On the role of genes relative to the environment in carcinogenesis

Abstract

Age-specific incidence rates for cancers of the female breast, ovary, corpus uteri, testis, prostate and male and female cancers of the stomach, colon, rectum, lung, and male and female Hodgkin's disease in 51 worldwide populations in the 1990 era were obtained from a published source. For each cancer in each population, rates were summed over all age-intervals from birth to 84 years to give total incidence. The ages at 1/10th- and 1/2-this total were identified and found to vary among cancers, but, for each cancer except Hodgkin's disease, to be rather invariant with population. Except for Hodgkin's disease, the impact of this population-independent influence is far greater than the impact of the total population-dependent influence. The nature of this population-independent aging influence can be explained as the resultant of two opposing processes. On the one hand, DNA damage accumulates with advancing age and this predisposes to malignant transformation. On the other hand, cellular proliferation rates decline with advancing age and this inhibits malignant transformation. With advancing age, the second process becomes dominant suggesting that beyond some advanced age, the risk of cancer declines with advancing age.

Introduction

In all eras and locations and in both genders, cancer is strikingly age-dependent (Turker, 2000). In the USA, approximately 60% of all cancers are diagnosed after age 65, although this age group represents only 12% of the population (Anonymous, 1997). Only 2% of all cancers are diagnosed under age 15 (Robison, 2000), although this age group represents more than 21% of the population. The cancers that are common in childhood, i.e. acute lymphoblastic leukemia, cerebellar and brain stem tumors, neuroblastoma, retinoblastoma, Wilms’ tumor, osteogenic and Ewing's sarcomas, and rhabdomyosarcoma are rare in older subjects (Robison, 2000). Two cancers that are common in middle-aged subjects, embryonal germ cell carcinomas of the testis (Nichols et al., 2000) and the nodular sclerotic subtype of Hodgkin's disease (Mauch and Armitage, 2000) are rare in younger and older subjects. The reason for this strong age-dependence is unknown. But there are only three possibilities: (1) environmental agents, (2) genes, or (3) some combination of the two.

There is little doubt that environmental agents cause cancer (Doll and Peto, 1981, Doll, 1999). The list of chemical, physical, and biological carcinogens is long and growing (Doll and Peto, 1981, Doll, 1999, Weston and Harris, 2000, Maltoni et al., 2000, Fine and Sodroski, 2000). The only question concerns the percentage of human cancer caused by these agents relative to the percentage that would be caused by genes acting in a carcinogen-free environment. Current estimates pin some 80% of the blame for human cancer on environmental agents (Doll and Peto, 1981, Doll, 1999, Lichtenstein et al., 2000, Hoover, 2000). I suggest these estimates are exaggerated.

If the environment were the principal cause of cancer, we would expect the risk of cancer to increase with exposure, and therefore, with age. This is clearly not the case for those cancers that exhibit peaks in incidence rates in childhood or middle-age. And although these cancers account for only a small percentage of all cancer, there are at least a dozen different cancers that exhibit such peaks. Any theory of carcinogenesis that portrays environmental agents as the principal cause of cancer must view each of these cancers as an exception to the rule.

If the environment were the principal cause of cancer, we would expect cancer incidence rates not only to increase with advancing age, but to increase at a continually increasing rate, i.e. to accelerate. This is not what is observed. The common cancers that have been studied show a continuous deceleration in incidence rate after approximately age 30 (Benson et al. 1996). Only a theory that portrays the action of genes as the principal problem has the potential both to encompass all cancers and to explain the continuous deceleration in incidence rate of the common cancers.

Genes are known to cause cancer in the absence of carcinogens and the list of such genes is also long and growing (Doll, 1999, Brose et al., 2000, Pierotti et al., 2000, Fearon and Vogelstein, 2000, Turker, 2000). But, by current estimates, the amount of cancer caused by such genes is small (Doll and Peto, 1981, Doll, 1999, Lichtenstein et al., 2000, Hoover, 2000). The methods supporting these estimates, however, are flawed. The most powerful argument against genes as the principal cause of cancer is based on twin studies (Lichtenstein et al., 2000, Hoover, 2000). Because the cancer concordance rate is only slightly higher for mono- than for dizygotic twins, it was concluded that ‘the environment has the principal role in causing sporadic cancer’ (Lichtenstein et al., 2000). What these authors fail to acknowledge is that 99.9% of human DNA is identical in all people (Collins and McKusick, 2001). If any part of this common portion of the genome were the principal cause of cancer, it would be invisible in twin studies. And a recent reassessment of twin and family data suggests that Lichtenstein et al. (2000) underestimated the role of segregating genes (Risch, 2001).

The known cancer-causing genes are located within the 0.1% of the genome that segregates. If these genes account for 1% of all cancers, a very conservative estimate (Brose et al., 2000) and if a similar proportion of cancer-causing genes were located in that 99.9% of the genome that doesn't segregate, then genes would easily be the principal cause of cancer.

Twins share environment as well as genes, and the fact that the cancer concordance rate among twins is small, indicates that shared environment is not a principal cause of cancer (Lichtenstein et al., 2000, Hoover, 2000). This conclusion is the same as that from spouse studies (Dong and Hemminki, 2001a). The definitive data on shared environment is from women with primary cancer in one breast. The unaffected breast has the same genes and, for most women, has been exposed in the same way to the same environment. What is the risk of developing a second primary in the contralateral breast? Small (0.7% per year) (Hoover, 2000) and, more importantly, decreasing with advancing age past menopause (Levi et al., 2001). A similar decline in risk of a second primary with advancing age has been observed in colon cancer and melanoma (Dong and Hemminki, 2001b). It is difficult to imagine how a decline in risk with advancing age can be reconciled with a theory that exposure to the environment is the principal cause of cancer.

Much of the work that launched the idea that environmental agents were more important than genes in human carcinogenesis was based on migrant studies. Migrants between communities of differing rates were found to adopt rates more typical of their new homes (Doll and Peto, 1981). The problem is that the relevant communities differed not only in cancer rates but also in stages of economic development. Cancers that exhibit decreasing rates as migrants move from less to more developed areas, e.g. stomach and liver cancers (Thomas and Karagas, 1987, Yu et al., 1991), do suggest a strong role for the environment. But most cancers exhibit increasing rates as people move from less to more developed communities (Doll and Peto, 1981, Thomas and Karagas, 1987, Yu et al., 1991, Ziegler et al., 1993), and then the conclusion must be tempered: ‘a number of possible explanations must be considered when interpreting international data, including potential differences in medical care systems, diagnostic acumen, differences in classification, thoroughness of cancer surveillance and reporting, the size of the population and accuracy of census, and the true occurrence of cancer among the various populations’ (Robison, 2000).

Because the quantification of avoidable cancer risks (Doll and Peto, 1981) is based on international comparisons of age-standardized incidence rates (ASR) and suffers from the same flaws as those migrant studies that find cancer rates to be higher in more- relative to less-developed countries (Doll and Peto, 1981), the quantitative role of genes relative to the environment in human carcinogenesis warrants revaluation.

Endogenous metabolic processes are known to be an important source of carcinogens (Ames and Gold, 1990, Fraga et al., 1990), and genes can influence the generation of endogenous carcinogens as well as the vulnerability of host cells to induced and spontaneous carcinogenesis. I suggest that information on the role of genes relative to the environment lies buried in the cross-sectional age-distributions of cancer incidence rates. Cross-sectional studies suffer from potential errors due to temporal trends in incidence rate. Older or younger birth cohorts may have more or less risk of cancer due to different histories of exposure to environmental carcinogens. When cross-sections of all birth cohorts are analyzed simultaneously on a common scale, these differences in risk due to differences in exposure can seem to be differences in risk due to differences in age (Dix, 1989). This would be of particular concern if environmental exposure were the principal cause of cancer. But temporal trends in incidence rates are insignificant for many cancers over at least the past 30 years (Thun and Wingo, 2000) and an analysis of stomach and lung cancers, which display opposite temporal trends, gave no evidence of distortion in the cross-sectional analysis (Dix, 1989). Since cohort studies are also plagued by potential errors and because data on cohorts is more limited (Dix, 1989), I suggest that cross-sectional age-incidence patterns are worthy of analysis.

Imagine that carcinogenesis can be divided into two components, one that is population-dependent and includes all genetic and environmental variables, and the other that is population-independent and includes all genetic and environmental constants. Which component contains the aging process? Until recently, there were no reference points on cancer age-incidence patterns that would permit comparisons among populations at equivalent points, and, thus, no way to estimate the magnitude of variation among populations at these points. The age at half total incidence, Age1/2, is such a reference point (Dix and Cohen, 1999). Age1/2 values for the most common cancers are rather invariant among populations that differ genetically and environmentally, and, for this reason, we concluded that Age1/2 must be determined by genes that remain constant between populations (Dix and Cohen, 1999, Leipold et al., in press). It is the purpose of this paper to demonstrate that this population-independent aging process has far greater impact on carcinogenesis than does the collection of population-dependent processes.