall prevalences are similar and are of the order of about 6%. This committee therefore accepts the 6% figure as reasonable for use in risk estimation in this report.
For chronic multifactorial diseases, the committee prefers to use the estimate of 65% obtained by Czeizel and colleagues (1988) in view of the fact that the estimate is based on 26 clear-cut disease entities defined by ICD (International Classification of Diseases) code numbers that were studied epidemiologically in a large population. This estimate was also used by UNSCEAR (1988, 1993, 2001) as the best available overall estimate for chronic diseases as a whole (excluding cancers). Included in the above estimate are heart or blood vessel-related diseases, together, about 25%. For the estimate of 60% mentioned in BEIR V (NRC 1990) under the heading “heart disease” no verifiable source or study is cited. Likewise, for cancers, BEIR V cites an estimate of 30%, again with no citation of the source or the types of cancers included. As mentioned earlier, both of these numbers represent round number approximations.
In the view of the BEIR VII committee, the inclusion of cancers in estimating the heritable risks of radiation is not meaningful at the present state of knowledge.
OCR for page 96
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
Estimates of Baseline Frequency of Chromosomal Disease
The BEIR V report (NRC 1990) and the UNSCEAR (1993) report assessed the baseline prevalence of chromosomal diseases to be of the order of about 0.4% in live births. The present committee sees no reason to alter this estimate.
Summary of Current Estimates of Baseline Frequencies of Genetic Diseases and Comparison with Those in BEIR V
Table 4-1 presents these comparisons showing that the current estimates for Mendelian diseases are higher than those used in 1990, while those for the other classes remain essentially unchanged.
The Doubling Dose
As discussed earlier DD is one of the important quantities used in the equation for the doubling dose method of risk estimation. Although the DD concept was formulated by Muller (1951, 1954, 1959) in the 1950s and several possible estimates and/or ranges of DDs were discussed in the BEAR report (NRC 1956), in UNSCEAR (1962), and in Lüning and Searle (1971), actual use of the method to obtain quantitative estimates of risk began only in 1972 (NRC 1972). Changes in the conceptual basis and database used for DD estimates from the mid-1950s to the early 1990s have recently been reviewed (Sankaranarayanan and Chakraborty 2000a).
TABLE 4-1 Estimates of the Baseline Prevalences of Genetic Diseases Used in BEIR VII and BEIR V
Disease Class
Baseline Prevalence Estimates per 106 Live Births
BEIR VII
BEIR V
Mendelian
Autosomal dominant
15,000
10,000
X-linked
1500
400
Autosomal recessive
7500
2,500
Chromosomal
~4000
~4000
Multifactorial
Congenital abnormalities
60,000
20,000–30,000
Chronic multifactorial
650,000
a
Other Disorders of Complex Etiology
Heart disease
b
600,000
Cancer
c
300,000
Selected others
b
300,000
aBEIR V included these diseases under “other disorders of complex etiology.”
bIncluded under chronic multifactorial diseases in BEIR VII.
cNot specifically considered in this chapter.
SOURCE: Table reproduced with permission from Chakraborty and others (1998b).
Table 4B-1 (see Annex 4B) summarizes the important developments. As evident from that Table, with one exception, most of the DD estimates used in risk estimation by UNSCEAR and the BEIR committees were based on data on both spontaneous and induced mutation rates in mice. The one exception was BEIR I (NRC 1972), which used data on spontaneous rate of mutations of human genes and induced rate of mutations in mouse genes. As discussed below, reevaluation of the assumptions underlying the use of mouse data on spontaneous mutation rate for DD calculations has shown that these are incorrect and that the use of human data on spontaneous mutation rates along with mouse data on induced rates is correct.
Incorrectness of the Assumption of Similarity of Spontaneous Mutation Rates in Mice and Humans—The Need to Use Human Spontaneous Mutation Rates for DD Calculations
Extrapolation of the mouse-based DD to humans for risk estimation implies the assumption that both the spontaneous and the induced rates of mutations are similar in the two species. The assumption of similarity of induced rates of mutations in both species is defensible on the grounds of generally similar gene organization, 70–90% homology in DNA sequence of genes, and substantial conservation of synteny for many chromosomal regions between humans and mice. However, the situation is different with respect to spontaneous mutations.
The reasons spontaneous mutation rates in humans are unlikely to be similar to those in mice have been discussed (Sankaranarayanan 1998). Briefly, these have to do with the differences in the number of cell divisions between the zygote and the mature germ cell in the two species. Vogel and Motulsky (1997) estimate that in human females, the number of cell divisions from zygote to the mature egg (Nf) is of the order of about 24. For the mouse female, estimates of Drost and Lee (1995) suggest that Nf is of the same order. So, from the standpoint of Nf, human and mouse females are similar.
In human males, however, the comparable number of cell divisions is much higher; it is about 30 until the age of puberty (taken to be 15 years), ~23 per year thereafter, and 6 for proliferation and meiosis. Thus, the number of cell divisions prior to sperm production (Nm) in a 20-year-old male can be estimated to be 30 + (5 × 23) + 6 = 151, increasing to 381 at age 30 years, 611 at age 40 years, and 841 at age 50 years (Crow 1999). The Nm/Nf thus increases with paternal age, being 6.3 at age 20, 15.9 at age 30, 25.5 at age 40, and 35.0 at age 50. In the male mouse, the number of cell divisions from zygote to sperm is of the order of about 62 at age 9 months, assuming a 9-month generation (Chang and others 1994; Drost and Lee 1995; Li and others 1996). The Nm/Nf ratio in the mouse is therefore 2.5 (i.e., 62/25), which is much lower than in humans. The committee notes that in
OCR for page 97
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
most mouse experiments, the parental animals were used at a rather uniform age (usually about 12 weeks), and the question of paternal age effects has not been specifically addressed.
Since most spontaneous mutations arise as a result of errors in DNA replication, one would expect that the mutation rate in human males would be higher than that in females and that there would be an increase in the likelihood of spontaneous germinal mutations with the age of the male (so-called paternal age effect). By and large, these expectations have been fulfilled. The literature on this subject and the recent evidence from molecular studies have been reviewed (Crow and Denniston 1985; Crow 1993, 1997, 1999; Vogel and Motulsky 1997; Sankaranarayanan 1998; Green and others 1999).
When one considers the large differences in life span between humans and mice and the paternal age effect for spontaneous mutations in humans, it is clear that extrapolation from short-lived mice to humans is unlikely to provide a reliable average spontaneous rate in a heterogeneous human population of all ages. This is one reason to abandon the use of the mouse data on spontaneous mutation rates in DD calculations and to use human spontaneous mutation rates instead. The following arguments support this: (1) estimates of spontaneous mutation rates in humans are unweighted averages of the rates in the two sexes (and therefore automatically incorporate sex differences and paternal age effects), and (2) the sex-averaged rate is relevant in the context of DD calculations (Sankaranarayanan 1998).
A second reason for not using the mouse spontaneous mutation rates for DD calculations is that the whole question of spontaneous mutation rates in mice has now assumed an unexpected complexity due to the noninclusion, until recently, of mutations that originated as germinal mosaics (resulting in progeny carrying the same mutation [“clusters”] in the following generation) in estimates of spontaneous mutation rates in the specific locus experiments (Russell and Russell 1996; Selby 1998a, 1998b; Russell 1999). According to Russell and Russell (1996), if mosaic data are included, the total spontaneous rate becomes twice that of 6.6 × 10−6 per gene based on mutations that arose singly. However, Selby (1998a, 1998b) has argued that (1) the data on clusters should be included in calculating the total spontaneous mutation rate; (2) his computer simulation studies (which incorporate clusters in his model) suggest an increase of the rate by a factor of about 5 compared to that based on mutations that arose singly; (3) the fivefold higher total spontaneous rate is the appropriate numerator in DD calculations; and (4) if paternal age effects are extrapolated from humans to mice, the estimate of spontaneous rate is even higher. In the view of this committee, the above argument cannot be sustained for humans for the following reasons:
First, while there is no doubt that a proportion of spontaneous mutations in human genes arise as germinal mosaics (and can potentially result in clusters in the following generation), the limited data available on mosaics and clusters at present preclude a quantitative assessment of their contribution to spontaneous mutation rates. The main relevance of germinal mosaicism in the human context is this: the parent who carries a mosaic mutation for an autosomal dominant or X-linked trait does not have a mutant phenotype and therefore would not be considered as having a risk of producing affected children. However, because his or her gonads contain mutant and normal cells, he or she may run the risk of having more than one progeny who carries the mutant gene (mutational “clusters”).
Second, if a substantial proportion of human mutations arise as germinal mosaics in one generation and result in clusters in the following generation, the frequencies of at least autosomal dominant and X-linked diseases also have to be corrected upwards to account for this possibility; there is no reliable way of doing this at present. The published estimates of human spontaneous mutation rates do not provide sufficient grounds for assuming that substantial proportions of mutations in the germ cells first arose as mosaics and subsequently resulted in clusters of mutations; if this had been the case, major increases in the frequencies of affected individuals from one generation to the next would have been observed, but this does not appear to be true. Further, family sizes in present-day human populations are limited (in fact, they are so small that there is almost never more than one affected offspring from a mating, in contrast to the situation in mice where large numbers of progeny are obtained from a single male). Both of these arguments support the view that mutational clusters are much less relevant in humans than in mice.
The advantages of using human spontaneous mutation rates for DD calculations are (1) they pertain to human disease-causing genes; (2) as mentioned earlier, the mutation rate estimates in humans, because they are averaged over both sexes, automatically include sex differences and paternal age effects; and (3) in mutation rate calculations, human geneticists count all mutants that arise anew irrespective of whether they were part of a cluster or not; if clusters had occurred, they would have been included. The committee therefore accepts the view that the use of human spontaneous rates and mouse induced rates for DD calculations (i.e., the procedure used in BEIR I; NRC 1972) is more logical, and it has assessed published data on spontaneous mutation rate in humans and induced rates of mutations in mice.
Doubling Dose Estimation Using Spontaneous Mutation Rates of Human Genes and Induced Rates of Mouse Genes
Estimation of the Average Spontaneous Mutation Rate of Human Genes
To calculate a representative average spontaneous mutation rate of human genes, the available estimates for indi-
OCR for page 98
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
vidual autosomal dominant diseases published by Childs (1981) and Vogel and Motulsky (1997) were used, irrespective of whether these diseases have high or low prevalence or high or low mutation rates. However, the analysis took into account the numbers of genes thus far known or estimated to underlie each of these disease phenotypes (Vogel and Motulsky 1997; Sankaranarayanan 1998; McKusick 2000). This represents an important departure from earlier estimates based on disease phenotypes alone, which generally assumed a one-to-one relationship between mutation and disease. Details of these diseases, estimates of mutation rates, and selection coefficients are given in Table 4-2. The (unweighted) average mutation rate derived from these data (for some 26 autosomal dominant phenotypes with an esti
TABLE 4-2 Database for Estimating Average Spontaneous Mutation Rate of Human Autosomal Genes Associated with Autosomal Dominant Diseases and Their Selection Coefficients(s)
Disease Phenotype
Estimated
No. of Loci
Mutation Rate (× 106)a
Selection Coefficient(s)b
Achondroplasia
1
11.0
0.8
Amelogenesis imperfecta
1
1.0
0
Aniridia
2
3.8
0.1
Apert’s syndrome
1
3.5
0
Blindness
9
10.0
0.7
Cataracts (early onset)
30
6.0
0.3
Cleft lip
1
1.0
0.2
Deaf mutism
15
24.0
0.7
Dentinogenesis imperfecta
2
1.0
0
Huntington disease
1
5.0
0.2
Hypercholesterolemia
1
20.0
0
Marfan syndrome
1
5.0
0.3
Multiple exotoses
3
7.7
0.3
Myotonic dystrophy
1
18.0
0.3
Neurofibromatosis
2
70.0
0.5
Osteogenesis imperfecta
2
10.0
0.4
Osteopetrosis
1
1.0
0.2
Otosclerosis
1
20.0
0
Polyposis of intestine
1
10.0
0.2
Polycystic kidney disease
2
87.5
0.2
Porphyria
2
1.0
0.05
Primary basilar impression
1
10.0
0.2
Rare diseases (early onset)
50
30.0
0.5
Retinoblastoma
1
8.7
0.5
Spherocytosis
1
22.0
0.2
Tuberous sclerosis
2
8.0
0.8
Total
135
Average
( 2.95 ± 0.64)
0.294
aFor some entries, mutation rate estimates are uncertain (see Childs 1981 for details).
bEstimated from reproductive fitness.
SOURCE: Childs (1981); Vogel and Motulsky (1997).
mated 135 loci) is (2.95 ± 0.64) × 10−6 per locus per generation. This figure is within the range of 0.5 × 10−5 to 0.5 × 10−6 per locus used in the 1972 BEIR I report (NRC 1972).
The list of autosomal dominant diseases used to provide the basis for the prevalence estimate (P in Equation (4-3)) encompasses many more than the 26 diseases used in the above calculations (Sankaranarayanan 1998); these other diseases could not be included in the present analysis because of lack of information on mutation rates. Further, the mutation rate estimates for X-linked phenotypes have not been included in these calculations; instead, it has been assumed that the average spontaneous mutation rate for autosomal dominant genes calculated above can also be used for X-linked genes. The justification for this assumption rests on the following lines of reasoning: (1) among Mendelian diseases, autosomal dominants constitute the most important group from the standpoint of genetic risks, and (2) although X-linked recessive diseases are also expected to respond directly to an increase in mutation rate, since their prevalence is an order of magnitude lower than that of autosomal dominants (i.e., 0.15% versus 1.5%) the assumption of similar spontaneous rates of mutations for autosomal dominants and X-linked recessives is unlikely to result in any significant underestimation of the total risk. In fact, for this reason, these two classes of diseases are considered together in risk estimation.
The Average Rate of Induced Mutations in Mice
To calculate the average rate of induced mutations in mice, the committee used all available data on rates of induced mutations in defined genes in mice; these relate to recessive specific locus mutations at 12 loci, biochemical mutations (null enzyme mutations, also recessive at a large number of loci), and autosomal dominant mutations at 4 loci incidentally detected in the course of the specific locus experiments. The data on these autosomal dominant mutations are all from studies carried out in Harwell; comparable data from Oak Ridge studies were unavailable. Inclusion of the data on dominant mutations in mutation rate calculations was dictated by the consideration that although the underlying genes were not well defined at the time these experiments were performed (but mutations were “frequently” observed and recorded, indicating that they were among the more radiation-mutable loci), we now know not only their identity (and the molecular nature of the mutations) but also their human counterparts (the mouse Sl, W, Sp, and T correspond to, respectively, the MGF, KIT, PAX3, and T genes in humans; see McKusick 2000). All of the data considered here come from experiments involving stem cell spermatogonia.
The data from female mice have not been used because there is uncertainty about whether mouse immature oocytes are a good model for assessing the mutational radiosensitivity of human immature oocytes (UNSCEAR 1988). The arguments rest on (1) the strikingly higher sensitivity of mouse
OCR for page 99
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
immature oocytes to radiation-induced killing (the majority are destroyed by 0.5 Gy; Oakberg and Clark 1964) in contrast to those of human and rhesus monkey immature oocytes, for which the dose required is at least 100 times higher (Baker 1971) and (2) the observations that no mutations were recovered from oocytes sampled 7 weeks after irradiation in contrast to the situation with mature and maturing oocytes (Russell 1965). In view of this uncertainty and in order not to underestimate the risk, the committee has used the assumption that the rate estimated for males will also be applicable to females.
Details of the data used are summarized in Tables 4-3A to 4-3C and are from experiments involving acute X-irradiation or from high-dose fractionated X-irradiation (usually two fractions separated by 24 h) appropriately normalized to acute X-irradiation conditions (see Table 4-3A, footnote d; and Table 4-3B, footnotes a and b) to permit easy comparisons. Table 4-3A shows that the average rate of induced mutations is highest at the original seven specific loci (3.03 × 10−5 per locus per gray) and is about one-third of the above at the six loci used in the experiments of Lyon and Morris (1969; i.e., 0.78 × 10−5 per locus per gray; one locus, a, is common to both sets). For various sets of biochemical loci at which null mutations have been scored, the estimates vary over a range from 0.24 × 10−5 to 1.64 × 10−5 per locus per gray. The average rate for dominant visible mutations is within the above range. The unweighted average of the induced mutation rates is 1.09 × 10−5 per locus per gray for acute irradiation. The use of this rate for DD calculations, however, is somewhat problematic since (1) there is overlap of one or more loci in different data sets; (2) in some studies (see footnote e, Table 4-3A), all of the loci involved could not be ascertained; and (3) there is no simple way of taking into account the interlocus variation and sampling variance of induced rates from the derived average estimate of 1.09 × 10−5 per locus per gray.
TABLE 4-3A Database for Calculating Rates of Induced Mutations in Mice
System
No. of Loci
Average Rate/Locus/Gy (× 105)
Reference
1. The 7-locus system (Lyon and others 1964) (3 and 6 Gy; acute X- or γ-irradiation or 3 + 3 Gy, 24 h interval)
7a
3.03
Phillips (1961);
Russell (1965, 1968);
Lyon and others (1972);
Cattanach and Rasberry (1994);
Pretsch and others (1994)
2. The 6-locus system (Lyon and others 1964) (6 Gy; acute X-irradiation)
6b
0.78
Lyon and Morris (1969)
3. Biochemical loci (recessive, null enzyme) (3 + 3 Gy, 24 h interval; X-rays)
12c
0.70d
Charles and Pretsch (1986);
Pretsch and others (1994)
4. Biochemical loci (recessive, null enzyme) (3 Gy, 3 + 3 Gy, 24 h interval and 6 Gy; X-rays)
32e
32
32
1.64
0.67d
0.24
Unpublished data of S.E. Lewis, cited in Neel and Lewis (1990)
5. Biochemical loci (recessive, null enzyme) (3 + 3 Gy, 24 h interval; X-rays)
4f
1.24d
Unpublished data of J. Peters, cited in Neel and Lewis (1990)
6. Dominant visibles (Sl, W, Sp and T)g (X rays)
4
0.44
See Table 4-3B
Unweighted average: 8.74/8 = 1.09 × 10−5 per locus per gray
NOTE: Data are from experiments involving irradiation of males (stem cell spermatogonia) and all rates are normalized to single acute X-irradiation conditions.
aa: non-agouti; b: brown; c: chinchilla; d: dilute; p: pink-eyed dilution; s: piebald; se: short ear; in the work of Pretsch and others (1994), with some strains, mutations at four or five of these loci were scored.
ba: non-agouti; bp: brachypodism; fz: fuzzy; ln: leaden; pa: pallid; pe: pearl.
cLdh1, Tpi, Gpi1, Pgk, G6pd1, G6pd2, Pk, Gr, Mod1, Pgam, Gapdh, Ldr.
dNormalized assuming additivity of the effect of dose fractionation.
eAcy1, Car2, G6pd1, Ggc, Es1, Es3, G6pd1, Gpi1, Hba, Hbb, Idh1, Ldh1, Ldh2, Mod1, Mod2, Np1, Pep2, Pep3, Pep7, Pgm1, Pgm2, Pgm3, Pk3, Trf (the identity of the other 8 loci could not be ascertained).
fHba, Hbb, Es3, Gpi1.
gSl: steel; W: dominant spotting; Sp: splotch; T: brachyury.
OCR for page 100
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
TABLE 4-3B Dominant Visible Mutations Recovered in the Course of Mouse Specific Locus Experiments (Spermatogonial Irradiation)
Expt No.
X-ray Dose (Gy)
Number of Progeny
Number of mutations at
Mutations per Locus per Gray (× 105)
Reference
Sl
W
Sp
T
Total
1
6 + 6 (8-week interval)
3,612
1
—
—
—
1
0.58a
Lyon and others (1964)
2
6
16,735
—
1
—
—
1
0.25
Lyon and Morris (1969)
3
5 + 5
7,168
1
—
—
—
1
0.35a
Cattanach and Moseley (1974)
Cattanach and others (1985)
4
3 + 3
7,645
2
—
—
—
2
1.09a
Cattanach and Rasberry (1994)
Cattanach and others (1985)
5
3 + 3
15,849
1
1
1
3
6
0.35b
Cattanach and Rasberry (1994)
Cattanach and others (1985)
6
6
10,897
1
—
—
—
1
0.38
Cattanach and Rasberry (1994)
7
6
19,285
1
—
—
—
1
0.22
Cattanach and Rasberry (1994)
8
1 + 9
10,318
1
—
—
1
2
0.24a
Cattanach and others (1985)
9
1 + 9
14,980
—
—
—
3
3
0.50a
Cattanach and others (1985)
Unweighted average: 3.96/9 = 0.44 per locus per gray
NOTE: Experiments were carried out during 1964–1994 in Harwell, England. All rates are normalized to single acute X-irradiation conditions.
aNormalized to single unfractionated irradiation conditions under the assumption of additivity of yields.
bNormalized to single unfractionated irradiation (by dividing the rate by 3) on the basis of observations of the enhancement of specific locus mutation frequency (in the same experiment by a factor of 3 [3H1 strain of mice]).
The committee therefore used the following approach to derive the average induced rate of mutations. All experimental data were first grouped by loci, so that an unweighted estimate of the locus-specific induced rates could be derived from the average of the estimates from all experiments involving each of the loci. Subsequently, these locus-specific rates were averaged across loci to arrive at the average induced mutation rate. This procedure permitted calculation of the standard error of the estimated rate that incorporated the sampling variability across loci as well as the variability of the rates in individual experiments. In this approach, unpublished data of Neel and Lewis (1990) were excluded since details of the identity of all the loci and the loci at which mutations were recovered were unavailable. Although fewer data were used (the total number of loci became 34), this approach was considered preferable since (1) no locus is double-counted while averaging over all loci, (2) the loci and the corresponding mutant phenotypes are clear, and (3) an estimate of the standard error of the mean (which takes into account both intra- and interlocus variability) can be given. These data permit an overall average estimate of (1.08 ± 0.30) × 10−5 per locus per gray (Table 4-3C). With a dose-rate reduction factor of 3 traditionally used1 (Russel 1965;
1
In the mouse, the dose-rate reduction factor of 3 for spermatogonial irradiations comes not only from the 6 Gy data of Dr. William Russell but also from the analysis of Dr. Tony Searle published in the Proceedings of the Cortina International Radiation Reseach Conference in 1967. Dr. Searle analyzed all of the chronic radiation data in the range from 37.5 to 861 R statistically and showed that the exposure-frequency relationship is linear and that the straight line of best fit could be described by
Y = 8.34 x 10−6 + 6.59 x 10−8X,
where Y is the yield of mutations and X is the exposure in roentgens. The slope is one-third of that for acute X-irradiation (300 and 600 R).
Further, the following statement from BEIR V (NRC 1990, p. 110) provides additional substantiation for the dose-rate reduction factor of 3: “The other important baseline value for spermatogonia is for the response to low dose-rate, low-LET irradiations … the rate is (7.3 ± 0.8)10−8/locus/rad for total doses between 35 and 900 rad (Ru82a). The dose-rate factor is 3.0 ± 0.4.”
OCR for page 101
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
TABLE 4-3C Locus-Specific Rates for Radiation-Induced Mutations in Mice Estimated from Data Tables 4-3A and 4-3B
Locusa
Rate per Gray (× 105)
SE (× 105)
pa
0
0
pe
0
0
G6pd1
0
0
G6pd1
0
0
Ldh2
0
0
Ldr
0
0
Pgk1
0
0
Tpi
0
0
Hba2
0
0
Hbb1
0
0
Hbb2
0
0
Gapdh
0
0
Pk
0
0
Mod1
0
0
Sp
0.04
0.04
W
0.15
0.12
Gpi
0.33
0.33
a
0.45
0.24
T
0.45
0.18
ln
0.67
0.67
Ldh1
0.97
0.69
se
0.97
0.33
Sl
1.31
0.51
bp
1.34
0.95
Es3
1.67
1.67
Hba1
1.67
1.67
c
1.90
0.48
Gr
2.19
1.40
b
2.35
0.52
fz
2.68
1.34
p
2.93
0.56
d
3.14
0.62
Pgam
3.91
1.93
s
7.59
0.89
Average rate (acute irradiation)
1.08
0.30b
Chronic irradiation
0.36
0.10b
NOTE: For raw data and their analysis, see Sankaranarayanan and Chakraborty (2000a).
aIn these calculations, two additional loci (Ldh2 in the experiments of Pretsch and others 1994; Hba2 in the experiments of Peters) have been included based on current evidence (Lewis and Johnson 1986).
bThe standard error of the average rate was calculated taking into account variation of the rates among loci as well as sampling variation of the experimental data for each locus.
Searle 1967), the rate for chronic low-LET radiation conditions becomes (0.36 ± 0.10) × 10−5 per locus per gray.
It is worth reiterating here that this is the first time an attempt has been made to use the mutation data coming not only from the 7 specific loci but also from all loci for which there are published data (a total of 34 loci; see Table 4-3C) taking into account interlaboratory and interexperimental variations in induced rates. Unfortunately, all of the data from biochemical loci and for dominant visibles were from experiments involving acute X- or fractionated X-irradiation experiments. In trying to put together all of these data, there was no alternative but to use the correction factors suggested by the authors of the respective papers to estimate the rate for chronic radiation conditions from the available data. The committee feels that the procedures adopted in estimating an induced rate of (0.36 ± 0.10) × 10−5 per gray are sound and that it is justifiable to use a single estimate for the induced rate of mutations.
THE DOUBLING DOSE ESTIMATE
With the estimates of (2.95 ± 0.64) × 10−6 per locus for the rate of origin of spontaneous mutations in humans and (0.36 ± 0.10) × 10−5 per locus per gray for induced mutations in mice, the DD becomes 0.82 ± 0.29 Gy. This new estimate is not very different from 1 Gy that has been used thus far and was based entirely on mouse data. The conceptual basis and the database used for estimating the average spontaneous and induced rates of mutations, however, are now different. The committee suggests retaining the use of 1 Gy for the DD estimate.
MUTATION COMPONENT OF GENETIC DISEASES
Background
As noted earlier, the MC is one of the quantities in the equation used to estimate risk of genetic disease using the doubling dose method (i.e., risk per unit dose = P × [1/DD] × MC, where P = baseline disease prevalence, 1/DD = the relative mutation risk per unit dose, and MC = the mutation component). The rationale for including MC in the risk equation is that the relationship between mutation and disease varies between different classes of genetic diseases—simple for autosomal dominant and X-linked diseases, slightly complex for autosomal recessive diseases, and very complex for multifactorial diseases—and the use of disease class-specific MC makes it possible to predict the impact of an increase in mutation rate on the frequencies of all classes of genetic diseases (Chakraborty and others 1998b; Denniston and others 1998; ICRP 1999).
General Definition
Let P be the disease prevalence before an increase in mutation rate and ΔP its change due to a Δm change in spontaneous mutation rate, m. The mathematical identity
(4-4)
formalizes the definition of MC. In this equation, since ΔP/P is the relative change in disease prevalence and Δm/m is the
OCR for page 121
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
FIGURE 4A-2 Comparisons of the distribution liability in the general population with those in relatives of affected individuals when there are differences in the prevalence of multifactorial disease, according to the multifactorial threshold model with the additional assumption of different thresholds for disease liability in the two sexes.
Based on the properties of the normal distribution of liability (made up of both genetic and environmental components) that underlies the MTM, methods have been developed to use data on the population frequency of a given multifactorial disease to predict the risk to relatives of those affected and to estimate, on the basis of correlation in liabilities between relatives, the relative contribution of genetic factors to the overall phenotypic variability summarized in the statistic called “heritability of liability” (h2).
Concept of Heritability
In quantitative genetics, the relative contributions of genetic and environmental factors to the overall phenotypic variation is assessed by analysis of variance (i.e., by estimating the total phenotypic variance, VP, and apportioning it into variance due to genetic factors, VG, and variance due to environmental factors, VE). Under the assumption that the genetic and environmental effects are independent of each other (i.e., they are not correlated), VP = VG + VE. The ratio VG/VP is called “broad-sense heritability of liability,” or “degree of genetic determination,” and is symbolized by hB2. It provides a measure of the relative importance of genotype as a determinant of phenotypic value (Smith 1975).
The genotypic variance VG can be subdivided into an additive component (VA) and a component to deviations from additivity. Additive genetic variance is the component attributable to the average effect of genes considered singly, as transmitted in the gametes. The ratio VA/VP is called “narrow-sense heritability,” or hN2, and expresses the extent to which the phenotypes exhibited by parents are transmitted to offspring, and it determines the magnitude of correlation between relatives. The nonadditive genetic variance is due to the additional effects of these genes when combined in diploid genotypes and arises from dominance (VD), interaction (epistasis, VI) between genes at different loci, and assortative mating (VAM). In the absence of these sources of genetic variance, hN2 = hB2. It is important to note that most of the heritability estimates for chronic diseases published in
OCR for page 122
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
the literature are broad-sense heritability of liability estimates and are in the range of about 0.3–0.8.
Other Models of Inheritance of Multifactorial Diseases
An important assumption of the MTM as discussed above is that a large number of factors, each with small effects, contributes to liability. However, the assumption of fewer contributing factors is also consistent with data from familial aggregation studies, and for this reason, it is not a good analytical tool for discriminating between different modes of inheritance. Consequently, attempts to fit the familial data to Mendelian models (with appropriate choice of assumptions on the numbers of loci, penetrance, dominance, etc.) or to a combination of major locus and polygenic models have been made, (e.g., Elston and Stewart 1971; Morton and MacLean 1974; Kendler and Kidd 1986); although these models are of interest in catalyzing the search for the genes involved, they are now largely superseded by molecular approaches that hold the potential for direct identification of the genes.
ANNEX 4B: THE DOUBLING DOSE
Table 4B-1 provides a broad overview of the data used during the past four decades for estimating doubling doses. It is worth noting that although the present unit for expressing absorbed radiation dose is gray (or sievert when considering radiations of different qualities), in reviewing the earlier estimates in this section the DDs are expressed in the same units employed in the original publications, namely, roentgens (R), rads, roentgen-equivalent-man (rem), grays, and sieverts. Note that for low-LET radiation (e.g., X-rays and γ-rays), 1 Gy = 100 rads ~ 100 R; 1 rem=1rad; and 1Sv=100 rem.
Briefly, the notion that the DD for genetic damage induced in human males at low-dose or chronic low-LET radiation conditions is likely to be of the order of about 100 R was already entertained in the early 1960s (UNSCEAR 1962). This estimate was guided by the findings (from mouse studies on recessive specific locus mutations) that chronic X-irradiation would be only about one-third as effective as acute X-irradiation in males and much less effective in females (Russell and others 1958, 1959). Consequently, it was suggested that the DD for chronic X-irradiation exposure conditions was probably at least three times that for acute X-irradiation (i.e., three times that of about 30 R suggested in the 1958 UNSCEAR report for acute X-irradiation or about 100 R).
In 1971, Lüning and Searle broadened the original concept of the DD to include not only mutations at defined gene loci, but also four other end points of genetic damage (semisterility, dominant visible mutations recovered in the course of studies on recessive specific locus mutations, autosomal recessive lethals, and skeletal abnormalities, all from experiments involving irradiation of male mice [spermatogonial stem cell irradiations]). They found that for acute X-irradiation of males, although individual estimates varied from 16 to 51 R (with wide confidence limits, except for specific locus mutations), the overall average was about 30 R. For low-dose or chronic low-LET radiation exposure, the suggestion was that it would be between three and four times that for acute X-irradiation (i.e., about 100 R). UNSCEAR, however, did not use the DD method in its 1972 report, but in all reports published until 1993, the mouse data-based estimate of 1 Gy has been used.
The BEIR I report (NRC 1972) introduced the concept that DD estimates must be based on the average spontaneous mutation rate of human genes and the average induced rate of mutations in mouse genes. In that report it was assumed that (1) the spontaneous mutation rate of human genes might be in the range of 0.5 × 10−6 to 0.5 × 10−5 per gene and (2) the sex-averaged rate of induced recessive mutations in mouse was about 0.25 × 10−7 per locus per rem for low-LET radiation conditions. With these estimates, a range of DDs from 20 to 200 rem was calculated.
The induced rate of 0.25 × 10−7 per locus per rem mentioned above was the unweighted average of the rate of 0.5 × 10−7 per locus per rem for males (at 12 loci, including 7 of the specific loci have been used in most mouse experiments and the additional 5 used in the studies of Lyon and Morris 1969) and that of zero assumed for females. It was noted, however, that the estimate of 0.25 × 10−7 per locus per rem might be too high for at least two reasons: (1) “the gene loci at which these studies were made, were to some extent preselected for mutability” and (2) “the rate of induction of dominant visible mutations in mice is lower than for recessives by at least an order of magnitude and dominant mutations constitute a substantial part of the human genetic risk.” This procedure of using human data on spontaneous mutation rates was driven by one of the principles stated by the committee—namely, that emphasis should be placed on human data when feasible—the implicit idea being that if the induced rate was extrapolated from mouse to humans, there would be one extrapolation uncertainty and if both spontaneous and induced rates were extrapolated to humans, there would be two such uncertainties.
When UNSCEAR (1977) first used the mouse data-based DD of 100 rads, it did not actually specify the induced rates. This was because the estimate of 100 rads was arrived at by assuming that the DD for low-LET chronic radiation conditions would be three times that of ~30 rads for high-dose-rate acute X-irradiation conditions (for five different end points; see Lüning and Searle 1971).
In BEIR III (NRC 1980), however, the committee abandoned the method that was used in BEIR I, namely, using human data on spontaneous mutation rates and mouse data on induced mutation rates in defined genes. The stated objection to the BEIR I method was that it mixed the induced rate of a set of mouse genes preselected for high mutability
OCR for page 123
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
TABLE 4B-1 Doubling Dose Estimates Used in Risk Estimation from the 1950s to the Early 1990s
Reference
DD
Radiation Conditions
Comments
1956 BEAR report (NRC 1956)
50–80 R
40 R
High dose rate (acute)
Guided more by general radiation genetic principles (established mostly from Drosophila studies) than by knowledge of mouse or human mutation rates and, therefore, nothing more than educated guesses; among the principles were (1) linear dose-effect relationship for induced mutations and (2) effect independent of dose rate or dose fractionation.
The general philosophy and “best” estimates of the Medical Research Council (MRC 1958) and UNSCEAR (1958) were roughly similar
UNSCEAR (1962)
100 R
Chronic
Based on mouse data on the reduced effectiveness of chronic γ-irradiation for the induction of specific locus mutations (Russell and others 1958); assumed that DD for males will be about 3 times that of 30 R assumed in UNSCEAR (1958) for acute X-irradiation conditions; noted that DD for females may be higher
Lüning and Searle (1971)
16–51 rads
~100 rads
Acute
Chronic
Based on mouse data for 5 different end points for males; no DD estimate provided for females
1972 BEIR report (NRC 1972)
20–200 rem
Chronic
Based on a range of spontaneous rates in humans (0.5 × 10−6 to 0.5 × 10−5) and a sex-averaged rate of induction of specific locus mutations of 0.25 × 10−7 per locus per rem in mice
Neel and others (1974)
46 rem (Petersen and others 1990)
125 rem (females)
Acute
Based on data on mortality of children born to A-bomb survivors through the first 17 years of life; assumed that for chronic irradiation, the DD for males might be 3 to 4 times 46 rem and as much as 1000 rem for females
Sankaranarayanan (1976); Searle (1976)
80–240 rads
Chronic
Based on mouse data for specific locus mutations induced in spermatogonia and in mature + maturing oocytes and dominant visibles and translocations induced in spermatogonia
UNSCEAR (1977)
100 rads
Chronic
Rationale stated as follows: “Examination of available evidence in the mouse suggests that the use of a 100-rad DD will not underestimate the risk. The ICRP Task Group has also this figure in its calculations …”
1980 BEIR report (NRC 1980)
50–250 rem
Chronic
Based on the “best substantiated” estimate of DD of 114 rem for spermatogonial irradiation of male mice and approximately halving and doubling the above estimate to arrive at the range of 50–250 rem
UNSCEAR (1982)
100 rads
Chronic
No change from the 1977 report
Neel and others (1982); Schull and others (1982)
60 ± 93 rem
135 ± 388 rem
535 ± 2416 rem
135 ± 156 rem
Acute
The first three estimates are based, respectively, on data on UPOs, survival through childhood, and sex chromosomal aneuploids in the Japanese studies; the authors considered that the weighted average of 135 ± 156 rem (last entry) should be multiplied by a factor of 3 to make it applicable to chronic radiation conditions
UNSCEAR (1986)
1 Gy
Chronic
No change from the 1977 report
UNSCEAR (1988)
1 Gy
Chronic
No change from the 1977 report
1990 BEIR report (NRC 1990)
100 rads
Chronic
Overall estimate based on mouse data (both sexes) on several different end points; most estimates given as ranges that vary by factors between about 2 and 30 (a reflection of differences in estimated spontaneous and induction rates); multiplication factors between 5 and 10 used when necessary to convert DD estimates for high-dose-rate irradiation to those for chronic irradiation
Neel and others (1990)
1.69–2.23 Sv
Acute
Composite estimates of “minimal DDs” (DDs at 95% lower confidence limits) compatible with Japanese results on UPOs, F1 mortality, F1 cancer, sex chromosomal aneuploids, and mutations altering protein charge or function; on the assumption of a dose-rate reduction factor of 2, the authors suggest that for chronic low-LET, low-level radiation, the figures are likely to be twice those estimated (i.e., about 3.4 to 4.5 Sv)
Neel and Lewis (1990)
1.35 Gy
Acute
Based on an analysis of mouse data on 7 mutational end points (spermatogonial irradiation experiments); the authors suggest that with the use of a dose-rate factor of 3, the DD will be about 3 Gy
UNSCEAR (Rabes and others 2000)
1 Gy
Chronic
No change from the 1977 report
SOURCE: Sankaranarayanan and Chakraborty (2000a).
OCR for page 124
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
with estimates of human spontaneous rates for more typical genes. The BEIR III committee adopted the view that it was preferable to use a DD estimate obtained from spontaneous and induced mutations in the same set of loci in the same species and used exclusively the data on the seven specific loci obtained in experiments with male mice. The figures used were 7.5 × 10−6 per locus for spontaneous rates and 6.6 × 10−8 per locus per rem for induced rates from which “the best substantiated” DD estimate of 114 R was calculated. To derive DDs for risk predictions, it approximately halved and doubled the above estimate of 114 R to obtain a range of 50 to 250 rem.
In BEIR V (NRC 1990), the committee again used primarily mouse data but included several additional end points in both sexes (dominant lethals, recessive lethals, dominant visibles, recessive visibles, reciprocal translocations, congenital malformations, and aneuploidy). On the basis of all these data, it concluded that “considering all endpoints together, the direct estimates of doubling dose for low dose rate radiation have a median value of 70–80 rad, indirect estimates based on high dose-rate experiments have a median value of 150 rad, and the overall median lies in the range of 100 to 114 rad. These estimates support the view that the doubling dose for low dose-rate, low-LET radiation in mice is approximately 100 rad for various genetic endpoints.”
Table 4B-1 also shows that the DD estimates made over the years based on genetic data from A-bomb survivors (Neel and others 1974, 1982, 1990; Schull and others 1981, 1982; Otake and others 1990; Neel 1998) were at least some three to four times that of 1 Gy used by UNSCEAR and the BEIR committee; the so-called Japanese DD estimates, however, were never used by the above committees. For the first time, the BEIR V (NRC 1990) report gave a formal “status” to the Japanese results by noting that “a doubling dose of 100 rem approximates the lower 95% confidence limit for the data from atomic bomb survivors in Japan and it is also consistent with the range of doubling doses in mice.”
ANNEX 4C: ASSUMPTIONS AND SPECIFICATIONS OF THE FINITE-LOCUS THRESHOLD MODEL
The assumptions and specifications of the FLTM have been discussed in detail by Denniston and colleagues (1998) and in the ICRP (1999) Task Group report. Briefly, the FLTM assumes that (1) the genetic component of liability of a chronic multifactorial disease is discrete and is determined by mutant alleles at a finite number (n) of autosomal gene loci; the total number of mutant alleles at these n loci in a given genotype is a random variable g; (2) the environmental component is continuous and represented by a random variable e, which has a Gaussian distribution with mean of zero and variance of Ve; (3) the total liability x = f(g) + e, where f(g) is a function of the number of mutant alleles in the n-locus genotype of the individual and e is the environmental effect; (4) individuals with liability exceeding the threshold T (i.e., x > T) are affected by the disease, and those for whom x < T are unaffected; and (5) unaffected individuals have a fitness of 1 and unaffected ones of (1 − s). The impact of an increase in total mutation rate as a result of radiation exposures—from m to m(1 + k), with k measuring the increase relative to the baseline—is assessed in terms of changes in heritability of liability (hx2), and consequent changes in the MC. This assessment was carried out by assuming that the effects of the mutant alleles are either additive or synergistic.
Unlike the case of Mendelian diseases, the algebraic formulations of the FLTM do not permit expressing the effects in the form of a single equation. However, the predictions of the model can be evaluated iteratively using the computer program that was developed for this purpose. The program is first run using a specified set of parameter values (mutation rate, selection coefficients, threshold, etc.) until the population reaches equilibrium between mutation and selection. Once this occurs, the mutation rate is increased either once or permanently corresponding to radiation exposure in one generation only or in every generation, and the computer run is resumed with the new mutation rate while the other parameters remain the same. The changes in mutation component and its relationship to heritability of liability are then examined in desired generations and at equilibrium. It is worth mentioning that the h2 estimates are not inputs but outputs of the program obtained using different combinations of s values, environmental standard deviation, and threshold.
ANNEX 4D: DIFFERENCES BETWEEN SPONTANEOUS DISEASE-CAUSING MUTATIONS IN HUMANS AND RADIATION-INDUCED MUTATIONS IN EXPERIMENTAL SYSTEMS
The molecular alterations recorded in spontaneous disease-causing mutations in humans include a wide variety ranging from base-pair changes to whole-gene deletions and some multigene deletions. Radiation-induced mutations studied in experimental systems (including the mouse), however, are often multigene deletions, although scored through the phenotype of the marker loci. The extent of the deletion varies with the locus and the genomic region in which it is located.
Spontaneous mutations arise through a number of different mechanisms, and most are dependent on the DNA sequence organization of the genes and their genomic context. In contrast, radiation-induced mutations originate through random deposition of energy in the cell. One can, therefore, assume that the initial probability of radiation inducing a deletion may not differ between different genomic regions. However, their recoverability in live-born offspring seems dependent on whether the loss of the gene or genomic region is compatible with viability in heterozygotes.
OCR for page 125
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
Spontaneous mutations can cause either loss or gain of function of the normal gene through different mechanisms. For example, loss-of-function mutations in genes that code for structural or regulatory proteins may result in dominant phenotypes through haploinsufficiency (i.e., a single normal gene is not sufficient for normal functioning) or through dominant negative effects (i.e., the mutant product interferes with the function of the normal gene in the heterozygote). While loss of function of a gene can result from a variety of molecular alterations including deletions, gain-of-function mutations are likely only when specific changes in the gene cause a given disease phenotype. Radiation-induced mutations, because they are often multigene deletions, cause loss of function through haploinsufficiency.
Despite the existence of a number of differences between spontaneous and radiation-induced mutations as outlined above, radiation mutagenesis studies with a variety of experimental systems have been very successful. The possible reasons for this are now becoming evident: although the choices of marker genes in early studies of induced mutations were dictated more by practical considerations (e.g., obtaining sufficient numbers of mutants, unambiguous identification through their respective phenotypes) than by their relevance to human genetic diseases, in retrospect it is clear that the “successful” mutation test systems have been those in which most of these marker genes, and the genomic regions in which they are located, are nonessential for the viability of heterozygotes (in vivo) or of the cell carrying the induced genetic change (in vitro). Consequently, induced mutations—predominantly deletions—could be recovered and studied. Most human genes, however, do not appear to be of this type.
ANNEX 4E: CRITERIA USED TO ASSIGN HUMAN GENES TO ONE OF THREE GROUPS FROM THE STANDPOINT OF THE RECOVERABILITY OF INDUCED MUTATIONS IN LIVE BIRTHS
The genes included in the analysis are a subset of those in which mutations cause autosomal dominant and X-linked diseases, which have provided the basis for the overall incidence estimates for these diseases discussed earlier (Sankaranarayanan 1998). Since not all of them fulfilled the requirements for inclusion (because of insufficient information about one or more of the following: gene size, structure, function, genomic context, etc.), only a subset could be used. The “gene-richness” or “gene poorness” of given genomic regions was assessed using the MIM (Medelian Inheritance in Man) gene maps that present the cytogenetic location of “disease genes” and other expressed genes in given cytogenetic bands (McKusick 2000.).
A gene is assigned to group 1 (induced deletions unlikely to be recovered and/or unlikely to cause the phenotype of the disease under study) when the phenotype of the naturally occurring disease is due to specific (1) gain-of-function mutations (e.g., the FGFR3 gene involved in achondroplasia); (2) trinucleotide repeat expansions (e.g., Huntington’s disease); (3) dominant negative mutations (e.g., the COL1A1 gene involved in osteogenesis imperfecta); and (4) restricted array of point mutations (e.g., mutations in the APOB gene involved in one form of familial hypercholesterolemia). Also included in this group are genes that are relatively small in size and located in putative gene-rich regions (e.g., the VMD2 gene in Best’s macular dystrophy).
The gene is assigned to group 2 (uncertain recoverability) when (1) it is large, it codes for an essential structural protein, and the known genetic changes are missense or nonsense mutations; (2) whole-gene deletions are rare; (3) whole-gene deletions are not rare, but the gene is located in a putative gene-rich region; and (4) information on these other genes and their function is insufficient (e.g., BRCA2; VHL [von Hippel-Lindau syndrome]).
Group 3 (potentially recoverable) includes genes that are generally large and constitutional deletions, some extending beyond the confines of genes, and translocations or inversions with breakpoints in the gene causing the disease phenotype are known despite the putative gene-rich nature of the genomic region (e.g., EXT1 [multiple exotoses]; RB1 [retinoblastoma]).
For X-linked genes, the assessment is based on whether the induced deletion will be compatible with viability in males and cause disease (since the loss of the whole X chromosome is compatible with viability but results in 45,X females).
ANNEX 4F: RADIATION STUDIES WITH EXPANDED SIMPLE TANDEM REPEAT LOCI IN THE MOUSE AND MINISATELLITE LOCI IN HUMAN GERM CELLS
Introduction
The mouse and human nuclear genomes, like those of other complex eukaryotes, contain a large amount of highly repeated DNA sequence families most of which are transcriptionally inactive (Singer 1982). Among these are the simple sequence repeats that are perfect or slightly imperfect tandem repeats of one or a few base pairs (bp). In the mouse genome, the tandem repeat loci are represented by (1) relatively short microsatellites (<500 bp) with a repeat size of 1 to 4 bp; (2) long expanded simple tandem repeats (0.5 to 16 kilobases, repeat size 4 to 6 bp); and (3) true minisatellites (0.5 to 10 kb) with repeat size of 14 to 47 bp (Gibbs and others 1993; Bois and others 1998a, 1998b; Blake and others 2000).
Mouse ESTRs
The ESTRs were originally called minisatellites but have recently been renamed to distinguish them from the much
OCR for page 126
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
more stable true minisatellites in the mouse genome (Bois and others 1998a, 1998b). The ESTRs are highly unstable (i.e., they manifest high spontaneous mutation rates) in both somatic and germ cells. The mutational changes are manifest as changes in the number of tandem repeat cores and, hence, allele length. The available data suggest that the ESTR instability is a replication- or repair-based process involving polymerase slippage similar to mechanisms suggested for microsatellite instability (Ellegren 2000).
Human Minisatellites
In contrast to mouse ESTRs, the minisatellites in humans consist of longer repeats (10 to 60 bp) that may span from about 0.5 kb to several kilobases and show considerable sequence variation along the array (Jeffreys and others 1991; 1994; May and others 1996; Buard and others 1998; Tamaki and others 1999; Stead and Jeffreys 2000; Vergnaud and Denoeud 2000). The majority of the classical minisatellites are GC rich. The fact that some of the human minisatellite loci studied are highly unstable and have very high spontaneous mutation rates of the order of a few percent is now well documented (Jeffreys and others 1985, 1988, 1995; Smith and others 1990; Vergnaud and Denoeud 2000). Mutation at these loci is almost completely restricted to the germline and is attributed to complex gene conversion-like events involving recombinational exchanges of repeat units between alleles (Jeffreys and others 1994; May and others 1996; Jeffreys and Neumann 1997; Tamaki and others 1999; Buard and others 2000; Stead and Jeffreys 2000; Vergnaud and Denoeud 2000).
Radiation Studies with Mouse ESTR Loci
The Loci Used
Two ESTR loci have been used thus far in mouse mutation studies, namely, the Ms6-hm, and Hm-2, both of which show multiallelism and heterozygosity within inbred strains. The Ms6-hm is <10 kb in size (varying greatly between different mouse strains) and consists of tandem repeats of the motif GGGCA. Linkage analysis localized Ms6-hm near the brown (b) coat color gene on chromosome 4. The germline mutation rate is about 2.5% per gamete (Kelly and others 1989). The Hm-2 locus is located on chromosome 9 and consists of GGCA tetranucleotide repeats with alleles containing up to 5000 repeat units (i.e., up to 5 kb). The germline mutation rate of this locus is estimated to be of the order of at least 3.6% (Gibbs and others 1993). As discussed below, Dubrova and colleagues studied mutation induction at both of the above loci, whereas the Japanese workers focused their attention only on the Ms6-hm locus.
Low-LET Radiation Studies
In the studies of Dubrova and colleagues (1993) involving irradiation of spermatagonial stem cells (0.5 and 1 Gy of γ-rays; CBA/H strain), significant increases in the frequencies of mutations at the Ms6-hm and Hm-2 loci were found. Subsequent work with X-irradiation doses of 0.5 and 1 Gy established that for mutations induced in the above cell stage, the dose-effect relationship was consistent with linearity (y = 0.111 + 0.338D), where D is the dose in grays (Dubrova and others 1998a, 1998b). From these data, the authors estimated that the DD for ESTR mutations induced in spermatogonia was 0.33 Gy for acute X-irradiation, similar to that reported for specific locus mutations in mice.
In the above work, spermatids were found to be insensitive to mutation induction, a finding at variance with those of Sadamoto and colleagues (1994) and Fan and coworkers (1995) with the C3H/HeN mouse strain. These authors showed that for Ms6-hm locus mutations, all male germ cell stages were sensitive (3 Gy of γ-irradiation). Nonetheless, both sets of studies demonstrated that increases in mutation frequencies could be detected at radiation doses and sample sizes substantially smaller than those used in conventional genetic studies with specific locus mutations.
High-LET Radiations Studies
Niwa and collegues (1996) found that acute neutrons from a 252Cf source (65% neutrons + 35% γ-rays) were 5.9, 2.6, and 6.5 times more effective, respectively, in spermatozoa, spermatids, and spermatogonia, than acute γ-irradiation in inducing mutations at the Ms6-hm locus. In similar studies, Dubrova and colleagues (2000a) noted that in spermatogonial cells, chronic neutrons also from a 252Cf source had a relative biological effectiveness of about 3 relative to chronic γ-irradiation (regression equations: y = 0.136 + 1.135D, neutrons; doses of 0.125, 0.25, and 0.5 Gy; y = 0.110 + 0.373D, γ-rays; doses of 0.5 and 1 Gy). Additionally (and not unexpectedly), they found that at the above γ-ray doses of 0.5 and 1 Gy, there was no dose-rate effect. It should be remembered that the lower effectiveness of chronic γ-irradiation recorded in earlier specific locus mutation studies (Russell and others 1958) occured at total doses of 3 and 6 Gy. This observation is in contrast to earlier results with specific locus mutations (Russell and others 1958) at 3 and 6 Gy showing that chronic γ-irradiation was only one-third as effective as acute X-irradiation in inducing specific locus mutations.
Mutation Induction at the ESTR Loci—An Untargeted Process Arising as a Result of Radiation-Induced Genomic Instability
One important conclusion that emerges from these studies is that mutation frequencies in the progeny of irradiated animals are too high to be accounted for by the direct induc-
OCR for page 127
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
tion of mutations at the loci studied (i.e., radiation induction of germline mutations at ESTR loci is an untargeted process). Dubrova and colleagues (1998a, 1998b) concluded that there might be two associated processes: structural damage elsewhere in the genome or in other sensor molecules and, subsequently, indirect mutation at ESTR loci. This nontargeted origin of radiation-induced mutations at the ESTR loci is reminiscent of the phenomenon of delayed radiation-induced genomic instability in somatic cells (discussed in Chapters 2 and 3). The experiments of Barber and colleagues (2000) showed that the ESTR mutations in unirradiated or irradiated mice are not associated with a general genome-wide increase in meiotic recombination rate.
Further support for the concept of the nontargeted origin of induced ESTR mutations comes from the work of Niwa and Kominami (2001). In their study, male mice received 6 Gy of γ-irradiation and were mated to unirradiated females to produce F1 progeny from irradiated spermatozoa and stem cell spermatogonia. As in their earlier studies, mutations at the Ms6-hm locus were studied. The mutant frequencies for the paternally derived allele increased to 22% and 19% in the F1 progeny from irradiated spermatozoa and spermatogonia, respectively (about a twofold increase over the control rate). The surprising finding was that the mutation frequency also was higher (20%) in the maternally derived allele in progeny descended from irradiated spermatozoa, but not from spermatogonia. The authors’ interpretation is that the introduction of damage into the egg by irradiated spermatozoa triggers genomic instability in zygotes and in embryos of subsequent developmental stages, and that this genomic instability induces untargeted mutation in cis (in the paternally derived allele) and in trans (in the unirradiated maternally derived allele).
Transgenerational Instability
Dubrova and colleagues (2000a) and Barber and coworkers (2002) provided additional evidence for the involvement of radiation-induced germline genomic instability in the origin of induced ESTR mutations. In these experiments involving chronic neutron irradiation (0.5 Gy) of spermatogonial stem cells, the mutation frequency in the F1 progeny was about sixfold higher than in the control. Breeding from the unirradiated F1 mice revealed that the mutation rate remained high in transmissions from both F1 males (6×) and F1 females (3.5×; scored in F2). A part of this increase is due to germline mosaicism in F1 animals, suggesting that paternal exposure to radiation results in a destabilization of ESTR loci in the germline of offspring and that some of the mutations occur sufficiently early in germline development for significant levels of mosaicism to arise. More importantly, this instability is transmissible through meiosis and mitosis to the F2 generation and appears to operate in trans in the F1 germline (i.e., affecting alleles not only from the exposed F0 male but also from the unexposed F0 female). The latter finding is similar to that of Niwa and Kominami (2001).
In subsequent experiments, Barber and colleagues (2002) confirmed the transgenerational effects of chronic neutron irradiation and extended the observations to acute X-irradiation. Additionally, the response of two other inbred mouse strains (C57BL/6 and BALB/c) was compared with that of the CBA/H strain used in their studies. The rationale for the comparisons rests on earlier findings that BALB/c and CBA/H mice show higher levels of radiation-induced genomic instability in somatic cells than C57BL/6 mice and that this difference can be attributed to the strain-specific polymorphism at the Cdkn2a (cyclin-dependent kinase inhibitor) and Prkdc (DNA-dependent protein kinase catalytic subunit) genes (Zhang and others 1998; Yu and others 2001).
In these experiments, (1) spermatogonial neutron (0.4 Gy) or X-irradiation (2.0 Gy) of CBA/H mice resulted in an increase in the mutation rate in both the F1 and the F2 generations (derived from unirradiated F1 males and females); however, although spermatid irradiation did not cause an increase in mutation rate in the F1 generation (which was also the case in their earlier work), there was a clear increase in mutation rate in the F2 progeny, suggesting that destabilization of the F1 germline occurs after fertilization, regardless of the stage of spermatogenesis exposed to radiation, and that the radiation-induced signal also persists and destabilizes the F2 germline; (2) transgenerational effects were also observed in neutron-irradiated (0.4 Gy) C57BL/6 and X-irradiated (1 Gy) BALB/c mice; and (3) there were clear differences in the levels of spontaneous and transgenerational instability in the order BALB/c > CBA/H > C57BL/6. In summary, these data permit the conclusion that the instability associated with radiation-induced germ cell mutations at the ESTR loci persist for at least two generations.
Direct Studies of ESTR Mutations in Mouse Sperm
In a recent paper, Yauk and colleagues (2002) have reported on mouse experiments involving single molecular polymerase chain reaction (PCR) analysis of genomic DNA for studying spontaneous and radiation-induced mutations at the Ms6-hm locus. These X-irradiated male mice (1 Gy) were killed 10 weeks postirradiation, and spermatozoa collected from caudal epididymis from the mice were screened for mutations. The findings were that (1) significant increases in mutation frequency could be detected, with the magnitude being similar to that established by conventional pedigree analysis, and (2) the majority of mutations resulted from small gains or losses of three to five repeat units.
OCR for page 128
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
Radiation-Induced Mutations at Human Minisatellite Loci
Studies After the Chernobyl Accident and Around the Semipalatinsk Nuclear Test Site
Dubrova and colleagues (1996) first reported on radiation-induced minisatellite mutations among children born between February and September 1994 to parents who were continuously resident in the heavily polluted rural areas of the Mogilev district of Belarus following the Chernobyl accident. Blood samples were collected from 79 families (father, mother, and child) for DNA analysis. The control sample consisted of 105 nonirradiated Caucasian families from the United Kingdom, sex-matched to the offspring of the exposed group. DNA fingerprints were produced from all families by using the multilocus minisatellite probe 33.15 and two hypervariable single-locus probes, MS1 and MS31. Additionally, most families were profiled with the minisatellite probes MS32 and CEB1. For the Mogilev families, the level of 137Cs contamination was used as a dose measure, and the families were divided according to the median 137Cs contamination levels into those inhabiting less contaminated areas (<250 kBq m−2) and those inhabiting more contaminated areas (>250 kBq m−2).
The data showed that the frequency of mutations (1) was higher by a factor of about 2 in the children of exposed families relative to control families and (2) showed a correlation with 137Cs contamination levels as demarcated above. The authors suggested that these findings were consistent with radiation induction of germline mutations but also noted that other nonradioactive contaminants from Chernobyl, such as heavy metals, could be responsible. These results have been subject to criticism on the grounds that the U.K. control population was ethnically and environmentally different and therefore inappropriate for comparisons (UNSCEAR 2001). Furthermore, from the data presented, it would seem that the estimated germline doses in the whole region remain sufficiently uncertain to question the true significance of an approximately twofold difference in mutation frequencies.
In a subsequent extension of the above study, Dubrova and colleagues (1997) recruited 48 additional families and used five additional probes and found that the data confirmed the approximately twofold higher mutation rate in exposed families compared to nonirradiated families from the United Kingdom. In these studies, (1) approximate individual doses for chronic γ-ray exposures were computed for 126 families in the exposed group using published data on the annual external and internal exposure to 137Cs in soil, milk, and vegetables and family histories after the Chernobyl accident; (2) the parental dose for each family was taken as the mean value of the paternal and maternal doses up to conception of the child; (3) families within the exposed group could be divided according to the median of the distribution, into less exposed (<20 mSv) and more exposed (>20 mSv); and (4) the mutation rate in the latter was significantly higher than in the former, and both were higher than in the unexposed UK controls.
Further evidence showing an increase in minisatellite mutation frequencies has also been obtained from two studies, one in the Kiev and Zhitomir regions of Ukraine that sustained heavy radioactive contamination after the Chernobyl accident (Dubrova and others 2002b) and another at the Semipalatinsk nuclear test site in Kazakhstan (Dubrova and others 2002a). In the Ukraine investigation, the control and exposed groups were composed of families containing children conceived before (n = 98) and after (n = 240) the Chernobyl accident. Eight hypervariable minisatellite probes (CEB1, CEB15, CEB25, CEB36, MS1, MS31, MS32, and B6.7) were used.
A statistically significant 1.6-fold increase in mutation rate was found in the germline of exposed fathers, whereas the maternal germline mutation rate was not elevated. More than 90% of the children in the exposed cohort came from the most heavily radioactively contaminated areas of Ukraine, with a level of surface contamination from 137Cs of >2 Ci/km2. According to gamma spectrometric measurements of radionuclide concentration in soil and measurements of external exposures (γ-exposure rate in air), the whole-body doses from external exposures did not exceed 50 mSv, and similar doses from the ingestion of 137Cs and 134Cs for the Ukrainian population were also reported. The authors note that that all of these doses are well below all known estimates of the DD for mammalian germline mutation of 1 Sv (Sankaranarayanan and Chakraborty 2000b; UNSCEAR 2001) and, therefore, cannot explain the 1.6-fold increase in mutation rate found in exposed families
Between 1949 and 1989, the Semipalatinsk site was the former Soviet Union’s premier test site for 456 nuclear tests; it was closed in 1991. The surrounding population was exposed mainly to the fresh radioactive fallout from four surface explosions conducted in 1949, 1951, 1953, and 1956, and the radioactive contamination outside the test zone currently is assessed to be low. A total of 40 three-generation families around the test site (characterized by the highest effective dose >1 Sv) along with 28 three-generation nonirradiated families from a geographically similar noncontaminated rural area of Kazakhstan were included in the study (Dubrova and others 2002a). Note that the above dose estimate cited in the paper is from Gusev and colleagues (1997; based mostly on external radiation), and the World Health Organization (WHO 1998) states that the estimates range from <0.5 Sv to 4.5 Sv. All parents and offspring were profiled with the eight hypervariable minisatellite probes previously used in the Belarus and Ukraine studies. The mutation rates in the P0 and F1 generations were established from the observed frequencies, respectively, in the F1 and F2 generations (controls and exposed progeny).
The findings were (1) in the controls, the spontaneous mutation rates in the P0 and F1 generations were similar;
OCR for page 129
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
(2) in the irradiated groups, the P0 rate was significantly higher (1.8-fold) and the F1 rate was nonsignificantly (1.5-fold) higher compared to controls; and (3) plotted against the parental year of birth (1950–1960, 1961–1965, and 1966–1974), the mutation rate in the exposed F1 generation showed a negative correlation (i.e., decreased) with the parental year at birth, with the highest rate in the 1950–1960 cohort (similar to that in the P0 families) and much lower in the later two time periods (similar to that in the control cohorts).
The authors have interpreted these findings as follows: (1) all P0 parents born between 1926 and 1948 would have been directly exposed to relatively high levels of radiation from the nuclear tests, and this would explain the 1.8-fold increase in mutation rate; (2) F1 parents born between 1950 and 1956 would be heterogeneous with respect to the doses received: some would also have been exposed to high radiation doses, while those born later would have received considerably lower doses, and this heterogeneity in the parental doses could explain the 1.5-fold increase in mutation rate; and (3) the negative correlation with the year of birth may reflect the decreased exposure after the decay of radioisotopes in the late 1950s and after the cessation of surface and atmospheric nuclear tests.
Other Population Studies
In the mid-1990s, subsequent to publication of the radiation studies with mouse ESTR loci discussed earlier, Kodaira and colleagues (1995) conducted a pilot feasibility study on germline instability in cell lines established from the children of atomic bomb survivors in Japan. The cell lines were from 64 children from the 50 most heavily exposed families (combined gonadal equivalent dose of 1.9 Sv) and 50 children from control families. Mutations at six minisatellite loci were studied using the following six probes: Pc-1, 8TM-18, ChdTC15, p8g3, 8MS1, and CEB1. A total of 28 mutations were found, but these were at the p8g-3, 8MS-1, and CEB-1 loci only, and there were no mutations at the other three loci. Twenty-two of these were in the controls (of 1098 alleles tested; 2%), and six were in children from irradiated parents (among 390 alleles; 1.5%). Thus, there was no significant difference in mutation frequencies between the control and the exposed groups. The use of probes 33.16 and 33.15 in subsequent work did not alter the above conclusion (Satoh and Kodaira 1996; Satoh and others 1996).
The discrepancy between the results of Kodaira and colleagues, on the one hand, and those of Dubrova and colleagues (1996, 1998b, 2000a, 2000b) in the Belarus and other cohorts discussed earlier appears real. To what extent this might be due to differences in type and duration of radiation exposure remains unclear. For instance, the A-bomb survivors were externally exposed to considerable acute doses of radiation, whereas in the Belarus, Ukraine, and Semipalatinsk studies the exposures were chronic (both internal and external). Secondly, in the case of A-bomb survivors, most of their children were born more than 10 years after the single, acute parental exposure; in Belarus and Ukraine, however, the affected areas have been irradiated constantly since the Chernobyl accident. Finally, the Japanese data are derived from families in which most of the children were born to parents of whom only one had sustained radiation; in the work of Dubrova and colleagues, the data pertain to children for whom both parents had been exposed to chronic irradiation.
Livshits and colleagues (2001) found that the children of Chernobyl cleanup workers (liquidators) did not show an elevated rate of minisatellite mutations compared to a Ukrainian control group. The dose estimate for the liquidators was <0.25 Gy but is subject to uncertainty (Pitkevich and others 1997), and the main exposure was from external γ-irradiation (with a relatively minor contribution from the intake of radionuclides) received as repeated small daily doses. Interestingly, children conceived within 2 months of the fathers’ employment had a higher mutation rate than those conceived more than 4 months after the fathers stopped working there. This would be consistent with an effect on cells undergoing spermatogenesis, but not on spermatogonial stem cells. However, none of these differences was statistically significant.
More recently, Kiuru and colleagues (2003) compared the frequencies of minisatellite mutations among children of 147 Estonian Chernobyl cleanup workers. The comparisons were within families (i.e., between children born before and after their fathers were exposed to radiation). The post-Chernobyl children (n = 155) were conceived within 33 months of their fathers’ return from Chernobyl; the “control” children were siblings (n = 148) born prior to the accident. Mutations were studied at eight minisatellite loci (CEB1, CEB15, CEB25, CEB36, MS1, MS31, MS32, and B6.7). The estimated mean dose to the workers was 100 ± 60 mSv, with fewer than 1.4% of the cohort receiving more than 250 mSv.
A total of 94 mutations (42 in the pre-Chernobyl group and 52 in the post-Chernobyl group) were found at the eight tested loci. Within-family (i.e., pre- and post-Chernobyl) comparisons of mutation rates showed that the post-Chernobyl children had a slightly but not significantly higher mutation rate (0.042 per band) than the pre-Chernobyl children (0.035 per band) with an odds ratio of 1.33 (95% CI: 0.80, 2.20). The available data do not permit an assessment of the extent to which differences in paternal age might have contributed to this difference. When the cleanup workers were subdivided according to their radiation doses, the mutation rate in children born to fathers with recorded doses of 200 mSv, showed a nonsignificant increase relative to their siblings; at lower doses there was no difference.
Weinberg and colleagues (2001) screened children born in families of cleanup workers (currently either in Ukraine or Israel) for new DNA fragments (‘mutations’) using “multisite DNA fingerprinting.” In contrast to the results of Livshits and colleagues (2001), they reported a sevenfold
OCR for page 130
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
increase in mutation rate in these children compared to those conceived before the Chernobyl accident and external controls. However, the mutants were detected using random amplified polymorphic DNA-PCR, an unreliable technology. These mutants were not validated and had no obvious molecular basis (Jeffreys and Dubrova 2001).
Studies of Cancer Patients
There are some limited data on minisatellite mutations detected directly in sperm sampled from cancer patients who have sustained radiotherapy and/or chemotherapy (Armour and others 1999; May and others 2000; Zheng and others 2000). All of these studies used the so-called small-pool PCR approach (SP-PCR) originally developed for the analysis of spontaneous mutations at human minisatellite loci (Jeffreys and others 1994). While this method can overcome the small sample size limitations encountered in pedigree analysis, a major shortcoming of the SP-PCR approach, compared to the pedigree approach, is the very large variation in spontaneous mutation rates of individual alleles at a single locus. Although SP-PCR can be used to evaluate the mutation rate in the same male before and after mutagenic treatment, it does not allow amplification of very large minisatellite alleles (longer than 5 kb), thus restricting mutation scoring to a subset of relatively small minisatellite sizes.
In the first of these studies (Armour and others 1999), sperm DNA of two men exposed to the anticancer drugs cyclophosphamide, etoposide, and vincristine, plus 2.2 Gy of X-rays (scattered radiation from mediastinal radiotherapy), were analyzed for mutations at the MS205 locus known to have a high germline mutation rate (~0.4–0.7% per gamete). There were no significant differences in mutation frequencies in the pretherapy and posttherapy samples (11 and 16 months, respectively, in the two individuals). Mutation rates were 0.38% versus 0.47% in the former and 0.10% versus 0.11% in the latter. It should be noted, however, that in mouse experiments, cyclophosphamide is mutagenic only in postmeiotic germ cells, etoposide (a topoisomerase II inhibitor) is mutagenic only in meiotic cells, and vincristine is not mutagenic, although it is known to prevent the assembly of tubulin into spindle fibers (Witt and Bishop 1996; Russell and others 1998).
In the second study (Zheng and others 2000), sperm DNA from 10 men treated for Hodgkin’s disease (with different combinations of chemotherapeutic agents plus 2.5 Gy of abdominal X-rays) were analyzed using the MS205 locus. Nine patients treated with either vinblastine or adriamycin and bleomycin did not show any increases in mutation frequency. Vinblastine binds to tubulin and, in mice, results in aneuploidy but not chromosome breakage or mutations. Adriamycin is an intercalating agent and an inhibitor of topoisomerase-II, and in mice, this compound is toxic to germ cells but does not cause mutations (Witt and Bishop 1996). Bleomycin, a radiomimetic agent, selectively targets mouse oocytes, but no mutation induction in male germ cells has been observed. The only patient treated with procarbazine + oncovin + prednisone (for six cycles with 3–4 week intervals between cycles) showed a slight increase in mutation frequency (1.14% versus 0.79%). Procarbazine is known to be mutagenic to mouse spermatogonia.
In the work of May and colleagues (2000), sperm DNA samples from three seminoma patients who underwent orchiectomy and external beam radiotherapy were used to study induction of mutations at the B6.7 and CEB1 loci. These men received 15 fractions of acute X-irradiation, with a total testicular dose (from scattered radiation) ranging between 0.4 and 0.8 Gy. No induced mutations were found.
ANNEX 4G: DOUBLING DOSES ESTIMATED FROM GENETIC DATA OF CHILDREN OF A-BOMB SURVIVORS
The most recent DD estimates consistent with the Japanese data are those of Neel and colleagues (1990). These were expressed as “end-point-specific minimal DDs” excluded by the data at specified probability levels and “most probable gametic DD” (note that all of these are for the acute radiation conditions obtained during the bombings). For example, the minimal DDs at the 95% probability level were the following: 0.05 to 0.11 Sv (F1 cancers); 0.18 to 0.29 Sv (UPO); 0.68 to 1.10 Sv (F1 mortality); 1.60 Sv (sex-chromosomal aneuploidy), and 2.27 Sv (electrophoretic mutations). When only UPO, F1 cancers, and F1 mortality were considered together, the estimated DD at the 95% probability level was 0.63 to 1.04 Sv. The comparable estimate for sex chromosomal aneuploidy and electrophoretic mutations considered together was 2.71 Sv.
The oft-quoted DD range of 1.69 to 2.23 Sv, called the “most probable gametic DD” by Neel and colleagues, was obtained by calculating overall spontaneous and induced “mutation rates” for the above-mentioned five end-points and obtaining a ratio of these two. The former was estimated by summing the five individual estimates of spontaneous rates (which yielded 0.00632 to 0.00835 per gamete) and the latter, likewise, by summing the individual rates of induction (which yielded 0.00375 per gamete per parental Sv). The ratio 0.00632-0.00835/0.00375 is the DD range which is 1.69 to 2.23 Sv. The overall DDs thus calculated were found to be between 1.69 Sv (i.e., 0.00632/0.00375) and 2.23 Sv (i.e., 0.00835/0.00375) for the acute radiation conditions during the bombings. In these estimates, the limits reflect biological uncertainties about the parameters, but do not take into account the additional error inherent in the estimation process itself, which must be relatively large (Neel and others 1990). With a dose-rate reduction factor of 2 (which was used) for chronic low-LET radiation conditions, the relevant DD becomes about 3.4 to 4.5 Sv. Note, how-
OCR for page 131
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2
ever, that the dose-rate reduction factor traditionally used by UNSCEAR and the BEIR committees is 3, based on specific locus mutation experiments with male mice.
For reasons discussed in the main text, the DDs estimated from these data cannot readily be compared with those used by UNSCEAR and the BEIR committees. However, the results with one indicator of damage used in the Japanese studies, namely, untoward pregnancy outcome, which includes stillbirths, congenital abnormalities, and early neonatal deaths, permit a crude comparison with the risk of congenital abnormalities estimated in this report. The rate of induction defined by the regression coefficient for UPO is (26.4 ± 27.7) × 10−4 per parental sievert, compared to the background risk of 500 × 10−4 assumed in the calculations. The risk of congenital abnormalities (estimated from mouse data in this document) is 60 × 10−4 per Gy−1 for acute X-irradiation, compared to the background risk (human data) of 600 × 10−4. Considering the uncertainties involved in both of these estimates, one can conclude that they are of the same order.
The other end points—namely, F1 mortality, F1 cancers, sex chromosomal aneuploidy, and electrophoretic mobility or activity mutations—that have been used in the Japanese studies have not been used in this report and so do not lend themselves to comparisons. It should be noted that the first two of the above (i.e., F1 mortality, F1 cancers) are multifactorial traits (similar to UPO), and their responsiveness to an increase in mutation rate will depend on the magnitude of the mutation-responsive component, which is quite small, as Neel and colleagues point out. Consequently, the rates of induced genetic damage underlying these traits are expected to be small, and increases will be undetectable with the available sample sizes at the relatively low radiation doses (about 0.4 Sv) sustained by most of the survivors.
The reasons for the lack of significant effects on sex chromosomal aneuploidy and electrophoretic mutations are different. There is no evidence from mouse studies that radiation is capable of inducing chromosomal nondisjunction (the principal basis for the origin of sex chromosomal aneuploidy). Since radiation is a poor inducer of point mutations, a priori one would not expect electrophoretic mutations to be induced by radiation to any great extent as they are known to be due to base-pair changes. Null enzyme mutations would be expected to be induced, but they are unlikely to be found at the low dose levels experienced by most survivors. Consequently, it is not surprising that the DD estimates of Neel and colleagues for these end points (1.60 Sv for sex-chromosomal aneuploids and 2.27 Sv for electrophoretic mutations) are higher than those for the other end points.
Representative terms from entire chapter:
low dose
