The segregation of cancer-causing genes in human populations

Mutation Research, 247 (1991) 191-198 © 1991 Elsevier Science Publishers B.V. 0027-5107/91/$03.50 ADONIS 0027510791000782

191

MUT 00052

The segregation of cancer-causing genes in human populations

William J. Schull University of Texas Health Science Center, Genetics Center, Houston, TX 77225 (U.S.A.)

Keywords: Gene, cancer-causing, segregation of; Cancer-causing genes, segregation of; Human populations, cancer-causing genes in

Summary

Cancer can arise through genetic damage of a variety of sorts, including recessive and dominant mutations, large chromosomal rearrangements, and the inability of cells to repair damaged DNA. Many of these events can be studied by standard methods of genetic analysis and thereby furnish the means to localize the gene to a specific region in the human genome. However, conventional methods of segregation analysis cannot provide the molecular and cellular understanding of the process of gene action essential to informed intervention. Here, recent advances in molecular biology, immunology and biochemistry hold promise of providing the understanding of how normal cells control their replication and why cancer cells do not. Heretofore these techniques have been largely restricted to modest-sized studies, but the requisite assays have now reached a level of development that makes practicable large clinical and population-based studies. Collectively, through these rapidly evolving techniques, we may eventually achieve the acquisition of new methods of prevention, diagnosis and therapy, and a better awareness of the events that order the lives of our cells.

One of the intriguing features of the world's mortality data is the striking similarity in the overall proportion of deaths in the developed nations attributed to cancer. Other common causes of death, such as cardiovascular or cerebrovascu- lar disease, do not exhibit the same degree of similarity. To illustrate the comparability, age- standardized death rates in 1984 for such countries as Canada, England, the Federal Republic of Germany, France, The Netherlands, New Zea- land, and the United States varied only from 132 cancer deaths per 100 000 population in the United

Correspondence: William J. Schull, Ph.D., University of Texas Health Science Center, Genetics Center, P.O. Box 20334, Houston, TX 77225 (U.S.A.).

States to 151 in England, and the difference is even smaller when the age-standardized rates are 'corrected' for differences in the coding of death certificates (Percy and Muir, 1989). Admittedly mortality at individual sites of malignancy, such as breast, colon, lung, ovary, and stomach, can and do vary appreciably, but this makes the comparability of the overall rates the more interesting, particularly when one bears in mind the diverse composition of some of these populations.

One can adduce many reasons why this correspondence in cancer mortality might exist. Cancer is a disease of middle and later fife generally, and individuals in these nations have longer life ex- pectancies, on average, hence greater numbers of years at risk of death from cancer. They also have higher levels of medical care, and reliable cancer

192

mortality statistics entail a certain degree of development of this care. It has further been argued that a convergence in life styles has occurred in these countries in the relatively recent past, and that to the extent that environmental factors are important in the etiology of malignancies exposure to these factors has become more similar. However, it warrants noting that a corresponding multi-national concordance in the occurrence of severe, so-called major congenital malformations led Neel (1958), some years ago, to suggest the existence of a genetic homeostatic mechanism. Obviously it is conjectural whether this or a related hypothesis is applicable to cancer, but the thought is inviting nonetheless given the normal functions of some of the genes that now appear to be involved in carcinogenesis.

The arguments just cited are without exception post hoc ergo propter hoc, and are overly simple from an etiologic perspectiv e in that cancer causation is seen either as genetic or environmental with little or no provision for an interaction of these factors. It is tacitly assumed that a given genotype will result in the same phenotype in all environ- ments, and from this assumption follow state- ments such as 80% of all cancers are environmental in origin. This disregards the wealth of evidence that points to the dependence of gene expression on the environment in which expression occurs, that is, a genotype that may be advanta- geous or at least neutral in one environment may be disadvantageous in another.

Cancer-predisposing genes: their identification

A gene is commonly said to be cancer-predisposing if the frequency of cancer, overall or site- specifically, is higher in individuals heterozygous or homozygous for the candidate gene than the frequency of cancer in the general population. Obviously this definition gives rise to a continuum of cancer-predisposing genes ranging from those where the elevation in cancer frequency is dramatic, and virtually every bearer of the gene(s) develops cancer, to those where the elevation is slight, demonstrable only with very large samples. Conventionally, such genes are identified either by formal segregation analysis or the association of the cancer with known genetic markers in popula-

tion studies. Segregation analysis is based on families, and seeks to ascertain the distribution of the disease within relatives, and ultimately to com- pare the observed distribution with one or more of those expected under various simple genetic hy- potheses. Association studies look for evidence of alleles with different frequencies among disease categories, and commonly involve either a search for distortions of simple population genetic equi- libria, such as the Hardy-Weinberg , or linkage disequilibrium or both. The notion or hypothesis that prompts such studies is simple. If there is no genetic basis to the disease in question, the latter should be equally represented among all marker alleles. If this is not true, there is presumptive evidence that the marker gene, or a closely linked one is involved in the causation of the disease. Ultimately one of the aims of both segregation analysis and association studies is to localize the gene to a specific chromosome and region using linkage analysis (Ott, 1985).

Each of these approaches has its own, often elegant statistical procedures (see, e.g., Morton, 1982; Smouse and Williams, 1982), but they have their limitations. For example, in association studies where the number of alleles at the marker locus is exceptionally large, as in the case of HLA or in the use of certain restriction fragment length poly- morphisms, chance or spurious associations are to be expected, and the separation of these from ' real ' associations is a non-trivial pursuit. It must also be borne in mind that in multiple-locus studies, if the number of haplotypes is large, some may not be represented at all even in very large samples, and thus certain gene associations cannot be tested. It is not my intention, however, to review the statistical procedures used in the demonstra- tion of the existence of cancer-predisposing genes, or, for that matter, to at tempt to catalog all of the genes that have been implicated. Rather it is my aim to address certain general issues using specific cancers to illustrate the point of interest. We shall look at common malignancies as well as the rare, or at least infrequent ones.

Cancer-predisposing genes of infrequent occurrence

Much of what is now known about the role and the nature of the action of genetic factors in

193

cancer susceptibility is based on studies of individuals with either dominant mutant phenotypes or homozygous for recessive mutations. Among the former are disorders such as retinoblastoma and multiple endocrine neoplasia, and among the latter are xeroderma pigmentosum, Fanconi's anemia, Bloom's syndrome, and ataxia telangiectasia. Pro- gress in the study of many of these has moved surprisingly rapidly in the last several years, and the genomic localization of the genes has pro- ceeded apace. Among these tumors arguably the most interesting has been retinoblastoma, but since subsequent speakers will undoubtedly describe the advances in our understanding of this malignancy, our remarks will be limited to just the recognition of the remarkably fruitful role that the Knudson two-mutation hypothesis has played in these advances (Knudson, 1986). However, the frequency of individuals with one or another of these dis- eases in most, if not all populations is low, and therefore their overall contribution to cancer incidence or mortality is relatively small although the gravity of the situation for the affected individuals and their families is obviously large.

For autosomal recessive traits, where the frequency of heterozygous carriers is much higher than that of the trait-bearers themselves, the epidemiological implications are much greater if the heterozygote has an increased risk of spontaneous neoplasms. At least in the case of ataxia telangiectasia (AT) there is evidence that this may be so (Swift and Chase, 1983; Swift et al., 1987). Swift and his colleagues have reported the relative risk is cancer among males and females presumably heterozygous for the AT gene to be 2.3, and 3.1, respectively. Among heterozygous women, breast cancer is the one most conspicuously elevated; the relative risk is 6.8. These authors have further argued that based on a presumed population frequency of 1.4% for heterozygotes (U.S. estimates range from 0.68 to 7.7%), some 9% (8.8 to be exact) of women with breast cancer are heterozygous for the AT gene. Whether this be so or not, and the findings are controversial, it il- lustrates the potential contribution of carriers of rare cancer-predisposing genes to the overall occurrence of cancer.

Comparative national data on the frequency of rare cancer genes, such as those associated with

AT, neuroblastoma, retinoblastoma, or xeroderma pigmentosum, are distressingly limited. Most of the information that we have is based on oppor- tunistic sampling or registries of uncertain com- pleteness, and few surveys would stand rigorous epidemiological scrutiny. It is obviously difficult under these circumstances to know whether the differences that are seen in the frequencies of these recessive phenotypes are real or conse- quences of dissimilar methods of ascertainment. However, one thing does seem to be clear, and that is these disorders are generally genetically heterogeneous. Save with the possible exception of Bloom's syndrome (Weksberg et al., 1988), more than one complementation group commonly exists and the frequencies of these are usually different in different populations (see e.g., Bridges, 1982; Fischer et al., 1985; Duckworth-Rysiecki et al., 1985; Lehmann, 1982; Jaspers et al., 1985).

Neuroblastoma is one of the childhood cancers that has been suggested to follow the Knudson two-mutation model (Knudson and Strong, 1972). The evidence for this is somewhat less persuasive now than previously, and although chromosomal abnormalities have been reported to my knowl- edge no one has been specific to the disease. However, reasonable incidence data are available on this cancer, and these data, though limited, suggest no clear difference among races within the United States (Davis et al., 1987). Further studies of the racial distribution are certainly warranted.

The terms multiple endocrine neoplasia cover a family of disorders characterized by hyperplasia of various organs derived from the neural crest. Two different forms are now recognized, MEN-1 and MEN-2, and both are transmitted in an autosomal dominant mode. MEN-1 has recently been mapped to a small segment, about 12 centimorgans long, on chromosome 11 (Nakamura et al., 1989), and MEN-2 to a position near the centromere of chromosome 10 (Mathew et al., 1987a; Simpson et al., 1987).

Genes and common malignancies

The literature on the role of segregating genes in the occurrence of most common cancers is formidable, often contradictory, and places as big

194

a burden on the reader's perseverance as per- spicacity. However, a familial aggregation of many of these has been repeatedly reported. This is certainly so with respect to breast cancer (see e.g., Dupont and Page, 1987, or Negri et al., 1988), cancer of the ovary (Mori et al., 1988), colorectal cancer (Knudson, 1986), and cancer of the lung (Heighway et al., 1986) to mention but a few. Indeed, Schwartz et al. (1988) have estimated that among cancers overall some 27% of the variability in cancer occurrence within families is due to familial factors. It should be noted that this is not synonymous with saying that 27% of the variability is genetic in origin, for most studies of aggregation do not, often cannot distinguish between genetic and shared environmental factors. This fact notwithstanding site-specific aggregation generally obtains irrespective of the magnitude of the age-adjusted mortality rate. For example, although mortality from ovarian cancer has been gradually rising in Japan, it remains among the lowest in the developed countries. Yet, a history of a breast, uterine or ovarian cancer in a mother or sister is one of the most striking risk factors associated with ovarian cancer in a Japanese woman (Mori et al., 1988). Similar findings have been reported in the United States where ovarian malignancies are substantially more common (Schildkraut and Thompson, 1988). Often, however, the association of a specific cancer with a family history of that cancer appears complex. As an illustration, Dupont and Page (1987) find that the cancer risk associated with a late age of first birth as contrasted with nulliparity is greater among women with a positive family history of breast cancer than those without. Whatever the nature of the interaction may be, credence is given to its reality by the comparabili ty of the findings in different studies, as for example, in the Ameri- can (Dupont and Page, 1987) and Italian studies (Negri et al., 1988) of breast cancer.

Cancers of the colon are of progressively greater importance as cause of mortality in the United States, and such cancers are often familial. More- over, at least 20% of common colon tumors have deletions of all or part of chromosome 5, and if the technical limitations of the assay are taken into account, this value could be as high as 40% (Solomon et al., 1987). And recently, Bodmer and

his colleagues (1987) have localized the gene for familial adenomatous polyposis to chromosome 5.

Distributed through the human genome are a number of genes, commonly designated as proto- oncogenes, whose roles in cancer causation are still poorly understood although in another form as oncogenes they can cause neoplastic growth and are, therefore, presumably not solely adventi- tious. Many of these oncogenes have a normal role to play in cellular proliferation and differentia- tion, and the specific role of a number is known. Their products are growth factors, growth factor receptors, t ransmembrane signaling proteins and the like. Some 50 or so of these oncogenes are already known (see review in Bishop, 1987), and the list continues to grow. Commonly alterations or amplification of these oncogene sequences are observed in malignant tumors. This has been seen with regard to cancers of the lung, colon, and bladder to mention a few where single-base-pair substitutions in the sequence have been described.

Hall and his colleagues (1989) have examined the relationship of 9 such sequences to the susceptibility to breast cancer; the specific sequences were HRAS, KRAS2, NRAS, INT2, MYB, MYC, MOS, RAF1, and ERBA2. Linkage analyses failed to suggest that polymorphism at any one of these loci was associated with susceptibility, and these authors, therefore, conclude that these oncogenes are not the primary sites of alterations leading to breast cancer. This analysis could not, of course, exclude the possibility that alterations at these sites were involved in tumor progression. More recently Genuardi et al. (1989), using recombinant D N A probes, have shown a high frequency of reduction to homozygosity at the highly polymor- phic locus D1. Z2, localized on lp36, in women with ductal breast carcinomas. This loss of hetero- zygosity was more frequent in patients with the characteristics of hereditary tumors - - strong family history, early diagnosis, multiple tumors or tumor foci - - than in patients with none of these characteristics. They speculate that a fundamental step in the pathogenesis of ductal carcinoma of the breast may be the inactivation of a tumor suppressor gene located on the distal portion of chromosome lp. Interestingly, this portion of this chromosome has also been implicated in the pathogenesis of neuroblastoma (Gilbert et al., 1984),

familial melanoma (Greene et al., 1983), and MEN-2 (Mathew et al., 1987b).

The relationship between chromosomal change and oncogenesis has also been extended by the finding of chromosomal fragile sites close to the location of known oncogenes (see e.g., Mitelman and Heim, 1988). Many such sites appear to be targets for various DNA-damaging agents, such as ionizing radiation. This has led to the suggestion that individuals with rearrangements at these fragile sites may have an increased predisposition to breakage that is heritable, manifesting itself in an increased risk of familial cancer. Whether this is so remains to be established; however, Mules and her colleagues failed to find an increased risk of cancer among the relatives of leukemic patients with chromosomal rearrangements at rare, heritable fragile site locations in their malignant cells (Mules et al., 1989).

Segregation of host susceptibility factors

For a variety of reasons, most cancer-oriented genetic research has focused on those genes which appear directly related to cancer etiology and little attention has been directed toward those cancers where other etiologic factors may be involved, but where genetic variability may still be instrumental in the occurrence of the cancer, or its natural course, or its response to therapy. To state it somewhat differently, little has been done to ex- plore inherited differences in host susceptibility. Let me illustrate the point I wish to make as follows.

Viewed globally, hepatocellular carcinoma is one of the world's most common cancers, and it has been estimated that in 75-90% of such in- stances the hepatitis B virus is the etiologic agent. It is known that hepatitis B virus DNA is almost universally incorporated into the genome of hepatocytes of individuals with chronic infections as manifested by positivity to HBsAg; yet other individuals apparently eliminate the virus and develop protective antibodies (anti-HBs). Why this should occur is not clear, but there is ample basis for speculating that this variability in response to the virus is genetic. Evolutionarily the HB virus is probably related to the retroviruses, and the existence and influence of viral restriction genes

195

have been extensively documented in animal mod- els. Some block retroviral incorporation into the host genome, still others apparently block viral replication. Little is yet known about the role that such genes may play in human hepatocellular carcinoma, or in human oncogenesis more generally, but it would certainly seem a fruitful area for research.

Similarly, gallbladder cancer although relatively infrequent in the United States appears elevated among most populations of Amerindian origin and admixed populations such as the Mexican- Americans. It is a particularly common cancer in certain South American populations, such as in Bolivia. These populations appear especially prone to cholecystitis and cholelithiasis, but the role of genetic factors in this proclivity is still unclear. Estimates of admixture in mixed populations provide one simple method of beginning the search for the relevant susceptibility genes.

Epidemiological studies have shown that human populations exposed to low-LET radiation exhibit an excess in a range of malignancies, and that, in general, these do not differ markedly, if at all, from those malignancies arising sponta- neously. This finding contrasts with that commonly seen with chemical agents where the neo- plasm is often organ specific, for example, asbes- tos and mesothelioma. The basis for this difference is not known, but it is commonly specu- lated that it stems from the diversity of effects on DNA of ionizing radiation as contrasted with the more specific interactions of chemicals with DNA. Of particular interest, however, is the existence of inherited disorders, such as AT, that are radiosen- sitive. This suggests, in turn, that we are not all equally likely to develop a malignancy following exposure to ionizing radiation. The implications of this are far-reaching. Extrapolations of risk have invariably been based on the supposition that we are all equally prone to develop a radiation-related malignancy, and that the probability of such an event is determined largely by dose and age at exposure. If differences in proneness exist, then these extrapolations may overestimate the risk to most individuals and substantially underestimate that to some. As yet, even among the survivors of the atomic bombing of Hiroshima and Nagasaki, we know little about the frequency or possible role

196

of inherited differences in the capacity to repair D N A damage in the increased risk we observe.

Induced germinal mutations and the occurrence of cancer

If now genes do play a significant role in the carcinogenic process, and certainly a growing body of information suggests this, and if mutat ion at the appropriate loci represents an initiating or converting event, then it is reasonable to suppose that the offspring of individuals exposed to muta- genic agents will exhibit higher rates of malignancy. One unequivocal mutagen is ionizing radiation, and it has been established that exposure to radiation increases the frequency of a variety of cancers in the exposed individual (see e.g., Shimizu et al., 1987, 1988). The atomic bombing of Hiroshima and Nagasaki precipitated a series of genetic studies that still continue. One aspect of these is a mortality surveillance of a cohort, ini- tially numbering about 53 000 individuals but now embracing some 75000 persons, most the offspring of exposed parents. Some years ago (see Schull et al., 1982), we examined the frequency of cancer mortality in the initial cohort. At that time, the average individual within this cohort was 17 years old, that is, had been exposed to 17 years of risk of death. Among the 3552 deaths that had occurred 72 were ascribed to malignancy. The most common of these latter causes of death was leukemia; indeed, 35 of the 72 were attributed to this disease. No clear trend in the occurrence of these causes of death with radiation was visible. Recently, as a result of the installation of a new system for computing the doses of individual survivors, further surveillance and the opportunity to supplement the mortality experience with incidence data, malignancies in this cohort have been examined anew (Yoshimoto et al., 1989). Among some 67 645 individuals where a DS86 dose could be assigned to their parents, 83 had died of a malignancy before the age of 20. A multiple linear regression analysis revealed no increase in childhood cancers among these individuals. However, examination of the data suggested that only some 2 4% of the tumors of childhood observed in the comparison groups are associated with an inherited genetic predisposition which would be ex-

pected to exhibit an altered frequency if the parental mutation rate were increased. Thus, in the statistical sense, the power of the data to detect an increase is small despite the size of the sample. Whether there will be an increase in adult-type malignancies with increasing dose remains to be seen, for few of these individuals have yet reached those ages in life where these cancers are common.

Conclusions

Where now do these somewhat disjointed thoughts and speculations lead us? First, it should be clear that cancer can arise through genetic damage of a variety of sorts; these include recessive and dominant mutations, large chromosomal rearrangements, and the inability of cells to repair damaged DNA. Many of these sorts of damage can be studied by conventional methods of genetic analysis and can culminate in the localization of the gene to a specific region in the human genome. Second, it should be equally clear that conventional methods of analysis cannot provide the molecular and cellular understanding to the process of gene action essential to informed intervention. Here, however, recent advances in molecular biology, immunology and biochemistry hold promise of providing the means to understand how normal cells control their replication and why cancer cells do not. Heretofore these techniques have been largely restricted to modest-sized studies, but as Taylor (1989) has cogently argued the requisite assays have now reached a level of development making practicable large clinical and population-based studies. Collectively, as Bishop (1987) has noted through these rapidly evolving techniques 'we may eventually achieve an even grander goal (than just the acquisition of new methods of prevention, diagnosis and therapy), to grasp the designs that order the lives of our cells'.

References

Bishop, J.M. (1987) The molecular genetics of cancer, Science, 235, 305-311.

Bodmer, W.E., C.J. Bailey, J. Bodmer, H.J.R. Bussey, A. Ellis, P. Gorman, F.C. Lucibello, V.A. Murday, S.H. Rider, P. Scambler, D. Sheer, E. Solomon and N.K. Spurr (1987)

Localization of the gene for familial adenomatous polyposis on chromosome 5, Nature (London), 328, 614-616.

Bridges, B.A. (1982) Some DNA repair-deficient human syn- dromes and their implications for human health, in: T. Sugimura, S. Kondo and H. Takebe (Eds.), Environmental Mutagens and Carcinogens, Liss, New York, pp. 47-57.

Davis, S., M.A.M. Rogers and T.W. Pendergrass (1987) The incidence and epidemiologic characteristics of neuroblastoma in the United States, Am. J. Epidemiol., 126, 1063-1074.

Duckworth-Rysiecki, G., K. Cornish, C.A. Clarke and M. Buchwald (1985) Identification of two complementation groups in Fanconi anemia, Somat. Cell. Mol. Genet., 11, 35-41.

Dupont, W.D., and D.L. Page (1987) Breast cancer risk associated with proliferative disease, age at first birth, and a family history of cancer, Am. J. Epidemiol., 125, 769-779.

Fischer, E., W. Keijzer, H.W. Thielmann, O. Popanda, E. Bohnert, L. Elder, E.G. Jung and D. Bootsma (1985) A ninth complementation group in xeroderma pigraentosum, Mutation Res., 145, 217-225.

Genuardi, M., H. Tsihira, D.E. Anderson and G.F. Saunders (1989) Distal deletion of chromosome Ip in ductal carcinoma of the breast, Am. J. Hum. Genet., 45, 73-82.

Gilbert, F., M. Feder, G. Balaban, D. Brangman, D.K. Lurie, R. Podolsky, V. Rinaldt, et al. (1984) Human neuro- blastomas and abnormalities of chromosomes 1 and 17, Cancer Res., 44, 5444-5449.

Greene, M.H., L.R. Goldin, W.H. Clark Jr., E. Lovrien, K.H. Draemer, M.A. Tucker, D.E. Elder, et al. (1983) Familial cutaneous malignant melanoma: autosomal dominant trait possibility linked to the Rh locus, Proc. Natl. Acad. Sci. (U.S.A.), 80, 6071-6075.

Hall, J.M., P.J. Zuppan, L.A. Anderson, B. Huey, C. Carter and M.C. King (1989) Oncogenes and human breast cancer, Am. J. Hum. Genet., 44, 577-584.

Heighway, J., N. Thatcher, T. Cerny and P.S. Hasleton (1986) Genetic predisposition to human lung cancer, Br. J. Cancer, 53, 453-457.

Jaspers, N.G.J., R.B. Painter, M.C. Paterson, C. Kidson and T. Inoue (1985) Complementation analysis of ataxia- telangiectasia, in: R.A. Gatti and M. Swift (Eds.), Ataxia- telangiectasia: Genetics, Neuropathology, and Immunology of a Degenerative Disease of Childhood, Liss, New York, pp. 147-162.

Knudson, A.G. (1986) Genetics of human cancer, Annu. Rev. Genet., 20, 231-251.

Knudson, A.G., and L.C. Strong (1972) Mutation and cancer: Neuroblastoma and pheochromocytoma, Am. J. Hum. Genet., 24, 514-532.

Lehmann, A.R. (1982) Three complementation groups in Cockayne syndrome, Mutation Res., 106, 347-356.

Mathew, C.G.P., K. Chin, D. Easton, K. Thorpe, C. Carter, G. Liou, S.L. Fong, C. Bridges, H. Haak, A. Nieuwenhnijzen- Kruzeman, S. Schifter, H. Hansen, H. Telenius, M. Tele- nius-Berg and B. Ponder (1987a) A linked genetic marker for multiple endocrine neoplasia type 2a on chromosome 10, Nature (London), 328, 527-528.

197

Mathew, C.G.P., B.A. Smith, K. Thorpe, Z. Wong, N.J. Royle, A.J. Jeffreys and B.A.J. Ponder (1987b) Deletion of genes on chromosome 1 in endocrine neoplasia, Nature (London), 328, 524-526.

Mitelman, F., and S. Helm (1988) Consistent involvement of only 71 of the 329 chromosomal bands of the human genome in primary neoplasia-associated rearrangements, Cancer Res., 48, 7115-7119.

Moil, M., I. Harabuchi, H. Miyake, J.T. Casagrande, B.E. Henderson and R.K. Ross (1988) Reproductive, genetic, and dietary risk factors for ovarian cancer, Am. J. Epide- miol., 128, 771-777.

Morton, N.E. (1982) Outline of Genetic Epidemiology, Karger, Basel, pp. viii and 247.

Mules, E.J., J.R. Testa, G.H. Thomas, H. Abbey and B.H. Cohen (1989) Cancer in relatives of leukemic patients with chromosomal rearrangements at rare (heritable) fragile-site locations in their malignant cells, Am. J. Hum. Genet., 44, 811-819.

Nakamura, Y., C. Larsson, C. Julier, C. Bystrom, B. Skogseid, S. Wells, K. Oberg, M. Carlson, T. Taggart, P. O'Connell, M. Leppert, J.M. Laouel, M. Nordenskjold and R. White (1989) Localization of the genetic defect in multiple endocrine neoplasia type 1 within a small region of chromosome 11, Am. J. Hum. Genet., 44, 751-755.

Neel, J.V. (1958) A study of major congenital defects in Japanese infants, Am. J. Hum. Genet., 10, 398-445.

Negri, E., C. La Vecchia, P. Bruzzi, G. Dardanoni, A. Decarli, D. Palli, F. Parazzini and M. Rosselli del Turco (1988) Risk factors for breast cancer: pooled results from three Italian case-control studies, Am. J. Epidemiol., 128, 1207-1215.

Ott, J. (1985) Analysis of Human Genetic Linkage, Johns Hopkins Press, Baltimore, MD.

Percy, C., and C.S. Muir (1989) The international comparability of cancer mortality data, Am. J. Epidemiol., 129, 934- 946.

Sasaki, M.S. (1989) Cytogenetic aspects of cancer-predisposing genes, Jpn. J. Hum. Genet., 34, 1-16.

Schildkraut, J.M., and W.D. Thompson (1988) Familial ovarian cancer: A population-based case-control study, Am. J. Epidemiol., 128, 456-466.

Schull, W.J., J.V. Neel, M. Otake, A. Awa, C. Satoh and H.B. Hamilton (1982) Hiroshima and Nagasaki: Three and a half decades of genetic screening, in: T. Sugimura, S. Kondo and H. Takebe (Eds.), Environmental Mutagens and Carcinogens, Liss, New York, pp. 685-700.

Schwartz, A.G., M. Boehnke and P.O. Moll (1988) Family risk index as a measure of familial heterogeneity of cancer risk, Am. J. Epidemiol., 128, 524-535.

Shimizu, Y., H. Kato, W.J. Schull, D.L. Preston, S. Fujita and D.A. Pierce (1987) Life span study report, II. Comparison of risk coefficients for site-specific cancer mortality based on the DS86 and T65D shielded kerma and organ doses, Radiat. Effects Res. Found. Tech. Rep., 12-87.

Shimizu, Y., H. Kato and W.J. Schull (1988) Life span study report, II. Part 2. Cancer mortality in the years 1950-85 based on the recently-revised doses (DS86), Radiat. Effects Res. Found. Tech. Rep., 5-88.

198

Simpson, N.E,. K. Kidd, P. Goodfellow, H. McDermid, S. Myers, J. Kidd, C. Jackson, A. Duncan, I. Farrer, K. Brasch, C. Castiglione, M. Genel, J. Gertner, C. Greenberg, J. Gusella, J. Holden and B. White (1987) Assignment of multiple endocrine neoplasia type 2A to chromosome 10 by linkage, Nature (London), 328, 529-530.

Smouse, P.E., and R.C. Williams (1982) Multivariate analysis of HLA-disease associations, Biometrics, 38, 757-768.

Solomon, E., R. Voss, V. Hall, W.F. Bodmer, J.R. Jass, A.J. Jeffreys, F.L. Lucibello, I. Patel and S.H. Rider (1987) Chromosome 5 allele loss in human colorectal carcinomas, Nature (London) 328, 616-619.

Swift, M., and C. Chase (1983) Cancer and cardiac deaths in obligatory ataxia-telangiectasia heterozygotes, Lancet, i, 1049-1050.

Swift, M., P.J. Reitnauer, D. Morrell, et al. (1987) Breast and

other cancers in families with ataxia-telangiectasia. New Engl. J. Med., 316, 1289-1294.

Taylor, J.A. (1989) Oncogenes and their applications in epidemiological studies, Am. J. Epidemiol., 130, 6-13.

Weir, B.S. (1979) Inferences about linkage disequilibrium, Bio- metrics, 35, 235-254.

Weksberg, R., C. Smith, L. Anson-Cartwright and K. Maloney (1988) Bloom syndrome: A single complementation group defines patients of diverse ethnic origin, Am. J. Hum. Genet., 42, 816-824.

Yoshimoto, Y., J.V. Neel, W.J. SchulL H. Kato, M. Soda, R. Eto and K. Mabuchi (1989) The frequency of malignant tumors during the first two decades of life in the offspring of atomic bomb survivors, Radiat. Effects Res. Found. Tech. Rep., x-89.

Documents

The segregation of cancer-causing genes in human populations