Hereditary Predisposition to Cancer

Hereditary Predisposition to Cancer ALFRED G. KNUDSON"

Institfite f ir Cancer Research, Fox Chase Cancer Center, Philadelphia, Pennsylvania 191 11, USA

ABSTRACT: Both hereditary and environmental factors influence the risk of cancer. Four risk categories, or oncodemes, can exist for a particular kind of cancer, depending upon the presence of neither, one, or both factors: (1) spontaneous, or background; (2) hereditary; (3) environmental; (4) in- teractive. In the second, mutation imparts a high relative risk, but a generally low attributable risk; in the fourth, the opposite obtains. The second oncodeme contains genes that are also important for the non-hereditary forms of the same cancer. Probably all forms of cancer exist in a dominantly heritable form. Most of the genes are tumor suppressors, although a few are oncogenes or DNA repair genes. The mutations are in most, if not all cases, maintained in a population by an equilibrium between mutation and selection.

Most of the cloned genes are expressed widely among tissues, yet there is typically some tumor specificity. Somatic mutations in second alleles at the relevant loci are necessary, but generally not sufficient for carcinogenesis, although they, in some instances, lead to the formation of benign precursor lesions. Further events are necessary for carcinogenesis. This is particularly true for carcinomas. The benign lesions appear to involve an increase in number of long-lived cells that can accumulate other mutations. For some tumors, physiologic events, such as tissue growth at puberty or proliferation of embryonic stem cells, may produce this effect. Mutations of DNA mismatch repair genes underscore the effect that changes in somatic mutation rates can have, especially in the risk for multi-event carcinomas. Conversely, these are the tumors that offer the greatest opportunity for prevention.

INTRODUCTION

Cancer may be considered as a somatic genetic disease; the genotypes of the lesions differ from the genotypes of the hosts. One may ask whether this difference is the cause or the effect of the carcinogenetic process, and the answer should probably be that much of the difference is an effect of cancer but that some of it is causative. Some of the clearest evidence derives fkom the fact that some mutations associated with cancer occur in the same genes whose homologues are the oncogenes of certain RNA tumor viruses, as is the case with the R A S and MTC families of genes. Other mutations inacti-

"Address for telecommunication: Phone, 215/728-3642; fax, 215/728-3105.

58

KNUDSON: HEREDITARY PREDISPOSITION 59

vate tumor-suppressor genes that can also be inactivated by protein products of the transforming genes of certain DNA tumor viruses; the R B I and P 5 3 genes exemplifl this phenomenon. Another kind of evidence comes from hereditary cancer, when the germ-line mutation occurs in a gene, such as RBI, 27’53, W T I , or APC, that is often mutated somatically in the non- hereditary form of the same tumor to which the germ-line mutation predisposes. In fact, we now recognize two large classes of cancer genes, oncogenes and anti-oncogenes (tumor-suppressor genes).

Of course, we also recognize that environmental agents can cause cancer. Does this recognition create a contradiction? We think not. It often happens that the same genes that are identified with hereditary cancer are also somatically mutant in “environmental” cancer; e.g., small cell carcinoma of the lung, nearly always associated with cigarette smoking, features somatic mutations in R B I and P53.

Mutations in both somatic and germinal cells can be increased above background rates by environmental agents, notably certain kinds of radiation and chemicals. What “causes” background rates is not completely known, but the causes seem to be universal, not subject to human manipu- lation, and therefore not “environmental” in the usual sense of the word. Thus, cosmic radiation contributes to background mutation rates. The importance of “background” mutations for cancer lies in the fact that as long as they occur we shall experience some fraction of the present incidence of cancer. So, while prevention offers the possibility of substantial reduction in age-specific cancer rates, we should not expect complete success, and we must retain an interest in treatment.

The familiar forms of hereditary predisposition to cancer are attributable to dominantly inherited germ-line mutations that are usually rare and produce high relative risks for one or more forms of cancer. A few rare recessively inherited diseases also predispose to cancer, but these are few in number, although interesting with respect to mechanism. Thus, xeroderma pig- mentosum, Fanconi’s anemia, Bloom’s syndrome, and ataxia telangiectasia (AT) involve defects in repairing certain kinds of genetic damage. All of these are rare conditions, so again the fraction of cases of cancer that can be attributed to them is small. However, the heterozygotes for AT, but not the other recessive disorders, are reported to be at increased risk of breast cancer, in which case they too can be described as having dominantly heritable cancer.’ Finally, there are predispositions with much lower relative risks that are imparted by polymorphic genes whose products react differently to environmental agents that can cause cancer; the attributable risks could be high for such conditions.

Out of these considerations comes the realization that we may identifl four categories of persons in any population with respect to the roles of heredity and environment in the causation of a particular kind of cancer. I have used the term oncodemes to describe these group^.^ The importance of

60 ANNALS NEW YORK ACADEMY OF SCIENCES

any oncodeme for a particular form of cancer varies greatly. Thus, for retinoblastoma the background and hereditary oncodemes probably account for most of the cases, while for lung cancer the environmental and in- teractive oncodemes probably do so. The dominantly inherited subset of the “hereditary” oncodeme is what I shall discuss here.

DOMINANTLY INHERITED CANCERS

Dominant inheritance of virtually every form of cancer has been reported. Only occasionally is this predisposition limited to one or two types, but on the other hand it never extends to all kinds of cancer. For example, hereditary paraganglioma and hereditary Wilms’ tumor are not usually associated with other tumors, and pheochromocytoma and medullary carcinoma of the thyroid are the only tumors featured in multiple endocrine neoplasia type 2 (MEN2). Persons with hereditary retinoblastoma frequently later develop osteosarcoma or soft tissue sarcomas, and those with the Li- Fraumeni syndrome (LFS), sarcomas, and breast cancer. Several kinds of carcinoma are featured in hereditary non-polypoid colon cancer (HNPCC), but embryonal tumors are not and sarcomas are uncommon. Neurectoder- mal and endocrine tumors are commonly featured in hereditary cancer.

Most of the dominantly heritable cancers are rare, that is, fewer than one per 10,000 persons born carry such a mutant gene, and this incidence changes slowly, if at all, with time. This phenomenon results fiom the estab- lishment of a mutational equilibrium, by which some mutant gene carriers die before completing reproduction and are replaced by new mutants. The survival value, a term applied to the reproductive fitness of an individual and which is assigned the value of unity for a “normal” person, can vary in this situation fiom zero to one, and reflects the probability of both early death and diminished fertility. For neurofibromatosis type 1 (NFl), this value is approximately 0.5, and new germ-line mutants account for 50 percent of all cases, thus perpetuating an equilibrium. We note, incidentally, that NF1 has a high incidence of one per 3000 persons and a correspondingly high germ- line mutation rate.

HNPCC and hereditary breast cancer due to mutation in BRCAl or BRCA2 stand in great contrast to these conditions, with unusually high in- cidences, of the order of 1-10 per 1000 persons. Even a small decrease in survival value should render these conditions less frequent, because mutation rates could not be high enough to compensate for gene losses. In fact, newly mutant cases have been extremely rare or nonexistent, and probably constitute less than 1 percent of total cases. However, there is selection against homozygotes, as is also the case for the rarer disorders. This is known through the construction of “knockout” heterozygous mice, which can then be bred to produce homozygotes. In nearly every instance the ho-

KNUDSON HEREDITARY PREDISPOSITION 61

mozygous animal dies from defective fetal development. In humans, this would translate to very small rates of mutant loss. For example, if the gene frequency for one of the conditions were then the heterozygote frequency would be 2 x and the homozygote frequency, 10“. Only two mutant genes would be lost per million births (or 2 million genes), which would require a mutation rate of only 10“ per generation to replace the loss. This, in turn, would create 2 x 10“ new heterozygotes per generation, only one per thousand total heterozygotes. Obviously, when selection oper- ates only against homozygotes, heterozygote frequencies can become quite high, and very few will be new mutants. It also means that a new mutation can by chance be either eliminated very soon after its occurrence or perpet- uated over very long periods. This latter phenomenon, the founder effect, has been observed already for both HNPCC and hereditary breast cancer due to BRCAl and BRCA2, the latter quite spectacularly so among Ameri- cans of Ashkenaszi Jewish de~cent .~

RETINOBIASTOMA: A PROTOTYPE

Even though a few of the forms of dominantly inherited cancer are rather common, the group of hereditary cancers as a whole is small, and probably constitutes less than 5 percent of all cases of cancer. Clearly they are important for the affected families, but what is their relationship to the common non-hereditary forms of cancer, including those for which environmental influences are so important? The first tumor to reveal an important relationship was retinoblastoma (Rb), a rare (five cases per lo5 births worldwide) tumor of children. Approximately 40 percent of cases are attributable to mutations in the germ line, and the remainder are “non-hereditary.” A large majority of the patients with germ-line cases do not provide a family history, because they represent new mutations; but 50 percent of their offipring are at risk of the tumor. What is the relationship between the two forms? First, we take note of the fact that a small fraction of gene carriers do not develop the tumor, which informs us that inheritance of the mutation is not a sufficient condition for oncogenesis. I proposed that a second mutation, occurring in the somatic retinoblasts during early life was necessary; retinoblastoma could be considered as a “two-hit” lesion, one in the germ line, one in somatic cells.4 Furthermore, the non-hereditary disease could also be considered as a “ t ~ ~ - h i t ” tumor, with both mutations occurring somatically, after conception. The only difference between the two forms was the time of occurrence of the first mutation.

The simplest explanation was that the two hits were mutations in the two copies of an Rb gene, now known as RBI. I found that the mean number of tumors per carrier was approximately three, and the numbers in a series of patients followed a Poisson distribution, as if by chance. Given the number

62 A N N U NEW YORK ACADEMY OF SCIENCES

of replications of retinoblasts during the development of the eye, it was ap- parent that the probability of the second event was of the same order of magnitude as somatic mutation rates. The non-hereditary cases could also be explained by ordinary “spontaneous” somatic mutation rates.5 Indeed, my estimate is that about 30 per cent of persons bear clones of “once-hit” cells in their eyes, but tumors do not usually develop because no second event occurs before the cells become postmitotic. The second events can be of several kinds, including intragenic mutations (point mutations, deletions, insertions, duplications), chromosomal deletion, chromosomal non-disjunc- tion and loss, and somatic recombination. The splendid work of Cavenee e t af.6 subsequently demonstrated these mechanisms, and the cloning of RBI has shown us that the hereditary and non-hereditary forms of this tumor are so related. Predisposition to cancer is dominantly inherited, but oncogenesis itself is recessive. Tumor formation results fiom loss or inactivation of both copies of a gene, whose normal allele may be considered as an antioncogene or tumor-suppressor gene.

PHENOTYPE: TUMOR SPECIFICITY

Most of the hereditary forms of cancers are attributable to heterozygosity for tumor-suppressor genes, but a few involve proto-oncogenes, while for another condition, HNPCC, DNA-mismatch repair genes. For all there is a set of tumors to which the host is predisposed, as noted earlier. What deter- mines this specificity? In one case, the answer is relatively simple. For the WTI gene, which is mutant in 10-15 percent of Wilms’ tumors, with a few percent being germ-line mutations, the explanation is a normal expression of the gene in a limited number of tissues, primarily those of the genitourinary t r a ~ t . ~ However, such an explanation is not a general one, since most of the cloned genes are widely expressed, and much of that expression occurs in tissues that give rise to tumors to which the mutant host is not predisposed. It is also noteworthy that predisposition extends disproportionately to some tumors that are much less common than the carcinomas. Thus the NFl gene predisposes to neurofibromas and neurofibrosarcoma, and the RET protooncogene mutation in multiple endocrine neoplasia type 2, to pheochromocytoma and medullary carcinoma of the thyroid. Even though 27’53 is often mutant somatically in the non-hereditary carcinomas, germ- line mutation predisposes especially to sarcomas, and just one carcinoma, that of the breast. Survivors of Rb who have germ-line mutations are also susceptible to sarcomas as second cancers occurring later.

The analysis of colon cancer, and of a strongly predisposing condition, fa- milial adenomatous polyposis (FAP), has been very helpful in this regard.8 FAP is so named for the precursor lesion, which commonly occurs by the


hundreds or even thousands in affected persons. These polyps contain cells that have sustained “second hits”, that is, the cells are homozygously defective for the responsible APC gene.9 Despite the very large number of polyps, only one or a few carcinomas are found in a given subject. At the cellular level malignant transformation is a very rare event. Comparison of genetic changes in polyps and carcinomas reveals that several other mutations have occurred in the latter that are not found in the former. One of the most frequent of these is in the Tp53 gene, yet germ-line mutation in this gene, which is found in many cases of the Li-Fraumeni syndrome (LFS), does not predispose to colon cancer to any great extent. Why does the APC gene, but not Tp53, predispose to colon carcinoma? The answer seems to lie with the nature of the polyp. It is clearly a hyperplastic lesion, which contains many more cells than the epithelial surface from which it arises. It appears that typical renewal type cell divisions are not the only kinds of division that have occurred; there is an increase in the number of symmetric replicative divisions, thereby increasing the number of tissue stem cells. This phenomenon has the effect of increasing the number of target cells available for transformation by subsequent events. The number of polyps is very large because just one somatic event must occur, as with hereditary retinoblastoma, but the number of carcinomas is small because several other events are necessary.

Why then do germ-line mutations in R B I yield a very high incidence of malignant tumors? Why is only one somatic event necessary in that situation? The most obvious difference between the tumors is that retinoblastoma arises in an embryonal tissue that undergoes enormous numbers of symmetric replicative cell divisions.’O Here mutation is a transforming event. No equivalent of an APC mutation is necessary to increase the number of target cells. Support for this proposition comes from the observation that a non-genetic condition, chronic ulcerative colitis, also predisposes to colon cancer. Here it seems that ulceration stimulates stem cell growth during repair. Support for this idea comes from the observation that carcinomas that arise in persons with ulcerative colitis show a high fiequency of TP53 mutations, but a very low frequency of APC mutations.” Stem cell proliferation can be accomplished in two different ways, one genetic and one non- genetic. Returning then to LFS, we can employ this theme to attempt to ex- plain the prominence of sarcomas and breast cancer in that syndrome. One explanation is that these tumors are the ones that arise in tissues that are un- dergoing growth during adolescence, again a situation whereby stem cell proliferation occurs. No additional mutation is required, because the proliferation occurs physiologically, as in embryonal tissues. Why Rb is not a feature of LFS is not obvious, although we note that 7F53 mutations are rarely, if ever, found in that tumor, so the gene may be irrelevant in that tissue.

How general is the idea that “two-hit” lesions in epithelial tissues are be-

64 ANNALS NEW YORK ACADEMY OP SCIENCES

nign precursors of carcinoma? We do not know the answer, but two conditions that produce renal carcinomas may bear on the question. One of these is the von-Hippel-Lindau syndrome (VHL), whose responsible gene has been cloned. Hundreds of very small tumors can be found in the kidneys of such persons, yet only a few progress to clinically important cancer. A similar phenomenon is found in the Eker rat, a rat model for dominantly inherited cancer which is due to germline mutation of the rat homologues of the human tuberous sclerosis type 2 (TSC2) gene.12J3 Again, numerous very small tumors only in some instances progress to invasive carcinoma, and these show additional cytogenetic abnormalities. These small tumors should, in both diseases, be labeled adenomas rather than carcinomas. As with adenomatous polyps, even the smallest tumors contain cells that have lost the second allele of the VHL and TSC2 genes, respectively. But these small tumors seem to be just “two-hit” lesions. In the Eker rat hrther evidence for this notion is provided by the induction of very small tumors by ionizing radiation. The dose-response relationship at 0, 3, 6, and 9 Gy is linear, suggest- ing that just one rate-limiting event (the second hit) is involved in the pro- duction of very small tumors.14

Not enough is known about precursors of carcinoma to conclude how general this two-hit precursor phenomenon may be. Intensive search is on for such lesions in breast, and the well-known precursors in prostate may soon be subjected to such analysis. Nevertheless, some epithelial tissues show such a phenomenon, and it may well be general for predisposing conditions that result from heterozygosity for a mutant tumor-suppressor gene. Precursor lesions are also known for at least one proto-oncogene (REn disorder, MEN2. Focal hyperplasia of C-cells in the thyroid medulla is charac- teristic of this condition, and there is a question whether their initiation may entail a second event.

The necessity of more events in some tissues than others probably also accounts for a well-known observation with respect to the tumor spectrum in persons carrying a germ-line mutation in RBI. Since the great majority of such persons are now survivors of Rb, they can survive to acquire tumors at other sites, most of which are soft-tissue sarcomas or osteosarcomas. The relative risk of osteosarcoma is of the order of 1000-fold times normal. For Rb itself the relative risk for one tumor is even greater, about 105-fold, since the carrier develops a mean of three tumors while just 3 per lo5 persons develop non-hereditary Rb. This may reflect the fact that many osteosarcomas have also sustained 27’53 mutations, or two more events. The impact of the germ-line mutation has been diluted. Furthermore, another tumor, small cell carcinoma of the lung, which nearly always shows mutations of both R B I and 72’53, but also has deletions in chromosomal arm 3p, the site of a putative tumor-suppressor gene, is not featured in these individuals; the pre- sumptive extra “hits” required for this tumor still hrther dilute the impact of the germ-line mutation.’O


IMPLICATIONS FOR PREVENTION

We take note here of our mutual interest in preventing cancer, and call attention to the idea that prevention may be very difficult for a tumor like retinoblastoma. The genetically predisposed individual might be identified, but since only one somatic event is required, and since it occurs before or soon after birth, there is little opportunity. O n the other hand, the multiple events on the paths to carcinoma permit opportunities for environmental influences to modulate the penetrance of the gene. In some instances the risk of cancer is greatly increased, but on the other hand intervention might reduce the rate of transit along this longer path.

One genetic condition illustrates dramatically the relationship between number of events and rate of progression along a path to cancer. This is HNPCC, which is most frequently caused by mutation in one of two DNA mismatch repair genes, MSH2 and MLHI. Here the tumor spectrum is very different from most other dominantly inherited predispositions. The typical tumors are carcinomas, especially of the colon, stomach, and endometrium, while embryonal tumors are not observed and sarcomas are uncommon.15 Why is this disease so different from all others? First we note that other DNA repair defects are recessively inherited, as one expects of a deficiency disease. But this disorder is like those that involve tumor-suppressor genes; oncogenesis depends upon loss of the second allele of MSH2 or MLH1.16 These cells undergo further mutations at a greatly increased rate, because they now have a repair defect. These rates may be more than one hundred times usual spontaneous somatic mutation rates,17 so they can then compensate for the requirement for an extra event at the beginning of the se- quence of events. However, if only two somatic events are necessary, as for non-hereditary Rb, this compensation cannot occur. If a somatic mutation rate is of the order of 10" per cell division, the two events give a product of lo-'* in normal tissues and low8 or so in HNPCC tissues after the second repair allele is mutated. But this mutation occurs at the usual rate of lo", giving a product of 10-14. Obviously the situation is very different if four events, rather than two, are necessary. The relevance of this phenomenon for prevention is that, just as more events multiply the effect of increasing mutation rate, they also multiply the effect of decreasing mutation rates. To the extent that cancer is caused by environmental mutagens, successful at- tempts to reduce exposure to them will affect most the common multi-hit carcinomas.

CONCLUSIONS

The dominantly inherited forms of cancer have already taught us much about genetic mechanisms operating in cancer, both hereditary and non-

66 ANNALS NEW YORK ACADEMY OF SCIENCES

hereditary. The tumor spectrum of these diseases is particularly selective for kinds of cancer that entail fewer somatic genetic events. On the other hand, one disease, HNPCC, shows a reverse phenomenon, which can be account- ed for by the magnifying effect of the increased somatic mutation rates in that condition. From this we can conclude that tumors with more somatic mutations are more subject to increases in incidence by environmental agents, and to decrease in incidence as a consequence of intervention.

REFERENCES

1. Swn, M., D. MORRELL, R B. MASSEY & C. L. CHASE. 1991. Incidence of cancer in 161 families affected by ataxia-telangiectasia. New En@. J. Med., 325: 1831-1836.

2. KNUDSON, A. G. 1985. Hereditary cancer, oncogenes, and antioncogenes. Cancer

3. STRUEWING, J. P., D. ABELIOVICH, T. PERETZ, N. AVISHAI, M. M. KABAcK, F. S. COLLINS & L. C. BRODY. 1995. The carrier frequency of the BRCAl 185del AG mutation is approximately 1% in Ashkenazi Jewish individuals. Nature Genetics 11:

4. KNUDSON, A. G. 1971. Mutation and cancer: Statistical study of retinoblastoma. Proc. Natl. Acad. Sci. USA 68: 820-823.

5 . HETHCOTE, H. W. & A. G. KNUDSON. 1978. Model for the incidence of embryonal cancers: Application to retinoblastoma. Proc. Natl. Acad. Sci. USA 75: 2453-2457.

6. CAVENEE, W. K., T. P. DRYJA, R A. PHILLIPS, W. F. BENEDICT, R GODBOUT, B. L. GALLIE, A. L. MURPHREE, L. C. STRONG & R L. WHITE. 1983. Expression of recessive alleles by chromosomal mechanisms in retinoblastoma. Nature 305:

7. PRITCHARD-JONES, K, S. FLEMING, D. DAVIDSON, W. BICKMORE & D. PORTEOUS. 1990. The candidate Wilms’ tumor gene is involved in genitourinary development. Nature 346: 194-197.

8. KINZLER, K. W. & B. VOGELSTEIN. 1996. Lessons from hereditary colorectal cancer. Cell 87: 159-170.

9. LEVY, D. B., K J. SMITH, Y. BEAZER-BARCLAY, S. R HAMILTON, B. VOGELSTEIN & K W. KINZLER. 1994. Inactivation of both APC alleles in human and mouse tumors. Cancer Res. 5 4 5953-5958.

10. KNUDSON, A. G. 1992. Stem cell regulation, tissue ontogeny, and oncogenic events. Sem. Cancer Biol. 3: 99-106.

11. TARMIN, L., J. YIN, N. HAWAZ, M. KOZAM, J, NOORDZIJ, L. ANTONIO, H.-Y. JIANG, 0. CHAN, K CYMES & S. J. MELTZER. 1995. Adenomatous polyposis coli gene mutations in ulcerative colitis-associated dysplasias and cancers versus sporadic colon neoplasms. Cancer Res. 55: 2035-2038.

12. YEUNG, R S., G.-H. XIAO, F. JIN, W.-C. LEE, J. R TESTA & A. G. KNUDSON. 1994. Predisposition to renal carcinoma in the Eker rat is determined by germline mutation of the tuberous sclerosis 2 (TSC2) gene. Proc. Natl. Acad. Sci. USA 91:

13. KOBAYASHI, T., Y. HIRAYAMA, E. KOBAYASHI, Y. KUBO & 0. HINO. 1995. A germline insertion in the tuberous sclerosis (Tzc2) gene gives rise to the Eker rat model of dominantly inherited cancer. Nature Genetics 9: 70-75.

14. HINO, O., A. J. P. KLEIN-SZANTO, J. J. FREED, J. R TESTA, D. Q. BROWN, M. VILEN- SKY, R S. YEUNG, K. D. TARTOF & A. G. KNUDSON. 1993. Spontaneous and radia-

45: 1437-1443.

198-200.

779-784.

11413-11416.


tion-induced renal tumors in the Eker rat model of dominantly inherited cancer. Proc. Natl. Acad. Sci. USA 90: 327-331.

15 . WATSON, P. & H. T. LYNCH. 1992. Extracolonic cancer in hereditary nonpolyposis colorectal cancer. Cancer 71: 677-688.

16. HEMMINKI, A,, P. PELTO-, J.-P. MECKLIN, H. J ~ V I N E N , R SALOVAARA, M. NYS- TROM-LAHTI, A. DE AL CHAPELLE & L. A. AALTONEN. 1994. Loss of the wild-type MLHl gene is a feature of hereditary nonpolyposis colon cancer. Nature Genetics 8: 405410.

17. BHATTACHAFWA, N. P., A. SKANDALIS, A. GANESH, J. GRODEN & M. MEUTH. 1994. Mutator phenotypes in human colorectal carcinoma cell lines. Proc. Natl. Acad. Sci. USA 91: 6319-6323.

Documents

Hereditary Predisposition to Cancer